Preparation of optically pure flurbiprofen via an integrated chemo-enzymatic synthesis pathway

Article abstract of DOI:10.1016/j.mcat.2019.01.024

In the synthesis of chiral molecules, the incorporation of enantioselective enzymatic conversions within the synthetic route often presents a useful approach. For the substitution of a chemical step with an enzymatic reaction, however, the complete synthetic route leading to and from this reaction needs to be considered carefully. An integrated approach, taking the possibilities and challenges of both types of conversions into account, can give access to chemo-enzymatic processes with great potential for effective synthesis strategies. We here report on the synthesis of enantiopure flurbiprofen using arylmalonate decarboxylase (AMDase, EC 4.1.1.76) in a chemo-enzymatic approach. Interestingly, practical considerations required shifting the enzymatic step to an earlier position in the synthetic route than previously anticipated. Engineered enzyme variants made it possible to obtain both (R)- and (S)-enantiomers of the target compound in excellent optical purity (>99%ee). The presented results underline that enzymes are most useful when they fit in a synthetic route, and that the optimization of biocatalytic steps and the planning of synthetic routes should be an integrated process.

Full text of DOI:10.1016/j.mcat.2019.01.024

MolecularCatalysis467(2019)135–142
Contents lists available at ScienceDirect
Molecular Catalysis
journal homepage: www.elsevier.com/locate/mcat
Preparation of optically pure flurbiprofen via an integrated chemo-
enzymatic synthesis pathway
Junichi Enokia,1, Max Linhorsta,1, Florian Buscha, Álvaro Gomez Baraibara, Kenji Miyamotob,
Robert Kouristc,⁎, Carolin Müggea,⁎
a Junior Research Group for Microbial Biotechnology, Ruhr-University Bochum, 44780, Bochum, Germany
b Department of Biosciences and Informatics, Keio University, 3-14-1 Hiyoshi, Yokohama, 223-8522, Japan
c Institute of Molecular Biotechnology, Graz University of Technology, Petersgasse 14, 8010, Graz, Austria
A R T I C L E I N F O
A B S T R A C T
Keywords:
Biocatalysis
Asymmetric catalysis
Protecting groups
C-C coupling
In the synthesis of chiral molecules, the incorporation of enantioselective enzymatic conversions within the
synthetic route often presents a useful approach. For the substitution of a chemical step with an enzymatic
reaction, however, the complete synthetic route leading to and from this reaction needs to be considered
carefully. An integrated approach, taking the possibilities and challenges of both types of conversions into ac-
count, can give access to chemo-enzymatic processes with great potential for effective synthesis strategies. We
here report on the synthesis of enantiopure flurbiprofen using arylmalonate decarboxylase (AMDase, EC
4.1.1.76) in a chemo-enzymatic approach. Interestingly, practical considerations required shifting the enzymatic
step to an earlier position in the synthetic route than previously anticipated. Engineered enzyme variants made it
possible to obtain both (R)- and (S)-enantiomers of the target compound in excellent optical purity (> 99%ee).
The presented results underline that enzymes are most useful when they fit in a synthetic route, and that the
optimization of biocatalytic steps and the planning of synthetic routes should be an integrated process.
Protein engineering
Introduction
considerable interest over the last years since it effectively promotes the
stereospecific decarboxylation of α-disubstituted malonic acids, re-
The production of drugs bearing a chiral center often relies on an
appropriate natural – chiral – source or makes use of laborious methods
of chiral synthesis that often include tedious steps of chiral resolution.
Another approach to introduce the chiral center may be the direct
asymmetric synthesis of intermediates by aid of highly effective en-
zymes [1]. Biocatalysis offers considerable advantages regarding se-
lectivity and sustainability, but the development of cost-efficient and
sustainable processes with enzymes requires a successful integration of
biocatalytic steps into synthesis routes. Since enzymatic reactions often
use a different substrate than the traditional chemical approach, the
availability of the starting material is as important as the efficiency of
the enzymatic reaction. An interesting case is presented by chiral α-
substituted carboxylic acids. This class of compounds embraces many
chiral building blocks and active pharmaceutical ingredients. While
numerous enzymes have been suggested to produce such compounds,
most suffer from narrow substrate spectra, insufficient enantioselec-
tivity or other limitations [2]. In this context, arylmalonate decarbox-
ylase from Bordetella bronchiseptica (AMDase, EC 4.1.1.76) has received
sulting in pure enantiomers of the respective monoacids [3–6]. It acts as
stereospecific decarboxylase to malonic acids bearing an aryl or vinyl
unit and a small substituent, such as methyl groups, at the quaternary α-
carbon atom. So far, all known natural AMDases exclusively produce
the (R)-enantiomer of the α-substituted monoacids due to a stereo-
specific protonation step, which is selectively conducted by a Cys-SH
function (Scheme 1) [7,8].
Switching the position of this Cys from one side of the substrate
binding site to the opposite by enzyme engineering gave access to
variants with selectivity towards the (S)-enantiomers [9]. While the
activity of the (S)-selective variants was significantly reduced, [12,13]
focused directed evolution [14–16] and rational design [3] overcame
this problem and even increased the activity of the (R)-selective var-
iants. Compounds of the profen class, α-arylpropionic acids like (R)- and
(S)-flurbiprofen or naproxen, could efficiently be produced by AMDase-
driven decarboxylation of the respective malonic acid precursors in
yields of > 98% and > 99%ee for the respective enantiomers [3,17].
