Engineering the promiscuous racemase activity of an arylmalonate decarboxylase

Article abstract of DOI:10.1002/chem.201001924

Variant G74C of arylmalonate decarboxylase (AMDase) from Bordatella bronchoseptica has a unique racemising activity towards profens. By protein engineering, variant G74C/V43A with a 20-fold shift towards promiscuous racemisation was obtained, based on a reduced activity in the decarboxylation reaction and a two-fold increase in the racemisation activity. The mutant showed an extended substrate range, with a 30-fold increase in the reaction rate towards ketoprofen. Molecular dynamics simulations and the substrate profile of the racemase indicate that the steric and polar effects of the substrate structure play a more dominant role on catalysis than mere kinetic α-proton acidity. The observation that the conversion of β,γ-unsaturated carboxylic acids does not lead to a rearrangement to form their α,β isomers indicates a concerted rather than a stepwise mechanism. Interestingly, a substrate bearing a nitro group instead of the carboxylic acid group on the α-carbon atom was also converted by the racemase.

Full text of DOI:10.1002/chem.201001924

FULL PAPER
DOI: 10.1002/chem.201001924
Engineering the Promiscuous Racemase Activity of an
Arylmalonate Decarboxylase
Robert Kourist, Yusuke Miyauchi, Daisuke Uemura, and Kenji Miyamoto*[a]
Abstract: Variant G74C of arylmalo-
nate decarboxylase (AMDase) from
Bordatella bronchoseptica has a unique
racemising activity towards profens. By
protein engineering, variant G74C/
V43A with a 20-fold shift towards pro-
miscuous racemisation was obtained,
based on a reduced activity in the de-
carboxylation reaction and a two-fold
increase in the racemisation activity.
The mutant showed an extended sub-
strate range, with a 30-fold increase in
the reaction rate towards ketoprofen.
Molecular dynamics simulations and
the substrate profile of the racemase
indicate that the steric and polar effects
of the substrate structure play a more
dominant role on catalysis than mere
kinetic a-proton acidity. The observa-
tion that the conversion of b,g-unsatu-
rated carboxylic acids does not lead to
a rearrangement to form their a,b iso-
mers indicates a concerted rather than
a stepwise mechanism. Interestingly, a
substrate bearing a nitro group instead
of the carboxylic acid group on the a-
carbon atom was also converted by the
racemase.
Keywords: enzymatic promiscuity ·
enzyme catalysis · protein engineer-
ing · racemases
Introduction
the enantiopreference of the enzyme.[5,6] Cofactor-indepen-
dent amino acid racemases are essential enzymes of patho-
AHCTUNGTRENNUNG
Enzymes that catalyse more than one chemical transforma-
tion with differences in the type of formed or cleaved bond
and in the catalytic mechanism are called promiscuous.[1,2]
The introduction of new enzyme activities by protein engi-
neering allows insights into the molecular determinants of
catalysis and how new enzymatic functions may have
emerged from a molecular scaffold during natural divergent
evolution. It also gives leads for the generation of enzymes
for new, unnatural reactions. Recent examples include the
conversion of a haeme-free bromoperoxidase into a lipase[3]
and of an esterase into an epoxide hydrolase.[4] Cys-decar-
boxylases and Cys-based cofactor-independent racemases
share a common aspartate transcarbamoylase fold. The en-
zymatic decarboxylation proceeds through a two-step mech-
anism (Scheme 1a). Cleavage of the pro-(R) carboxylic
group, which is the rate-determining step, is caused by its
destabilisation in the so-called hydrophobic pocket.[5] The
following protonation mediated by a cysteine is decisive for
genic organisms and, thus, potential targets for the treat-
ment of infectious diseases.[7] Consequently, their molecular
mechanism has received much attention in the last few
years. Recent calculations for glutamate racemase (GluR)[8]
and proline racemase (ProR)[7] suggest that the deprotona-
tion and protonation is done simultaneously by two cys-
teines situated in an opposite orientation in the active site
(Scheme 1b, path b2). The introduction of a second Cys74
residue into arylmalonate decarboxylase (AMDase) from
Bordatella bronchoseptica conferred a promiscuous racemis-
ing activity towards 2-arylaliphatic acids on it while the de-
carboxylating activity was retained.[6]
The application of racemases in organic synthesis has
been mainly focused on the preparation of amino acids and
the use of mandelate racemase, which is confined to sub-
strates bearing an a-hydroxy group.[9] Although the enzy-
matic racemisation of arylaliphatic acetyl-CoA-thioesters
has been known for a long time,[10] AMDase variant G74C
has a unique racemising activity towards the free 2-arylpro-
pionic acids. These so-called “profens” play an important
role in the pharmaceutical industry as non-steroidal anti-in-
flammatory drugs.[11] Chemical racemisation at high temper-
atures and under strongly basic conditions[12] for the recy-
cling of unwanted enantiomers in the synthesis of enantio-
pure 2-arylpropionic acids has disadvantages that are of in-
creasing significance. Enzymatic racemisation offers a green,
[a] Dr. R. Kourist, Y. Miyauchi, Prof. D. Uemura, Prof. K. Miyamoto
Department of Biosciences and Informatics
Keio University, 3-14-1 Hiyoshi, 2238522 Yokohama (Japan)
Fax : (+81)45-566-1783
E-mail: kmiyamoto@bio.keio.ac.jp
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem201001924.
