Highly active mutants of carbonyl reductase S1 with inverted coenzyme specificity and production of optically active alcohols

Article abstract of DOI:10.1271/bbb.69.544

A wild type NADPH-dependent carbonyl reductase from Candida magnoliae (reductase S1) has been found not to utilize NADH as a coenzyme. A mutation to exchange the coenzyme specificity in reductase S1 has been designed by computer-aided methods, including three-dimensional structure modeling and in silico screening of enzyme mutants. Site-directed mutagenesis has been used to introduce systematic substitutions of seven or eight amino acid residues onto the adenosine-binding pocket of the enzyme according to rational computational design. The resulting S1 mutants show NADH-dependency and have lost their ability to utilize NADPH as a coenzyme, but retain those catalytic activities. Kinetic parameter V_max and K_m, values of those mutants for NADH are 1/3- to 1/10-fold those of the wild type enzyme for NADPH. As a model system for industrial production of optically active alcohols, the S1 mutants can be applied to an asymmetric reduction of ketones, cooperating with a coenzyme-regeneration system that uses an NAD-dependent formate dehydrogenase.

Full text of DOI:10.1271/bbb.69.544

Biosci. Biotechnol. Biochem., 69 (3), 544–552, 2005
Highly Active Mutants of Carbonyl Reductase S1 with Inverted
Coenzyme Specificity and Production of Optically Active Alcohols
y
Souichi MORIKAWA,1; Takahisa NAKAI,1 Yoshihiko YASOHARA,2
3
Hirokazu NANBA,2 Noriyuki KIZAKI,2 and Junzo HASEGAWA
1Life Science Research Laboratories, Kaneka Corporation, 1-8 Miyamae-machi, Takasago-cho,
Takasago, Hyogo 676-8688, Japan
2Fine Chemicals Research Laboratories, Kaneka Corporation, 1-8 Miyamae-machi, Takasago-cho,
Takasago, Hyogo 676-8688, Japan
3Life Science RD Center, Kaneka Corporation, 1-8 Miyamae-machi, Takasago-cho, Takasago,
Hyogo 676-8688, Japan
Received October 5, 2004; Accepted November 30, 2004
A wild type NADPH-dependent carbonyl reductase
from Candida magnoliae (reductase S1) has been found
not to utilize NADH as a coenzyme. A mutation to
exchange the coenzyme specificity in reductase S1 has
been designed by computer-aided methods, including
three-dimensional structure modeling and in silico
screening of enzyme mutants. Site-directed mutagenesis
has been used to introduce systematic substitutions of
seven or eight amino acid residues onto the adenosine-
binding pocket of the enzyme according to rational
computational design. The resulting S1 mutants show
NADH-dependency and have lost their ability to utilize
NADPH as a coenzyme, but retain those catalytic
activities. Kinetic parameter Vmax and Km values of
those mutants for NADH are 1/3- to 1/10-fold those of
the wild type enzyme for NADPH. As a model system
for industrial production of optically active alcohols, the
S1 mutants can be applied to an asymmetric reduction
of ketones, cooperating with a coenzyme-regeneration
system that uses an NAD-dependent formate dehydro-
genase.
production system for optically active alcohols with a
coenzyme-regenerating system containing glucose de-
hydrogenase (GDH) and glucose to recycle NADP+ to
NADPH (Scheme 1).4) Though GDH can regenerate
both NADH and NADPH, this enzyme produces the
equivalent of gluconic acid as a by-product, and occa-
sionally it causes side-effects to the main reaction,
resulting in the formation of an unfavorable optical
isomer. On the other hand, formate dehydrogenase
(FDH) is also useful as a regenerating enzyme for a
coenzyme, but FDH interacts only with NADH.5)
Therefore, FDH is not able to couple with an
NADPH-dependent carbonyl reductase such as S1.
But, there are several advantages to employing FDH in
an enzymatic reduction system. NADH is less expensive
than NADPH, and the only resulting by-product from
recycling NAD+ to NADH is carbon dioxide, FDH
being inert to the substrate of the main reaction.6) Hence
we therefore attempted to modify the coenzyme depend-
ency in S1 from NADPH-specific to NADH. We report
herein the rational computational design of S1 mutants
and the properties of the resulting mutants, and also
show the synthesis of an optically active alcohol with
the NADH-dependent S1 mutant utilizing the NADH
regeneration system with FDH.
