The role of the oligosaccharide binding cleft of rice BGlu1 in hydrolysis of cellooligosaccharides and in their synthesis by rice BGlu1 glycosynthase


  • Salila Pengthaisong,

    1. Schools of Biochemistry and Chemistry, Institute of Science, Suranaree University of Technology, Nakhon Ratchasima 30000, Thailand
    Search for more papers by this author
  • Stephen G. Withers,

    1. Centre for High-throughput Biology (CHiBi) and Department of Chemistry, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z1
    Search for more papers by this author
  • Buabarn Kuaprasert,

    1. Research Facility Division, Synchrotron Light Research Institute (Public Organization), Nakhon Ratchasima 30000, Thailand
    Search for more papers by this author
  • Jisnuson Svasti,

    1. Department of Biochemistry and Center of Excellence in Protein Structure and Function, Faculty of Science, Mahidol University, Bangkok 10400, Thailand
    Search for more papers by this author
  • James R. Ketudat Cairns

    Corresponding author
    1. Schools of Biochemistry and Chemistry, Institute of Science, Suranaree University of Technology, Nakhon Ratchasima 30000, Thailand
    • Schools of Biochemistry and Chemistry, Institute of Science, Suranaree University of Technology, 111 University Avenue, Muang District, Nakhon Ratchasima 30000, Thailand
    Search for more papers by this author


Rice BGlu1 β-glucosidase nucleophile mutant E386G is a glycosynthase that can synthesize p-nitrophenyl (pNP)-cellooligosaccharides of up to 11 residues. The X-ray crystal structures of the E386G glycosynthase with and without α-glucosyl fluoride were solved and the α-glucosyl fluoride complex was found to contain an ordered water molecule near the position of the nucleophile of the BGlu1 native structure, which is likely to stabilize the departing fluoride. The structures of E386G glycosynthase in complexes with cellotetraose and cellopentaose confirmed that the side chains of N245, S334, and Y341 interact with glucosyl residues in cellooligosaccharide binding subsites +2, +3, and +4. Mutants in which these residues were replaced in BGlu1 β-glucosidase hydrolyzed cellotetraose and cellopentaose with kcat/Km values similar to those of the wild type enzyme. However, the Y341A, Y341L, and N245V mutants of the E386G glycosynthase synthesize shorter pNP-cellooligosaccharides than do the E386G glycosynthase and its S334A mutant, suggesting that Y341 and N245 play important roles in the synthesis of long oligosaccharides. X-ray structural studies revealed that cellotetraose binds to the Y341A mutant of the glycosynthase in a very different, alternative mode not seen in complexes with the E386G glycosynthase, possibly explaining the similar hydrolysis, but poorer synthesis of longer oligosaccharides by Y341 mutants.


Glycosynthases are mutants of glycosidases in which the catalytic nucleophile of a retaining glycosidase has been replaced. They can be used for the synthesis of specific oligosaccharides in quite attractive yields. The catalytic nucleophile of these mutants is typically replaced by a nonnucleophilic residue such that the mutant cannot form a reactive glycosyl–enzyme intermediate for hydrolysis or transglycosylation.1 However, when a glycosyl fluoride of the opposite anomeric configuration to that of the parent substrate is present as a glycosyl donor, the mutant enzymes are able to transfer the glycosyl moiety to acceptor alcohols without hydrolysis of the products2 (Fig. 1). Moreover, a glycosynthase was recently derived from an inverting glycosidase, which ordinarily hydrolyzes the glycosidic bond by a single displacement mechanism.3 While high yields and selectivities are key features of the glycosynthase reaction, mutagenesis can also play an important role in the improvement and alteration of the catalytic activities of glycosynthases.4–9

Figure 1.

Mechanism of rice BGlu1 E386G glycosynthase.

Exoglycosynthases, glycosynthases derived from exoglycosidases, have moderate substrate specificity and regioselectivity and can synthesize short chain oligosaccharides (di-, tri-, and tetra-oligosaccharides) that have various glycosidic linkages.10–12 Endoglycosynthases, which are derived from endoglycosidases, generally have high regioselectivity and can catalyze the synthesis of specific glycosidic linkages. They can synthesize longer oligosaccharides than exoglycosynthases because they have long glycone-binding sites that accommodate longer glycosyl donor substrates.13–16

