Dimeric Crystal Structure of Rabbit l-Gulonate 3-Dehydrogenase/λ-Crystallin: Insights into the Catalytic Mechanism

Article abstract of DOI:10.1016/j.jmb.2010.06.069

l-Gulonate 3-dehydrogenase (GDH) is a bifunctional dimeric protein that functions not only as an NAD⁺-dependent enzyme in the uronate cycle but also as a taxon-specific λ-crystallin in rabbit lens. Here we report the first crystal structure of GDH in both apo form and NADH-bound holo form. The GDH protomer consists of two structural domains: the N-terminal domain with a Rossmann fold and the C-terminal domain with a novel helical fold. In the N-terminal domain of the NADH-bound structure, we identified 11 coenzyme-binding residues and found 2 distinct side-chain conformers of Ser124, which is a putative coenzyme/substrate-binding residue. A structural comparison between apo form and holo form and a mutagenesis study with E97Q mutant suggest an induced-fit mechanism upon coenzyme binding; coenzyme binding induces a conformational change in the coenzyme-binding residues Glu97 and Ser124 to switch their activation state from resting to active, which is required for the subsequent substrate recruitment. Subunit dimerization is mediated by numerous intersubunit interactions, including 22 hydrogen bonds and 104 residue pairs of van der Waals interactions, of which those between two cognate C-terminal domains are predominant. From a structure/sequence comparison within GDH homologues, a much greater degree of interprotomer interactions (both polar and hydrophobic) in the rabbit GDH would contribute to its higher thermostability, which may be relevant to the other function of this enzyme as λ-crystallin, a constitutive structural protein in rabbit lens. The present crystal structures and amino acid mutagenesis studies assigned the role of active-site residues: catalytic base for His145 and substrate binding for Ser124, Cys125, Asn196, and Arg231. Notably, Arg231 participates in substrate binding from the other subunit of the GDH dimer, indicating the functional significance of the dimeric state. Proper orientation of the substrate-binding residues for catalysis is likely to be maintained by an interprotomer hydrogen-bonding network of residues Asn196, Gln199, and Arg231, suggesting a network-based substrate recognition of GDH.

Full text of DOI:10.1016/j.jmb.2010.06.069

doi:10.1016/j.jmb.2010.06.069
J. Mol. Biol. (2010) 401, 906–920
Available online at www.sciencedirect.com
Dimeric Crystal Structure of Rabbit L-Gulonate
3-Dehydrogenase/λ-Crystallin: Insights into
the Catalytic Mechanism
Yukuhiko Asada1, Chizu Kuroishi2, Yoko Ukita2,
Rie Sumii3, Satoshi Endo3, Toshiyuki Matsunaga3,
Akira Hara3 and Naoki Kunishima1
⁎
1Protein Crystallography
Research Group, RIKEN
SPring-8 Center, Harima
Institute, 1-1-1 Kouto,
Sayo-cho, Sayo-gun,
L-Gulonate 3-dehydrogenase (GDH) is a bifunctional dimeric protein that
functions not only as an NAD+-dependent enzyme in the uronate cycle but
also as a taxon-specific λ-crystallin in rabbit lens. Here we report the first
crystal structure of GDH in both apo form and NADH-bound holo form.
The GDH protomer consists of two structural domains: the N-terminal
domain with a Rossmann fold and the C-terminal domain with a novel
helical fold. In the N-terminal domain of the NADH-bound structure, we
identified 11 coenzyme-binding residues and found 2 distinct side-chain
conformers of Ser124, which is a putative coenzyme/substrate-binding
residue. A structural comparison between apo form and holo form and a
mutagenesis study with E97Q mutant suggest an induced-fit mechanism
upon coenzyme binding; coenzyme binding induces a conformational
change in the coenzyme-binding residues Glu97 and Ser124 to switch their
activation state from resting to active, which is required for the subsequent
substrate recruitment. Subunit dimerization is mediated by numerous
intersubunit interactions, including 22 hydrogen bonds and 104 residue
pairs of van der Waals interactions, of which those between two cognate C-
terminal domains are predominant. From a structure/sequence comparison
within GDH homologues, a much greater degree of interprotomer
interactions (both polar and hydrophobic) in the rabbit GDH would
contribute to its higher thermostability, which may be relevant to the other
function of this enzyme as λ-crystallin, a constitutive structural protein in
rabbit lens. The present crystal structures and amino acid mutagenesis
studies assigned the role of active-site residues: catalytic base for His145 and
substrate binding for Ser124, Cys125, Asn196, and Arg231. Notably, Arg231
participates in substrate binding from the other subunit of the GDH dimer,
indicating the functional significance of the dimeric state. Proper orientation
of the substrate-binding residues for catalysis is likely to be maintained by
an interprotomer hydrogen-bonding network of residues Asn196, Gln199,
and Arg231, suggesting a network-based substrate recognition of GDH.
© 2010 Elsevier Ltd. All rights reserved.
Hyogo 679-5148, Japan
2SR System Biology Research
Group, RIKEN SPring-8
Center, Harima Institute,
1-1-1 Kouto, Sayo-cho,
Sayo-gun, Hyogo 679-5148,
Japan
3Laboratory of Biochemistry,
Gifu Pharmaceutical University,
Mitahora-higashi,
Gifu 502-8585, Japan
Received 31 March 2010;
received in revised form
24 June 2010;
accepted 30 June 2010
Available online
8 July 2010
Keywords: 3-hydroxyacyl-CoA dehydrogenase; coenzyme; induced fit;
X-ray crystallography; site-directed mutagenesis
Edited by R. Huber
Introduction
*Corresponding author. E-mail address:
kunisima@spring8.or.jp.
Abbreviations used: GDH, L-gulonate
3-dehydrogenase; λCRY, λ-crystallin; HAD,
3-hydroxyacyl-CoA dehydrogenase; PDB, Protein Data
Bank; ASA, accessible surface area; WT, wild type.
L-Gulonate 3-dehydrogenase (GDH; EC 1.1.1.45)
is an NAD+-dependent enzyme in the uronate
cycle, an alternative glucose metabolic pathway
that plays essential roles in the biosynthesis of
glucuronide, glycosaminoglycan, and ascorbic acid.
The enzyme oxidizes L-gulonate to 3-dehydro-L-
0022-2836/$ - see front matter © 2010 Elsevier Ltd. All rights reserved.