Key words: carbonyl reductase; Candida magnoliae;
coenzyme specificity; protein engineering
A reductase S1 (S1) is the carbonyl reductase isolated
from Candida magnoliae AKU4643. It is applied to an
asymmetric reduction of ethyl 4-chloro-3-oxobutanoate
(COBE) to optically pure ethyl (S)-4-chloro-3-hydroxy-
butanoate ((S)-CHBE).1,2) The wild type S1 is a member
of the short-chain dehydrogenase and reductase (SDR)
family;3) it shows strict NADPH-dependency and cannot
use NADH as a coenzyme. It has been applied to a good
Several successful mutations and redesigns to ex-
change the coenzyme specificity of dehydrogenase or
reductase using nicotinamide coenzyme have been
reported, including glutathione reductase and dihydroli-
poamide dehydrogenase (of the flavoprotein disulphide
oxidoreductase family),7,8) isocitrate dyhydrogenase and
isopropylmalate dehydrogenase (of the decarboxylating
dehydrogenase family),9) lactate dehydrogenase,10) leu-
y
To whom correspondence should be addressed. Fax: +81-794-45-2459; E-mail: Souichi.Morikawa@kn.kaneka.co.jp
Abbreviations: 1D, one-dimensional; 3D, three-dimensional; (S)-CHBE, ethyl (S)-4-chloro-3-hydroxybutanoate; COBE, ethyl 4-chloro-3-
oxobutanoate; CR, carbonyl reductase; DEE, dead-end elimination; FDH, formate dehydrogenase; GDH, glucose dehydrogenase; S1 (or reductase
S1), NADPH-dependent carbonyl reductase from Candida magnoliae; S1Mx, S1M1, S1M2, etc., mutant(s) of reductase S1; SDR, short-chain
dehydrogenase and reductase

Carbonyl Reductase S1 Mutants with Inverted Coenzyme Specificity
545
Scheme 1. Asymmetric Reduction of Ketone (COBE).
cine dehydrogenase,11) and mouse lung carbonyl reduc-
tase (of the SDR family).12) These results were obtained
using empirical design methods derived from detailed
analysis of amino acid sequences (one-dimensional (1D)
structure) and/or atomic coordinates, as determined by
X-ray analysis (three-dimensional (3D) structure). Con-
sidering the results of these empirical design methods,
we have introduced a rational design method on the
basis of the 3D structures of proteins and ligands (see
methods section) to redesign the coenzyme specificity of
the S1 mutant.
Meanwhile, the thermostability of the enzyme should
be considered in order to design the mutants available
for practical use. It was assumed in this study that the
thermostability of the protein can be evaluated by
comparing the difference in the free energy change ꢀG
between the energy GN of the properly folded native
state (N state) and the energy GD of the unfolded
denatured state (D state) (eq. 4).
ꢀG ¼ GN ꢀ GD
ð4Þ
The ꢀG is, in general, a free-energy change in the
molecular system in the folding process of the protein.
To compare the thermostability between a mutant and its
template protein, the template usually being a wild type
protein, it is necessary to compare the difference in the
energy ꢀꢀG between the folding energies, ꢀGmutant and
ꢀGwild-type, of the mutant and wild type proteins
respectively (eq. 5).
Materials and Methods
Summary of rational design. In the catalytic reaction
with an enzyme, the ꢀGbind value of the enzyme and the
substrate (or the coenzyme) molecule is generally
equivalent to the Michaelis–Menten kinetic parameter
Km value of the enzyme for the substrate, according to
the thermodynamic equation (1) in a simple enzymatic
reaction process (2).
ꢀꢀG ¼ ꢀGmutant ꢀ ꢀGwild-type
ð5Þ
The design method for the S1 mutants in this study
involved in silico screening methods,13) and was
composed entirely of semi-automatic computational
processes for (1) generation of mutant protein sequences
from a template (wild type) protein sequence according
to the input condition, (2) modeling of three-dimen-
sional structures (i.e., atomic coordinates of side-chains
of amino acid residues) of those generated mutants with
a dead-end elimination (DEE) algorithm,14) and (3)
evaluation of free-energy differences (ꢀꢀGbind) in the
ligand-binding (here, coenzyme-binding) processes and
evaluation of free-energy differences (ꢀꢀG) in the
folding processes between those mutants, by calculating
roughly the AMBER molecular mechanics potential
energy,15) including the solvation energy term16) in
vacuo, and the change in the conformational entropy
terms17,18) by fixation of the main-chain and side-chains
in the folding process.
ꢀGbind ¼ RT ln Km
ð1Þ
Here, ꢀGbind is the change in free energy in the process
of binding the enzyme and substrate (or cofactor), Km is
the equilibrium constant (dissociation constant) of these
molecules, and R and T are the gas constant and the
temperature respectively.
Km
kcat
E þ S ꢀ ES ! E þ P
ð2Þ
Here, E, S, and P are the enzyme, substrate, and product
molecule respectively, the substrate and product mole-
cules possibly being coenzymes or cofactors. The
Michaelis–Menten kinetic parameters Km and kcat are
the equilibrium constant (the dissociation constant) of E
and S, and the velocity constant, respectively. The
binding affinity of the enzyme mutants for the coenzyme
can then be evaluated by comparing the difference
(ꢀꢀGbind) of each of the ꢀGbind values between the
mutants (ꢀGbind-mutant) and the wild type (template)
protein (ꢀGbind-wild-type) (eq. 3).
Computer programs. The search for related SDR
sequences and the multiple-alignment were performed
by the BLAST program19) and the Clustal-X program20)
respectively. The molecular graphics program Swiss
ꢀꢀGbind ¼ ꢀGbind-mutant ꢀ ꢀGbind-wild-type
ð3Þ

546
S. MORIKAWA et al.
PDB Viewer21) was used in the homology modeling of
S1. Computational screenings were performed with the
Shrike program.13) The conditions of these calculations
are included in the results section below.
free extract of haloketone-resistant FDH22) (780 U),
6.3 ml of 5 M formic acid, 1.0 g of sodium formate, and
25 mg NAD+ in a total volume of 30 ml. To the aqueous
solution, 30 ml of butyl acetate and 6.0 g of COBE were
added, and the mixture was stirred at 30 ꢁC. The pH of
the mixture was adjusted to 6.5 with 5 M formic acid.
