Solution structure of the chitin-binding domain of Bacillus circulans WL-12 chitinase A1.

The three-dimensional structure of the chitin-binding domain (ChBD) of chitinase A1 (ChiA1) from a Gram-positive bacterium, Bacillus circulans WL-12, was determined by means of multidimensional heteronuclear NMR methods. ChiA1 is a glycosidase that hydrolyzes chitin and is composed of an N-terminal catalytic domain, two fibronectin type III-like domains, and C-terminal ChBD(ChiA1) (45 residues, Ala(655)-Gln(699)), which binds specifically to insoluble chitin. ChBD(ChiA1) has a compact and globular structure with the topology of a twisted beta-sandwich. This domain contains two antiparallel beta-sheets, one composed of three strands and the other of two strands. The core region formed by the hydrophobic and aromatic residues makes the overall structure rigid and compact. The overall topology of ChBD(ChiA1) is similar to that of the cellulose-binding domain (CBD) of Erwinia chrysanthemi endoglucanase Z (CBD(EGZ)). However, ChBD(ChiA1) lacks the three aromatic residues aligned linearly and exposed to the solvent, which probably interact with cellulose in CBDs. Therefore, the binding mechanism of a group of ChBDs including ChBD(ChiA1) may be different from that proposed for CBDs.

Chitinase (EC 3.2.1.14) is a glycosyl hydrolase that catalyzes the hydrolytic degradation of chitin, a fibrous insoluble polysaccharide made of ␤-1,4-N-acetyl-D-glucosamine residues. Chitinases are found in a wide variety of organisms that possess chitin as a constituent (fungi, insects, and crustaceans) and organisms that do not possess chitin as well (bacteria, plants, and vertebrates). The roles of chitinases in these organisms are diverse (1). Invertebrates require chitinases for partial degradation of old exoskeletons. Fungi produce chitinases to modify chitin, which is used as an important cell wall component. Bacteria produce chitinases to digest chitin and utilize it as carbon and energy sources. It is suggested that the production of chitinases by higher plants is a part of defense mechanisms against fungal pathogens (2).
Bacillus circulans WL-12 is a Gram-positive bacterium identified as being lytic for yeast and fungal cell walls (3). The bacterium has been reported to secrete multiple chitinases into culture medium containing chitin as an inducer (4,5). Among these chitinases, A1 encoded by the chiA gene is thought to be the key enzyme in the chitinase system of this bacterium, because chitinase A1 (ChiA1) 1 is produced most abundantly and exhibits the highest activity as to the hydrolysis of colloidal chitin (4) and a high affinity to insoluble chitin. ChiA1 (M r ϭ ϳ74,000) contains three discrete functional domains: an Nterminal catalytic domain (CatD) (417 amino acid residues), a tandem repeat of fibronectin type III-like (FnIII) domains (duplicate 95 residues), and a C-terminal chitin-binding domain (45 residues) (6). Such a structural form of discrete catalytic and substrate-binding domains (7) as in ChiA1 has also been observed for some other polysaccharide-degrading enzymes, including cellulases (8,9), xylanases (10), amylases, and a ␤-1,3-glucanase (11).
The C-terminal chitin-binding domain (ChBD ChiA1 ) is required for ChiA1 to bind specifically to insoluble chitin and to hydrolyze it efficiently (4,6). We have found that ChBD ChiA1 does not bind to chito-oligosaccharides or soluble derivatives of chitin but does bind to insoluble or crystalline chitin. 2 So far, two roles of the cellulose-binding domains (CBDs) of cellulases in their hydrolysis activities have been suggested: (i) CBDs may enhance the cellulase activities by concentrating the cellulases on a cellulose surface (13,14), and (ii) CBDs may disrupt noncovalent interactions including hydrogen bonds between adjacent glucose units (15). The chitin-binding domains (ChBDs) are also considered to have similar roles to those of CBDs, but the roles of ChBDs remain unclear. Although a large number of studies have been performed on the structures of the CBDs of cellulases (8, 16 -21), little is known about the tertiary structures of the ChBDs of chitinases.
