Structural dissection of the reaction mechanism of cellobiose phosphorylase

Abstract

Cellobiose phosphorylase, a member of the glycoside hydrolase family 94, catalyses the reversible phosphorolysis of cellobiose into α-D-glucose 1-phosphate and D-glucose with inversion of the anomeric configuration. The substrate specificity and reaction mechanism of cellobiose phosphorylase from Cellvibrio gilvus have been investigated in detail. We have determined the crystal structure of the glucose-sulphate and glucose-phosphate complexes of this enzyme at a maximal resolution of 2.0 Å (1 Å=0.1 nm). The phosphate ion is strongly held through several hydrogen bonds, and the configuration appears to be suitable for direct nucleophilic attack to an anomeric centre. Structural features around the sugar-donor and sugar-acceptor sites were consistent with the results of extensive kinetic studies. When we compared this structure with that of homologous chitobiose phosphorylase, we identified key residues for substrate discrimination between glucose and N-acetylglucosamine in both the sugar-donor and sugar-acceptor sites. We found that the active site pocket of cellobiose phosphorylase was covered by an additional loop, indicating that some conformational change is required upon substrate binding. Information on the three-dimensional structure of cellobiose phosphorylase will facilitate engineering of this enzyme, the application of which to practical oligosaccharide synthesis has already been established.

INTRODUCTION

The enzymes involved in oligo- and poly-saccharide degradation and synthesis can be categorized into two major classes of Carbohydrate-Active enZymes, GH (glycoside hydrolases) and GT (glycosyl transferases) [1] (see CAZy website at http://www.cazy.org/CAZY). Enzymes of a third class, phosphorylases, catalyse phosphorolysis at glycosidic bonds, generating glycosylphosphates. This reaction is reversible, since the energy of a glycosyl–phosphate bond is not as high as that of a glycosyl–nucleotide bond, which acts as the sugar-donor for general GTs. Therefore, phosphorylases can be employed for both the synthesis and degradation of sugar chains.

Phosphorylases are found in both the GT and GH classes. GT-type phosphorylases have a similar fold and catalytic mechanism as GT enzymes. For example, glycogen phosphorylase (GT-35) has a GT-B fold [2–5] and also exhibits a strong resemblance in its active site to typical GT enzymes, such as OtsA (GT-20) [6] and glycogen synthase (GT-5) [7]. GH-type phosphorylases, such as GH-13, GH-65 and GH-94, have structures and catalytic mechanisms similar to those of GH enzymes. For example, maltose phosphorylase (GH-65) and ChBP [chitobiose (GlcNAcβ1–4GlcNAc) phosphorylase; GH-94] share a glucoamylase-like (α/α)₆-barrel fold [8,9], and their proposed reaction mechanisms are similar to those of hydrolytic enzymes with the exception of molecules attacking the C1 atom (the hydroxyl ion in hydrolytic enzymes; the phosphate in phosphorylases).

CBP (cellobiose phosphorylase) catalyses the phosphorolysis of cellobiose into Glc 1-P (α-D-glucose 1-phosphate) and D-glucose, with inversion of the anomeric configuration. CBP also possesses cellobiose synthesis activity, using Glc 1-P as its sugar-donor and glucose as its sugar-acceptor. CBP has been classified into the GH-94 family, along with homologous ChBP, CDP (cello-dextrin phosphorylase), and cyclic-1,2-glucan synthase. Since none of these showed hydrolytic activity, GH-94 enzymes were formerly regarded as GT enzymes (GT-36). However, determination of the three-dimensional structure of VpChBP (chitobiose phosphorylase from Vibrio proteolyticus) revealed that it bore structural and mechanistic similarity to GH-15 glucoamylase and GH-65 maltose phosphorylase [8,10]. These findings led to the reclassification of GT-36 into a novel GH family, GH-94.