One drawback of this synthesis is the intrinsic instability of the
⁎ Corresponding authors.
E-mail addresses: kourist@tugraz.at (R. Kourist), carolin.muegge@rub.de (C. Mügge).
1 Both authors contributed equally to this work.
https://doi.org/10.1016/j.mcat.2019.01.024
Received 1 November 2018; Received in revised form 17 January 2019; Accepted 21 January 2019
2468-8231/©2019ElsevierB.V.Allrightsreserved.

J. Enoki et al.
MolecularCatalysis467(2019)135–142
Scheme 1. Synthesis of chiral α-disubstituted carboxylic acids from α-disubstituted malonic acid precursors with (R)- and (S)-selective AMDase variants.
malonic acids 2, resulting in partial spontaneous decarboxylation and
therefore decreasing the ee of the final isolated product 3. Furthermore,
the classical synthesis of flurbiprofen 3a proceeds via malonic acid
dialkyl esters 1a (Scheme 2). These are then cleaved by protolysis, re-
sulting in ester hydrolysis and simultaneous decarboxylation of 2a,
therefore producing a racemic product rac-3a [18,11,10]. Efforts to
produce (R)-flurbiprofen (R)-3a from the dimethyl ester 1aMe resulted
in a maximal 92%ee after biocatalysis with AMDase [19]. Slightly
higher ee values (up to 96%ee) have been obtained with additional,
laborious, steps of racemic resolution [20]. To make use of the high
stereoselectivity of AMDase-based decarboxylation, the preparation of
the α-disubstituted malonic acid 2 is a crucial issue: the aforementioned
classical pathway will not suffice in producing a malonic acid 2 without
the presence of at least some decarboxylated product rac-3. We have
recently reported on the use of benzyl esters as alternative protecting
groups since they can be cleaved by hydrogenolysis instead of protolysis
[21]. Using this approach, we produced malonic acid precursors 1Bn to
flurbiprofen and naproxen without undesired spontaneous decarbox-
ylation [3].
Even though this was an important milestone in the incorporation of
the AMDase-based decarboxylation into a production process, the effi-
cient synthesis of the malonic acid benzyl esters 1Bn was challenging
and required optimization. We therefore set out to develop an in-
tegrated chemo-enzymatic approach for the production of optically
pure (R)- and (S)-flurbiprofen 3a based on the enzymatic decarbox-
ylation of α-disubstituted malonic acids 2 and their preparation from
their dibenzyl ester precursors 1Bn, taking into account the whole
synthesis pathway. In this context, it is of utmost importance to keep
the balance between enzyme effectivity and an efficient overall syn-
thetic route. Such an approach required to re-consider the position of
the enzymatic asymmetrization within the route. This was made pos-
sible since enzyme engineering has provided variants with an improved
activity for the synthesis of both enantiomers of different α-methyl ar-
ylpropionic acids 3 [3,14,15,19]. We report here on the full synthesis of
enantiopure flurbiprofen 3a by keeping in mind issues of catalyst
compatibility, practicability and efficiency.
of metal catalysts, organic compounds and strong acids have been re-
ported.
In our aim to work under the most economically and ecologically
benign conditions possible, we found the use of Pd/C as a readily
available and fairly stable hydrogenolysis catalyst a proper choice for
the cleavage of the two benzyl ester functions that would replace the
ethyl esters in our process. In addition, the same catalyst can be used in
an efficient Suzuki coupling, where sodium tetraphenyl borate serves as
fourfold phenyl group donor [11,22]. We therefore decided to design a
synthesis pathway that would include
(a) the Suzuki coupling of a phenyl group to a suitable halogenated
aromatic precursor,
(b) the hydrogenolysis of malonic acid dibenzyl esters, both catalyzed
by the heterogenous Pd/C catalyst, and
(c) the stereoselective decarboxylation of a malonic acid by AMDase
at strategically relevant steps of the process.
These three steps could be taken from one central intermediate of
the process, namely dibenzyl α-(4-bromo-3-fluorophenyl)-α-methyl-
malonate 1bBn (Scheme 3). From there, two different reaction se-
quences are in principle thinkable. The first route in the synthetic order
(a)-(b)-(c) (vi-vii-viii in Scheme 3) is closely related to the route re-
ported by Hylton [10] and includes the decarboxylation of flurbiprofen
malonic acid 2a by AMDase as the final and already established step
[3]. The second route in the synthetic order (b)-(c)-(a) (vii-viii-vi in
Scheme 3) was also considered, even though in this case the reactivity
of AMDase towards α-(4-bromo-3-fluorophenyl)-α-methylmalonic acid
2b had not been investigated so far.
