Crystal Structure and Mutational Analysis of Isomalto-dextranase, a Member of Glycoside Hydrolase Family 27*

Background: Arthrobacter globiformis T6 isomalto-dextranase (AgIMD) hydrolyzes a polysaccharide, dextran, but is classified into glycoside hydrolase family (GH) 27, which includes mainly α-galactosidases and α-N-acetylgalactosaminidases. Results: The crystal structure of AgIMD was determined. Conclusion: AgIMD has features found in GH13, GH31, and GH66 enzymes. Significance: The results provide insights into the evolutionary relationships among GH13, -27, -31, -36, and -66. Arthrobacter globiformis T6 isomalto-dextranase (AgIMD) is an enzyme that liberates isomaltose from the non-reducing end of a polymer of glucose, dextran. AgIMD is classified as a member of the glycoside hydrolase family (GH) 27, which comprises mainly α-galactosidases and α-N-acetylgalactosaminidases, whereas AgIMD does not show α-galactosidase or α-N-acetylgalactosaminidase activities. Here, we determined the crystal structure of AgIMD. AgIMD consists of the following three domains: A, C, and D. Domains A and C are identified as a (β/α)8-barrel catalytic domain and an antiparallel β-structure, respectively, both of which are commonly found in GH27 enzymes. However, domain A of AgIMD has subdomain B, loop-1, and loop-2, all of which are not found in GH27 human α-galactosidase. AgIMD in a complex with trisaccharide panose shows that Asp-207, a residue in loop-1, is involved in subsite +1. Kinetic parameters of the wild-type and mutant enzymes for the small synthetic saccharide p-nitrophenyl α-isomaltoside and the polysaccharide dextran were compared, showing that Asp-207 is important for the catalysis of dextran. Domain D is classified as carbohydrate-binding module (CBM) 35, and an isomaltose molecule is seen in this domain in the AgIMD-isomaltose complex. Domain D is highly homologous to CBM35 domains found in GH31 and GH66 enzymes. The results here indicate that some features found in GH13, -31, and -66 enzymes, such as subdomain B, residues at the subsite +1, and the CBM35 domain, are also observed in the GH27 enzyme AgIMD and thus provide insights into the evolutionary relationships among GH13, -27, -31, -36, and -66 enzymes.

The most notable feature of AgIMD is that the primary structure of the enzyme is homologous to those of ␣-galactosidases and ␣-N-acetylgalactosaminidases, despite the fact that AgIMD hydrolyzes a polymer of glucose. AgIMD is classified as a member of glycoside hydrolase family (GH) 27 in the CAZy database (7). Another enzyme family, GH36, comprises mainly ␣-galactosidases and ␣-N-acetylgalactosaminidases. The three-dimensional structures of the two families, GH27 and GH36, share a common structural core (8,9), a catalytic (␤/␣) 8 barrel domain, and the two families are demonstrated to be evolutionarily related (10). It is interesting to note that GH31 also has a similar catalytic (␤/␣) 8 barrel domain, and the three families, GH27, GH31, and GH36, are categorized into clan GH-D. The main GH31 members function as ␣-glucosidases and ␣-xylosidases (11,12), but an ␣-galactosidase that belongs to GH31 has been reported recently (13). Therefore, the structure-function relationships of clan GH-D is complicated, and the study of AgIMD is useful for better understanding of clan GH-D.
In addition to the unique phylogenetic position of the catalytic domain, AgIMD has another intriguing architectural feature belonging to carbohydrate-binding module (CBM) 35. The CBM35 domains have been shown to share a highly similar fold, although there is considerable diversity in the biological functions of the enzymes possessing CBM35 (14,15). We have previously constructed the expression system of AgIMD in Escherichia coli (16), and the enzyme has been crystallized (17). Here, we determined the crystal structure of AgIMD and compared it with those of GH27, GH31, and other related enzymes.

