The crystal structure of an inverting glycoside hydrolase family 9 exo-β-D-glucosaminidase and the design of glycosynthase

Article abstract of DOI:10.1042/BJ20150966

Exo-β-D-glucosaminidase (EC 3.2.1.165) from Photobacterium profundum (PpGlcNase) is an inverting GH (glycoside hydrolase) belonging to family 9.We have determined the three-dimensional structure of PpGlcNase to describe the first structure-function relationship of an exo-type GH9 glycosidase. PpGlcNase has a narrow and straight active-site pocket, in contrast with the long glycan-binding cleft of a GH9 endoglucanase. This is because PpGlcNase has a long loop, which blocks the position corresponding to subsites -4 to -2 of the endoglucanase. The pocket shape of PpGlcNase explains its substrate preference for a β1,4-linkage at the non-reducing terminus. Asp¹³⁹, Asp¹⁴³ and Glu555 in the active site were located near the β-O1 hydroxy group of GlcN (D-glucosamine), with Asp¹³⁹ and Asp¹⁴³ holding a nucleophilic water molecule for hydrolysis. The D139A, D143A and E555A mutants significantly decreased hydrolytic activity, indicating their essential role. Of these mutants, D139A exclusively exhibited glycosynthase activity using α-GlcN-F (α-D-glucosaminyl fluoride) and GlcN as substrates, to produce (GlcN)₂. Using saturation mutagenesis at Asp139, we obtained D139E as the best glycosynthase. Compared with the wild-type, the hydrolytic activity of D139E was significantly suppressed (<0.1%), and the F^--release activity also decreased (<3%). Therefore the glycosynthase activity of D139Ewas lower than that of glycosynthases created previously from other inverting GHs. Mutation at the nucleophilic water holder is a general strategy for creating an effective glycosynthase from inverting GHs.However, for GH9, where two acidic residues seem to share the catalytic base role, mutation of Asp139 might inevitably reduce F.-release activity.

Full text of DOI:10.1042/BJ20150966

Biochem. J. (2016) 473, 463–472 doi:10.1042/BJ20150966
463
The crystal structure of an inverting glycoside hydrolase family
9 exo-β-D-glucosaminidase and the design of glycosynthase
Yuji Honda*1, Sachiko Arai*, Kentaro Suzuki†, Motomitsu Kitaoka‡ and Shinya Fushinobu†
*Ishikawa Prefectural University, 1-308 Suematsu, Nonoichi, Ishikawa 921-8836, Japan
†Department of Biotechnology, The University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan
‡National Food Research Institute, National Agriculture and Food Research Organization, 2-1-12 Kannondai, Tsukuba, Ibaraki 305-8642, Japan
D-glucosaminyl fluoride) and GlcN as substrates, to produce
(GlcN)2. Using saturation mutagenesis at Asp139, we obtained
D139E as the best glycosynthase. Compared with the wild-type,
the hydrolytic activity of D139E was significantly suppressed
(<0.1%), and the F−-release activity also decreased (<3%).
Therefore the glycosynthase activity of D139E was lower than that
of glycosynthases created previously from other inverting GHs.
Mutation at the nucleophilic water holder is a general strategy for
creating an effective glycosynthase from inverting GHs. However,
for GH9, where two acidic residues seem to share the catalytic
base role, mutation of Asp139 might inevitably reduce F−-release
activity.
Exo-β-D-glucosaminidase (EC 3.2.1.165) from Photobacterium
profundum (PpGlcNase) is an inverting GH (glycoside hydrolase)
belonging to family 9. We have determined the three-dimensional
structure of PpGlcNase to describe the first structure–function
relationship of an exo-type GH9 glycosidase. PpGlcNase has
a narrow and straight active-site pocket, in contrast with the
long glycan-binding cleft of a GH9 endoglucanase. This is
because PpGlcNase has a long loop, which blocks the position
corresponding to subsites − 4 to − 2 of the endoglucanase. The
pocket shape of PpGlcNase explains its substrate preference for
a β1,4-linkage at the non-reducing terminus. Asp139, Asp143 and
Glu555 in the active site were located near the β-O1 hydroxy
group of GlcN (D-glucosamine), with Asp139 and Asp143 holding
a nucleophilic water molecule for hydrolysis. The D139A,
D143A and E555A mutants significantly decreased hydrolytic
activity, indicating their essential role. Of these mutants, D139A
exclusively exhibited glycosynthase activity using α-GlcN-F (α-
Key words: exo-β-D-glucosaminidase, D-glucosamine, glycoside
hydrolase family 9, glycosynthase, inverting glycoside hydrolase,
Photobacterium profundum.
substrates respectively [13,14]. Applying this strategy to other
inverting GHs, we created glycosynthases from an inverting GH95
1,2-α-L-fucosidase from Bifidobacterium bifidum and an inverting
GH19 chitinase from the moss Bryum coronatum [15,16].
GlcNases (exo-β-D-glucosaminidases) (EC 3.2.1.165) hydro-
lyse the β-O-glycosidic linkage between GlcN (D-glucosamine)
residues in chitosan or chitosan oligosaccharide [(GlcN)n;
n = degree of polymerization] from the non-reducing end
(http://www.chem.qmul.ac.uk/iubmb/enzyme/). GlcNases are
categorized into the GH2, GH9 and GH35 families [12].
Retaining GlcNases isolated from various micro-organisms, e.g.
Thermococcus kodakaraensis KOD1, Amycolatopsis orientalis,
Aspergillus species, Pyrococcus horikoshii and Trichoderma
reesei PC-3-7 belong to GH2 and GH35 [17–21]. The crystal
structure of GH2 GlcNase from A. orientalis [18] has been
reported in [22]. We have previously reported that a GH9 enzyme
PBPRA0520 from Photobacterium profundum SS9 (PpGlcNase)
is an inverting GlcNase, which produces an α anomer of GlcN
[23]. We are interested in the catalytic mechanism and conversion
of glycosynthase from the inverting GlcNase, but its three-
dimensional structure is unknown.
