Structure of the thermophilic l-Arabinose isomerase from Geobacillus kaustophilus reveals metal-mediated intersubunit interactions for activity and thermostability

Article abstract of DOI:10.1016/j.abb.2016.02.033

Thermophilic l-arabinose isomerase (AI), which catalyzes the interconversion of l-arabinose and l-ribulose, can be used to produce d-tagatose, a sugar substitute, from d-galactose. Unlike mesophilic AIs, thermophilic AIs are highly dependent on divalent metal ions for their catalytic activity and thermostability at elevated temperatures. However, the molecular basis underlying the substrate preferences and metal requirements of multimeric AIs remains unclear. Here we report the first crystal structure of the apo and holo forms of thermophilic Geobacillus kaustophilus AI (GKAI) in hexamer form. The structures, including those of GKAI in complex with l-arabitol, and biochemical analyses revealed not only how the substrate-binding site of GKAI is formed through displacement of residues at the intersubunit interface when it is bound to Mn²⁺, but also revealed the water-mediated H-bonding networks that contribute to the structural integrity of GKAI during catalysis. These observations suggest metal-mediated isomerization reactions brought about by intersubunit interactions at elevated temperatures are responsible for the distinct active site features that promote the substrate specificity and thermostability of thermophilic AIs.

Full text of DOI:10.1016/j.abb.2016.02.033

Archives of Biochemistry and Biophysics 596 (2016) 51e62
Contents lists available at ScienceDirect
Archives of Biochemistry and Biophysics
journal homepage: www.elsevier.com/locate/yabbi
Structure of the thermophilic -Arabinose isomerase from Geobacillus
L
kaustophilus reveals metal-mediated intersubunit interactions for
activity and thermostability
Jin Myung Choic 1, Yong-Jik Lee b , Thinh-Phat Cao , Sun-Mi Shin , Min-Kyu Park ,
a
,
,
1
a
b
b
d
e
c
f
Han-Seung Lee , Eric di Luccio , Seong-Bo Kim , Sang-Jae Lee , Sang Jun Lee ,
a
,
**
, Dong-Woo Lee b
, *
Sung Haeng Lee
a
Department of Cellular and Molecular Medicine, Chosun University School of Medicine, Gwangju, 61452, South Korea
School of Applied Biosciences, Kyungpook National University, Daegu, 41566, South Korea
Major in Food Biotechnology, Silla University, Busan, 46958, South Korea
School of Life Sciences, Kyungpook National University, Daegu, 41566, South Korea
CJ CheilJedang, Life Ingredient & Material Research Institute, Suwon, 16495, South Korea
Microbiomics & Immunity Research Center, Korea Research Institute of Bioscience and Biotechnology, Daejeon, 34141, South Korea
b
c
d
e
f
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 9 November 2015
Received in revised form
Thermophilic
ribulose, can be used to produce
thermophilic AIs are highly dependent on divalent metal ions for their catalytic activity and thermo-
stability at elevated temperatures. However, the molecular basis underlying the substrate preferences
and metal requirements of multimeric AIs remains unclear. Here we report the first crystal structure of
the apo and holo forms of thermophilic Geobacillus kaustophilus AI (GKAI) in hexamer form. The struc-
L
-arabinose isomerase (AI), which catalyzes the interconversion of
L-arabinose and L-
D
-tagatose, a sugar substitute, from -galactose. Unlike mesophilic AIs,
D
27 February 2016
Accepted 29 February 2016
Available online 4 March 2016
tures, including those of GKAI in complex with L-arabitol, and biochemical analyses revealed not only
Keywords:
how the substrate-binding site of GKAI is formed through displacement of residues at the intersubunit
L-arabinose isomerase
interface when it is bound to Mn2 , but also revealed the water-mediated H-bonding networks that
contribute to the structural integrity of GKAI during catalysis. These observations suggest metal-
mediated isomerization reactions brought about by intersubunit interactions at elevated temperatures
are responsible for the distinct active site features that promote the substrate specificity and thermo-
stability of thermophilic AIs.
þ
Crystal structure
Intersubunit interaction
Substrate specificity
Thermostability
©
2016 Elsevier Inc. All rights reserved.
1. Introduction
past decade, several AIs have been characterized and commercially
utilized to produce -tagatose, a new antidiabetic and obesity
D
L
-Arabinose isomerase (AI) (E.C. 5.3.1.4), encoded by the araA
gene of the arabinose operon, catalyzes the isomerization of
arabinose to -ribulose in -arabinose catabolism [1e4]. During the
control nutraceutical [5] that has also been ‘generally recognized as
safe (GRAS)' as a sweetener [6e8]. Although many attempts have
L-
L
L
been made to use microbial AIs to produce
galactose derived from lactose hydrolysis [1,8,9], mesophilic AIs
yield insufficient amounts of -tagatose for commercial purposes
1,10,11]. Because the isomerization of aldose to ketose is an
endothermic reaction [7,12], the biological conversion of -galac-
tose to -tagatose using thermostable AIs was found to be advan-
D-tagatose from D-
D
[
Abbreviations: AI,
L-arabinose isomerase; CD, circular dichroism; DSC, differ-
D
ential scanning calorimetry; EDTA, ethylenediaminetetraacetic acid; FI,
isomerase; H-bond, hydrogen bond; ICP, inductively coupled plasma; MR, molec-
ular replacement.
Corresponding author.
* Corresponding author.
E-mail addresses: sunglee@chosun.ac.kr (S.H. Lee), leehicam@knu.ac.kr
D.-W. Lee).
These authors contributed equally to this work.
L-fucose
D
tageous, as shown by the feasibility of the process, the low viscosity
of substrates, the high conversion yields, and the minimal microbial
contamination [1,7]. Thus, thermostable enzymes, such as
AIs from Thermotoga neapolitana [13], Thermotoga maritima [7],
Thermoanaerobacter mathranii [14], and Thermoanaerobacter
*
*
(
1
http://dx.doi.org/10.1016/j.abb.2016.02.033
003-9861/© 2016 Elsevier Inc. All rights reserved.
0

52
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
ꢀ
saccharolyticum [15] were predicted to yield higher amounts of
tagatose than mesophilic AIs [10,16e18]. Indeed, T. neapolitana AI
D
-
over-expressed GKAI was heated at 60 C for 30 min in prior to
affinity chromatography, which facilitates further purification
procedures and increases protein purity. All the mutants in this
study were generated by using a standard site-directed mutagen-
esis technique.
(
TNAI) and T. maritima AI (TMAI) isomerized
D
-galactose to
D
2þ
-
ꢀ
tagatose at 80 C in the presence of divalent metal ions such as Co
2
þ
or Mn with yields of ꢁ60% [7,13,19]. Notably, such (hyper)ther-
mophilic AIs were found to be highly dependent on metal ions for
both activity and structural stability, which distinguished them
from their mesophilic counterparts [20].
2.2. Enzyme assay
By contrast to mesophilic AIs, which show little activity in the
AI activity was assayed by measuring the increase in L-ribulose.
isomerization of
miscuous substrate preferences of thermophilic AIs that exhibit
unusually high activity for -galactose remains unclear [19,21].
