An antifungal chitosanase/glucanase isolated from the soil bacterium Paenibacillus sp. IK-5 has two CBM32 chitosan-binding modules (DD1 and DD2) linked in tandem at the C-terminus. In order to obtain insights into the mechanism of chitosan recognition, the structures of DD1 and DD2 were solved by NMR spectroscopy and crystallography. DD1 and DD2 both adopted a β-sandwich fold with several loops in solution as well as in crystals. On the basis of chemical shift perturbations in 1H-15N-HSQC resonances, the chitosan tetramer (GlcN)4 was found to bind to the loop region extruded from the core β-sandwich of DD1 and DD2. The binding site defined by NMR in solution was consistent with the crystal structure of DD2 in complex with (GlcN)3, in which the bound (GlcN)3 stood upright on its non-reducing end at the binding site. Glu14 of DD2 appeared to make an electrostatic interaction with the amino group of the non-reducing end GlcN, and Arg31, Tyr36 and Glu61 formed several hydrogen bonds predominantly with the non-reducing end GlcN. No interaction was detected with the reducing end GlcN. Since Tyr36 of DD2 is replaced by glutamic acid in DD1, the mutation of Tyr36 to glutamic acid was conducted in DD2 (DD2-Y36E), and the reverse mutation was conducted in DD1 (DD1-E36Y). Ligand-binding experiments using the mutant proteins revealed that this substitution of the 36th amino acid differentiates the binding properties of DD1 and DD2, probably enhancing total affinity of the chitosanase/glucanase toward the fungal cell wall.

INTRODUCTION

Interactions between soil bacteria and fungal pathogens play important roles in sustainably managing agricultural production from soil [1]. Fungal cell walls contain several kinds of glucans, chitin and chitosan, which are polysaccharides composed of glucose, N-acetylglucosamine and GlcN (glucosamine) respectively as their main components [2,3]. Soil bacteria antagonize pathogenic fungi via antibiotics as well as their own enzymes, which have the ability to hydrolyse polysaccharides in the fungal cell wall [4]. Strategies for the biological control of fungal pathogens have often been designed on the basis of bacterial polysaccharide hydrolases [5,6]. Paenibacillus species isolated from soil have been recognized to produce glucanases, chitinases and chitosanases, which are involved in antagonism towards pathogenic fungi [7]. Among these enzymes, we focused on a chitosanase/glucanase produced by a Paenibacillus sp. IK-5 strain, which hydrolyses the β-1,4-glycosidic linkages of chitosan and glucan. The enzyme exhibited a modular structure consisting of a GH8 (glycoside hydrolase 18) family catalytic module, fibronectin type-III module and two CBMs (carbohydrate-binding modules) [8], as shown in Figure 1(A). These CBMs were referred to as DD1 and DD2 (discoidin domains 1 and 2) respectively on the basis of sequence similarities to discoidins I and II from Dictyostelium discoideum [9,10]. Since no other chitosanases investigated to date have these CBMs, the chitosanase/glucanase may have a higher affinity for chitosan and hence for fungal cell walls. This chitosanase/glucanase has also been reported to lose its chitosan-binding ability with the truncation of CBMs [11]. Thus these two CBMs have been speculated to target the enzyme to chitosan polysaccharide, thereby promoting catalytic action. A subgroup of discoidin domains possessing the carbohydrate-binding ability has now been classified in the CBM32 family [12].

Arrangement and amino acid sequences of DD1 and DD2 in Paenibacillus sp. IK-5 chitosanase/glucanase

Figure 1
Arrangement and amino acid sequences of DD1 and DD2 in Paenibacillus sp. IK-5 chitosanase/glucanase

(A) Schematic representation of the full-length structure of a chitosanase/glucanase from Paenibacillus sp. IK-5. Numerals are the numbers in the amino acid sequence. (B) Alignments of the amino acid sequences and secondary structures of DD1 and DD2 observed in the crystal structures. α-Helix is coloured magenta, and β-strands are coloured blue for DD1 and orange for DD2. β4 and β5 of DD1 and DD2 are designated β4(1), β4(2), β5(1) and β5(2) respectively to more easily differentiate the locations of the individual strands in the crystal structures (see Figure 3C). Identical amino acid residues are shown with a purple background. Red boxes in the sequences indicate amino acid residues interacting with chitosan oligosaccharide (see Figure 6).

Figure 1
Arrangement and amino acid sequences of DD1 and DD2 in Paenibacillus sp. IK-5 chitosanase/glucanase

(A) Schematic representation of the full-length structure of a chitosanase/glucanase from Paenibacillus sp. IK-5. Numerals are the numbers in the amino acid sequence. (B) Alignments of the amino acid sequences and secondary structures of DD1 and DD2 observed in the crystal structures. α-Helix is coloured magenta, and β-strands are coloured blue for DD1 and orange for DD2. β4 and β5 of DD1 and DD2 are designated β4(1), β4(2), β5(1) and β5(2) respectively to more easily differentiate the locations of the individual strands in the crystal structures (see Figure 3C). Identical amino acid residues are shown with a purple background. Red boxes in the sequences indicate amino acid residues interacting with chitosan oligosaccharide (see Figure 6).

In our previous study [13], we produced three proteins, i.e. DD1, DD2 and a protein comprising DD1 and DD2 in tandem (DD1+DD2), using an Escherichia coli expression system, and these three proteins were characterized for their binding properties by thermal unfolding experiments, ITC (isothermal titration calorimetry) and NMR titration experiments using chitosan oligosaccharides, (GlcN)n (n=2–6), as the ligands. Although DD1 and DD2 belong to the same family, CBM32, and share high sequence similarities (74%, Figure 1B), these binding experiments revealed that DD1 and DD2 have different binding specificities [13]. DD1 binds chitosan oligosaccharides, but not cello-oligosaccharides or laminarioligosaccharides, whereas DD2 binds all three oligosaccharide types with different affinities. In addition, the binding affinity of DD1 towards (GlcN)n was higher than that of DD2 by approximately 2.0 kcal/mol (1 kcal=4.184 kJ) of binding free energy change [13]. Homology-modelled structures of DD1 and DD2 have been constructed and compared with each other in order to structurally rationalize the differences in binding properties. However, we failed to fully explain these differences. It is now highly desirable to solve the three-dimensional structures of DD1 and DD2.

In the present study, the structures of DD1 and DD2 were solved by NMR spectroscopy and crystallography in order to define structural differences between these two CBMs. Amino acid residues predicted to differentiate the binding properties of DD1 and DD2 were mutated, and the mutated proteins were then characterized. The results obtained from these experiments provide insights into the ligand-recognition mechanism and hence the role of these two CBMs in the enzymatic degradation of carbohydrates in fungal cell walls.

EXPERIMENTAL

Materials

Chitosan oligosaccharides (GlcN)n (n=2–6) were purchased from Seikagaku Biobusiness. E. coli BL21(DE3) pLacI and Rosetta (DE3) pLacI cells and the expression vector pET Blue-1 were from Novagen. The Ni2+-affinity resin COSMOGEL His-Accept was purchased from Nacalai Tesque. Sephacryl S-100 HR was from GE Healthcare. All other reagents were of analytical grade.

Protein preparation

For the production of non-labelled DD1 or DD2 protein, E. coli cells harbouring the plasmid (pETB-DD1 or pETB-DD2) [13] were grown in LB medium to an attenuance of 0.6 at 600 nm before being induced with 1 mM IPTG. After induction, growth was continued at 18°C for 18 h. Cells were harvested by centrifugation, suspended in 10 mM Tris/HCl (pH 8.0), and disrupted with a sonicator. After cell debris was removed by centrifugation at 10000 g for 15 min, the supernatant was dialysed against the same buffer, and adsorbed on to an Ni2+-affinity column. The adsorbed proteins were eluted with 10 mM Tris/HCl (pH 8.0) containing 250 mM imidazole, and separated further using a column of Sephacryl S-100 HR equilibrated with 10 mM Tris/HCl (pH 8.0) containing 0.1 M NaCl. Fractions exhibiting a single protein band on SDS/PAGE were collected as a purified recombinant protein. M9 medium containing 15NH4Cl and/or [13C]glucose was used to produce 15N- and/or 13C-labelled proteins. Protein concentrations were determined by reading absorbance at 280 nm using molar absorption coefficients calculated from the equation proposed by Pace et al. [14].