INTRODUCTION
GHs (glycoside hydrolases) are divided into two main classes,
which either retain or invert stereochemistry, based on the
anomeric form of the new reducing end of the product following
hydrolysis [1–3]. Retaining GHs can sometimes catalyse
transglycosylation reactions because of their double displacement
mechanism, whereas inverting GHs cannot do this [2]. With
inverting GHs, a nucleophilic water molecule participates in
the reaction initially, and the reaction is apparently irreversible.
Glycosynthases are versatile tools for the synthesis of various
oligosaccharides and glycosides [4]. The original glycosynthase
reaction was achieved using an inactive nucleophilic mutant of
a retaining GH and an opposite anomer of glycosyl fluoride as
the donor substrate. This system synthesizes a new glycosidic
bond and produces an oligosaccharide, with enzymatic hydrolysis
suppressed by the active-site mutation. Various glycosynthases
derived from retaining GHs have been reported since 1998
[5–9]. We have reported previously the first glycosynthase derived
from an inverting GH enzyme, REX (reducing end xylose-
releasing oligoxylanase) (EC 3.2.1.156) [10,11], which belongs
to the GH8 family based on the CAZy (Carbohydrate-Active
enZyme) database [12]. The glycosynthase from the inverting
GH reaction was catalysed by enzymes containing mutations
around the nucleophilic water in the active site, using α-xylobiosyl
fluoride (α-X2-F) and xylose (X1) as the donor and acceptor
The GH9 family mainly consists of endoglucanases
(EC 3.2.1.4) (http://www.cazypedia.org/index.php/Glycoside_
Hydrolase_Family_9/) with the crystal structure of at least 12
members of GH9 being reported (Figure 1). In addition to
PpGlcNase, two members of GH9, Cel9D from Fibrobacter
Abbreviations: Fs-Cel9D, CelD from Fibrobacter succinogenes; GH, glycoside hydrolase; α-Glc-F, α-glucosyl fluoride; GlcN, D-glucosamine; GlcNase,
exo-β-D-glucosaminidase; α-GlcN-F, α-D-glucosaminyl fluoride; α-(GlcN)2-F, β-D-glucosaminyl-(1→4)-α-D-glucosaminyl fluoride; MPD, 2-methyl-2,4-
pentanediol; PpGlcNase, GH9 enzyme PBPRA0520 from Photobacterium profundum SS9; REX, reducing end xylose-releasing oligoxylanase; Tf-Cel9A,
Thermobifida fusca Cel9A.
1
To whom correspondence should be addressed (email honda@ishikawa-pu.ac.jp).
ꢀ
c
2016 Authors; published by Portland Press Limited

464
Y. Honda and others
of sodium methoxide. Other reagents were of analytical grade and
commercially available.
Synthesis of α-GlcN-F
1,3,4,6-Tetra-O-acetyl-2-deoxy-2-trifluoroacetamide-D-glucose
(3) was synthesized from compound 2 with acetic anhydride and
sodium acetate according to a standard procedure. Compound
3 (2 g) was dissolved in 4 ml of pyridinium poly(hydrogen
fluoride) and the solution was left for 6 days at 25◦C [27]. The
reactant was extracted with chloroform, then the extract was
stirred in crushed ice and water. Subsequently, the extract
was washed with saturated NaHCO3 and water using a separating
funnel. The organic layer, which contained 1,3,4,6-tetra-O-
acetyl-2-deoxy-2-trifluoroacetamide-D-glcosyl fluoride (4), was
dried over MgSO4 then evaporated to yield 4 as a thick liquid
(1.5 g). Compound 4 was used for N- and O-deacetylation
without further purification. Compound 4 (0.41 g) was dissolved
in dry methanol (10 ml) and the solution was bubbled with dry
NH3 for 30 min at 0◦C. After 7 days at ambient temperature, the
solution was concentrated under reduced pressure. The product,
α-GlcN-F (α-D-glucosaminyl fluoride, 1), was purified by silica
gel dry chromatography with acetonitrile and methanol (7:3, v/v)
as the eluent. The solvent was evaporated from the eluate to yield
1 (0.12 g, yield 66%). The NMR spectrum of 1 was obtained
using a JEOL ECX-400P spectrometer operating at 295 K with a
13C frequency of 100.53 MHz and 19F frequency of 376.17 MHz.
13C-NMR ([2H4]methanol) δ108.28 (d, J1,F = 220.5 Hz, C1),
Figure 1 Phylogenetic tree of GH9 enzymes
Source organisms of the proteins are as follows: Aa-Cel9A, Alicyclobacillus acidocaldarius;
Cb-Cel9A, Caldicellulosiruptor bescii (formerly Anaerocellum thermophilum); Cc-Cel9M
and Cc-Cel9G, Clostridium cellulolyticum; Ct-Cbh9A, Ct-Cel9T and Ct-Cel9I, Clostridium
thermocellum; Fs-Cel9D Fibrobacter succinogenes; Ef-EG2, Eisenia fetida; Lc-CelG, uncultured
bacteriumfrommetagenome;Nt-EG, Nasutitermestakasagoensis;PBPRA0520, Photobacterium
profundum; Rt-Cel9A, Ruminiclostridium thermocellum; Tf-Cel9A, Thermobifida fusca;
VC0615, Vibrio cholerae; VP2484, Vibrio parahaemolyticus. Crystal structures of all proteins
except for VC0615 and Fs-Cel9D were determined. VP2484 was not characterized.
75.23 (C5), 73.81 (C3), 69.48 (C4), 60.77 (C6), 55.68 (d, J2,F
25.0 Hz, C2). 19F-NMR ([2H4]methanol) δ − 149.95 (dd, JF,H2
34.7 Hz, JF,H1 = 81.0 Hz).
=
=
Construction of PpGlcNase mutants
The gene encoding wild-type PpGlcNase in a pET30 (Novagen)
vector, which incorporated a His6 sequence at the C-terminus,
was constructed as described previously [23]. The genes
encoding D139A, D143A and E555A mutants of PpGlcNase
were constructed using wild-type PpGlcNase as a template
for site-directed mutagenesis. PCR was performed using KOD
polymerase (Toyobo) with site-specific primers as listed in
Supplementary Table S1. The reactant was digested by DpnI.