Chem. Eur. J. 2011, 17, 557 – 563
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
557

Results and Discussion
Molecular dynamics (MD) sim-
ulations can shed light on the
accommodation of the substrate
in the active centre and the
extent to which steric restric-
tions might affect the catalysis.
The recently solved structures
of AMDase in its empty form
and with an inhibitor[5,14] and of
AMDase mutant G74C (Pro-
tein Data Bank (PDB) entry
3IXM[15]) show that residues 74
and 188 are situated in analo-
gous positions to those of the
catalytic machinery of Cys-
based racemases. The dianionic
transition state is stabilised by a
network of hydrogen bonds to
Thr75, Ser76, Tyr126 and
Gly189 in the so-called dioxy-
Scheme 1. Schematic representation of a) decarboxylation and b) racemisation catalysed by AMDase variants.
The hydrophobic pocket is highlighted green, the dioxyanion hole is coloured blue. For Cys-racemases, a con-
certed mechanism (path b2) has been suggested.
AHCTUNGTRENNUNG
a
sustainable alternative. However, the activity of AMDase
G74C towards sterically demanding profens is very limited.
Protein engineering is an efficient tool to widen the sub-
strate range of enzymes.[13] The residues of the hydrophobic
pocket have a strong influence on the decarboxylating activ-
ity of the enzyme and were chosen as targets for a potential
increase of the racemising activity.[14] For glutamate race-
mase (GluR), it has been suggested[8] that co-catalytic resi-
dues increase the nucleophilicity of the catalytic Cys resi-
dues by hydrogen bonding in order to facilitate the abstrac-
tion of the a proton. In AMDase G74C, the kinetic a-
proton acidity is increased by a tight hydrogen-bond net-
work in the so-called dioxyanion hole.[7] We were interested
in the extent to which nucleophile activation might occur in
AMDase G74C and whether the catalytic activity might also
be increased by the introduction of co-catalytic residues into
the active site. Finally, we were interested in whether the
racemase and its variants would accept compounds other
than profens, for instance, b,g-unsaturated carboxylic acids
and substrates bearing a different electron-withdrawing
group (EWG) on the chiral centre.
Cys188 residue is bent away from the substrate. In the struc-
ture of mutant G74C/C188S (PDB entry 3IXL),[15] residues
Cys74 and Ser188 are oriented towards inhibitor 1a in the
active site. The structure in PDB entry 3IXL was in silico
back-mutated to G74C, minimised and used as a starting
structure for MD simulations.
The a-carbon atom of the substrate lies almost exactly on
the line joining the two sulfur atoms at distances of only 3.5
and 3.7 ꢁ to Cys74 and Cys188, respectively. Important cri-
teria for productive binding are the stability of the hydrogen
bonds in the dioxyACTHNUTRGNEUNGanion hole and the distances from the Cys
residues to the a proton (d1, Table 1) and the a-carbon
atom (d2, Table 1), respectively. This is important for access
of the nucleophile to the chiral centre (Figure 1). The angle
À
À
g (S Ca S) of the line joining the two sulfur atoms with the
Ca atom should be close to 1808 for an initiation of the
transition state, in which the tetrahedral geometry of the Ca
atom has to change to a planar configuration with a straight
À À À À
S H Ca H S axis. In the case of achiral substrates, each of
the two Ha atoms can be abstracted. Compounds 1b, 10a,
13–16b, 1e and 1g (see Scheme 2) were modelled into the
active site in a similar manner to 1a. In MD simulations
over 500 ps each, the stability differed considerably, with the
most stable structure being that with 13b (Figure 1,
Table 1). In all of the simulations, no indication was found
for an interaction of the catalytic cysteines with neighbour-
ing basic residues, as suggested by Puig et al. for glutamate
racemase.[8] Interestingly, in all cases, the donating cysteine
(Cys188 with the S enantiomer, Cys74 with the R enantio-
mer) formed a hydrogen bond with the carboxylic group of
the substrate. The role of this hydrogen bond for the mecha-
Herein, we report the generation of AMDase variant
G74C/V43A with an up to 30-fold increased activity towards
profens and report the substrate ranges of AMDase G74C
and the new variant towards a wider variation of substrates,
including aryl, pyridyl and vinylaliphatic carboxylic acids
and nitro compounds.