NMR spectroscopy and solution structure determinations

NMR data were acquired on a Bruker Avance 500 MHz spectrometer controlled with TopSpin 3.0 software and equipped with a triple-resonance pulsed-field-gradient cryoprobe head. All measurements were carried out at 300 K. The protein solution, 0.2 mM 15N-labelled or 13C/15N-double-labelled DD1 or DD2, was dialysed against 10 mM Tris/HCl (pH 7.0), containing 10% 2H2O. Protein backbone resonances were assigned using 1H-15N-HSQC and three-dimensional NMR spectra, including HNCO, HNCACO, HNCACB and CBCA(CO)NH [15,16]. In order to assign side chain resonances, the 13C/15N-double-labelled DD1 or DD2 preparation was dialysed against 100% 2H2O, followed by the acquisition of 1H-13C-HSQC, HCCH-TOCSY and CCH-TOCSY spectra [17]. In structure determinations, NOE-derived distance restraints were obtained from 15N-edited NOESY and 13C-edited NOESY, both of which were collected with a 120 ms mixing time [18]. Backbone torsion angle constraints were derived from chemical shift values of the main-chain resonances using the program TALOS+ [19]. All NMR spectra were processed using NMRPipe [20] and analysed using Sparky software (Goddard and Kneller, https://www.cgl.ucsf.edu/home/sparky/). The CYANA program (version 2.1) [21] was used to calculate the structure, starting with 1000 randomized conformers. Ten structures with the lowest target functions were selected, and their quality was assessed with PROCHECK [22]. Judging from the interstrand NOEs in the β-sheets, hydrogen bonds were identified during the sequential assignment and confirmed by hydrogen–deuterium exchange experiments (results not shown). For the hydrogen bonds that were structurally satisfied in the preliminary structure calculation only with NOE-derived restraints, the hydrogen bond constraints were added to the distance constraints: NH-O (1.8–2.0 Å) and N-O (2.7–3.0 Å). NMR structural statistics are presented in Table 1. The final set of restraints for DD1 contained 878 non-redundant unambiguous NOEs and 60 pairs of hydrogen bond restraints, plus 255 backbone dihedral restraints. The final set for DD2 contained 805 non-redundant unambiguous NOEs and 44 pairs of hydrogen bond restraints, plus 276 backbone dihedral restraints.

Table 1
Structural statistics for the ten final NMR structures
 DD1 DD2 
NMR distance and dihedral restraints 
 Total unambiguous NOE 878 805 
 Intraresidue 275 201 
 Interresidue   
  Sequential 308 279 
  Medium-range 50 44 
  Long-range 245 281 
 Hydrogen-bond restraints 60 44 
 Dihedral angle restraints   
  Phi constraints 135 153 
  Psi constraintes 120 123 
Structure statistics 
 Violations (mean±S.D.)   
  Distance constraints (Å) 0.0201±0.0013 0.0306±0.0022 
  Dihedral angle constraints (°) 0.5469±0.0486 0.6058±0.0691 
  Maximum dihedral angle violation (°) 4.11±0.72 4.49±0.29 
  Maximum distance constraint violation (Å) 0.29±0.05 0.42±0.07 
 Average pairwise rmsd   
  Heavy atoms (Å) 1.07±0.08 1.55±0.19 
  Backbone atoms (Å) 0.64±0.14 1.03±0.19 
Ramachandran statistics (%) 
 Residues in most favourable regions 78.2 67.4 
 Residues in additional allowed regions 20.2 28.0 
 Residues in generously allowed regions 1.6 4.6 
 Residues in disallowed regions 
 DD1 DD2 
NMR distance and dihedral restraints 
 Total unambiguous NOE 878 805 
 Intraresidue 275 201 
 Interresidue   
  Sequential 308 279 
  Medium-range 50 44 
  Long-range 245 281 
 Hydrogen-bond restraints 60 44 
 Dihedral angle restraints   
  Phi constraints 135 153 
  Psi constraintes 120 123 
Structure statistics 
 Violations (mean±S.D.)   
  Distance constraints (Å) 0.0201±0.0013 0.0306±0.0022 
  Dihedral angle constraints (°) 0.5469±0.0486 0.6058±0.0691 
  Maximum dihedral angle violation (°) 4.11±0.72 4.49±0.29 
  Maximum distance constraint violation (Å) 0.29±0.05 0.42±0.07 
 Average pairwise rmsd   
  Heavy atoms (Å) 1.07±0.08 1.55±0.19 
  Backbone atoms (Å) 0.64±0.14 1.03±0.19 
Ramachandran statistics (%) 
 Residues in most favourable regions 78.2 67.4 
 Residues in additional allowed regions 20.2 28.0 
 Residues in generously allowed regions 1.6 4.6 
 Residues in disallowed regions 

The NMR, atomic co-ordinates, chemical shifts and restraints have been deposited in the PDB under accession codes 2RV9 for DD1 and 2RVA for DD2, and in the BMRB under accession codes 11591 for DD1 and 11592 for DD2.

Crystallization and data collection

DD1 was crystallized at 20°C in a reservoir solution consisting of 0.1 M ammonium sulfate, 0.01 M magnesium chloride, 0.05 M Mes (pH 5.6) and 20% (w/v) PEG 8000. The sitting-drop set up by the addition of 1 μl of reservoir solution to 1 μl of protein solution (3.0 mg/ml) dialysed against distilled water was equilibrated against 100 μl of reservoir solution. Regarding data collection, the crystal was transferred and briefly soaked in cryoprotectant solution consisting of the reservoir solution containing 20% ethylene glycol, before being cryocooled in a nitrogen stream at 95 K. Data were collected on NW-12A beamline at the Photon Factory (Ibaraki, Japan) using an ADSC Q210 CCD (charge-coupled device) detector with a wavelength of 1.00000 Å (1 Å=0.1 nm) at a cryogenic temperature (95 K). In order to obtain SeMet (selenomethionine)-containing DD1, Val110 was mutated to methionine using the oligonucleotide primers listed in Supplementary Table S1, and the mutated gene was transformed into E. coli B834(DE3). The cells were cultivated and the SeMet-labelled protein (DD1-SeMet-V110M) was produced by the same procedure described for the production of wild-type protein. DD1-SeMet-V110M was crystallized at 20°C in a sitting drop consisting of 1 μl of protein solution (5.4 mg/ml) dialysed against distilled water and 1 μl of reservoir solution containing 0.2 M sodium fluoride and 20% (w/v) PEG 3350 (pH 7.3), which was equilibrated against 100 μl of reservoir solution. The crystal was soaked briefly in cryoprotectant solution consisting of the reservoir solution containing 20% (v/v) ethylene glycol. After cryocooling the crystal at 95 K, data collection was performed at a wavelength of 0.97865 Å on the beamline BL-17A at the Photon Factory using a DECTRIS PILATUS3 S 6M hybrid photon-counting detector.

DD2 crystals grew in a sitting drop consisting of 1 μl of protein solution (5.2 mg/ml) and 1 μl of reservoir solution containing 0.2 M ammonium sulfate, 0.1 M Bis-Tris (pH 5.5) and 25% PEG 3350, which was equilibrated against 100 μl of the reservoir solution at 20°C. After soaking the crystal in the cryoprotectant solution (reservoir solution containing 30% PEG 5000) and cryocooling at 95 K, data were collected on the beamline BL-17A at the Photon Factory using the ADSC Q270 CCD detector with a wavelength of 0.98000 Å. DD2 (1.35 mg/ml) was co-crystallized with (GlcN)3 at a molar ratio of 1:10 protein/sugar. The crystal grew in a sitting drop consisting of 1 μl of protein/sugar solution and 1 μl of the same reservoir solution as that used for the crystallization of DD2. Data collection was conducted following the same procedure described above for DD2, except that 15% ethylene glycol was added to the reservoir solution instead of 30% PEG 5000 for cryoprotection. All data were integrated and scaled with HKL2000 [23].

Determination of crystal structures

The crystal structure of DD1-SeMet-V110M was initially solved by SAD (single-wavelength anomalous dispersion) phasing [24] using AutoSol [25] in the program suite Phenix [26]. The other three structures were solved by the molecular replacement method from Phaser [27] in the program suite CCP4 [28]. The crystal structure of DD1-SeMet-V110M was used as a template for obtaining the structure of DD1, which was in turn used as a template for obtaining the structure of DD2 in complex with (GlcN)3. The structure of the DD2–(GlcN)3 complex was then used as a template for obtaining the structure of DD2. Three protein molecules were packed in the crystallographic asymmetric unit of the crystal structure of DD1, and two molecules were packed in each asymmetric unit of the other three crystal structures. Models were improved by several rounds of refinement with the combination of REFMAC [29] in the CCP4 program suite [28], Phenix.refine [30] in the Phenix program suite and manual rebuilding with COOT [31]. These structures were refined to Rwork/Rfree of 14.3/17.2, 20.8/24.9, 20.0/22.6 and 17.3/21.0 at resolutions of 1.40, 1.20, 1.29 and 1.65 Å for DD1, DD1-SeMet-V110M, DD2 and the DD2–(GlcN)3 complex respectively. The quality of the four structures was checked with RAMPAGE [32]. Protein residues in the favoured/allowed/disallowed region of the Ramachandran plot were as follows: 98.5/1.5/0.0% for DD1, 98.1/1.9/0.0% for DD1-SeMet-V110M, 97.7/1.9/0.4% for DD2, and 97.7/2.3/0.0% for the DD2–(GlcN)3 complex. DD2 contained one residue (Tyr120 of the A chain) in an outlier region. However, the plot in the disallowed region was not far from the allowed region, and the structure of DD2 was satisfactorily superimposed with the DD2–(GlcN)3 complex structure, in which no residues were found in the disallowed region. We regarded the DD2 structure thus obtained as reliable. Data collection and refinement statistics for all four structures are presented in Table 2. 