A glycoside hydrolase family 1 β-D-glucosidase from rice (BGlu1, systematically called Os3BGlu7) has been expressed in Escherichia coli and characterized.17 Rice BGlu1 hydrolyzes β-1,3- and β-1,4-linked oligosaccharides and pyridoxine 5′-O-β-D-glucoside, and also has high transglucosylation activity to produce these same compounds as products.18 Moreover, kinetic subsite analysis of cellooligosaccharide hydrolysis indicated that rice BGlu1 has subsites for binding of at least six β-1,4-linked D-glucosyl residues. Glycine, alanine, and serine mutants of rice BGlu1 nucleophile, E386, have been generated and the E386G mutant, acting as glycosynthase, catalyzed the most rapid accumulation of transglycosylation products.19 The unique property of this enzyme compared to previous glycosynthases derived from exoglycosidases is that it could synthesize much longer chains (of at least 11 β-1,4-linked glucosyl residues) than do other exoglycosidase-derived glycosynthases. This was hypothesized to be a consequence of an extended active site in rice BGlu1, as represented by the outer subsites in Figure 1. The X-ray crystal structures of wild type BGlu1 and of its E176Q acid–base mutant in complexes with oligosaccharides revealed that the glucosyl residues were primarily bound by aromatic-sugar stacking interactions and water-mediated hydrogen bonds with several residues along a long active site cleft, except for the nonreducing terminal glucosyl residue that was distorted through strong hydrogen bonding interactions with the surrounding amino acids.20, 21

To assess the role of the long oligosaccharide-binding cleft of rice BGlu1 E386G glycosynthase in synthesis of long oligosaccharides, its structures alone and in complexes with the donor substrate α-glucosyl fluoride or the products cellotetraose and cellopentaose were determined by X-ray crystallography. The importance of the interactions in the long oligosaccharide binding cleft for hydrolysis and glycosynthase production of long cellooligosaccharides was investigated by site-directed mutagenesis of residues in this cleft in the wild type BGlu1 and BGlu1 E386G glycosynthase. Enzymatic and structural analysis of these mutants showed flexibility in the binding of cellooligosaccharides for hydrolysis, while synthesis of long cellooligosaccharides by the glycosynthase appeared more sensitive to loss of an outer glucosyl-residue binding amino acid.

Results and Discussion

3D structures of BGlu1 E386G glycosynthase complexed with substrates

To assess the structural basis for the efficient glycosynthase activity of BGlu1 E386G, its structures alone and in complex with α-glucosyl fluoride (α-GlcF) were determined by X-ray crystallography. Crystals of the apo BGlu1 E386G glycosynthase and of its complexes were isomorphous with wild type BGlu1 crystals.20 Diffraction data statistics and model parameters for the structures are summarized in Tables I and II.

Table I. Data Collection Statistics
  • Numbers in parentheses are outer shell parameters.

  • a

    Rsym= ΣhklΣi|Ii(hkl) − 〈I(hkl)〉|/Σhkl ΣiI(hkl).

Space groupP212121P212121P212121P212121P212121P212121
Unit cell parameters (Å, °)a = 79.4a = 79.9a = 79.3a = 79.5a = 79.4a = 79.3
b = 100.8b = 100.8b = 101.0b = 100.8b = 101.5b = 101.3
c = 127.4c = 127.1c = 127.2c = 127.2c = 127.3c = 127.4
α = β = γ = 90°α = β = γ = 90°α = β = γ = 90°α = β = γ = 90°α = β = γ = 90°α = β = γ = 90°
No. of molecules per ASU222222
Resolution range (Å)30–2.20 (2.28–2.20)30–1.95 (2.02–1.95)30–1.60 (1.66–1.60)30–1.58 (1.64–1.58)30–2.10 (2.18–2.10)30–1.90 (1.97–1.90)
No. of unique observations51,49873,721134,069139,83060,07880,727
Redundancy5.6 (5.4)6.8 (4.8)4.7 (4.6)6.0 (6.0)7.3 (7.2)7.3 (7.3)
Completeness (%)97.5 (92.6)97.8 (92.9)99.1 (97.3)100.0 (100.0)99.9 (100.0)99.3 (98.9)
I/σ(I)13.9 (4.4)21.4 (3.9)19.0 (3.5)23.7 (3.9)13.0 (4.3)16.7 (4.6)
Rsym (%)a11.4 (46.3)7.9 (39.5)6.8 (45.8)7.1 (45.5)15.0 (46.9)10.8 (44.3)
Table II.  Refinement Statistics
 E386GE386G_ α-GlcFE386G_C4E386G_C5S334A_C4Y341A_C4
  • a

    Rfactor = (Σ|Fo| − |Fc|/Σ|Fo|).

  • b

    Based on 5% of the unique observations not included in the refinement.