558
www.chemeurj.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 557 – 563

Engineered Racemase Activity of a Decarboxylase
FULL PAPER
with GluR,[8] positions Val13, Pro14 and Gly190 were identi-
fied as the corresponding sites in AMDase for nucleophile
activation. Variants bearing Asp, Glu and His residues in
these positions were generated by site-directed mutagenesis.
However, G74C/V13D, G74C/P14D, G74C/G190H, G74C/
G190E and G74C/V13D/G190H proved to be inactive. Pre-
sumably, these mutations led to a disturbance of the com-
plex hydrogen-bond system in the active site. This also con-
firms the observation that introduction of serine in position
188 in the variant G74C/C188S lowers the decarboxylating
activity of AMDase relative to that of G74C and that man-
delic acid is a poor substrate for the racemase G74C;[6] in
both cases, the catalytic activity of AMDase is reduced by
the introduction of hydrogen-bond donors into the catalytic
centre.
The so-called hydrophobic pocket formed by Val43 and
Met159 contributes to the destabilisation of the leaving car-
boxyl group in the decarboxylation reaction. Molecular dy-
namics studies of 1b in the active site of AMDase G74C
hinted that Val43 (with 4 ꢁ distance) and Met159 (with
3.9 ꢁ distance) are very closely situated to the smaller sub-
stituent of the substrate and, thus, are potential determi-
nants of the racemising activity. Several mutants bearing ex-
changed amino acids in positions 43 and 159 were generated
by site-specific mutagenesis. Exchange of M159 by leucine
resulted in a variant with increased activity, whereas Ala in
this position led to a decrease (Table 2). G74C/V43 variants
Table 1. Molecular dynamics simulations of arylaliphatic carboxylic acids
in the active site of AMDase G74C. The distances from the Cys residues
to the a-proton (d1) and the a-carbon atom (d2) are indicated as bold
À
À
dashes; the angle g of the axis S Ca S and the H-bonds of the dioxyan-
ion hole are indicated by dashes
Substrate
Average
g [8]
Productive frames
[%][a]
Relative activity
[%][b]
1a [pro-(R)]
(R)-1b
(S)-1b
10a
(R)-13b
(S)-13b
(R)/(S)-14b
(R)/(S)-15b
(R)-16b
(S)-16b
(R)/(S)-1e
(R)/(S)-1g
158
133
145
164
164
154
82/57[a]
140
100
5
5
24
82/1[a]
85
19
0
0
54
1
767
n.d.[c]
n.d.[c]
155
58
2
0
135
n.d.[c]
n.d.[c]
0
0
0
0
[a] States that allow the initiation of the reaction with stable H-bonds in
the dioxyanion hole, d1,d2<4.8 ꢁ and g>1508. For achiral substrates 1a
and 10a, two possible reactions leading to abstraction of the pro-(R)/pro-
(S) hydrogen atoms are considered. [b] Relative activity of AMDase
G74C determined by enzymatic deuteration. [c] n.d.: not determined.
The substrate turns in the active site and cannot be accessed by the cata-
lytic machinery.
Table 2. Racemising activity of G74C variants with a modified hydropho-
bic pocket.
Mutant
G74C
G74C/V43A 1.1
G74C/V43G n.c.[b]
G74C/V43N n.c.
Activity[a] [UmgÀ1
]
Mutant
Activity[a] [UmgÀ1
]
n.c.[b]
0.05
0.6
G74C/V43Q
G74C/V43T
G74C/M159A 0.08
G74C/M159L 1.1
[a] Determined in the enzymatic deuteration of (R,S)-1b. [b] n.c.: no con-
version.
bearing a glycine (V43G) or a large polar residue (Asn,
Gln) lost activity in both the racemisation and decarboxyla-
tion reactions (Table 2). Threonine in this position also
yielded only residual activity (Table 2). Interestingly, variant
G74C/V43A had an almost two-fold increased relative effi-
ciency in the racemisation of 1b but only 8% relative effi-
ciency in the decarboxylation reaction (Table 3). The race-
misation closely resembles the second step of the enzymatic
decarboxylation. In G74C/V43A, the rate of the first step of
the decarboxylation reaction is reduced and that of the
second is increased. Both changes were kcat based. The 20-
fold shift from the native decarboxylation reaction to the
promiscuous racemisation (Table 3) confirms the important
role of the hydrophobic pocket in the decarboxylation.