Experimental Procedures
Construction of Expression Plasmids for Wild-type AgIMD and the Mutants-An expression plasmid, pETG2DspϪ, which is a derivative of pET3a(ϩ), has been previously obtained (16). To facilitate the purification of AgIMD, an expression plasmid of His-tagged AgIMD was constructed. pETG2DspϪ was digested with NdeI and BamHI, and the fragment was ligated into the NdeI-BamHI site of pET28a(ϩ), resulting in plasmid pET28a-AgIMD. Mutant enzymes were generated by site-directed mutagenesis with a QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). To construct expression plasmids for mutants, oligonucleotides 5Ј-C TGG TAC GAG GAC GGA AGG GCC GCG AAT ATT GGG CAG GTC-3Ј (for D207A), 5Ј-TCC TGG TAC GAG GAC GGA AGG GCC GCG GCC ATC GGG CAG G-3Ј (for D207A/ N209A); and 5Ј-GAA GTT TCG CTC GTA GCG CCG CAT ATG TTC-3Ј (for M243A) and their complementary strands were used as primers (mutated amino acid residues are underlined). All the constructs used were verified by DNA sequencing.
Expression and Purification of AgIMD-E. coli BL21(DE3) cells harboring pET28a-AgIMD were grown in 1 liter of Luria-Bertani (LB) medium containing kanamycin (50 g ml Ϫ1 ) to A 600 ϭ 0.6 -0.8, and then induced with isopropyl ␤-D-thiogalactopyranoside at a final concentration of 0.1 mM; the incubation was then continued for 18 h at 20°C. The cells were harvested by centrifugation at 4000 ϫ g for 5 min, resuspended in 40 ml of 20 mM imidazole in 20 mM Tris-HCl buffer (pH 8.0), and disrupted by sonication. The cells and supernatants of the cultures were separated by centrifugation at 12,000 ϫ g for 20 min. The supernatant obtained was applied onto a nickel-nitrilotriacetic acid (Ni-NTA)-agarose (Qiagen, Hilden, Germany) column equilibrated with 20 mM imidazole in 20 mM Tris-HCl buffer (pH 8.0). After washing the column with the same buffer, the enzyme was eluted with 100 mM imidazole in 20 mM Tris-HCl buffer (pH 8.0). The active fractions were collected, and the buffer was replaced by 20 mM Tris-HCl buffer (pH 8.0) using an Amicon Ultra-15 centrifugal filter unit (Merck Millipore, Darmstadt, Germany). The His tag was cleaved by thrombin (1 unit per 1 mg of AgIMD) at 20°C for 16 h and applied onto a Ni-NTA column with the same procedure, and the flowthrough fraction was collected. The AgIMD mutants were expressed and purified in a manner identical to that used for the wild-type enzyme.
Crystallization, Data Collection, and Model Building-The enzyme was crystallized at 20°C using the hanging drop vapor diffusion method, where 1 l of AgIMD (30 mg ml Ϫ1 ) was mixed with the same volume of well solution containing 50 mM sodium acetate buffer (pH 3.6), 16% (w/v) polyethylene glycol 8000, and 50 mM potassium dihydrogen phosphate. For phase determination, the crystal was first transferred for 1 min to a solution containing 50 mM sodium acetate buffer (pH 3.6) and 20% polyethylene glycol 8000, and was then transferred for 48 h to a solution containing 10 mM lead(II) acetate, 50 mM sodium acetate buffer (pH 3.6), and 20% polyethylene glycol 8000. The obtained crystal was transferred to a cryo-solution of 20% glycerol, 50 mM sodium acetate buffer (pH 3.6), and 20% polyethylene glycol 8000. Crystals of AgIMD-isomaltose and AgIMDpanose were obtained by soaking in well solutions containing 30% (w/v) isomaltose and panose, respectively, for a few seconds. The solution containing the ligand also acted as a cryoprotectant. Diffraction data were collected on the beamlines AR-NW12A and AR-NE3A at the Photon Factory (Tsukuba, Japan). All data were processed and scaled using HKL2000 (18). The initial phases were calculated from the single wavelength anomalous dispersion data set using the AutoSol program in the PHENIX suite (19). The coarse model obtained was applied for the molecular replacement method with MOLREP (20) in the CCP4 program suite (21). The model was refined using REFMAC5 in the CCP4 suite, and manual adjustment and rebuilding of the model were carried out using the program COOT (22). Solvent molecules were introduced using the program ARP/wARP (23). Validation of the structures was performed using the MolProbity server (24). Figures were prepared using PyMOL, Caver (26), and LigPlot (27). The data collection and refinement statistics are summarized in Table 1.