The synthesis of 1bBn was explored on different routes, with the left
route in Scheme 3 being in close analogy with the industrial procedure
[10]. All steps were re-drawn, using dibenzyl methylmalonate 5Bn in-
stead of its diethyl counterpart 5Et. The final Sandmeyer-type conver-
sion of 1fBn was not successful in the isolation of 1bBn in our hands. The
compound was therefore prepared via an alternative route from 1-
bromo-2-fluoro-4-iodobenzene 6 and dibenzyl malonate 8Bn in two
steps involving Cu(I) catalyzed coupling [23] to give 1gBn and sub-
sequent methylation of the α-carbon atom (Scheme 3, right route). The
direct coupling of α-methyl dibenzyl malonate 5Bn with 6 (or 7) by this
procedure did not succeed (data not shown), possibly due to hindered
interaction of the in situ generated catalyst complex with the methyl
substituted 5Bn. In parallel, also the homologous compound without the
fluorine substituent 1cBn was synthesized via the same two-step
pathway and used for comparison.
Results and discussion
Establishment of the synthetic route
Preparations of flurbiprofen have been reported by different
methods [18,11,10,19]. They all include the formation of an aryl-aryl
bond to create the biphenyl unit and the protolysis of a malonic dialkyl
ester, ultimately leading to a racemic product. These steps are central
processes in the overall pathway and require several catalysts; The use
Next, we investigated the two possible routes to prepare (R)- and
(S)-flurbiprofen 3a. Following the pathway analogous to the one
136

J. Enoki et al.
MolecularCatalysis467(2019)135–142
Scheme 2. Reported procedures for the production of rac-flurbiprofen 3a ac-
cording to Hylton [10] (left) and Lu et al. (right) [11]. Reaction conditions: i,
NaOH, DMF, r.t., 4.5 h; ii, H2, Pd/C, EtOAc, r.t., 2.5 h; iii, [a] NaNO2, H2O,
benzene, 70 °C, [b] HOAc, benzene, 70 °C, 8 h, [c] 85% H2SO4; iv, [a] NaOH,
EtOH, r.t., 6 h, [b] HCl; v, HOAc, reflux, 17 h, vi, HOAc, H2SO4, H2O, reflux,
24 h; vii, H2, Pd/C, EtOH, r.t., 16 h; viii, [a] NaNO2, 40% HBr, H2O, 0 °C, [b]
CuBr, 40% HBr, 60 °C; ix, Ph4BNa, Na2CO3, Pd/C, reflux, 1 h.
Scheme 3. Explored synthesis routes towards optically pure flurbiprofen 3a.
Reaction conditions: i, NaOH, DMF, 0 °C, 12 h; ii, Fe/HCl, NH4Cl, EtOH/H2O,
65 °C, 12 h; iii, [a] NaNO2, 48% HBr, H2O, 0 °C, [b] CuBr, 48% HBr, reflux; iv,
Cs2CO3, CuI/picolinic acid, dioxane, r.t. 24 h; v, NaH, CH3I, DMF, 0 °C to r.t.,
4 h; vi, Ph4BNa, Na2CO3, Pd/C, H2O or H2O/dioxane, reflux, 1 h; vii, H2, Pd/C,
EtOAc, 5 h; viii, AMDase crude cell extract, Tris−HCl pH 8.5, 30 °C, 1 h.
reported by Hylton, [10] a phenyl group had to be coupled to the
bromo substituent in 1bBn by Suzuki coupling. Even though the reac-
tion took place to some degree, it was not possible to achieve full
conversion and an acceptable yield of 1aBn under conditions suitable
for this particular Suzuki coupling procedure (data not shown). A
possible reason for this is the low solubility of 1bBn in water at the one
hand and the low solubility of the reactant sodium tetraphenylborate in
less polar solvents on the other hand: It was not possible to dissolve
both components in the same mixture to a sufficient degree, even when
using organic solvents as additives. Therefore, even if the following
steps have been properly described and performed before, [3] this route
had to be considered impractical for the overall goal to achieve an ef-
ficient synthesis of flurbiprofen 3a. Also, using the initially reported
benzyl coupling method from 1fBn to 1aBn (Step iii in Scheme 2) [10]
was out of the question for us due to the very unfavorable and ecolo-
gically risky conditions (amongst others, the use of benzene as solvent).
We therefore turned our attention to the alternative route, demanding
an efficient activity of AMDase towards substrates that have previously
not been investigated.
The second possible route required the cleavage of the benzyl esters
from 1bBn to produce the malonic acid 2b. Interestingly, the hydro-
genolysis of 1bBn occurred less easily as the one of 1aBn: Cleavage of the
Bn ester from 1aBn had been achieved within 5 h at 0 °C, with higher
reaction temperatures leading to partial decarboxylation of the malonic
acid 2a [3]. Contrary, 2b was produced in 99% isolated yield upon
hydrogenolysis for 5 h at 25 °C, with no detectable spontaneous dec-
arboxylation to rac-3b. This presents a significant advantage over the
first-described route since the stability of the malonic acid in absence of
AMDase is a crucial factor for obtaining a high final ee value. The
catalyst Pd/C was easily filtered off, making it available for re-use.
Next, the malonic acid was subjected to enzymatic decarboxylation
137

J. Enoki et al.
MolecularCatalysis467(2019)135–142
by different variants of AMDase. By converting 1 mmol (291 mg, in
50 mL reaction volume) of 2b in presence of AMDase, both enantiomers
of the arylpropionic acid were produced in > 99%ee and isolated yields
of 99 and 95% for (R)-3b and (S)-3b, respectively. This was achieved
within 1 h of reaction with crude cell extract containing AMDase in
Tris-HCl buffer (pH 8.5), followed by extraction and crystallization of
the pure products. In order to closer investigate the efficiency of this
particular reaction, the decarboxylation of all structural analogs 2a-d
was conducted with the different AMDase variants (cf. below).