Structure of L-Gulonate 3-Dehydrogenase
907
Fig. 1. The reaction catalyzed
by GDH.
gulonate (Fig. 1) and is distributed in a variety
of tissues of mammals1–3 and Drosophila melano-
gaster.4 GDH exhibits additional dehydrogenase
activity towards several other organic acids with 3-
hydroxyl groups, such as L-3-hydroxybutyrate and
L-threonate,1,4 and is also called L-β-hydroxyacid
dehydrogenase. Recently, a cDNA sequence for
GDH from rabbit liver5 was demonstrated to be
identical with that of the taxon-specific λ-crystallin
(λCRY), which is a constitutive structural protein
in rabbit lens.6 Thus, in rabbits, GDH is a
bifunctional protein. In humans, however, GDH
is distributed in nonlens tissues, and its expression
level is downregulated in hepatocellular carcinoma
tissues from patients.7 Although the GDHs from
rabbits and humans share a high sequence identity
(84%) and are almost identical in enzymatic
properties and molecular sizes as a homodimer of
36-kDa subunits,5 the rabbit enzyme is remarkably
stable against heat/urea denaturation when com-
pared to the human enzyme.
Mammalian GDH shares a marginal sequence
identity of around 20% with the NAD+-dependent
3-hydroxyacyl-CoA dehydrogenase (HAD), an en-
zyme critical for oxidative fatty-acid metabolism in
mammals.8 To date, crystal structures of HADs from
humans and pigs have been reported.9,10 HAD is
also dimeric, and its subunit consists of two distinct
domains: N-terminal and C-terminal domains that
are responsible for coenzyme binding and subunit
dimerization, respectively. Based on the crystal
structure of the human HAD/NAD+/acetoacetyl-
CoA ternary complex, Barycki et al. proposed a
catalytic mechanism of HAD11 in which His158 was
implicated as a general base to abstract a proton
from the 3-hydroxyl group of substrate, Glu170
neutralized the positive charge on His158 after the
proton abstraction, Ser137 interacted with the
substrate and coenzyme to orient them for catalysis,
and Asn208 stabilized the reaction product together
with Ser137. However, a later study using muta-
genesis and crystallography suggested that Glu170
was not directly involved in the catalysis but was
required for the proper orientation of the catalytic
His158 and for the structural integrity of the human
enzyme.12 The roles of other active-site residues of
HAD, including Ser137 and Asn208, should be
investigated further.
GDHs (His145, Ser124, Glu157, and Asn196, respec-
tively). A site-directed mutagenesis study of the
rabbit GDH has shown that His145, Asn196, and
Ser124 are critical for the catalytic function of this
enzyme.5 This supports an earlier suggestion that
λCRY/GDH belongs to the same structural family
of HADs.6 However, interestingly, replacement of
Glu157 with Gln did not result in apparent kinetic
alterations on the rabbit GDH, in contrast to a
significant decrease in the activity of the human
HAD by a corresponding mutation of Glu170 to
Gln.12 Therefore, further information based on the
crystal structure of GDH would provide better
understanding and characterization of the function-
al nature of this nascent group of enzymes. In this
study, we present the first crystal structures of rabbit
GDH in both apoenzyme form and NADH-bound
holoenzyme form at resolutions of 1.70 and 1.85 Å,
respectively, to clarify the structural difference
between GDH and HAD, as well as to identify the
residues of GDH that are crucial for coenzyme
specificity, substrate recognition, and catalytic
mechanism.
Results and Discussion
Quality of the models
The crystal structure of the rabbit GDH apoen-
zyme was determined at 1.70 Å resolution, yielding
final Rcryst and Rfree factors of 18.3% and 20.6%,
respectively. The asymmetric unit contains a mono-
meric protomer of the enzyme. The final model of
the apoenzyme structure covers amino acid residues
7−316 with well-defined electron densities, while six
N-terminal residues and three C-terminal residues
were not modeled due to structural disorder. A total
of 419 water molecules were included in the final
refinement. In a stereochemistry analysis using the
program PROCHECK,13 no residue is found in
generously allowed or disallowed regions of the
Ramachandran plot, with the exception of Phe193,
which resides neatly in well-defined electron densi-
ties without steric clashes. The crystal structure of
the enzyme in complex with NADH was deter-
mined at 1.85 Å resolution, yielding final Rcryst and
Rfree factors of 16.8% and 20.0%, respectively. The
final model of the NADH-bound holoenzyme
contains a GDH protomer (residues 7−316), an
NADH molecule, and 452 water molecules in the
In spite of the low sequence identity between
GDH and HAD, the catalytically important residues
(His158, Ser137, Glu170, and Asn208) of human
HAD are completely conserved in rabbit and human

908
Structure of L-Gulonate 3-Dehydrogenase
Table 1. Summary of refinement statistics
hydrophobic contacts between their side-chain
carbons within van der Waals distance (Supple-
mentary Fig. 3). Lys63 and Tyr151 are conserved in
rabbit, human, and mouse GDHs (Fig. 3). The C-
terminal domain of rabbit GDH adopts a novel
entirely helical architecture (Fig. 2a) and dominates
subunit dimerization. The first six α-helices (α8−
α12) form a tightly packed bundle with the cognate
α-helices of the other subunit, whereas the terminal
long helix α13 somewhat protrudes from the core
bundle (Fig. 4; Supplementary Fig. 1). The central
helix α8 penetrates the core bundle completely to
reach the N-terminal domain of the other subunit,
which allows remarkable intersubunit interactions
between α9a of one subunit and β8 of the other
subunit. In addition, α11 is uniquely bent by
approximately 45° at Ser256, when compared to
the other straight helices. The two domains are
connected by a linker region (β8−α8 loop, residues
186−194), which adopts a hairpin-like conformation
(Fig. 2a and b). This conformation is stabilized by
eight hydrogen-bonding interactions (Supplemen-
tary Table 1). In addition, 31 residue pairs of van
der Waals interactions also contribute to the
connection of the two domains (Supplementary
Table 2). These residues are well conserved in other
mammalian GDHs (Fig. 3).
Apoenzyme
NADH bound
Space group
Resolution range (Å)
C2
30.0–1.70
(1.78–1.70)
C2
30.0–1.85
(1.93–1.85)
Cell dimensions
a (Å)
71.81
69.08
72.02
69.52
65.10
102.7
b (Å)
c (Å)
65.64
β (o)
102.7
Number of unique reflections
34,455 (4274)
18.3 (25.2)
20.6 (27.6)
22.5
26,578 (3246)
16.8 (20.1)
20.0 (22.2)
18.9
Rcryst (%)a
Rfree (%)b
Wilson B-factor (Å2)
Average B-factor (Å2)c
Ramachandran geometry
Most favored (%)
Allowed (%)
20.0
16.2
93.0
6.6
0.4
0.0
91.9
7.4
0.4
0.4
Generously allowed (%)
Disallowed (%)
rmsd from ideality
Bond length (Å)
Bond angles (o)
Dihedral (o)
0.005
1.2
20.9
0.81
0.006
1.2
20.2
0.83
Improper (o)
a
Rcryst =∑|Fobs(hkl)−Fcalc(hkl)|/∑|Fobs(hkl)|.
b
Rfree is Rcryst calculated for 5% of the data set not included in
the refinement.
c
Protein atoms were used for the calculation.
A structural similarity search using the DALI
server15 was performed between the refined model
of rabbit GDH protomer and the coordinates
available in the Protein Data Bank (PDB). The
search did not provide any overall structural
homologue. Instead, another DALI analysis using
the N-terminal domain of the rabbit GDH as search
query identified a number of oxidoreductases:
human HAD9 (PDB accession code 2HDH; Z-
score=18.6; rmsd=1.6 Å), Aquifex aeolicus prephe-
nate dehydrogenase16 (PDB accession code 2G5C;
Z-score=16.7; rmsd=2.4 Å), Streptococcus pyogenes
Δ1-pyrroline-5-carboxylate reductase17 (PDB acces-
sion code 2AHR; Z-score=14.3; rmsd=2.6 Å), Pseu-
domonas aeruginosa class I acetohydroxy acid
isomeroreductase18 (PDB accession code 1NP3; Z-
score=14.3; rmsd=2.6 Å), and Thermus thermophilus
HB8 3-hydroxy-isobutyrate dehydrogenase19 (PDB
accession code 2CVZ; Z-score=14.0; rmsd=2.5 Å).
These enzymes show a low sequence identity
(b21%) with the rabbit GDH and are characterized
by an N-terminal Rossmann fold and an enzyme-
specific C-terminal domain. Thus, a DALI analysis
using the C-terminal domain of the rabbit GDH did
not provide any significant similarity with known
structures (Z-scores b2.6). When the Cα trace of the
rabbit GDH is compared with that of the human
HAD, the N-terminal domain and the domain
linker region are well superimposed onto each
other, with the exception of the differences in the
length of the α3 helix and in the orientation of the
α2−α3 loop (Fig. 2c). By contrast, the structures of
the C-terminal domains are quite different between
these two enzymes, in good agreement with the fact
that their sequence identity is low especially in the
C-terminal regions (Fig. 3).
asymmetric unit. The nine terminal residues are
disordered as seen in the apoenzyme structure. With
the exception of Asp8 and Phe193, which locate
neatly in well-defined electron densities without
steric clashes, no other residue is found in gener-
ously allowed or disallowed regions of the Rama-
chandran plot. Statistics from crystallographic
analysis are summarized in Table 1. Structural
descriptions herein, if not stated differently, denote
those observed in the 1. 85-Å crystal structure of the
enzyme–NADH complex.
Overall structural architecture
The GDH protomer consists of two distinct
domains: the N-terminal α/β dinucleotide-binding
domain (residues 7−185) and the C-terminal α-
helical domain (residues 195−316) (Fig. 2a). The N-
terminal domain contains eight β-strands (β1−β8)
and seven α-helices (α1−α7), and displays a typical
α/β dinucleotide binding motif, the so-called
Rossmann fold.14 The first six strands of the β-
sheet are in a parallel orientation, and the remain-
ing two strands (β7 and β8) run in the opposite
direction. In the N-terminal domain, the first six
strands are surrounded by exterior α-helices to
form a hydrophobic core. Of the loops connecting
consecutive secondary structures, the α2−α3 and
β6−β7 loops are spatially located in close proximity
to each other, in contrast with their distant rela-
tionship in primary structure. This loop–loop
contact is mediated by the hydrogen-bonding
interactions of Lys63 (backbone O and Nζ) with
Tyr151 (Oη and backbone O, respectively) and 10

