Structural Elucidation of the Cyclization Mechanism of α-1,6-Glucan by Bacillus circulans T-3040 Cycloisomaltooligosaccharide Glucanotransferase*

Background: Cycloisomaltooligosaccharide glucanotransferase catalyzes an intramolecular transglucosylation reaction and produces cycloisomaltooligosaccharides from dextran. Results: The crystal structure of Bacillus circulans T-3040 cycloisomaltooligosaccharide glucanotransferase was determined. Conclusion: The enzyme structures complexed with isomaltooligosaccharides and cycloisomaltooctaose revealed the molecular mechanism of action. Significance: CBM35 functions in the product size determination and substrate recruitment. Bacillus circulans T-3040 cycloisomaltooligosaccharide glucanotransferase belongs to the glycoside hydrolase family 66 and catalyzes an intramolecular transglucosylation reaction that produces cycloisomaltooligosaccharides from dextran. The crystal structure of the core fragment from Ser-39 to Met-738 of B. circulans T-3040 cycloisomaltooligosaccharide glucanotransferase, devoid of its N-terminal signal peptide and C-terminal nonconserved regions, was determined. The structural model contained one catalytic (β/α)8-barrel domain and three β-domains. Domain N with an immunoglobulin-like β-sandwich fold was attached to the N terminus; domain C with a Greek key β-sandwich fold was located at the C terminus, and a carbohydrate-binding module family 35 (CBM35) β-jellyroll domain B was inserted between the 7th β-strand and the 7th α-helix of the catalytic domain A. The structures of the inactive catalytic nucleophile mutant enzyme complexed with isomaltohexaose, isomaltoheptaose, isomaltooctaose, and cycloisomaltooctaose revealed that the ligands bound in the catalytic cleft and the sugar-binding site of CBM35. Of these, isomaltooctaose bound in the catalytic site extended to the second sugar-binding site of CBM35, which acted as subsite −8, representing the enzyme·substrate complex when the enzyme produces cycloisomaltooctaose. The isomaltoheptaose and cycloisomaltooctaose bound in the catalytic cleft with a circular structure around Met-310, representing the enzyme·product complex. These structures collectively indicated that CBM35 functions in determining the size of the product, causing the predominant production of cycloisomaltooctaose by the enzyme. The canonical sugar-binding site of CBM35 bound the mid-part of isomaltooligosaccharides, indicating that the original function involved substrate binding required for efficient catalysis.

pounds. Thus, cyclodextrins aid drug delivery and protect compounds from degradation (6,7). Cyclodextrins are also used as room refreshers.
Cycloisomaltooligosaccharides, also known as cyclodextrans (CIs), are similar cyclic oligomers of glucose molecules to cyclodextrins, but glucoses in CIs are linked with ␣-1,6-glucosidic bonds (8). CIs are produced from dextran via an intramolecular transglycosylation catalyzed by the enzyme cycloisomaltooligosaccharide glucanotransferase (CITase; EC 2.4.1.248). CIs of 7-17 glucose units (CI-7 to CI-17; named as per the number of glucose units present) are produced by Bacillus circulans T-3040 CITase (BcCITase (9 -12)). The dominant product of BcCITase is cycloisomaltooctaose (CI-8), secondary CI-7, and the amount reduces in accordance with the higher degree of polymerization for CI-9 to CI-17. CIs are highly water-soluble and have a central hydrophobic cavity that is similar to cyclodextrins. Hence, the inclusion complex-forming ability is expected to be similar to that observed for cyclodextrins and that of CI-10 against Victoria blue B, as has been reported (10,11). CIs are also known as strong inhibitors of glucansucrase, showing anti-plaque activity (13). Thus, CIs appear to be novel bio-nanomaterials applicable to various bio-industries as well as cyclodextrins.
