Crystal Structures of the Catalytic Domain of a Novel Glycohydrolase Family 23 Chitinase from Ralstonia sp. A-471 Reveals a Unique Arrangement of the Catalytic Residues for Inverting Chitin Hydrolysis

Background: Chitinase C from Ralstonia sp. A-471 (Ra-ChiC) is a chitinase that was first found in glycohydrolase family 23. Results: The crystal structure of Ra-ChiC exhibited a tunnel-shaped conformation in its active site. Conclusion: The tunnel-shaped conformation is essential for a unique arrangement of the catalytic residues and substrate specificity. Significance: This is the first report on the tunnel-shaped binding site of an inverting chitinase. Chitinase C from Ralstonia sp. A-471 (Ra-ChiC) has a catalytic domain sequence similar to goose-type (G-type) lysozymes and, unlike other chitinases, belongs to glycohydrolase (GH) family 23. Using NMR spectroscopy, however, Ra-ChiC was found to interact only with the chitin dimer but not with the peptidoglycan fragment. Here we report the crystal structures of wild-type, E141Q, and E162Q of the catalytic domain of Ra-ChiC with or without chitin oligosaccharides. Ra-ChiC has a substrate-binding site including a tunnel-shaped cavity, which determines the substrate specificity. Mutation analyses based on this structural information indicated that a highly conserved Glu-141 acts as a catalytic acid, and that Asp-226 located at the roof of the tunnel activates a water molecule as a catalytic base. The unique arrangement of the catalytic residues makes a clear contrast to the other GH23 members and also to inverting GH19 chitinases.

Chitin, a ␤-1,4 polymer of N-acetyl-D-glucosamine (NAG), 3 is one of the most abundant biopolymers in nature. It is easily obtained from natural resources such as shrimp and crab shell waste. It is estimated that at least 10 gigatons (1 ϫ 10 13 kg) of chitin are synthesized each year, of which the vast majority is discarded (1). Chitin could be a significant biological resource because N-acetyl-chitooligosaccharides ((NAG) n ) and chitooligosaccharides, products of chitin hydrolysis, have a variety of biological functions including antibacterial activity and antitumor activity (2). In nature, chitin is degraded by chitinases (EC 3.2.1.14) that belong to glycohydrolase (GH) families 18 and 19 (3,4). These two families do not have sequence or structural similarities and act via different catalytic mechanisms. GH family 18 chitinases are distributed in various organisms from microbes to human. Their catalytic domain has an (␣/␤) 8 -barrel structure and utilizes a double-displacement mechanism, which results in the net retention of an anomeric configuration (retaining mechanism) (5)(6)(7). In contrast to GH18 chitinases, GH family 19 chitinases are found only in plants and some bacteria (8,9). Their overall structures show high ␣-helical contents (10 -12), and possess an active site topology similar to those of lysozymes from hen (GH22), goose (GH23), and phages (GH24) (13)(14)(15). GH family 19 chitinases utilize a single-displacement mechanism, which results in an inverted anomeric configuration (inverting mechanism) (16).
We previously cloned the gene encoding a chitinase from a moderate thermophilic strain Ralstonia sp. A-471 (Ra-ChiC) (17). Ra-ChiC comprises a putative signal peptide (residues 1-35), a chitin-binding domain (residues 36 -80), an interdomain linker (residues 81-102), and a catalytic domain (residues 103-252). Ra-ChiC produces ␣-anomers by hydrolyzing ␤-1,4glycosidic linkages, indicating that the enzyme is an inverter like GH family 19 chitinases. Ra-ChiC is stable up to 50°C and retains considerably high activity when incubated with high concentrations of organic solvents such as 50% acetonitrile, 50% methanol, and 50% ethanol. This enzyme is also stable over pH range 5.0 -10.0. In addition, the enzyme is also active toward various chitin derivatives, including ethylene glycol chitin, CM-chitin, and soluble chitin (50% deacetylated chitin). Because the enzymatic activity is barely affected by metal ions, except for Hg ϩ , Ra-ChiC could be useful for industrial applications (17).