Next, Escherichia coli XL1-Blue cells were transformed by
pET30 for D139A, D143A and E555A. The pET30 plasmids
were extracted and the sequences were checked. Subsequently,
E. coli BL21(DE3) cells were transformed with the plasmids. The
recombinant proteins were expressed in LB medium as described
previously [23]. Transformants of the genes encoding PpGlcNase
and its mutants were incubated in LB medium at 37◦C. After the
medium reached an attenuance of 0.6–0.8 at 600 nm, IPTG was
added, then the cultures were incubated for 24 h at 20◦C.
succinogenes (Fs-Cel9D) and VC0615 from Vibrio cholerae,
have been reported to have exo-type glucan 1,4-β-glucosidase
(EC 3.2.1.74) activity [24,25]. However, Hunt et al. [26] have
suggested re-annotating VC0615 as GlcNase because of the
inability of V. cholerae to grow on cellobiose, up-regulation of
the gene by growth on chitin and functions of the adjacent genes
as chitin-metabolic enzymes. The endoglucanases and exo-type
glycosidases in GH9 are clearly separated in a phylogenetic tree
(Figure 1), sharing a very low amino acid sequence identity
(<20%). The crystal structure of VP2484 protein from Vibrio
parahaemolyticus RIMD 2210633 is available through a structural
genomics project (New York SGX Research Center for Structural
Genomics), but it has not yet been characterized.
In the present study, we aim to determine the three-dimensional
structure of PpGlcNase and describe the structure–function
relationship of an exo-type GH9 glycosidase for the first time.
We also convert this inverting PpGlcNase into a glycosynthase
through the saturation mutagenesis of active-site residues.
Saturation mutagenesis and site-directed mutagenesis at Asp139
Saturation mutagenesis at Asp139 in PpGlcNase was performed
by the PCR method as described above. The following
primers were used: forward primer, 5ꢁ-GGTGGTTGGTAT-
NNNGCATCAGGTGACGTA-3ꢁ, and reverse primer, 5ꢁ-
GTCACCTGATGCNNNATACCAACCACCATG-3ꢁ (N = A, T,
G and C). D139C, D139G, D139H, D139I, D139K, D139P,
D139T and D139W were also constructed using site-specific
primers as listed in Supplementary Table S1. The recombinant
proteins were expressed in LB medium as described above.
MATERIALS AND METHODS
Materials
(GlcN)2 was obtained from Seikagaku. N-trifluoroacetyl-
D-glucosamine (2) was purchased from Tokyo Chemical
Industry Co. α-Glc-F (α-glucosyl fluoride) was prepared
by O-deacetylation of 2,3,4,6-tetra-O-acetyl-α-D-glucopyranosyl
fluoride (Tokyo Chemical Industry Co.) with a catalytic amount
ꢀ
c
2016 Authors; published by Portland Press Limited

Glycosynthase based on the crystal structure of an inverting exo-β-D-glucosaminidase
465
Table 1 X-ray crystallography
Purification of the recombinant proteins
The recombinant proteins were purified by Ni-NTA (Ni2 +
Values in parentheses are for the highest resolution shell.
-
nitrilotriacetate)–agarose (Qiagen) resin column chromatography
as described previously [23]. The purity of the proteins was
determined by SDS/PAGE [28].
Dataset
Ligand-free
GlcN complex
Data collection
PDB code
Space group
5DGQ
P21
5DGR
P21
Unit cell parameters
Crystallography
˚
a (A)
61.9
102.3
90.3
97.7
62.3
102.6
89.8
97.2
Crystallization trials were performed at 20◦C over 2 weeks,
using the sitting-drop vapour-diffusion method. The crystals were
obtained by mixing 1.0 μl of protein solution, which contained
8.64 mg/ml PpGlcNase protein, 20 mM Tris/HCl (pH 7.0) and
10 mM GlcN, with 1.0 μl of reservoir solution containing 10%
PEG6000, 0.1 M Hepes/NaOH (pH 7.5) and 5% MPD (2-methyl-
2,4-pentanediol). The ligand-free crystal was cryoprotected in
reservoir solution with a high concentration of MPD (17.5%, v/v).
The GlcN complex crystals were prepared by soaking the crystals
in the reservoir solution supplemented with 100 mM GlcN and
17.0% (v/v) MPD. Diffraction experiments were conducted at
beamline BL5A at the Photon Factory, High Energy Accelerator
Research Organization (KEK, Tsukuba, Japan). Crystals were
flash-cooled at 95 K in a stream of nitrogen gas. Diffraction data
were collected at 1.0 Å (1 Å = 0.1 nm). Data were processed
using the HKL2000 program [29]. Molecular replacement
was performed using MOLREP [30]. Manual model building
and molecular refinement were performed using Coot [31] and
Refmac5 [32]. Data collection and refinement statistics are given
in Table 1. Molecular graphics were prepared using PyMol
(DeLano Scientific, http://www.pymol.org).
˚
b (A)
˚
c (A)
◦
β ( )
˚
Resolution (A)
Total reflections
Unique reflections
Completeness (%)
Redundancy
50.00–1.90 (1.97–1.90)
322306
87933
99.7 (100.0)
3.7 (3.8)
15.3 (3.5)
10.0 (28.9)
50.00–1.90 (1.93–1.90)
320621
88029
99.5 (100.0)
3.6 (3.7)
19.0 (3.2)
Mean I/σ(I)
Rsym (%)
7.9 (28.6)
Refinement
Resolution (A)
˚
31.87–1.90
83390
15.7/20.6
10005
28.90–1.90
83494
16.2/20.4
9891
Number of reflections
R-factor/Rfree (%)
Number of atoms
Average B factor (A )
2
˚
Protein (chain A/B)
GlcN (chain A/B)
Na+ (chain A/B)
Water
18.0/20.8
–
12.3/19.8
27.0
23.6/24.0
19.9/17.7
26.4/26.6
30.9
RMSD from ideal values
˚
Bond lengths (A)
0.023
1.978
0.023
2.020
◦
Bond angles ( )
Ramachandran plot (%)
Favoured
Allowed
97.4
2.4
97.0
2.9
Outlier
0.2
0.1
Analysis of the reaction products
The products of enzymatic reactions towards α-GlcN-F and GlcN
at 30◦C were separated by TLC on a silica gel 60F254 plate
(5.0 cm×7.5 cm; Merck) with a solvent system of propan-1-ol and
28% ammonia solution (2:1, v/v). The products were analysed
using a HPLC system connected to a Shodex Asahipak NH2-
50 4E column (4.6 mm×250 mm; Showa Denko). Elution was
performed with acetonitrile and water (78:22, v/v) at a flow rate of
1.0 ml/min at 30◦C. The products were detected using a Corona®
CAD (charged aerosol detector) (ESA).
in 0.1 M Mes/NaOH (pH 7.0) at 37◦C. (GlcN)2 in the solution
was detected using the HPLC method described above.