  • c

    Ramachandran values were determined from PROCHECK.22

Resolution range (Å)30–2.2030–1.9530–1.6030–1.5830–2.1030-1.90
No. of amino-acid residues944944944944944944
No. of protein atoms760876087608760876067594
No. of water molecules651833912922781902
Refined carbohydrateNoneα-GlcFCellotetraoseCellopentaoseCellotetraoseCellotetraose
No. of carbohydrate atomsNone24135112135135
No. of other hetero atoms354735353535
Rfactor (%)a19.817.219.117.417.417.0
Rfree (%)b24.020.720.819.220.619.8
Ramachandran statisticsc      
 Most favored regions (%)88.789.889.189.490.089.3
 Additionally allowed regions (%)11.310.210.910.610.010.7
 Outlier regions (%)
R.m.s.d. from ideality      
 Bond distances (Å)0.0110.0130.0080.0100.0100.010
 Bond angle (°)1.351.391.361.381.331.32
Mean B factors (Å2)      
 All protein atoms16.016.613.511.310.611.1
 Hetero atoms35.533.024.319.927.526.1
 Carbohydrate atomsNone14.042.937.636.130.3
 Subsite −1 (A/B)None14.0/13.930.5/31.221.8/21.928.5/33.018.6/19.5
 Subsite +1 (A/B)NoneNone39.9/39.633.0/34.835.4/37.621.8/23.9
 Subsite +2 (A/B)NoneNone42.5/42.138.8/41.140.1/40.727.0/27.4
 Subsite +3 (A/B)NoneNone45.7/46.142.5/44.043.5/44.340.3/40.9
 Subsite +4 (A/B)NoneNoneNone46.9/48.2NoneNone

Structures of the E386G glycosynthase, both cocrystallized and soaked with 10 mMpNP-β-cellobioside (pNPC2), an acceptor substrate, as well as with both 10 mM α-GlcF and 10 mMpNPC2 were determined, but pNPC2 was not observed in the active site. Nonetheless, α-GlcF was observed in the active site of ternary complex crystals and, if the crystals were soaked for longer periods, the crystal mosaicities generally increased and the crystals broke down after soaking for ∼3 h, suggesting they may have produced long oligosaccharides that pushed the crystal matrix apart.

The α-GlcF complex structure is the first description of an α-GlcF donor substrate complex with a glycosynthase. The α-GlcF binds to the glycosynthase in a relaxed 4C1 chair conformation stacked onto the indole ring of W433 at the -1 subsite in the same position as the 2-deoxy-2-fluoro-α-D-glucosyl moiety (G2F) in its covalent complex with wild type BGlu1 (PDB: 2RGM)20 and made all the hydrogen bonds previously observed for that complex. The active site residues R86, N175, and N313, as well as the fluorine and O2 of the α-GlcF substrate interact with a single water molecule [Fig. 2(A,B)] that is also found in the structure of the covalent intermediate of BGlu1 with G2F, at the same position as the nucleophile carboxylate oxygen of apo wild type BGlu1 [Fig. 2(C), PDB: 2RGL].20

Figure 2.

Comparison of active sites of rice BGlu1 E386G glycosynthase and free BGlu1 and its 2-fluoroglucosyl enzyme intermediate. (A) The electron densities of α-GlcF and water in the active sites of E386G (yellow carbons). The electron density (Fo - Fc) maps were calculated with all heteroatoms omitted from the active site, and are shown in brown mesh at +3σ. An active site water molecule is seen in the wild type covalent intermediate as well as in the glycosynthase mutant complex with α-GlcF. This water is bound at the same location as the catalytic nucleophile in the apo wild type enzyme. The superimposition of rice BGlu1 E386G (yellow carbons) and apo BGlu1 (gray carbons) with the structure of the 2-fluoroglucosyl enzyme intermediate (purple carbons) is shown in wall-eyed stereo in (C). The α-GlcF is represented as balls and sticks and the water molecule is represented as a ball. Oxygen is shown in red, nitrogen in blue, and fluorine in green. Distances (in Å) are shown from the one water molecule to the surrounding polar groups and the mutated residue. No other water molecules were observed in this part of the active site.

The BGlu1 E386G glycosynthase activity was threefold higher than that of E386S and 19-fold higher than E386A glycosynthases.19 It was previously speculated that the high activity of the glycine glycosynthase of Agrobacterium sp. β-glucosidase4 may be due to binding of a water in a position similar to that of the serine glycosynthase hydroxyl group. However, there was no evidence for a water molecule in this position in the BGlu1 E386G structures. The position of the one water molecule found near the α-GlcF ligand was conserved with that of a water observed to stabilize the covalent complex of the BGlu1 with 2-deoxy-2-fluoroglucose and was similar to that of the water in the structure of Man26A β-mannanase E320G glycosynthase complexed with mannobiose.23 The BGlu1 E386G mutant is likely an efficient glycosynthase due to the flexible positioning of this water molecule and surrounding protein atoms, as well as other dynamic issues that cannot be fully understood from the current static X-ray crystallographic structural model of the initial Michaelis complex structure.24

Removal of either the catalytic base that activates the nucleophilic water or the hydroxyl that coordinates binding of the same water molecule to disrupt hydrolysis but allow transglycosylation, can produce a glycosynthase from an inverting glycosidase.3, 6 The prominent position of the water stabilizing the α-GlcF suggests that modulating binding of a critical water molecule may be important for production of an efficient glycosynthase from a retaining glycosidase, as well.