CgCBP (CBP from Cellvibrio gilvus) is the most studied enzyme in the GH-94 family [11–17]. CgCBP consists of 822 amino acid residues with a 33% identity with the amino acid structure of VpChBP. Its substrate specificity, especially its sugar-acceptor specificity in synthetic reactions, has been investigated in detail. Knowledge of the characteristics of this enzyme enabled us to synthesize several hetero-oligosaccharides {e.g. the branched trisaccharides, [Glcβ1–4(Glcα1–6)Glc)]} [17]. To better understand the function of CgCBP in practical oligosaccharide synthesis, it is necessary to determine its three-dimensional structure. In the present paper, we report the three-dimensional structure of two sugar–anion complexes (glucose–sulphate and glucose–phosphate) of CgCBP. The latter is the first such determination of a sugar–phosphate complex structure for an inverting phosphorylase. These structures have enabled us to examine the structural and functional differences between CgCBP and VpChBP.

MATERIALS AND METHODS

Sample preparation, crystallization and data collection

Production and purification of the CgCBP protein were performed as described in [18]. CgCBP crystals were obtained at 25 °C using the sitting-drop vapour-diffusion method by mixing 2 μl of a protein solution (10 mg/ml) with 2 μl of a reservoir solution consisting of 1.5 M ammonium sulphate, 0.1 M Mes/NaOH (pH 6.5) and 5 mM glucose (for CgCBP–SO₄), or 0.8 M sodium/potassium phosphate (pH 8.2) and 10 mM glucose at 4 °C (for CgCBP–PO₄). The crystals were transferred to a reservoir solution containing glycerol [20% (v/v) for CgCBP–SO₄, and 40% (v/v) for CgCBP–PO₄] and flash-cooled in a stream of cold nitrogen gas at 100 K. X-ray-diffraction data sets were collected using synchrotron radiation [beam line BL-5A and NW12A, Photon Factory, High Energy Accelerator Research Organization (KEK), Tsukuba, Japan], and these data sets were processed and scaled using HKL2000 [19]. Data collection and processing statistics are shown in Table 1.

Table 1. Crystallographic data collection and refinement statistics.

	Crystal
Parameter	CgCBP–SO4	CgCBP–PO4
Data collection statistics
Beamline	PF BL-5A	PF-AR NW12A
Wavelength (Å)	1.0000	1.0000
Space group	P21	P21
Unit-cell parameters
a (Å)	84.8	85.3
b (Å)	98.4	98.8
c (Å)	104.0	105.1
β (°)	102.7	102.5
Resolution (Å)*	50.00–2.00 (2.07–2.00)	50.00–2.10 (2.18–2.10)
Measured reflections	392667	377058
Unique reflections	111886	99298
Completeness (%)	97.0 (92.1)	100.0 (100.0)
Mean I/σ	8.0 (2.8)	8.1 (3.3)
Rmerge (%)†	6.8 (30.7)	9.4 (30.3)
Refinement statistics
Resolution range (Å)	49.21–2.00	41.86–2.10
Rfactor/Rfree‡ (%)	17.6/21.3	18.0/22.0
No. of protein atoms	12856	12856
No. of solvent atoms	1074	931
No. of heteroatoms	34	46
Average B-factors (Å2)	23.7	22.0
Protein (Chain A/B)	22.8/23.7	20.7/22.7
Glucose	23.8/30.6	24.1/32.8
Sulphate	24.9/24.7
Phosphate		19.5/18.0
Glycerol		25.3/27.9
Water	28.5	25.8
Rmsd bond lengths (Å)	0.005	0.006
Rmsd bond angles (°)	1.3	1.3

*Values in parentheses are for the highest resolution shell.

†R_merge=Σ_hΣ_i|I(h,i)−〈I(h)〉|/Σ_hΣ_iI(h,i), where I(h,i) is the intensity of the ith measurement of reflection h and 〈I(h)〉 is the average value over multiple measurements.

‡Calculated using a test data set; 5% of total data randomly selected from the observed reflections.