CHBE concentrations were monitored by gas chroma-
tography. The optical purity of CHBE was analyzed by
HPLC, as described previously.23)
Site-directed mutagenesis and expression. Substitu-
tions for DNA sequences to produce the S1 mutants
were introduced into plasmid pNTS1, which carries the
S1 gene from Candida magnoliae inserted into
pUCNT.2,4) The recombinant plasmids of containing
mutated cDNAs used in this study were constructed as
follows: 167bp oligonucleotide fragments having the
sequences of designed mutants (abbreviated as S1Mx,
S1M1 to S1M8) were synthesized and inserted into the
3.2-kb DNA fragment, which was prepared by the
digestion of pNTS1 with EcoO109I. The mutated
plasmids pNTS1Mx were introduced into E. coli
HB101. E. coli HB101 carrying pNTS1Mx (pNTS1M1
to pNTS1M8) was inoculated into a test tube containing
2 ml of 2xYT medium4) supplemented with 50 mg/ml
ampicillin, followed by incubation at 37 ꢁC for 15 h.
Cells were harvested from 1.5 ml of culture broth by
centrifugation following suspension in 1.5 ml of 100 mM
potassium phosphate buffer (pH 6.5), and then disrupted
by ultrasonication. The cell debris was removed by
centrifugation. The supernatant was recovered as cell-
free extract for the enzyme assay. The S1M1 and S1M4
mutants were purified by procedure of Wada et al.,1)
which involves ammonium sulfate precipitation,
DEAE-Sepharose column chromatography, and Phen-
yl-Sepharose column chromatography, to homogeneity
on SDS polyacrylamide gel electrophoresis.
Results
Multiple alignments of S1 and related SDRs
Amino acid sequences similar to the S1 sequence
were obtained by the BLAST program applied to the
nonredundant protein sequence database from NCBI.
These sequences described or reported as having
NADPH/NADH dependency were then selected. The
resulting sequences were multiply aligned by the
Clustal-X program, and those with known 3D structures
were manually realigned by superimposition of their 3D
structures on the molecular graphics program Swiss
PDB Viewer. The result of multiple alignments of S1
and the related SDRs is shown in Fig. 1. The regions
bound to the adenosine ribose moiety of the coenzyme
(from Pro-14 to Gly-81 in S1) are shown in Fig. 1.
Homology modeling of the 3D structure of S1
A 3D structure model of S1 was prepared for the
subsequent in silico screening. The 3D structures of
1YBV at the 14–215 residue positions and 1FMC at the
197–249 positions from the Protein Data Bank (PDB)
were used as the template 3D stuructures to model S1.
These corresponded to the 18–227 and 228–279 posi-
tions of S1 respectively. Several insertions, deletions,
and substitutions of amino acid residues were performed
with the program Swiss PDB Viewer. Neither insertions
nor deletions were entered in secondary structures such
as ꢀ-helices and ꢁ-sheets based on the conformational
information. The preferable rotamers (sets of conforma-
tions) in the side-chains were evaluated by the DEE
algorithm method for two enzyme–coenzyme complexes
with NADPH and NADH respectively. Strains were then
removed based on energy minimization calculations in
vacuo with atomic restraints for the main-chain and
nicotinamide heavy atoms. The resulting S1 model is
shown in Fig. 2.
Enzyme assay and kinetic analysis. S1 and the S1Mx
mutants were assayed spectrophotometrically. The
standard reaction mixture, which consisted of 100 mM
potassium phosphate buffer (pH 6.5), 0.167 mM
NAD(P)H, 1 mM ethyl 4-chloroacetoacetate (COBE),
and enzyme solution in a total volume of 3.0 ml, was
monitored for a decrease in absorbance at 340 nm. One
unit of enzyme activity was defined as the amount of
enzyme catalyzing the oxidation of 1 mmol of NAD(P)H
per min at 30 ꢁC. Protein was measured with a protein
assay kit from Bio-Rad, using bovine serum albumin as
the standard. The kinetic constants, Km and Vmax, in the
COBE reduction were calculated from double-reciprocal
Lineweaver–Burk plots with a fixed saturated concen-
tration of substrate or coenzyme. The thermostability of
S1 and the mutants was measured with the residual
catalytic activity in the COBE reduction after heat
treatment of the enzyme solution (pH 7.0, potassium
phosphate buffer) at 30, 35, 40, 45, 50, 55, and 60 ꢁC for
30 min each.
Screening of S1 mutants in silico
Several sets of computational screenings were per-
formed with the Shrike program to search for reductase
mutants having lower coenzyme affinity to NADPH,
higher coenzyme affinity to NADH, and sufficient
thermostability. The amino acid residue positions to
mutate were selected from neighbors of adenosine 20-
phosphate ester (in NADPH) or adenosine 20-OH (in
NADH) of the coenzyme, according to an analysis of
that modeled S1 structure. Here, the conserved motif
Reduction of COBE with S1M4 and formate dehy-
drogenase (FDH) in a two-phase system. The aqueous
solution of the reaction mixture consisted of 0.82 ml of
the cell-free extract of S1M4 (340 U), 3.0 ml of the cell-

Carbonyl Reductase S1 Mutants with Inverted Coenzyme Specificity
547
Fig. 1. Multiple Alignment of the Short-Chain Dehydrogenase/Reductase (SDR) Family.