Table 2
Data collection and refinement statistics

Rmerge=Σ|IavgIi|/ΣIi. Rwork=Σ|F0Fc|/ΣF0 for reflections of the working set. Rfree=Σ|F0Fc|/ΣF0 for reflections of the test set (5.0% of total reflections).

 DD1 DD1-SeMet-V110M DD2 DD2–(GlcN)3 
Data collection 
 Space group PP21 P212121 P212121 
 Cell dimensions     
  a, b, c (Å) 42.1, 45.6, 57.6 28.2, 65.3, 69.9 38.5, 66.0, 117.3 38.4, 65.5, 117.5 
  α, β, γ (°) 111.7, 105.4, 97.9 90.0, 97.2, 90.0 90.0, 90.0, 90.0 90.0, 90.0, 90.0 
 Wavelength (Å) 1.00000 0.97865 0.98000 0.98000 
 Resolution (Å) 50–1.40 (1.42–1.40) 50–1.20 (1.22–1.20) 50–1.29 (1.31–1.29) 50–1.65 (1.68–1.65) 
Rmerge 0.031 (0.069) 0.123 (0.220) 0.083 (0.248) 0.093 (0.584) 
II 71.3 (34.2) 28.4 (4.47) 43.9 (15.8) 20.3 (2.75) 
 Completeness (%) 95.1 (91.6) 99.8 (97.0) 96.4 (79.9) 100 (99.9) 
 Redundancy 7.9 (8.0) 6.1 (4.1) 9.7 (8.9) 11.1 (7.7) 
Refinement 
 Resolution (Å) 50.26–1.40 47.55–1.20 58.7–1.29 58.74–1.65 
 Number of reflections 65749 75932 69638 34596 
Rwork/Rfree 14.3/17.2 20.8/24.9 20.0/22.6 17.3/21.0 
 Number of atoms     
  Protein 3258 2184 2043 2045 
  Ligand/Ion 96 − 18 71 
  Water 625 564 477 309 
 Average B-factors (Å2    
  Protein 4.59 9.78 11.7 15.5 
  Ligand/ion 21.5 − 25.8 28.2 
  Water 21.8 23.2 27.5 29.9 
 RMSD     
  Bond length (Å) 0.0205 0.0301 0.0305 0.0214 
  Bond angles (°) 2.4143 2.4659 2.5151 2.1203 
 DD1 DD1-SeMet-V110M DD2 DD2–(GlcN)3 
Data collection 
 Space group PP21 P212121 P212121 
 Cell dimensions     
  a, b, c (Å) 42.1, 45.6, 57.6 28.2, 65.3, 69.9 38.5, 66.0, 117.3 38.4, 65.5, 117.5 
  α, β, γ (°) 111.7, 105.4, 97.9 90.0, 97.2, 90.0 90.0, 90.0, 90.0 90.0, 90.0, 90.0 
 Wavelength (Å) 1.00000 0.97865 0.98000 0.98000 
 Resolution (Å) 50–1.40 (1.42–1.40) 50–1.20 (1.22–1.20) 50–1.29 (1.31–1.29) 50–1.65 (1.68–1.65) 
Rmerge 0.031 (0.069) 0.123 (0.220) 0.083 (0.248) 0.093 (0.584) 
II 71.3 (34.2) 28.4 (4.47) 43.9 (15.8) 20.3 (2.75) 
 Completeness (%) 95.1 (91.6) 99.8 (97.0) 96.4 (79.9) 100 (99.9) 
 Redundancy 7.9 (8.0) 6.1 (4.1) 9.7 (8.9) 11.1 (7.7) 
Refinement 
 Resolution (Å) 50.26–1.40 47.55–1.20 58.7–1.29 58.74–1.65 
 Number of reflections 65749 75932 69638 34596 
Rwork/Rfree 14.3/17.2 20.8/24.9 20.0/22.6 17.3/21.0 
 Number of atoms     
  Protein 3258 2184 2043 2045 
  Ligand/Ion 96 − 18 71 
  Water 625 564 477 309 
 Average B-factors (Å2    
  Protein 4.59 9.78 11.7 15.5 
  Ligand/ion 21.5 − 25.8 28.2 
  Water 21.8 23.2 27.5 29.9 
 RMSD     
  Bond length (Å) 0.0205 0.0301 0.0305 0.0214 
  Bond angles (°) 2.4143 2.4659 2.5151 2.1203 

The crystallographic atomic co-ordinates and structural factors have also been deposited in the PDB with accession codes 4ZXE for DD1, 4ZY9 for DD1-SeMet-V110M, 4ZZ5 for DD2, and 4ZZ8 for DD2 in complex with (GlcN)3.

Expression and purification of mutant proteins

Site-directed mutagenesis of the 36th amino acid was performed in DD1 and DD2 using pETB-DD1 and pETB-DD2 [13] as a template respectively by means of the QuikChange® site-directed mutagenesis kit (Stratagene). The mutagenic primers used for DD1-E36Y, DD1-E36Q, DD2-Y36E and DD2-Y36Q are listed in Supplementary Table S1. All mutated genes were sequenced to confirm the presence of the desired mutation and absence of unintended mutations. The mutated plasmids pETB-DD1-E36Y and pETB-DD1-E36Q were each transformed into E. coli BL21(DE3), whereas pETB-DD2-Y36E and pETB-DD2-Y36Q were each transformed into E. coli Rosetta(DE3). E. coli cells harbouring the plasmid pETB-DD1, pETB-DD1-E36Y, pETB-DD1-E36Q, pETB-DD2, pETB-DD2-Y36E or pETB-DD2-Y36Q were grown to obtain the corresponding mutated proteins following the same procedure as that employed for obtaining wild-type proteins.

NMR titration experiments

15N-labelled DD1, DD1-E36Y, DD1-E36Q, DD2, DD2-Y36E or DD2-Y36Q solution (60–100 μM) in 50 mM sodium acetate buffer (pH 5.0) containing 10% 2H2O was titrated with the incremental addition of (GlcN)4, and chemical shift perturbations were monitored by measuring 1H-15N-HSQC spectra. Chemical shift changes in HSQC signals induced by oligosaccharide binding (Δδ) were calculated using the equation:

 
formula
1

where ΔNH and ΔN represent the observed shifts in the 1H-axis and 15N-axis respectively.

Isothermal titration calorimetry (ITC)

Non-isotoplabelled DD1, DD1-E36Y, DD1-E36Q, DD2, DD2-Y36E and DD2-Y36Q solutions (63–92 μM) in 50 mM sodium acetate buffer (pH 5.0) were degassed and their concentrations were determined. (GlcN)4 was dissolved in the same buffer (1.5–10 mM), and the solution pH was adjusted to 5.0. (GlcN)4 solution was then degassed and loaded into a syringe, whereas the protein solution (0.2028 ml) was loaded into the sample cell. A calorimetric titration was performed with an iTC200 system (Microcal) at 298 K. Aliquots (1.0–2.0 μl) of ligand solution were added to the sample cell with a stirring speed of 1000 rev./min. Titrations were completed after 19–39 injections. When the c value (c=nKassoc•[M]t; where n is the stoichiometry, Kassoc is the association constant, and [M]t is the initial protein concentration) was less than the range of 10<c<100, we confirmed that all requirements proposed by Turnbull and Daranas [33] were fulfilled in the experiments. In the analysis of ITC data, Origin software installed in the ITC instrument was used. The One Set of Sites model was employed to fit the data. The n values were fixed at 1.0 for DD2, whereas the n value for DD1 was optimized together with the other variables, ΔHo and Kassoc.