Phase calculation and refinement

Phase calculation was performed by molecular replacement with program MOLREP [20] in the CCP4 program suite [21] against the data set of CgCBP–SO₄. The crystal structure of VpChBP (Protein Data Bank entry 1V7W) was used as a search model. Program ARP/wARP [22] was used for automatic model building. The structure of CgCBP–PO₄ was refined starting from the refined CgCBP–SO₄ structure. Visual inspection of the models was performed using XtalView [23]. Simulated annealing, energy minimization, and individual B factor refinement were carried out using CNS1.1 [24]. The Figures were prepared using MOLSCRIPT [25], RASTER3D [26], XtalView, SPOCK [27], and PyMol [http://www.pymol.org]. Superimposition of CgCBP and VpChBP structures were carried out with program TOPP [29] in the CCP4 program suite.

Enzyme assays and kinetic analysis

1,5-Anhydroglucitol, isomaltose, and melibiose were purchased from Sigma–Aldrich. Measurement of the synthetic reaction was carried out as described previously [11]. The initial rate was assayed at 37 °C in 50 mM Tris/HCl (pH 7.0) by measuring the amount of phosphate released from 10 mM Glc 1-P with a suitable concentration of a glucosyl acceptor [30]. The kinetic parameters were calculated by regressing the experimental data by the curve fit method using GraFit Version 4.0 (Erithacus Software, Surrey, U.K.).

RESULTS AND DISCUSSION

Crystallography and overall structure

CgCBP crystals were obtained either complexed with sulphate (CgCBP–SO₄) or complexed with phosphate (CgCBP–PO₄). CgCBP–SO₄ and CgCBP–PO₄ structures were determined at 2.0 and 2.1 Å (1 Å=0.1 nm) resolution to R factor (R_free) of 17.6% (21.3%) and 18.0% (22.0%) respectively (Table 1). The crystals belong to the space group P2₁ and contain two subunits per asymmetric unit (Figure 1A). The structures of the four subunits (CgCBP–SO₄ chain A/B and CgCBP–PO₄ chain A/B) are almost identical. The rmsd (root-mean square deviation) for Cα atoms between all pairs of these subunits was within 0.3 Å.

Figure 1. Ribbon diagrams of the CgCBP–PO₄ structure.

(A) CgCBP dimer in an asymmetric unit. One subunit is shown in grey. (B) CgCBP–PO₄ monomer (N-terminal domain, blue; linker helices, green; α-helix barrel domain, yellow; C-terminal domain, red). Bound glucose, glycerol and phosphate are shown as ball-and-stick models.

The two molecules in an asymmetric unit form a dimer (Figure 1A), which seems to correspond to the dimeric form in solution [31]. The rmsd for Cα atoms between the dimer was 0.27 Å, both in CgCBP–SO₄ and CgCBP–PO₄. The two subunits are tightly connected through hydrophobic interactions and a number of hydrogen bonds. Upon dimer formation, 3200 Å² of solvent-accessible surface area (12% of the total monomer surface) is buried. This dimer form is similar to that of VpChBP. Since the residues involved in dimer interaction are conserved, as well as contributing to the formation of the active-site pocket (see below), it is likely that GH-94 CBPs from other sources also form dimers, although some CBPs have been reported to be monomeric in solution [32,33].

Figure 1(B) shows a ribbon diagram of the monomeric structure of CgCBP–PO₄, consisting of 822 amino acid residues from the original N-terminal methionine residue; an additional sequence, carrying a His₆-tag at the N-terminal end, is disordered. Structurally, the CgCBP monomer exhibits a complex architecture, consisting of four distinct regions, similar to that of VpChBP: an N-terminal domain (residues 1–289), a helical linker region (residues 290–320), an α-helix barrel domain (residues 329–745), and a C-terminal domain (residues 321–328 and 746–822). The rmsd between CgCBP and VpChBP is 1.58 Å for the Cα atoms of 766 residues.

Complex structures with SO₄ and PO₄

We have previously reported the ternary (protein–sugar–anion) structure of VpChBP, complexed with two GlcNAc, at subsites −1 and +1, and an SO₄ anion [8]. The SO₄-binding site was located in the vicinity of the N-termini of two helices, making it also suitable for other anions, such as chloride or PO₄. Due to the similar shapes and electrostatic properties of phosphate and sulphate ions, it is likely that the SO₄-binding site is also the PO₄-binding site.