Measurement of Enzymatic Activity and Protein-The enzymatic activity of AgIMD was measured as described (16). Briefly, the activity of dextran T2000 (GE Healthcare, Chalfont St. Giles, UK) was measured in 100 mM sodium acetate buffer (pH 5.3) for 30 min at 30°C. Preparation and measurement of the cleavage of p-nitrophenyl ␣-D-isomaltoside (pNP-IM) was also performed as described (16). The protein concentration was determined by measuring the absorbance at 280 nm, using the molar extinction coefficient (1 mg/ml ϭ 2.487) calculated by the ExPASy ProtParam server. Kinetic parameters were calculated by nonlinear regression analysis using KaleidaGraph (Synergy Software, Reading, PA).

Results and Discussion
Overall Structure of AgIMD-The recombinant AgIMD was expressed in E. coli and affinity-purified by Ni-NTA-agarose chromatography. The N-terminal His tag was then removed by thrombin cleavage, and the protein was crystallized. The crystals belonged to the tetragonal space group I4 1 with one monomer in the asymmetric unit. The structure was determined using the single wavelength anomalous dispersion technique with a crystal soaked in lead acetate solution. A coarse model of AgIMD was initially built with a 2.5-Å resolution dataset, and the model was further refined using a 1.44-Å resolution dataset ( Table 1). The 2F o Ϫ F c electron density contoured at 1 showed continuous density for almost all the main chain atoms, but the N-terminal segment, GSHMATAVTARPGV, was not visible. The Ramachandran plot calculated with the Molprobity server shows that only one residue, Asp-312, was identified as an outlier. Asp-312 is one of the key amino acid residues in the active site, and the electron density for this residue was well defined.
The structure of AgIMD consists of three domains, designated domain A (residues 21-367), domain C (368 -466), and domain D (467-606), as well as an extra loop comprising residues 11-20 in the N terminus (Fig. 1A). Domain A is composed of a (␤/␣) 8 -barrel, and domain C is made up of an antiparallel ␤-structure. This two-domain architecture is commonly found in GH27 enzymes (25). An extra structural component (residues 123-178) is present in domain A, and it is here designated subdomain B because of its resemblance to those found in GH13 and GH31 (described below). Domain D has been classified as CBM35 in the CAZy database, and like other CBM35 structures (14,15) domain D is composed of a ␤-sandwich fold.
A structural similarity search of domain A (including subdomain B), domain C, and domain D was performed using the DALI server (28). Domain A shows a significant similarity to the domains of GH27, GH36, GH31, and GH13 enzymes, all of which consist of (␤/␣) 8 -barrel folds ( Table 2). Among the GH27 enzymes, similarities with ␤-L-arabinopyranosidases (25,29) were relatively higher than those with ␣-N-acetylgalactosaminidases and ␣-galactosidases (30). Domain C is similar to the antiparallel ␤-sheet domains of various GH family enzymes. The results of the similarity search of domain D were markedly different from those of domains A and C, and other GH27 members do not possess CBM35, as described below.

