Structure-guided directed evolution of alkenyl and arylmalonate decarboxylases

Full text of DOI:10.1002/anie.200904112

Angewandte
Chemie
DOI: 10.1002/anie.200904112
Enzyme Mechanisms
Structure-Guided Directed Evolution of Alkenyl and Arylmalonate
Decarboxylases
Krzysztof Okrasa, Colin Levy, Matthew Wilding, Mark Goodall, Nina Baudendistel,
Bernhard Hauer, David Leys,* and Jason Micklefield*
The asymmetric decarboxylation of prochiral a-disubstituted
malonic acids is an attractive route to produce homochiral a-
substituted carboxylic acids. Various metal-catalysed and
organocatalytic methods have been explored for related
enantioselective decarboxylative protonation (EDP) reac-
tions.[1] However, the synthetic EDP methods can only deliver
chiral carboxylic acids or esters with moderate enantioselec-
tivities (ee).[1] On the other hand, enzymatic decarboxylation
offers
a cleaner, more sustainable and more efficient
approach to deliver carboxylic acids of very high ee. Notably,
the arylmalonate decarboxylase (AMDase) isolated from
Bordetella bronchiseptica[2] has been shown to be highly
effective in the decarboxylation of a-arylmalonates to give
enantiomerically pure a-arylcarboxylic acids (Scheme 1).[3,4]
The fact that the AMDase is highly robust and does not
require co-factors further increases the potential of this
enzyme for biocatalysis. Despite this, the pool of substrates
accepted by the enzyme is limited to malonates that possess a-
aryl substituents. Indeed
a number of other bacterial
AMDases have been discovered from sequence similarity
searches,[5] or selective enrichment experiments.[6] While all
these enzymes could catalyse the decarboxylation of a-
arylmalonates, none have so far been demonstrated to
decarboxylate malonates which do not possess a-aryl sub-
stituents.
In order to rationalise this, and guide the development of
new enzymes which could catalyse a wider range of reactions,
we solved the X-ray crystal structure of the B. bronchiseptica
AMDase.[5] Our first high-resolution structure[7] reveals a
“dioxyanion hole” motif, which can donate six hydrogen
bonds to potentially stabilize a putative high-energy enedio-
late intermediate, along with a small hydrophobic cavity likely
favouring the formation of the neutral CO2. From this we
Scheme 1. Proposed mechanism of the AMDase.[5] Preferred malonate
substrates for the AMDase possess a larger substituent (L=aryl or
alkenyl), which binds to a solvent-accessible cavity, and a smaller
substituent (S=H, Me, OH, NH2). The neutral leaving group CO2
develops within the small hydrophobic pocket and the resulting
enediolate is stabilised by the dioxyanion hole. In the case of the
model substrate 2-methyl-2-phenylmalonate (L=Ph, S=Me) decarbox-
ylation has been shown to occur through loss of the pro-R carboxylate
and subsequent protonation of the si-face by Cys188, resulting in
overall inversion of configuration.[5]
with structures of Asp/Glu racemases,[8] we proposed that
conserved dioxyanion hole motifs could have evolved to
stabilise common enediolate intermediates in these mecha-
nistically related enzymes. Indeed, up until our structure of
the AMDase, there was no evidence to suggest how the
AMDase or Asp/Glu racemases stabilised common putative
high-energy enediolate intermediates.[5] However, our pro-
posal was based on a structure which only possessed
phosphate as an active-site ligand.[5] To lend further support
to our proposal, we here present a second B. bronchiseptica
AMDase structure with a mechanism-based inhibitor that
more closely resembles the enediolate intermediate bound to
the active site of the enzyme. This allowed us to rationalise the
mechanism of this simple yet intriguing enzyme and design
new and effective malonate substrates for the AMDase. Also,
the new structure guided directed evolution of new AMDase
mutants with significantly enhanced catalytic efficiency with a
range of substrates.
were able to suggest
a mechanism for the AMDase
(Scheme 1). From comparison of the AMDase structure[5]
[*] Dr. K. Okrasa, M. Wilding, M. Goodall, Prof. Dr. J. Micklefield
School of Chemistry, The University of Manchester
Manchester Interdisciplinary Biocentre
131 Princess Street, Manchester M1 7ND (UK)
E-mail: jason.micklefield@manchester.ac.uk
Dr. C. Levy, Dr. D. Leys
Faculty of Life Science, The University of Manchester
Manchester Interdisciplinary Biocentre
131 Princess Street, Manchester M1 7ND (UK)
E-mail: david.leys@manchester.ac.uk
Dr. N. Baudendistel, Prof. Dr. B. Hauer
BASF SE, GVF/E, 67056 Ludwigshafen (Germany)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.200904112.