The structures of at least three chitinases have already been determined. They are (i) endochitinase from barley seeds, Hordeum vulgare (22), (ii) hevamine with combined chitinase and lysozyme activities from a plant, Hevea brasiliensis (23), and (iii) chitinase A from a Gram-negative soil bacterium, Serratia marcescens (24). However, the two plant chitinases do not contain ChBDs that are separated from their catalytic domains. It is suggested that the remaining chitinase, S. marcescens chitinase A, contains a putative ChBD in its N-terminal region (ChiN) (residues Ala 24 -His 137 ) (1), but no sequence similarity exists between ChiN and ChBD ChiA1 (see "FnIII Do-mains"). Therefore, the three-dimensional structure of a ChBD whose function has clearly been identified is necessary for elucidation of its binding mechanism with chitin and its role in the subsequent catalytic activity. Further, some CBDs studied so far have the ability to bind to soluble cello-oligosaccharides (17) and chitin (25) as well as crystalline cellulose. On the other hand, ChBD ChiA1 only binds to insoluble chitin; specifically, it does not bind to soluble chito-oligosaccharides, soluble chitin derivatives, or cellulose. 2 Thus, it is important to approach the mechanism by which this rather small domain consisting of 45 amino acids exhibits this binding specificity to insoluble chitin from the viewpoint of its tertiary structure.
Here we present the three-dimensional structure of the chitin-binding domain (Ala 655 -Gln 699 ) of B. circulans WL-12 chitinase A1 (ChBD ChiA1 ) determined by NMR using a uniformly 15 N-labeled domain. To the best of our knowledge, this is the first reported structure of a ChBD that binds specifically to chitin. Analysis involving structures and sequential alignments allowed us to suggest that the chitin-binding mechanism of ChBD ChiA1 is different from the currently proposed cellulosebinding mechanism of the CBDs of cellulases, which are supposed to interact with cellulose via three aromatic rings arranged linearly on the surfaces of CBDs.

EXPERIMENTAL PROCEDURES
Sample Preparation-Cells of Escherichia coli BL21(DE3) harboring the expression pET3a plasmid (Novagen, Madison, WI) with the ChB-D ChiA1 gene were grown in M9 minimal medium containing 100 g/ml ampicillin, 0.1% glycerol, 4.0 g/liter D-glucose, and 0.5 g/liter 15 NH 4 Cl as a sole nitrogen source. The cells were incubated at 30°C with shaking, and expression of the protein was induced by the addition of 0.5 mM isopropyl ␤-D-thiogalactopyranoside for 24 h. The cells were collected by centrifugation and disrupted by sonication in 100 mM Tris-HCl buffer (pH 8.0) containing 10 mM phenylmethylsulfonyl fluoride and 1 mM EDTA. Proteins in the soluble fraction were collected by ammonium sulfate precipitation (60% saturation) and then dialyzed against 5 mM sodium phosphate buffer (pH 6.0). The solution was applied to a chitin affinity column (Chitin EX, Funakoshi, Japan) equilibrated previously with the same buffer. After washing the column with 20 mM sodium phosphate buffer (pH 6.0) containing 1 M sodium chloride and with 20 mM sodium acetate buffer (pH 5.5), the protein was eluted with 20 mM acetic acid (pH 3.0). About 15 mg of purified and uniformly 15 N-labeled ChBD ChiA1 was obtained from 1 liter of an M9 minimal culture. The obtained protein has a T7 tag consisting of 14 residues (H 2 N-Met-Ala-Ser-Met-Thr-Gly-Gly-Gly-Gly-Met-Gly-Arg-Gly-Ser-) derived from the expression vector at its N terminus. The protein was stable and did not form a precipitate, at least upon changes of the solution in the pH range of 3-9, a NaCl concentration range of 20 -1000 mM, and a temperature range of 4 -40°C. The samples for most of the NMR measurements comprised 2.0 mM 15 N-labeled ChBD ChiA1 in 10 mM KH 2 PO 4 -K 2 HPO 4 (pH 6.0) and 10 mM deuterated dithiothreitol in 90% H 2 O, 10% 2 H 2 O or 1.2 mM 15 N-labeled ChBD ChiA1 in 100 mM KH 2 PO 4 -K 2 HPO 4 (pH 6.0) and 10 mM deuterated dithiothreitol in 99.8% 2 H 2 O. Protein concentrations were estimated using the calculated molar absorption coefficient at 280 nm (⑀ 280 ϭ 20,970). The calculated weight of the molecule, without the T7 tag region, is 5,018.