RESULTS

Structures of DD1 and DD2

In DD1 and DD2, the backbone superimposition of ten structures with the lowest values of the target function (Figure 2) revealed excellent convergence, especially in the core β-sandwich region (RMSDs for DD1, 0.64±0.14 Å for backbone atoms and 1.07±0.08 Å for heavy atoms; RMSDs for DD2, 1.03±0.14 Å for backbone atoms and 1.55±0.08 Å for heavy atoms). The other structural statistics are listed in Table 1. DD1 adopted a β-sandwich fold characterized by two opposing β-sheets, which appeared to be formed by five antiparallel β-strands and three antiparallel β-strands respectively (Figure 2, upper panel). Five upward loops, which appeared to contain two short antiparallel β-strands and one short helix, and two downward loops were attached to the exterior portion of the structure. The DD2 structure also adopted a β-sandwich fold, which was closely similar to that of DD1 (Figure 2, lower panel). Five upward loops with one short helix and two downward loops were similarly found in the exterior portion.

Stereo view of the backbone superimposition of ten calculated NMR-based structures, which gave the lowest values for the target function, for DD1 (top) and DD2 (bottom) respectively

Figure 2
Stereo view of the backbone superimposition of ten calculated NMR-based structures, which gave the lowest values for the target function, for DD1 (top) and DD2 (bottom) respectively

N and C represent N-terminal and C-terminal ends of the proteins respectively. The structures were drawn using PyMOL (http://www.pymol.org).

Figure 2
Stereo view of the backbone superimposition of ten calculated NMR-based structures, which gave the lowest values for the target function, for DD1 (top) and DD2 (bottom) respectively

N and C represent N-terminal and C-terminal ends of the proteins respectively. The structures were drawn using PyMOL (http://www.pymol.org).

We then obtained the crystal structures of DD1 and DD2 in their free forms. Structural statistics are listed in Table 2. Secondary structures identified from the crystal structure are aligned with the amino acid sequences in Figure 1(B). Comparison of the overall crystal structures between DD1 and DD2 revealed a close similarity (RMSD 0.459 Å) (Figure 3A). The β-core structure of DD1 was formed by two β-sheets, each of which was composed of four antiparallel β-strands, β1, β2, β4(1) and β6, and two antiparallel β-strands, β3 and β5(1), respectively. The core structure of DD2 was formed by a five-stranded antiparallel β-sheet, β1, β2, β4(2), β5(2) and β6, and a two-stranded antiparallel β-sheet, β3 and β7. They were less similar to that reported for the CBM32 module in α-N-acetylglucosaminidase from Clostridium perfringens (PDB code 4AAX) with RMSDs of 0.819 Å to DD1 and 0.903 Å to DD2 [34], and markedly different from that of the CBM32 in polygalacturonase from Yersinia enterocolitica (PDB code 2JDA) with RMSDs of 2.386 Å to DD1 and 0.775 Å to DD2 [35]. The superimpositions of the solution and the crystal structures for individual modules are shown in Figures 3(B) and 3(C) respectively. In DD1 and DD2, the structures obtained by NMR spectroscopy were highly similar to those obtained by crystallography, especially in the β-core structure. However, conformational differences between solution and crystal structures were found in the three loop regions: 14–18, 35–40 and 75–82 (Figure 1B). These loops in the crystal structure were packed closer to the β-core structure, whereas those in the solution structure were more loosely packed to the core structure and fluctuated.

Comparison of the crystal structures of DD1 and DD2 to their NMR solution structures

Figure 3
Comparison of the crystal structures of DD1 and DD2 to their NMR solution structures

(A) Stereo view of the superimposition of crystal structures of DD1 (green) and DD2 (orange). Secondary structures are labelled according to the nomenclature shown in Figure 1(B). A conserved Ca2+ ion (green sphere) in DD1 was completely overlapped with that in DD2, and an additional Ca2+ (orange sphere) was observed only for DD2. (B and C) Stereo views of the superimposition of crystal (green/orange) and NMR-based structures (blue/yellow) for DD1 and DD2 respectively.

Figure 3
Comparison of the crystal structures of DD1 and DD2 to their NMR solution structures

(A) Stereo view of the superimposition of crystal structures of DD1 (green) and DD2 (orange). Secondary structures are labelled according to the nomenclature shown in Figure 1(B). A conserved Ca2+ ion (green sphere) in DD1 was completely overlapped with that in DD2, and an additional Ca2+ (orange sphere) was observed only for DD2. (B and C) Stereo views of the superimposition of crystal (green/orange) and NMR-based structures (blue/yellow) for DD1 and DD2 respectively.

A structural metal ion was found in DD1, and two ions were found in DD2. In Figure 3(A), metal ions were modelled as Ca2+, which is commonly found in CBMs belonging to the CBM32 family [34,35]. Heptahedral co-ordination of the metal (green sphere) was observed with backbone oxygen atoms of Arg22, Ser27, Thr30 and Trp124, and side-chain oxygen atoms of Asp25, Thr30 and Glu125 of DD1. DD2 heptahedrally co-ordinated with the conserved metal ion (green sphere) through the backbone oxygen atoms of Lys22, Asn27, Thr30 and Trp124, and side-chain oxygen atoms of Asp25, Thr30 and Glu126. DD2 also co-ordinated with an additional ion (orange sphere) through the backbone oxygen atoms of Gly94 and Ile96, the side-chain oxygen atoms of Asp93 and Asp97, and three bound waters.

Ligand-binding sites of DD1 and DD2

In order to define ligand-binding sites, (GlcN)4 was titrated into 15N-labelled DD1 or DD2 protein solution and chemical shift perturbations were monitored in 1H-15N-HSQC experiments. Since the molecular size of (GlcN)4 appeared to be sufficient to occupy the entire binding sites of DD1 and DD2 [13], this tetramer was used in an interaction analysis. When (GlcN)4 was titrated into the DD1 solution, the chemical shifts of several resonances were perturbed, as shown in Figure 4(A). The (GlcN)4 titration to DD1 was conducted until the molar ratio of (GlcN)4 to DD1 reached 10:1. No further changes in chemical shifts were observed when the molar ratio was more than 5. Resonances showed two types of chemical exchanges, fast or intermediate on the NMR timescale. The resonances of Glu14, Ala33, Glu36, Ala63 and Ser122 were significantly shifted by (GlcN)4 binding, and Δδ values were greater than 0.1 p.p.m. (Figure 4B). The intensities of the resonances of Gly93, Gly94 and Gly120 gradually decreased due to their line-broadening, and were finally beyond recognition. Figure 4(C) shows the solution structure of DD1, in which the amino acid residues responding to ligand binding are highlighted. The chitosan oligosaccharide is likely to bind to the loop region (top of the β-core structure) of DD1, including Glu14, Ala33 and Glu36 on the β1/β2 loop, Ala63 on the β3/β4(1) loop, and Ser122 in the C-terminal region following β6.

NMR-based titration experiments for DD1

Figure 4
NMR-based titration experiments for DD1

(A) 1H-15N-HSQC spectra of DD1 in the absence or presence of (GlcN)4. The titration was conducted in 50 mM sodium acetate buffer (pH 5.0) containing 10 % 2H2O. The solution temperature was 300 K. The molar ratios of (GlcN)4/DD1 were 0:1 (red), 1:1 (orange), 5:1 (green) and 10:1 (blue). (B) Chemical shift changes (Δδ) in backbone resonances affected by the addition of (GlcN)4 (molar ratio 5:1). Magenta bars are amino acids whose resonances shifted by more than 0.1 p.p.m. Blue bars are amino acids whose resonances lost their intensities due to line broadening. (C) NMR-based structure of DD1. Left: view from the side of the β-sandwich. Right: view from the top of the β-sandwich. Amino acid residues responding to the addition of (GlcN)4 are highlighted. Colours correspond to those in (B).

Figure 4
NMR-based titration experiments for DD1

(A) 1H-15N-HSQC spectra of DD1 in the absence or presence of (GlcN)4. The titration was conducted in 50 mM sodium acetate buffer (pH 5.0) containing 10 % 2H2O. The solution temperature was 300 K. The molar ratios of (GlcN)4/DD1 were 0:1 (red), 1:1 (orange), 5:1 (green) and 10:1 (blue). (B) Chemical shift changes (Δδ) in backbone resonances affected by the addition of (GlcN)4 (molar ratio 5:1). Magenta bars are amino acids whose resonances shifted by more than 0.1 p.p.m. Blue bars are amino acids whose resonances lost their intensities due to line broadening. (C) NMR-based structure of DD1. Left: view from the side of the β-sandwich. Right: view from the top of the β-sandwich. Amino acid residues responding to the addition of (GlcN)4 are highlighted. Colours correspond to those in (B).