The regions that bound to the adenosine ribose moiety of the coenzyme are shown in (a) NAD-dependent SDRs and (b) NADP-dependent
SDRs. The residue position number is shown for S1 reductase. The mutated positions in this study are marked with asterisks (*). The conserved
residues GXXXGXG are shown as bold characters. The residues Arg, Lys, and Asp that bound to 20,30-diol of NADH or to 20-phosphate of
NADPH are shown as bold and italic characters.
residues were excluded, and the residues finally selected
to mutate were Ser-41, Ser-42, Ser-43, Tyr-47, Trp-63,
Tyr-64, Asn-65, and Ser-66 of S1. These residues are
positioned far from the binding pocket of the substrate
compound and the reaction center of S1 in 3D space. As
such, mutations of these residues should not influence
the substrate specificity or the catalytic activity of the
enzyme.
First, single-point mutations to the residues were
calculated so as to limit residue types in the subsequent
multiple-point mutation calculation. The residue types to
be mutated to all included 20 natural amino acids in each
position. The results of calculating the binding energy to
the coenzyme and the thermostability of S1 mutants,
which were mutated at residue Tyr-64, are shown in
Table 1. It was assumed that mutation from Tyr-64 to
Ala, Asn, or Asp would confer NADH-dependency to
those S1 reductase mutants. These mutations were
therefore selected and used in the following multiple-
point mutation. The calculated mutation from Tyr-64 to
Cys or Met also showed NADH-dependency, but the
side-chains of these residues have weak reactivity and
lesser robustness to halogeno-compounds such as COBE
and CHBE, and these mutations were therefore not
selected. The mutation from Tyr-64 to Leu or Val
showed NADH-dependency, but the calculated thermo-
stability of these holo-form (NADH complex) mutants
was less than that of the wild type reductase, and these
mutations were therefore not selected. Energy-decom-
position analysis of the calculated mutants produced the
following results: mutation from Tyr-64 to Asn, Asp,
and Cys caused better electrostatic interaction between
the enzyme and the coenzyme, and mutation to Ala,
Met, Leu, and Val caused better van der Waals
interactions within the enzyme itself (better packing of
its hydrophobic core). On the other hand, the calculated
results for the mutations to Tyr-64 showed worse
NADPH-dependency than the wild type, showing
weaker electrostatic and van der Waals interactions
between the enzyme and the coenzyme or within the
enzyme itself. The residue types at other positions were
limited by similar considerations and analysis (data not
shown). Significantly better coenzyme-dependency and
thermostabilty were obtained by the electrostatic and
van der Waals interactions between the enzyme and
coenzyme, the same as the mutation at Tyr-64, as
determined by the decomposed energy terms in the
results of the MM calculations.

548
S. MORIKAWA et al.
were the same as those of the single-point mutations,
and the residue types were selected and limited to
particular types according to the results of the single-
point mutations as: Ser-41 to Ala, Ser-42 to Ala or Arg,
Ser-43 to Gly, Gln, or Arg, Tyr-47 to Arg, Trp-63 to Ile,
Tyr-64 to Asp, Asn-65 to Ile or Val, and Ser-66 to Leu
or Asn respectively, including their wild type residues.
A total of 1,152 mutated sequences and their 3D
structures were generated in the Shrike program. Of
these, 144 mutants showed worse thermostabilty in the
NADH holo-form than the wild type (ꢀꢀG > 100:0
kcal/mol). This destabilization was caused primarily by
the unfavorable van der Waals contacts of the mutated
residues. The remaining 1,008 mutants were extracted
and analyzed as candidates.
A summary of the calculated results is shown in
Fig. 3 as scatter diagrams. Of the 1,008 generated
mutants, 807 with possibly better NADH affinity and
worse NADPH affinity were obtained disregarding their
thermostability, according to the calculated coenzyme-
binding energy (Fig. 3a). Meanwhile, many calculated
mutants showed worse thermostability in NADH-bound
holo-forms (Fig. 3b), and the mutants of only 306
sequences showed better thermostability than the wild
type sequence disregarding their coenzyme specificity.
Consequently, the mutants of 277 sequences showed
both better NADH-dependency and better thermostabil-
ity versus the wild type enzyme. The S1 mutants of 8
sequences (S1Mx) were finally selected from these 277
according to the orders of coenzyme specificity and
thermostability. Their mutated residues are shown in
Table 2. Essentially, mutations to Ser-41, Ser-42, Ser-
43, Trp-63, Tyr-64, and Asn-65 appear to give NADH-
dependency, and mutations to Tyr-47 and Ser-66 appear
to give thermostability in the NADH-bound holo-form,
considering the calculated results.
Fig. 2. Modeled Reductase S1 3-Dimensional Structure.
A complex structure of S1 reductase and NADPH created by
homology modeling from the X-ray structures (see ‘‘Results’’) is
shown as a ribbon diagram. The NADPH molecule is represented as
CPK, and the eight residues to be mutated, position numbers 41–43,
47, and 63–66, are represented as ball-and-stick.