These findings are in agreement with a previous study
where saturation mutagenesis of position 43 failed to pro-
duce enzyme variants with increased decarboxylating activi-
ty.[14] The combination of both variants in G74C/V43A/
Figure 1. a) (S)-13b (blue) modelled in the active site of AMDase G74C
after 500 ps of molecular dynamic simulations. Hydrogen bonds in the di-
oxyanion hole and between Cys188 and the substrate are indicated as
yellow dashes. The two nucleophilic cysteine residues, the dioxyanion
and Val43 (red) are highlighted as sticks. b) The vinyl moiety of 10a
(orange) can rotate freely in the active site and the hydrogen atoms
remain accessible for the catalytic Cys residues. c and d) Steric restric-
tions induce non-productive accommodation of (S)-14b (cyan; c) and
(S)-15b (green; d), which in turn leads to a lower enzyme activity. (For
substrate structures, see Scheme 2.)
nism is not clear. It can contribute to stabilisation of the do-
nated proton close to the Ca atom, which leads to initiation
of the reaction. Based on sequence and structure alignments
Chem. Eur. J. 2011, 17, 557 – 563
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
559

K. Miyamoto et al.
Table 3. Reaction specificity of AMDase variants in the decarboxylation (dec) or racemisation (rac) reactions.
confirmed that the deuteration
of (R)-1b proceeds with com-
plete racemisation. Interesting-
ly, 1a, which has previously
been found to be an inhibitor
of wild-type AMDase in the de-
carboxylation reaction,[15] is also
a substrate of the promiscuous
racemase. Several factors were
identified to play a crucial role
in the catalysis:
Mutant
Decarboxylation[a]
Racemisation[b]
Relative
rac/
kcat
KM
kcat/KM
relative
value [%]
kcat
KM
kcat/KM
relative
value [%]
dec
values
[sÀ1
]
[mm] [mmÀ1 sÀ1
]
[sÀ1
]
[mm] [mmÀ1 sÀ1
]
G74C
G74C/
V43A
G74C/
M159L
0.27
0.022 11.4
10.4
0.026
0.0019
100
8
0.6
1.1
30.2
29.7
0.020
0.037
100
185
1
23
0.063 18.3
0.0034
14
1.0
30.0
0.033
165
12
[a] Decarboxylation of phenylmalonic acid. [b] Racemisation of (R)-1b, as determined by chiral HPLC.
Table 4. Activity of AMDase variants G74C and G74C/V43A in the
deuteration of aryl, pyridyl and vinylcarboxylic acids.
M159L leads to a reduced racemising activity, which indi-
cates that the variant effect is not additive. G74C and
G74C/V43A differ in their substrate recognition (Figure 2,
Substrate
G74C
G74C/V43A
activity
[UmgÀ1
relative value activity relative value
[a]
[a]
]
[%]
[UmgÀ1
]
[%]
1b
1a
2a
3a
4a
5a
6a
7a
8a
0.6
100
140
162
117
23
343
217
317
350
33
1.1
1.3
1.5
1.2
183
217
250
199
35
417
600
233
350
67
0.84
0.97
0.70
0.14[b]
2.06
1.3
1.9
2.1
0.2
0.21[b]
2.5
3.6
1.4
2.1
0.4
9a
10a
11a
1c
1d
1e
0.03
2.6
5
433
8
0
0
0.07
4.2
12
700
40
0.7
0
7
0.05
n.c.
n.c.
0.04
n.c.
4.6
0.35
0.01
n.c.
n.c.
n.c.
n.c.
n.c.
3.7[c]
0.24
0.004
n.c.
0.04
n.c.
12.6
1.1
1 f
7
0
12b
13b
14b
15b
16b
1g
1h
1i
1j
0
767
58
1.7
0
0
0
0
0
616
2100
183
50
0
Figure 2. Activities of AMDase G74C (black) and AMDase G74C/V43 A
(white) and the relative increase in the activity of G74C/V43A (grey) to-
wards pharmacologically important profens. The relative activity has
been determined by comparison to the conversion of 1b by G74C
(0.6 UmgÀ1; Table 2).
0.3
n.c.
n.c.
n.c.
n.c.
n.c.
8.6[c]
0
0
0
0
Table 4). The increased activity of the latter towards 1c and
1d can be explained by the larger space in the hydrophobic
pocket. We were pleased to find that large profens are con-
verted with improved activity. The 30-fold increase in activi-
ty towards ketoprofen (15b) is striking and shows the better
ability of the variant to provide space for these large sub-
strates. The increase in reaction rate improves the synthetic
utility of the enzyme because it allows the use of shorter re-
action times and lower catalyst loadings.
Isotope-exchange experiments allow a real-time determi-
nation of racemase activity towards a wide range of sub-
strates. Their use eliminates the need to establish conditions
for chiral analytics for every substrate or to obtain enantio-
pure material, both of which are difficult in many cases. The
method also allows the determination of the activity towards
non-chiral substrates.[16] The substrate range of AMDase
mutants G74C and G74C/V43A was determined by monitor-
ing the exchange of the a proton by deuterium with the use
of NMR spectroscopy (Scheme 2, Table 4). HPLC analysis
1k
1433
[a] Determined by enzymatic deuteration. [b] Corrected for the spontane-
ous racemisation with t(1/2) =100 min. [c] Corrected for the spontaneous
racemisation with t(1/2) =200 min.