Structures of BGlu1 E386G glycosynthase complexed with cellooligosaccharides

The oligosaccharides cellotetraose and cellopentaose, which could serve as glycosynthase acceptors or products, were soaked into the E386G crystals and the structures were determined, with the X-ray diffraction statistics shown in Tables I and II. The observed electron densities, shown in Figure 3(A), and Supporting Information Figures S1 and S2, indicated that these oligosaccharides bound with the nonreducing residue in the -1 site as glycosynthase products or hydrolase substrates, as previously seen for the E176Q acid–base mutant,21 rather than in the position of acceptor substrates. To obtain sufficient ligand occupancies in the active site, the crystals were soaked in cryoprotectant saturated with the cellotetraose or cellopentaose, rather than the 2 mM cellotetraose and 1 mM cellopentaose used for the E176Q mutant. This resulted in the binding of an extracellotetraose on the surface of the protein between Molecule A of one asymmetric unit and Molecule B of another asymmetric unit, but this is unlikely to be of any biological significance.

Figure 3.

Cellooligosaccharide binding in the active site of BGlu1 E386G and BGlu1 E386G/Y341A. (A) The 2Fo - Fc electron density in the active site of BGlu1 E386G with cellopentaose. The 2Fo - Fc map was calculated with heteroatoms in the active sites, and are shown in green mesh at +1σ. The ligands from the final structures are shown in ball and stick representation. The carbons are colored in yellow and oxygen is shown in red. The interacting amino acids in the active sites are shown with carbons in pink. (B) Superimposition of the active sites of the BGlu1 E386G glycosynthase complex with cellopentaose (pink) and the BGlu1 E176Q acid–base mutant complex with cellopentaose (cyan). Superimposition of these structures with the cellotetraose complexes can be seen in Supporting Information Figure S3(A). (C) Diagram of the strong hydrogen bonds between the glucose residues of cellopentaose, surrounding amino acid residues and the network of water molecules (W) in the active site of BGlu1 E386G. Hydrogen bonds with measured distances of 3.2 Å or less are shown as dashed lines to indicate the hydrophilic interactions. The direct hydrogen bond between N245 and Glc2, which is not seen in the complexes of cellotetraose and cellopentaose with BGlu1 E176Q, is shown as a darker dashed line. Hydrogen bonding for the cellotetraose complex was similar, as shown in Supporting Information Figure S3(B). (D) The 2Fo - Fc electron density in the active site of BGlu1 E386G/Y341A with cellotetraose. (E) Superimposition of the active sites of BGlu1 E386G (violet carbons), BGlu1 E386G/S334A (yellow carbons) and BGlu1 E386G/Y341A (green carbons) in complexes with cellotetraose. The direct hydrogen bonds between the glucosyl moieties in the +1 to +3 subsites of the active site of BGlu1 E386G/Y341A were identified as O6 of Glc4 with Gln187 Nε and Oε and a weak interaction of O3 of Glc2 with Asn245 Nδ, which are shown as dashed lines. The ligands are represented as balls and sticks. (F) Diagram of the strong hydrogen bonds in active site of the complex of the E386G/Y341A mutant with cellotetraose, showing the flip of the glucosyl residues in subsites +1 to +3 that results in a completely different network. The direct hydrogen bonds with Q187 and shortened hydrogen bonds from N175 and Y315 are shown in darkened dashed lines with their distances in Å marked on the lines.

The comparisons of the structures of BGlu1 E386G in complex with cellotetraose and cellopentaose with those of the BGlu1 E176Q mutant in complex with cellotetraose and cellopentaose (PDB: 3F5J and 3F5K)21 in Figure 3(B) and Supporting Information Figure S3(A) reveal very similar positions of the amino acid residues, except the E176 (the acid/base, which is mutated to glutamine in the BGlu1 E176Q complex structures) and N245 side chain positions were shifted slightly. Likewise the Glc1, Glc2, Glc3, and Glc4 glucosyl residues from the nonreducing ends of cellotetraose and cellopentaose at subsites −1, +1, +2, and +3 in the active site of BGlu1 E386G adopt similar positions in the two complexes, although Glc5 in the active site of BGlu1 E386G had lower electron density than that in the cellopentaose complex with BGlu1 E176Q, and a slightly different position [Fig. 3(B)].

Hydrogen bonding interactions between the glucosyl residues of cellotetraose or cellopentaose, ordered water molecules and amino acid residues in the active site of BGlu1 E386G are shown in Figure 3(C). The hydrogen bonds and aromatic-sugar stacking interactions of the Glc3, Glc4, and Glc5 residues at subsites +2, +3, and +4 are similar to those in the BGlu1 E176Q complex with cellopentaose. The slight shifts in the positions of the oligosaccharide sugar moiety and of the N245 side chain allowed the N245 Nδ to form hydrogen bonds with Glc2 O6, at subsite +1, and Glc3 O2, at subsite +2, rather than the bonds to Glc3 O2 and O3 observed in the E176Q complexes.21 This shift reflects an inherent plasticity in oligosaccharide binding by this enzyme, as demonstrated by the hydrogen bonding of N245 to sugars in either the +1 subsite (in BGlu1 E176Q with laminaribiose) or the +2 subsite (in BGlu1 E176Q with cellopentaose and cellotetraose),21 or both (in the BGlu1 E386G structures presented here).