Interactions with Isomaltose and Panose in the Active Site-
The crystal structure of the AgIMD-isomaltose complex is almost isomorphous with that of the unliganded form, and clear electron density maps (F o Ϫ F c , 3 ) were obtained for two isomaltose molecules (Fig. 1, B and C). One of the molecules is seen in the active site, and the other molecule is bound to domain D (Fig. 1A). The active site structure of AgIMDisomaltose was compared with that of unliganded AgIMD. GH27 enzymes employ a retaining mechanism, and two Asp residues act as catalytic residues (31,32). The structural homology with GH27 ␣-galactosidases indicates that Asp-197 and Asp-265 are identified as a nucleophile and acid/base catalyst, respectively. Based on the position of the catalytic residues, isomaltose is bound to subsites Ϫ2 and Ϫ1 in AgIMD-isomaltose, and the two glucose residues are labeled Glc Ϫ2 and Glc Ϫ1 ( Fig. 2A). The binding of isomaltose is accompanied by conformational changes in Glu-81, Val-242, and Met-243. The side chain of Glu-81 in unliganded AgIMD points away from the active site, whereas atoms OE1 and OE2 of Glu-81 in AgIMDisomaltose form hydrogen bonds with atoms O6 and O4 of Glc-2, respectively. Also, the main chain atom O of Val-242 is not oriented toward the active site in unliganded AgIMD, whereas in the AgIMD-isomaltose, significant conformational differences of Val-242 and Met-243 are observed, and atom O of Val-242 interacts with atoms O2 of Glc Ϫ1 via a water molecule. We also determined the crystal structure of AgIMD complexed with a trisaccharide, panose (Glc-␣(136)-Glc-␣(134)-Glc). AgIMD hydrolyzes panose, albeit inefficiently, to produce isomaltose and glucose (4). Electron density for two ligand molecules was seen in AgIMD-panose (Fig. 1, D and E), and as with AgIMD-isomaltose, one molecule was found in the active site, and the other molecule was located on domain D. In the active site, three glucose residues were modeled at subsites Ϫ2, Ϫ1, and ϩ1. The glucose residues at subsites Ϫ2 and Ϫ1 found in AgIMD-panose and those found in AgIMD-isomaltose are well superimposed (Fig. 3A). The obtained structural models of domain D in the AgIMD-panose complex is nearly identical to that observed in the AgIMD-isomaltose complex, because a glucose residue at the reducing end was disordered, and the molecule bound to domain D was modeled as isomaltose (Fig. 1E).
Tunnel Structures-There are two tunnels, designated tuunel-1 and tunnel-2, near the methionine residue at position 243 in the active site of AgIMD, and several water molecules are found in the tunnels (Fig. 3B). The methionine residue is oxidized based on the interpretation of the electron density map (Fig. 1, F-H), and thus the residue is described as Sme243 in this paper. There is another possibility that residue 243 is present in two conformations simultaneously. However, the distance between an atom in the side chain of the 243th residue and atom N of Ser-200 is 3.0 Å, and also that between the same atom in the side chain of the 243th residue and atom N of Trp-201 is 3.0 Å. The observation suggests that hydrogen bonds are likely to be present between these atoms, and thus the 243th residue is identified as methionine sulfoxide. It is interesting to note that an elaborate conformational change is seen in tunnel-1. It is unclear whether Sme243 was oxidized spontaneously or by x-ray radiation. A GH27 ␣-galactosidase from Trichoderma reesei showed a 12-fold increase in activity when treated with H 2 O 2 (33), and Met-258 is considered to be possibly oxidized based on the structure modeling (34). However, the residue corresponding to Met-258 in T. reesei ␣-galactosidase is identified as Phe-313 in AgIMD. In a GH26 ␤-mannanase from the termite Reticulitermes speratus, a methionine sulfoxide residue, Sme85, is located in the active site (35), but the role of Sme243 is obviously different from that of Sme85 in the GH26 ␤-mannanase, because Sme243 does not form any hydrogen bonds with the ligand molecule (isomaltose or panose) in the ligand-bound form, unlike in GH26 ␤-mannanase.
It is also difficult to explain the role of the tunnels of AgIMD. It is unlikely that the row of water molecules present in the tunnels directly interacts with the acid-base catalyst, Asp-265, as the distance between atom OD1 of Asp-265 and the row of the first water molecule is more than 10 Å. Similar tunnel structures have been reported to be observed in GH13 (36,37), GH48 (38), and GH68 (39). These water paths have been proposed to function as a water drain and/or a water reservoir, and the tunnels of AgIMD might have similar roles.