RESULTS AND DISCUSSION
Crystallography of the inverting GlcNase
The crystal structure of PpGlcNase was determined by molecular
replacement using the VP2484 protein structure (62.5% sequence
identity, PDB code 3H7L) as a search model. Both the ligand-
free and GlcN complex structures were determined at 1.9 Å
resolution. The crystal contained a dimer in the asymmetric unit
(Figure 2). The molecular masses of PpGlcNase, as deduced
from the amino acid sequence, estimated by SDS/PAGE and
estimated by MALDI–TOF-MS were 65.5, 60 and 66.7 kDa
respectively [23]. A peak corresponding to the dimer (133 kDa)
was also observed by the MALDI–TOF-MS analysis, suggesting
that dimers are formed. When the crystallographic dimer interface
was analysed, approximately 10% of the entire surface area of
the dimeric molecule (3870 Å2 of 38200 Å2) was buried. There
are 17 hydrogen bonds and four salt bridges, and the estimated
ꢀiG value is − 23.4 kcal/mol (1 kcal = 4.184 kJ). The dimer
Enzyme assay
The (GlcN)2 hydrolytic reactions were performed in a 0.05 M
Mes/NaOH (pH 7.0) at 37◦C. A portion of the solution was
periodically boiled for 5 min to terminate the enzymatic reaction.
The products were quantified by HPLC as described above. The
kinetic parameters of the hydrolytic reactions were determined
by regressing the experimental data using the Michaelis–Menten
equation.
F−-release activity from α-GlcN-F and α-Glc-F by the enzymes
was determined in 0.1 M Mes/NaOH (pH 7.0) at 37◦C. The
reaction was monitored using a fluoride ion-selective electrode
(9609BN OrionTM, Thermo Fisher Scientific) connected to a
portable pH ISE Meter model 290A + (Thermo Fisher Scientific).
The rate constant of the spontaneous release of F− from α-
GlcN-F was determined to be 7.4×10− 6 s− 1 with a half-life of
9.4×104 s under the above conditions. To determine the (GlcN)2
concentrations of the glycosynthase reactions by D139A, D139E,
D139G, D139N, D139S and D139T mutants, the enzymatic
reaction containing 20 mM α-GlcN-F and GlcN was performed
is also connected by disulfide bonds between Cys117 and Cys275
.
The two polypeptides within the dimer are almost identical, with
RMSDs for Cα atoms of less than 0.13 Å. The monomer structure
of PpGlcNase consists of an N-terminal fibronectin type III
fold of unknown function and a C-terminal catalytic (α/α)6 barrel
domain. The structure of PpGlcNase is similar to the ligand-free
structure of uncharacterized VP2484 protein (RMSD of 0.71 Å
for 523 Cα atoms).
ꢀ
c
2016 Authors; published by Portland Press Limited

466
Y. Honda and others
Figure 2 Overall structure of PpGlcNase
Chain A polypeptide is shown in rainbow colours (blue and red for the N- and C-termini, respectively), and chain B is shown in grey. Bound GlcN molecules are shown as green sticks. Side chains
of Cys117 and Cys275, which form disulfide bonds between the dimer, are shown as magenta sticks. Side chains of the active site residues are also shown as sticks.
Figure 3(A) shows the active site of PpGlcNase with the
GlcN complex (green) and ligand-free (magenta) structures
superimposed. There is no significant movement of the residues
around the active site between these structures. Clear electron
density for a β-GlcN molecule (blue mesh) was observed in
the GlcN complex structure. The hydroxy groups of the GlcN
were assigned as glucan 1,4-β-glucosidases [24,25], whereas
PpGlcNase exhibited high GlcNase activity [23].
Figure 4 compares the molecular surfaces of PpGlcNase and Tf-
Cel9A. The Tf-Cel9A endoglucanase has a typical endo-type long
cleft extending from subsites − 4 to + 2 [35], whereas PpGlcNase
has a narrow and straight pocket. The pocket shape can explain
the substrate preference of PpGlcNase, which is specific for a
β1,4-linkage at the non-reducing terminus [23]. In the exo-type
PpGlcNase structure, subsites − 4, − 3 and − 2 are blocked by
a long loop connecting α10 and α11 (Glu412–Asn451) as shown
in Figure 5 (orange). Three residues involved in the recognition
of the non-reducing end (O3 and O4 hydroxy groups) of GlcN
(Tyr445, Trp446 and Gln448) are located in this loop.
molecule form hydrogen bonds with the side chains of Asp143
,
His150, Tyr445, Trp446, Gln448, Asn524 and Glu555. Trp557 provides
an aromatic stacking platform for the sugar ring of GlcN. The
2-amino group forms a hydrogen bond with Tyr147. Gln161 and
Phe231 also located near this position. There are three residues
(Asp139, Asp143 and Glu555) near the β-O1 hydroxy group. In the
ligand-free structure, a water molecule bridges Asp143 and Asp139
by hydrogen bonds. This water is located near the anomeric
C1 atom of GlcN, where it can act as a nucleophile for the
inverting hydrolysis. Glu555 forms a water-mediated hydrogen
bond with the β-O1 hydroxy group of GlcN. This may indicate
a ‘Grotthuss’ mechanism involving a water-mediated proton
donation as previously proposed for the general base residue
in inverting GH6 cellulases [33]. However, the general acid
glutamate residue in crystal structures of GH9 endoglucanases
interacts directly with the scissile glycosidic bond oxygen [34].
The proton donation mechanism of PpGlcNase, whether water-
mediated or direct from the general acid residue, needs further
investigation. On the basis of an azide rescue test, Li et al. [35]
determined the catalytic residues of Thermobifida fusca Cel9A
(Tf-Cel9A) as Asp58 (catalytic base) and Glu424 (catalytic acid).