The cit gene encoding BcCITase has been cloned, and its nucleotide sequence has been determined (14). According to amino acid sequence analysis, the enzyme belongs to the glycoside hydrolase family 66 (GH66). In the CAZy database (15), the following three CITases are listed: B. circulans T-3040; B. circulans U155 (9), and Paenibacillus sp. 598K (12). Most of the other enzymes in the family are dextranases, which hydrolyze dextran to produce linear isomaltooligosaccharides (IGs). Dextranases from streptococci, especially from Streptococcus mutans (SmDex), have been the most extensively studied among the GH66 proteins, and biochemical studies using site-directed mutagenesis revealed that Asp-385 in SmDex was essential for catalytic activity (16). Similarly, Asp-243 of dextranase from Thermotoga lettingae (17) and Asp-308 of BcCITase (18), both corresponding to Asp-385 in SmDex, are implicated as catalytic residues. We have solved the crystal structure of the fragment from Gln-100 to Ile-732 of SmDex, devoid of its N-and C-terminal variable regions, as the first structure of a GH66 protein, and we displayed that the conserved region in all GH66 proteins included three domains as follows: the domain N with an immunoglobulin fold, the catalytic domain A with a (␤/␣) 8 barrel structure, and the domain C with tandemly repeated Greek key motifs (19). Structural analysis using the ligandbound structures verified the catalytic nucleophile Asp-385 and also identified the catalytic acid/base Glu-453. Besides these GH66 core structural domains, BcCITase has two extra carbohydrate-binding module family 35 (CBM35) domains; one is inserted in the middle part of the conserved region, and this domain is common in CITases (BcCBM35-1); another is in the C-terminal region (BcCBM35-2) (20,21). The detailed function of these domains remains to be elucidated, although the presence of the first domain has been shown to stabilize the protein and to enhance the production of CI-8 (21).
In this work, we present the crystal structure of a stable C-terminally truncated mutant of BcCITase and its ligand-bound structures, and we discuss the detailed reaction mechanism of the CITase by comparison with related structures. The novel function of the inserted domain of CBM35 in the product size determination is also addressed.
Protein Expression and Crystallization-We constructed two enzyme forms containing His 6 purification tags at either the N terminus (BcCITase-NH) or the C terminus (BcCITase-CH). BcCITase-NH was composed of a single polypeptide chain of 721 amino acids (18 -738), where the N terminus 18 MGSSHH-HHHHSSGLVPRGSHM 38 was derived from the expression vector, including the purification tag and thrombin cleavage site. BcCITase-CH was composed of a single polypeptide chain of 710 amino acids , where the N-terminal Met-37 and Gly-38 and the C-terminal 739 LEHHHHHH 746 were derived from the expression vector. Both constructs contained enzyme residues Ser-39 -Met-738 (molecular mass of 78 kDa), and the N-terminal signal peptide and the C-terminal variable region were omitted. Both protein forms were expressed in Escherichia coli BL21(DE3) cells, purified by His 6 tag nickel affinity chromatography and crystallized as reported previously (21,24). BcCITase-NH with a concentration of 7 mg/ml was crystallized by the sitting drop vapor diffusion method using a precipitant solution composed of 1.0 M sodium acetate trihydrate, 0.1 M HEPES, pH 7.5, 0.05 M cadmium sulfate hydrate, and 12%(v/v) glycerol, at 293 K. BcCITase-CH with a concentration of 20 mg/ml was crystallized using a precipitant solution composed of 1.6 M ammonium sulfate, 0.1 M MES, pH 6.5, and 10% Site-directed mutagenesis, Asp-308 to alanine, of BcCITase-CH was performed using the QuikChange mutagenesis kit (Stratagene, Agilent Technologies, Inc., Santa Clara, CA). For incorporation of the selenomethionine (SeMet) enzyme, BcCITase-NH proteins were expressed in E. coli B834(DE3) cells grown in LeMaster medium supplemented with 25 mg/liter seleno-Lmethionine (Wako Pure Chemical, Tokyo, Japan). Expression and purification were conducted in a similar manner to the native enzyme.
Data Collection and Structural Determination-Data for the native and its SeMet derivative crystals of BcCITase-NH protein were collected at 95 K to 2.3 Å resolution at beamline BL-6A and to 2.8 Å resolution at beamline BL-5A (Photon Factory, High Energy Accelerator Research Organization, Tsukuba, Japan). Data for BcCITase-CH, its D308A mutant, and ligand complexes were collected at 95 K at the Photon Factory.