Angew. Chem. Int. Ed. 2009, 48, 7691 –7694
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
7691

Communications
The natural substrate of the B. bronchiseptica AMDase is
through formation of an enediolate intermediate, which is
entirely consistent with the previous proposal.[5]
not known, however phenylmalonic acid (1a) is the best
substrate that has been described to date. Based on this, and
the fact that phosphate was previously shown to bind to the
dioxyanion hole,[5] benzylphosphonate (2) was selected as a
potential mechanism-based inhibitor. Benzylphosphonate (2)
is both sterically and electronically similar to the proposed
enediolate intermediate 3 resulting from decarboxylation of
phenylmalonate (Figure 1). In support of this, kinetic studies
Asymmetric decarboxylation of aliphatic a-disubstituted
malonates would be highly desirable. To rationalise why these
substrates are not accepted by the enzyme, a model of 2-
methyl-2-propylmalonate bound to the AMDase active site
was generated (Figure S3). This substrate fits well within the
active site, suggesting that electronic rather than steric effects
must account for the substrate selectivity in this case. Indeed,
the AMDase structures (Figure 1) indicate that the stacking
interaction between Pro14 and Gly189–Gly190 amide bond
necessitate that the phenyl group must be coplanar with the p-
system of the enediolate. Therefore, in addition to the H-
bonding with the dioxyanion hole, delocalisation of electron
density into the phenyl group is likely to further help stabilise
the enediolate. To explore this, malonate 5c with a trans-
propenyl substituent was prepared by modification of liter-
ature procedures (Scheme S1).[9] Notably, this substrate 5c is
decarboxylated by the B. bronchiseptica AMDase to give the
R-configured acid 6c, with 99% ee (Table 1). Kinetic param-
eters were determined for this transformation, using the
method described previously.[5,10] The results revealed that the
catalytic efficiency for conversion of the trans-propenyl
derivative 5c is lower than for the 2-methyl-2-phenylmalo-
nate (5a) (Table 1).
In light of this promising result, a range of substrates
possessing alkenyl substituents were prepared, through
modification of literature procedures (Scheme S1).[9,11]
Remarkably, all of these compounds proved to be substrates
for the AMDases giving decarboxylation products with very
high enantioselectivities, albeit with slightly lower catalytic
efficiencies compared with the corresponding phenylmalo-
Table 1: Kinetic parameters for AMDase catalysed decarboxylations.
Figure 1. A) Active-site structure of the B. bronchiseptica AMDase with
the benzylphosphonate inhibitor bound to the active site. This was
derived by soaking crystals of the enzyme with benzylphosphonate.
B) Model of the AMDase as derived from the original phosphate-
bound structure[5] by docking a putative enediolate intermediate 3 into
the active site using the H-bonds of the dioxyanion hole to establish
orientation. Residues of the dioxyanion hole are coloured in green in
structure A and blue in structure B. The Tyr126, which forms a sixth
H-bond of the dioxyanion hole has been removed from these struc-
tures for clarity.
[a]
Subst.
Km[a] [mm]
kcat[a] [sÀ1
]
kcat/Km
Product[c]
(config.)
ee[b]
[%]
[sÀ1 mmÀ1
]
revealed benzylphosphonate to be a competitive inhibitor of
the AMDase (Ki = 5.2 mm). Consequently, co-crystallisation
of the AMDase with the benzylphosphonate was undertaken.