According to amino acid sequence conservation and the structural
feature, we assigned Glu555 of PpGlcNase as the general acid
residue; Asp139 and Asp143 are likely to be the catalytic base
and nucleophilic water binding residue respectively. Almost all
of these residues at the active site are conserved in the exo-
Glycosynthase mutants
On the basis of the active-site structure of PpGlcNase, the
three catalytic residue candidates, Asp139, Asp143 and Glu555
,
were replaced individually by alanine residues. The hydrolytic
activities towards 1.0 mM (GlcN)2 by D139A, D143A and E555A
were 1.3×10− 4, 5.0×10− 4 and 2.5×10− 4 s− 1 respectively.
The hydrolytic activities of the mutants were significantly
decreased compared with that of wild-type (19.6 s− 1). Because
suppression of hydrolysis is the primary requirement for creating
a glycosynthase [13,14], we checked glycosynthase reactions by
the mutants using α-GlcN-F (a ‘wrong’ glycosyl fluoride) and
GlcN as the donor and the acceptor substrates respectively. As
shown in Figure 6(A), D139A produced (GlcN)2 and α-(GlcN)2-F
[β-D-glucosaminyl-(1→4)-α-D-glucosaminyl fluoride], whereas
D143A and E555A exhibited no glycosynthase activity after
10 days (results not shown).
type GH9 glycosidases (Figure 3B). The exception is Phe231
,
To investigate the effect of different side chains at position 139,
we constructed Asp139 mutants by saturation mutagenesis. TLC
analysis indicated the hydrolytic activity of all of Asp139 mutants
which is replaced by tyrosine in VC0615 and VP2484. From the
crystal structure, we cannot explain why VC0615 and Fs-Cel9D
ꢀ
c
2016 Authors; published by Portland Press Limited

Glycosynthase based on the crystal structure of an inverting exo-β-D-glucosaminidase
467
Figure 3 Active-site structures of PpGlcNase in GlcN complex (green) and ligand-free (magenta) structures (A), and partial amino acid sequence alignment
of exo-type GH9 glycosidases (B)
In (A), a |Fo| − |Fc| omit electron density map for GlcN contoured at 3.5σ is shown. The catalytic residues, a nucleophilic water holder, residues near the 2-amino group of GlcN and other residues
at the active site are indicated by red, magenta, blue and black characters and marks respectively.
towards (GlcN)2 decreased drastically compared with wild-type
PpGlcNase (results not shown). Figure 6(B) shows the TLC
analysis of reaction products from α-GlcN-F and GlcN by Asp139
mutants. D139A, D139E, D139G, D139N, D139S and D139T
produced (GlcN)2 and α-(GlcN)2-F after 21 h, suggesting that
these mutants exhibited glycosynthase activity. Neither (GlcN)3
nor its fluoride was detected, probably because (GlcN)2 and
α-(GlcN)2-F were not effective acceptors of the glycosynthase
reaction as the positive subsites do not seem to be suitable
for binding these disaccharides. The hydrolytic activities of
PpGlcNase for (GlcN)3, (GlcN)4 and (GlcN)5 were no larger than
that for (GlcN)2, indicating that binding of (GlcN)n at subsites
+ 2∼ + 4 did not increase the activity [23].
the best glycosynthase as it accumulated the largest amount of
(GlcN)2.
Figure 7 shows a possible reaction scheme of glycosynthase
from PpGlcNase. The release of a F− from α-GlcN-F is coupled
with the glycosidic bond formation, and the slight residual
hydrolytic activity of the glycosynthase (active site) mutant can
degrade the product. Therefore, to acquire a better glycosynthase,
it is important to minimize the decrease in F−-release activity
while maximizing the decrease in hydrolytic activity [13,14]. The
kinetic analysis of these activities of wild-type, D139A, D139E
and D139S is summarized in Table 3. Judging by the kcat values, the
hydrolytic activities of the mutants were in the following order:
D139A<D139E<D139S. In the kinetic analysis of F−-release
activity, all enzymes, except for D139E, exhibited an almost linear
S–v relationship up to 27 mM substrate concentration, so their
kcat/Km values were determined by regressing the linear plots.
D139E exhibited similar kcat values between the hydrolytic and
the F−-release activities. The kcat/Km values of F−-release activity
by D139A and D139S decreased to less than 0.4% of that of the
wild-type, but D139E displayed 3% of the wild-type reaction.
The GlcN concentration did not affect the F−-release activity
from α-GlcN-F by the enzymes (results not shown). The wild-type
Next, we analysed the reaction products from D139A, D139E,
D139G, D139N, D139S and D139T. In the HPLC analysis
of the glycosynthase reaction, we could only determine the
(GlcN)2 concentration because the peaks of α-(GlcN)2-F and
GlcN overlapped. Table 2 summarizes the (GlcN)2 concentrations
from glycosynthase reactions using α-GlcN-F and GlcN as
substrates with six Asp139 mutants. Of these, D139E, D139A
and D139S produced a significant amount of (GlcN)2 after 10 h,
but the amount decreased after 24 h for D139S. D139E was
ꢀ
c
2016 Authors; published by Portland Press Limited

468
Y. Honda and others
Table 2 (GlcN)2 concentration from α-GlcN-F and GlcN by D139A, D139E, D139G, D139N, D139S and D139T reaction after 10 and 24 h
◦
The enzymatic reactions were performed in 20 mM α-GlcN-F and GlcN with 5 μM of each enzyme at 37 C. (GlcN)2 concentrations were determined by HPLC as described in the text. Values in
parentheses indicate percentage yield of (GlcN)2 from 20 mM α-GlcN-F.
Reaction time (h)
D139A
D139E
D139G
D139N
D139S
D139T
10
24
3.0 mM (15%)
4.7 mM (24%)
5.3 mM (27%)
6.0 mM (30%)
0.3 mM (1.5%)
0.3 mM (1.5%)
0.1 mM (0.5%)
0.1 mM (0.5%)
2.4 mM (12%)
0.7 mM (3.5%)
0.2 mM (1.0%)
0.2 mM (1.0%)
Figure 4 Molecular surfaces of exo-type PpGlcNase (A) and Tf-Cel9A endoglucanase (B, PDB code 4TF4)
PpGlcNase has a function-unknown fibronectin type III (FnIII) domain at its N-terminus, whereas Tf-Cel9A has a CBM3c domain at its C-terminus. The central barrel domains of the two structures
are shown in the same orientation.