1) Functional group: In the chemical base-catalysed racemi-
sation, esters and particularly thioesters are favoured
over carboxylic acids due to their higher kinetic proton
acidity.[17] In the MD simulations, neither ketone 1g nor
ester 1i could be accommodated in the so-called dioxy-
AHCTUNGTREGaNNUN nion hole, which is essential for carbanion stabilisation.
Consequently, AMDase G74C is inactive towards 1g, 1i
and 1j. Interestingly, nitro compound 1k was converted
with an enzymatic activity significantly higher than the
non-enzymatic background reaction (t(1/2) =200 min).
Compound 1k is isoelectronic to 1a. In the case of
enoate reductases, it is well known that a nitro group can
substitute the carboxylic acid group for activation of the
560
www.chemeurj.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 557 – 563

Engineered Racemase Activity of a Decarboxylase
FULL PAPER
4-toluyl (in 5a) or 4-nitro-
phenyl (in 6a) substituent
are rather small, despite the
different effects of the sub-
stituent on the acidity. How-
ever, the most striking ex-
amples are with very large
substrates. Table 1 lists the
results of MD simulations of
several substrates with
L
substituents of different
sizes. The MD runs show
that the large naphthyl
group of 13b can be accom-
modated without any steric
restrictions. It has a higher
stability than 1b. This ac-
counts, in combination with
the high level of carbanion
Scheme 2. AMDase G74C catalysed deuteration of aryl, pyridyl and vinylaliphatic carboxylic acids and nitro
compounds. S and L denote the small and large substituents; EWG denotes an electron-withdrawing group.
HEPES: 2-[4-(2-hydroxyethyl)-1-piperazinyl]-ethanesulfonic acid.
Ca atom.[18] To the best of our knowledge, this is the first
time that this has also been shown for cofactor-free race-
mases.
stabilisation, for the excellent activity of the enzyme to-
wards 13b. Substrates 14b and 15b have also been antici-
pated to have a high kinetic a-proton acidity. However,
in the MD simulations they did not fit into the active site
(Figure 1c). Steric restrictions are the cause of the re-
duced catalytic efficiencies of 58 and 2% of AMDase
G74C towards 14b and 15b, respectively. A similar case
has been observed for mandelate racemase towards sub-
strates bearing a naphthyl group.[20] The carboxyl group
of 10a is anchored by hydrogen bonding in the dioxyan-
ion hole. Its small vinyl moiety can freely move and even
turn by 1808C (Figure 1b), whereas the aromatic sub-
stituents of the other substrates are fixed into one posi-
tion (Figure 1a). Nevertheless, despite the high mobility
of 10a, the Ha proton is accessible, which leads to 82%
of productive states throughout the MD simulation. The
low activity of 5% is most likely the result of the lower
ability of the vinylic group to stabilise the carbanion. As
a final point, it is difficult to elucidate why the racemase
does not convert ibuprofen (16b). Also, the analogue a-
methyl-a-(4-isobutylphenyl)malonic acid is not a sub-
strate of wild-type AMDase (data not shown). In con-
trast to both enantiomers of 14b and 15b, 16b apparent-
ly fits nicely into the active site of AMDase (Table 1). In
docking experiments 16b adopted an identical binding
configuration to that of 1a. (R)-16b remained stable
during the MD simulations and obviously has an Ha
acidity very similar to that of 1b, which makes the lack
of activity of AMDase G74C very confusing. The free ro-
tation of the voluminous and dynamic isobutyl group
might lead to a strong steric hindrance. Differences in
substrate recognition between 16b and 1b have recently
also been reported for variants of lipase A from Candida
antarctica, which underlines that isobutylphenyl- and
phenyl-substituted propionic acids can be very different
substrates for enzymes.[21]
2) Role of the smaller substituent S: The low activity to-
wards 1 f can be attributed to decreased acidity of the
Ha atom. Substrates bearing groups larger than a methyl
or hydroxy group are affected by steric restrictions. Also,
9a is converted with clearly decreased activity, which in-
dicates steric hindrance at the Ca atom. Surprisingly, the
difference between 1a and 1b is rather small when it is
considered that that the wild-type AMDase decarboxy-
lates a-methyl-a-phenylmalonic acid with 20-fold re-
duced activity compared to the reaction with phenylma-
lonic acid.[19] The fact that G74C/V43A has considerably
increased activity towards 1c and 1d relative to the ac-
tivity of G74C confirms that unfavourable steric interac-
tions with residues of the hydrophobic pockets limit the
size of S. Compound 1e, bearing a voluminous S sub-
stituent, is not converted by any of the variants.