D
-galactose to
D
-tagatose, the reason for the pro-
Unless otherwise noted, the standard reaction mixture contained
50 mM MOPS (4-morpholinepropanesulfonic acid) buffer (pH 7.5 at
room temperature), 0.2 mL of enzyme preparation at a suitable
D
Mutational studies [22,23] have suggested that the metal-
dependent catalysis and thermostability of hyperthermophilic AIs
at elevated temperatures may be due to structurally region-specific
evolution, because the amino acid sequences of thermostable AIs
are highly similar (ꢁ70%) to those of mesophilic AIs. Notably, bio-
physical data and electron microscopy analysis have indicated that
dilution, 0.1 M
1.25 mL. The mixtures were incubated at appropriate temperatures
for 20 min and stopped by chilling on ice. -Ribulose was quantified
by the cysteine-sulfuric acid-carbazole method [33], and the
absorbance was measured at 560 nm. One unit of isomerase activity
L-arabinose, and distilled water to a final volume of
L
was defined as the amount of enzyme that produces 1
mmol of
Escherichia coli
L-arabinose isomerase (ECAI) is composed of hex-
product per min under the assay conditions.
americ subunits [24,25], although its crystal structure showed it
was composed of asymmetric units with a trimeric architecture
2.3. Metal content analysis and circular dichroism
[
26]. Despite the observed discrepancies between studies con-
cerning the oligomeric state, the overall scaffold of ECAI is quite
similar to the hexameric crystal structure of E. coli -fucose isom-
To verify the metal content of AIs, column-purified enzymes
were rendered metal-free and then reconstituted with pure metals
as described [21,22]. The divalent metal contents of the as-isolated
and EDTA-treated samples were determined by high-resolution
inductively coupled plasma (ICP)-mass spectrometry on a Plas-
maQuad 3 instrument at the Korea Basic Science Institute, Kyung-
pook National University. To verify the conformational integrity of
AI, circular dichroism (CD) was determined using a Jasco J-810
spectropolarimeter with a Peltier temperature-controlled cuvette
holder. The CD spectra of enzyme samples in a cuvette with a 0.1 cm
path length were recorded in the far-UV region (190e240 nm).
Scans were collected five times at 0.1 nm intervals with a 1 nm
bandwidth. Each spectrum was corrected by subtracting the spec-
trum of the solution containing the buffer used, with and without
L
erase (ECFI) [26,27], based on the observation that the crystallo-
graphic two fold axis of the ECAI structure shows a homo hexamer
with D3 symmetry. In light of this, structural information about
thermophilic AIs is especially important for understanding the
molecular basis of thermostability and substrate specificity,
compared with that of its mesophilic counterparts. Using a struc-
tural genomics approach, we recently attempted to predict the
substrate-binding site of a thermophilic AI, despite the lack of 3D-
structure information [28]. Our bioinformatics-guided analyses
suggested that residues located at the interface between the sub-
units of multimeric AI would be important for catalytic cavity.
Therefore, in this study, we expressed the araA gene from the
thermophilic bacterium Geobacillus kaustophilus [29e31] in E. coli,
purified the recombinant G. kaustophilus AI (GKAI), characterized it,
and determined for the first time the three-dimensional structure
2
þ
.
Mn
2.4. Differential scanning calorimetry (DSC) measurement
2
þ
of a thermophilic AI in the presence or absence of Mn and L-
arabitol. Together with site-directed mutagenesis analysis, our data
provide insight into the relationship between AI structure and
function, especially regarding substrate preference, metal re-
quirements, and thermostability, as well as providing clues to the
molecular evolution of sugar isomerases that led to thermal
adaptation at elevated temperatures.
Calorimetric measurements were performed using a VP-DSC
microcalorimeter (Microcal Inc. GE Healthcare, Northampton,
USA). All scans were run at pH 7.4 in 1 mM potassium phosphate
ꢀ
ꢀ
buffer, in a temperature range from 20 to 100 C at a rate of 90 C/h.
The cell volume was 0.8 mL. The apo-GKAI (160 M) incubated with
1 mM Mn (Mn -GKAI) for 2 h at 25 C. The potassium phosphate
buffer was used for baseline scans. The apparent T values of GKAI
m
2þ
2þ
ꢀ
m
2
. Materials and methods
with and without metal ions were determined. Data from Microcal
were analyzed using the software Origin 8.0.
2.1. Expression and purification of GKAI
2.5. Kinetics
Bacterial genomic DNA was isolated using genomic DNA
extraction kits (Qiagen, Germany), according to the manufacturer's
instructions. The araA gene from G. kaustophilus was amplified by
PCR using forward and reverse primers designed to include NdeI
and HindIII restriction sites for cloning into pET-15b [32]. The araA
gene of G. kaustophilus was ligated into the NdeI and HindIII sites of
pET-15b, yielding pET-15b-GKAI. The expression vector was used to
transform E. coli BL21(DE3), and the gene encoding GKAI was over-
expressed in E. coli BL21(DE3) by IPTG induction. The resulting
recombinant DNA in the expression vector contained a 6ꢂ His-tag
and a thrombin cleavage site at the N-terminus.
The kinetic parameters of wild-type and mutant AIs were
determined in the same reaction mixtures as described above,
except that AIs were assayed over 1 min to obtain the initial reac-
tion rates. The concentrations of L-arabinose ranged from 0 to
40 mM. Kinetic results were obtained by fitting the data with a
Michaelis-Menten equation using the software Origin 8.0.
2.6. Protein crystallization and X-ray crystallography
Purified GKAI was concentrated to 6e9 mg/mL in an appropriate
2
þ
N-terminal hexa-histidine tagged GKAI was purified by Ni-
affinity chromatography, followed by thrombin cleavage and size-
exclusion chromatography [20,32]. Typically, lysate containing
buffer with or without Mn and crystallized by using a hanging-
drop vapor diffusion technique. The ligand complex of GKAI was
2
þ
obtained by soaking Mn -GKAI crystals in 10 mM L-arabitol. X-ray

J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
53
experiments were performed at Pohang Light Source (PLS),
beamline 5C and 7A (Pohang, South Korea) using an ADSC Q315r
and ADSC Q270 CCD detectors, respectively, and all three structures
of GKAI were determined by molecular replacement (MR) using a
monomeric apo form of ECAI structure (2AJT) as a search model
showed that the molecular mass of the native protein was an
estimated 336 kDa (Fig. 1A), indicating that the heterologously
expressed GKAI is a homohexamer, a finding consistent with that of
ECAI [25]. High-resolution ICP-mass spectrometry showed that,
following purification, the recombinant GKAI contained an esti-
2
þ
[
26]. A subsequent refinement and model building process reduced
the Rwork and Rfree values to the reliable ranges for all the three
models (Table 1). In particular, the -arabitol PDB file was created
using the PHENIX program LIGANDFIT [34] and fitted into mF
density consistent with the ligand during the building of the
mated 0.06 ± 0.001 equivalents of Mn per subunit. Other divalent
2
þ
2þ
2þ
2þ
2þ
2þ
2þ
2þ
metal ions (e.g., Co , Ca , Cd , Cu , Mg , Fe , Ni , and Zn
)
L
were not detected in the purified enzyme. Further, GKAI that had
o
-DF
c
been treated with EDTA and subsequently isolated by dialysis
2
þ
L
2þ
-
contained only trace amounts of metals, including Mn
arabitol -Mn2 -GKAI model. Similar to ligand-bound GKAI, Mn
þ
(<0.01 ± 0.003 equivalents per monomer), indicating EDTA is an
effective chelator of AI-bound metal cations.
ion was also found in the electron density map for both of holo- and
þ
L
-arabitol -Mn2 -GKAI.
The temperature dependence of the recombinant enzyme was
determined by incubating the apo-GKAI at various temperatures
2þ
2.7. Accession numbers
for 20 min in the presence and absence of 1 mM Mn (Fig. 1B).
Under these assay conditions, GKAI exhibited maximal activity at
ꢀ
2þ
ꢀ
Database: structural data are available in Protein Data Bank,
60 C in the presence of 1 mM Mn . Incubation of GKAI at 60 C for
2
þ
www.pdb.org under the accession number 4R1O (apo-GKAI), 4R1P
holo-GKAI), and 4R1Q (holo-GKAI in complex with -arabitol).