Comparison of the Catalytic Domains of AgIMD and GH27 ␣-Galactosidase-The catalytic domain of GH27 human ␣-galactosidase (hGAL) is one of the most structurally homologous proteins to that of AgIMD, and its catalytic mechanism has been extensively studied (30,32). The structures of the catalytic domains of AgIMD and hGAL were superimposed (Fig. 4A). Although the fold of AgIMD is basically identical to that of hGAL, three components comprising residues 123-178, 204 -216, and 276 -289 were not found in hGAL, and thus these components are designated here as subdomain B, loop-1, and loop-2, respectively. Subdomain B of AgIMD is located between the third ␤-strand (␤3) and the third ␣-helix (␣3) of the (␤/␣) 8 -barrel. A short loop comprising residues 136 -151 is present instead in the corresponding position of hGAL, and this loop is described as loop-A in this paper. Loop-1 of AgIMD is present between ␤4 and ␣4, and loop-2 of AgIMD is observed as an insertion of ␣6.
To compare the residues involved in each subsite of AgIMD and those of hGAL, residues interacting with panose (Fig. 3C) and the corresponding residues in hGAL are listed (Table 3). At subsite Ϫ1, four residues are conserved between AgIMD and hGAL (Asp-77/92, Tyr-120/134, Arg-261/277, and Asp-312/ 266), despite the fact that glucose binds at subsite Ϫ1 of AgIMD, whereas galactose binds at subsite Ϫ1 of hGAL. Residues at subsite Ϫ2 of AgIMD are completely different from those of hGAL. At subsite Ϫ2 of AgIMD, atoms OE1 and OE2 of Glu-81 directly form hydrogen bonds with O6 and O4 of Glc Ϫ2, respectively, and also atom N of Trp-79 directly forms hydrogen bonds with O6 and O5 of Glc Ϫ2 (Fig. 3C). The tryptophan residue, Trp-79, appears to be the key residue of subsite Ϫ2 and is involved in substrate stacking interactions (Fig. 5A). In contrast, ␣-galactosidase has been reported to be an enzyme that catalyzes the hydrolysis of galactosyl residues from the non-reducing end of a variety of oligosaccharides and polysaccharides, and thus subsite Ϫ2 is unnecessary for hGAL. In fact, the position equivalent to the Glc Ϫ2 binding cleft of AgIMD is occupied by residues Cys-142 and Ala-143, which are part of loop-A in hGAL, and therefore no binding cleft for subsite Ϫ2 is found in hGAL (Fig. 5B).
Residues involved in subsite ϩ1 of AgIMD are also different from those of hGAL. The report of the hGAL-melibiose structure indicated that few interactions with the glucose portion of melibiose have been found (32). However, five residues appear to be involved in the binding of Glc ϩ1 in AgIMD. Atom O2 and atom O3 of Glc ϩ1 directly form hydrogen bonds with atom OD2 of Asp-207 and atom NE1 of Trp-285, respectively, suggesting that Glc ϩ1 binds tightly to AgIMD (Fig. 3C). The substrate, isomalto-oligosaccharide, is expected to form a helixlike structure like panose, and Phe-198 is located at the center of the helical spiral of the substrate (Fig. 5A). Side chains of Asp-207, Asn-209, and Trp-285, which are involved in the binding of Glc ϩ1 of panose, are likely to form hydrogen bonds with Glc ϩ1 of isomalto-oligosaccharide either directly or  Site-directed Mutagenesis of Residues Unique to AgIMD-To assess the role of Asp-207, Asn-209, and Sme243, which are unique to AgIMD, alanine mutants D207A, D207A/N209A, and M243A were constructed. The kinetic parameters of wildtype and mutant AgIMD were determined (Table 4). For the small synthetic saccharide, pNP-IM, the K m values of all the mutants increased only slightly (less than 2-fold). The k cat values of D207A and D207A/N209A for pNP-IM decreased but not significantly (0.37-and 0.39-fold, respectively), whereas the M243A mutation drastically affected the k cat value for pNP-IM (8.2 ϫ 10 Ϫ3 -fold). For dextran, the K m values of all the mutants were almost identical to that of wild-type enzyme, and the k cat values decreased significantly (0.02-to 0.06-fold). The ratio of k cat /K m values of wild type:D207A:D207A/N209A:M243A for pNP-IM was 100:25:24:0.4, whereas that for dextran was 100: 3.6:5.6:2.1. The results suggest that Asp-207 is important for the catalysis of the polysaccharide, dextran, whereas Sme243 plays a significant role in the catalysis of both oligosaccharides and dextran.