Data collection statistics are summarized in Table 1. Sugarbound BcCITase-CH D308A mutant crystals were prepared by soaking the enzyme crystal into a precipitant drop containing 1-5% (w/v) sugar ligands for 10 min to 1 h and 10% ethylene glycol for about 10 min, before the diffraction experiments. All diffraction data were processed using HKL2000 (25). Initial structural determination was conducted by the multiple-wavelength anomalous dispersion method. Selenium atom positions in the SeMet derivative of BcCITase-NH were determined, and phases were calculated using the programs SOLVE/RESOLVE (26,27). Further manual model building and refinement were performed for the native crystal using ARP/wARP (28), REFMAC5 (29), and COOT (30). Structural models included calcium and cadmium ions, which were distinguished by the anomalous signals and peak sizes. The crystal structures of the BcCITase-CH D308A mutant complexed with ligands were determined by the molecular replacement method using MOLREP (31) and refined with REFMAC5 and COOT. The structure of the ligand-free BcCITase-CH could not be refined because of crystal twinning, probably being transformation twins containing tetragonal and orthorhombic crystal systems. Structural refinement statistics are summarized in Table 2. Stereochemistry of the models was analyzed with the Rampage program (32), and structural drawings were prepared using the PyMOL program (DeLano Scientific LLC, Palo Alto, CA).

RESULTS
Overall Structure-The crystal structure of BcCITase-NH was determined by the multiwavelength anomalous dispersion method at 2.3 Å resolution. The BcCITase-NH model includes a single protein molecule, as well as the surrounding water and glycerol molecules and sodium, calcium, cadmium, and acetate ions. The N-terminal nine residues Met-18 -His-26 were not identified because of a lack of electron density. Successively, four ligand complex structures of the BcCITase-CH D308A mutant with IG-6 (BcCITase⅐IG-6), IG-7 (BcCITase⅐IG-7), IG-8 (BcCITase⅐IG-8), and CI-8 (BcCITase⅐CI-8) were determined ( Table 2). The crystals of BcCITase⅐IG-6 and BcCITase⅐IG-7 belong to the tetragonal space group of P4 1 2 1 2, whereas the BcCITase⅐IG-8 and BcCITase⅐CI-8 crystals belong to the less symmetrical orthogonal space group of P2 1 2 1 2 1 . This difference might have been caused by the structural rearrangement upon binding of relatively large ligands at the catalytic cleft and the substrate-binding site in BcCBM35-1, although the crystals were obtained from identical conditions. The tetragonal and orthorhombic crystals contained one and two protein molecules in the asymmetric unit, respectively. The N-terminal five residues Met-37-Ser-41 and the C-terminal five residues His-742-His-746 were not identified because of a lack of electron density. Besides the sugar and water molecules, the structural model contained sodium, calcium, and sulfate ions, as well as MES and ethylene glycol molecules.
The core structure of BcCITase is composed of four domains. The ribbon model of the BcCITase structure for chain B of the BcCITase⅐CI-8 complex is shown in Fig. 1. As three of them are conserved in the C-terminal truncated mutant of SmDex belonging to GH66 (Fig. 2) (19), the domain names are designated in accordance with those of SmDex; the N-terminal BcCITase⅐IG-8

BcCITase⅐CI-8
Space group  (19), in which the second and fifth strands are both split into two strands.
Domain A (Ser-140 -Ala-419 and Thr-547-Asp-614) folded into a (␤/␣) 8 -barrel. The loop3, located between the 3rd ␤-strand A␤3 and the 3rd ␣-helix A␣3 (see the names in the secondary structures of Fig. 1B), is the longest with 47 residues, and it was found to be compactly folded into a small block with the loop4 in a manner similar to that observed for the SmDex structure. The structure of the barrel was also almost identical with that of SmDex.
Domain C (Ser-615-Met-738) was composed of a tandemly repeated Greek key motifs. The overall fold of domain C was similar to that of SmDex ( Fig. 2) (19), except that BcCITase lacked the ␣-helix located between C␤3 and C␤4. The electron density at the loop region between C␤1Ј and C␤1 was poor, and this region had variable structures between the models. Domain arrangement of BcCITase was similar to that of SmDex but different from that of CGTases that consisted of five domains and belonged to GH13 (33). Besides CGTases, significant similarities were not observed with the structures of amylomaltase belonging to GH77 (34 -37) or 4-␣-glucanotransferase belonging to GH57 (38) that produce larger cyclic ␣-1,4-glucan or cycloamylose.