3) Role of the larger substituent L: On the one hand, L is a
crucial factor governing the ability of resonance stabilisa-
tion of the dianionic transition state.[20] This is nicely con-
firmed by the low activity towards 10a and the higher
relative activities towards phenylic 1a, phenylvinylic 11a,
and naphthylic 13b, activities that increase in accordance
with the higher ability of carbanion stabilisation. The
lower Ca acidity of 12b bearing an O-aryl substituent is
reflected in a lack of enzymatic activity. On the other
hand, L can cause steric restrictions that lead to an in-
creased enzyme activity. The individual contributions of
kinetic a-proton acidity and steric interactions differ
from case to case. If analogue compounds are consid-
ered, kinetic a-proton acidity generally plays a less im-
portant role than the substrate structure. For instance, 4a
undergoes a measurable deuteration under the reaction
conditions (t1/2 =100 min), which indicates high kinetic a-
proton acidity. Nevertheless, the enzymatic conversion is
slower than that of its (stable) analogues 2a and 3a.
Also, the differences in activity with a 4-pyridyl (in 2a),
4) Chemoselectivity of AMDase G74C: It was anticipated
that the racemisation of b,g-unsaturated carboxylic acids
Chem. Eur. J. 2011, 17, 557 – 563
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
561

K. Miyamoto et al.
would lead to an intramolecular rearrangement to form
the a,b-unsaturated isomers. Nevertheless, conversion of
10a and 11a exclusively afforded the a,a-dideuterated
products. Compound 10a is stable in the presence of
active site. As a final point, the high chemoselectivity in the
conversion of b,g-unsaturated carboxylic acids indicates a
concerted mechanism of AMDase G74C.
diisopropylACHTUNGTRENNUNGamine at pD 12. Upon treatment with 30%
NaOD in D2O, it undergoes complete rearrangement
into crotonic acid in 1 h. This is the preferred species due
to its conjugated double-bond system. In contrast, no re-
arrangement to crotonic acid was observed with NMR
spectroscopy during the G74C/V43A-catalysed conver-
sion of 10a with up to 80% conversion in 48 h. The high
chemoselectivity of the racemase indicates that the re-
deuteration of the Ca atom by the second cysteine pro-
ceeds much faster than a possible deuteration at the C4
atom. This is the case in a concerted mechanism, in
which the subtraction of the proton proceeds simultane-
ously with the deuteration (Scheme 1, path b2). Unsatu-
rated alkanoic acids have recently attracted attention
due to their possible application for the production of bi-
ofuels.[22] Moreover, novel processes for the generation
of biofuels involve b,g-unsaturated carboxylic acids as in-
termediates.[23] Herein, we report a method for the
highly selective generation of their a,a-dideuterated de-
rivatives, which might be useful for mechanistic studies.
The synthetic value of the enzymatic deuteration was
shown when the G74C/V43A-catalysed deuteration af-
forded 2,2-d-11a with excellent d purity (47 mg, 93%
yield, 98% content of d isomer).
Experimental Section
Molecular modelling: Molecular dynamic studies were performed in a
periodic waterbox by using the YASARA software (version 7.4.22) and
an AMBER99 force-field with long-range electrostatics with a cut-off at
7.86 J (Particle Mesh Ewald). The force-field of the substrate was ob-
tained by using AutoSMILES force-field parameter assignment. After ad-
dition of the solvent, cell neutralization and prediction of pKa values,
simulations of 500 ps were performed at 358C, pH 7 and a solvent density
of 0.997 gLÀ1. Each simulation step contained a 1.25 fs step for inter- and
intramolecular forces. The structure in PDB entry 3IXL was in silico
back-mutated to G74C by manual removal of mutation C188S and the
structure was minimised.[15] This structure represents a closed conforma-
tion containing phenylacetic acid (1a) in the active site. It was preferred
to the structure in PDB entry 3IXM[15] because residue Cys188 is moved
3.8 ꢁ away from the active centre in the latter structure. The other sub-
strates were generated manually and minimised prior to molecular dy-
namics simulations. Docking studies were done by using Accelrys Discov-
ery Studio 2.1 software (Accelrys Inc, San Diego, USA). Images were
generated by using Pymol software (DeLano Scientific, San Carlos,
USA).
Expression and purification of AMDase mutants: Escherichia coli
TOP10 cells were used as hosts for the transformation of the pBAD plas-
mid bearing the codon-optimised gene of AMDase as a fusion protein
with a C-terminal His tag.[24] The strains were grown in Luria–Bertani
(LB) liquid media or on LB agar plates supplemented with 100 mgmLÀ1
ampicillin at 378C until an optical density of 0.5 was achieved. AMDase
production was then induced by addition of arabinose (final concentra-
tion 0.025% v/v). After 14 h of further incubation at 308C, cells were har-
vested by centrifugation (15 min, 48C, 8000 g) and washed twice with
tris(hydroxymethyl)aminomethane (Tris) buffer (50 mL, 50 mm, pH 8.0).
Cells were resuspended in Tris buffer (20 mL) containing 10 mm imida-
zole and DNAse and were disrupted by sonification with cooling over
ice. Cell debris was removed by centrifugation (15 min, 48C, 8000 g).