In the case of DD2, the number of resonances affected by the addition of (GlcN)4 was greater than that in DD1, as shown in Figure 5(A). (GlcN)4 titration to DD2 was conducted until the molar ratio of (GlcN)4 to DD2 reached 25:1. No further changes in chemical shifts were observed at higher molar ratios. The sensitivity of DD2 to (GlcN)4 binding was markedly less than that of DD1, suggesting that the binding affinity of DD2 to (GlcN)n is lower than that of DD1. Resonances showed a fast exchange on the NMR timescale. The resonances of Ile13, Glu14, His18, Arg31, Trp32, Ala33, Gly37, Ala119, Tyr120 and Ser123 were significantly shifted by (GlcN)4 binding, and Δδ values were greater than 0.1 p.p.m. (Figure 5B). The resonances of Ala35, Tyr36 and Gly121 were line-broadened and beyond recognition. As shown in Figure 5(C), the binding site of DD2 appeared to include the 13–18 region on the β1/α1 loop, the 31–37 region on the α1/β2 loop and the 119–123 region on the β6/β7 loop. In DD1 and DD2, (GlcN)4 bound to an almost identical loop region extruded upward from the β-core structure.

NMR-based titration experiments for DD2

Figure 5
NMR-based titration experiments for DD2

(A) 1H-15N-HSQC spectra of DD2 in the absence or presence of (GlcN)4. The titration was conducted in 50 mM sodium acetate buffer (pH 5.0) containing 10 % 2H2O. The solution temperature was 300 K. The molar ratios of (GlcN)4/DD2 were 0:1 (red), 1:1 (orange), 5:1 (green) and 10:1 (blue). (B) Chemical shift changes (Δδ) in backbone resonances affected by the addition of (GlcN)4 (molar ratio 25:1). Magenta bars are amino acids whose resonances shifted by more than 0.1 p.p.m. Blue bars are amino acids whose resonances lost their intensities due to line broadening. The green zone indicates the amino acid residues with missing assignments. (C) NMR-based structure of DD2. Left: view from the side of the β-sandwich. Right: view from the top of the β-sandwich. Amino acid residues responding to the addition of (GlcN)4 are highlighted. Colours correspond to those in (B).

Figure 5
NMR-based titration experiments for DD2

(A) 1H-15N-HSQC spectra of DD2 in the absence or presence of (GlcN)4. The titration was conducted in 50 mM sodium acetate buffer (pH 5.0) containing 10 % 2H2O. The solution temperature was 300 K. The molar ratios of (GlcN)4/DD2 were 0:1 (red), 1:1 (orange), 5:1 (green) and 10:1 (blue). (B) Chemical shift changes (Δδ) in backbone resonances affected by the addition of (GlcN)4 (molar ratio 25:1). Magenta bars are amino acids whose resonances shifted by more than 0.1 p.p.m. Blue bars are amino acids whose resonances lost their intensities due to line broadening. The green zone indicates the amino acid residues with missing assignments. (C) NMR-based structure of DD2. Left: view from the side of the β-sandwich. Right: view from the top of the β-sandwich. Amino acid residues responding to the addition of (GlcN)4 are highlighted. Colours correspond to those in (B).

Crystal structure of DD2 in complex with (GlcN)3

We successfully obtained the crystal structure of DD2 in complex with (GlcN)3. As shown in Figure 6, bound (GlcN)3 was vertically oriented toward the binding site, and stood upright on the non-reducing end (G3). The reducing end GlcN (G1) of the bound (GlcN)3 exhibited poorer electron density than that of G3. The average B-factor values for the individual sugar residues were 30.7, 25.0 and 14.5 Å2 from G1 to G3 respectively. These results indicated that DD2 predominantly recognizes G3 of the bound (GlcN)3, whereas G1 is hardly recognized by DD2. Figure 6 also shows the amino acid residues of DD2 interacting with the ligand. The side-chain carboxylate of Glu14 appeared to interact strongly with the amino group of G3 probably through an electrostatic interaction. The phenolic hydroxy group of Tyr36 also interacted with the amino group of G3. The carboxy oxygen atom of the side chain of Glu14 interacted with the hydroxy group on C3 of G3. The guanidyl side chain of Arg31 formed three hydrogen bonds with the two hydroxy groups on C3 and C4 of G3. The side-chain carboxylate group of Glu61 also formed two hydrogen bonds with the hydroxy groups on C4 and C6 of G3. The phenolic hydroxy group of Tyr36 interacted with the hydroxy group on C6 of the middle GlcN residue (G2). In addition, the aromatic side chain of Tyr120 appeared to form a hydrophobic interaction with the pyranose ring G3. The interaction mechanism shown in Figure 6 is fully consistent with the electron densities of the bound GlcN residues, average B-factor values and NMR titration data shown in Figure 5. Since the metal-binding site is located separately from the ligand-binding site (Figure 3A), the chitosan-binding appears to be independent of the bound metal, as reported for the other CBM32 modules [35]. Comparison of crystal structures between the free and bound states of DD2 revealed that the main-chain conformation was hardly affected by (GlcN)3 binding (RMSD 0.092 Å). However, small conformational changes were identified in the side chains of Glu14 and Tyr36, which interacted with the ligand. The Glu14 side chain rotated at the carboxy end and the phenolic side chain of Tyr36 came slightly closer to bound (GlcN)3.

Stereo view of the crystal structure of DD2 in complex with (GlcN)3

Figure 6
Stereo view of the crystal structure of DD2 in complex with (GlcN)3

The (GlcN)3 molecule is displayed by a stick model coloured yellow. The simulated annealing-omit map of bound (GlcN)3 is coloured blue and contoured at 1.5 σ. Amino acid residues interacting with (GlcN)3 are highlighted by a stick model coloured cyan. Possible interactions are represented by broken lines. G1, G2, and G3 represent the reducing-end, middle and non-reducing end GlcN residues of bound (GlcN)3 respectively.

Figure 6
Stereo view of the crystal structure of DD2 in complex with (GlcN)3

The (GlcN)3 molecule is displayed by a stick model coloured yellow. The simulated annealing-omit map of bound (GlcN)3 is coloured blue and contoured at 1.5 σ. Amino acid residues interacting with (GlcN)3 are highlighted by a stick model coloured cyan. Possible interactions are represented by broken lines. G1, G2, and G3 represent the reducing-end, middle and non-reducing end GlcN residues of bound (GlcN)3 respectively.

Unfortunately, we failed to obtain the crystal structure of DD1 in complex with (GlcN)n. However, amino acid residues interacting with (GlcN)3 in DD2 (Glu14, Arg31, Tyr36, Glu61 and Tyr120) are conserved in DD1, except that Tyr36 is replaced by glutamic acid in DD1, as shown in Figure 1(B). In addition, the binding site of DD1 defined by the NMR titration experiments (Figure 4C) was almost identical with that of DD2 (Figure 5C). Thus the binding mode of DD1 to (GlcN)n may basically be similar to that of DD2.

Site-directed mutagenesis

As described above, Tyr36 in DD2 is replaced by glutamic acid in DD1 (Figure 1B). In order to examine the effects of substitutions of the 36th amino acid of DD1 and DD2, mutations were introduced at this site in both proteins, and the mutant proteins were characterized. Glu36 of DD1 was mutated to tyrosine and glutamine (DD1-E36Y and DD1-E36Q) respectively, and Tyr36 of DD2 was mutated to glutamic acid and glutamine (DD2-Y36E and DD2-Y36Q) respectively.

NMR-based titration experiments

The 1H-15N-HSQC spectra of DD1-E36Y, DD1-E36Q, DD2-Y36E and DD2-Y36Q exhibited profiles similar to those of the corresponding wild-type proteins; individual amide resonances were not affected by mutations, except for the main-chain NH resonance of the mutated amino acid. This result suggests that none of the mutants underwent major conformational changes. Titration experiments were conducted until the molar ratio of (GlcN)4 to the protein reached 1:1 in order to observe clear differences in affinity between the wild-type and mutants. As shown in Supplementary Figure S1(A), the amino acid residues of DD1 responding to the addition of (GlcN)4 were similar to those in Figure 4(B). However, when (GlcN)4 was titrated into DD1-E36Y solution, chemical shift changes (Δδ) were reduced and markedly smaller than those of DD1, as shown in Supplementary Figure S1(B). Effects of (GlcN)4 binding to DD1-E36Q on the amide resonances were also smaller than those of DD1 (Supplementary Figure S1C).

The resonances of DD2 were not significantly affected by (GlcN)4 binding when the molar ratio of (GlcN)4 to DD2 was at 1:1 (Figure S2A). In contrast, the addition of (GlcN)4 to DD2-Y36E strongly affected the resonances of amino acids located in the upward loops (Supplementary Figure S2B). Amino acid residues responding to (GlcN)4 binding were similar to those in Figure 5(B). The mutation of Tyr36 to glutamic acid in DD2 appeared to enhance the affinity for chitosan oligosaccharides. The addition of (GlcN)4 to DD2-Y36Q did not significantly affect the resonances of the protein (Figure S2C).