Structure of BcCBM35-1-Domain B (Gly-420 -Gly-546) folded into a ␤-jellyroll, comprising BcCBM35-1, and this structure was inserted between A␤7 and A␣7 of the catalytic domain, where the corresponding region in SmDex was loop 7 with an ␣-helix fold (Figs. 1 and 3A). BcCBM35-1 protruded from the catalytic domain on the catalytic cleft side. Structural comparison using the Dali server showed that this domain had the highest similarity with CBM35 structures of Amycolatopsis orientalis exo-␤-D-glucosaminidase CsxA (Chi-CBM35, PDB code 2VZP) and Clostridium thermocellum rhamnogalacturonan acetyl esterase Rgae12A (Cthe_3141) (Rhe-CBM35, PDB code 2W1W), with a root mean square difference of 1.4 Å (39). These CBM35 structures have two calcium ion-binding sites (Fig.   3B). One of them is located between the end of B␤1 and the beginning of B␤1Ј. This site is conserved among CBM35s, except for Cellvibrio japonicus endo-␤-1,4-mannanase (Mac5C-CBM35, PDB code 2BGO (40)) and Rhe-CBM35 (39). In BcCBM35-1, this calcium ion-binding site is conserved and formed by Glu-424, Glu-426, Thr-443, Gly-446, and Asp-541 (Fig. 3, A and C). In the BcCITase-NH structure, the calcium ion was replaced by a cadmium ion, which was derived from the crystallization conditions. The second calcium ion-binding site of CBM35s is located at the tip of the fold on two loops between B␤1Ј and B␤2 and between B␤7 and B␤8, which is conserved in all CBM35 structures determined and contributes to substrate recognition. In BcCBM35-1, however, the main chain of these loops was structured differently from those of the other CBM35s, and the asparagine residue on the first loop was replaced by Ala-414 (Fig. 3D). Concomitantly, the second calcium-binding site was not conserved, and the metal ion was not observed. Instead, this site acted as one of the sugar-binding sites with affinity toward isomaltooligosaccharides, as described later (Fig. 3E).
Sugar Complex Structures of BcCITase-In the crystal structure of BcCITase-NH, the peptide tag used for protein expression and purification was maintained without protease digestion and stacked into the catalytic cleft of the adjacent molecule in the crystal, preventing ligands from accessing the active site in the soaking experiments. To obtain stable ligand⅐protein complex structures, the catalytic site mutant recombinant BcCITase-CH was used. Simultaneously, for the purpose of the soaking experiments of the protein⅐sugar complex, a point mutation of nucleophile Asp-308 to alanine was introduced, because the CITase mutants in which the nucleophile was mutated showed no enzyme activity (11,21). The ligands were observed mainly at two sites, the catalytic cleft in domain A (Fig. 1A, site A-1) and the sugar-binding site of BcCBM35-1 (Fig. 1A, site B-1). BcCITase-CI-8 contained an additional CI-8 molecule in the intermolecular cavity surrounded by loop 3 of domain A and the bound CI-8 molecule at the catalytic cleft (Fig. 1A, site A-2). This CI-8 bound loosely, as judged from the B-factor of the bound CI-8 molecule and appeared to be an artifact due to crystal packing. However, it was embedded within the exposed hydrophobic residues of Tyr-243, Tyr-255, Tyr-274, Tyr-276, and Tyr-319 on loop3 and loop4 (Fig. 1B), and thus on the opposite side of the catalytic cleft. Therefore, these hydrophobic residues might function as a surface-binding site, as has been observed for barley ␣-amylases (41).