In the final step, Suzuki coupling with sodium tetraphenylborate in
the presence of Pd/C afforded enantiopure 3a with > 99%ee for both
enantiomers under retention of the stereocenter. Isolated yields were
99% and 98% for (R)-3a resp. (S)-3a. The results prove the stability of
the stereocenter under coupling conditions and show the high effec-
tivity of the coupling method under aqueous conditions. Again, work-
up of the final step was straightforward, since a simple precipitation-
filtration sequence allowed us to recover the catalyst and isolate the
pure product. It should be noted that the coupling of other aryl units
should also be possible via this method. Since Flurbiprofen is the only
pharmaceutically relevant drug in this class with a biphenyl-pharma-
cophore, we focused solely on the coupling of an unsubstituted phenyl
group.
result in a more effective decarboxylation and as well change the
binding of the substrate respective intermediate by limiting its flex-
ibility within the active site.
Results of the biocatalytic activity investigations are summarized in
Fig. 2 and Table S2. Phenyl malonate 2d lacks the α-methyl group and
is a fast-converted substrate for both (R)-selective variants. It is striking,
however, that the (S)-selective variants are far less active, with the
variant CLG-IPL only showing 8% relative activity (28 U · mg−1) of the
wildtype (345 U · mg−1). In contrast, this variant is considerably faster
than the wildtype in the conversion of 2a. This demonstrates that the
effect of amino acid substitutions in the hydrophobic pocket is very
specific for individual substrates. While the steric effect of the missing
α-methyl group in phenyl malonate 2d affected the variants differently,
the comparison of 2b with 2c clearly demonstrates that the presence of
the m-fluoro atom in 2b has an activating effect on the decarboxylation
catalyzed by all variants. The activating effect of the m-fluoro sub-
stituent for all AMDase variants resulted in a fourfold (for CLG-IPL and
IPLL) to eightfold (for CLG) improvement in the conversion of 2b vs. 2c.
In view of the desired synthesis of optically pure 3a, we were
pleased to find that (R)-selective AMDase IPLL showed activity in the
conversion of 2b (480 U · mg−1) comparable to 2a (522 U · mg−1), for
which the variant has initially been engineered [3]. This observation
might be explained by a comparable size as well as roughly similar
electronic effects ([+M] and [-I]) of the para phenyl (2a) and bromo
(2b) substituents. The compounds thus seem to be similar substrates to
variant AMDase IPLL, whilst their differences weigh more to the other
investigated AMDase variants. (S)-selective AMDase CLG-IPL converted
2b with a specific activity of 41 U · mg−1, which is satisfactory for
synthetic applications. As this variant is faster than the wildtype in the
conversion of 2a, but has one third of the wildtype’s relative activity
towards 2b, there is still potential for improvement by further en-
gineering that is specific for the newly identified interesting substrate
2b. The results show clearly that the availability of a reasonable variant
library is crucial for the quick identification of enzyme mutants with
high activity and the desired enantiopreference for the asymmetric
decarboxylation of different α-arylmalonates, and thus significantly
contributes to solving specific synthetic problems.
Development of the biocatalytic step
Key to the success of the overall route was the availability of opti-
mized AMDase variants to efficiently produce both enantiomers of the
α-arylpropioic acids 3. The engineering of the hydrophobic pocket of
the enzyme’s active site had generated AMDase variants with improved
activity for the synthesis of several arylpropionic acids, yet the effects
were rather specific for different substrates. As it was difficult to an-
ticipate which enzyme variants would be most suitable for the con-
version of the intermediate 2b, we screened several AMDase mutants
available in our laboratory. It is known that electron-donating sub-
stituents (like isobutyl) at the aromatic system decrease the enzyme
activity, [14,24] which is attributed to a lower stability of the inter-
mediately formed enediolate [25–27]. Electron-withdrawing sub-
stituents increase the stability of the enediolate and accelerate the re-
action. A typical example is flurbiprofen malonate 2a that gradually
undergoes spontaneous decarboxylation at neutral pH. Steric effects of
substituents at the aromatic site are more difficult to rationalize.
Nevertheless, AMDase and its variants have the capacity to produce
arylpropionic acids with large aromatic substituents. We reasoned that
the presence of a p-bromo-atom in 2b would not be sterically challen-
ging, and the overall weak electron-donating effect ([-I] ≈ [+M])
would be more than compensated by the electron-withdrawing effect
([-I] > [+M]) of the m-fluoro atom.
To understand the effect of the substituents on the enzymatic ac-
tivity, we compared α-(p-bromo-m-fluorphenyl)-α-methylmalonate 2b
with α-(p-bromophenly)-α-methylmalonate 2c and α-phenylmalonate
2d as a reference. We investigated the specific activity of the four most
promising AMDase variants available in our laboratory, namely the (R)-
selective wildtype (AMDase wt) and improved variant IPLL [3] as well
as the (S)-selective variants G74C/M159 L/C188 G (AMDase CLG) [15]
and its improved variant G74C/M159 L/C188 G/V43I/A125 P/V156 L
(AMDase CLG-IPL), [14] towards the crucial malonic acid 2b as well as
to structural homologs 2a, c and d.