Structure of L-Gulonate 3-Dehydrogenase
909
Fig. 2. Structure of rabbit GDH
protomer. (a) Ribbon drawing of
the NADH-bound GDH structure.
The N-terminal domain, the linker
region, and the C-terminal domain
are shown in pink, green, and light
blue, respectively. The NADH mol-
ecule bound is shown as a stick
model. The bent helix α11 is indi-
cated by a red arrow. (b) Close-up
view of the linker region. The
residues in linker form hydrogen
bond (dotted lines) to those in the
N-terminal domain (red), the C-
terminal domain (blue), and the
linker region (black). (c) Stereo
representation of the Cα trace of
GDH (brown) superimposed onto
the structure of human HAD
(cyan). The NADH molecule
bound in GDH and NAD+ and the
acetoacetyl-CoA molecule bound in
HAD are depicted as stick models.
Dimerization interface
This observed crystal-state dimer is consistent with
our previous biochemical study that used a gel-
filtration method, which revealed that the enzyme
functions as a homodimer in solution state.5 The
dimeric structure can be generated by applying a
Interprotomer interactions in the crystal lattice
clearly suggest that the rabbit GDH assembles into
an apparent dimer (Fig. 4; Supplementary Fig. 1).

Fig. 3. Structure-based sequence alignments of rabbit GDH with homologues. The amino acid sequence of the rabbit GDH (accession no. AB359905) was aligned with those of
the human GDH (accession no. AK024041), mouse GDH (accession no. AB359906), and human HAD (accession no. 2515250A). Residues identical with those of the rabbit GDH are
denoted by hyphens. Alignment gaps are indicated by closed circles. The N-terminal mitochondrial targeting signal (nine residues) of HAD is excluded. Secondary structure
elements of the rabbit GDH and the human HAD are depicted above and below their sequences, respectively. Functionally important residues in the rabbit GDH are boxed. The
dimer interface residues are distinguished by colors: red, residues involving both hydrogen bonds and van der Waals interactions (33 residues); green, residues involving only van
der Waals interactions (61 residues).

Structure of L-Gulonate 3-Dehydrogenase
911
Fig. 4. Diagrams of the GDH homodimer showing important intersubunit interactions. The right model is rotated by
90° with respect to the left. One subunit is shown in brown, and the other subunit is shown in gray. Bound NADH
molecules are depicted as stick models. The enlargement (box) on the left is a cylinder drawing of the C-terminal
dimerization domain. The enlargement on the right shows a ribbon drawing of the linker region and nearby helices on the
other subunit. Hydrogen-bonding interactions are indicated by dotted lines.
crystallographic 2-fold symmetry operation to the
protomer in the asymmetric unit. Thus, we define a
prime as representing an affiliation with the other
subunit of the crystal-state dimer with a perfect 2-
fold axis. A molecular surface analysis using the
program Connolly20 shows that 5186 Å2 of accessi-
ble surface area (ASA) per protomer is buried by the
dimer interface. The buried area at the dimer
interface corresponds to 29.6% or 55.9% of ASA for
the total protomer or the C-terminal domain,
respectively, suggesting a tight dimer association.
Subunit dimerization is mediated by numerous
intersubunit interactions, including 22 hydrogen
bonds and 104 residue pairs of van der Waals
interactions (Supplementary Tables 3 and 4).
Subunit dimerization of GDH is mediated by
residues mainly on the C-terminal domain through
interprotomer homodomain (C-domain to C-do-
main) interactions. C-domain helices are arranged
so as to make anti-parallel helical pairs between α8
and α8′, α9 and α9′, and α13 and α13′ (Fig. 4, left).
Interface interactions at the central helical pair
(α8−α8′) are maintained by a hydrogen bond
between Arg197 and Glu205′ and by four residue
pairs of van der Waals interactions between Arg197
and Glu205′/Leu209′, Ala201 and Ala201′, Ile202
and Ile202′. Residues on α8 also interact with those
in the other structural elements of the counterpart
subunit through 2 hydrogen bonds at Arg197 and
Gln199 and 18 residue pairs of van der Waals
interactions. It should be noted that the side chain of
Gln199 forms an interprotomer hydrogen-bonding
network with putative substrate-binding residues
Asn196 and Arg231′, whose possible role will be
discussed in a later section. Although the terminal
helix α13 (residues 292−315) is located away from
the main body of the C-domain core (α8−α12), it
contributes to subunit association through many
interactions with the C-domain residues on the other
subunit. These interactions include 7 hydrogen
bonds and 23 residue pairs of van der Waals
interactions. Other helices and loops in the C-
terminal domain also make interprotomer homo-
domain interactions, including 5 hydrogen bonds
and 34 residue pairs of van der Waals interactions.
Part of the N-terminal domain (β7 and β8) and
the linker region (β8−α8 loop) are additionally
involved in subunit dimerization through inter-
protomer heterodomain interactions, including 7
hydrogen bonds and 25 residue pairs of van der
Waals interactions. Noteworthy, the charged
hydrogen bonds between Asp191 and Arg261′
may stabilize the bent conformation of α11′ (Fig. 4,
right).
Thermostabilization of rabbit GDH
The number of intraprotomer hydrogen bonds
of the rabbit GDH and the human HAD is 358 and
342, respectively. The number of intraprotomer