In the catalytic cleft, four glucose residues out of six were observed in the BcCITase⅐IG-6 complex structure, whereas in the other complex structures, the whole molecules were observed  -1, B-2. B, topological diagram of BcCITase as follows: ␣-helices, 3 10 -helices, and ␤-strands, as filled cylinders, shaded cylinders, and filled arrows, respectively. Catalytic residues, bound calcium, and sodium ions were placed. (Fig. 4A). In BcCBM35-1, four or five glucose moieties are modeled in an almost identical conformation in the sugar-binding site of BcCITase⅐IG-6, BcCITase⅐IG-7, and BcCITase⅐IG-8 complex structures (Fig. 1A, site B-1). In the BcCITase⅐CI-8 complex structure, the bound CI-8 molecules were observed differently between two nonsymmetrically related molecules. Five glucose moieties were observed at the same positions with the BcCITase⅐IG-8 complex structure in chain B, whereas all eight glucoses of the CI-8 molecule were observed in chain A (Fig. 3, E and F).
Sugar-binding Structure in the Catalytic Site-In the BcCITase⅐IG-6 complex structure, four glucose residues (Glc-1 to Glc-4 from the reducing end) were observed at four subsites (Ϫ1 to Ϫ4 from the catalytic site) only in one side of the active site, similar to those in SmDex (Fig. 4, B and C) (19), and the other two glucose moieties (Glc-5 and Glc-6) were disordered. In the BcCITase⅐IG-8 complex, the entire IG-8 molecule (Glc-1 to Glc-8 from the reducing end) was observed, and the four glucose moieties from the reducing end were located at the same position with those of the BcCITase⅐IG-6 complex (Fig. 5,  A and B). The glucose moieties at the subsites Ϫ1 (Glc-1) and Ϫ2 (Glc-2) were recognized via hydrogen bonds by the enzyme, and Glc-2 was buried in the pocket made by hydrophobic residues Leu-206, Phe-207, Tyr-233, Met-235, Phe-268, Leu-275, Met-310, and Tyr-581, in a manner similar to SmDex (Fig. 2) (19). The anomeric hydroxyl group of Glc-1 adopted a ␤-configuration in both complexes. The differences from SmDex are found in His-188 and Glu-580 of BcCITase, both of which were replaced by alanine in SmDex. Glc-3 and Glc-4 were mainly recognized by the hydrophobic residues Tyr-581 and Phe-207. The other four glucose moieties Glc-5 to Glc-8 in the BcCITase⅐IG-8 complex were observed to extend outwards from the catalytic cleft (Fig. 1A, site B-2). Glc-5 to Glc-7 were located in the hydrophobic straight between domains A and B, each had one direct hydrogen bond to protein residues but did not show tight binding architectures. Glc-8 was buried in the pocket made by aromatic residues Tyr-499, Phe-501, Trp-514, and Tyr-515 in domain B with its O6 atom at the bottom and recognized directly via the formation of four hydrogen bonds with Gln-496, Tyr-499, and Trp-514. This glucose-binding pocket was located at the side of BcCBM35-1, involving the above residues on B␤5 and B␤6. This second sugar-binding site was not conserved in the other CBM35 structures reported.
In the BcCITase⅐CI-8 complex, CI-8 (Glc-1 to Glc-8) was bound to the catalytic cleft (Fig. 5, C and D). The bound CI-8 took a distorted cyclic structure along the catalytic surface that consisted of loops 3 and 4 encircling the side chain of Met-310. Glc-1 and Glc-2 were located at the same position with those of the bound IG-6 and IG-8. Glc-8 of CI-8 was situated adjacent to Glc-1 and over the catalytic site. A covalent ␣-1,6-bond was also observed between the Glc-1 O1 atom and the Glc-8 O6 atom. Glc-3 was located at subsite Ϫ3, as observed for the bound IG-6 and IG-8, but the orientation of the sugar ring was different, and it was found to have formed a hydrogen bond between Glu-580 and the O3 atom. Glc-4, Glc-5, and Glc-6 had no direct hydrogen bonds with the protein and were situated on the hydrophobic surface that included residues Val-266, Phe-268, Met-310, and Arg-313. Glc-7 was located on the main chain of Gly-311-Gln-312 and bound by one hydrogen bond with the side chain of Asp-357. Glc-8 was located on the Trp-382 side chain, halfcovered by Met-310 and bound by one hydrogen bond with the side chain of Glu-580. These two glucoses, Glc-7 and Glc-8, were considered to lie at subsites ϩ1 and ϩ2, respectively. The subsites for the bound CI-8 were designated after the nomenclature of the subsites for CGTase (42) as subsites Ϫ1, Ϫ2, Ϫ3, Ϫ4c, Ϫ5c, Ϫ6c, ϩ2, and ϩ1 (Figs. 4A and 5C).