Conclusions
By shifting the enzymatic reaction to an earlier step of the route
than previously anticipated, we developed an efficient preparation of a
suitable α-arylmalonate as key intermediate for the chemo-enzymatic
synthesis of optically pure flurbiprofen. The higher stability of 2b
compared to 2a towards spontaneous decarboxylation facilitates the
deprotection of its precursor considerably. By the identification of two
AMDase variants with high activity and outstanding enantioselectivity
in the decarboxylation of 2b, we could establish a synthetic route of
pure (S)- and (R)-flurbiprofen starting from simple and inexpensive
starting materials. This underlined the importance of mutant libraries
or enzyme collections for the rapid identification of suitable biocatalyst
variants.
The results furthermore show that an integrated optimization of
synthesis strategies and biocatalysis is an efficient approach for the
development of synthetic routes leading to optically pure ingredients.
Experimental section
Fig. 1 shows the composition of the substrate binding pocket for the
wildtype and the three engineered AMDase variants. It demonstrates
how the (R)- and (S)-selective variants can perform a stereospecific re-
protonation from a distinct side of the planar intermediate by swapping
the position of the catalytic cysteine from the pro-(R)-side (C188) to the
pro-(S)-side (C74) of the substrate binding site. All additional amino
acid exchanges only alter the shape and size of the binding pocket, as
there is no change in polarity or charge. By the introduction of the IPL
(L) mutation set, the size of the hydrophobic pocket is altered. This may
General experimental conditions
Syntheses were carried out under nitrogen atmosphere with anhy-
drous solvents using standard Schlenk techniques when required,
otherwise regular air-exposed conditions were used. If not indicated
otherwise, all chemicals used for synthesis and protein expression were
obtained from standard commercial suppliers (Sigma Aldrich, Alfa
Aesar, TCI, Fisher Scientific, abcr, VWR, Acros Organics) in the highest-
138

J. Enoki et al.
MolecularCatalysis467(2019)135–142
Fig. 1. Illustrations of the substrate binding
sites of the four investigated AMDase variants.
The graphics are based on the AMDase protein
structure PDB 3IXL [13] with phenylacetate as
ligand. Amino acid sequences were adapted
manually and energy-minimized (cf. sup-
porting information). Top left, (R)-selective
AMDase wt; Bottom left, (R)-selective im-
proved variant AMDase IPLL; Top right, (S)-
selective variant AMDase CLG; Bottom right,
(S)-selective improved variant AMDase CLG-
IPL.
Fig. 2. Specific activities of the four investigated AMDase variants towards the model substrates 2a-d.
available purity grade. As analytical product standard, rac-flurbiprofen
rac-3a (100% purity, ChemicalPoint, Deisenhofen, Germany) was pur-
chased. Palladium on charcoal (Pd/C, 10 wt-%, 50% water wet) was
purchased from Acros Organics. Solvents for anhydrous procedures
were purchased in its anhydrous form in sealed flasks. Light petroleum
(b.p. 40–65 °C) and ethyl acetate (EtOAc) used for synthesis were of
technical grade.
preparative scale have been reported earlier[3].
General analytic conditions
NMR analyses were carried out on Bruker Avance DPX-200 (200.1/
50.3 MHz for 1H/13C{1H}) or DRX-400 (400.13/100.61 MHz for
1H/13C{1H}) spectrometers. Data is given relative to TMS and refer-
enced to the solvent residual signal. Signal assignment was confirmed
by H,H-COSY, HSQC and HMBC experiments. FTIR spectra (ATR) were
recorded on a Bruker Tensor 27 device, equipped with a Pike MIRacle
micro ATR unit. HPLC chromatograms and mass spectra were obtained
Compounds 1cBn, 1eBn, 1fBn, 1hBn, 2c, 3c as well as rac-3b and rac-
3c as racemic standards for enantiopurity determination were prepared
as described in the supporting information. Preparations of compounds
2a from 1aBn and of enantiopure (R)- resp. (S)-3a from 2a in
139