AMDase variants were purified by His-tag purification by using an Ni se-
pharose high-performance column (50 mL, GE Healthcare, Uppsala,
Sweden) according to the instructions of the manufacturer. Protein elu-
tion was performed by using Tris buffer (50 mm, pH 8.0) with an imida-
zole gradient from 10 to 250 mm imidazole. AMDase variants eluted at
imidazole concentrations of around 90 mm. Proteins were also analysed
with a 12% separating sodium dodecylsulfate polyacrylamide gel. After
electrophoresis, the gel was stained with Coomassie brilliant blue. The
protein concentration was determined by using the Biorad assay accord-
ing to the instructions in the manual. Fractions containing AMDase were
pooled and concentrated by using centricons. Deuterated enzyme solu-
tion was prepared by washing the protein in centricons with deuterated
HEPES buffer (10 mm, pH 8.0).
In conclusion, protein engineering, molecular dynamics
simulations and the comparison of different substrates indi-
cate that steric and polar effects of the substrate structure
play a more dominant role on catalysis than mere kinetic a-
proton acidity. This is particularly illustrated in the examples
of analogue substrates 2–4a and 7–9a and profens 1b and
13–15b. It can be argued that a very stable accommodation
of the carboxylic group and an optimal configuration of the
Cys dyad towards the substrate are needed to decrease the
energy barrier of the transition state sufficiently for cataly-
sis. This is more important than the difficulty of abstracting
the a proton.
Conclusion
Protein engineering gave rise to a racemase variant with up
to 30-fold improved activity in the racemisation of bulky
profens. The 20-fold shift of the activity of the variant to-
wards promiscuous racemisation underlines the role of the
hydrophobic pocket as a determinant of the decarboxylating
and racemising activity in AMDase.
We have shown for the first time that a range of 2-aryl,
heteroaryl and vinylacetic acid derivatives are well-accepted
substrates for G74C racemase and its variants. Interestingly,
the racemase accepts nitro compounds as substrates. The a-
proton acidity of a substrate is less important for catalysis
than efficient accommodation of the transition state in the
Site-directed mutagenesis: Complementary primers bearing the nucleo-
tides to be changed and the vector pBAD-AMDase encoding for the
codon-optimised gene of AMDase were used for the polymerase chain
reaction (PCR) with Pyrococcus furiosus (Pfu) DNA polymerase and the
following reaction conditions: 1) 958C, 30 s; 2) 18 cycles: 958C, 50 s, then
54–608C, 60 s, then 728C, 360 s. The PCR mixture was treated afterwards
with DpnI restriction endonuclease to digest non-methylated template
DNA. The mutated plasmids were transformed into competent E. coli
Top10 cells. Mutant G74C/V43A/M159L was generated by using muta-
genic primers bearing the mutation V43A on a plasmid bearing the muta-
tion M159L. For the mutagenic primers, please refer to the Supporting
Information. Mutant G74C/V13D/G190H was generated by using a meg-
aprimer[25] generated from a forward primer bearing mutant V13D and a
reverse primer bearing mutant G190H.
562
www.chemeurj.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Chem. Eur. J. 2011, 17, 557 – 563

Engineered Racemase Activity of a Decarboxylase
FULL PAPER
NMR spectroscopy assay for enzymatic activity: Purified enzyme solution
(2 mL, 2.5 mgmLÀ1, 50 mL) in deuterium oxide was added to substrate so-
lution (15 mm) in deuterated HEPES buffer (10 mm, pH 8.0) to a final
volume of 800 mL in an NMR spectroscopy tube. Chiral substrates were
used as racemates. The reaction was monitored in the NMR machine at
378C by following the decrease of the signal of the a-hydrogen atom or
the doublet signal of the a-methyl group. Control experiments confirmed
that no spontaneous racemisation occurred. Only in the case of 4a, did
the enzymatic reaction rate have to be corrected for non-enzymatic race-
misation with t(1/2) =100 min.
[4] H. Jochens, K. Stiba, C. Savile, R. Fujii, J. G. Yu, T. Gerassenkov,
R. J. Kazlauskas, U. T. Bornscheuer, Angew. Chem. 2009, 121, 3584–
3587; Angew. Chem. Int. Ed. 2009, 48, 3532–3535.
[5] K. Okrasa, C. Levy, B. Hauer, N. Baudendistel, D. Leys, J. Mickle-
field, Chem. Eur. J. 2008, 14, 6609–6613.
[6] Y. Terao, K. Miyamoto, H. Ohta, Chem. Commun. 2006, 3600–3602.
[7] A. Rubinstein, D. T. Major, J. Am. Chem. Soc. 2009, 131, 8513–
8521.
[8] E. Puig, E. Mixcoha, M. Garcia-Viloca, A. Gonzalez-Lafont, J. M.
Lluch, J. Am. Chem. Soc. 2009, 131, 3509–3521.