Comparison of the Catalytic Domains of AgIMD and the GH13, GH31, and GH36 Enzymes-A search with the DALI server indicated that highly homologous proteins belonging to GH13, GH31, and GH36 are identified as Thermoactinomyces vulgaris ␣-amylase I (PDB code 2D0F; hereafter TVAI) (40), E. coli YicI (PDB code 1XSK; EcYicI) (11), and Lactobacillus acidophilus ␣-galactosidase (melibiase; PDB code 1ZY9; LaMel36A) (9), respectively ( Table 2). The (␤/␣) 8 -barrel domains of these proteins were superimposed, indicating that the backbone folds are similar among these enzymes (Fig. 4B). It is noteworthy that subdomain B of AgIMD shares structural homology with those of TVAI and EcYicI (Fig. 4, C-E), despite the low similarities in their primary structures (ϳ10%). In GH13 and GH31 enzymes, subdomain B is found between ␤3 and ␣3 of the barrel structure, which is the same position in AgIMD. The superimposition of TVAI-glucopentasaccharide, EcYicI-fluoroxylopyranosyl intermediate, and LaMel36A-galactose shows that the position and orientation of the pyranose ring at subsite Ϫ1 of AgIMD is almost identical to those of EcYicI and LaMel36A but is different from that of TVAI, and the structural similarity of the active site between AgIMD and TVAI is low (Fig. 4B).
The active sites of AgIMD, hGAL, EcYicI, and LaMel36A were compared (Fig. 5). In hGAL and LaMel36A, Trp-47(hGAL)-Trp-340(LaMel36A) is located near atom O4 of galactose at subsite Ϫ1, and Tyr-40(AgIMD) and Phe-515 (EcYicI) are located near atom O4 of Glc Ϫ1 in AgIMD and EcYicI, respectively. The finding suggests that this difference is important for the specificity for the C4-epimeric configuration of pyranose. At subsite Ϫ1, the residue equivalent to Phe-198 of AgIMD is found in EcYicI (Phe-417). Also, EcYicI possesses an extra domain in the N terminus, and residues in this N-terminal domain (Asp-185 and Tyr-194) appear to participate in subsite ϩ1, and thus, unlike in ␣-galactosidase, the N-terminal domain of EcYicI may perform a function similar to Asp-207, Asn-209, and Trp-285 of AgIMD. These observations led us to the conclusion that some key residues found in AgIMD, Phe-198 and Asp-207, are found in only GH31 enzymes and are not conserved in GH27 and GH36 enzymes. Domain D, CBM35-One of the most intriguing features of the structure of AgIMD is the presence of domain D at the C terminus. Structural homology analysis using DALI indicated that the highest Z-score was for a CBM35 domain of an uncharacterized protein Lmo2446 from Listeria monocytogenes EGD-e (PDB code 4KWU). This Lmo2446 protein showed the highest homology to a GH31 protein, 3-␣-isomaltosyltransferase from Sporosarcina globispora N75 (42), among the characterized proteins. The second highest Z-score match was for a CBM35 domain of a GH66 protein, cycloisomalto-oligosaccharide glucanotransferase from Bacillus circulans T-3040 (BcCIT) (43). An isomaltose molecule was identified in the F o Ϫ F c omit map of both AgIMD-isomaltose and AgIMD-panose (Fig. 1, C and E), and the two glucose residues are labeled Glc-(a) and Glc-(b) from the non-reducing end (Fig. 6B).