We were able to obtain two protein–sugar–anion complex structures of CgCBP, CgCBP–SO₄ and CgCBP–PO₄, in which the SO₄ and PO₄ were derived from the precipitants used for crystallization. The SO₄ or PO₄ was located at the anion-binding site, and a glucose moiety was located at subsite +1 (Figures 2A and B). Kinetic analysis has shown that the binding affinity of subsite −1 for glucose is much weaker; the K_i of glucose to subsites −1 and +1 has been estimated to be 170 mM and 3.0 mM respectively [34]. The glucose molecule at subsite +1 takes on a chair (⁴C₁) conformation in both of the CgCBP complex structures. This glucose overlaps with the GlcNAc molecule at subsite +1 in the VpChBP structure (Figure 2C), enabling us to compare sugar recognition at this site in detail (see below). Interestingly, the CgCBP–PO₄ contained a glycerol at subsite −1 (Figure 2B; see below), whereas subsite −1 of CgCBP–SO₄ was empty.

Figure 2. Active site structure of CgCBP.

(A) Wireframe model of CgCBP–SO₄, and the unbiased |F_obs| − |F_calc| electron-density map of the bound glucose and phosphate (3.0σ). The residues (single-letter amino acid codes are used) involved in subsite formation (R362, W488, E649, Y653, E659, and Q165 from the adjacent subunit), sulphate recognition (H666), and catalysis (D490) are labelled. Glucose takes on the β-anomer configuration and forms a hydrogen bond with E649 (green broken line). (B) Wireframe model of CgCBP–PO₄, and the unbiased |F_obs| − |F_calc| electron-density map of the bound glucose, glycerol, and phosphate (3.0σ). (C) Stereo view of CgCBP–PO₄ and VpChBP superimposed at the active site. The backbone traces of CgCBP and VpChBP are coloured yellow and black, respectively. The bound glucose, glycerol, and phosphate ion in CgCBP (yellow), and the GlcNAc molecules and sulphate ion in VpChBP (black) are shown as wireframe models. Subsites −1 and +1 are labelled (−1) and (+1), respectively. The structurally conserved residues are shown as a wireframe model and labelled in the order corresponding to CgCBP and VpChBP. The 495–513 loop in CgCBP is coloured red. The residues related to N-acetyl/hydroxy group recognition at the C-2 position in each subsite are indicated by arrowheads. The oxygen atom suitable for direct attack on the C-1 atom of glucose at subsite −1 is circled in pink. Single-letter amino acid codes are used.

The bound forms of SO₄ and PO₄ were almost identical. Among the highly conserved residues involved in anion recognition are Arg³⁵¹, His⁶⁶⁶, Thr⁷³¹, Gly⁷³², and Thr⁷³³ (Figure 3). Interestingly, the orientations of the bound SO₄ and PO₄ are different, in that one P–O bond points to His⁶⁶⁶, whereas one S–O bond points in the opposite direction (Figure 3). The pH values for crystallization (pH 8.2 for CgCBP–PO₄ and pH 6.5 for CgCBP–SO₄) suggest that the anions are probably in states with two negative charges (HPO₄²⁻ and SO₄²⁻). Therefore, the different orientations of the bound anions may be caused by an interaction with glycerol, which was present only in CgCBP–PO₄, not by different charge states.

Figure 3. Stereo view of superimposition around the anion binding site of CgCBP–PO₄ and CgCBP–SO₄.

The phosphate in CgCBP- PO₄ and the sulphate in CgCBP–SO₄ are coloured pink and green respectively. Broken lines indicate hydrogen bonds between CgCBP and the anion (red for phosphate and cyan for sulphate). The two oxygen atoms held by T731, G732 and R351 are indicated by arrowheads. Single-letter amino acid codes are used.