J. Enoki et al.
MolecularCatalysis467(2019)135–142
from LC–MS (ESI) runs, using a Shimadzu LCMS-8030 device, equipped
with a SPD-M20 A diode array detector (scan range: 200–800 nm) and
using a Phenomenex Kinetex C18 reversed-phase column (2.6 μm,
100 A, 100 × 2.1 mm) including a Phenomenex SecurityGuard ULTRA
C18 guard column. HPLC runs were performed at an isocratic flow of
39.95% water, 0.1% formic acid and 59.95% acetonitrile at a flow rate
of 0.5 mL · min−1. MS experiments were performed using both, positive
and negative, electrospray ionization. The following standard instru-
ment parameters were used: Nebulizing gas flow 3 L · min−1, dilution
line temperature 250 °C, heat block temperature 400 °C, drying gas flow
15 L · min−1, CID gas 230 kPa, ion gauge vacuum 1.7·10−3 Pa, Interface
(30 mL, anhydrous), was added dropwise under nitrogen atmosphere
whilst cooling the system to 0 °C in an ice bath. After stirring the re-
action mixture for 1 h at 0 °C, iodomethane (815 μL, 13.1 mmol) was
added and the reaction mixture stirred for additional 3 h, while the
temperature was allowed to reach room temperature. The reaction
mixture was quenched by addition of NH4Cl (30 mL, 10% aqueous so-
lution) and subsequently extracted with EtOAc (3 × 30 mL). The com-
bined organic phases were dried over anhydrous Na2SO4 and the sol-
vents evaporated. Finally, the resulting colorless oil was purified by
silica gel column chromatography (mobile phase: light petroleum/
EtOAc 20:1 v/v, Rf(1bBn) ≈ 0.3). Isolated yield: 2.03 g (99%), colorless
oil. 1H NMR (400 MHz, CDCl3, δ ppm) 7.37 (dd, J = 8.5, 7.3 Hz, 1H, Ar-
voltage 4.5 kV. Mass detection occurred in scan mode (scan range: (
)
3
2
150–350 or 300-600 m/z, depending on analyte) and in single ion mode
(SIM) set to m/z values of the [M+H]+/[M+2+H]+ and [M−H]-/[M
+2-H]- ions of the respective analytes. For high-resolution mass data
(HRMS), samples were dissolved in 50% acetonitrile with (for positive
ionization) or without (for negative ionization) 0.1% formic acid and
directly injected into a Synapt G2-S-HDMSE mass spectrometer (Waters,
Milford, Massachusetts, USA) in combination with an ESI-LockSpray™-
Source (Waters). Spectra were recorded for 3 min in positive or nega-
tive high-resolution mode with the following settings: capillary voltage,
3 or 2 kV, respectively; cone voltage, 40 V; source temperature, 100 °C;
cone gas flow, 50 L · h−1; desolvation gas flow, 500 L · h−1; desolvation
temperature, 150 °C. MS spectra were recorded within a mass range of
50–800 m/z with a scan time of 1 s. Leucine-encephalin was injected
every 60 s as a lock mass using a capillary voltage of 3 kV. Data were
recorded and analyzed using the MassLynx™ software (Waters).
H), 7.27 – 7.11 (m, 10H, Ar-H), 7.06 (dd, JHH = 10.2, JHH = 2.3 Hz,
1H, Ar-H), 7.00 – 6.82 (m, 1H, Ar-H), 5.07 (s, 4H, Ar−CH2), 1.78 (s,
3H, CH3); 13C NMR (100 MHz, CDCl3, δ ppm) 170.4 (COOR), 159.9,
157.7 (Ar-Cq), 139.5 (Ar-Cq), 134.8, (Ar-Cq), 132.9 (Ar-C), 128.8 (Ar-C),
124.5 (Ar-C), 127.9 (Ar-C), 115.9 (Ar-C), 108.6 (Ar-Cq), 67.5
(Bn−CH2), 58.4 ((Cq), 22.2 (CH3). IR (ATR,
cm−1): 3034.1, 2997.4,
˜
ν
2950.4, 1727.4, 1673.2, 1623.5, 1577.8, 1483.9, 1455.0, 1414.0,
1383.9, 1321.9, 1210.8, 1178.8, 1108.9, 1074.8, 1038.2, 952.0, 926.1,
889.4, 877.7, 842.7, 802.0, 764.4, 725.5, 695.2, 632.6, 620.4. HRMS
(ESI+, m/z): found 471.0591 [M+H]+, calcd. 471.0607. rp-HPLC
(C18): tR = 1.76 min (> 92%).
2. -(4-bromo-3-fluorophenyl)-2-methylmalonic acid 2b
For cleavage of the benzyl esters, 1bBn (1.89 g, 4 mmol) was dis-
solved in EtOAc (10 mL) and mixed with Pd/C (189 mg, 0.02 mol-eq.,
10 wt-%, 50% water wet). The reaction mixture was stirred under an
atmosphere of H2 (1 bar) at room temperature for 5 h. The Pd/C cata-
lyst was removed by filtration over a pad of diatomaceous earth and the
solvent evaporated under reduced pressure at room temperature. The
reddish powder was purified by crystallization from EtOAc and light
petroleum. Isolated yield: 873 mg (75%), white crystals. 1H NMR
(400 MHz, MeOD, δ ppm) 7.59 (dd, 3JHH = 8.5, 7.4 Hz, 1H, Ar-H), 7.34
Dibenzyl-2-(4-bromo-3-fluorophenyl)-malonate 1gBn
A Schlenk flask was charged with dibenzylmalonate 8Bn (2.88 g,
10.13 mmol), 1-bromo-2-fluoro-4-iodobenzene 6 (3.00 g, 10.13 mmol)
and Cs2CO3 (9.90 g, 30.39 mmol), degassed, secured and dissolved in
1,4-dioxane (50 mL, anhydrous). Simultaneously, picolinic acid
(124.34 mg, 1.01 mmol) and CuI were mixed and dissolved in 1,4-di-
oxane (10 mL, anhydrous) in another Schlenk flask, upon which an
intensive red color appeared. The catalyst solution was added to the
reaction mixture and stirred at room temperature for 24 h. During that
time, the mixture’s color changed from an initial red to yellow or green.