120 min in the presence of 1 mM Mn caused no loss of AI activity,
ꢀ
(
L
whereas 50% of activity was loss after incubation at 70 C for
2
9.2 min (Fig. 1C). Consequently, the apo-GKAI was 1.5- to 9-fold
ꢀ ꢀ
3
. Results
less stable than the holo-GKAI at 60 C and 70 C, respectively,
2
þ
indicating that GKAI was thermostabilized by Mn , a finding
consistent with thermophilic AIs [7,20,21]. Purified GKAI had high
activity at pH 7.5 (Fig. 1D). As shown in Fig. 1E, EDTA-treated GKAI
showed significantly decreased activity (less than 25% compared
3.1. Biochemical characteristics of thermophilic GKAI
The araA gene (1.5 kb) encoding GKAI was cloned and success-
2
þ
fully expressed in E. coli BL21(DE3) as an N-terminal 6ꢂ His-tagged
fusion protein with a thrombin cleavage site upon induction with
IPTG as described previously [32]. GKAI was purified by heat
with GKAI in the presence of 1 mM Mn ) in the absence of divalent
2
þ
metal ions. However, addition of 1 mM Mn to the apo-GKAI
restored enzymatic activity, until it was 84% that of As-isolated
GKAI, consistent with the divalent metal ion requirements of
other thermophilic AIs [7,14,15,19]. Other divalent metal ions were
tested, including Co , Mg , Ca , Fe , Ni , Zn , and Cu , with
GKAI strongly activated in the presence of Mg (117% compared
treatment, a Ni2 -chelate affinity chromatography, and thrombin
þ
treatment to remove the His-tag residues. The apparent M
protein was estimated to be 56 kDa by SDS-PAGE, consistent with
the value of M (56,236) calculated from the presumptive amino
r
of the
2
þ
2þ
2þ
2þ
2þ
2þ
2þ
2
þ
r
2
þ 2þ 2þ
acid sequence. Subsequent size-exclusion gel chromatography
with Mn ), Ca (106%), and Co (58%). These results indicate
Table 1
Statistics for data collection and refinement of GKAI.
Mn2 -GKAI (4R1P)
þ
GKAI-Mn
2þ
- L
-arabitol (4R1Q)
Apo-GKAI (PDB ID 4R1O)
Data collection
Beam line
Space group
PAL-7A
P2
PAL-5C
C2
PAL-5C
1
2
1
2
1
P2
1 1 1
2 2
Wavelength (Å)
Unit-cell parameters (Å
1.00150
0.97951
1.00001
ꢀ
)
ɑ ¼ 116.72, b ¼ 140.75, c ¼ 215.82
ɑ ¼ 223.66, b ¼ 153.12, c ¼ 91.24
ɑ ¼ 118.51, b ¼ 146.26, c ¼ 215.68
ꢀ
ꢀ
ꢀ
ꢀ
a
¼
b
¼
g
¼ 90.00
a
¼
g
¼ 90.00 ,
b
¼ 103.85
a
¼
b
¼
g
¼ 90.00
Resolution (Å)
Total reflections
Unique reflection
Completeness (%)
Redundancy
50e2.40 (2.44e2.40)
982633
50e2.30 (2.34e2.30)
629574
50e2.25 (2.29e2.25)
2123676
133329
96.2 (94.1)
7.4 (7.1)
11.5 (45.9)
14.091 (3.75)
45.41
131054
99.5 (99.1)
4.8 (4.7)
14.4 (83.3)
18.33 (2.81)
45.31
175982
99.2 (99.4)
12.1 (11.4)
11.2 (45.9)
20.576 (5.29)
48.23
R-mergea (%)
Average I/s (I)
Solvent content (%)
Matthews coefficient (Å 3Da
Refinement
ꢃ1
)
2.25
2.25
2.37
R
work/Rfree (%)
16.43/21.58
2970
1182
15.56/21.29
15.38/20.12
Protein residues
Waters
Mn
2962
1171
6
2972
1645
6
L
-arabitol
6
RMSD
Angle ( )
ꢀ
1.091
0.007
25.6
1.094
0.008
28.3
1.071
0.007
29.8
Length (Å)
2
Wilson B factor (Å )
Ramachandran plot
Most favored regions (%)
Allowed regions (%)
outliers
96.48
3.45
0.07
97.19
2.61
0.20
97.37
2.57
0.07
Values in parentheses correspond to highest resolution shell.
a
R-merge ¼ Shkl
S
i
i
jI (hkl)e〈I(hkl)〉j/Shkl
i i i
S I (hkl), where I (hkl) is the observed intensity and 〈I(hkl)〉 is the average intensity of symmetry-related observations.

54
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
Fig. 1. Biochemical and biophysical characterization of GKAI. (A) Size-exclusion chromatography of recombinant GKAI. A superdex 200 16/60 column was run at a flow rate of
.0 mL/min, and the elution profile was monitored at 280 nm. The column was calibrated with blue dextran (2000 kDa), thyroglobulin (669 kDa), ferritin (440 kDa), aldolase
1
(
158 kDa), conalbumin (75 kDa), and ovalbumin (44 kDa) as standards. (B) Effect of temperature on recombinant GKAI activity. The temperature dependence of recombinant GKAI
ꢀ
was determined by incubating the apo and holo enzymes at various temperature (30e90 C) for 20 min. (C) Time course of irreversible thermal inactivation of purified GKAI. After
ꢀ
ꢀ
ꢀ
2þ
2þ
various incubation at 60 C (square), 65 C (circle), and 70 C (triangle) in the absence of Mn or in the presence of 1 mM Mn , aliquots were withdrawn, and their residual
activities were measured under the standard conditions. Open and closed symbols represent data points for the apo and holo enzymes, respectively. (D) Effect of pH on recombinant
GKAI activity. The pH dependence of recombinant GKAI was determined by cysteine-sulfuric acid-carbazole method to calculate the product after 20 min incubation at various pH
values. (E) Effect of divalent metal ions on GKAI activities. Reaction mixtures contained GKAI as metal-free enzymes (0.1 mg/ml) in the presence of 1 mM each metal ions. Activity
was measured at 60 C in triplicate under standard assay conditions without addition of 1 mM Mn . (F) The rates for GKAI-catalyzed isomerization at various concentrations of L-
arabinose as the substrate.
ꢀ
2þ
that the catalytic activity of GKAI is strongly dependent on divalent
of six identical subunits.
metal ions, a finding previously reported for thermophilic AIs
A ribbon representation of the overall monomer structure of
GKAI is shown in Fig. 2C. The root mean square deviations for C
a
[
7,20,21]. Kinetic parameters were determined by steady-state ki-
netic analyses with -arabinose and -galactose as substrates. GKAI
reactions yielded typical Michaelis-Menten curves, with the
steady-state kinetic values of K , Vmax, kcat, and kcat/K for -arab-
inose and -galactose as substrates (Fig. 1F). Although the kcat/K
value of GKAI is much greater for
the enzyme showed higher activity in converting
L
D
atoms and Z-scores of DALI search [35] for monomeric apo-GKAI
were 0.76 Å and 51.9, respectively, with monomeric apo-ECAI
(PDB code 2AJT, 60% identity), and 2.9 Å and 31.8, respectively,
with apo-ECFI (1FUI, 11% identity), suggesting that the overall
scaffold of GKAI is quite similar to those of sugar isomerases such as
ECAI and ECFI (Fig. 2D and E). Indeed, each monomeric subunit of
GKAI can be divided into three domains: the N-terminal domain (N
terminus-Leu177), the central domain (Lys178-Gly325), and the C-
terminal domain (Lys326-end), with each subunit composed of 17
m
m
L
D
m
L-arabinose than for
D
-galactose,
D
-galactose to
D
-
tagatose than other mesophilic counterparts [1,10,11,20].