[9] U. Felfer, M. Goriup, M. E. Koegl, U. Wagner, B. Larissegger-
Schnell, K. Faber, W. Kroutil, Adv. Synth. Catal. 2005, 347, 951–961.
[10] D. Ouazia, S. L. Bearne, Anal. Biochem. 2010, 398, 45–51.
[11] A. M. Evans, Eur. J. Clin. Pharmacol. 1992, 42, 237–256.
[12] E. J. Ebbers, G. J. A. Ariaans, J. P. M. Houbiers, A. Bruggink, B.
Zwanenburg, Tetrahedron 1997, 53, 9417–9476.
[13] R. J. Kazlauskas, U. T. Bornscheuer, Nat. Chem. Biol. 2009, 5, 526–
529.
[14] K. Okrasa, C. Levy, M. Wilding, M. Goodall, N. Baudendistel, B.
Hauer, D. Leys, J. Micklefield, Angew. Chem. 2009, 121, 7827–7830;
Angew. Chem. Int. Ed. 2009, 48, 7691–7694.
[15] R. Obata, M. Nakasako, Biochemistry 2010, 49, 1963–1969.
[16] U. M. Babu, R. B. Johnston, L. C. McNeff, Anal. Biochem. 1975, 63,
208–213.
[17] P. J. Um, D. G. Drueckhammer, J. Am. Chem. Soc. 1998, 120, 5605–
5610.
Enzymatic deuteration of 11a: Purified solution of AMDase G74C/
V43A (2 mL, 2.3 mgmLÀ1) was added to a solution of 11a (50 mg,
0.30 mmol) in deuterated HEPES buffer (20 mL, 10 mm, pD 8.0), and the
reaction mixture was shaken at 378C for 48 h. The reaction was quenched
by addition of DCl (5m) and the mixture was extracted three times with
diethyl ether. The combined organic layers were dried over MgSO4 and
the solvents were removed under reduced pressure to furnish the crude
product, which was further purified by column chromatography (silica
gel with n-hexane/ethyl acetate as the mobile phase) to yield 2,2-d-11a
(46.8 mg, 0.28 mmol, 93% yield, 98% d purity according to 1H NMR
spectroscopy). The spectroscopic data were satisfactory: 1H NMR
(CDCl3): d=9.9 (brs, 1H), 7.23 (m, 2H), 6.7 (d, J=15.9 Hz, 1H),
6.2 ppm (d, J=15.9 Hz, 1H); 13C NMR (CDCl3): d=178.2 (CH3), 136.6
(Ar C), 134.0 (CH), 128.8, 127.9, 126.3 (Ar C), 120.7 (CH), 37.5 ppm (q,
J=19.9 Hz, CD2).
[18] R. Stuermer, B. Hauer, M. Hall, K. Faber, Curr. Opin. Chem. Biol.
2007, 11, 203–213.
[19] K. Miyamoto, H. Ohta, Eur. J. Biochem. 1992, 210, 475–481.
[20] U. Felfer, U. T. Strauss, W. Kroutil, W. M. F. Fabian, K. Faber, J.
Mol. Catal. B: Enzym. 2001, 15, 213–222.
[21] K. Engstrçm, J. Nyhlen, A. G. Sandstrçm, J. E. Backvꢂll, J. Am.
Chem. Soc. 2010, 132, 7038–7042.
[22] R. Palkovits, Angew. Chem. 2010, 122, 4434–4436; Angew. Chem.
Int. Ed. 2010, 49, 4336–4338.
[23] J.-P. Lange, R. Price, P. M. Ayoub, J. Louis, L. Petrus, L. Clarke, H.
Gosselink, Angew. Chem. 2010, 122, 4581–4585; Angew. Chem. Int.
Ed. 2010, 49, 4479–4483.
Acknowledgements
R.K. gratefully acknowledges the Japanese Society for the Promotion of
Science (JSPS) for a stipend (P-09010). The authors would like to thank
Dr. Osamu Ohno from the Department of Chemistry, Keio University,
Yokohama, for inspiring scientific discussions and support in the isotope-
exchange experiments.
[24] H. Aygꢃn, S. Wojczewski, M. Kircher, S. Rosmus, US 2010/0144010,
2005.
[25] K. Miyazaki, M. Takenouchi, Biotechniques 2002, 33, 1033–1038.
[1] U. T. Bornscheuer, R. J. Kazlauskas, Angew. Chem. 2004, 116, 6156–
6165; Angew. Chem. Int. Ed. 2004, 43, 6032–6040.
[2] K. Hult, P. Berglund, Trends Biotechnol. 2007, 25, 231–238.
[3] B. Chen, Z. Cai, W. Wu, Y. L. Huang, J. Pleiss, Z. L. Lin, Biochem-
istry 2009, 48, 11496–11504.
Received: July 8, 2010
Published online: November 9, 2010
Chem. Eur. J. 2011, 17, 557 – 563
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chemeurj.org
563