Ra-ChiC is the first chitinase classified as GH family 23 (CAZy database) (18), which has been recognized to include goose-type (G-type) lysozymes, peptidoglycan lyases, and peptidoglycan lytic transglycosylases. Ra-ChiC cleave the ␤-1,4glycosidic bonds between NAG residues in chitin, whereas the other GH23 members act toward the bonds between N-acetyl-D-muramic acid (NAM) and NAG in peptidoglycan. Because of the chemical similarities of their substrates, some chitinases can hydrolyze peptidoglycans less efficiently in addition to chitin and, conversely, some lysozymes can hydrolyze chitin (19 -22). However, Ra-ChiC exhibits only chitinase activity but not the activity toward peptidoglycan (17). In addition, Ra-ChiC differs from G-type lysozymes in their splitting mode of NAG oligosaccharide substrates, (NAG) n : for example, the NAG hexamer substrate, (NAG) 6 , is split predominantly into (NAG) 3 ϩ (NAG) 3 by G-type lysozymes, whereas the same substrate is split also into (NAG) 2 ϩ (NAG) 4 by Ra-ChiC. The subsite structure of Ra-ChiC appears to differ from those of G-type lysozymes.
Inverting glycoside hydrolases have been recognized to require two acidic amino acids for their catalytic reaction; one acts as a general acid and the other as a general base. The distance between the carboxyl groups of acidic residues was proposed to be ϳ8 -10 Å (23,24). G-type lysozymes have a highly conserved glutamic acid residue, Glu 73 , as a general acid (25)(26)(27). In addition, Helland et al. (28) suggested that Asp 101 in Atlantic cod G-type lysozyme (AcLYZ), which is at a distance of 8.5 Å from Glu 73 acts as a general base in the catalytic reaction. Sequence alignments of Ra-ChiC and AcLYZ suggested that Glu 73 and Asp 101 of AcLYZ are conserved in Ra-ChiC as Glu 141 and Glu 162 , respectively (17), which are the best candidates for catalytic residues. However, this is only a suggestion from sequence comparison, and the catalytic residues of the unusual chitinase, Ra-ChiC, remains to be elucidated by the structural and mutation analyses.
Here we report the binding specificity of Ra-ChiC as determined by NMR spectroscopy and the first crystal structures for the catalytic domain of Ra-ChiC in the presence and absence of (NAG) 2 , an E141Q mutant of the catalytic domain in the presence of (NAG) 4 , and an E162Q mutant of the catalytic domain in the presence and absence of (NAG) 2 . We found that Ra-ChiC has a unique substrate-binding site compared with other GH family 23 enzymes. In addition, we carried out a mutation analysis of acidic amino acid residues located at the active site, and found that Glu 141 is a catalytic acid. Contrary to the suggestion from the sequence comparison, Asp 226 rather than Glu 162 was found to act as a catalytic base. Based on structural and mutation analyses, we also discuss a catalytic mechanism for Ra-ChiC.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Protein expression and purification of Ra-ChiC segments, which included residues 36 -252 (Ra-ChiC mature ) and 89 -252 (Ra-ChiC cat ) with an N-terminal His 6 tag, were previously described (29). The mutants derived from Ra-ChiC mature and Ra-ChiC cat were produced by site-directed mutagenesis using a Stratagene QuikChange II kit (Stratagene, La Jolla, CA). The forward and reverse oligonucleotide primers are shown in supplemental  Table S1. The mutants E141Q, E141A, E141D, E141N, E162Q, E162A, E162D, E162N, D226N, and D226A were created using the synthesized oligonucleotide primers. Selenomethionine-labeled Ra-ChiC cat was overexpressed using the methionine pathway inhibition in LeMaster Broth (30 -32). The E141Q, E162Q, and selenomethionine Ra-ChiC cat proteins were purified using the same procedure as wild-type Ra-ChiC cat . To produce 15 N-labled Ra-ChiC mature , Escherichia coli cells harboring the expression plasmid were grown in M9 minimal medium containing 15 N-NH 4 Cl (0.5 g/liter). Production and purification of the labeled Ra-ChiC mature was done as described above.
Crystallization-Purified Ra-ChiC cat was concentrated to 4.5-6.5 mg/ml for crystallization. Crystallization conditions are shown in supplemental Table S2. Crystallization was carried out by the hanging-drop vapor diffusion method at 20°C. All drops included equal volumes of the protein and reservoir solutions. Crystals of E141Q and E162Q mutants were obtained by the micro-seeding method using wild-type crystals as a seed, and the (NAG) n -bound forms were obtained by the soaking method.