3.2. Overall architecture of GKAI
a
-helices and 18
structurally similar to nucleotide binding domain, is found in the N-
terminal domain, where five parallel -strands ( 1- 5) alternative
with five -helices ( 1e 3 and 5e 6). Although the organization
b-strands (Fig. 2C). A typical Rossmann fold,
The crystal structures of the apo form of thermophilic GKAI and
of its holo form with and without -arabitol as a substrate analogue
b
b
b
L
a
a
a
a
a
were determined at resolutions of 2.40 Å, 2.30 Å and 2.25 Å,
respectively. In these structures, most amino acids, as well as metal
ions and substrates, were well fitted to their respective electron
density maps with reliable statistics (Table 1). All the solved
structures contained six subunits (M1, M2, M3, M4, M5, and M6),
which formed a sandwich-like homo-hexamer found in an asym-
metric unit with two main non-crystallographic symmetries: the 3-
fold symmetry in the homo-trimer and the 2-fold symmetry be-
tween the two trimers (Fig. 2A and B). The subunits of the hex-
of overall secondary structural elements in the central domain is
similar to that of the N-terminal domain, it only contains four
parallel
appearing to be Rossmann folds. The C-terminal domain contains
an antiparallel -barrel composed of six -strands ( 10e 11 and
14e 17). The three designated domains of each monomeric GKAI
b-strands (b6eb9), with links within helices (a6ea13) not
b
b
b
b
b
b
forms interfaces with two other monomers, allowing them to
cooperatively participate in the formation of active sites in func-
tional homo-hexamers (see below for details).
americ structures of apo-GKAI, Mn2 -GKAI, and Mn -GKAI bound
þ
2þ
a
to
L
-arabitol were found to be superimposable, with C r.m.s. de-
viations of 0.27e0.62 Å, 0.20e0.30 Å, and 0.18e0.25 Å, respectively.
Although the crystal packing of mesophilic ECAI contains a trimer
in the asymmetric unit [26], our size-exclusion chromatography
results (Fig. 1A), together with structural and biochemical data,
strongly suggested that GKAI may consist of a hexamer composed
3.3. The metal-binding site of GKAI
The initial active sites of GKAI structures have been proposed as
corresponding to the putative active site of apo-ECAI [26]. As shown
in Fig. 3A, the cavity between the N-terminal domain of one

J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
55
Fig. 2. Overall structure of GKAI. Top view (A) and side view (B) of the hexamer structure. The six molecules are colored in salmon (M1), yellow (M2), aquamarine (M3), green (M4),
2þ
slate (M5), and violet (M6). (C) Ribbon representation of detailed subunit structure of GKAI in complex with Mn . The structure of GKAI is comprised of the N-terminal domain
shown in green, the central domain in cyan, and the C-terminal domain in blue. The small orange sphere indicates metal ion. (D) Superimposed structure of monomer of apo-GKAI
(
4R1O) with that of apo-ECAI (2AJT). Each monomer is colored in deep salmon and white, respectively. (E) Alignment of the amino acid sequences of AIs from G. kaustophilus, E. coli,
and T. maritima. Secondary structure elements are shown with cylinders (
figure legend, the reader is referred to the web version of this article.)
a-helices) and arrows (b-strands) above the sequence. (For interpretation of the references to colour in this

56
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
subunit, consisting of the
a
-helices (
a
4,
a
5),
b
1e
a
1 loop, and the
replace the water molecules coordinated by four residues (Glu307,
Glu332, His349, and His448) (Fig. 3A). Upon the entry of the sub-
b
3e 3 loop from one subunit (M1 subunit) and loops in the central
a
and C-terminal domains from the neighboring subunit (M2), is like
a substrate-binding site. Notably, the neighboring subunit seems to
strate analogue L-arabitol into the active site, additional two water
molecules are pushed out and three residual water molecules
appeared to stabilize the substrate by forming strong H-bond
provide the metal and substrate-binding residues on
9e 13 and 15e 17. The active sites of both holo-GKAI and
arabitol bound GKAI around six inter-subunit clefts exhibit Mn
ions, with strong differences between electron density maps fol-
b6ea7,
2
þ
b
a
a
b
L
-
networks, together with Mn (Fig. 3A).
2
þ
3.4. The substrate-binding site of GKAI
2
þ
lowed by subsequent refinement. Since we added Mn to apo-
GKAI in screening crystals, the metal ions bound to GKAI were
The difference Fourier maps (mF
o c
-DF ) revealed a strong elec-
assumed to be Mn2 . Moreover, the ICP spectrum of an as-isolated
þ
tron density for bound -arabitol as the substrate analogue deep in
L
2
þ
form also showed the presence of Mn . The bound metal ions are
located at the cleft region of each subunit and coordinated in a
distorted tetrahedral form with Glu307, Glu332, His349, and
the intersubunit cleft formed among the three domains and with
coordination distances of atoms in the enzyme-substrate com-
plexes (Fig. 3). Each active site bound to L-arabitol was located in a
cleft between the C-terminal domain and the central domains of
the subunit adjacent to the other subunit interface (i.e., the N-ter-
2
þ
ε
His448 (Fig. 3B). The distance between Mn and the N atoms of
His349 and His448 were 2.54 Å and 2.45 Å, respectively, and the
distance between Mn2 and the O atoms of Glu307 and Glu332
were 2.43 Å and 2.47 Å, respectively. As shown in Fig. 4 and Table 2,
circular dichroism (CD) data demonstrated that site-directed
mutagenesis on the four residues involved in metal binding did
not affect the secondary structures of individual single mutant AIs
except H448A. Nevertheless, these mutants exhibited very little
þ
d2
minal domain) (Fig. 3B). The O1 atom of
L
-arabitol coordinates with
Mn , Glu307, and His 448, and forms H-bonds with His447
2.62 Å) and W1 (3.09 Å). The O2 atom of -arabitol coordinates
2
þ
(
L
2
þ
with Mn , Glu307, Glu332, and His349, and forms H-bonds with
W2 (3.14 Å). These coordinations are likely to help to fix the sub-
strate in the proper conformation for the hydride or proton shift
between the C1 and C2 of the substrate [36]. W1 and W2 are likely
key water molecules responsible for the hydride or proton shift in
the proposed catalytic mechanism, because they form H-bonds
m
activity and/or low catalytic efficiencies (kcat/K ) even in the
presence of Mn2 . Thus these results clearly demonstrated that the
metal-binding site of GKAI consisted of the four residues above. As
shown in Fig. 3A, six water molecules at the active site of apo-GKAI
were apparently H-bonded to hydrophilic residues, including
þ
with the O1 and O2 atoms of
O3 atom of -arabitol H-bonds with His129 of the neighboring
subunit (M2), and W3 forms multiple H-bonds with O4 atom of
L-arabitol, respectively. In addition, the
L
2
þ
Glu307, His447, and Glu332. In holo-GKAI, Mn
appeared to
L-
Fig. 3. Metal-binding and substrate-binding sites of GKAI. (A) The substrate binding site of GKAI at interface between two subunits are depicted as dotted rectangle box. Each of the
monomers (M1 and M2) are colored in salmon and yellow, respectively. For simplification, other subunits are shown as surface structure in gray. Close-up view of the water-
mediated H-bonds in the active sites of apo-, holo-, and L-arabitol bound GKAIs. Residues interacting with metal are depicted as sticks and the bound L-arabitol is shown in
2
þ
cyan stick. Mn and water molecules are shown as slate and red spheres, respectively. H-bonds are shown as dashed line in black. Ligands and side chains are colored by atom type
2
þ
(
oxygen in red and nitrogen in blue, respectively). (B) The mF -DF electron density map of Mn (left panel) and L-arabitol (right panel) contoured at 2s and represented to blue
o c
2þ
color. H-bonds are shown as dashed line. Residues interacting with metal and L-arabitol in GKAI are depicted as sticks. Mn and water molecules are shown as purple and red
spheres, respectively. The bound L-arabitol is shown in cyan stick. L-Arabitol in the active site of GKAI interacts with residues from M1 (salmon) and M2 (yellow) subunits and water
molecules. Ligands and side chains are colored by atom type (oxygen in red, nitrogen in blue, and phosphorous in orange, respectively). (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
57
Table 2
ꢀ
Kinetic parameters for GKAI (wild-type) and its single mutant AIs at 60 C.
b
ꢃ1
catb/k
ꢃ1
ꢃ1
Enzyme
K
m
(mM)
Vmax (U/mg)
k
cat (min
)
k
m
(mM min )
Wild-type 11.3 ± 1.4a 19.8 ± 0.4
1113.5
264.3
410.5
106.8
NA
81.9
24.7
31.1
3.0
Q17A
10.7 ± 0.9
13.2 ± 0.4
35.4 ± 8.7
4.7 ± 0.1
7.3 ± 0.5
1.9 ± 0.4
NA
H18A
L19A
Y20A
c
NA
NA
F84A
13.8 ± 6.8
4.4 ± 0.5
21.8 ± 4.8
NA
23.1 ± 1.0
9.9 ± 5.2
NA
8.4 ± 0.2
16.6 ± 0.1
NA
1.0 ± 0.3
2.8 ± 0.2
3.7 ± 0.1
NA
12.5 ± 0.5
2.8 ± 0.1
NA
7.8 ± 0.1
7.2 ± 0.1
NA
56.2
4.1
35.8
9.5
M186A
V189A
E307A
E332A
Y334A
H349A
M350A
I371A
H447A
H448A
157.5
208.1
NA
703.0
157.5
NA
438.6
404.9
NA
NA
30.4
15.9
NA
52.2
24.4
NA
NA
NA
NA
NA
a
Data are means ± standard deviation.