ITC experiments

DD1, DD1-E36Y, DD1-E36Q, DD2, DD2-Y36E and DD2-Y36Q were employed for (GlcN)4-binding experiments using ITC. The titration of (GlcN)4 to these proteins was performed at 25°C and pH 5.0. Individual thermograms obtained from titrations were successfully fitted with a one-site binding model, as shown in Figure 7, and thermodynamic parameters were obtained as listed in Table 3. (GlcN)4 was found to bind to DD1 with a Kassoc of 4.7×105 M−1. Binding was clearly enthalpy-driven (ΔH°=−8.9 kcal/mol) with a small entropy loss (−TΔS°=1.1 kcal/mol). Regarding DD1-E36Y, Kassoc was determined to be 5.6×103 M−1, which was markedly lower than that of DD1. The interaction between DD1-E36Y and (GlcN)4 was also enthalpy-driven to a similar extent to that of DD1 (ΔH°=−8.3 kcal/mol), whereas the entropy loss was higher (−TΔS°=3.2 kcal/mol) than that of DD1. The substitution of tyrosine for Glu36 was disadvantageous for chitosan oligosaccharide binding. (GlcN)4 bound to DD1-E36Q with Kassoc 2.1×104 M−1, which was lower than that of DD1, but higher than that of DD1-E36Y. This lower affinity was also derived predominantly from the higher entropy loss (−TΔS°=3.2 kcal/mol).

Table 3
Thermodynamic parameters for (GlcN)4 binding to DD1, DD1-E36Y, DD1-E36Q, DD2, DD2-Y36E and DD2-Y36Q obtained from ITC profiles shown in Figure 7 
Protein n Kassoc (M−1ΔH° (kcal/mol) ΔS° (cal/mol/°) TΔS° (kcal/mol) ΔG° (kcal/mol) 
DD1 1.2±0.1 (4.7±0.4)×105 −8.9±0.1 −3.8 1.1 −7.8 
DD1-E36Y 1.0 (5.6±0.1)×103 −8.3±0.1 −10.7 3.2 −5.1 
DD1-E36Q 0.8±0.1 (2.1±0.1)×104 −9.1±0.1 −10.7 3.2 −5.9 
DD2 1.0 (5.8±0.2)×103 −11.0±0.1 −19.7 5.8 −5.2 
DD2-Y36E 1.0 (3.1±0.1)×104 −9.2±0.1 −10.6 3.2 −6.0 
DD2-Y36Q 1.0 (2.1±0.1)×103 −11.2±0.2 −22.5 6.7 −4.5 
Protein n Kassoc (M−1ΔH° (kcal/mol) ΔS° (cal/mol/°) TΔS° (kcal/mol) ΔG° (kcal/mol) 
DD1 1.2±0.1 (4.7±0.4)×105 −8.9±0.1 −3.8 1.1 −7.8 
DD1-E36Y 1.0 (5.6±0.1)×103 −8.3±0.1 −10.7 3.2 −5.1 
DD1-E36Q 0.8±0.1 (2.1±0.1)×104 −9.1±0.1 −10.7 3.2 −5.9 
DD2 1.0 (5.8±0.2)×103 −11.0±0.1 −19.7 5.8 −5.2 
DD2-Y36E 1.0 (3.1±0.1)×104 −9.2±0.1 −10.6 3.2 −6.0 
DD2-Y36Q 1.0 (2.1±0.1)×103 −11.2±0.2 −22.5 6.7 −4.5 

Thermograms and binding isotherms with theoretical fits obtained for the binding of (GlcN)4 to 88 μM DD1 (A), 63 μM DD1-E36Y (B), 72 μM DD1-E36Q (C), 84 μM DD2 (D), 92 μM DD2-Y36E (E) and 72 μM DD2-Y36Q (F)

Figure 7
Thermograms and binding isotherms with theoretical fits obtained for the binding of (GlcN)4 to 88 μM DD1 (A), 63 μM DD1-E36Y (B), 72 μM DD1-E36Q (C), 84 μM DD2 (D), 92 μM DD2-Y36E (E) and 72 μM DD2-Y36Q (F)

The buffer used was 50 mM sodium acetate buffer (pH 5.0). Titrations to DD1 were conducted with 1.5 mM (GlcN)4 solution, whereas 10 mM (GlcN)4 solution was used for the other proteins.

Figure 7
Thermograms and binding isotherms with theoretical fits obtained for the binding of (GlcN)4 to 88 μM DD1 (A), 63 μM DD1-E36Y (B), 72 μM DD1-E36Q (C), 84 μM DD2 (D), 92 μM DD2-Y36E (E) and 72 μM DD2-Y36Q (F)

The buffer used was 50 mM sodium acetate buffer (pH 5.0). Titrations to DD1 were conducted with 1.5 mM (GlcN)4 solution, whereas 10 mM (GlcN)4 solution was used for the other proteins.

The Kassoc value for DD2 was markedly lower than that obtained for DD1 by two orders of magnitude, 5.8×103 M−1. The lower binding ability of DD2 was mainly derived from the higher entropy loss (−TΔS°=5.8 kcal/mol) than that of DD1. In the case of DD2-Y36E, the Kassoc value was 3.1×104 M−1, which was higher than that of DD2. The mutation of Tyr36 to glutamic acid in DD2 reduced the entropy loss (−TΔS°=3.2 kcal/mol), and consequently enhanced binding affinity to (GlcN)4. This enhancement was not intensive because of entropy–enthalpy compensation [36]. However, the mutation of Tyr36 to glutamine in DD2 elevated the entropy loss (−TΔS°=6.7 kcal/mol), and, accordingly, the binding affinity of DD2-Y36Q to (GlcN)4 was lower than that of DD2.

DISCUSSION

DD1 and DD2 have been classified into the CBM32 family, which is known as the largest and most diverse family. The members of this family are often found as modules associated with bacterial enzymes [37]. CBM32 modules found in multimodular enzymes from C. perfringens [34,3840] were reported previously to bind to a wide variety of carbohydrate ligands, such as galactose, N-acetylgalactosamine, N-acetylglucosamine and N-acetyl-lactosamine. Polygalacturonate and mannan polysaccharides have also been known to bind to a CBM32 module in polygalacturonase from Y. enterocolitica [35] and a CBM32 module in mannanase from C. thermocellum [41] respectively. Most CBM32 members adopt a β-sandwich fold, and several loops are extruded from the β-core structure. However, the extruded loops that create the carbohydrate-binding site adopt different conformations, and the amino acid sequences of the loop regions vary for individual members [37]. These differences may lead to the diverse binding specificities reported for CBM32 members. In the present study, a structural analysis of DD1 and DD2 showed that the binding site for (GlcN)n is located in the upward loop region including Glu14, Arg31, Glu36(DD1)/Tyr36(DD2), Glu61 and Tyr119(DD1)/Tyr120(DD2) for both binding modules, and also that substitution of the 36th amino acid differentiates the binding properties between DD1 and DD2. Bound (GlcN)3 was found to stand upright on its non-reducing end GlcN, which was predominantly recognized by the binding site (Figure 6). DD1 and DD2 belong to Type C CBMs [42], similar to the other CBM32 modules. Thus the chitosanase/glucanase may initially bind to the non-reducing end region of the chitosan polysaccharide chain, which may be translocated to the GH8 catalytic module to undergo enzymatic hydrolysis. The enzyme may then processively release small oligosaccharides (dimers and trimers) as enzymatic products from the non-reducing end of the substrate. Thus DD1 and DD2 may be important for the processive action of the enzyme.

In the binding site of a CBM32 of N-acetyl-β-hexosaminidase from C. perfringens, two aromatic residues create a hydrophobic pocket, and stack with galactose units [38]. Aromatic residues in a CBM32 of N-acetylglucosaminidase from C. perfringens also play a key role in the binding of N-acetylglucosamine units [39]. Carbohydrate-active proteins often employ aromatic residues for their interactions [43]. However, alteration of the 36th amino acid from glutamic acid to tyrosine reduced affinity to (GlcN)n, whereas the reverse substitution enhanced affinity (Table 3). The aromatic moiety at this site may be disadvantageous for (GlcN)n binding. Amino groups with positive charges in chitosan oligosaccharides favour binding to a hydrophilic moiety, particularly, carboxylates with negative charges [44,45]. The binding site of DD1 or DD2 includes three or two glutamic acids: Glu14, Glu36/Tyr36 and Glu61 respectively. Among these residues, the side chain of Glu14 or Glu36(DD1) appeared to interact most strongly with the amino group of the non-reducing end GlcN (Figure 6), and plays a key role in chitosan recognition. In fact, the proteins with three glutamic acids at the binding sites (DD1 and DD2-Y36E) exhibited higher affinity towards (GlcN)nG° in Table 3).