Data Collection, Processing, Phasing, and Structure Refinement-Diffraction data were collected at 100 K with 20% (v/v) 2-methyl-2,4-pentanediol in the mother liquid as a cryoprotectant at beamlines BL38B1 and BL41XU at SPring-8 (Harima, Japan) and at beamlines NW12A, NE3A, BL-1A, BL-5A, and BL-17A at Photon Factory (Tsukuba, Japan). The data sets for the selenomethionine crystal were collected at 0.9641 Å (remote (high)), 0.9950 Å (remote (low)), 0.9789 Å (peak), and 0.9730 Å (edge) for multiple wavelength anomalous diffraction phasing (supplemental Table S3). Data were processed and scaled using the HKL2000 suite (33). Initial phasing, phase improvement, and auto model building were performed by SOLVE, DM, and buccaneer, respectively (34 -36). The crystals belonged to space group P6 1 22 with unit cell parameters a ϭ b ϭ 100 Å, c ϭ 243 Å and contained four Ra-ChiC cat molecules (MolA, MolB, MolC, and MolD) in the asymmetric unit. Refinement was carried out using REFMAC5, including refinement of atomic displacement parameters by translation, liberation, and screw method, with each monomer in the asymmetric unit treated as a single translation, liberation, and screw group (37). Manual model modification was performed by Coot (38). Data collection statistics and refinement parameters are given in Table 1.
Catalytic Activity against Ethylene Glycol Chitin-According to the method described previously (39), the enzyme activities of wild-type and 10 mutants of Ra-ChiC mature were determined using ethylene glycol chitin as the substrate in 0.1 M sodium phosphate buffer, pH 6.0, at 37°C. One unit of activity was defined as the amount of enzyme that liberated 1 mol of the reducing sugar per minute.
Catalytic activity against (NAG) n -Time courses of (NAG) n (n ϭ 3, 4, 5, and 6) hydrolysis catalyzed by wild-type Ra-ChiC mature were determined by quantifying the amounts of substrate and products by HPLC. The enzymatic reaction was performed in 50 mM acetate buffer, pH 6.0, at 40°C for a given incubation time. Enzyme and substrate concentrations were 0.82 M and 4.75 mM, respectively. The reaction mixture was rapidly chilled in liquid nitrogen to terminate the enzymatic reaction, and then applied to HPLC using a column of TSK-gel NH 2 -60 (Tosoh Co.). Elution was carried out with 70% (v/v) acetonitrile at a flow rate of 0.8 ml/min. (NAG) n fragments were monitored by absorption at 220 nm. From the peak area obtained by HPLC, (NAG) n concentrations at each reaction time were calculated using a standard curve obtained with an authentic saccharide solution, and plotted against the reaction time.

RESULTS
Interaction Analysis by NMR Spectroscopy-1 H-15 N heteronuclear single quantum coherence spectra of the 15 N-labeled Ra-ChiC mature were obtained in the presence of (NAG) 2 (0.6 or 1.2 mM) or NAG-NAM-AiQ (0.6 or 1.2 mM). The overlaid spectra are shown in Fig. 1. Only a trivial change was observed, when the NAG-NAM-AiQ was added to the protein solution (Fig.  1A). However, the numerous resonances were affected upon addition of (NAG) 2 (Fig. 1B). This indicates that Ra-ChiC specifically interacts with the chitin oligomer, but not with the peptidoglycan fragment.
Crystal Structures of Wild-type Ra-ChiC cat -The structure of wild-type Ra-ChiC cat (WT) contains four molecules per asymmetric unit, and these molecules are superposed each other with root mean square deviation (r.m.s. deviation) values from 0.3 to 0.4 Å for the 152 corresponding C␣ atoms. The N-terminal region of MolA penetrates to and interacts with an adjacent molecule, MolB, and electron densities are observed from R free was calculated from the test set (5% of the total data).
In the structure of WT complexed with (NAG) 2 , WT-NAG2, electron densities corresponding to (NAG) 2 molecules are clearly observed in Ϫ3 to Ϫ2 and ϩ1 to ϩ2 subsites (Fig. 3A,  left), using the nomenclature proposed by Davies et al. (40). Although the overall structure of WT was not affected upon (NAG) 2 binding (r.m.s. deviation value of 0.3 Å for 153 C␣  (45). Acidic residues located in the active site are shown in red. Conserved residues are shown by the magenta background. Residues that form the secondary structures of Ra-ChiC and AcLYZ are indicated above and below, respectively. L-7 is indicated by a red box. atoms), only the main chain of Gly 147 -Gly 150 on loop 3 (L-3) had slightly moved away (Ϸ1.1 Å) from the disaccharide bound at ϩ1 to ϩ2 subsites (supplemental Fig. S1).