Table 1. Calculated Binding Energy to Coenzyme and Thermo-
stability
binding energy
ꢀꢀGbind (kcal/mol)
thermostability
ꢀꢀG (kcal/mol)
NADH complex
mutated residues
at position 64
NADH
NADPH
Ala
Arg
Asn
Asp
Cys
Gln
Glu
Gly
His
ꢀ21:3
11.9
39.7
60.1
42.2
66.0
18.5
16.2
45.0
39.9
41.7
45.3
58.2
36.6
44.6
41.9
48.4
45.9
17.8
74.0
0.0
ꢀ5:2
ꢀ6:4
0.0
ꢀ18:7
ꢀ11:1
ꢀ8:9
0.3
ꢀ6:9
ꢀ15:2
ꢀ22:7
ꢀ9:2
ꢀ4:4
13.7
Expression and analysis of site-directed mutants
The selected S1 mutants were obtained by site-
directed mutagenesis to the wild type S1. The resulting
activities of these mutated enzymes, not purified but as a
cell-free extraction, are shown in Table 2. All the
designed mutants showed NADH dependency and lost
their ability to utilize NADPH as a coenzyme. The
ability to catalyze the asymmetric reduction of COBE
was also retained in these eight S1 mutants, which gave
(S)-CHBE with an optical purity of >99% e.e. These
catalytic activities, however, were weaker than those of
the wild type enzyme.
Among the mutants obtained, the S1M4 mutant
showed better activity than others, so it was purified
and its enzymatic properties were determined. Mean-
while, the S1M1 mutant was also purified because it was
considered that the Arg-43 residue of S1M4 might cause
or retain the NADPH-dependency of the mutant,
according to the empirical analysis and comparison of
the 1D structures (see Fig. 1), and the effect of the
residue position of 43 was to be examined. The kinetic
ꢀ4:4
7.6
15.4
3.8
Ile
20.9
15.5
Leu
Lys
Met
Phe
Pro
ꢀ17:0
ꢀ3:4
ꢀ15:5
7.3
11.6
ꢀ4:6
ꢀ3:0
25.6
ꢀ6:3
0.4
Ser
ꢀ6:7
ꢀ30:4
12.8
Thr
Trp
Tyr (wt)
Val
ꢀ5:3
ꢀ1:5
0.0
0.0
1.6
ꢀ12:4
46.4
The results of the mutation at residue position 64 of reductase S1 are shown.
The negative ꢀꢀGbind and ꢀꢀG of the mutants mean better binding affinity
and thermostability versus the wild type respectively.
Next, multiple-point mutations were calculated for
these same residue positions to identify the desirable
(i.e., better affinity to NADH) single-point mutations.
The residue positions in the multiple-point mutations

Carbonyl Reductase S1 Mutants with Inverted Coenzyme Specificity
549
Fig. 3. Calculated Properties of Reductase S1 Mutants.
The results of the 8-point mutations are shown as a scatter diagram. Each virtual mutant is plotted as a small dot, the wild type (template) S1 as
a big circle, and the 8 mutants (S1M1 to S1M8) as big triangles respectively. (a) The binding affinity of mutants to NADPH and NADH vs. the
wild type is shown. The negative ꢀꢀGcalc of the mutants means better coenzyme affinity versus the wild type. Mutants plotted further to the left
and higher in this diagram are expected to be NADH specific regardless of their thermostability and chemical stability. (b) The thermostability of
the holo form (NADH) and the binding affinity of mutants to NADH vs. the wild type is shown. The negative ꢀꢀGcalc of the thermostability
means better stability versus the wild type. Mutants plotted further to the left and lower in this diagram are expected to be thermostable and better
NADH binding regardless of their NADPH affinity.
Table 2. Calculated and Experimental Properties of Reductase S1 Mutants
calculated
binding energy
ꢀꢀGbind (kcal/mol)
calculated
experimental
activity
Enzyme
mutated residues
41–43/47/63–66
thermostability
ꢀꢀG (kcal/mol)
NADH complex
for COBE (U/ml)
/coenzyme
NADH
NADPH
S1 (wt)
S1M1
S1M2
S1M3
S1M4
S1M5
S1M6
S1M7
S1M8
SSS/Y/WYNS
AAQ/Y/IDIN
AAQ/Y/IDVL
AAG/Y/IDIL
AAR/Y/IDIN
AAQ/R/IDIN
AAR/R/IDIN
ARS/Y/IDIN
ARS/R/IDIN
0.0
ꢀ18:9
ꢀ14:1
ꢀ25:3
ꢀ36:1
ꢀ24:7
ꢀ23:1
ꢀ22:1
ꢀ15:2
0.0
13.2
23.5
33.1
30.2
36.9
18.9
6.9
0.0
ꢀ9:9
ꢀ3:6
ꢀ5:7
ꢀ4:0
ꢀ9:1
ꢀ9:3
ꢀ14:7
ꢀ2:0
6.35/NADPH
1.11/NADH
0.37/NADH
1.17/NADH
3.02/NADH
0.37/NADH
1.17/NADH
0.51/NADH
0.08/NADH
22.4
The results for the experimentally obtained reductase S1 mutants are shown. The calculated negative ꢀꢀGbind and ꢀꢀG of the mutants mean better binding affinity
and thermostability versus the wild type, respectively. The experimental activity was determined with crude enzyme extracted solution.
Table 3. Kinetic Parameters of Wild Type and Designed Reductase S1 Mutants
activity for
COBE (U/mg)
/coenzyme
Km for coenzyme
(mM)
Km for COBE
(mM)
kcat for coenzyme
(secꢀ1
Enzyme
)
NADPH
NADH
S1 (wt)
S1M1
S1M4
27.8/NADPH
2.8/NADH
9.9/NADH
16.7
N.D.