However, effects of the mutations on the thermodynamic parameters for chitosan binding were more intensive in the entropic term (−TΔS°) than in the enthalpic term (ΔH°) (Table 3), suggesting that the higher affinities of DD1 and DD2-Y36E may not only derive from enhanced electrostatic or hydrogen-bonding interactions. In the crystal structure of DD1 having three glutamic acids (Glu14, Glu36 and Glu61) in the ligand-binding site, the Glu14 side chain rotated away from the amino group of G3 (Supplementary Figure S3), and was unlikely to interact with the protein. Instead, the Glu36 side chain became closer to the amino nitrogen of G3, and appeared to interact with it. The electrostatic interaction with the amino group of G3 may be rearranged from Glu14 to Glu36 in DD1. The rearrangement may result in entropy gain for the Glu14 side chain, while keeping a comparable extent of enthalpy-driven interactions. Similar rearrangement may be possible in the DD2-Y36E mutant. This situation may be responsible at least partly for the smaller effect on ΔH° but larger effect on −TΔS°.

In contrast with DD1, DD2 has broad specificity for carbohydrates, as described in the Introduction. This difference in specificity may also be caused by the substitution of the 36th amino acid (glutamic acid/tyrosine). Aromatic residues in the binding site may enhance affinity to the non-polar parts of the ligand, but reduce affinity to the polar groups. DD2 with Tyr36 has been shown to interact with cello-oligosaccharides and laminarioligosaccharides, as well as (GlcN)n [13]. A combination of DD1 and DD2 with different specificities may allow the chitosanase/glucanase to bind not only to chitosan, but also to glucans, thereby enhancing total binding affinity of the enzyme to its target, namely the more extended region of complex carbohydrates in the fungal cell wall.

AUTHOR CONTRIBUTION

Shoko Shinya, Hisashi Kimoto, Hideo Kusaoke and Tamo Fukamizo designed the research. Shoko Shinya, Shigenori Nishimura and Yoshihito Kitaoku performed the research. Shoko Shinya, Shigenori Nishimura, Yoshihito Kitaoku, Tomoyuki Numata, Takayuki Ohnuma and Tamo Fukamizo analysed the data. Shoko Shinya and Tamo Fukamizo wrote the paper.

We thank Ms Hiromi Oi for assisting with protein purification and crystallization.

FUNDING

This work was supported by “Strategic Project to Support the Formation of Research Bases at Private Universities: Matching Fund Subsidy from the MEXT (Ministry of Education, Culture, Sports, Science and Technology, Japan)”, 2011–2015 [grant number S1101035]. S.S. was supported by a Research Fellowship for Young Scientists from the Japan Society for the Promotion of Science [grant number 25-3639].

Abbreviations

     
  • CBM

    carbohydrate-binding module

  •  
  • CCD

    charge-coupled device

  •  
  • DD

    discoidin domain

  •  
  • GH8

    glycoside hydrolase 8

  •  
  • GlcN

    glucosamine

  •  
  • ITC

    isothermal titration calorimetry

  •  
  • SeMet

    selenomethionine

References

References
1
Haas
D.
Défago
G.
Biological control of soil-borne pathogens by fluorescent pseudomonads
Nat. Rev. Microbiol.
2005
, vol. 
3
 (pg. 
307
-
319
)
[PubMed]
2
Latgé
J.P.
The cell wall: a carbohydrate armour for the fungal cell
Mol. Microbiol.
2007
, vol. 
66
 (pg. 
279
-
290
)
[PubMed]
3
Free
S.J.
Fungal cell wall organization and biosynthesis
Adv. Genet.
2013
, vol. 
81
 (pg. 
33
-
82
)
[PubMed]
4
Budi
S.W.
van Tuinen
D.
Arnould
C.
Dumas-Gaudut
E.
Gianinazzi-Pearson
V.
Gianinazzi
S.
Hydrolytic enzyme activity of Paenibacillus sp. strain B2 and effect of antagonistic bacterium on cell wall integrity of two soil-borne pathogenic fungi
Appl. Soil Ecol.
2000
, vol. 
15
 (pg. 
191
-
199
)
5
Kishore
G.K.
Pande
S.
Podile
A.R.
Biological control of late leaf spot of peanut (Arachis hypogaea) with chitinolytic bacteria
Phytopathology
2005
, vol. 
95
 (pg. 
1157
-
1165
)
[PubMed]
6
Theis
T.
Stahl
U.
Antifungal proteins: targets, mechanisms and prospective applications
Cell. Mol. Life Sci.
2004
, vol. 
61
 (pg. 
437
-
455
)
[PubMed]
7
Aktuganov
G.
Melentjev
A.
Galimzianova
N.
Khalikova
E.
Korpela
T.
Susi
P.
Wide-range antifungal antagonism of Paenibacillus ehimensis IB-X-b and its dependence on chitinase and β-1,3-glucanase production
Can. J. Microbiol.
2008
, vol. 
54
 (pg. 
577
-
587
)
[PubMed]
8
Kimoto
H.
Kusaoke
H.
Yamamoto
I.
Fujii
Y.
Onodera
T.
Taketo
A.
Biochemical and genetic properties of Paenibacillus glycosyl hydrolase having chitosanase activity and discoidin domain
J. Biol. Chem.
2002
, vol. 
277
 (pg. 
14695
-
14702
)
[PubMed]
9
Aragão
K.S.
Satre
M.
Imberty
A.
Varrot
A.
Structure determination of discoidin II from Dictyostelium discoideum and carbohydrate binding properties of the lectin domain
Proteins
2008
, vol. 
73
 (pg. 
43
-
52
)
[PubMed]
10
Mathieu
S.V.
Aragão
K.S.
Imberty
A.
Varrot
A.
Discoidin I from Dictyostelium discoideum and Interactions with oligosaccharides: specificity, affinity, crystal structures, and comparison with discoidin II
J. Mol. Biol.
2010
, vol. 
400
 (pg. 
540
-
554
)
[PubMed]
11
Kimoto
H.
Akamatsu
M.
Fujii
Y.
Tatsumi
H.
Kusaoke
H.
Taketo
A.
Discoidin domain of chitosanase is required for binding to the fungal cell wall
J. Mol. Microbiol. Biotechnol.
2010
, vol. 
18
 (pg. 
14
-
23
)
[PubMed]
12
Cheng
Y.-M.
Hsieh
F.-C.
Meng
M.
Functional analysis of conserved aromatic amino acids in the discoidin domain of Paenibacillus β-1,3-glucanase
Microb. Cell Fact.
2009
, vol. 
8
 pg. 
62
 