The tunnel-shaped cavity formed by L-7 represents a substrate-binding site. NAG residues that occupied the Ϫ2 and ϩ1 subsites are almost buried within this cavity (Fig. 3A, left). In the Ϫ2 subsite, six direct and four water-mediated hydrogen bonds were observed between the NAG residue and WT ( Table 2, supplemental Fig. S2). Glu 162 , which is a conserved acidic residue in G-type lysozymes strongly contributes to substrate recognition at the Ϫ2 subsite by forming three direct hydrogen bonds. In the ϩ1 subsite, seven direct and one water-mediated hydrogen bonds were observed. The side chain oxygen (O⑀2) of Glu 141 , a highly conserved glutamic acid, is 2.7 Å from the O4 atom of the NAG residue at the ϩ1 subsite. In contrast, NAG residues at the Ϫ3 and ϩ2 subsites are exposed to solvent (Fig.  3A, left), and recognition of these NAG residues are relatively loose. An NAG residue at the Ϫ3 subsite forms only one direct and three water-mediated hydrogen bonds. An NAG residue at the ϩ2 subsite is recognized by this enzyme via two direct hydrogen bonds and one water-mediated hydrogen bond. In fact, in the WT-NAG2 structure, the average B-factor values of NAG residues at the Ϫ3 and ϩ2 subsites are higher than those at the Ϫ2 and ϩ1 subsites (Ϫ3 subsite: 35 4 are observed well at Ϫ3 to ϩ1 subsites except for a part of an NAG residue at the ϩ1 subsite (Fig. 3A,  right). The overall structure of E141Q-NAG4 is essentially identical with that of the WT structure (r.m.s. deviation value of 0.2 Å for 153 C␣ atoms), whereas a slight conformational change is observed in L-7 and L-3 (supplemental Fig. S1). The maximum shift of 0.8 and 0.5 Å were confirmed at the C␣ atom position of Leu 148 in L-3 and Pro 231 in L-7, respectively. Because such a conformational change in L-7 is not observed in the WT-NAG2 structure, those conformational changes in L-3 and L-7 seem to be derived from the NAG residue binding at ϩ1 and Ϫ1 subsites, respectively. In addition, the positions of NAG residues at the Ϫ3 and Ϫ2 subsites are slightly different between the WT-NAG2 and E141Q-NAG4 structures. The mean distances of the corresponding atoms of NAG residues at the Ϫ3 and Ϫ2 subsites in these structures are 0.4 and 0.6 Å, respectively, when the overall structures are superposed. These shifts of NAG residues may be influenced by the presence of an NAG residue at the Ϫ1 subsite.  From the structural differences noted above, the interactions between this enzyme and NAG residues at these subsites observed in the E141Q-NAG4 structure are also slightly different from those in the WT-NAG2 structure. In the E141Q-NAG4 structure, an additional hydrogen bond was observed between O6 of the NAG residue at the Ϫ3 subsite and O␥1 of Thr 165 , whereas a direct hydrogen bond and two water-mediated hydrogen bonds observed at O1 and O6 of the NAG residue at the Ϫ2 subsite were not observed ( Table 2, supplemental Fig. S3). The NAG residue at the Ϫ1 subsite is completely buried within the tunnel-shaped cavity of E141Q and adopts a chair conformation (Fig. 3A, right). Direct hydrogen bonds are observed between O1 and N⑀2 of Gln 141 , N2 and O␦1 of Asp 226 , O3 and the main chain carbonyl of Asn 215 , and O6 and O⑀1 of Gln 141 . In addition, two water-mediated hydrogen bonds were observed at O7 of the NAG residue ( Table 2).