The corresponding crystal structure shows the phosphonate
dianion 2 bound to the active site by the same six hydrogen
bonds that make up the dioxyanion hole (Figure 1A). More-
over, the phenyl group of the phosphonate side chain is
positioned in the larger solvent-accessible cavity, with the
phenyl ring stacking, through van der Waals interactions,
between the Gly189–Gly190 amide bond and the edge of the
Pro14 residue. This is similar to a model that was generated,
based on the first X-ray crystal structure of the AMDase
(Figure 1B and Figure S1 in the Supporting Information),
with the enediolate intermediate 3 docked to the dioxyanion
hole.[5] This structure thus provides additional evidence that
the AMDase catalyses decarboxylation of arylmalonates
1a
5a
5b
5c
5d
5e
7a
7b
7e
9a
9b
10.7Æ0.8[a]
26.9Æ2.1
7.8Æ0.6
12.8Æ1.0
14.4Æ1.1
3.5Æ0.3
30.2Æ2.4
9.1Æ0.7
7.3Æ0.6
n.d.[d]
316Æ36[a]
279Æ32
42.8Æ4.9
46.2Æ5.3
23.9Æ2.7
5.2Æ0.6
101Æ12
47.6Æ5.5
20.7Æ2.4
n.d.[d]
29.6Æ6.0[a]
10.4Æ2.1
5.5Æ1.1
3.6Æ0.8
1.7Æ0.3
1.5Æ0.3
3.3Æ0.7
5.2Æ1.1
2.8Æ0.6
n.d.[d]
4a (–)
6a (R)
6b (R)
6c (R)
6d (R)
6e (R)
8a (R)
8b (R)
8e (R)
10a (R)
10b (R)
–
99
99
99
99
99
96
98
76
33[d]
61[d]
n.d.[d]
n.d.[d]
n.d.[d]
[a] Kinetic parameters for B. bronchiseptica AMDase.[5] [b] Product ee
determined by chiral-phase GC. [c] Configuration was confirmed by
comparison with available homochiral standards or optical rotations
from the literature. [d] The absence of kinetic parameters and the low ee
observed are due to competing non-enzymatic decarboxylation.
7692 www.angewandte.org
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 7691 –7694

Angewandte
Chemie
nate derivatives. Interestingly, the alkenyl substrates with a-
methyl (5b–e) or a-hydroxy groups (7b, 7e) possess lower Km
values than the corresponding phenyl derivatives (5a, 7a).
This suggests alkenyl substrates, particularly those with more
substituted double bonds (5e, 7e), have higher affinity for the
AMDase. On the other hand, the turnover number is lower
for the alkenyl substrates compared with corresponding
phenyl derivatives. Given that the decarboxylation step to
generate the enediolate is most likely rate-limiting, this could
be explained by the greater delocalisation that is afforded by
phenyl compared with alkenyl substituents. Indeed the lowest
turnover numbers are observed for the substrates with more
electron-donating alkyl-substituted double bonds (e.g. 5e,
7e), which would be predicted to result in greater destabili-
sation of the transition state leading to the enediolate. Similar
results were also obtained with the AMDase from Mesorhi-
zobium sp.[5] (Table S1).
Previously the AMDase-catalysed decarboxylation of 2-
amino-2-phenylmalonate was reported to give (R)-phenyl-
glycine in 96% ee.[12] In our hands, this proved difficult to
reproduce as the background non-enzymatic decarboxylation
was found to compete with the enzyme-catalysed reaction,
which is in line with literature observations.[13] Despite this, 2-
amino-2-vinylmalonate (9b) was prepared (Scheme S1) and
incubated with the AMDase. Whilst significant background
decarboxylation was again observed, it was possible to obtain
(R)-vinylglycine (10b) in 61% ee. It is likely that the ee could
be improved using a milder and more efficient synthesis of 9b
that does not involve aggressive base-catalysed hydrolysis in
the final step. It may also be possible to increase the rate of
the enzymatic decarboxylation, through the directed evolu-
tion of the AMDases.