What is the role of the CBM35 domain D in AgIMD? CBM35(AgIMD), CBM35(Lmo2446), and CBM35(BcCIT) were superimposed and the CBM35 domains, and the catalytic domains are only illustrated in Fig. 6D to clarify the relations between these two domains. Despite the remarkable similarity of the CBM35 domains, the positions of the catalytic domains   (14). A surface model of AgIMD shows that there is a cleft, seeming suitable for binding a polysaccharide chain, near the sugar-binding site of CBM35(AgIMD) (Fig.  6E). However, this cleft is disconnected to the catalytic cleft, and thus it is unlikely that a linear polysaccharide chain directly moves from CBM35(AgIMD) to the catalytic cleft. Dextran has been reported to contain ␣-1,3and occasionally ␣-1,2and ␣-1,4-branched linkages (1, 44), and also this polysaccharide is known to function as a component of bacterial biofilm matrix (44). Therefore, the polysaccharide chains could adopt complicated structures and tend to be tangled with each other. Although the overall structure is different, a GH16 ␤-agarase has an extra substrate-binding site, and an unwinding mechanism for agarose chains has been proposed (45). It is likely that CBM35(AgIMD) could help to unwind the tangled polysaccharide chains, similar as proposed in the ␤-agarase. A BLAST homology search was carried out to elucidate the physiological function of AgIMD, resulting in some bacteria possessing genes homologous to the AgIMD gene. The most homologous genes were the Caci_6974 gene from Catenulispora acidiphila DSM44928 (48% similarity) (46) and the BN506_02253 gene from Bacteroides cellulosilyticus CAG:158 (45% similarity), although no CBM35 domain was found in Caci_6974 and BN506_02253. A gene encoding GH66 putative endodextranase (Caci_6973 from C. acidiphila) or GH97 putative ␣-glucosidase (BN506_02255 from B. cellulosilyticus) was found near the putative isomalto-dextranase genes, whereas no gene encoding GH70 putative dextransucrase was identified in these bacteria. Some bacteria preferentially utilize uncommon sugars such as cyclodextrins, which has been proposed to be beneficial for surviving in a competitive environment (47). It is likely that Caci_6974 and BN506_02253 are involved in efficient utilization of dextran as a special energy source. Although some dextranases have been reported to influence the formation of bacterial biofilm (48), Caci_6974 and BN506_02253 may not participate in the biofilm formation because the organisms do not possess GH70 enzymes. AgIMD is also likely to play a role similar to that of Caci_6974 and BN506_02253, and perhaps CBM35(AgIMD) is an apparatus for accelerating the hydrolysis of the complicated polysaccharide structure.
CBMs are often found in biomass-degrading enzymes, and engineering of CBMs is expected to be beneficial for the production of biofuels (41,49). Despite the high homology with other CBM35 domains, CBM35(AgIMD) has unique features, such as no calcium-binding site and the presence of a long loop. Thus, our results provide new information for biotechnological engineering of CBMs.
Conclusions-The crystal structure of AgIMD was determined. Although the structure most closely resembles GH27 enzymes, domain A of AgIMD has subdomain B, loop-1, and loop-2, all of which are not found in GH27 hGAL. The fold of subdomain B is basically identical to those of GH31 EcYicI and GH13 TVAI. Four residues at subsite Ϫ1 in AgIMD are con- , and other key residues are illustrated. Colors: magenta, catalytic Asp residues; cyan, residues at subsite Ϫ1; blue, residues at plus subsites; green, other key residues. The subsite numbers of bound saccharides are labeled.

TABLE 4
Kinetic parameters of wild-type and mutant AgIMD served in hGAL, whereas the residues involved in subsites Ϫ2 and ϩ1 in AgIMD are completely different from those in hGAL. Site-directed mutagenesis showed that Asp-207 at subsite ϩ1 in AgIMD is important for the catalysis of the polysaccharide dextran, and the corresponding aspartic acid residue, Asp-185, is present in GH31 EcYicI. The structural feature of domain D, CBM35(AgIMD), is highly homologous to those of CBM35 (Lmo2446) and CBM35(BcCIT), which are domains present in the GH31 enzyme Lmo2446 and the GH66 enzyme BcCIT, respectively. These observations lead us to the conclusion that AgIMD has some features found in GH31, GH13, and GH66 enzymes, despite the fact the overall structure most closely resembles GH27 enzymes. GH27, -31, -36, and -66 enzymes have been proposed to share a common origin with those of the GH13 family, and the architecture of AgIMD appears to provide such evidence.