Other residues that make obvious interactions with the oligosaccharides include S334 and Y341, residues that interact primarily with Glc4 and Glc5 at the +3 and +4 subsites. The Y341 residue interacts via a water-mediated hydrogen bond to the glycosidic oxygen between subsite +2 and +3 and aromatic-sugar stacking interactions at subsites +3 and +4, while the side chain and α-carbonyl of S334 form water-mediated hydrogen bonds at subsite +3.

Oligosaccharide-binding residue mutants

Based on the observed interactions between S334 and Y341 and the oligosaccharides, the BGlu1 mutants S334A, Y341A, and Y341L were generated to investigate how they would affect hydrolysis and synthesis of cellooligosaccharides. The activity of the N245V mutant, which was previously shown to have higher Km values for pNP-β-D-glucopyranoside (pNPGlc), cellobiose, and cellotriose than does wild type BGlu1,21 was also investigated with longer oligosaccharides. Only slightly higher values of Km and slightly lower catalytic efficiency values (kcat/Km) for cellotriose, cellotetraose, and cellopentaose were seen for the BGlu1 S334A, Y341A, and Y341L mutants compared to wild type BGlu1 (Table III). The BGlu1 N245V mutant had a 10-fold lower kcat/Km value than wild type for cellotriose, but only threefold lower for cellotetraose and cellopentaose. Kinetic subsite analysis by the method of Hiromi et al.25 (with the assumption of an intrinsic kcat) verified that the S334A, Y341A, and Y341L mutations had little effect on the apparent affinity at the +3 and +4 subsites. Even the very low efficiency of the N245V mutant in the hydrolysis of cellotriose appeared to be compensated by binding subsequent glucosyl residues in cellotetraose and cellopentaose, giving the mutant a high apparent subsite +3 affinity.

Table III.  Kinetic Parameters of BGlu1 Wild Type and Mutants for Hydrolysis of Cellotriose, Cellotetraose, and Cellopentaose
SubstrateKinetic parametersBGlu1S334AY341AY341LN245V
  1. Subsite affinities were calculated by the method of Hiromi et al.25Ai = −ΔG = RTln[(kcat/Km)n/(kcat/Km)n − 1] as a means of estimating the effect of each successive sugar residue.

CellotrioseKm (mM)0.50 ± 0.050.62 ± 0.050.97 ± 0.070.97 ± 0.0511.2 ± 0.8
kcat (s−1)15.6 ± 0.620.3 ± 0.726.5 ± 0.723.3 ± 0.536.1 ± 0.9
kcat/Km (M−1 s−1)31,00033,00027,00024,0003,230
CellotetraoseKm (mM)0.37 ± 0.040.52 ± 0.040.70 ± 0.050.54 ± 0.035.3 ± 0.2
kcat (s−1)23.7 ± 1.129.1 ± 1.235.5 ± 1.026.5 ± 0.4103.5 ± 1.6
kcat/Km (M−1 s−1)64,00056,00058,00049,00020,000
Subsite affinity (subsite +3, kJ mol−1)
CellopentaoseKm (mM)0.28 ± 0.020.36 ± 0.030.44 ± 0.020.36 ± 0.023.83 ± 0.18
kcat (s−1)27.5 ± 1.231.8 ± 1.532.3 ± 0.625.6 ± 0.4139 ± 3
kcat/Km (M−1 s−1)98,00088,00073,00071,00036,300
Subsite affinity (subsite +4, kJ mol−1)


Previously, we hypothesized that the long cellooligosaccharide-binding cleft of rice BGlu1 allowed its glycosynthase to synthesize longer oligosaccharides than do other exoglycosidase-based glycosynthases.19 Because the mutations generated were in the oligosaccharide binding site and had small effects on cellooligosaccharide hydrolysis, the same mutations of E386G glycosynthase were tested to see whether they affected the production of long oligosaccharides in transglycosylation reactions with α-GlcF donor and pNPC2 acceptor. When the E386G glycosynthase and its four mutants, S334A, Y341A, Y341L, and N245V, were compared with α-GlcF donor and pNPC2 acceptor at molar ratios of 1:2, 1:1, 2:1, and 5:1 transparent product solutions resulted for E386G/Y341A, E386G/Y341L, and E386G/N245V, but the reactions of 2:1 donor:acceptor with E386G and 5:1 donor to acceptor with E386G and E386G/S334A resulted in precipitated products (data not shown). These insoluble products were previously shown to comprise cellooligosaccharides of up to 11 glucosyl residues for the E386G glycosynthase.19