912
Structure of L-Gulonate 3-Dehydrogenase
Table 2. Contribution of hydrophobicity to thermal
electrostatic interactions of the rabbit GDH and the
human HAD is 8 and 11, respectively. These data
suggest that the contribution of intraprotomer
hydrogen bonding or intraprotomer electrostatic
interactions to the thermal stability of the rabbit
GDH is similar to their contribution to the human
HAD. On the other hand, the numerous interproto-
mer interactions at the dimer interface of the rabbit
GDH, including 22 hydrogen bonds, are in contrast
to much fewer interactions at the human HAD
interface in which only 4 hydrogen bonds are pre-
sent.9 In addition, a unique feature for the rabbit
GDH is the involvement of the linker region (β8−
α8 loop) in dimerization through interprotomer
heterodomain interactions including 4 hydrogen
bonds. In this region of the human HAD, only
Lys200 and Asp226′ make a hydrogen bond,9 which
corresponds to that formed by Lys188 and Glu212′
in the rabbit GDH. Moreover, the ratio of the dimer
interface area to the C-terminal domain surface
area in the rabbit GDH (55.9%) is much higher than
that in the human HAD (32.3%). These structural
comparisons suggest that the dimer association of
the rabbit GDH is much stronger than that of the
human HAD.
Unfortunately, there is no report of an experimen-
tal comparison of the folding stabilities of the rabbit
GDH and the human HAD. In order to estimate
quantitatively the thermal stability of these proteins,
we calculated semiempirically the Gibbs energy
differences (ΔG) upon denaturation from hydro-
phobic interactions (Table 2). The ΔG values for
the rabbit GDH were higher than those for the
human HAD by 65, 992, and 864 kJ/mol in the
monomeric state, dimeric state, and dimer interface,
respectively. These data show that hydrophobic
interactions are stronger in the rabbit GDH than in
the human HAD, especially when the dimer inter-
faces are compared.
On the other hand, within mammalian GDHs, the
human enzyme is reported to be much less stable
against denaturation by heat or urea when compared
to the rabbit enzyme.5 As shown in alignments of
their amino acid sequences (Fig. 3), 7 residues
required for interprotomer hydrogen-bonding inter-
actions (Ser182, Leu187, Asp191, Asp226, Arg261,
Ser265, and Met316) and 12 residues required for
interprotomer hydrophobic interactions (Ser182,
Val184, Leu187, Ile190, Asp191, Asp226, Met259,
Arg261, Lys264, Ser265, Val277, and Met316) in the
rabbit GDH are not conserved in the human enzyme.
Particularly, the replacement of Asp191 and Arg261
with Ala and His, respectively, in the human GDH
may impair the dimer interface between the β8−α8
loop and the α11 helix, since the strong interactions
between Asp191 and Arg261 would contribute
substantially to dimer stability (Fig. 4, right).
Collectively, a much greater degree of interprotomer
interactions (both polar and hydrophobic) in the
rabbit GDH is likely to enhance its thermostability as
compared with the homologous proteins. The higher
stability of the rabbit GDH may be favorable for its
other function as the taxon-specific λCRY, which is a
stability
Monomeric
state
Dimeric
state
Dimer
interface
Protein
ΔG ΔΔG ΔG ΔΔG ΔG ΔΔG
Rabbit GDH (3ADO) 2995
0
7255 1266
0
0
Human HAD (1F14) 2930 −65 6263 −992 402 −864
ΔG and ΔΔG values are expressed in kilojoules per mole. PDB
accession codes used for the calculation are expressed in
parentheses. See Materials and Methods in detail.
constitutive structural protein probably requiring
such extra thermostabilization.
Coenzyme binding mode
In the structure of GDH complexed with NADH,
the coenzyme is bound to the Rossmann fold of the
N-terminal domain with well-defined electron den-
sities (Fig. 5a). The NADH-binding residues 13−18
represent a consensus sequence Gly-X-Gly-X-X-Gly,
which is one of the amino acid fingerprints for NAD
(P) binding.14,21 A Cα superposition between the
apoenzyme model and the NADH-bound model
provided an rmsd of 0.2 Å. Significant deviations
upon NADH binding are observed in the N-
terminal regions of α1 and α2 and in the β6−α6
loop. Especially, the side-chain orientations of
Leu16, Val17, Arg40, Gln41, Glu97, Ser124, and
Cys125 are different between the two forms
(Fig. 5b). Among these residues, the side chains of
Gln41, Glu97, and Ser124 are directly involved in
coenzyme binding. While the Nɛ2 of Gln41 is
hydrogen bonded to the carboxylate of Asp36 in
the apoenzyme, it makes a hydrogen bond with the
3′-OH of the adenine ribose of NADH and interacts
with the pyrophosphate moiety of coenzyme via a
water molecule (Fig. 5b and c). The carboxylate
oxygen atoms Oɛ1 and Oɛ2 of Glu97 are hydrogen
bonded to the Oγ of Ser123 and the backbone N of
Ser124, respectively, in the apoenzyme. In the
NADH-bound structure, the conformation of the
Glu97 side chain is significantly changed so that its
carboxylate forms additional interactions with the
nicotinamide ribose and Cys125 (Figs. 5c and 6). The
protein–coenzyme hydrogen bonds are summarized
in Fig. 5c. These residues, except for Leu16 and
Val17, are conserved in other mammalian GDHs
(Fig. 3). A hydrogen bond is found between the side
chain of Asn148 and the carboxyamide oxygen of
the nicotinamide moiety, which is also recognized
by the protein main chains (Leu16 and Ser20) via a
water molecule. The hydroxyl groups of the
nicotinamide ribose interact with the side chains of
Lys102, Glu97, and Ser124. The pyrophosphate
moiety interacts with residues 14−18. The hydroxyl
groups of the adenine ribose are recognized by the
carboxylate of Asp36, in addition to the above-
mentioned recognition by Gln41. There is no polar
interaction between the adenine moiety and the
protein residues.

Structure of L-Gulonate 3-Dehydrogenase
913
Fig. 5 (legend on next page)

914
Structure of L-Gulonate 3-Dehydrogenase
data on its reaction stereochemistry are available
to date.
In our previous report, the role of Asp36 in
coenzyme specificity was partly understood with its
replacement with arginine in the D36R mutant,
which showed dual coenzyme specificity for both
NAD+ and NADP+.5 To establish the roles of the
newly identified coenzyme-binding residues in the
current crystallographic study, we replaced Gln41
and Glu97 with Asn and Gln, respectively. In
addition, the double-mutant enzyme D36R/Q41N
was prepared to examine cooperation between
Asp36 and Gln41 in coenzyme specificity. The
kinetic alterations by these mutations are summa-
rized in Table 3. The single mutation E97N virtually
abolished enzyme activity, indicating a critical role
for Glu97 in catalysis. On the other hand, the single
mutation Q41N had few effects on the steady-state
kinetics for both L-gulonate and NAD(P)+. How-
ever, the double mutation D36R/Q41N resulted in
an almost complete switch of coenzyme specificity;
the NAD+-linked reaction showed a large decrease
in Vmax/Km value, whereas a significant increase
was observed in the NADP+-linked reaction. The
ratio of Vmax/Km for the NADP+-linked reaction to
the NAD+-linked reaction was 48, which was much
higher than the values for mutations D36R and
Q41N (9.5 and 0.005, respectively). This indicates a
synergistic effect between Asp36 and Gln41 for the
determination of cofactor specificity, through their
cooperative interactions with the hydroxyl groups
of the adenine ribose, although the contribution may
be predominant in Asp36.
Fig. 6. Alternative conformations of Ser124 and
Cys125 in the rabbit GDH. The active-site residues are
superimposed: the apoenzyme form (gray) and the
NADH-bound form (brown). The side-chain conformation
of Glu97 is different between the two forms; its torsion
angles χ1 and χ2 in the NADH-bound form are −175° and
75°, respectively, whereas the respective values are −67°
and −66° in the apoenzyme form. In the NADH-bound
form, the side chains of Ser124 and Cys125 show two
alternate conformations: conformer-1 and conformer-2.
Hydrogen-bonding interactions specifically derived from
conformer-1 and conformer-2 are represented as dotted
lines shown in blue and red, respectively. The remaining
constitutive hydrogen bonds among the nicotinamide
ribose, Glu97, and Cys125 are shown in green dotted lines.
In the apoenzyme model, two hydrogen bonds among
Glu97, Ser123, and Ser124 are indicated by gray dotted
lines.
Bound NADH molecule adopts an extended form
in which the adenine moiety is in anti conformation
and the nicotinamide moiety is in syn conformation.
Both ribose rings have C2′-endo puckering. The
bottom panel in Fig. 5a shows an enlarged view of
the pyridine ring of nicotinamide. Fortunately, a
high resolution of the crystal structure clearly
revealed a boat conformation of the pyridine ring,
indicative of the reduced state of the coenzyme. The
carboxyamide nitrogen of nicotinamide forms a self-
hydrogen bond (2.95 Å) with one of the pyrophos-
phate oxygens (Fig. 5c). Notably, the B-face of the
nicotinamide ring is exposed to the putative sub-
strate-binding pocket, whereas the A-face packs
against nonpolar residues Val17 and Pro146 (Fig.
5b). In addition, the syn conformation of the nicotin-
amide ring is observed in the crystal structures of
human HAD9 and other oxidoreductases22 that
transfer the B-face 4-pro-S hydride ion. Thus, from
a structural point of view, GDH would be a B-side-
specific dehydrogenase, although no biochemical
Induced fit of Ser124 and Cys125 upon
coenzyme binding
Interestingly, in the NADH-bound crystal, the side
chains of Ser124 and neighboring Cys125 show
considerable extra electron densities (Supplementary
Fig. 4), indicating the existence of two alternative side-
chain conformations: conformer-1 and conformer-2.
Such conformers were observed neither in other
residues of the holoenzyme structure nor in all
residues of the apoenzyme structure that was
solved at a slightly higher resolution. The conform-
er-1 Oγ of Ser124, which is similar to that in the
apoenzyme, is hydrogen bonded to the 2′-OH of
the nicotinamide ribose (3.17 Å), but is positioned
away from the catalytic base His145 (Fig. 6). By
contrast, in conformer-2 of Ser124, the Cα−Cβ bond
rotates by about 120° when compared to conform-
er-1, allowing the Oγ atom to locate more closely to
both the 2′-OH of nicotinamide ribose (2.67 Å) and
Fig. 5. Coenzyme binding modes. (a) The final (2Fo −Fc) electron density at 1.85 Å resolution contoured at 2σ for the
NADH molecule in the rabbit GDH/NADH complex. The coenzyme model is shown as a stick model. The bottom panel
shows an enlarged view of the pyridine ring of NADH. (b) Comparison of residues around the coenzyme-binding pocket
in the apoenzyme (gray) and the GDH/NADH complex (brown). The coenzyme and residues are depicted as stick
models. (c) Schematic representation of the NADH binding mode in rabbit GDH and human HAD. A cyan wire NADH
model of the HAD/NADH complex (PDB accession code 1F17) is superimposed on a brown stick NADH model of the
GDH/NADH complex. Potential hydrogen bonds are indicated with broken lines. Shaded boxes represent corresponding
residues in HAD.