In PO₄ recognition, three hydrogen bonds, i.e. those with the side-chain hydroxy group of Thr⁷³¹, the main-chain NH group of Gly⁷³², and the N^εH group of Arg³⁵¹, hold two of the four oxygen atoms of PO₄ (indicated by arrowheads in Figure 3). The other two oxygen atoms, which are positioned differently from those of SO₄, form two additional hydrogen bonds, one with the N^δ atom of His⁶⁶⁶ and the other, which is held by Gln⁷¹² and glycerol (circled in Figure 2C), appears to be suitable for the direct attack on the C-1 atom of glucose at subsite −1. The structure of the CgCBP–PO₄ complex supports our hypothesis that the reaction mechanism of inverting phosphorylases begins with direct nucleophilic attack by phosphate on the anomeric centre [8].

Sugar-donor specificity (subsite −1)

Subsites −1 and +1 in the phosphorolytic reaction of CBP act as sugar-donor and sugar-acceptor sites in the synthetic reaction, respectively. Although we could not obtain a complex structure of CgCBP with glucose at subsite −1 (sugar-donor site), a glycerol, which was derived from the cryoprotectant, was bound to subsite −1 in CgCBP–PO₄. Glycerol occasionally occupies sugar-binding subsites [35]. The carbon and oxygen atoms of glycerol in CgCBP almost completely overlap the C-4 to C-6 and O-4 to O-6 atoms of GlcNAc in VpChBP (Figure 2C). The average distance between these atoms was 0.42 Å (maximum 0.65 Å) when the polypeptide chains of CgCBP and VpChBP were superimposed by the least-squares method. Although the sugar ring at subsite −1 may form a reaction intermediate with an oxocarbonium ion-like conformation during phosphorolysis [8], the shift of the C-4, C-5, and C-6 positions is likely to be relatively small. Therefore, the structure of the complex with glycerol does not conflict with the reaction mechanism.

Residues surrounding subsite −1 are basically conserved between CgCBP and VpChBP. At the C-2 position of the substrate sugar, however, they carry a different group, hydroxy in CBP and N-acetyl in ChBP. In the VpChBP structure, Arg³⁴³ and Asp³⁵⁰ form hydrogen bonds with the N-acetyl group (Figure 2C). These residues are conserved in CgCBP (Arg³⁶² and Asp³⁶⁸), but the arginine side chain is significantly close to the sugar substrate when the two structures are superimposed. This arginine side chain sterically hinders an N-acetyl group at this site (indicated by a black arrowhead in Figure 2C). These structural features at this subsite are consistent with the substrate specificities of these enzymes, in that CBP cannot accommodate GlcNAc 1-P, whereas ChBP can accommodate both GlcNAc 1-P and Glc 1-P [36].

Sugar-acceptor specificity (subsite +1)

The electron density of glucose bound to subsite +1 (sugar-acceptor site) was clear in shape. Therefore, we could confidently determine the conformation of the glucose. When we compared the structures of CgCBP and VpChBP, we found that the Asp^490/492, Lys^658/636, Glu^659/637, and Gln^712/690 residues were conserved. In addition, the Gln^165/168 residue, in the N-terminal domain of the other monomer, was conserved to form the active-site pocket (Figure 2C). As with the O-1 atom at the reducing end of cellobiose, which shows strict β-anomer specificity [37], the reducing end O-1 atom of glucose formed a hydrogen bond with Glu⁶⁴⁹ of CgCBP (green broken line in Figures 2A and 2B), and electron density showed that the glucose moiety was exclusively in the β-anomer configuration. In contrast, VpChBP, which shows no anomer specificity at the reducing-end GlcNAc, did not form hydrogen bonds with the O-1 atom of GlcNAc, and electron density showed a mixture of the α- and β-anomer configurations [8]. CgCBP and VpChBP also show distinct specificities for β-substituted sugar-acceptors. Unlike VpChBP, which can phosphorolyse β-substituted chitobioses such as o-nitrophenyl chitobiose [38], CgCBP cannot [34]; this is possibly due to the existence of a loop, consisting of residues 495–513, in CgCBP (see below).