PpGlcNase exhibited only a slight F−-release activity towards α-
Glc-F (kcat/Km 0.0078 s− 1·mM− 1) with Asp139 mutants displaying
no detectable activity towards α-Glc-F.
reaction [14]. Table 4 shows the F/H ratio based on the kcat/Km
values of the F−-release and hydrolytic activities of the wild-type
and Asp139 mutants. The F/H ratio of D139A was 6-fold higher
than that of D139E. The F−-release and hydrolytic activities
of D139A were 9% and 2% of those of D139E respectively.
D139S also displayed 15% of the F−-release activity of D139E,
but its hydrolytic activity was similar. The decrease in F−-
release activity for D139S and D139A reflected the efficiency
of the glycosynthase reaction. Therefore we concluded that
A concept for creating glycosynthases from inverting GHs
The ratio of F−-release activity compared with hydrolytic activity
(F/H ratio) is a good indicator of the efficiency of a glycosynthase
ꢀ
c
2016 Authors; published by Portland Press Limited

Glycosynthase based on the crystal structure of an inverting exo-β-D-glucosaminidase
469
Table 3 Kinetic parameters of (GlcN)2 hydrolysis and F−-release activity from α-GlcN-F by wild-type PpGlcNase, D139E, D139S and D139A
(a)
(GlcN)2 hydrolysis
kcat (s− 1
)
Km (mM)
kcat/Km (s− 1 · mM− 1
)
Proportion of kcat/Km (%)
+
+
+
Wild-type
D139E
D139S
D139A
87.0 2.4
0.071 0.004
1.8 0.2
1.8 0.3
49.0 3.7
0.040 0.006
100
−
−
−
+
+
+
0.08
0.07
0.001
−
−
−
+
+
+
0.54 0.04
15.0 1.9
0.036 0.002
−
−
−
+
+
+
0.0047 0.0006
7.7 1.9
0.00062 0.00007
−
−
−
(b)
Fl−-release activity from α-GlcN-F
kcat (s− 1
)
Km (mM)
–
kcat/Km (s− 1 · mM− 1
)
Proportion of kcat/Km (%)
+
Wild-type
D139E
D139S
D139A
–
0.73 0.04
100
2.9
0.4
0.3
−
+
+
+
0.073 0.005
–
3.5 0.9
0.021 0.004
−
−
−
+
–
0.0031 0.0002
−
+
–
–
0.0019 0.0001
−
Table 4 kcat/Km values between F−-release activity and (GlcN)2 hydrolysis
by wild-type PpGlcNase and D139E, D139S and D139A
F−-release
activity (F,
Hydrolysis
(H,
Enzyme
s− 1 · mM− 1
)
s− 1 · mM− 1
)
F/H ratio
Wild-type
D139E
D139S
D139A
0.730
0.021
0.0031
0.0019
49.0
0.02
0.53
0.09
3.1
0.040
0.036
0.00062
D139E was the best glycosynthase of the Asp139 mutants created
from GlcNase. However, the F/H ratio of D139E of 0.53 was
significantly lower than that of a glycosynthase from REX. The
best glycosynthase from REX (Y198F) exhibited an F/H ratio
of 78.3, but that of the wild-type was 0.1 [14]. The difference
was due to the drastic decrease in the F−-release activity through
the mutation of PpGlcNase. Thus the glycosynthase reaction by
D139E mutants from PpGlcNase was not efficient compared with
the glycosynthase from REX.
Previously, we proposed that mutation at the nucleophilic water
holder is a general strategy for creating efficient glycosynthases
from inverting GHs because such a mutation did not cause a
decrease in the F−-release activity [36]. This strategy was suitable
for the glycosynthase mutant of GH19 chitinase [16] (Figure 8B).
On the other hand, in the case of GH9 enzymes, the situation is
clearly different from those of GH8 REX and GH19 BC-ChiA,
because there are two aspartate residues holding the nucleophilic
water (Figure 8A). A chemical rescue study on a distant GH9
member, Tf-Cel9A, has indicated that one of the two, a residue
corresponding to Asp139, acts as the catalytic base [35]. However,
as we assume that the two aspartate residues share the role
of general base in some GH9 members, we cannot confidently
assign the nucleophilic water holder for targeted mutation to
create a glycosynthase in such cases. Because the glutamate side
chain of the D139E mutant is longer by one methylene group
compared with the wild-type enzyme, the nucleophilic water is
presumably displaced. This might have caused the large decrease
in the F−-release activity observed for D139E (<3% of the wild-
type). However, compared with the other mutants (D139A and
D139S), the F−-release activity of D139E was approximately 10-
fold higher (Table 3). Maintaining the charge balance at the active
Figure 5 Loop structures of exo-type PpGlcNase (A) and Tf-Cel9A
endoglucanase (B) viewed from the same orientation as in Figure 4
Loops connecting α2–α3 (blue), α4–α5 (cyan), α6–α7 (green), α8–α9 (yellow), α10–α11
(orange) and α12–α13 (magenta) are coloured. Side chains of the active-site residues and
residues blocking the subsite − 2 of PpGlcNase (Tyr445, Trp446 and Gln448, orange) are shown
as sticks.
ꢀ
c
2016 Authors; published by Portland Press Limited

470
Y. Honda and others
Figure 6 α-GlcN-F and GlcN reaction by wild-type PpGlcNase and its mutants
(A) D139A, D143A and E555A reactions towards α-GlcN-F and GlcN. The reactions were performed in 20 mM α-GlcN-F in the absence (lanes 1 and 2) and presence of 20 mM GlcN (lanes 3 and 4)
◦
for 42 h at 30 C. Odd lanes are not enzymatic reactions, and even lanes are enzymatic reactions (D139A, 27 μM; D143A, 4 μM; E555A, 14 μM). Other conditions are described in the text. Roman
numerals indicate degree of polymerization of standard GlcN. (B) α-GlcN-F and GlcN reaction by wild-type PpGlcNase and Asp139 mutants. The reactions were performed in 20 mM α-GlcN-F and
◦
20 mM GlcN with 4 μM of each enzyme for 21 h at 30 C. Other conditions are described in the text. Roman numerals indicate the degree of polymerization of standard GlcN. Ct is no enzymatic
reaction. All Asp139 mutants, including wild-type, are described using the single-letter amino acid code to indicate the residue at the 139th position.