N.D.
N.D.
160
62
3.0
3.9
2.6
870 (NADPH)
64 (NADH)
120 (NADH)
The Km and the kcat for the coenzyme were determined under the condition of 10 mM COBE as a reactant, and N.D. in the Km column means that the catalytic activity
is not detected with each coenzyme. The Km for COBE was determined under the condition of 5 mM coenzyme.
parameters of S1M1, S1M4, and the wild type S1 are
shown in Table 3. The coenzyme-dependency of these
mutants was inverted (see Km for coenzyme), and their
binding-affinity for the substrate was retained (see Km
for COBE), although the catalytic activity was slightly
weakened (see kcat for coenzyme). The thermostability
of the mutants was not changed versus the wild type.
After heat treatment, they retained their catalytic activity

550
S. MORIKAWA et al.
Fig. 4. Synthesis of (S)-CHBE by Mutated S1 (S1M4).
(a) The reaction scheme of a water–organic two-phase asymmetric reduction system cooperating with the coenzyme regeneration system is
shown. The reactant COBE and the product CHBE are in an organic layer, and the enzymes (S1M4, FDH), coenzyme NADH, and formate are in
a water layer. (b) The reaction profile as the conversion rate of the product CHBE from COBE is shown.
up to 45 ꢁC and then lost it at temperatures over 50 ꢁC
(data not shown).
at that residue position (Tyr-64). Asp at this position
might be critical to NADH-dependency because the
carboxylate of the Asp side-chain can form a ‘‘bi-dental’’
strong electrostatic nonbonded interaction to adenosine
20,30-diol of NADH. In comparison, the anionic side-
chain of Asp repulses the adenosine 20-phosphate anion
of NADPH. S1 does not have these features in its amino
acid sequence, which would confer distinct NADPH- or
NADH-dependency. It is not clear how the S1 enzyme
distinguishes between NADPH and NADH based on
analysis of its 1D structure, but, according to our
comparisons with the 1D structures of SDRs, a mutation
of Tyr-64-Asp onto S1 is considered to be the best
choice to invert the coenzyme-dependency of S1. On the
other hand, the mutation of Thr-38-Asp onto mouse lung
carbonyl reductase (CR), another member of the SDR
family, was studied by Nakanishi et al.,12) and that
mutant showed NADH-dependency. Thr-38 in mouse
lung CR occupies the same position as Tyr-64 in S1 in
the 3D structure, so the mutation Tyr-64-Asp onto S1 is
also expected to invert the coenzyme specificity.
The correlation between the difference in the calcu-
lated NADH-binding energy ꢀꢀGbind and the exper-
imental activity of the S1 mutants was found to be semi-
quantitative (its correlation coefficient R ¼ 0:88) with
the regression analysis (Table 2). Hence it was assumed
that the regained NADH-binding energy ꢀGbind, which
is equivalent to the binding constant (dissociation
constant) Km of the enzyme for NADH (eq. 1), made a
significant contribution to the NADH-dependency to
those S1 mutants. Meanwhile, it was difficult to analyze
the individual influences of those mutated residues on
the properties of the S1 mutants, as these multiple-point
mutations were introduced to the residues adjacent to
Reduction of COBE with S1M4 and formate dehy-
drogenase (FDH) in a two-phase system
The results of the asymmetric reduction of COBE by
S1M4 reductase and NADH in an organic/water two-
phase system in cooperation with the NADH regenerat-
ing system of FDH and formate are shown in Fig. 4. The
catalytic reduction of COBE (162 g/l) with S1M4 gave
CHBE (163 g/l), and the turnover rate of NAD+ based
on the amounts of NAD+ added and CHBE formed was
approximately 1,000 mol/mol. The conversion rate from
COBE to CHBE was >99% after 3.5 h (Fig. 4b), and the
optical purity of the resulting alcohol was >99% e.e. In
this two-phase system, S1M4 showed good tolerance to
the organic solvent as well as the wild type S1.
Discussion
The conserved residues GXXXGXG are known as the
nicotinamide-coenzyme binding motif in the SDR
family,3) and S1 has the same motif (Gly-40, Gly-44,
and Gly-46). In NADPH-dependent SDRs (Fig. 1b), Arg
or Lys that bind to adenosine 20-phosphate of NADPH
with an electrostatic interaction, are almost always
conserved, but S1 has different residues at the corre-
sponding residue positions (Ser-43, Tyr-47, and Asn-
65). Several NADH-dependent SDRs also have Arg or
Lys at these positions, so these cationic residues might
be preferable for, but not critical to, NADPH-depend-
ency. Meanwhile, in NADH-dependent SDRs (Fig. 1a),
Asp, which binds to adenosine 20,30-diol of NADH, is
almost always conserved, but S1 has a different residue

Carbonyl Reductase S1 Mutants with Inverted Coenzyme Specificity
551
Table 4. Calculated Binding Energy to Coenzyme and Thermo-
stability
mutation of Asn-65-Val was estimated to give better
NADH-dependency than Asn-65-Ile under the condition
of the single-point mutation, but the calculated enzy-
matic properties of S1M2 (Asn-65-Val) and S1M1 (Asn-
65-Ile) were inverted in the multiple-point mutation
(Table 2), so that S1M2 might show worse specific
activity than S1M1. The mutations of Ser-66-Asn and
Ser-66-Leu were estimated to give poor coenzyme
specificity but slightly better thermostability, and those
effects were unclear in the multiple-point mutants but
might have resulted in slightly better thermostability in
the holo-form S1Mx. Consequently, the effects of each
of the single-point mutations must accumulate and
balance one another more effectively in the amino acid
sequence of S1M4 than in the other S1Mx.