The kcat is the number of substrate molecules reacted per active per min.
NA, no activity.
b
c
holo-GKAI with L-arabitol can recognize substrates with different
configurations of C4, C5, O4, and O5 by using Tyr20, Leu19 and
Gln17, and vice versa (Figs. 5 and 3B). For example, O4 H-bonds with
Tyr20, and C5 is coordinated with Leu19. In addition, Tyr20 and
Leu19 of holo-GKAI are proximal to O4 of
Fig. 3B). By contrast, when the structure of
was superimposed on that of apo-GKAI without
corresponding Leu19 and Tyr20 residues of apo-GKAI were located
.95 Å and 5.29 Å away from O4 atoms of -arabitol, respectively
Fig. 5). Thus, the substrate recognition at the 4 and 5 positions of
arabitol is thought to be less favorable for apo-GKAI than that of
L-arabitol, respectively
2
þ
(
L
-arabitol -Mn -GKAI
L-arabitol, the
8
L
(
L
-
-
L
arabitol for holo-GKAI, a finding consistent with the significant
lower enzyme activity of apo enzyme than that of the holo enzyme
(Fig. 1). In addition, CD data demonstrated that site-directed
mutagenesis on the substrate binding residues above did not lead
to any significant perturbation of the structural integrity of each
single mutant AI. Nevertheless, these mutant AIs exhibited less
activity and/or lower catalytic efficiencies than the wild-type
enzyme, which was probably due to lacking of any conforma-
tional changes in substrate binding (Table 2 and Fig. 4C). Therefore,
substrate binding by GKAI involves the residues Met186 (
loop), Met350 ( 11e 14 loop), Ile371 ( 12- 13 loop), and His447
15e 17 loop) of the M1 subunit, and Gln17 (M2), Leu19 (M2), and
Tyr20 ( 1e 1 loop, M2), in the N-terminal domain of the neigh-
boring subunit.
b6ea7
b
a
b
b
(a
b
Fig. 4. The influence of mutations on the structural integrity and activity of GKAI. (A)
SDSePAGE of purified GKAI (WT) and its single mutants. (B) The CD spectra of the
wild-type and single mutant AIs. (C) The effects of single mutations on AI activity. The
AI activity of the wild-type enzyme for L-arabinose as a substrate was taken as 100%.
b
a
3.5. Effect of metal ions on the structural stability of hexameric
arabitol, Tyr334 (M1), Asn123 (M2), and Gln126 (M2) (Fig. 3A).
Met186 and Met350 are coordinated by the O5 and O4 atoms,
respectively, recognizing the hydroxyl groups at the 4 and 5 posi-
GKAI
To investigate the effect of metal ions on the structural stability
tions of
L
-arabitol, and C4 exhibits hydrophobic interactions with
of GKAI, we determined the oligomeric state of the apo and holo
forms of GKAI using size-exclusion chromatography. As shown in
Fig. 6A, as-isolated GKAI showed predominantly a hexameric
structure, whereas apo-GKAI showed two distinct populations such
as an intact hexamer and a partially unfolded and/or dissociated
Phe280 (M1), Ile371 (M1), and Leu19 (M2) (Fig. 3B). Remarkably, O4
and O5 also form H-bonds with Tyr20 (M2, 2.56 Å) and Gln17 (M2,
2
.93 Å), respectively, strongly indicating that the substrate-binding
coordination occurs through intersubunit interactions.
The crystal structures of GKAI and its complex with
a substrate analogue showed that the enzyme undergoes large
displacements of several amino acids, reflecting conformational
changes in the course of catalysis. Remarkably, structural differ-
ences between the apo and holo forms of GKAI are observed in the
recognition of atoms at the 4 and 5 positions of
the observation that the structure of holo-GKAI was highly super-
imposed on that of
-arabitol bound GKAI in the presence of Mn2þ
2
þ
to apo-GKAI shifted
L
-arabitol as
form. Remarkably, the addition of Mn
significantly the oligomerization equilibrium toward an intact
hexamer ( ) as a major population, indicating that metal ions play
a
6
an important role in the oligomerization of GKAI.
To further assess the effect of metal ions on the structural sta-
bility of the apo and holo forms of GKAI, we used differential
scanning calorimetry (DSC) to determine their melting tempera-
L-arabitol. Based on
L
,
m
tures (T ) (Fig. 6B). Intriguingly, thermal denaturation of apo-GKAI

58
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
2
þ
Fig. 5. Conformational change of the
GKAI moves toward M1 subunit upon Mn -binding. Displacement of residues on the
b
1e
a
1 loop involved in the active site of GKAI and ECAI upon Mn -binding. Compared with apo-GKAI, the
b1ea
1 loop of M2 subunit in holo-
2
þ
b1e
a
1 loop is shown as dashed arrows in black. Side chains of M1 subunit (salmon) and M2
(
yellow) residues that are involved in ligand coordination are represented as sticks. In contrast, such a conformational change is not observed in the corresponding
b
1e 1 loop
a
regions of apo-ECAI (2AJT) and holo-ECAI (2HXJ). Ligands and side chains are colored by atom type (oxygen in red and nitrogen in blue, respectively). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
ꢀ
revealed
a
major
midpoint
of
thermal
transition
other by 30 ,
a
7 in the central domain of one subunit and the loop
ꢀ
ꢀ
(
T
m
¼ 71.09 ± 0.56 C) with a minor one at 65.17 ± 1.44 C, whereas
structure between
b
6 and 7 interact with the dimeric junction re-
a
ꢀ
holo-GKAI exhibited two thermal transitions at 78.94 ± 0.11 C and
gion formed by two domains of the other subunits (i.e., the central
domain in one subunit and the N-terminal domain in the other
subunit) (Fig. 7, M2-M5 interface). Along with the contact area be-
tween the central domains, the same residues in the same loop face
each other in the 3D arrangement, with nine H-bonds and five strong
salt bridges (i.e., Arg181:Asp184, Asp184:Arg181, Lys196:Glu193,
Asp214:His284, and His284:Asp214) (Fig. 7, M1-M5 interface and
7
1.88 ± 0.34, respectively. Notably, the major and minor transition
temperatures of holo-GKAI coincided with the T values of As-
isolated GKAI, which were higher than the T values of apo-GKAI
Fig. 6B). Together with gel-filtration analysis, the DSC data thus
m
m
(
clearly indicate that the holo form of GKAI is more thermody-
namically stable than its apo form, suggesting that the binding of
metal ions significantly contribute to the thermal stability of mul-
timeric GKAI.
Table S1). Two short a-helices, 7 and 3, run antiparallel, contacting
each other via a salt bridge between Gln200 and Lys88 (Fig. 7, M2-M5
interface). Another important feature of the hexameric formation was
3.6. Structural basis for inter-subunit interactions
that two a-helices (a2s) between the N-terminal domains of each
subunit run antiparallel to each other by the salt bridges (Arg62:-
Glu72 and Arg98:Glu96) and there are H-bondings among Arg62,
Asn69, and Glu72 (Fig. 7, M2-M5 interface and Table S1). Additional
contacts could also be deduced from active site, with salt bridges
formed between Lys88:Asn208 and Asp195:Arg140.