NMR Spectroscopy-All NMR experiments were performed with a Bruker DRX500 or DRX800 spectrometer equipped with a triple resonance ( 1 H, 15 N, and 13 C) probe with a self-shielded triple axis gradient coil. Most spectra were recorded at 310 K. For 1 H and 15 N resonance assignments (26), two-dimensional 15 N-1 H HSQC, two-dimensional TOCSY (with a mixing time of 50 ms), three-dimensional 15 N-edited TOCSY (with a mixing time of 70 ms), and two-dimensional H(NN)H TOCSY and three-dimensional (H)NNH TOCSY (with mixing times of 308 ms) (27), spectra were acquired. The numbers of complex points and spectral widths in the three-dimensional (H)NNH TOCSY experiment with the DRX500 were 20, 620.2 Hz ( 15 N, F 1 , and F 2 ) and 681, 8012.8 Hz ( 1 H, F 3 ). For dihedral angle constraints, a two-dimensional HMQC-J spectrum (26) was acquired with the DRX800. The 15 N dimension was recorded with an acquisition time of 403 ms (400 complex data sets). The digital resolution was 0.5 Hz after zero-filling and subsequent Fourier transformation. For interproton distance constraints (26), twodimensional NOESY and three-dimensional 15 (26). The three-dimensional 15 N-edited TOCSY experiment included the sensitivity enhancement and gradient echo methods for the indirect 15 N dimension (26). All other indirect dimensions were recorded in the States-TPPI manner (26). The NMR data were processed and analyzed using the nmrPipe (28) and Pipp (29) software packages, respectively. Stereospecific assignments of the methyl groups of the leucine and valine residues were achieved with 15% fractionally 13 C-labeled ChBD ChiA1 dissolved in 99.8% 2 H 2 O as described (30).
Structure Calculation-The NOE connectivities derived from strong, medium, and weak cross-peaks were categorized and assumed to correspond to the upper limits for interproton distances of 3.0, 4.0, and 5.0 Å, respectively. Pseudoatom corrections were applied to upper bound constraints involving methyl, methylene, and aromatic ring protons as described (31). The NOE list was not filtered to remove restraints of no structural value. The distance constraints for the hydrogen bonds were applied for slowly exchanging amides, i.e. 2.8 -3.3 Å for N-O pairs and 1.8 -2.3 Å for H-O pairs. A pair of these constraints was used for each hydrogen bond to restrict the length and angle of the bond (32, 33). The ␤-strands were aligned on the basis of the interstrand NOE connectivities among H ␣ and H N atoms. Hydrogen bond pairs between the amide hydrogens with slow chemical exchange rates and carbonyl oxygens were determined in accordance with the alignment of the ␤ strands. The backbone torsion angles, , were estimated from the scalar 3 J HNH␣ coupling constants derived from the HMQC-J spectrum. The angle constraints used were Ϫ65 Ϯ 25°for 3 J HNH␣ Ͻ4.7 Hz, Ϫ120 Ϯ 40°for 3 J HNH␣ Ͼ8.5 and Ͻ9.9 Hz, and Ϫ120 Ϯ 20°for 3 J HNH␣ Ͼ10.0 Hz. At the stage of structural constraint collection, the program DYANA version 1.5 (molecular dynamics in a torsion angle space with 25,000 steps) was used (34). Finally, an ensemble of 100 structures of ChBD ChiA1 was calculated with the program X-PLOR version 3.851 (35), the standard simulated annealing protocol (sa.inp file) with 18,000 steps at a high temperature and 9,000 steps for the cooling being used. The T7 tag region at the N terminus was excluded from the structure calculations. The best 30 structures were analyzed with MOLMOL (36) and with AQUA and PROCHECK-NMR (37) software. The secondary structures were determined on the basis of the main chain hydrogen bond patterns, the interstrand NOE connectivities, and the results of AQUA and PROCHECK-NMR analyses (37).