The mutated residue positions were at a distance from
the substrate-binding pocket of S1Mx, so it was difficult
to explain their weakened catalytic activity. The kcat and
Vmax values of the enzyme for a substrate such as COBE
could not be estimated by the design method reported
here. These mutations might cause some perturbation of
the substrate-binding pocket of the mutated reductases
versus the wild type. As a result, the orientation and
distance between the nicotinamide moiety of the
coenzyme and the spatial configuration of the substrate
molecule for the enzyme would become unsuitable to
the catalytic reduction, possibly resulting in weaker
activity. However, the results of the CHBE synthesis
from COBE (Fig. 4b) showed that S1M4 has high
catalytic activity and availability to use the NADH
coenzyme and is applicable to the asymmetric reduction
of ketones with the NADH coenzyme, as intended. The
rational design method in this work must be applicable
to exchanging the coenzyme specificity or substrate
specificity of other enzymes.
residue
wild-type mutated
calculated
binding energy
ꢀꢀGbind (kcal/mol)
calculated
thermostability
ꢀꢀG (kcal/mol)
NADH complex
NADH
NADPH
Ser-41
Ser-42
Ala
Ala
Arg
Gln
Gly
Arg
Arg
Ile
ꢀ22:9
ꢀ11:9
ꢀ3:4
19.9
16.4
52.9
82.2
33.1
52.2
35.7
52.4
58.2
66.0
44.2
39.0
18.0
34.1
ꢀ8:8
ꢀ13:6
ꢀ4:5
16.4
Ser-43
2.8
0.8
25.4
16.3
Tyr-47
Trp-63
Tyr-64
Asn-65
20.7
ꢀ10:9
9.0
ꢀ12:1
ꢀ11:1
ꢀ4:4
ꢀ13:8
7.0
Asp
Ile
ꢀ6:9
14.1
Val
Asn
Leu
13.1
Ser-66
ꢀ5:2
ꢀ7:2
10.8
The results of the one-point mutation at residue positions 41, 42, 43, 47, 63,
64, 65, and 66 of reductase S1 (see also Table 2) are shown. The negative
ꢀꢀGbind and ꢀꢀG of the mutants means better binding affinity and
thermostability versus the wild type, respectively.
each other. As such, the simple energy-decomposition
analysis gave unclear and complicated results (data not
shown). The individual effects of each mutated residue
in the S1 mutants were estimated with the results of
calculated single-point mutations at those residues. The
summarized results are shown in Table 4.
The common mutations of Ser-41-Ala, Trp-63-Ile,
and Tyr-64-Asp to S1Mx were estimated to give better
NADH-dependency, and those mutants were estimated
to give lower nonbonded interaction energies (electro-
static, van der Waals, and solvation energy terms) based
on the energy-decomposition analysis. The shape of
their coenzyme-binding pocket might therefore be fitted
to an NADH molecule, and it was considered that these
mutations might result in substantial effects on the
inverted coenzyme specificity. The mutation of Tyr-47-
Arg was estimated to give better thermostability but
worse NADH-dependency. As a result, S1M5, S1M6,
and S1M8 showed weaker activity with NADH as a
coenzyme than did S1M1, S1M4, and S1M7 respective-
ly. Both mutations of Ser-42-Ala and of Ser-42-Arg
were estimated to give better NADH-dependency and
thermostability. The former, however, was more effec-
tive in its enzymatic properties than the latter, so that
S1M7 and S1M8 might show weaker activity for COBE
than the other S1Mx. The mutations of Ser-43-Gln, Ser-
43-Gly, and Ser-43-Arg were estimated to give some-
what worse NADH affinity under the condition of the
single-point mutation, but they were estimated to give
even worse NADPH affinity, which might affect the
NADH-specificity. The mutation of Ser-43-Arg was
estimated to give rather better NADH affinity than Ser-
43-Gln, so that S1M4 might show better activity than
S1M1, although the analysis of the 1D structures of the
SDRs showed that the mutation of Ser-43-Arg appeared
to be preferable to NADPH-dependency (Fig. 1b). The
References
1) Wada, M., Kataoka, M., Kawabata, H., Yasohara, Y.,
Kizaki, N., Hasegawa, J., and Shimizu, S., Purification
and characterization of NADPH-dependent carbonyl
reductase, involved in stereoselective reduction of ethyl
4-chloro-3-oxobutanoate, from Candida magnoliae. Bio-
sci. Biotechnol. Biochem., 62, 280–285 (1998).
2) Yasohohara, Y., Kizaki, N., Hasegawa, J., Wada, M.,
Kataoka, M., and Shimizu, S., Molecular cloning and
overexpression of the gene encoding an NADPH-
dependent carbonyl reductase from Candida magnoliae,
involved in stereoselective reduction of ethyl 4-chloro-3-
oxobutanoate. Biosci. Biotechnol. Biochem., 64, 1430–
1436 (2000).