Based on the AMDase structure, five key active-site
residues were identified, and divided into two groups, for
saturation mutagenesis. The first group includes residues from
the large solvent-exposed aryl/alkenyl-binding pocket
(Pro14, Pro15 and Gly190) and the second group form the
smaller, more hydrophobic cavity (Val43 and Met159)
(Figure S4A). The two groups of active-site residues were
separately mutated, using the NNK degenerate codon,
following a strategy that is similar to iterative saturation
mutagenesis (ISM).[14] Following this strategy, mutation of the
aryl binding pocket group (Pro14, Pro15 and Gly190) led in
all cases to a first generation of single mutants of each residue
(Figure S4B). Those single mutants, which proved to be active
decarboxylases, were then subjected to a second round of
mutagenesis. Similarly, the second generation of active double
mutants were mutated to generate a library of triple mutants.
Phenylmalonic acid was used for initial screening, because
turnover of this substrate by AMDases is considerably faster
than any of the disubstituted malonates. Using this substrate,
mutant AMDases can be screened rapidly in 96-well plates,
using the pH indicator bromothymol blue (BTB). This assay
allows mutants to be selected visually through colour change,
and also accurate kinetic parameters can be determined as
described previously.[5]
further mutagenesis of Val43, but no active double mutants
were isolated. Despite this, we were able to generate a library
of single, double and triple mutants with significantly
improved efficiency for phenylmalonate decarboxylation
(Table 2). In light of this, successful mutants were further
screened with the pool of aryl and alkenylmalonates that had
been synthesised. Notably, the M159V mutant shows a 51-fold
improvement in activity with phenylmalonate 1a. Such a large
Table 2: Kinetic parameters of the selected mutant AMDases.
Subst. Mutants[b]
Km [mm]
kcat [sÀ1
]
kcat/Km
Rel.
act.[a]
[sÀ1 mmÀ1
]
1a
M159V
0.3Æ0.02 450Æ52
0.43Æ0.03 246Æ28
1499Æ290 51
570Æ115 19
11
M159G
P14V+P15G
M159S
M159C
M159C
G190A
3.5Æ0.3
5.0Æ0.4
1143Æ131 326Æ62
386Æ44
77.2Æ16
2.6
12.1Æ1.0 489Æ56
40.4Æ8.7 1.4
5a
5b
5e
7a
3.2Æ0.3
0.8Æ0.1
1.6Æ0.1
47.8Æ5.5 14.8Æ3.6 1.4
20.8Æ2.4 25.2Æ4.9 4.6
M159V
4.6Æ0.5
2.9Æ0.5
6.6Æ1.4
6.5Æ1.4
2.1
1.9
1.9
1.7
1.5
G190S+P14V+P15G 14.3Æ1.1 94.4Æ11
P14V+P15G
M159C
15.3Æ1.2 99.8Æ11
4.1Æ0.3
8.0Æ0.6
23.4Æ2.7 5.8Æ1.1
34.4Æ4.0 4.3Æ0.9
7e
P14V+P15G
[a] Relative activity was calculated using: Rel. act. = (kcat/Km of mutant)/
(kcat/Km of wild-type). [b] B. bronchiseptica AMDase mutants were charac-
terised by DNA sequencing and ESI-MS.
increase in activity, from a single-point mutant, suggests that
phenylmalonate is not the natural substrate for the B.
bronchiseptica AMDase. Indeed, nature has presumably
sampled this mutation as the Mesorhizobium sp. AMDase
possesses V rather than M at the position corresponding to
159, the only difference between the active sites of both
enzymes. The M159V mutant also shows a 2-fold increase in
catalytic efficiency with the new alkenyl malonate 5e.
Modelling of M159V and also M159G mutations suggests
these appear to generate more space around the catalytic
C188 residue, which can result in elevated flexibility of this
residue and closer interaction with bound substrate and/or
enediolate. The double mutant P14V+ P15G proved to be
one of the most versatile variants, since it has enhanced
activity towards three different substrates, phenylmalonate
(1a) (11-fold), 2-hydroxy-2-phenylmalonic acid (7a) (1.9-
fold) and the corresponding alkenyl derivative 7e (1.5-fold).