RESULTS
Resonance Assignments-The amide 1 H and 15 N resonances were assigned at first using two-dimensional H(NN)H TOCSY and three-dimensional (H)NNH TOCSY spectra (27), magnetization being transferred from an amide 15 N spin to those of both preceding and following residues through a 15 N homonuclear Hartmann-Hahn mixing (TOCSY) scheme. Thus, the two-dimensional and three-dimensional spectra provided observable resonances at the frequencies of ( Hi , Hj ) and ( Ni , Nj , Hj ), respectively, where i equals j Ϫ 1, j, or j ϩ 1. Figs. 1,  A and B, show examples of the sequential connectivities of the amide groups of residues (Tyr 662 -Tyr 670 ) in the two-dimensional H(NN)H TOCSY and three-dimensional (H)NNH TOCSY spectra, respectively. The combination of these twodimensional and three-dimensional spectra provided secure amide sequential connectivities, which amounted to 63% of the expected connectivities. Next, additional information regarding 1 H N -1 H N distances and amino acid types was obtained from the three-dimensional 15 N-edited NOESY and three-dimensional 15 N-edited TOCSY spectra, respectively. This information compensated for the ambiguity of the amide sequential connectivities that were not resolved in the above 3 J NN -based experi-ments. Finally, we accomplished the assignment of all the amide resonances except for those of Asn 671 , Gly 672 , the T7 tag region, and three proline residues. The region containing Asn 671 and Gly 672 may undergo a conformational exchange, because these residues form the turn connecting the ␤ 2 -and ␤ 3 -strands (see "Structure Description"). Fig. 1C shows a twodimensional 15 N-1 H HSQC spectrum with the amide resonance assignments. The well dispersed resonance peaks in both the 1 H and 15 N dimensions reflect the stable structure of ChBD- ChiA1 . The proton resonances of the ␣ sites and side chains were assigned using the two-dimensional TOCSY and three-dimensional 15 N-edited TOCSY spectra. The methyl groups of the four leucine and two valine residues were assigned stereospecifically using 15% fractionally 13 C-labeled ChBD ChiA1 . The methyl groups of the other two residues, Leu 678 and Val 658 , could not be assigned stereospecifically owing to their overlapping 1 H chemical shifts. Overall, we assigned 95% of the expected 1 H and 15 N resonances of the main chain and side chains.
Constraints for the Structure Calculation-The distance constraints based on NOE were extracted from two-dimensional NOESY and three-dimensional 15 N-edited NOESY spectra with mixing times of 150 ms. By comparing the intensities of well resolved cross-peaks in a series of preliminary two-dimensional NOESY spectra obtained with different mixing times, i.e. 50, 100, 150, 200, and 250 ms (data not shown), we judged that the mixing time of 150 ms did not cause severe spin diffusion during the NOE transfer of magnetization. Fig. 2A summarizes the sequential and medium-range NOE connectivities along with the secondary structures, 1 H ␣ chemical shift indices (38), amide hydrogen exchange rates, and 3 J HNH␣ coupling constants observed for ChBD ChiA1 . The ␤-strand regions are characterized by the strong intensities of the NOE crosspeaks between ␣ protons and the amide protons of the subsequent residues (d ␣N(i, i ϩ 1) ). The locations of the ␤-strands are also indicated by positive deviations of the 1 H ␣ chemical shift values from those observed in random coils, which are represented by the 1 H ␣ chemical shift index of ϩ1 (38), slow rates of exchange of the amide protons with the solvent, and 3 J HNH␣ coupling constants larger than 8.5 Hz. Interstrand NOE connectivities were observed among H ␣ and H N atoms ( 1 H N -1 H N , 1 H ␣ -1 H N , and 1 H ␣ -1 H ␣ ), as shown by the arrows in Fig. 2B. Overall, 493 NOE-based distance, 20 hydrogen bond, and 33 dihedral angle constraints were collected and used for the structure calculations (Table I).