3) Persson, B., Krook, M., and Joernvall, H., Character-
istics of short-chain alcohol dehydrogenases and related
enzymes. Eur. J. Biochem., 200, 537–543 (1991).
4) Kizaki, N., Yasohara, Y., Hasegawa, J., Wada, M.,
Kataoka, M., and Shimizu, S., Synthesis of optically
pure ethyl (S)-4-chloro-3-hydroxybutanoate by Esche-
richia coli transformant cells coexpressing the carbonyl
reductase and glucose dehydrogenase genes. Appl.
Microbiol. Biotechnol., 55, 590–595 (2001).

552
S. MORIKAWA et al.
5) Kataoka, M., Kita, K., Wada, M., Yasohara, Y.,
The dead-end elimination theorem and its use in protein
side-chain positioning. Nature, 356, 539–542 (1992).
15) Weiner, S. J., Kollman, P. A., Nguyen, D. T., and Case,
D. A., An all atom force field for simulation for proteins
and nucleic acids. J. Comput. Chem., 7, 230–252 (1986).
16) Stouten, P. F. W., Froemmel, C., Nakamura, H., and
Sander, C., An effective solvation term based on atomic
occupancies for use in protein simulations. Mol. Simu-
lation, 10, 97–120 (1993).
Hasegawa, J., and Shimizu, S., Novel bioreduction
system for the production of chiral alcohols. Appl.
Microbiol. Biotechnol., 62, 437–445 (2003).
6) Nanba, H., Takaoka, Y., and Hasegawa, J., Purification
and characterization of formate dehydrogenase from
Ancylobacter aquaticus strain KNK607M, and cloning
of the gene. Biosci. Biotechnol. Biochem., 67, 720–728
(2003).
7) Scrutton, N. S., Berry, A., and Perham, N., Redesign of
the coenzyme specificity of a dehydrogenase by protein
engineering. Nature, 343, 38–43 (1990).
8) Bocanegra, J. A., Scrutton, N. S., and Perhama, R. N.,
Creation of an NADP-dependent pyruvate dehydrogen-
ase multienzyme complex by protein engineering. Bio-
chemistry, 32, 2737–2740 (1993).
17) D’Aquino, J., Gomez, J., Hilser, V. J., Lee, K. H., and
Amzel, L. M., Magnitude of the backbone conforma-
tional entropy change in protein folding. Proteins, 25,
143–156 (1996).
18) Doig, A. J., and Sternberg, M. J. E., Side chain
conformational entropy in protein. Protein Science, 4,
2247–2251 (1995).
9) Chen, R., Greer, A., and Dean, A. M., A highly active
decarboxylating dehydrogenase with rationally inverted
coenzyme specificity. Proc. Natl. Acad. Sci. U.S.A., 92,
11666–11670 (1995).
10) Holmberg, N., Ryde, U., and Buelow, L., Redesign of
the coenzyme specificity in L-lactate dehydrogenase
from Bacillus stearothermophilus using site-directed
mutagenesis and media engineering. Protein Engineer-
ing, 12, 851–856 (1999).
11) Galkin, A., Kulakova, L., Ohshima, T., Esaki, N., and
Soda, K., Construction of a new leucine dehydrogenase
with preferred specificity for NADP+ by site-directed
mutagenesis of the strictly NAD+ specific enzyme.
Protein Engineering, 10, 687–690 (1997).
12) Nakanishi, M., Matsuura, K., Kaibe, H., Tanaka, N.,
Nonaka, T., Mitsui, Y., and Hara, A., Switch of
coenzyme specificity of mouse lung carbonyl reductase
by substitution of threonine 38 with aspartic acid. J. Biol.
Chem., 272, 2218–2222 (1997).
19) Altschul, S. F., Gish, W., Miller, W., Myers, E. W., and
Lipman, D. J., Basic local alignment search tool. J. Mol.
Biol., 215, 403–410 (1990).
20) Thompson, J. D., Gibson, T. J., Plewniak, F.,
Jeanmougin, F., and Higgins, D. G., The ClustalX
windows interface: Flexible strategies for multiple
sequence alignment aided by quality analysis tools.
Nucleic Acids Res., 24, 4876–4882 (1997).
21) Guex, N., and Peitsch, M. C., SWISS-MODEL and the
Swiss-PDB Viewer: An environment for comparative
protein modeling. Electrophoresis, 18, 2714–2723
(1997).
22) Nanba, H., Takaoka, Y., and Hasegawa, J., Purification
and characterization of an ꢀ-haloketone-resistant
formate dehydrogenase from Thiobacillus sp. strain
KNK65MA, and cloning of the gene. Biosci. Biotechnol.
Biochem., 67, 2145–2153 (2003).
23) Yasohara, Y., Kizaki, N., Hasegawa, J., Takahashi, S.,
Wada, M., Kataoka, M., and Shimizu, S., Synthesis of
optically active ethyl 4-chloro-3-hydroxybutanoate by
microbial reduction. Appl. Microbiol. Biotechnol., 51,
847–851 (1999).
13) Morikawa, S., and Nakai, T., Japan Kokai Tokkyo Koho,
2001-184381 (July 6, 2001).
14) Desmet, J., De Maeyer, M., Hazes, B., and Lasters, I.,