This double mutation provides more space and flexibility in
the large aryl-binding pocket, which can better accommodate
the larger alkenyl side chain of 7e as well as the phenyl groups
of 1a and 7a. Another notable mutant, G190A, showed a 4.6-
fold increase in activity towards the smallest aliphatic
substrate, 2-vinyl-2-methylmalonic acid (5b). Presumably
the G190A mutation decreases the size of the aryl-binding
pocket consequently providing tighter binding to the smaller
substrate 5b. Indeed, the Km of the G190A mutant for this
substrate is 0.8 mm, compared with 7.8 mm observed for the
wild-type enzyme. Finally, analysis of the chiral products
obtained from the decarboxylation with the mutant enzymes
showed that there was no change in ee compared with the
reaction catalysed by the wild-type AMDase.
In the case of the small binding pocket, we were able to
generate single mutants of Met159, but no mutants of Val43
could be isolated. Single mutants of Met159 were subjected to
Angew. Chem. Int. Ed. 2009, 48, 7691 –7694
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 7693

Communications
481; d) K. Matoishi, M. Ueda, K. Miyamoto, H. Ohta, J. Mol.
Catal. B 2004, 27, 161 – 168.
In summary, we have obtained a new crystal structure of
the B. bronchiseptica AMDase with benzylphosphonate
bound in the active site, suggesting that an enediolate is the
most probable intermediate formed during decarboxylation.
Although the presence of a dioxyanion hole stabilises the
enediolate, this appears insufficient for decarboxylation to
ensue, and viable substrates also need either an aryl or alkenyl
group in conjugation with the enediolate p-system to further
stabilise the developing negative charge concomitant with
decarboxylation. Other substituents possessing a p-system,
such as alkynyl groups, would presumably facilitate enzymatic
decarboxylation. These findings allowed us to design a library
of novel a-alkenyl malonate substrates which were decar-
boxylated with very high ee by the AMDase. Many of the
corresponding products are difficult to prepare, in such high
ee, using non-enzymatic asymmetric synthesis methods.
Furthermore, the double bond in these products is an ideal
handle for further synthetic modification, with the newly
installed a-stereogenic centre providing a means of control-
ling the configuration of additional stereocentres.[15–17] The a-
aryl and a-alkenylmalonate AMDase substrates are also very
cheap and readily available from natural sources. In addition
to malonic acid, many of the substrates can be derived from
ketomalonate (Scheme S1) which is available from oxidation
of glycerol, a by-product of biodiesel production.[18] Coupled
with the key environmentally benign enzymatic decarboxy-
lation, this makes for a more sustainable route to these key
chiral precursors. Finally, we used our refined inhibitor-bound
structure of the AMDase (PDB code 3IP8)[19] to guide the
directed evolution of new decarboxylases, with up to 50-fold
increase in activity. Moreover, mutants were selected which
showed improved catalytic activity with a range of a-aryl and
a-alkenyl malonates. Indeed, we have demonstrated that
AMDase displays considerable plasticity in the substrate
binding pocket, and is thus ideal for further rational redesign
and directed evolution.
[4] a) Y. Ijima, K. Matoishi, Y. Terao, N. Doi, H. Yanagawa, H.
Ohta, Chem. Commun. 2005, 877 – 879; b) Y. Terao, Y. Ijima, K.
Miyamoto, H. Ohta, J. Mol. Catal. B 2007, 45, 15 – 20.
[5] K. Okrasa, C. Levy, B. Hauer, N. Baudendistel, D. Leys, J.
Micklefield, Chem. Eur. J. 2008, 14, 6609 – 6613.
[6] a) K. Miyamoto, Y. Yatake, K. Tamura, Y. Terao, H. Ohta, J.
Biosci. Bioeng. 2007, 104, 263 – 267; b) Y. Yatake, K. Miyamoto,
H. Ohta, Appl Microbiol Biotechnol. 2008, 78, 793 – 799.
[7] Another B. bronchiseptica AMDase structure was also reported
at around the same time: E. B. Kuettner, A. Keim, M. Kircher, S.
Rosmus, N. Strꢀter, J. Mol. Biol. 2008, 377, 386 – 394. This
structure is proposed to possess four independent subunits and
differs considerably from our structure. Attempts to align the
two structures proved problematic. Also, there is no ligand
bound to any of the subunits, in the structure from Kuettner
et al., and none of the key active-site residues, apart from Cys188
are identified. Moreover, the Cys188 residue had been modified
by mercaptoethanol, presumably an artefact of the crystallisa-
tion process employed by Kuettner et al. From our work, the
conditions of crystallisation were shown to be very important.