Structure Determination-Solution structures were calculated through the standard simulated annealing protocol in the program X-PLOR 3.851 (35). Out of 100 calculated structures, 30 final structures, which showed the lowest energy values, no distance constraint violation of Ͼ0.3 Å, and no dihedral angle constraint violation of Ͼ5°, were selected for further analyses. Superpositioning of these 30 structures is shown in Fig. 3A, and a summary of the constraints and structural statistics is given in Table I. The backbones converged well, as indicated by the r.m.s. deviation of 0.326 Å from the mean structure for the backbone CЈ, C ␣ , N atoms of all the residues except for the N-terminal Ala 655 and C-terminal Gln 699 residues. The r.m.s. deviation for all heavy atoms in the same regions is 0.700 Å. The Asn 671 and Gly 672 residues, whose amide signals could not be observed in 1 H-15 N HSQC spectra, are in the loop region connecting the ␤ 2 -and ␤ 3 -strands, and the calculated coordinates of the loop diverged due to the few constraints in the loop.
Almost the same structures were obtained upon calculation with the program DYANA 1.5 (34) with the same constraints as those used in the X-PLOR calculation. The average DYANA target function value of the best 30 structures selected out of the 100 calculated ones was 0.32 Ϯ 0.10 (one S.D.), and the r.m.s. deviations of all 60 structures that were calculated with X-PLOR and DYANA were 0.354 Å for the backbone atoms and 0.713 Å for all of the heavy atoms in the same regions as   2. Summary of the structure information obtained in the NMR experiments. A, summary of the sequential and medium range NOE connectivities, secondary structures, chemical shift indices, amide hydrogen exchange rates, 3 J HNH␣ coupling constants, and solvent accessibility values for ChBD ChiA1 . The NOE connectivities are represented by bars, the size of which indicates the NOE intensity (strong, medium, or weak). The notation d ␣N(i, i ϩ 1) , for example, represents the connectivity between the ␣ proton resonance of a residue (i) and the amide proton resonance of the subsequent residue (i ϩ 1) in the sequence. Amide protons that were exchanged slowly at pH described above.
Structure Description-ChBD ChiA1 has a compact and globular structure, as shown in Fig. 3B. It contains two antiparallel ␤-sheets (Fig. 2C). One sheet is composed of three strands designated as ␤ 2 (Gln 666 -Tyr 670 ), ␤ 3 (Lys 673 -Cys 677 ), and ␤ 5 (Trp 696 -Leu 698 ), while the other is composed of two strands designated as ␤ 1 (Thr 660 -Tyr 662 ) and ␤ 4 (His 681 -Ser 683 ). No region characteristic of an ␣-helix exists. ␤ 5 consists of three residues and forms an antiparallel ␤-sheet with the three Cterminal residues of ␤ 3 . Fig. 2B shows the hydrogen bond networks and NOE connectivities among these five ␤-strands that were used for the structure calculation. The two antiparallel ␤-sheets formed by these hydrogen bond networks fold into the topology of a twisted ␤-sandwich with an angle of about This hydrophobic core and the hydrogen bond networks observed among the ␤-strands make the overall structure rigid and compact. Such a character was also expected from the presence of the amide hydrogen atoms that were not exchanged with the solvent for as long as 18 h at pH 6.0 and 298 K (the residues with closed circles in Fig. 2A). ChBD ChiA1 contains only three charged residues (ϩLys 673 , ϩLys 676 , and ϪGlu 688 ), and the surface of the molecule is dominated by noncharged residues, as shown in Fig. 3C.
The similarity of the tertiary structure of ChBD ChiA1 to known structures was examined with the DALI server version 2.0 (39), pairs with Z scores of more than 2.0 being assumed to be similar. The results showed that the structure of ChBD ChiA1 is similar to that of the CBD of endoglucanase Z (CBD EGZ ) (Protein Data Bank code 1aiw) secreted by a Gram-negative and plant pathogenic bacterium, Erwinia chrysanthemi (8), as indicated by the Z score of 2.8 and r.m.s. deviation (C ␣ atoms) of 2.4 Å (Fig. 4). No other similar structure (characterized by a Z score Ͼ 2.0) was found in the DALI data base. The sequence of CBD EGZ is also similar to that of ChBD ChiA1 (Fig. 5) (CLUSTAL W (40) alignment score, 19%) (see "Discussion").