A hexameric structure consists of two identical trimers (i.e., upper
and lower sides) stacked through head-to-tail contacts between
monomers (Fig. 7). The PDBePISA calculations revealed that buried
surface (total surface) areas of hexameric apo-, holo-, and
L
-arabitol-
2
2
bound GKAI were 39,340 (89,030) Å , 40,120 (89,520) Å , and 46,230
CASTP (http://cast.engr.uic.edu) analysis revealed that the contact
between the C-terminal domain of one subunit and the N-terminal
domain of another subunit in apo-GKAI (M1-M2) created a major
2
(
89,070) Å , respectively. Every subunit has three regions interacting
with two adjacent monomers on the same side and one from the
other side, with most of these reactions occurring through H-bonds
and salt bridges (Fig. 7 and Table S1). Residues in the C-terminal
domain of each subunit are in contact with those of the N-terminal
domain of the neighboring subunit on the same side. In addition to
these head-to-tail interactions on the same side, each subunit of a
trimer was in contact with two monomeric subunits in the other
trimer. Since two plates (a trimer per plate) are twisted around each
3
cavity for substrate binding (483.3 Å ). The solvent accessible area
2
(SAS) of the surface of the cavity (386.9 Å ) is about 17% of the total
2
SAS (2285 Å ) (Table S3). This space is surrounded by the loop
structures between
domains in a subunit and the contact regions between the
10) of the C-terminal domain and the -helix ( 4) and 4e
of the N-terminal domain in a neighboring unit (Fig. 7, M2-M1
b
-strands 11 and 12 of the central and C-terminal
-strand
5 loop
b
a
(b
a
a
a

J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
59
Fig. 6. (A) Size-exclusion chromatography for the apo and holo forms of GKAI. The Superdex 200 10/300 column (GE healthcare) was calibrated with protein size standards. HMW,
high molecular weight; LMW, low molecular weight. (B) DSC data for the apo (blue line) and holo (red line) forms of GKAI. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
interface). Other subunits with neighboring subunits (e.g., M2-M3
and M3-M1) were also arranged in the same manner. Consequently,
the cavity can be divided into two distinct regions. One region forms
an active site, with most residues on the loops from two subunits
coordinating with metal and ligand, whereas the other region,
extending from the active site, shows several direct interactions
directly correlated with conversion efficiency, need to be further
improved through protein engineering for a cost-effective biolog-
ical process. Closer scrutiny of our crystal structures reveals the
general principles of the novel structural AI fold and the molecular
mechanism underlying metal-mediated catalysis and thermosta-
bility via intersubunit interactions. During aldose-ketose isomeri-
zation, one hydrogen is transferred as a proton between the
electronegative oxygens of the hydroxyl and carbonyl groups of the
aldose, whereas transfer of the second hydrogen from C-2 to C-1
may occur directly as a hydride (Fig. S1). Both isomerization can
between
and the
Table S1). This space was formed by a combination of a
15 in the C-terminal domain and 5 from the N-terminal domain.
Two important modes of inter-chain interactions are present inside
the region. The penetration of a long side chain of Arg132 in 5 of the
N-terminal domain allowed it to form a salt bridge and an H-bond
with Asp333 (in 10) and Tyr336 in the C-terminal domain of a
neighboring monomer, respectively (Fig. 7, M2-M1 interface).
Although away from the region, His105 in 4 projecting into space is
H-bonded to Tyr336. In addition, Arg132 links two H-bonds formed
between His129 in 5 and Asp333, which creates serial interactions
b
-strands (part of the
-helix of the N-terminal domain of each subunit (Fig. 7 and
-barrel and
b-barrel) of the C-terminal domain
a
b
a
a
2
þ
occur under aqueous conditions, with Mn stabilizing the transi-
tion state for both reactions, suggesting that metal ion catalysis of
isomerization provides a simple explanation for the requirement by
GKAI of divalent metal ions [36]. The enzyme activity of thermo-
stable AIs was highly dependent on metal ions for catalysis [7].
Moreover, in contrast to mesophilic AI, thermophilic AI undergoes
a
b
b
2
þ
conformational changes induced by Mn at elevated temperatures
[20]. Indeed, the crystal structures of the apo and holo forms of
a
of two partitions in the region. These “serial alternative or staggered
chain interactions” seem to tighten two adjacent subunits, providing
the active site architecture surrounded by loops. In addition, the M2-
M1 intersubunit interactions may be further stabilized by an addi-
tional H-bond between Arg145 in the M2 subunit and the carbonyl
oxygen of Leu442 in the M1 subunit.
GKAI with and without bound L-arabitol elucidate the metal-
dependent catalysis of AIs. Each monomer of hexameric GKAI un-
dergoes large displacements of several amino acids upon metal and
substrate binding (Fig. 5), supporting previous findings that the
2
þ
binding of Mn to GKAI converts the inactive form of the enzyme
into an active form at elevated temperatures. We thus conclude
2
þ
that GKAI may exhibit sequential, ordered kinetics, with Mn
binding first, followed by -arabinose.
4
. Discussion
L
Despite the structural similarity of GKAI and ECAI, catalysis
resistant to thermal inactivation is an intriguing issue. As described
above, divalent metal ions significantly contributed to the overall
architecture of GKAI (Fig. 6B), strongly indicating that metal-
Although (hyper)thermophilic AIs are quite advantageous for
the commercial production of
D-tagatose from D-galactose [13e18],
their substrate specificities and thermostabilities, which are

60
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
Fig. 7. Cavity formation of hexameric GKAI via intersubunit interactions. Interfaces between upper two subunits (M1 and M2) and lower one (M5) subunit are shown as rectangle
dotted boxes. In M2-M5 interface, each monomer contacts in an antiparallel fashion, and forms salt bridges and H-bonds between 2e 2 and 3 ( 7)e 7 ( 3) helices, respectively
left panel). M1 and M5 subunits are in contact with each subunit through identical loop faces' interaction (right panel). Extended region for intersubunit interactions by residues of
a
a
a
a
a
a
(
M1 and M2 subunits is shown as dashed lines (upper panel). The M1, M2 and M5 subunits are depicted as ribbon representations, and colored in salmon, yellow, and slate,
respectively. Non-conserved Arg residues in GKAI are shown as red colored number. H-bonds and salt bridges are shown as dashed lines in black. (For interpretation of the ref-
erences to colour in this figure legend, the reader is referred to the web version of this article.)
ꢀ
mediated transition of GKAI from the partially unfolded and/or
dissociated form to an intact hexameric form is quite prerequisite
not only for catalysis but also for the maintenance of oligomeric
states of GKAI at elevated temperatures. Many studies have sug-
gested that several factors are mainly involved in the thermosta-
bility of (hyper)thermophilic enzymes: for example, amino acid
composition and changed environment of the active site surface
salt bridges, with more free energy of ~100 kJ/mol at 70 C.
High resolution structures of GKAI also clearly showed the H-
bonding networks at the active site in detail. Novel H-network
formation by additional water-mediated H-bonds between amino
acid residues and the substrate was found to stabilize the active site
of a thermophilic xylanase during catalysis [44]. As shown in Fig. 3,
the active site of apo-GKAI shows 2e3 more water-mediated H-
bonds than that of apo-ECAI (2AJT), suggesting that thermophilic AI
may tighten substrate binding with more H-bonding networks.
[37,38]. These factors are highly associated with the proportion of
charged residues (especially Arg) on interfaces and water-mediated
H-bonds at the active site of an enzyme. For example, the Arg
content of GKAI, >2%, is higher than that of other mesophilic sugar
isomerizing enzymes, such as ECAI and ECFI (Table S2). Indeed,
non-conserved Arg62 and Arg98 in the N-terminal domain of one
subunit were found to be in contact with Glu72 and Glu96 in the N-
terminal domain of another subunit, perhaps contributing to sub-
unit interactions through four salt bridges since two trimeric units
are in inverse contact, a contact not observed in mesophilic AIs
Thus additional H-bonds found in the GKAI bound to L-arabitol may
provide supportive evidence, that those water molecules play a
major role in maintaining the integrity of the active site of the
enzyme at elevated temperatures. Although several residues,
including Leu19 and Tyr20 (Leu16 and Tyr17 in ECAI), are
conserved, only the side chains in GKAI underwent significant
conformational changes, resulting in its interaction with the sub-
strate via H-bonds with water molecules. Moreover, we observed
new water-mediated H-bonds to previously unknown substrates.