Crystal Structures of the E162Q Mutant of Ra-ChiC cat -In the structure of the E162Q mutant in the absence of (NAG) n , E162Q-HEPES, electron density corresponding to the HEPES molecule derived from the mother liquid for crystallization was observed over the Ϫ3 to Ϫ1 subsite of three E162Q molecules (MolA, MolC, and MolD) in the asymmetric unit. As was observed in the WT structure, the N-terminal region of MolA penetrates to the substrate-binding site of MolB. In MolA and MolD of E162Q-HEPES, the characteristic structural change observed at ␣3 and L-7 results in a slight exposure of the substrate-binding site compared with other structures (supplemental Fig. S1). In contrast, this exposure is not observed in the MolC structure despite the presence of a HEPES molecule. The C␣ distances of Gln 162 to Ser 227 and Asn 156 to Tyr 228 in MolD are 2.4 and 1.4 Å, respectively, longer than those in the WT-NAG2 structure and, as a result, hydrogen bonding interactions between these residues are broken in the E162Q-HEPES structure as shown in Fig. 3B.
In the structure of the E162Q mutant complexed with (NAG) 2 , E162Q-NAG2, electron densities corresponding to (NAG) 2 molecules were observed in the Ϫ3 to Ϫ2 and ϩ1 to ϩ2 subsites analogous to the WT-NAG2 structure. In contrast to the MolA and MolD in the E162Q-HEPES structure, the structure of E162Q-NAG2 is quite similar to the WT-NAG2 structure including L-3, L-7, ␣3, and NAG molecules (r.m.s. deviation value of 0.1 Å for 153 C␣ atoms) (supplemental Fig. S1). The hydrogen bonding pattern between NAG residues and WT in the WT-NAG2 structure (listed in Table 2) is conserved in the E162Q-NAG2 structure even with the substitution of Glu 162 to Gln (Fig. 4, A and B).
Mutation Analysis of Ra-ChiC-Mutation analyses of Glu 141 and two acidic residues (Glu 162 and Asp 226 ) located within 10 Å distance from Glu 141 were performed to determine the catalytic residues of Ra-ChiC. We prepared 10 single mutants of Ra-ChiC mature (E141A, E141Q, E141D, E141N, E162A, E162Q,  E162D, E162N, D226A, and D226N) and determined their chitinolytic activities against glycol chitin ( Table 3). The E141A, E141Q, and E141D mutations resulted in Ͼ250-fold decreased chitinolytic activity, and the E141N mutation decreased the activity Ͼ100-fold. In addition, two mutants relevant to Asp 226 (D226A and D226N) also decreased the catalytic activities by over 100-fold. These results indicated that Glu 141 and Asp 226 have an important role in Ra-ChiC catalytic activity. In contrast, the E162A, E162D, and E162N mutants retained moderate activity, and the catalytic activity of the E162Q mutant was 2.5-fold higher than that of wild-type. Steady state kinetics of the enzymatic reaction toward glycol chitin did not afford accurate values of k cat and K m for the mutant enzymes, except for Glu 162 mutants, because of their low activity.
Catalytic Activity of Ra-ChiC against (NAG) n -Time courses of NAG oligomer hydrolysis catalyzed by Ra-ChiC mature were determined by HPLC, and the results are shown in Fig. 5. Specific activities toward (NAG) 3 , (NAG) 4 , (NAG) 5 , and (NAG) 6 , were calculated to be 16.6, 31.6, 52.6, and 52.6 mM/min/mg, respectively, based on the initial velocity of the substrate degradation. (NAG) 3 was hydrolyzed into (NAG) 2 ϩ NAG (Fig.  5D). As reported in a previous article (17), the rate of ␤-(NAG) 3 hydrolysis was higher than that of ␣-(NAG) 3 , and the products (NAG) 2 and NAG were rich in ␣and ␤-forms, respectively.  Because Ra-ChiC is an inverting enzyme, (NAG) 3 is supposed to bind predominantly to Ϫ2, Ϫ1, and ϩ1 subsites. (NAG) 4 was hydrolyzed into (NAG) 2 ϩ (NAG) 2 and (NAG) 3 ϩ NAG (Fig.  5C). Because the products (NAG) 3 and NAG were rich in ␣and ␤-forms, respectively (17), (NAG) 4 is supposed to bind to Ϫ2, Ϫ1, ϩ1, and ϩ2 subsites and Ϫ3, Ϫ2, Ϫ1, and ϩ1 subsites. Thus the affinity at the glycon-binding site (negatively numbered subsites) is likely higher than that at the aglycon-binding site (positively numbered subsites). Profiles of the product formations from the substrates (NAG) 5 and (NAG) 6 were quite complicated because of the subsequent breakdown of the products of (NAG) 3 , (NAG) 4 , and (NAG) 5 (Fig. 5, A and B). In the early reaction stage, however, simple hydrolysis of the substrates was observed. (NAG) 5 was hydrolyzed frequently into (NAG) 2 ϩ (NAG) 3 , and into (NAG) 4 ϩ NAG in a lesser extent. All five glycosidic bonds in (NAG) 6 were able to be hydrolyzed by these enzymes. No evidence for transglycosylation was found from the time course data. The mutant enzymes except Glu 162 mutants (E162A, E162Q, E162D, and E162N) did not hydrolyze the NAG oligomers at all.