Figure 7 Schematic representation of glycosynthase reaction by mutants from PpGlcNase
Oligosaccharides were produced when acceptor substrate binds to + 1 subsite. α-GlcN-F also acts as an acceptor substrate.
ꢀ
c
2016 Authors; published by Portland Press Limited

Glycosynthase based on the crystal structure of an inverting exo-β-D-glucosaminidase
471
Figure 8 Schematic description of the catalytic residues of GH9 PpGlcNase (A) in comparison with those of GH8 Rex and GH19 BsChiA (B)
In the case of PpGlcNase (A), Glu555 forms a water-mediated hydrogen bond with the β-O1 hydroxy group of GlcN. The proton donation mechanism of PpGlcNase, whether water-mediated or direct
from the general acid residue, needs further investigation.
centre may stabilize the transition state of the F−-release, which
potentially has an oxocarbenium ion-like character.
3
Gebler, J., Gilkes, N.R., Claeyssens, M., Wilson, D.B., Be´guin, P., Wakarchuk, W.W.,
Kilburn, D.G., Miller, R.C., Warren, R. and Withers, S.G. (1992) Stereoselective hydrolysis
catalyzed by related β-1,4-glucanases and β-1,4-xylanases. J. Biol. Chem. 267,
12559–12561 PubMed
Williams, S.J. and Withers, S.G. (2000) Glycosyl fluorides in enzymatic reactions.
Carbohydr. Res. 327, 27–46 CrossRef PubMed
Mackenzie, L.F., Wang, Q., Warren, R.A.J. and Withers, S.G. (1998) Glycosynthases:
mutant glycosidases for oligosaccharide synthesis. J. Am. Chem. Soc. 120, 5583–5584
CrossRef
In conclusion, we have determined the three-dimensional
structure of an inverting PpGlcNase, providing the first structural
description of an exo-type glycosidase in GH9. We have also
achieved the conversion of an inverting PpGlcNase into a
glycosynthase, although the glycosynthase activity was greatly
decreased compared with REX and the GH19-inverting chitinase.
4
5
6
Malet, C. and Planas, A. (1998) From β-glucanase to β-glucansynthase: glycosyl
transfer to α-glycosyl fluorides catalyzed by a mutant endoglucanase lacking its catalytic
nucleophile. FEBS Lett. 440, 208–212 CrossRef PubMed
7
8
Hancock, S.M., Vaughan, M.D. and Withers, S.G. (2006) Engineering of glycosidases and
glycosyltransferases. Curr. Opin. Chem. Biol. 10, 509–519 CrossRef PubMed
Kitaoka, M., Honda, Y., Fushinobu, S., Hidaka, M., Katayama, T. and Yamamoto, K. (2009)
Conversion of inverting glycoside hydrolases into catalysts for synthesizing glycosides
employing a glycosynthase strategy. Trends Glycosci. Glycotechnol. 21, 23–39
CrossRef
AUTHOR CONTRIBUTION
Sachiko Arai and Yuji Honda analysed the enzymatic reactions. Kentaro Suzuki and Shinya
Fushinobu performed the X-ray crystallography. Motomitsu Kitaoka discussed the data of
the enzymatic reactions. Yuji Honda and Shinya Fushinobu designed the study and wrote
the paper.
9
Cobucci-Ponzano, B., Strazzulli, A., Rossi, M. and Moracci, M. (2011) Glycosynthases in
biocatalysis. Adv. Synth. Catal. 353, 2284–2300 CrossRef
10 Honda, Y. and Kitaoka, M. (2004) A family 8 glycoside hydrolase from Bacillus
halodurans C-125 (BH2105) is a reducing end xylose-releasing exo-oligoxylanase. J.
Biol. Chem. 279, 55097–55103 CrossRef PubMed
ACKNOWLEDGEMENTS
We are grateful to Ms Marie Torimoto for her technical assistance in the study.
11 Fushinobu, S., Hidaka, M., Honda, Y., Wakagi, T., Shoun, H. and Kitaoka, M. (2005)
Structural basis for the specificity of the reducing end xylose-releasing exo-oligoxylanase
from Bacillus halodurans C-125. J. Biol. Chem. 280, 17180–17186
CrossRef PubMed
12 Lombard, V., Golaconda Ramulu, H., Drula, E., Coutinho, P.M. and Henrissat, B. (2014)
The carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res. 42,
D490–D495 CrossRef PubMed
13 Honda, Y. and Kitaoka, M. (2006) The first glycosynthase derived from an inverting
glycoside hydrolase. J. Biol. Chem. 281, 1426–1431 CrossRef PubMed
14 Honda, Y., Fushinobu, S., Hidaka, M., Wakagi, T., Shoun, H., Taniguchi, H. and Kitaoka,
M. (2008) Alternative strategy for converting an inverting glycoside hydrolase into a
glycosynthase. Glycobiology 18, 325–330 CrossRef PubMed
15 Wada, J., Honda, Y., Nagae, M., Kato, R., Wakatsuki, S., Katayama, T., Taniguchi, H.,
Kumagai, H., Kitaoka, M. and Yamamoto, K. (2008) 1,2-α-L-Fucosynthase: a
glycosynthase derived from an inverting α-glycosidase with an unusual reaction
mechanism. FEBS Lett. 582, 3739–3743 CrossRef PubMed
FUNDING
This work was supported in part by a grant-in-aid for Young Scientists (B) from the
Ministry of Education, Culture, Sports, Science and Technology of Japan [grant number
22780095 (to Y.H.)].
REFERENCES
1
Koshland, D.E. (1953) Stereochemistry and the mechanism of enzymatic reactions. Biol.