Structure of L-Gulonate 3-Dehydrogenase
915
Table 3. Alteration of kinetic parameters for coenzymes and L-gulonate by mutations
D36Ra
Q41N
Value
D36R/Q41N
Value
Ratiob
E97Q
Value
Parameter
WTa
Value
Ratiob
Ratiob
Ratiob
NAD+-linked activity
Km NAD+ (mM)
Km l-gulonate (mM)
Vmax (U/mg)
0.010
0.18
3.1
1.3
14
1.3
1.0
130
78
0.028 0.001
1.3 0.1
5.9 0.2
209
3
2.6 0.4
3.3 0.3
0.65 0.02
0.25
260
19
0.054 0.06
1.6 0.1
5
9
7
0.4
2
0.7
0.2
0.007 0.001 0.002
Vmax/Km NAD+
314
0.003
0.0006
0.13
0.0004
NADP+-linked activity
Km NADP+ (mM)
Km L-gulonate (mM)
Vmax (U/mg)
0.67
2.4
0.36
0.54
0.20
0.3
8
0.86 0.13
14 1.2
0.83 0.02
0.97
1
6
2
2
0.12 0.01
2.6 0.2
1.5 0.1
12
0.1
1
NDc
NDc
NDc
NDc
—
—
—
—
18
1.9
9.5
5
3
Vmax/Km NADP+
17
22
NADP+/NAD+ ratio of Vmax/Km
0.001
9.5
9500
0.005
5
48
48,000
—
—
a
Taken from Ishikura et al.5
b
Ratio of mutant to WT.
c
Activity was not detectable.
the catalytic base. In the neighboring Cys125, the
conformer-1 Sγ locates within a hydrogen-bonding
distance (2.98 Å) of the Oɛ2 of Glu97. However, in
conformer-2 of Cys125, the Cα−Cβ bond rotates
similarly to Ser124, resulting in relocation of the Sγ
atom to occupy a cavity space created by the 1-to-2
conformational change in Ser124. Related cavity
formation by the 1-to-2 conformational change in
Cys125 is compensated for by the recruitment of a
water molecule, Wat640, that forms a new hydro-
gen bond with the Oɛ2 of Glu97 (2.93 Å). Notably,
B-factor values of these residues do not increase
upon NADH binding, indicating an unchanged
degree of flexibility (Table 4). Therefore, coenzyme
binding most likely provides additional conformer-
2 to Ser124 and Cys125, rather than destabilizing
the structure of these residues. In our previous
report, a mutagenesis study suggested a critical role
for Ser124 in the catalytic function of GDH.5
Furthermore, the conformer-2 Oγ of Ser124 is
located within hydrogen-bonding distance of the
3-OH of the modeled substrate L-gulonate, as
described in a later section. Therefore, the side
chain of Ser124 would move to its conformer-2 state
from its conformer-1 state in the enzyme reaction,
which helps subsequent substrate binding.
Since the N-terminal domains of the rabbit GDH
and the human HAD are structurally similar, their
coenzyme binding modes are compared by super-
position (Fig. 5c). Although the conformations of
NADH molecules in the two enzymes are generally
similar, slight differences are observed in the
orientations of the nicotinamide moiety. This is
probably due to a difference in recognition residues
for the carboxyamide moiety of nicotinamide: three
residues Leu16, Ser20, and Asn148 in GDH, and
only Asn161 in HAD.9,11 In addition, the interaction
of Gln41 with the hydroxyl groups of the adenine
ribose in the GDH structure differs from that of
the corresponding residue (Gln46) of HAD. The
coenzyme-binding residues of GDH outnumber
those of HAD. This may be relevant to the higher
cofactor affinity of GDH compared to that of HAD;
the dissociation constants at pH 7.0 for NAD+ and
NADH in GDH are 16 and 0.4 μM, respectively,5
whereas the corresponding values in HAD are 86
and 0.7 μM.9 Furthermore, the coenzyme-induced
conformational change in Ser124 found in this study
was not observed in the studies of human
HAD.9,11,12 In our comparison of the human HAD
crystal structures, including the apoenzyme form
(PDB accession code 1F14), NADH-bound form
Table 4. Occupancy and B-factor of atoms relevant to
NADH binding
NADH bound
Apoenzyme
B-factor
B-factor
Atom
Occupancy
(Å2)
Occupancy
(Å2)
Ser124
Cys125
Glu97
N
1.0
1.0
1.0
1.0
1.0
0.5
12.25
13.61
14.35
15.62
13.91
15.45
1.0
1.0
1.0
1.0
1.0
1.0
13.14
13.46
15.37
16.36
16.63
18.63
Cα
C
O
Cβ
Oγ
conformer-1
Oγ
0.5
13.25
—
—
conformer-2
N
1.0
1.0
1.0
1.0
1.0
0.5
15.63
18.20
15.91
17.78
20.15
26.80
1.0
1.0
1.0
1.0
1.0
1.0
16.37
19.91
19.10
22.78
21.67
32.84
Cα
C
O
Cβ
Sγ
conformer-1
Sγ
0.5
24.42
—
—
conformer-2
N
Cα
C
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
11.15
11.23
12.19
12.07
10.17
11.50
11.93
10.90
13.17
13.71
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
—
18.61
20.29
20.44
22.01
21.50
25.21
24.39
25.30
26.96
—
O
Cβ
Cγ
Cδ
Oɛ1
Oɛ2
NADHa
a
Forty-four atoms were used for the calculation.