To quench the reaction, NH4Cl (50 mL, 10% aqueous solution) was
added and residual Cs2CO3 neutralized by addition of HCl (37%,
dropwise). The intensively stained red reaction mixture was extracted
with EtOAc (3 × 30 mL) and the combined organic phases extracted
with NaHCO3 (saturated solution, 3 × 30 mL), dried over anhydrous
Na2SO4, filtered and the solvents evaporated. The resulting yellowish
oil slowly crystallized at room temperature. Residual impurities (mainly
educts) were removed by washing the crude crystals with a mixture of
ice-cold EtOAc and hexane (1:1 v/v). Finally, the product was re-crys-
tallized from boiling EtOAc. Isolated yield: 3.39 g (73%), white crystals.
1H NMR (400 MHz, CDCl3, δ ppm) 7.56 (dd, 3JHH = 8.3, 7.1 Hz, 1H, Ar-
H), 7.46 – 7.18 (m, 11H, Ar-H), 7.09 (m, 1H, Ar-H), 5.21 (d,
JHH = 1.6 Hz, 4H, Ar−CH2), 4.72 (s, 1H, CH); 13C NMR (100 MHz,
CDCl3, δ ppm) 166.9 (COOR), 160.2, 157.7 (Ar-Cq), 134.9 (Ar-Cq),
133.5 (Ar-C), 128.8 (Ar-C), 127.1 (Ar-C), 126.3 (Ar-C), 117.8 (Ar-C),
109.4, 109.2 (Ar-Cq), 100.0 (Ar-Cq), 67.7 (Bn−CH2), 56.9 (CH); IR
3
2
3
(dd, JHH = 10.7, JHH = 2.2 Hz, 1H, Ar-H), 7.21 (dd, JHH = 8.5,
2JHH = 2.3 Hz, 1H, Ar-H), 1.84 (s, 3H, CH3); 13C NMR (100 MHz,
MeOD, δ ppm) 174.2 (COOH), 161.1, 158.7 (Ar-Cq), 142.6 (Ar-Cq),
134.0 (Ar-C), 126.0 (Ar-C), 116.1 (Ar-C), 108.6, 108.4 (Ar-Cq), 59.2
(Cq), 22.7 (CH3); IR (ATR,
cm−1): 2993.3, 2877.7, 2627.0, 2529.9,
˜
ν
1697.9, 1575.4, 1487.4, 1463.5, 1405.2, 1286.6, 1268.4, 1247.1,
1179.6, 1132.5, 1115.8, 1077.3, 1035.5, 911.5, 853.2, 808.9, 781.8,
746.8, 724.8, 696.3, 685.5, 643.1; HRMS (ESI-, m/z): found 290.9659
[M−H]-, calcd. 290.9668; rp-HPLC (C18): tR = 0.68 min (> 99%).
(S)- and (R)-2-(4-bromo-3-fluorophenyl)-propanoic acid (S)-3b and (R)-3b
by AMDase catalysis in preparative scale
A 250 mL round bottom flask was charged with Tris−HCl buffer
(Tris(hydroxymethyl)-aminomethan−HCl, 43 mL, 50 mM Tris−HCl
pH 8.5, 300 mM NaCl). 2b (291 mg, 1 mmol) and NaOH (1 M aqueous
solution, 2 mL, 2 mmol) were added stepwise in an alternating manner.
Subsequently the pH of the solution was re-adjusted to pH 8.5 with
NaOH (1 M aqueous solution). After addition of excess crude cell ex-
tract containing AMDase (5 mL, Escherichia coli (E. coli) BL21 (DE3)
transformed with pET28a AMDase IPLL resp. pET28a AMDase CLG-IPL;
cf. below for expression procedure), the reaction mixture was stirred for
1 h at 30 °C. The reaction was quenched by addition of HCl (37%,
20 mL) and the mixture extracted with EtOAc (3 × 10 mL), the com-
bined organic phases dried over anhydrous Na2SO4 and the solvents
evaporated. Finally, the resulting brownish powder was crystallized
from EtOAc and light petroleum. Isolated yields: (S)-3b: 235 mg
(95%), > 99%ee, cream colored crystals; (R)-3b, 247 mg (99%), > 99%
ee, cream colored crystals; 1H NMR (400 MHz, MeOD, δ ppm) 7.55 (dd,
3JHH = 8.3, 3JHH = 7.3 Hz, 1H, Ar-H), 7.18 (m, 1H, Ar-H), 7.06 (m, 1H,
(ATR,
cm−1): 3067.2, 3037.4, 2945.3, 1748.7, 1719.8, 1577.5,
˜
ν
1496.5, 1483.0, 1455.6, 1423.4, 1381.4, 1318.1, 1294.0, 1228.5,
1191.5, 1153.3, 1141.5, 1081.8, 1042.5, 1000.3, 952.8, 937.2, 904.6,
889.6, 868.4, 837.8, 819.7, 789.4, 761.8, 742.6, 72.6, 692.8, 640.7,
619.0; HRMS (ESI+, m/z): found 457.0458 [M+H]+, calcd. 457.0451;
rp-HPLC (C18): tR = 1.47 min (> 98%).