Rev. 28, 416–436 CrossRef
2
Sinnott, M.L. (1990) Catalytic mechanism of enzymic glycosyl transfer. Chem. Rev. 90,
1171–1202 CrossRef
ꢀ
c
2016 Authors; published by Portland Press Limited

472
Y. Honda and others
16 Ohnuma, T., Fukuda, T., Dozen, S., Honda, Y., Kitaoka, M. and Fukamizo, T. (2012) A
glycosynthase derived from an inverting GH19 chitinase from the moss Bryum
coronatum. Biochem. J. 444, 437–443 CrossRef PubMed
25 Park, J.K., Wang, L.X., Patel, H.V. and Roseman, S. (2002) Molecular cloning and
characterization of a unique β-glucosidase from Vibrio cholerae. J. Biol. Chem. 277,
29555–29560 CrossRef PubMed
17 Tanaka, T., Fukui, T., Atomi, H. and Imanaka, T. (2003) Characterization of an
exo-β-D-glucosaminidase involved in a novel chitinolytic pathway from the
hyperthermophilic archaeon Thermococcus kodakaraensis KOD1. J. Bacteriol. 185,
5175–5181 CrossRef PubMed
18 Coˆte´, N., Fleury, A., Dumont-Blanchette, E., Fukamizo, T., Mitsutomi, M. and Brzezinski,
R. (2006) Two exo-β-D-glucosaminidases/exochitosanases from actinomycetes define a
new subfamily within family 2 of glycoside hydrolases. Biochem. J. 394, 675–686
CrossRef PubMed
19 Ike, M., Isami, K., Tanabe, Y., Nogawa, M., Ogasawara, W., Okada, H. and Morikawa, Y.
(2006) Cloning and heterologous expression of the exo-β-D-glucosaminidase-encoding
gene (gls93) from a filamentous fungus, Trichoderma reesei PC-3–7. Appl. Microbiol.
Biotechnol. 72, 687–695 CrossRef PubMed
20 Liu, B., Li, Z., Hong, Y., Ni, J., Sheng, D. and Shen, Y. (2006) Cloning, expression and
characterization of a thermostable exo-β-D-glucosaminidase from the hyperthermophilic
archaeon Pyrococcus horikoshii. Biotechnol. Lett. 28, 1655–1660
26 Hunt, D.E., Gevers, D., Vahora, N.M. and Polz, M.F. (2008) Conservation of the chitin
utilization pathway in the Vibrionaceae. Appl. Environ. Microbiol. 74, 44–51
CrossRef PubMed
27 Hayashi, M., Hashimoto, M. and Noyori, R. (1984) Simple synthesis of glycosyl fluorides.
Chem. Lett. 1747–1750 CrossRef
28 Laemmli, U.K. (1970) Cleavage of structural proteins during the assembly of the head of
bacteriophage T4. Nature 227, 680–685 CrossRef PubMed
29 Otwinowski, Z. and Minor, W. (1997) Processing of X-ray diffraction data collected in
oscillation mode. Methods Enzymol. 276, 307–326 CrossRef
30 Vagin, A. and Teplyakov, A. (1997) MOLREP: an automated program for molecular
replacement. J. Appl. Crystallogr. 30, 1022–1025 CrossRef
31 Emsley, P., Lohkamp, B., Scott, W.G. and Cowtan, K. (2010) Features and development of
Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 CrossRef PubMed
32 Murshudov, G.N., Vagin, A.A. and Dodson, E.J. (1997) Refinement of macromolecular
structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53,
240–255 CrossRef PubMed
CrossRef PubMed
21 Li, S., Wang, C. and Xia, W. (2009) Expression, purification, and characterization of
exo-β-D-glucosaminidase of Aspergillus sp. CJ22–326 from Escherichia coli.
Carbohydr. Res. 344, 1046–1049 CrossRef PubMed
33 Koivula, A., Ruohonen, L., Wohlfahrt, G., Reinikainen, T., Teeri, T.T., Piens, K.,
Claeyssens, M., Weber, M., Vasella, A., Becker, D. et al. (2002) The active site of
cellobiohydrolase Cel6A from Trichoderma reesei: the roles of aspartic acids D221 and
D175. J. Am. Chem. Soc. 124, 10015–10024 CrossRef PubMed
34 Schubot, F.D., Kataeva, I.A., Chang, J., Shah, A.K., Ljungdahl, L.G., Rose, J.P. and Wang,
B.C. (2004) Structural basis for the exocellulase activity of the cellobiohydrolase CbhA
from Clostridium thermocellum. Biochemistry 43, 1163–1170 CrossRef PubMed
35 Li, Y., Irwin, D.C. and Wilson, D.B. (2007) Processivity, substrate binding, and
mechanism of cellulose hydrolysis by Thermobifida fusca Cel9A. Appl. Environ.
Microbiol. 73, 3165–3172 CrossRef PubMed
22 van Bueren, A.L., Ghinet, M.G., Gregg, K., Fleury, A., Brzezinski, R. and Boraston, A.B.
(2009) The structural basis of substrate recognition in an exo-β-D-glucosaminidase
involved in chitosan hydrolysis. J. Mol. Biol. 385, 131–139 CrossRef PubMed
23 Honda, Y., Shimaya, N., Ishisaki, K., Ebihara, M. and Taniguchi, H. (2011) Elucidation of
exo-β-D-glucosaminidase activity of a family 9 glycoside hydrolase (PBPRA0520) from
Photobacterium profundum SS9. Glycobiology 21, 503–511 CrossRef PubMed
24 Qi, M., Jun, H.S. and Forsberg, C.W. (2008) Cel9D, an atypical 1,4-β-D-glucan
glucohydrolase from Fibrobacter succinogenes: characteristics, catalytic residues, and
synergistic interactions with other cellulases. J. Bacteriol. 190, 1976–1984
CrossRef PubMed
36 Hidaka, M., Fushinobu, S., Honda, Y., Wakagi, T., Shoun, H. and Kitaoka, M. (2010)
Structural explanation for the acquisition of glycosynthase activity. J. Biochem. 147,
237–244 CrossRef PubMed
Received 4 September 2015/25 November 2015; accepted 27 November 2015
Accepted Manuscript online 30 November 2015, doi:10.1042/BJ20150966
ꢀ
c
2016 Authors; published by Portland Press Limited