Differences in C-2 group specificity at subsite +1 can be clearly explained by the respective structural features of the two enzymes. The substrate-binding pocket of VpChBP is locally hydrophobic (indicated by a green arrowhead in Figure 2C). The Val⁶³¹ residue is involved in the formation of this small hydrophobic pocket, which contains the methyl portion of the N-acetyl group of GlcNAc. In CgCBP, the area corresponding to the small hydrophobic pocket of VpChBP is filled by a bulky residue, Tyr⁶⁵³, which forms a hydrogen bond with the bound glucose. These residues, valine in ChBP and tyrosine in CBP, are fully conserved in each enzyme group.

At subsite +1 of CgCBP, the kinetic parameters for several analogues of glucose have been determined (Table 2). The Gibbs energy of activation difference (the ΔΔG values) and the net strengths of hydrogen-bonding interactions can be estimated from the differences in catalytic efficiency of enzyme for each analogue compared with that for glucose [32,39]. The ΔΔG values of 1,5-anhydroglucitol (4.4 kcal/mol; 1 kcal=4.184 kJ) and 2-deoxyglucose (3.3 kcal/mol), relative to D-glucose, suggest that the O-1 and O-2 hydroxy groups form strong interactions (probably via hydrogen bonds) with CgCBP. As with some reducing-end-specific carbohydrate-acting enzymes, recognition of the anomeric hydroxy group of CgCBP seems to be necessary to increase recognition position for fixing the reducing end unit [40]. In contrast, the ΔΔG value of 6-deoxyglucose (1.3 kcal/mol) indicates that the recognition of the O-6 hydroxy group was relatively weak. In addition, removal of the hydroxymethyl group at C-6 decreased k_cat/K_m 125-fold, with a loss of binding energy of 3.0 kcal/mol. Therefore we estimate that C-6 stabilizes substrate binding by 1.7 (3.0–;1.3) kcal/mol, probably via a hydrophobic interaction. Similar kinetic results have been reported for CuCBP (CBP from Cellulomonas uda) [32]. By determining the kinetic parameters for the binding of the 3-deoxy and deoxyfluoro analogues of glucose, the interaction between glucose and CuCBP was predicted to occur through the formation of hydrogen bonds between each hydroxy group and a negatively charged (at O-1, O-3, and O-4) or non-charged (at O-2 and O-6) group, as well as by a hydrophobic interaction at C-6.

Table 2. Kinetic parameters for the synthesis reactions of CgCBP with different glucosyl acceptors, and interactions in the complex structure.

The ΔΔG is the Gibbs energy of activation for k_cat/K_m for the enzymatic glucosyl transfer to an acceptor, relative to transfer to D-glucose, was calculated according to ΔΔG=RTln[(k_cat/K_m)_D-glucose/(k_cat/K_m)_acceptor], whereby R is 1.987 cal·mol⁻¹·K⁻¹ and T is 310.15 K [39]. Values in parentheses are taken from the results of CuCBP (T is 303.15 K) [32]. Values for 1,5-anhydroglucitol, isomaltose, gentiobiose and melibiose were determined in the present study.

		Km (mM)	kcat (s−1)	kcat/Km (s−1·M−1)	ΔΔG (kcal/mol)		Hydrogen-bond pair	Distance (Å)
	Glucose [11]	2.1	98	46800
O-1	1,5-Anhydroglucitol	89	5.3	59	4.11	(>7)	Glu649 Oϵ	2.6
O-2	Mannose [11]	115	39	342	3.03	(2.65)	Tyr653 Oη	2.8
	2-Deoxyglucose [11]	168	37	220	3.30	(2.24)	Glu659 Oϵ1	3.0
	Glucosamine [11]	13	15	1160	2.28	(2.19)
O-3	(3-Deoxyglucose)					(4.01)	Lys658 Nζ	3.0
							Glu659 Oϵ2	2.7
O-4							Asp490 Oδ	2.6
O-5	5a-Carba-β-D-glucopyranose [12]	55	93	1680	2.05		Gln165
							(other monomer) Nϵ	3.2
O-6	6-Deoxyglucose [11]	24	139	5780	1.29	(0.77)	None
	Xylose [11]	84	31	372	2.98	(2.15)
	Isomaltose	71	1.9	27
	Gentiobiose	83	7.2	87
	Melibiose	93	2.3	24