DISCUSSION
Comparison of ChBD ChiA1 and CBDs-Chitin differs chemically from cellulose only in that each C2 hydroxyl (-OH) group in cellulose is replaced by an acetamide (-NHCOCH 3 ) group in chitin. Thus, the mechanism by which ChBD ChiA1 binds to chitin was expected to be similar to the mechanism by which CBDs bind to cellulose. The most accepted model for the binding of CBDs to cellulose is that aromatic rings arranged in the flat face of a CBD are stacked on every other pyranose ring of polysaccharides through hydrophobic interactions (20). The involvement of aromatic residues in the interactions has been observed using NMR (41), site-directed mutagenesis (16,25), and chemical modification (42). In CBD EGZ , for example, three aromatic residues, Trp 18 , Trp 43 , and Tyr 44 , the latter two of which are conserved widely in CBDs, are localized on one face of CBD EGZ , and are exposed to the solvent, as shown in Fig. 4A  (8). Recently, Simpson et al. confirmed the involvement of these three aromatic residues in the interaction with cellulose on site-directed mutagenesis of the intact endoglucanase Z (also referred to as Cel5) (25).
Despite the similarity between the overall topologies of CB-D EGZ and ChBD ChiA1 , ChBD ChiA1 lacks a region that corresponds to the loop ranging from Val 14 to Gln 22 of CBD EGZ (colored purple in Fig. 4A). This loop region of CBD EGZ contains Trp 18 , which is one of the three aromatic residues that are involved in the interaction with cellulose. Moreover, the other two aromatic residues, which correspond to Trp 43 and Tyr 44 in CBD EGZ , are also missing or replaced by another aromatic residue, His 681 , in ChBD ChiA1 (Fig. 4B).
We recently showed that ChBD ChiA1 does not interact with soluble substrates such as hexa-N-acetylchitohexaose, 40% deacetylated chitin, carboxymethyl chitin, or ethylene glycol chitin by means of both NMR chemical shift mapping experiments and isothermal titration calorimetry. 2 Moreover, ChBD ChiA1 binds specifically to chitin but not to cellulose. Thus, ChBD ChiA1 probably recognizes the solid surface conformation of crystallin chitin but not that of cellulose or the flexible conformations of soluble substrates. On the other hand, the binding specificities of the CBDs of some cellulases are rather broad. CBD-CBHI , CBD CBHII (41), CBD EGI (17), and CBD XYLA (10) bind to both crystalline cellulose and soluble cello-oligosaccharides, and CBD EGZ (CBD Cel5 ) (25), CBD CbpA (43), CBD CBHII (44), CBD Cex (18), CBD CenA (45), and Cip-CBD (20) bind to both cellulose and chitin. Thus, we suggest that the binding mechanism of ChBD-ChiA1 is different from that of CBDs.
Two Groups of ChBDs-Brun et al. (8) pointed out that the stWWst motif, which corresponds to Ala 41 , Asn 42 , Trp 43 , Tyr 44 , Thr 45 , and Ala 46 in CBD EGZ , is not conserved in ChBD ChiA1 . Fig. 5 shows the sequence alignment of ChBD ChiA1 with the domains of other chitinases and CBD EGZ . In CBD EGZ and the five chitinases listed on the lower lines (Fig. 5, vi-xi), the Trp-Trp, Tyr-Trp, or Trp-Tyr (Arom-Arom) sequence is conserved at the site where CBD EGZ holds the Trp 43 -Tyr 44 sequence. On the contrary, ChBD ChiA1 and the four chitinases listed on the upper lines (i-v) have no such sequence at the corresponding site. The lack of the Arom-Arom sequence in ChBD ChiA1 is more apparent in the tertiary structure shown in Fig. 4B. Therefore, the chitinases in the group including ChB-D ChiA1 (the ChBD ChiA1 group) probably have different binding surfaces for substrates from those of the chitinases and cellulases in the other group including CBD EGZ (the CBD EGZ group).