The soluble products were monitored by thin-layer chromatography (TLC, Fig. 4 and Supporting Information Fig. S4) and HPLC (Supporting Information Table SI and Supporting Information Fig. S5). The S334A mutant has similar glycosynthase activity to E386G and synthesized long pNP-oligosaccharide products up to pNP-β-cellonanaoside (pNPC9), as judged by LC-MS analysis (Supporting Information Fig. S5). In TLC, the longer products appeared as a dark spot at the origin. The N245V mutant synthesizes small but significant amounts of pNP-β-cellopentaoside (pNPC5) for all reactions and longer pNP-cellooligosaccharides for the reaction at the 5:1 donor:acceptor ratio, based on TLC (Supporting Information Fig. S4), but pNP-celloheptaoside (pNPC7) was the longest product detected in the soluble products of the 1:1 ratio reaction (Supporting Information Table SI). However, the Y341A and Y341L mutants synthesized only very small amounts of longer products (pNPC5) unless the donor to acceptor ratio was increased to 5:1, although a tiny peak of pNPC6 could be detected by LC-MS of the 1:1 ratio reactions (Supporting Information Table SI). Small and similar amounts of the hydrolysis product pNPGlc were also seen in the HPLC of E386G glycosynthase and its mutants, due to ∼0.1% BGlu1 E386 β-glucosidase contamination in the expression system (Pengthaisong et al., unpublished). Thus, the BGlu1 E386G/Y341A, E386G/Y341L, and E386G/N245V glycosynthases synthesize shorter pNP-cellooligosaccharides than does the unmodified E386G glycosynthase and the E386G/S334A mutant glycosynthase under the same conditions.

Figure 4.

TLC detection of products of transglucosylation reactions between α-GlcF donor and pNPC2 acceptor at a molar ratio of 1:1. The reactions were incubated with E386G and E386G mutants in NH4HCO3 (pH 7.0) at 30°C for 16 h. The soluble products were separated by TLC, using EtOAc-MeOH-water (7:2.5:1) as a developing solvent and detected by exposure to 10% sulfuric acid in ethanol followed by charring to detect carbohydrates. Lane 1, glucose; Lane 2, α-GlcF; Lane 3, pNPC2; Lane 4, control reaction without enzyme; Lanes 5–9, reactions with glycosynthase enzymes: Lane 5, E386G; Lane 6, E386G/S334A; Lane 7, E386G/Y341A; Lane 8, E386G/Y341L; and Lane 9, E386G/N245V. The TLC detection of the products of the reactions between α-GlcF donor and pNPC2 acceptor at molar ratios of 1:2, 2:1, and 5:1 are shown in Supporting Information Figure S4. Based on the LC–MS analysis (Supporting Information Table SI and Supporting Information Fig. S5) and Hommalai et al.,19 the products were identified as pNPC3, pNPC4, pNPC5 and longer pNP-cellooligosaccharides. [Color figure can be viewed in the online issue, which is available at]

Another hydrolysis product, glucose, which is produced from both α-GlcF and pNP-oligosaccharides, was more evident in the aglycone binding cleft mutants (Fig. 4). Because only a small amount of pNPGlc was produced from pNPC2 hydrolysis in each case (<1% of the total pNP-oligosaccharides, Supporting Information Table SI), and the binding site mutants of the hydrolase are slightly less efficient at release of Glc from pNP-oligosaccharides, this is unlikely to result from hydrolysis of pNP oligosaccharides. Therefore, the extra glucose in the mutants is likely a result of the less efficient acceptor binding, resulting in an increased partitioning of the reaction to hydrolysis of α-GlcF relative to transglycosylation.

Structures of oligosaccharide-binding site mutants

The relatively small effects of mutation of Y341 on the hydrolysis of oligosaccharides was initially surprising, particularly for those with glucose residues that would stack on its aromatic ring at subsites +3 and +4. To explain the effects of these mutations, the X-ray crystal structures of E386G glycosynthase mutants complexed with cellotetraose and cellopentaose were investigated by X-ray crystallography. Cellotetraose was observed in the active site of the BGlu1 E386G/S334A and E386G/Y341A mutants. The glucosyl residues of cellotetraose and amino acid residues in the active site of the structure of the BGlu1 E386G/S334A glycosynthase mutant were similar to those in the cellotetraose complex of BGlu1 E386G [Fig. 3(E)], while the Glc2, Glc3, and Glc4 residues at subsites +1, +2, and +3 in the structure of the Y341A glycosynthase mutant were flipped nearly 180° compared to the structure of BGlu1 E386G with cellotetraose [Fig. 3(D–F)]. The Glc3 and Glc4 residues in the active site of the BGlu1 E386G/Y341A mutant moved from their positions in the active site of BGlu1 E386G, so that O6 of Glc4 directly hydrogen bonds to Q187 Nε at the new subsite +3, while N245 may now form only a weak interaction with Glc2 O3 (3.3 Å), rather than the hydrogen bonds with Glc2 and Glc3 observed in the E386G complex described earlier.