DISCUSSION
We have already reported that Ra-ChiC belonging to the GH-23 family hydrolyzes ethylene glycol chitin, carboxyl methyl chitin, and other soluble chitin but not the cell wall of Micrococcus lysodeikticus (17). Because the specificity was judged only from the data obtained with the polysaccharide substrates, it was still uncertain if Ra-ChiC interacts only with the chitinous substrates. Thus, we tested (NAG) 2 and NAG-NAM-AiQ for NMR titration experiments. As shown in Fig. 1, the backbone N-H resonances were strongly affected by the addition of (NAG) 2 but not by NAG-NAM-AiQ. Ra-ChiC indeed recognizes the chitin oligomer, but not the fragment derived from the lysozyme substrate, peptidoglycan.
Then we determined several structures for the catalytic domain of Ra-ChiC in the presence or absence of NAG oligomers. We also performed mutation analyses based on this structural information. The structure-function relationships provided important information for understanding the detailed catalytic mechanism and the unique substrate recognition mechanism of Ra-ChiC.
The overall structure of Ra-ChiC cat is similar to that of AcLYZ with a r.m.s. deviation value of 1.4 Å (Fig. 6A), and Glu 141 in Ra-ChiC corresponds to the catalytic acid Glu 73 in G-type lysozymes with respect to both the primary and tertiary structures (Figs. 2B and 6B). Indeed, the side chain atom of Gln 141 in the E141Q-NAG4 structure is located close enough to the O1 glycosidic oxygen for donating proton (Fig. 4C). Furthermore, mutation analyses showed that all Glu 141 mutations caused significant decreases in catalytic activities toward ethylene glycol chitin (Table 3). We tried to determine the steady state k cat and K m values. However, the very low activity of the Glu 141 mutants did not allow the accurate determination of the values for the glycol chitin substrate. For the NAG-oligomer substrates, no hydrolysis was observed for the Glu 141 mutants. On the other hand, binding experiments have been conducted by means of isothermal titration calorimetry, and we confirmed that the free energy changes for NAG oligomer binding were not affected by the mutation of Glu 141 . 4 Thus, we believe that the low activities of the Glu 141 mutants are exclusively derived from a decrease in k cat . These results indicated that Glu 141 of Ra-ChiC could act as a general acid. The isothermal titration calorimetry data will be reported elsewhere.
The carboxyl oxygen atoms of Glu 162 and Asp 226 are ϳ8.7 and 8.3 Å, respectively, from that of Glu 141 . A water molecule is observed between Glu 162 and Asp 226 as is the case for a catalytic water observed between Asp 101 and Asp 90 in the AcLYZ-NAG structure (PDB code 3GXR) in which (NAG) 2 and (NAG) 3 are bound at the Ϫ3 to Ϫ2 and ϩ1 to ϩ3 subsites (Fig. 6B). This water molecule is 3.8 Å from the C1 atom of the NAG residue at the Ϫ1 subsite in the E141Q-NAG4 structure, and consequently defined as the catalytic water (Fig. 4C). The catalytic water may be situated close to the C1 atom to be suitable for nucleophilic attack by the conformational change through the transition state. Asp 226 is located at L-7, and the apparent counterpart of Asp 226 is likely Glu 159 in AcLYZ based on the primary structure (Fig. 2B). However, based on the crystal structure (Fig.  6B), Asp 226 does not have any counterpart, and the carboxyl oxygen atom of Asp 226 is located only close to Asp 90 , which is involved in positioning the catalytic water in AcLYZ. By comparison, Glu 162 is located at a structurally analogous position to that of Asp 101 in AcLYZ, a putative general base. Both Glu 162 of Ra-ChiC and Asp 101 of AcLYZ are next to the highly conserved sequence found among GH family 23 members (Gly 157 -Gln 160 in Ra-ChiC numbering) (41), and sequence alignments of these enzymes indicated that Glu 162 of Ra-ChiC are equivalent to Asp 101 of AcLYZ (Fig. 2B). Nevertheless, mutation analysis showed that Glu 162 is not essential for the catalytic activity of Ra-ChiC (Table 3). Because Glu 162 forms direct hydrogen bonds with the NAG residue at the Ϫ2 subsite through its main and side chains in addition to a hydrogen bond with the catalytic water (Fig. 4), Glu 162 would be involved in both NAG recognition and positioning of this water. These roles of Glu 162 in Ra-ChiC correspond to those of Asp 90 in AcLYZ.