Since individual H-bonds have an average free energy of 5.5 kJ/mol
(~1.3 kcal/mol), contributing to enzyme stabilization [45], energy
greater than ~20 kJ/mol may better stabilize the transition state of
substrate in GKAI than in a mesophilic AI during catalysis. Although
further investigations are needed, with additional structural
studies on thermophilic enzymes, our findings provide a valuable
example of the thermal stabilization of substrate mediated by
substantial H-bonding, inasmuch as there were no such confor-
mational changes in the corresponding residues in ECAI and ECFI.
Therefore, we suggest that the extensive salt bridges in combina-
tion with H-bonds near the active sites may play important roles in
supporting and maintaining the integrity of the catalytic site during
functional assembly of hexameric AI monomers.
(Fig. 7, M2-M5 interface). In addition, Arg145 was found to
contribute to the H-bond with the peptide carbonyl oxygen of
Leu442 (Fig. 7, M2-M1 interface). These results are consistent with
previous findings, suggesting that more salt bridges/ion pairs
(
network of salt bridges) and H-bonds are key to increased ther-
mostability [39,40]. Intriguingly, the breakdown of two inter-
subunit salt bridges was sufficient to have an adverse effect on the
thermostability of hexameric thermostable glutamate de-
hydrogenases [41]. A single salt bridge rather than hydrophobic
interaction can provide free energy of ~22 kJ/mol, contributing to
the stabilization of protein folding at elevated temperatures, inas-
much as the stabilization of cellular proteins requires a free energy
of ~50 kJ/mol [42,43]. In this regard, it was not surprising that GKAI
was more stable than ECAI, inasmuch as the former had additional

J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
61
AIs have very low structural and sequence homology with other
phosphosugar and non-phosphosugar isomerases, indicating that
their structures are dissimilar, with no clear structure such as a
general ketol isomerase fold. Therefore, the lack of both structural
and sequence conservation suggests that AI is generally unrelated
to these sugar isomerases. However, the apparent conservation of
the overall fold in GKAI and ECFI crystal structures, despite their
low sequence similarity, suggests that the overall fold may be
conserved among AIs, with the overall fold therefore called an AI
fold. Fig. 8 shows the superimposed substrate-binding structures of
a clear boundary between M1 and M2 with a relatively small
2
contact area of 977 Å . In addition, the
enough to cover the substrate-binding site of the neighboring
subunit (Fig. 5). In ECAI with -ribitol, the corresponding residue is
Leu18 (Leu19 in GKAI), located 4.3 Å from C5 of -ribitol ( -arabitol
b1ea1 loop is not proximal
D
D
L
in GKAI), with a large space between Leu18 and the bound sub-
strate. In addition, His18 of GKAI is partly involved in electrostatic
interactions by changing its side-chain conformation following the
binding of a substrate, resulting in favorable interactions with the
substrate. Thus the unique side-chain conformation around the
hydroxyl group of C5, formed by residues Gln17, Leu19, and Tyr20
GKAI with bound
ECFI with bound inhibitor
involved in metal binding are conserved, and the interactions of O1,
O2 and O3 of -arabitol in GKAI are very similar to those in ECAI,
L-arabitol, ECAI with bound
D-ribitol (4F2D), and
L-fucitol. The amino acid residues
from the
b1ea1 loop of GKAI, recognizes L-arabitol as the most
suitable substrate, leading to the high substrate specificity of GKAI.
2þ
L
In the absence of Mn , however, the substrate-binding site of GKAI
loosely recognizes both -arabitol and -galactose as substrates.
This is consistent with the enzymatic activity of ECAI being much
lower for -galactose than that of GKAI.
Since -arabinose and -galactose have the same configurations
of asymmetric carbon atoms as -arabitol, their substrate-enzyme
suggesting that the catalytic mechanism of metal-mediated proton
transfer proposed for ECAI [46] also occurs for GKAI. In addition, the
L
D
2
þ
interactions between O1, O2, and O3 of
L-fucitol and ECFI with Mn
D
are also quite similar to those of GKAI. In ECFI forming a crystal-
lographic hexamer [27], the substrate-binding site for a neigh-
L
D
L
boring molecule consists partly of the
b1ea
1 loop, and an
interactions are thought to be almost identical, but their enzyme
extensively opened accessible surface is formed between M1 and
M2, as found in GKAI (Fig. 8). In ECAI, however, the corresponding
activities are significantly different. The structural variation be-
tween
L
-arabitol and
L-arabinose is found at the C5 position, where
b
1e
a1 loop covers the bound metal ions in the same subunit, with
L-arabinose has a hydroxyl group instead of an aldehyde group and
D-galactose has no group (Fig. 8). In the binding of D-galactose, a
lack of hydrophobic interactions of C6 with the enzyme found in
GKAI/ -arabitol may reduce enzyme activity toward -galactose.
Displacements of several amino acid residues restructure the
substrate-binding site, so that -galactose as the substrate is more
L
D
D
accessible to mobilization. Because of the structural similarity of its
substrates, which may be due to the sharing of hydroxyl groups at
C3eC4 in a cis-configuration, AI can also catalyze the isomerization
of
D-galactose to D-tagatose [1,47].
5. Conclusion
In this study, we determined the functional and structural
characteristics of a thermophilic -arabinose isomerase (AI) that
catalyzes the isomerization of -arabinose ( -galactose) to -ribu-
lose ( -tagatose). Despite the fact that thermophilic AIs are used to
L
L
D
L
D
produce an FDA-approved GRAS natural sweetener as well as a new
antidiabetic and obesity control neutraceutical, little is known
about the molecular basis underlying the substrate preferences and
metal requirements of multimeric AIs because there has been no
information about the structural and functional properties of these
AIs up until now. We expressed and purified an AI from the ther-
mophilic bacterium G. kaustophilus, and determined for the first
time the crystal structures of its apo and holo forms as hexamers as
well as that of holo-AI in complex with L-arabitol. The crystal
structures not only provide a molecular basis for understanding the
formation of the substrate-binding pockets through metal-
mediated inter-subunit interactions, but also furnish a substrate-
binding model for understanding isomerization reactions at
elevated temperatures, which are distinct from those of its meso-
philic counterparts.
Fig. 8. Close-up view of the active sites of GKAI, ECAI, and ECFI. All three proteins are
partly shown with a transparent surface and ribbon representation. Oxygen, nitrogen
and sulfur residues are shown in red, blue and yellow, respectively. All residues shown
are identical and/or corresponding to each other in GKAI, ECAI, and ECFI. Active site
residues are shown as sticks in their subunit colors and are labeled. One subunit of
GKAI and its neighboring subunit are depicted in yellow and whale, respectively. The
substrate analogue L-arabitol is shown as cyon sticks. One subunit of ECAI and its
neighboring subunit are depicted in white and orange, respectively. The inhibitor D-
ribitol is shown as green sticks. One subunit of ECFI and its neighboring subunit are
depicted in slate and pink, respectively. The substrate analogue L-fucitol is shown as
yellow sticks. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
Footnotes
This work was supported by the National Research Foundation
of
Korea
(NRF)
grants
(2014R1A2A2A01006765
&
2013R1A1A2057465) to Dr. Dong-Woo Lee and to Dr. Sung Haeng
Lee, respectively, funded by the Ministry of Science, ICT and Future
Planning (MSIP), Korea, and by a Grant 311042-05-1-HD120 from
the Korea Institute of Planning & Evaluation for Technology (iPET)
funded by the Ministry for Food, Agriculture, Forestry and Fisheries
to Dr. Dong-Woo Lee.

6
2
J.M. Choi et al. / Archives of Biochemistry and Biophysics 596 (2016) 51e62
[20] D.W. Lee, E.A. Choe, S.B. Kim, S.H. Eom, Y.H. Hong, S.J. Lee, H.S. Lee, D.Y. Lee,
Acknowledgments
Y.R. Pyun, Arch. Biochem. Biophys. 434 (2005) 333e343.