The alternative binding mode observed for cellotetraose in the structures of the Y341A glycosynthase mutant appears to allow the hydrolytic reaction to occur, since the -1 subsite is filled correctly. However, synthesis of oligosaccharides by glycosynthase requires the oligosaccharide acceptor to bind with the nonreducing residue in the weakly binding +1 subsite, where the shift in position may be more destabilizing. The fact that only the initial binding mode is observed in BGlu1 E176Q and E386G crystals without additional mutations suggests that it is the most stable binding mode. However, the small effects of the Y341 mutants on hydrolysis suggest that the energy of binding in the second mode observed in the BGlu1 E386G/Y341A mutant complex with cellotetraose may be only slightly less favorable than that in the previously observed mode. This suggests that the wild type BGlu1 can bind cellooligosaccharides in more than one position, which may allow efficient funneling of these substrates into the active site.


In conclusion, the crystal structure of the BGlu1 E386G glycosynthase with α-GlcF showed a stabilizing water molecule at the position of the nucleophile of apo BGlu1 β-glucosidase, the flexible positioning of which is likely the key to high activity in the E386G glycosynthase. Plasticity of the active site also appears to be important for binding oligosaccharides, since N245 can make direct hydrogen bonds in either the +1 or +2 subsite, and cellotetraose glucosyl residues may bind with their faces in different orientations. The greater effect of loss of Y341 on glycosynthase activity suggests that the binding of acceptors is not as flexible, so the effects of mutation of aglycone-binding residues on hydrolysis and transfer reactions may not be strictly comparable.


α-GlcF, α-glucosyl fluoride; DP, degree of polymerization; ESI-MS, electrospray ionization-mass spectrometry; G2F, 2-deoxy-2-fluoro-α-D-glucosyl moiety; Glc, glucose; IMAC, immobilized metal affinity chromatography; LC-MS, liquid chromatography-mass spectrometry; pNP, para-nitrophenyl; pNPC2, pNP-β-cellobioside; pNPC3, pNP-β-cellotrioside; pNPC4, pNP-β-cellotetraoside; pNPC5, pNP-β-cellopentaoside; pNPC6, pNP-β-cellohexaoside; pNPC7, pNP-β-celloheptaoside; pNPC8, pNP-β-cellooctaoside; pNPC9, pNP-β-cellonanaoside; pNPGlc, pNP-β-D-glucopyranoside.

Materials and Methods

Mutation of rice BGlu1 and BGlu1 E386G

The mutations for BGlu1 N245V and the BGlu1 E386G (previously designated E414G, based on its position in the BGlu1 precursor protein) have been described previously.17, 19 The mutations S334A, Y341A, Y341L, E386G/S334A, E386G/Y341A, E386G/Y341L, and E386G/N245V were made in the rice bglu1 and rice bglu1 E386G cDNA17, 19 with the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The primers used were as follows: 5′-G ACA CCG ACG AGT TAC GCA GCC GAT TGG CAG-3′ and its reverse complement for S334A; 5′-C GAT TGG CAG GTT ACC GCT GTT TTT GCG AAA AAC GGC-3′ and its reverse complement for Y341A; 5′-CC GAT TGG CAG GTT ACC CTT GTT TTT GCG AAA AAC GGC-3′ and its reverse complement for Y341L; and 5′-AA GTT GGA ATA GTT CTG GAC TTC GTA TGG TAT GAA GCT CTT TCC AAC TC-3′ and its reverse complement for N245V (the mutated nucleotides are underlined).

Protein expression, purification, and crystallization

The recombinant proteins of BGlu1, its glycosynthase mutants and the mutants described above were expressed and purified as previously described for wild type BGlu1.17, 20 The purified protein concentration was determined by measuring absorbance at 280 nm. Extinction coefficients ε280 of 113,560 M−1 cm−1 for most of the BGlu1 proteins and ε280 of 112,280 M−1 cm−1 for BGlu1 Y341A and Y341L were calculated by the method of Gill and von Hippel.26

The BGlu1 E386G and its mutants were crystallized with and without ligands, including 10 mM α-GlcF, 10 mMpNPC2, both 10 mM α-GlcF and 10 mMpNPC2, 2 mM cellotetraose and 2 mM cellopentaose by hanging drop vapor diffusion with microseeding, optimized in 16–26%, polyethylene glycol monomethyl ether (PEG MME) 5000, 0.12–0.26 M (NH4)2SO4, and 2–5 mg mL−1 rice BGlu1 in 0.1M MES (pH 6.7) at 288 K, as previously described.20 Before flash cooling in liquid nitrogen, the crystals with 10 mM α-GlcF, 10 mMpNPC2, and both 10 mM α-GlcF and 10 mMpNPC2 were soaked in cryosolution (18% (v/v) glycerol in precipitant solution) containing the same concentrations of ligands for 1–5 min, while the crystals with cellotetraose and cellopentaose were soaked in cryo solution saturated with the ligands (∼75–100 mM cellotetraose or 25 mM cellopentaose).