Mutation analysis also indicated that Asp 226 is critical for catalysis. Substitution of Asp 226 not only with Ala but also with Asn caused significant decreases in catalytic activity, suggesting that the side chain carboxyl group of Asp 226 is important for catalytic reactions. In the wild-type structure, the hydrogen bonding distance between Asp 226 and the catalytic water is somewhat long (3.5 Å), whereas the distance between Glu 162 and the water is 2.6 Å (Fig. 4A). However, Asp 226 could play the role of a general base, as the pK a value of Asp 226 , which is calculated as Ϫ2.7 to ϳ2.0, would be decreased by electrostatic interactions with the side chain of neighbor Arg 230 (Fig. 4). Interestingly, similar electrostatic interactions are also constructed between the putative general base Asp 101 and the neighboring Arg 103 in the AcLYZ structure. These interactions are observed for the catalytic residues in some inverting enzymes, including GH family 6 cellulase (PDB code 2BVW), GH family 8 cellulase (PDB code 1KWF), and GH family 46 chitosanase (PDB code 1CHK). Therefore, we propose a catalytic mechanism for Ra-ChiC based on structural analyses, enzymatic assay, and a theoretical model (supplemental Fig.  S4). Glu 141 donates a hydrogen atom to the scissile glycosidic bond. A water molecule that interacts with Asp 226 and Glu 162 via hydrogen bonds is activated by the general base of Asp 226 . Then, the nucleophilc water attacks the C1 atom of the NAG residue at the Ϫ1 subsite. As a result, an NAG oligomer with an ␣-anomeric configuration at the reducing end is produced. It should be noted that the general base Asp 226 is acting from a roof of the tunnel-shaped cavity formed by the loop, L-7. This is a unique arrangement of the catalytic residues for inverting enzymes.
As stated above, the NAG residue at the Ϫ1 subsite is completely buried within the tunnel-shaped cavity of E141Q and adopts a chair conformation. Although the general catalytic mechanism of inverting glycosidases is known, the transition state and the conformations adopted by the pyranosyl ring of the Ϫ1 sugar are still indefinite. Some reports on the inverting GH8 enzymes suggested that the Ϫ1 sugar ring displays an unfavorable boat 2,5 B conformation (42,43). We have been trying to get insight into the structural changes of the Ϫ1 sugar along the reaction pathway based on the crystal structure of the enzyme-NAG oligomer complexes. Unfortunately, however, the Ϫ1 NAG residue in the E141Q-NAG4 structure exhibits a favorable 4 C 1 chair conformation. It appears that the chair conformation of the Ϫ1 NAG residue reflects a nonproductive state, which does not undergo the catalytic reaction. We could not get any insight into the transition state from the crystal structures obtained in this study. Further crystallographic analysis is in progress using various mutant Ra-ChiCs.
We defined five substrate-binding subsites (Ϫ3 to ϩ2) in Ra-ChiC cat based on crystal structures, whereas it is well known that G-type lysozymes have six subsites. A comparison of the crystal structures for WT-NAG2 and AcLYZ-NAG illustrated that they have similar overall folds; in particular, five helices were superposed well (Fig. 6A). However, the substrate-binding site of Ra-ChiC appears to be shorter than that of AcLYZ because of a lack of two helices in its N-terminal region (␣1 and FIGURE 6. Structural comparison between Ra-ChiC cat and AcLYZ. A, superposition of the overall structures of WT-NAG2 (light green) and AcLYZ-NAG (orange). Bound NAG residues in WT-NAG2 and AcLYZ-NAG are shown as yellow and gray stick models, respectively. B, stereo diagram of active site superposition of the WT-NAG2 (light green) and AcLYZ-NAG (orange) structures. Acidic residues at each active site are shown as stick models. The catalytic water molecules observed in the WT-NAG2 and AcLYZ-NAG structures are shown as red and magenta spheres, respectively. AcLYZ residue numbers are indicated by asterisks.