916
Structure of L-Gulonate 3-Dehydrogenase
(PDB accession code 1F17), and NAD+-bound form
(PDB accession code 3HAD), the side chain of
Ser137, corresponding to the Ser124 of GDH,
shows only one conformation, which is not similar
to the two conformers seen in the GDH/NADH
complex. By contrast, the orientation of the Ser137
side chain in the human HAD/NAD+/acetoacetyl-
CoA ternary complex (PDB accession code 1F0Y) is
almost identical with that of the conformer-2 Ser124
of GDH, further supporting that conformer-2 is the
catalytically active state of Ser124. Thus, the
proposed conformational change in Ser124 induced
by coenzyme binding may be a unique and
important step in GDH catalysis.
The role of Glu97 in GDH catalysis is intriguing
because it also shows a substantial conformational
change upon coenzyme binding (Fig. 6). Notably, in
the present mutagenesis study, the E97Q mutation
only moderately affects the coenzyme Km, but
results in an increase in substrate Km and a marked
drop in Vmax. This clearly suggests the role of Glu97
as a trigger for the conformational changes in Ser124
and Cys125 in the coenzyme induced-fit mechanism
proposed. The reaction catalyzed by GDH follows
the compulsory ordered mechanism, in which the
coenzyme binds to the enzyme first and leaves last.5
The current findings provide structural evidence on
the mechanistic steps of the ordered binding of the
coenzyme and substrate, although determination of
the ternary complex will be needed to completely
understand the coenzyme induced-fit mechanism.
equivalent histidine residue in the human HAD
(His157) has been assigned as the catalytic base
involved in hydride transfer11 and since Arg231 is
suggested in this study as a key residue that binds
the carboxylate moiety of gulonate. The modeled
substrate molecule resides in the positively charged
active-site pocket of the rabbit GDH dimer. The C-3
atom of L-gulonate is considered as the reactive
center, since the hydride at this position should be
transferred to the C-4 atom of NAD+ during the
enzymatic reaction to produce the product 3-
dehydro-L-gulonate (Fig. 1). Thus, L-gulonate was
modeled into the putative active site so that its C-3 is
3.4 Å away from the C-4 atom of coenzyme. The L-
gulonate C-3 hydroxyl points toward His145 and
Ser124 and is located within potential hydrogen-
bonding distances of these two residues.
Reaction mechanism of GDH involving a
network-based substrate recognition
In our previous mutagenesis study on the rabbit
GDH, the H145Q mutation produced an inactive
enzyme, and the S124A mutation resulted in
significant decreases in substrate affinity and Vmax
value.5 The crucial roles of His145 and Ser124 as
catalytic base and substrate/product-binding resi-
due, respectively, are further confirmed by the
present observation of these residues in the crystal
structure and by the modeling study of the substrate
complex. While the imidazole ring of His145 does
not interact with the side chain of Ser124, it is
positioned to form a hydrogen bond with a
carboxylate oxygen of Glu157 (Fig. 7a). However,
in our previous report, the mutagenesis of E157Q
did not result in a significant alteration of kinetic
constants for the coenzyme and the substrate, except
that it caused a subtle impairment in the heat
stability of the enzyme.5 To clarify the interaction
between His145 and Glu157, we replaced Glu157 of
the rabbit GDH with Asp or Asn, in which one-
carbon-shorter side chains would split potential
hydrogen bonds between these two residues. As a
result, both E157D and E157N mutants were found
to be inactive enzymatically, although their intrinsic
fluorescence was decreased by the addition of NAD
(H), indicative of their ability to bind the coenzyme.
This suggests that Glu157 participates in the proper
orientation of the imidazole ring of the catalytic base
His145 by forming a hydrogen bond. Probably, no
interaction with the His145 side chain is present in
the E157D and E157N mutants, whereas Oɛ1 of the
replaced Gln in the E157Q mutant is still capable of
forming an effective hydrogen bond with the
imidazole ring of His145 to orient the catalytic
base. Thus, a properly oriented His145 seems to
work as the catalytic base without any assistance
from acidic or hydroxyl amino acids in the GDH
reaction. This is similar to the recently modified
catalytic mechanism of human HAD,12 but differs
from that of the catalytic His-Asp dyad for glucose
6-phosphate dehydrogenases23 and malate/lactate
dehydrogenases,24 and from that of the catalytic
Active site and substrate recognition
A putative GDH active site was identified as an
interdomain pocket that is adjacent to the nucleo-
tide-binding site (Fig. 7a). Structural information on
the enzyme–NADH complex facilitated this identi-
fication process. The nicotinamide moiety of NADH
and the catalytically important residues Ser124,
His145, and Asn196 are found in this pocket.
Importantly, the putative active site is located at
the dimerization interface; a large portion of the
binding pocket is maintained by one subunit,
including the regions of the β6−α6 loop, β6, β7,
and α8, whereas α9b′ and the α10′−α11′ loop
complete the pocket in the other subunit. Analysis
of this pocket revealed a highly polar active site
accommodating a small hydrophilic substrate mol-
ecule. Briefly, Ser124, Cys125, His145, Val147, and
Asn148 are found at the base of the pocket, whereas
Asn196, Gln199, Arg231′, and Asn244′ line the walls
of the active site. These nine residues are conserved
in rabbit, human, and mouse GDHs.
To gain further insight into the interactions
between the substrate and GDH, we modeled an
L-gulonate molecule into the putative active site
(Fig. 7b) using structural information from the
active-site residues and the C-4 atom of nicotin-
amide moiety as guide. In particular, the positions of
residues His145 and Arg231 were useful in orienting
the 3-hydroxyl group and the carboxyl group,
respectively, of L-gulonate molecule, since the

Structure of L-Gulonate 3-Dehydrogenase
917
Fig. 7. Putative GDH active site at the domain/subunit interface. Secondary structural elements (ribbons) and
residues from one subunit are shown in brown, and those from the other subunit and NADH are shown in gray. (a)
Diagram of the active-site residues. The residues and the NADH molecule are depicted as stick models. A dotted line
indicates the hydrogen bond between the imidazole ring of His145 and the carboxylate of Glu157. (b) Close-up view of the
modeled L-gulonate molecule (light blue), the surrounding protein residues, and the NADH molecule. Hydrogen-
bonding network among Asn196, Gln199, and Arg231′, as well as possible hydrogen bonds between the substrate and the
protein residues, is shown as black dotted lines. The blue dotted line represents a putative hydride transfer pathway. The
red dotted line represents a putative proton transfer pathway.
His-Ser-Asp triad for enzymes in the new NAD(P)-
dependent oxidoreductases family.25
interprotomer hydrogen-bonding network invol-
ving Asn196, Gln199, and Arg231′ conducts the
passage of substrate into the active site by assisting
with the proper orientation of the side chain of
Arg231′. To confirm this hypothesis, we prepared an
R231M mutant of GDH and found that it was
completely inactive; no enzyme activity was ob-
served from an excess amount (0.2 mg) of homoge-
neous enzyme preparation. These results, together
with the previous mutagenesis study on Asn196,
demonstrate the importance of network-based
substrate recognition in the regulation of GDH
reaction. Since the hydrogen-bonding network
involves residues from both subunits, the functional
It was reported that a mutation at Asn196 of the
rabbit GDH into Asp or Gln also produces almost
inactive enzymes.5 In the current crystal structure,
Asn196 is far (N5 Å) away from His145, and its side
chain Nδ2 is hydrogen bonded to the C-5 and C-6
hydroxyls of the modeled L-gulonate (Fig. 7b).
Notably, we identified a hydrogen-bonding net-
work among the side chains of Asn196 and Gln199
from one subunit and the side chain of Arg231 from
the other subunit. Since the Nη2 of Arg231′ is within
electrostatic interaction distance of the C-1 carboxy-
late of the modeled L-gulonate, it is likely that the