We searched for the residues of ChBD ChiA1 that may be involved in the binding to chitin. The criteria for the search were that (i) such residues are well conserved at least in the ChBD ChiA1 group (lines i-v in Fig. 5); (ii) they exist on the surface of the molecule; and (iii) they are hydrophobic or aromatic residues contributing to the hydrophobic interaction with chitin. The last criterion is based on the observation that more ChBD ChiA1 bound to chitin at a pH nearer to its isoelectric point (about 9.0) or in the presence of more NaCl in the solution. 2 The residues that meet these criteria are His 681 (27.1), Thr 682 (34.4), Trp 687 (22.5), Pro 689 (38.6), and Pro 693 (45.5), the values in parentheses indicating the solvent accessibility. These residues are localized on one face of the molecule, which is different from the interaction surface of CBD EGZ , as shown in Fig. 6. Further, Trp 687 and Pro 693 are specific residues to the ChBD-ChiA1 group, i.e. they are not conserved in the CBD EGZ group (lines vi-xi in Fig. 5). In CBD EGZ , the three aromatic residues involved in the interaction with cellulose, Trp 18 (49.2), Trp 43 (36.0), and Tyr 44 (45.9%), are exposed much on the surface of the molecule, but in ChBD ChiA1 no aromatic residue with such high solvent accessibility was found. His 681 (27.1) and Trp 687 (22.5%) are the two aromatic residues that are most exposed to the solvent. Instead, Thr 682 (34.4) and Pro 689 (38.6%) in ChB-D ChiA1 have much higher solvent accessibility values than the corresponding residues of CBD EGZ , Thr 45 (5.0) and Pro 49 (0.9%), respectively. Thus, in the ChBD ChiA1 group, residues such as threonine and proline as well as aromatic residues may also be involved in the interaction with chitin. Since ChBD ChiA1 binds only to the crystalline form of chitin, the domain is expected to recognize the characteristic conformation of crystallin chitin, such as extended polysaccharide chains aligned in parallel to each other. Thus, the binding specificity of ChBD-ChiA1 to solid chitin is probably due to the different binding mechanism from that of CBD EGZ despite their similar tertiary structures.
We anticipate that the hydrophobic effect is the major contribution to the interaction between ChBD ChiA1 and chitin, as mentioned above. However, since ChBD ChiA1 is eluted from a chitin affinity column at pH 3.0, the negatively charged unique residue, Glu 688 , may also be involved in the interaction (Fig.   3C). The involvement of polar residues (Gln, Asn) in the interaction with cellulose through hydrogen bonding to oxygen atoms or hydroxyl groups of glucose moieties has been observed in the CBDs (20,46). FnIII Domains-The N-terminal domain of chitinase A from a soil bacterium, S. marcescens, is likely to interact with chitin because the intact S. marcescens chitinase A exhibits significant chitin-binding activity (47) and comprises this N-terminal domain (ChiN) (residues Ala 24 -His 137 ) and a C-terminal catalytic domain (residues Val 159 -Ala 442 and Asp 517 -Val 563 ), which is significantly homologous to the catalytic domain of B. circulans chitinase A1 (ChiA1). However, no similarity was found in the sequence (CLUSTAL W (32) alignment score, 10%), overall topology, or arrangements of the aromatic rings between ChiN and ChBD ChiA1 . ChiN (114 residues, 11 ␤-strands) is more than twice the size of ChBD ChiA1 (45 residues, five ␤-strands), and the tertiary structure of ChiN rather resembles those of the FnIII modules found in various animal proteins despite no sequence similarity between ChiN and FnIII domains (48). Although ChiA1 also contains two consecutive FnIII domains between the catalytic (CatD) and binding (ChBD) domains (49), these FnIII domains, interestingly, do not bind to chitin (6). Since ChiN has nine aromatic residues, some of them may be involved in the interaction with chitin in the similar way to how the three consecutive aromatic rings of CBDs interact with cellulose.
A Disulfide S-S Bond-CBD EGZ has a disulfide bridge (Cys 4 -Cys 61 ) between the two extremities of the domain, whereas ChBD ChiA1 has no disulfide bridge at the corresponding site (Fig. 4). To be precise, Cys 61 of CBD EGZ corresponds to the outside of the C-terminal end of ChiA1, and no cysteine residue exists around the region in ChiA1 corresponding to the region containing Cys 4 in CBD EGZ . As a protein engineering approach, the introduction of a disulfide bridge into ChiA1 would increase the stability of the protein against changes in heat, pH, ionic strength, etc. more and would make a mutated protein more appropriate for various industrial uses.