␣2 of AcLYZ), resulting in a lack of a sixth subsite (ϩ3 subsite) in Ra-ChiC. Enzymatic assays for Ra-ChiC using (NAG) n showed that the activity of Ra-ChiC against (NAG) 5 was equivalent to that against (NAG) 6 . Thus, we concluded that the Ra-ChiC catalytic domain have five substrate-binding subsites (Ϫ3 to ϩ2) unlike G-type lysozymes. As noted above, the recognition of the NAG residues occupying the Ϫ3 and ϩ2 subsites is looser than those of the other NAG residues. Therefore, although Ra-ChiC has five subsites at maximum, the number of core subsites might be only three (Ϫ2, Ϫ1, and ϩ1). This proposition is consistent with the data obtained by HPLC-based time courses of NAG oligomer degradation; that is, (NAG) 3 binds predominantly to subsites Ϫ2, Ϫ1, and ϩ1, (NAG) 4 binds to subsites Ϫ3, Ϫ2, Ϫ1, and ϩ1 and subsites Ϫ2, Ϫ1, ϩ1, and ϩ2. When (NAG) 6 is hydrolyzed by Ra-ChiC, this enzyme produces (NAG) 4 , (NAG) 3 , and (NAG) 2 , and minimal amounts of (NAG) 5 and NAG (Fig. 5A). This phenomenon would be caused by multiple binding modes of (NAG) 6 due to the weak interactions with the sugar residues at both ends of the substratebinding site. L-7 in Ra-ChiC has the characteristic conformation compared with G-type lysozymes. The tunnel-shaped conformation of Ra-ChiC makes the Ϫ1 subsite quite narrow (Fig. 3A). Ra-ChiC does not have lysozyme activity that cleaves the glycosidic bond between NAM and NAG residues in peptidoglycans, components of bacterial cell walls. Because NAM has an additional large substituent group (lactic acid group) on C3 of its pyranose ring compared with NAG, NAM binding to such a narrow space in Ra-ChiC might be impossible. In fact, steric hindrance took place between the lactic acid group of the NAM molecule and L-7 when the NAG residue was replaced at the Ϫ1 subsite in the E141Q-NAG4 structure (Fig. 7A). For binding of the NAM molecule to the Ϫ1 subsite, a large conformational change of L-7 would be needed, but it may cause rearrangement of Asp 226 and result in inactivation of the enzyme. Although some lysozymes cleave the NAM-NAG bond, steric hindrance also took place between the NAM molecule and Ser 153 and Gln 160 of Ra-ChiC when the NAG residue at the ϩ1 subsite in the E141Q-NAG4 structure was replaced by the NAM residue (Fig. 7B). In nature, the natural substrates of lysozyme, peptidoglycans, form more bulky mesh-like layers consisting of glycan chains cross-linked with oligopeptides. Therefore, it appears that peptidoglycan binding to Ra-ChiC would be more difficult. By comparison, G-type lysozymes have open conformations at their active sites and could readily accommodate NAM in their Ϫ1 subsite. In fact, a hydroxyl group attached to C3 of the NAG residue occupying the Ϫ1 subsite orients toward solvent in the crystal structure of GEWL in complex with (NAG) 3 (PDB code 154L). Thus, we could replace the NAG residue with a NAM residue without steric hindrance (Fig. 7C). Hence, the characteristic L-7 conformation appears to generate the substrate specificity of Ra-ChiC.
In contrast, the tunnel-shaped conformation needs to be opened to accommodate (NAG) n in an endo-spliting manner. In fact, a semi-opened conformation resulting from structural shifts of L-7 and ␣3 in the E162Q-HEPES structure suggested that the tunnel-shaped conformation can be opened (Fig. 3B). This mechanism has also been proposed for the catalytic reac-tion of the processive endo-cellulase Cel48F, which has a tunnel-shaped active site (44).

CONCLUSION
Ra-ChiC is a chitinase that was first found in GH family 23 (goose-type lysozyme). The crystal structure of Ra-ChiC exhibited a tunnel-shaped conformation in its active site, which makes a clear contrast to the other GH23 members. The catalytic base is located at the roof of the tunnel, forming a unique arrangement of the catalytic residues for inverting chitin hydrolysis and determining the substrate specificity. This is the first report on the tunnel-shaped binding site of an inverting chitinase.