These estimates of the strength of interaction were consistent with the structural features of the enzyme. For example, the O-1 and O-2 hydroxy groups of glucose form one strong and two relatively weak hydrogen bonds with CgCBP, respectively, consistent with the estimated ΔΔG values (Table 2). The hydrogen bond between O-2 and Tyr⁶⁵³ appears to be weaker than the interactions at O-1 and O-3, probably because Tyr⁶⁵³ possesses a non-charged side chain. The O-3 hydroxy group of glucose in CgCBP forms a direct hydrogen bond with Glu⁶⁵⁹, and the short distance (2.7 Å) indicates that CgCBP strictly recognizes the C-3 equatorial hydroxy group of glucose. Lys⁶⁵⁸ also contributes to the interaction at O-3. The O-4 hydroxy group interacts with Asp⁴⁹⁰, which acts as a general acid in phosphorolysis. In the synthesis reaction, Asp⁴⁹⁰ would assist deprotonation of the O-4 hydroxy group of the sugar-acceptor molecule, as predicted for CuCBP. The O-6 hydroxy group does not form any direct hydrogen bonds with CgCBP, but there is a hydrophobic interaction with the guanidium plane of Arg³⁶² (Figure 2C). Consistently, the ΔΔG value for O-6 is smaller than that of any other hydroxy groups, and replacement with hydrogen increases K_m. The O-6 atom in CgCBP is exposed to the solvent, and there is a large open space ahead of it. This structural feature accounts for the ability of CgCBP to synthesize branched trisaccharides by accepting O-6 substituted disaccharides, such as isomaltose (Glcα1–6Glc), gentiobiose (Glcβ1–6Glc), and melibiose (Galα1–6Glc), at this site (Table 2) [17].

Active-site-pocket formation

The active sites of VpChBP and CgCBP are located at the dimer interface between the α-helix barrel domain and the N-terminal domain of the adjacent subunit (Figure 1A). The dimer interaction forms a pocket-type active site, which is appropriate for short disaccharide substrates such as chitobiose and cellobiose [8]. Therefore, the substrate specificity of GH-94 enzymes with regard to the degree of polymerization, e.g. the difference between CBP and CDP, seems to be determined by subunit interaction. Figures 4(A) and 4(B) show the molecular surfaces of VpChBP and CgCBP respectively. Although both enzymes phosphorolyse disaccharide substrates, the shape of the entrance to their active sites is different. In VpChBP, the active site pocket is relatively ‘open’, and the substrates are able to gain access without any structural change of the enzyme. In contrast, the active-site entrance of CgCBP is ‘closed’, in that it is covered by a long loop consisting of residues 495–513 (loop coloured red in Figure 2C). This loop, which is found only in CBPs, but not in ChBPs or CDPs, in GH-94, exhibits a relatively high B-factor (30 Å² in average), indicating that it is flexible. Therefore, this loop may undergo some conformational change on substrate binding. This structural feature may provide a solution to the problem of discrepancy in the proposed substrate binding order of four very similar CBPs from Cellvibrio gilvus, Clostridium thermocellum, Cellulomonas uda, and Thermotoga maritima (sequence identity >61% for each pair) [31,32,41,42]. All of these enzymes are likely to exhibit a random-ordered Bi Bi mechanism, since the movement of the loop covering the active site may have affected inhibition patterns.

Figure 4. Active-site-pocket formation in VpChBP (a) and CgCBP (b).

The molecular surface of one subunit and a ribbon diagram of the adjacent subunit are shown. (a) The bound GlcNAc molecules in VpChBP are shown as a wireframe model. (b) The molecular surface is rendered transparent to show the bound glucose beneath loop 495–513 covering the active-site-pocket.