918
Structure of L-Gulonate 3-Dehydrogenase
gulonate, and enzyme in a total volume of 2.0 ml. One unit
of activity was defined as the amount of enzyme that
catalyzes the formation of 1 μ mol/min NAD(P)H at 25 °C.
The kinetic constants for the coenzymes were determined
using a saturating concentration (more than 3-fold of the
Km values) of L-gulonate from experiments performed in
triplicate.
importance of the dimeric state of GDH is empha-
sized further.
The GDH/λCRY family
The 10 functionally important residues are per-
fectly conserved in the rabbit, human, and mouse
GDHs, whereas only five residues (Ser124, His145,
Asn148, Glu157, and Asn196) are conserved in the
human HAD (Fig. 3). These five invariant residues
may play common important roles in the reactions
of both GDH and HAD. However, while GDH is
similar to HAD with respect to their N-terminal
domain structures with catalytic amino acids, they
quite differ from each other in their C-terminal
domain structures, interprotomer interactions, and
mechanisms of substrate recognition. For instance,
the presence of the large C-terminal helix α13 near
the substrate-binding pocket of GDH may explain
why this enzyme does not accept large substrate
molecules,1,4,5 unlike HAD, which has a broad
substrate specificity for CoA derivatives with 4−16
acyl-chain carbons.26 Furthermore, although Asn196
is conserved in the human HAD at position 208, the
hydrogen-bonding network including this residue
for substrate recognition was not observed in the
crystal structure of HAD.11 In addition, the induced-
fit mechanism upon coenzyme binding is observed
only in GDH. Therefore, our findings cast doubt on
the previous assumption that GDH/λCRY and
HAD belong to the same enzyme family.6 Amino
acid sequences of the human, rabbit, and mouse
GDHs show high full-length identities (N76%) with
λCRY homologues from orangutan, dog, pig, cow,
and rat (accession nos. Q5RDZ2, XP_543175,
NP_999046, AF480862, and AY040223, respective-
ly). Most of the coenzyme-binding residues and
active-site residues of rabbit GDH identified in the
current study are conserved in these homologues,
suggesting their identity in the same family. In
conclusion, we propose a new GDH/λCRY family
that is distinct from the conventional HAD super-
family. The knowledge presented here will be the
molecular basis for understanding the proteins in
this new family.
Site-directed mutagenesis, expression, purification,
and crystallization
Mutagenesis was performed using a QuickChange site-
directed mutagenesis kit (Stratagene) and an expression
pRset vector harboring the cDNA for rabbit GDH,5 in
accordance with the protocol described by the manufac-
turer. The mutagenic primers were designed to produce
the mutant enzymes of GDH (D36R, Q41N, E97D, E196N,
and R231M) by replacing the respective codons in the
cDNA with those for the replaced amino acids. The cDNA
for D36R/Q41N was similarly prepared by replacing the
codon for Gln41 in the D36R mutant cDNA with AAC,
using the primers for preparing the Q41N mutant. The
complete coding regions of the cDNAs were sequenced
with a CEQ2000XL DNA sequencer (Beckman Coulter) to
confirm the presence of the desired mutations and to
ensure that no other mutation had occurred. The wild-
type (WT) and mutated cDNAs were expressed in
Escherichia coli BL21(DE3) cells, and the recombinant
enzymes were purified from 20,000g supernatants of cell
homogenates (each from 1 L of culture) by assaying
L-gulonate dehydrogenase activity or by detecting the
36-kDa protein band by SDS-PAGE, as described
previously.5,27 SDS-PAGE of the purified WT and mutants
showed a single protein band corresponding a molecular
mass of 36 kDa (Supplementary Fig. 2). The protein
concentration was determined by the method of Bradford
using bovine serum albumin as protein standard.28
Crystals of the purified recombinant rabbit GDH and its
NADH complex were obtained by the oil microbatch
method, as described previously.29
Structure determination, refinement, and evaluation
For data collection, crystals were directly mounted on
cryoloops from crystallization drops and frozen in cold
nitrogen gas stream at 100 K. The diffraction data for the
apoenzyme were collected using a Rigaku R-AXIS V
image plate detector and synchrotron radiation at beam-
line BL26B130 of SPring-8 (Japan). The diffraction data for
the GDH/NADH complex were collected in-house using
a Rigaku R-AXIS VII image plate detector and a MicroMax
007 generator operating with a copper target. All
measured diffraction spots were indexed, integrated,
and scaled using the HKL2000 program package,31 as
described previously.29
Materials and Methods
Chemicals
The rabbit GDH structure in apo form was solved by a
molecular replacement technique using the program
MOLREP.32 Chain A of HAD from Archaeoglobus fulgidus
(PDB accession code 1ZEJ) was used as search model. To
minimize the model bias caused by the search model, we
applied the initial phase to a phase improvement process
using the prime-and-switch protocol implemented in the
program RESOLVE.33 The resultant electron density map
was of high quality and clearly interpretable. The atomic
model of GDH was built using the graphic program
QUANTA (Accelrys, Inc., San Diego, CA, USA) and
refined using the program CNS, version 1.1.34 Several
cycles of model building and refinement yielded the final
Chemicals were purchased from Sigma-Aldrich Chem-
ical Company, Fluka, and Merck. Crystallization kits were
purchased from Hampton Research. L-Gulonate was
synthesized as described previously.4
Assay of enzyme activity
The dehydrogenase activity of rabbit GDH was assayed
by measuring the rate of change in coenzyme fluorescence
at 455 nm from excitation at 340 nm.5 The standard
reaction mixture consisted of 50 mM 4-morpholinepropa-
nesulfonic acid–NaOH (pH 7.0), 1 mM NAD+, 10 mM L-

Structure of L-Gulonate 3-Dehydrogenase
919
model. The crystal structure of the rabbit GDH/NADH
complex was solved by a difference Fourier technique
based on the apoenzyme structure. A model for the
NADH molecule bound to the enzyme was built into the
difference electron density map, and refinement of the
NADH-bound structure was carried out as described for
the apoenzyme. The final refinement statistics for the two
enzyme forms are given in Table 1. The same set of Rfree
reflections was used for the refinement of both structures.
The stereochemical quality of the final structures was
checked using the program PROCHECK.13
Analysis of the secondary structure was performed with
the program DSSP.35 Fold similarity searches were
performed with the DALI server.14 Superimpositions of
protein models were performed using the program
LSQKAB.36 The ASA of protein molecule was calculated
using the program Connolly.20 The denaturation Gibbs
energy (ΔG) of a protein due to hydrophobic effect was
estimated by the following equation from ASA values of
the protein model:37,38
References
1. Smiley, J. D. & Ashwell, G. (1961). Purification and
properties of β-L-hydroxy acid dehydrogenase: II.
Isolation of β-keto-L-gluconic acid, an intermediate in
L-xylulose biosynthesis. J. Biol. Chem. 236, 357–364.
2. Koundakjian, P. P. & Snoswell, A. M. (1972). 3-
Hydroxy acid dehydrogenases in sheep tissues.
Biochem. J. 127, 449–452.
3. Goode, D., Lewis, M. E. & Crabbe, M. J. C. (1996).
Accumulation of xylitol in the mammalian lens is
related to glucuronate metabolism. FEBS Lett. 395,
174–178.
4. Borack, L. I. & Sofer, W. (1971). Drosophila β-L-
hydroxy acid dehydrogenase. Purification and prop-
erties. J. Biol. Chem. 246, 5345–5350.
5. Ishikura, S., Usami, N., Araki, M. & Hara, A. (2005).
Structural and functional characterization of rabbit
and human L-gulonate 3-dehydrogenase. J. Biochem.
137, 303–314.
6. Mulders, J. W., Hendriks, W., Blankesteijn, W. M.,
Bloemendal, H. & de-Jong, W. W. (1988). λ-Crystallin,
a major rabbit lens protein, is related to hydroxyacyl
coenzyme A dehydrogenases. J. Biol. Chem. 263,
15462–15466.
7. Chen, J., Yu, L., Li, D., Gao, Q., Wang, J., Huang, X.
et al. (2003). Human CRYL1, a novel enzyme-crystallin
overexpressed in liver and kidney and downregulated
in 58% of liver cancer tissues from 60 Chinese patients,
and four new homologs from other mammalians.
Gene, 302, 103–113.
ΔG = 0:178ΔASAnonpolar − 0:013ΔASApolar
where ΔASAnonpolar and ΔASApolar represent the differ-
ences in ASA values of a protein upon denaturation in
nonpolar (C/S) and polar (N/O) atoms, respectively. The
ASA value of the native state was calculated from a native
crystal structure, and that of the denatured state was
calculated from an extended structure that was generated
from the native structure using the program Insight II
(Accelrys, Inc.). Hydrogen bond was defined as a proper
donor–acceptor interaction within a distance range of
2.2 Åbd≤3.4 Å. Electrostatic interaction was defined as an
interaction between a pair of oppositely charged atoms
(Asp/Glu/C-terminus and Arg/Lys/His/N-terminus),
with a distance of not more than 5.0 Å. Van der Waals
interaction was defined as an interatomic interaction with
a distance range of 2.65 Åbd≤4.0 Å. All visualization and
image production were prepared using the program
PyMOL†.
8. Schulz, H. (1991). Beta oxidation of fatty acids.
Biochim. Biophys. Acta, 1081, 109–120.
9. Barycki, J. J., O'Brien, L. K., Bratt, J. M., Zhang, R.,
Sanishvili, R., Strauss, A. W. & Banaszak, L. J. (1999).
Biochemical characterization and crystal structure
determination of human heart short chain L-3-hydro-
xyacyl-CoA dehydrogenase provide insights into
catalytic mechanism. Biochemistry, 38, 5786–5798.
10. Barycki, J. J., O'Brien, L. K., Birktoft, J. J., Strauss, A.
W. & Banaszak, L. J. (1999). Pig heart short chain L-3-
hydroxyacyl-CoA dehydrogenase revisited: sequence
analysis and crystal structure determination. Protein
Sci. 8, 2010–2018.
11. Barycki, J. J., O'Brien, L. K., Strauss, A. W. &
Banaszak, L. J. (2000). Crystallographic and spectro-
scopic evidence for distinct conformations of L-3-
hydroxyacyl CoA dehydrogenase. J. Biol. Chem. 275,
27186–27196.
PDB accession numbers
The atomic coordinates and structure factor amplitudes
of the rabbit GDH in the apoenzyme form and the NADH-
bound form were deposited in the RCSB PDB with
accession codes 3ADO and 3ADP, respectively.
12. Barycki, J. J., O'Brien, L. K., Strauss, A. W. &
Banaszak, L. J. (2001). Glutamate 170 of human L-3-
hydroxyacyl-CoA dehydrogenase is required for
proper orientation of the catalytic histidine and
structural integrity of the enzyme. J. Biol. Chem. 276,
36718–36726.
13. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &
Thornton, J. M. (1993). PROCHECK: a program to
check the stereochemical quality of protein structures.
J. Appl. Crystallogr. 26, 283–291.
Acknowledgements
This work (HTPF50004) was supported by the
“National Project on Protein Structural and Function-
al Analyses” funded by the Ministry of Education,
Culture, Sports, Science, and Technology of Japan.
14. Rossmann, M. G., Moras, D. & Olsen, K. W. (1974).
Chemical and biological evolution of nucleotide-
binding protein. Nature, 250, 194–199.
Supplementary Data
15. Holm, L. & Sander, C. (1993). Protein structure
comparison by alignment of distance matrices.
J. Mol. Biol. 233, 123–138.
16. Sun, W., Singh, S., Zhang, R., Turnbull, J. L. &
Christendat, D. (2006). Crystal structure of prephenate
dehydrogenase from Aquifex aeolicus: insights into the
catalytic mechanism. J. Biol. Chem. 281, 12919–12928.
Supplementary data associated with this article
can be found, in the online version, at doi:10.1016/
j.jmb.2010.06.069
† http://pymol.sourceforge.net