[PubMed]
13
Shinya
S.
Ohnuma
T.
Yamashiro
R.
Kimoto
H.
Kusaoke
H.
Anbazhagan
P.
Juffer
A.H.
Fukamizo
T.
The first identification of carbohydrate binding modules specific to chitosan
J. Biol. Chem.
2013
, vol. 
288
 (pg. 
30042
-
30053
)
[PubMed]
14
Pace
C.N.
Vajdos
F.
Fee
L.
Grimsley
G.
Gray
T.
How to measure and predict the molar adsorption coefficient of a protein
Protein Sci.
1995
, vol. 
4
 (pg. 
2411
-
2423
)
[PubMed]
15
Kay
L.E.
Ikura
M.
Tschudin
R.
Bax
A.
Three-dimensional triple-resonance NMR spectroscopy of isotopically enriched proteins
J. Magn. Reson.
1990
, vol. 
89
 (pg. 
496
-
514
)
16
Grzesiek
S.
Bax
A.
Correlating backbone amide and site chain resonances in larger proteins by multiple relayed triple resonance NMR
J. Am. Chem. Soc.
1992
, vol. 
114
 (pg. 
6291
-
6293
)
17
Bax
A.
Clore
G.M.
Gronenborn
A.M.
1H-1H correlation via isotropic mixing on 13C magnetization, a new three-dimensional approach for assigning 1H and 13C spectra of 13C-enriched proteins
J. Magn. Reson.
1990
, vol. 
88
 (pg. 
425
-
431
)
18
Marion
D.
Driscoll
P.C.
Kay
L.E.
Wingfield
P.T.
Bax
A.
Gronenborn
A.M.
Clore
G.M.
Overcoming the overlap problem in the assignment of proton NMR spectra of larger proteins by use of three-dimensional heteronuclear 1H-15N Hartmann-Hahn-multiple quantum coherence and nuclear Overhauser-multiple quantum coherence spectroscopy: application to interleukin 1β
Biochemistry
1989
, vol. 
28
 (pg. 
6150
-
6156
)
[PubMed]
19
Shen
Y.
Delaglio
F.
Cornilescu
G.
Bax
A.
TALOS+: a hybrid method for predicting protein backbone torsion angles from NMR chemical shifts
J. Biomol. NMR
2009
, vol. 
44
 (pg. 
213
-
223
)
[PubMed]
20
Delaglio
F.
Grzesiek
S.
Vuister
G.W.
Zhu
G.
Pfeifer
J.
Bax
A.
NMRPipe: a multidimensional spectral processing system based on UNIX pipes
J. Biomol. NMR
1995
, vol. 
6
 (pg. 
277
-
293
)
[PubMed]
21
Güntert
P.
Automated NMR structure calculation with CYANA
Methods Mol. Biol.
2004
, vol. 
278
 (pg. 
353
-
378
)
[PubMed]
22
Laskowski
R.A.
MacArthur
M.W.
Moss
D.S.
Thornton
J.M.
PROCHECK: a program to check the stereochemical quality of protein structures
J. Appl. Crystallogr.
1993
, vol. 
26
 (pg. 
283
-
291
)
23
Otwinowski
Z.
Minor
W.
Processing of X-ray diffraction data collected in oscillation mode
Methods Enzymol.
1997
, vol. 
276
 (pg. 
307
-
326
)
24
Hendrickson
W.A.
Teeter
M.M.
Structure of the hydrophobic protein crambin determined directly from the anomalous scattering of sulphur
Nature
1981
, vol. 
290
 (pg. 
107
-
113
)
25
Terwilliger
T.C.
Adams
P.D.
Read
R.J.
McCoy
A.J.
Moriarty
N.W.
Grosse-Kunstleve
R.W.
Afonine
P.V.
Zwart
P.H.
Hung
L.W.
Decision-making in structure solution using Bayesian estimates of map quality: the PHENIX AutoSol wizard
Acta Crystallogr. D Biol. Crystallogr.
2009
, vol. 
65
 (pg. 
582
-
601
)
[PubMed]
26
Adams
P.D.
Afonine
P.V.
Bunkóczi
G.
Chen
V.B.
Davis
I.W.
Echols
N.
Headd
J.J.
Hung
L.W.
Kapral
G.J.
Grosse-Kunstleve
R.W.
, et al. 
PHENIX: a comprehensive Python-based system for macromolecular structure solution
Acta Crystallogr. D Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
213
-
221
)
[PubMed]
27
McCoy
A.J.
Grosse-Kunstleve
R.W.
Storoni
L.C.
Read
R.J.
Likelihood-enhanced fast translation functions
Acta Crystallogr. D Biol. Crystallogr.
2005
, vol. 
61
 (pg. 
458
-
464
)
[PubMed]
28
Bailey
S.
The CCP4 suite: programs for protein crystallography
Acta Crystallogr. D Biol. Crystallogr.
1994
, vol. 
50
 (pg. 
760
-
763
)
[PubMed]
29
Murshudov
G.N.
Vagin
A.A.
Dodson
E.J.
Refinement of macromolecular structures by the maximum-likelihood method
Acta Crystallogr. D Biol. Crystallogr.
1997
, vol. 
53
 (pg. 
240
-
255
)
[PubMed]
30
Afonine
P.V.
Grosse-Kunstleve
R.W.
Echols
N.
Headd
J.J.
Moriarty
N.W.
Mustyakimov
M.
Terwilliger
T.C.
Urzhumtsev
A.
Zwart
P.H.
Adams
P.D.
Towards automated crystallographic structure refinement with phenix.refine
Acta Crystallogr. D Biol. Crystallogr.
2012
, vol. 
68
 (pg. 
352
-
367
)
[PubMed]
31
Emsley
P.
Cowtan
K.
Coot: model-building tools for molecular graphics
Acta Crystallogr. D Biol. Crystallogr.
2004
, vol. 
60
 (pg. 
2126
-
2132
)
[PubMed]
32
Lovell
S.C.
Davis
I.W.
Arendall
W.B.
III
de Bakker
P.I.
Word
J.M.
Prisant
M.G.
Richardson
J.S.
Richardson
D.C.
Structure validation by Cα geometry: φ, ψ and Cβ deviation
Proteins
2002
, vol. 
50
 (pg. 
437
-
450
)
33
Tumbull
W.B.
Daranas
A.H.
On the value of c: can low affinity systems be studied by isothermal titration calorimetry?
J. Am. Chem. Soc.
2003
, vol. 
125
 (pg. 
14859
-
14866
)
[PubMed]
34
Ficko-Blean
E.
Stuart
C.P.
Suits
M.D.
Cid
M.
Tessier
M.
Woods
R.J.
Boraston
A.B.
Carbohydrate recognition by an architecturally complex α-N-acetyl-glucosaminidase from Clostridium perfringens
PLoS One
2012
, vol. 
7
 pg. 
e33524
 
[PubMed]
35
Abbott
D.W.
Hrynuik
S.
Boraston
A.B.
Identification and characterization of a novel periplasmic polygalacturonic acid binding protein from Yersinia enterolitica
J. Mol. Biol.
2007
, vol. 
367
 (pg. 
1023
-
1033
)
[PubMed]
36
Chodera
J.D.
Mobley
D.L.
Entropy–enthalpy compensation: role and ramifications in biomolecular ligand recognition and design
Annu. Rev. Biophys.
2013
, vol. 
42
 (pg. 
121
-
142
)
[PubMed]
37
Abbott
D.W.
Eirín-López
J.M.
Boraston
A.B.
Insight into ligand diversity and novel biological roles for family 32 carbohydrate-binding modules
Mol. Biol. Evol.
2008
, vol. 
25
 (pg. 
155
-
167
)
[PubMed]
38
Ficko-Blean
E.
Boraston
A.B.
The interaction of a carbohydrate-binding module from a Clostridium perfringens N-acetyl-β-hexosaminidase with its carbohydrate receptor
J. Biol. Chem.
2006
, vol. 
281
 (pg. 
37748
-
37757
)
[PubMed]
39
Ficko-Blean
E.
Gregg
K.J.
Adams
J.J.
Hehemann
J.H.
Czjzek
M.
Smith
S.P.
Boraston
A.B.
Portrait of an enzyme, a complete structural analysis of a multimodular β-N-acetylglucosaminidase from Clostridium perfringens
J. Biol. Chem.
2009
, vol. 
284
 (pg. 
9876
-
9884
)
[PubMed]
40
Fujita
M.
Tsuchida
A.
Hirata
A.
Kobayashi
N.
Goto
K.
Osumi
K.
Hirose
Y.
Nakayama
J.
Yamanoi
T.
Ashida
H.
Mizuno
M.
Glycoside hydrolase family 89 α-N-acetylglucosaminidase from Clostridium perfringens specifically acts on GlcNAcα1,4Galβ1R at the non-reducing terminus of O-glycans in gastric mucin
J. Biol. Chem.
2011
, vol. 
286
 (pg. 
6479
-
6489
)
[PubMed]
41
Mizutani
K.
Fernandes
V.O.
Karita
S.
Luís
A.S.
Sakka
M.
Kimura
T.
Jackson
A.
Zhang
X.
Fontes
C.M.
Gilbert
H.J.
Sakka
K.
Influence of a mannan binding family 32 carbohydrate binding module on the activity of the appended mannanase
Appl. Environ. Microbiol.
2012
, vol. 
78
 (pg. 
4781
-
4787
)
[PubMed]
42
Boraston
A.B.
Bolam
D.N.
Gilbert
H.J.
Davies
G.J.
Carbohydrate-binding modules: fine-tuning polysaccharide recognition
Biochem. J.
2004
, vol. 
382
 (pg. 
769
-
781
)
[PubMed]
43
Fernández-Alonso
M.C.
Cañada
F.J.
Jiménez-Barbero
J.
Cuevas
G.
Molecular recognition of saccharides by proteins: insights on the origin of the carbohydrate-aromatic interactions
J. Am. Chem. Soc.
2005
, vol. 
127
 (pg. 
7379
-
7386
)
[PubMed]
44
Katsumi
T.
Lacombe-Harvey
M.E.
Tremblay
H.
Brzezinski
R.
Fukamizo
T.
Role of acidic amino acid residues in chitooligosaccharide-binding to Streptomyces sp. N174 chitosanase
Biochem. Biophys. Res. Commun.
2005
, vol. 
338
 (pg. 
1839
-
1844
)
[PubMed]
45
Tremblay
H.
Yamaguchi
T.
Fukamizo
T.
Brzezinski
R.
Mechanism of chitosanase–oligosaccharide interaction: subsite structure of Streptomyces sp. N174 chitosanase and the role of Asp57 carboxylate
J. Biochem.
2001
, vol. 
130
 (pg. 
679
-
686
)
[PubMed]