In the BcCITase⅐IG-7 complex, IG-7 (Glc-1 to Glc-7) was bound to the catalytic cleft in a similar manner to CI-8 in the BcCITase⅐CI-8 complex (Fig. 4, A and D). Glc-1 to Glc-4 were located at the same position with CI-8, and Glc-6 and Glc-7 were located at the positions of Glc-7 and Glc-8 of CI-8. Glc-5 was located on the side chain plane of Arg-313 that corresponded to the mid-point of Glc-5 and Glc-6 of CI-8, forming the shorter ␣-1,6-glucan chain between Glc-4 and Glc-6. The anomeric hydroxyl group of Glc-1 adopted a ␤-configuration, and no covalent bond was observed with the O6 atom of Glc-7, a distance of 3.3 Å apart.
Sugar-binding Structure in BcCBM35-1-Domain B, BcCBM35-1, bound isomaltooligosaccharides at two sites. The second site  was located at the side of the domain and recognized the nonreducing end glucose of IG-8, which extended from the catalytic site and worked as subsite Ϫ8, as mentioned above. Another site was the canonical sugar-binding site located on the top of the domain, where most CBM35s recognize their cognate sugars, involving the second calcium ion (Fig. 3D). This first site mainly recognizes one glucose moiety through stacking interactions with Trp-506 and hydrogen bonds by the side chains of His-439 and Asn-539, and the main chain N atom of Gln-450 (Fig. 3E). His-439, Trp-506, and Asn-539 are highly conserved among CBM35s, including Chi-CBM35 and Xyl-CBM35 (39), and are engaged in uronic sugar recognition in an identical manner. In these structures, an arginine residue on loop6 (the loop between ␤6 and ␤7) and the second conserved calcium ion interact with the carboxylic group of glucuronic acid. Isomaltooligosaccharides do not have a carboxylic group, and this position was occupied by the ␣-1,6-glucosyl bond to the adjacent glucose moiety. Thus, the arginine residue and calcium ion are not present in BcCBM35-1, because of the structural changes of the loops. Because of these structural changes, the first site of BcCBM35-1 could bind to the midpoint of the linear isomaltooligosaccharides. One glucose moiety on the reducing end side and two glucoses on the nonreducing end side of the bound glucose moiety (Glc-914, Fig. 3E) were commonly observed to link through ␣-1,6-glucosyl bonds in all ligand⅐complex structures. Intriguingly, the entire CI-8 molecule was observed in chain A of the BcCITase⅐CI-8 complex. The bound CI-8 took a constricted but symmetric cyclic structure, like a teacup, containing three inner hydrogen bonds that formed the bottom of the cup (Fig. 3F). The CI-8 molecule was found to wrap up the side chain of Trp-506, as if CI-8 had formed an inclusion complex.