21] D.W. Lee, Y.H. Hong, E.A. Choe, S.J. Lee, S.B. Kim, H.S. Lee, J.W. Oh, H.H. Shin,
Y.R. Pyun, FEBS Lett. 579 (2005) 1261e1266.
[
We thank the Beamline (PLS-5C) staff at the Pohang Light Source
for technical support during data collection. We wish to confirm
that there are no known conflicts of interest associated with this
publication.
[22] Y.H. Hong, D.W. Lee, Y.R. Pyun, S.H. Lee, J. Agric. Food Chem. 59 (2011)
2939e12947.
23] S.J. Lee, Y.J. Lee, S.B. Kim, S.K. Kim, D.W. Lee, Appl. Environ. Microbiol. 78
2012) 8813e8816.
1
[
(
[
[
[
[
24] J.W. Patrick, N. Lee, J. Biol. Chem. 243 (1968) 4312e4318.
25] J.W. Patrick, N. Lee, J. Biol. Chem. 244 (1969) 4277e4283.
26] B.A. Manjasetty, M.R. Chance, J. Mol. Biol. 360 (2006) 297e309.
27] J.E. Seemann, G.E. Schulz, J. Mol. Biol. 273 (1997) 256e268.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.abb.2016.02.033.
[28] Y.J. Lee, S.J. Lee, S.B. Kim, S.H. Lee, D.W. Lee, FEBS Lett. 588 (2014) 1064e1070.
[
29] T.N. Nazina, T.P. Tourova, A.B. Poltaraus, E.V. Novikova, A.A. Grigoryan,
A.E. Ivanova, A.M. Lysenko, V.V. Petrunyaka, G.A. Osipov, S.S. Belyaev,
M.V. Ivanov, Int. J. Syst. Evol. Microbiol. 51 (2001) 433e446.
References
[30] H. Takami, Y. Takaki, G.J. Chee, S. Nishi, S. Shimamura, H. Suzuki, S. Matsui,
I. Uchiyama, Nucleic Acids Res. 32 (2004) 6292e6303.
[
[
[
1] P.S.J. Cheetham, A.N. Wootton, Enzyme Microb. Technol. 15 (1993) 105e108.
2] I. Sa-Nogueira, H. de Lencastre, J. Bacteriol. 171 (1989) 4088e4091.
3] I. Sa-Nogueira, T.V. Nogueira, S. Soares, H. de Lencastre, Microbiology 143 (Pt.
[31] D.W. Lee, D. Kim, Y.J. Lee, J.A. Kim, J.Y. Choi, S. Kang, J.G. Pan, Proteomics 13
(2013) 2278e2282.
[32] T.P. Cao, J.M. Choi, S.J. Lee, Y.J. Lee, S.K. Lee, Y. Jun, D.W. Lee, S.H. Lee, Acta
Crystallogr. F. Struct. Biol. Commun. 70 (2014) 108e112.
3) (1997) 957e969.
[
[
[
[
4] R. Schleif, Trends Genet. 16 (2000) 559e565.
5] Y. Lu, G.V. Levin, T.W. Donner, Diabetes Obes. Metab. 10 (2008) 109e134.
6] G.V. Levin, J. Med. Food 5 (2002) 23e36.
7] D.W. Lee, H.J. Jang, E.A. Choe, B.C. Kim, S.J. Lee, S.B. Kim, Y.H. Hong, Y.R. Pyun,
Appl. Environ. Microbiol. 70 (2004) 1397e1404.
8] D.K. Oh, Appl. Microbiol. Biotechnol. 76 (2007) 1e8.
9] P. Kim, Appl. Microbiol. Biotechnol. 65 (2004) 243e249.
[33] Z. Dische, E. Borenfreund, J. Biol. Chem. 192 (1951) 583e587.
[34] P.D. Adams, R.W. Grosse-Kunstleve, L.W. Hung, T.R. Ioerger, A.J. McCoy,
N.W. Moriarty, R.J. Read, J.C. Sacchettini, N.K. Sauter, T.C. Terwilliger, Acta
Crystallogr. D. Biol. Crystallogr. 58 (2002) 1948e1954.
[35] S. Dietmann, J. Park, C. Notredame, A. Heger, M. Lappe, L. Holm, Nucleic Acids
Res. 29 (2001) 55e57.
[36] R.W. Nagorski, J.P. Richard, J. Am. Chem. Soc. 123 (2001) 794e802.
[37] C. Vieille, G.J. Zeikus, Microbiol. Mol. Biol. Rev. 65 (2001) 1e43.
[38] T.J. Taylor, I.I. Vaisman, BMC Struct. Biol. 10 (Suppl. 1) (2010) S5.
[39] D. Maes, J.P. Zeelen, N. Thanki, N. Beaucamp, M. Alvarez, M.H. Thi,
J. Backmann, J.A. Martial, L. Wyns, R. Jaenicke, R.K. Wierenga, Proteins 37
(1999) 441e453.
[40] S. Kumar, C.J. Tsai, R. Nussinov, Protein Eng. 13 (2000) 179e191.
[41] C. Vetriani, D.L. Maeder, N. Tolliday, K.S. Yip, T.J. Stillman, K.L. Britton,
D.W. Rice, H.H. Klump, F.T. Robb, Proc. Natl. Acad. Sci. U. S. A. 95 (1998)
12300e12305.
[
[
[
10] H.J. Roh, P. Kim, Y.C. Park, J.H. Choi, Biotechnol. Appl. Biochem. 31 (Pt. 1)
(
2000) 1e4.
[
[
[
11] K. Yamanaka, Methods Enzymol. 41 (1975) 458e461.
12] Y.B. Tewari, D.K. Steckler, R.N. Goldberg, Biophys. Chem. 22 (1985) 181e185.
13] Y.H. Hong, D.W. Lee, S.J. Lee, E.A. Choe, S.B. Kim, Y.H. Lee, C.I. Cheigh, Y.R. Pyun,
Biotechnol. Lett. 29 (2007) 569e574.
[
[
14] F. Jorgensen, O.C. Hansen, P. Stougaard, Appl. Microbiol. Biotechnol. 64 (2004)
8
16e822.
15] X.G. Hung, W.C. Tseng, S.M. Liu, W.S. Tzou, T.Y. Fang, Biochem. Eng. J. 83
2014) 121e128.
(
[42] R. Jaenicke, FEMS Microbiol. Rev. 18 (1996) 215e224.
[
[
16] B.C. Lim, H.J. Kim, D.K. Oh, Appl. Biochem. Biotechnol. 149 (2008) 245e253.
17] M. Rhimi, R. Ilhammami, G. Bajic, S. Boudebbouze, E. Maguin, R. Haser,
N. Aghajari, Bioresour. Technol. 101 (2010) 9171e9177.
[43] P.L. Privalov, S.J. Gill, Adv. Protein Chem. 39 (1988) 191e234.
[44] A.L. van Bueren, S. Otani, E.P. Friis, K.S. Wilson, G.J. Davies, Acta Crystallogr.
Sect. F. Struct. Biol. Cryst. Commun. 68 (2012) 141e144.
[
18] M. Liang, M. Chen, X. Liu, Y. Zhai, X.W. Liu, H. Zhang, M. Xiao, P. Wang, Appl.
Microbiol. Biotechnol. 93 (2012) 1469e1474.
19] B.C. Kim, Y.H. Lee, H.S. Lee, D.W. Lee, E.A. Choe, Y.R. Pyun, FEMS Microbiol.
Lett. 212 (2002) 121e126.
[45] B.A. Shirley, P. Stanssens, U. Hahn, C.N. Pace, Biochemistry 31 (1992) 725e732.
[46] S. Banerjee, F. Anderson, G.K. Farber, Protein Eng. 8 (1995) 1189e1195.
[47] K. Izumori, Y. Ueda, K. Yamanaka, J. Bacteriol. 133 (1978) 413e414.
[