920
Structure of L-Gulonate 3-Dehydrogenase
17. Nocek, B., Chang, C., Li, H., Lezondra, L., Holzle, D.,
Collart, F. & Joachimiak, A. (2005). Crystal structures
of Δ1-pyrroline-5-carboxylate reductase from human
pathogens Neisseria meningitides and Streptococcus
pyogenes. J. Mol. Biol. 354, 91–106.
18. Ahn, H. J., Eom, S. J., Yoon, H. J., Lee, B. I., Cho, H. &
Suh, S. W. (2003). Crystal structure of class I
acetohydroxy acid isomeroreductase from Pseudomo-
nas aeruginosa. J. Mol. Biol. 328, 505–515.
19. Lokanath, N. K., Ohshima, N., Takio, K., Shiromizu, I.,
Kuroishi, C., Okazaki, N. et al. (2005). Crystal
structure of novel NADP-dependent 3-hydroxyisobu-
tyrate dehydrogenase from Thermus thermophilus HB8.
J. Mol. Biol. 352, 905–917.
20. Connolly, M. L. (1993). The molecular surface pack-
age. J. Mol. Graphics, 11, 139–141.
28. Bradford, M. M. (1976). A rapid and sensitive method
for the quantitation of microgram quantities of protein
utilizing the principle of protein–dye binding. Anal.
Biochem. 72, 248–254.
29. Asada, Y., Kuroishi, C., Ukita, Y., Sumii, R., Endo, S.,
Matsunaga, T. et al. (2008). Crystallization and prelim-
inary X-ray crystallographic analysis of rabbit L-gulonate
3-dehydrogenase. Acta Crystallogr. Sect. F, 64, 228–230.
30. Ueno, G., Kanda, H., Hirose, R., Ida, K., Kumasaka, T.
& Yamamoto, M. (2006). RIKEN structural genomics
beamlines at the SPring-8; high throughput protein
crystallography with automated beamline operation.
J. Struct. Funct. Genomics, 7, 15–22.
31. Otwinowski, Z. & Minor, W. (1997). Processing of
X-ray diffraction data collected in oscillation mode.
Methods Enzymol. 276, 307–326.
21. Bellamaciana, C. R. (1996). The nicotinamide nucleo-
tide binding motif: a comparison of nucleotide
binding proteins. FASEB J. 10, 1257–1269.
22. Tanaka, N., Nonaka, T., Nakamura, K. T. & Hara, A.
(2001). SDR: structure, mechanism of action, and
substrate recognition. Curr. Org. Chem. 5, 89–111.
23. Cosgrove, M. S., Naylor, C., Paludan, S., Adams, M. J.
& Levey, H. R. (1998). On the mechanism of the
reaction catalyzed by glucose 6-phosphate dehydro-
genase. Biochemistry, 37, 2759–2767.
24. Goward, C. R. & Nicholls, D. J. (1994). Malate
dehydrogenase: a model for structure, evolution,
and catalysis. Protein Sci. 3, 1883–1888.
25. Goto, M., Muramatsu, H., Mihara, H., Kurihara, T.,
Esaki, N., Omi, R. et al. (2005). Crystal structures
of Δ1-piperideine-2-carboxylate/Δ1-pyrroline-2-
carboxylate reductase belonging to a new family
of NAD(P)H-dependent oxidoreductases. J. Biol.
Chem. 280, 40875–40884.
26. Kobayashi, A., Jiang, L. L. & Hashimoto, T. (1996).
Two mitochondrial 3-hydroxyacyl-CoA dehydro-
genases in bovine liver. J. Biochem. 119, 775–782.
27. Nakanishi, M., Deyashiki, Y., Ohshima, K. & Hara, A.
(1995). Cloning, expression and tissue distribution of
mouse tetrameric carbonyl reductase. Identity with
an adipocyte 27-kDa protein. Eur. J. Biochem. 228,
381–387.
32. Vagin, A. & Teplyakov, A. (1997). MOLREP: an
automated program for molecular replacement.
J. Appl. Crystallogr. 30, 1022–1025.
33. Terwilliger, T. C. (2003). SOLVE and RESOLVE
automated structure solution and density modifica-
tion. Methods Enzymol. 374, 22–37.
34. Brünger, A. T., Adams, P. D., Clore, G. M., DeLano,
W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).
Crystallography and NMR system (CNS): a new
software suite for macromolecular structure determi-
nation. Acta Crystallogr. Sect. D, 54, 905–921.
35. Kabsch, W. & Sander, W. (1983). Dictionary of protein
secondary structure: pattern recognition of hydrogen-
bonded and geometrical features. Biopolymers, 22,
2577–2637.
36. Kabsch, W. (1976). A solution for the best rotation to
relate two sets of vectors. Acta Crystallogr. Sect. A, 32,
922–923.
37. Funahashi, J., Takano, K., Yamagata, Y. & Yutani, K.
(1999). Contribution of amino acid substitutions at
two different interior positions to the conformational
stability of human lysozyme. Protein Eng. 12, 841–850.
38. Tanaka, H., Chinami, M., Mizushima, T., Ogasahara,
K., Ota, M., Tsukihara, T. et al. (2001). X-ray crystalline
structures of pyrrolidone carboxyl peptidase from
a hyperthermophile, Pyrococcus furiosus, and its Cys-
free mutant. J. Biochem. 130, 107–118.