DISCUSSION
Molecular Mechanism of Action for BcCBM35-1-A previous deletion mutant study (21) revealed that the presence of BcCBM35-1 is relevant to the predominant production of CI-8 by BcCITase, and our present ligand complex structures proved that BcCBM35-1 was involved in CI-8 production as shown in Fig. 6. The BcCITase⅐IG-8 complex structure showed that the IG-8 molecules bound in the catalytic cleft extending from subsite Ϫ1 at the catalytic site to subsite Ϫ8 in BcCBM35-1, whereas the BcCITase⅐CI-8 complex structure showed that the bound CI-8 was located in the catalytic cleft over the catalytic site. This implies that these structures correspond to the enzyme⅐substrate complex (Fig. 6, step B) and the enzyme⅐product complex (Fig. 6, step D) when this enzyme produced the product CI-8. These observations provide the CI-8 producing mechanism with the aid of BcCBM35-1 as follows. BcCBM35-1 binds the nonreducing end glucose of the substrate at the second sugar-binding site (Fig. 6,  step A). The bound ␣-1,6-glucan folds into the catalytic cleft along the subsites Ϫ8 to Ϫ1 with eight glucose moieties occupied (Fig. 6,  step B). The ␣-glucosidic linkage between Glc Ϫ1 and Glc ϩ1 is cleaved at the catalytic site, and the glucose at subsite Ϫ1 of the rest is held by the catalytic nucleophile as an enzyme-substrate intermediate, at the same moment the nonreducing end glucose is released from subsite Ϫ8 and occupies subsite ϩ1 by folding into a circular conformation around Met-310 (Fig. 6, step C). The glu-cose at subsite ϩ1 then functions as an acceptor of the transglucosylation to complete the cyclization reaction (Fig. 6, step D). The mutational experiments at subsite Ϫ8 (Y515A or Y515G), in which the mutants showed reduced production of CI-8 (data not shown), also support the finding that BcCBM35-1 enhanced CI-8 production. Similar observations that the carbohydrate-binding module is involved with the product size have been reported for CGTases. The small domain E, belonging to CBM20, adjacent to the active site of CGTase has two sugar-binding sites, and a substrate-binding site with lower affinity guides the substrate to the active site (43). Mutational studies of CGTase showed that the amino acid residues at donor subsites Ϫ3, Ϫ6, and Ϫ7 influence the product sizes (3,44). The BcCITase⅐IG-7 complex structure showed that the bound IG-7 was located in the catalytic cleft with a circular structure, and IG-7 appeared to be the snapshot just before the cyclization reaction when this enzyme produced CI-7, with the nonreducing end glucoses Glc-6 and Glc-7 docked at subsites ϩ2 and ϩ1, respectively. The O6 atom of Glc-7 at subsite ϩ1 is the closest transglucosylation acceptor with a proximity of 3.3 Å from the C1 atom at subsite Ϫ1. It also implied that IG-7 could not reach subsite Ϫ8 from the catalytic site, and that CI-7 was produced without the aid of subsite Ϫ8 in BcCBM35-1. The inability to produce CI-6 or smaller CIs can also be attributed to the structure of catalytic subsites. In the BcCITase⅐IG-7 and BcCITase⅐CI-8 complex structures, the catalytic site bound the circular sugar ligand with at least four subsites, namely Ϫ2, Ϫ1, ϩ1, and ϩ2, occupied, although in the BcCITase-IG-6 structure, bound IG-6 was observed only in Ϫ1 to Ϫ4 subsites. This indicates that the cyclization reaction involves at least four subsites Ϫ2, Ϫ1, ϩ1, and ϩ2, and IG-6 appears to be too short to occupy these subsites in a circular structure.
Sugar Binding in BcCBM35-1-The original function of BcCBM35-1 appears to be the binding of the substrate for efficient recruitment, as the canonical sugar-binding site (Figs. 1A, site B-1, and 3E)) was revealed to bind the mid-part of ␣-1,6glucan, showing the type B glycan chain-binding architecture among the three sugar-binding types of carbohydrate-binding modules (45). This first binding site was located on the opposite side from the catalytic site in the three-dimensional structure of BcCITase, and hence its ligand binding appears to occur independently from the catalytic machinery. It may work to select the long substrate that could reach to the catalytic site so that CIs could be produced by the enzyme. Conversely, the second binding site, subsite Ϫ8 (Figs. 1A, site B-2, and 5A), showed the type C small sugar-binding architecture. BcCBM35-1 is the first example among the solved CBM35 structures that has two sugar-binding sites, although the presence of two sugar-binding sites has been reported for CBM6s, a related family of CBM35 (46,47). More interestingly, this site is unique to BcCBM35-1 among CBM35s and is not conserved in BcCBM35-2 in BcCITase or the CBM35s of CITase from Paenibacillus sp. 598K strain (12). Paenibacillus CITase does not predominantly produce CI-8, but CI-7 instead, and this observation is consistent with the concept that this enzyme does not possess the second sugar-binding site.