The use of mercaptoethanol, in addition to modifying Cys188
which would completely abolish the ability of this residue to
function as a general acid, may also have effected the internal
stabilisation of the protein leading to the four misfolded
subunits. As a result, no reasonable structural or mechanistic
conclusions can be drawn from this.
[8] T. Lundqvist, S. L. Fisher, G. Kern, R. H. Folmer, Y. Xue, D. T.
Newton, T. A. Keating, R. A. Alm, B. L. de Jonge, Nature 2007,
447, 817 – 822.
[9] S. Tsuboi, K. Muranaka, T. Sakai, A. Takeda, J. Org. Chem. 1986,
51, 4944 – 4946.
[10] a) L. E. Janes, A. C. Lowendahl, R. J. Kazlauskas, Chem. Eur. J.
1998, 4, 2324 – 2331; b) A. Banerjee, P. Kaul, R. Sharma, U. C.
Banerjee, J. Biomol. Screening 2003, 8, 559 – 565.
[11] a) D. Grꢁe, L. Vallerie, R. Gre, L. Toupet, I. Washington, J. Org.
Chem. 2001, 66, 2374 – 2381; b) M. Calꢂ, M. Begtrup, Synthesis
2002, 63 – 66.
[12] K. Tamura, Y. Terao, K. Miyamoto, H. Ohta, Biocatal. Biotrans-
form. 2008, 26, 253 – 257.
[13] J. W. Thanassi, Biochemistry 1970, 9, 525 – 532.
[14] a) M. T. Reetz, J. D. Carballeira, A. Vogel, Angew. Chem. 2006,
118, 7909 – 7915; Angew. Chem. Int. Ed. 2006, 45, 7745 – 7751;
b) M. T. Reetz, J. D. Carballeira, Nat. Protoc. 2007, 2, 891 – 903.
[15] a) E. J. Corey, T. Hase, Tetrahedron Lett. 1979, 20, 335 – 338;
b) J.-M. Garnier, S. Robin, R. Guillot, G. Rousseau, Tetrahe-
dron: Asymmetry 2007, 18, 1434 – 1442.
[16] G. T. Crisp, P. T. Glink, Tetrahedron 1992, 48, 3541 – 3556.
[17] G. T. Pearce, W. E. Gore, R. M. Silverstein, J. Org. Chem. 1976,
41, 2797 – 2803.
[18] a) Y. Zheng, X. Chen, Y. Shen, Chem. Rev. 2008, 108, 5253 –
5277; b) R. Ciriminna, M. Pagliaro, Adv. Synth. Catal. 2003, 345,
383 – 388.
[19] The atomic coordinates for B. bronchiseptica AMDase with
benzylphosphonate bound to the active site dioxyanion hole has
been deposited in the Protein Data Bank (accession number
3IP8).
Received: July 24, 2009
Published online: September 8, 2009
Keywords: decarboxylase · directed evolution · enediolates ·
enzyme mechanisms · inhibitors
.
[1] a) J. Blanchet, J. Baudoux, M. Amere, M.-C. Lasne, J. Rouden,
Eur. J. Org. Chem. 2008, 5493 – 5506; b) M. Amere, M.-C. Lasne,
J. Rouden, Org. Lett. 2007, 9, 2621 – 2624; c) J. T. Mohr, T.
Nishimata, D. C. Behenna, B. M. Stoltz, J Am. Chem. Soc. 2006,
128, 11348 – 11349.
[2] K. Miyamoto, H. Ohta, J. Am. Chem. Soc. 1990, 112, 4077 – 4078.
[3] a) K. Miyamoto, H. Ohta, Biocatalysis 1991, 5, 49 – 60; b) K.
Miyamoto, H. Ohta, Appl. Microbiol. Biotechnol. 1992, 38, 234 –
238; c) K. Miyamoto, H. Ohta, Eur. J. Biochem. 1992, 210, 475 –
7694 www.angewandte.org
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 7691 –7694