Data collection and processing

Preliminarily, diffraction was tested in a nitrogen cryostream from a 700 series Cryostream Cooler (Oxford Cryosystems, Oxford, England), using a Cu Kα rotating anode source mounted on a MicroSTAR generator operating at 45 kV and 60 mA connected to Rayonix SX-165 CCD detector at the Synchrotron Light Research Institute (SLRI, Nakhon Ratchasima, Thailand). Datasets were collected with 1.0 Å wavelength X-rays and an ADSC Quantum 315 CCD detector on the BL13B1 beamline at the National Synchrotron Radiation Research Center (NSRRC in Hsinschu, Taiwan). The crystals were maintained at 105 K during data collection with a nitrogen cold stream (Oxford Instruments). Data were processed and scaled with the HKL-2000 package.27

Structure solution and refinement

The crystals of glycosynthases were isomorphous with wild type BGlu1 crystals,20 allowing the structures to be solved by rigid body refinement with the free BGlu1 structure (PDB: 2RGL) in REFMAC528 with the two molecules in the asymmetric unit refined as independent domains. The refinement was executed with REFMAC5 with tight noncrystallographic symmetry (NCS) restraints and model building with Coot.29 Water molecules were added with the Coot and ARP/wARP programs in the CCP4 suite. Glucosyl residues were built into the electron densities in the shapes that fit the densities best (4C1 relaxed chairs or 1S3 skew boats) and refined. The refined sugar residue coordinates were assigned their final conformation designation according to their Cremer-Pople parameters,30 calculated by the Cremer-Pople parameter calculator of Dr. Shinya Fushinobu (University of Tokyo, The occupancy of cellopentaose binding in the active site of E386G was refined to 0.7 by setting at different values and refining to find the occupancy that minimized the values of the temperature factors (B-factors) of the ligand and the free residual factor (Rfree). The final models were analyzed with PROCHECK22 and MolProbity.31 The figures of protein structures were generated in PyMol (Schrödinger LC).

Oligosaccharide hydrolysis

Wild type and mutant BGlu1 enzymes were purified by immobilized metal affinity chromatography (IMAC), enterokinase digestion and IMAC, as previously described.20 The activities of the enzymes toward cellotriose, cellotetraose, and cellopentaose were assayed in 50 mM sodium acetate buffer (pH 5.0) at 30°C, as previously described.17 The kinetic parameters Vmax and Km were calculated by nonlinear regression of Michaelis–Menten plots with the Grafit 5.0 computer program (Erithacus Software, Horley, UK) and divided by the protein concentrations to determine the apparent kcat and kcat/Km. Relative subsite affinities were determined by the method of Hiromi et al.25

Oligosaccharide synthesis by BGlu1 glycosynthase enzymes

The BGlu1 E386G glycosynthase and its four mutants, BGlu1 E386G/S334A, BGlu1 E386G/Y341A, BGlu1 E386G/Y341L, and BGlu1 E386G/N245V, were purified as described above and the buffer was exchanged with 50 mM phosphate buffer (pH 6.0). The proteins were incubated with α-GlcF donor and pNPC2 acceptor in 150 mM ammonium bicarbonate buffer (pH 7.0) at 30°C for 16 h.19 The reaction mixture was centrifuged at 13,000g for 5 min. The enzymes in the supernatant were removed by centrifugal filtration (Microcon YM-10, Millipore, USA) and the soluble products were analyzed by electrospray ionization-mass spectrometry (ESI-MS) and monitored by TLC (silica gel 60 F254, Merck, Germany) using 7:2.5:1 ethyl acetate (EtOAc)-methanol (MeOH)-water as solvent. Plates were visualized under ultraviolet (UV) light and by exposure to 10% sulfuric acid in ethanol followed by charring. Five microliters of the products from the reactions with 1:1 molar ratios of α-GlcF donor:pNPC2 acceptor were loaded onto a ZORBAX carbohydrate column (4.6 mm × 250 mm, Agilent, USA) connected to an Agilent 1100 series LC-MS. The column was eluted with a linear gradient from 90 to 50% acetonitrile in water over 30 min at a flow rate of 1 mL min−1. The eluted peaks were detected at 300 nm with a UV–visible diode array detector and the product masses determined by mass spectrometry.


The E386G glycosynthase vector was originally generated by Greanggrai Hommalai and Watchalee Chuenchor and the authors are grateful to these investigators whose ideas contributed to the project. Portions of this research were conducted at the National Synchrotron Radiation Research Center.