Structural Determinant for Cyclization Activity-GH66 enzymes contain mainly two enzymes, endo-dextranase and CITase. Because all CITases have the inserted CBM35 (20) compared with dextranases, it had been considered that CITase activity might be attributed to CBM35, but the relationship remains unclear. Recently, both Bacteroides thetaiotaomicron dextranase lacking this inserted domain and Paenibacillus sp. endo-dextranases containing the CBM35 domain have been reported to show CI producing activity, although the activity of these two enzymes is reduced when compared with the activity of BcCITase (48,49). Furthermore, deletion mutants, in which BcCBM35-1 is removed from BcCITase, still retain approximately one-twentieth of the CI producing activity of the wildtype enzyme (21). Therefore, CITase activity cannot be attributed wholly to CBM35, although CBM35 enhances the CITase activity. The crystal structures of BcCITase showed that BcCBM35-1 might contribute to the determination of product size preferences, as well as the usual task of CBMs in substrate recognition for efficient catalysis. BcCBM35-1 is distal from the cleavage site, and it would be difficult for the domain to affect the cyclization reaction.
Structural comparison of the catalytic cleft between BcCITase and SmDex displayed an obvious difference in the widths of the subsites ϩ1 and ϩ2, showing that the cleft of BcCITase is narrower than that of SmDex (Fig. 3, A and B). At subsite ϩ1, Met-310 and Glu-580 in BcCITase, corresponding to Ile-387 and Ala-559 in SmDex, are involved in acceptor recognition and are possible candidates that enhance transglycosylation. At subsite ϩ2, Gly-311, Gln-312, and Arg-313 provided van der Waals contact with the glucose in BcCITase, whereas the main chain trace of Gly-388 -Asn-390 had shifted to make the cleft wider in SmDex.
Met-310, which is located at the center of the bound IG-7 or CI-8 molecules in these complex structures, appears to play a role similar to that of Tyr-195 of B. circulans 251 CGTases (50,51). In this enzyme, substrate oligosaccharides are wrapped around this tyrosine residue upon the cyclization reaction, and mutational experiments of this residue cause a change in product profiles (50). In contrast with CGTases, GH13 ␣-amylases, which hydrolyze linear chains of ␣-1,4-glucans, have smaller residues at this position (52). In GH66, methionine is conserved only in CITases, whereas this residue is replaced by isoleucine in streptococcal dextranases and a variety of hydrophobic residues are found in other enzymes. Paenibacillus dextranase has tryptophan at this position and synthesizes relatively large CI molecules (48). Thus, the amino acid residues at this position might function in the determination of product size.
Glu-580 hydrogen bonded to the glucose at subsites ϩ1 in the BcCITase-CI-8 and BcCITase⅐IG-7 complex structures. Interestingly, Glu-580 of BcCITase is also conserved in CITases and Paenibacillus dextranase, but this residue is replaced by alanine in streptococcal dextranases. Amino acid residues at the acceptor subsites ϩ1 and ϩ2 in CGTases have been shown to play important roles in the cyclization reaction. Mutational studies cause the conversion of enzymatic activities, from CGTases to maltogenic amylase (53) or from Novamyl, a maltose-forming maltogenic amylase, to a cyclodextrin-forming enzyme (54). With regard to BcCITase, the significance of residues at subsites ϩ1 and ϩ2 to their enzymatic activities requires further clarification, and our biochemical studies have not yet provided a clear solution on this issue. Thus, more careful and detailed structural and biochemical studies on BcCITase are necessary to clarify the factors determining the CI producing activity.
Only a few bacteria are known to possess CITases that produce CIs from dextran, which is not an abundant natural resource synthesized from sucrose by some lactic acid bacteria (55). CIs appear to be by-products that finally degraded to glucoses by CITase for energy use, and the biological function of CIs in bacteria remains unclear. CIs, however, show profitable properties for human life. They inhibit glucan synthesis by glucansucrase, applicable for dental caries prevention (13), and show inclusion complex-forming abilities similar to that observed for cyclodextrins that are used in various ways (10,11). Furthermore, CIs show higher water solubility and a wider range of sizes from 7 to 17 degrees of polymerization (21). This is larger than that found for cyclodextrins that have three forms, and CIs are therefore expected to expand the use of cyclic glucans. Here, we have determined the crystal structure of CITase, elucidating its catalytic mechanism and providing the structural basis for enzymatic improvement. We hope that the elucidation of biological, chemical, and physical properties of CIs is accelerated and widespread use of CIs is developed.