Dibenzyl-2-(4-bromo-3-fluorophenyl)-2-methylmalonate 1bBn
A Schlenk flask was charged with NaH (272 mg, 6.78 mmol, 60 wt-
% in petroleum) and 1gBn (2.00 g, 4.37 mmol), dissolved in DMF
3
3
Ar-H), 3.77 (q, JHH = 21.0, JHH = 7.2 Hz, 1H, CH), 1.47 (d,
140

J. Enoki et al.
MolecularCatalysis467(2019)135–142
for AMDase CLG and CLG-IPL) was added to start the biocatalysis. The
reactions were quenched by addition of acetonitrile (250 μL) either
after 2 min or 5 min. Simultaneously, negative control samples without
enzyme solution were withdrawn after 0 min and 5 min.
3JHH = 7.2, JHH = 1.1 Hz, 3H, CH3); 13C NMR (100 MHz, MeOD, δ
ppm) 175.3 (COOH), 173.5 (Ar-Cq) 159.9, 157.0 (Ar-Cq), 142.5 (Ar-Cq),
133.0 (Ar-C), 124.1 (Ar-C), 114.0 (Ar-C), 107.2 (Ar-Cq), 44.3 (CH), 17.1
2
(CH3); IR (ATR,
cm−1): 2984.7, 2941.3, 2883.2, 2612.5, 2543.3,
˜
ν
2232.6, 2058.2, 1697.0, 1602.8, 1577.5, 1484.2, 1461.9, 1421.8,
1384.3, 1336.6, 1278.4, 1241.5, 1210.7, 1165.0, 1151.3, 1079.1,
1061.8, 1041.7, 1008.9, 947.2, 924.7, 869.2, 853.7, 830.3, 816.9,
794.5, 743.6, 714.3, 706.4, 674.8, 646.1, 634.3; MS (ESI-, m/z): 245/
247 [M−H]-/[M+2-H]-; Rp-HPLC (C18): tR((S)-3b) = 0.75 min
(> 99%); tR((R)-3b) = 0.76 min (> 99%).
Afterwards samples were measured by HPLC without any additional
preparation. HPLC chromatograms were obtained from a Knauer Azura
HPLC system (Knauer, Berlin, Germany). A NUCLEODUR C18 Pyramid
column (5 μm; 4.6 × 250 mm; Macherey-Nagel, Düren, Germany) was
used for the stationary phase. As mobile phases, HPLC-grade solvents
(Fisher Scientific) were used. Mobile phase A consisted of acetonitrile
with 0.05% TFA and mobile phase B of ultrapure water with 0.05%
TFA. All runs were monitored at a detection wavelength of 254 nm.
HPLC methods were optimized according to the substrate/product
pairs’ requirements as given in detail in the supporting information.
Enzymatic activities were calculated by linear regression of the area
values for the product 3 including all time points with a conversion rate
of less than 10%. The concentration of product 3 was corrected by the
value of the respective negative control and plotted versus the time. For
detailed data, cf. the Supporting information.
(S)- and (R)-2-(2-fluoro-biphenyl-4-yl)-propanoic acid (S)-3a and (R)-3a
Depending on the desired product, (S)-3b or (R)-3b (200 mg,
0.8 mmol) were dissolved in water (15 mL) together with sodium tet-
raphenylborate (139 mg, 0.4 mmol), Na2CO3 (172 mg, 1.6 mmol). After
addition of Pd/C (85 mg, 0.05 mol-eq., 10 wt-% 50% water wet), the
reaction mixture was stirred for 2 h under reflux conditions. The reac-
tion was quenched and 3a precipitated by addition of HCl (37%,
10 mL). The suspension was filtered on a pad of diatomaceous earth and
the precipitated 3a re-dissolved by washing with EtOAc, followed by
precipitation with light petroleum. Isolated yields: (S)-3a: 196 mg
(98%), > 99%ee, cream colored crystals; (R)-3a, 199 mg (99%), > 99%
ee, cream colored crystals; 1H NMR (400 MHz, CDCl3, δ ppm) 7.56 –
Determination of optical purity of compounds 3
Isolated product powders (ca. 1 mg) were dissolved in 600 μL MTBE.
100 μL Methanol and 25 μL trimethylsilyl-diazomethane were added
and the mixture incubated at room temperature for 30 min. 5 μL Acetic
acid were added, the solvents evaporated and the residue re-dissolved
in 200 μL ethyl acetate. Samples were then analyzed by GC-FID using a
Shimadzu GC Plus 2010 device; using a FS-Hydrodex-β-6TBDM column
(Macherey-Nagel, 25 m × 0.25 mm ID), with methods having a tem-
perature profile optimized for each compound (cf. supporting in-
formation) with an injection split of 1/20. A peak difference of at least
0.5 min was achieved to ensure baseline separation. The ee values were
determined from the relative peak areas according with standard pro-
cedures. Errors of ee value determination were extrapolated from the
area difference obtained from the respective racemic authentic stan-
dard. From this, the precision of ee values was found to be 0.3 %.
3
6.98 (m, 8H, Ar-H), 3.72 (q, JHH = 7.2 Hz, 1H, CH), 1.49 (d,
3JHH = 7.2 Hz, 3H, CH3); 13C NMR (100 MHz, CDCl3, δ ppm) 179.3
(COOH), 160.9, 158.5 (Ar-Cq), 140.9 (Ar-Cq), 135.4 (Ar-Cq), 130.9 (Ar-
C), 128.9 (Ar-Cq), 128.6 (Ar-C), 127.8 (Ar-C), 123.5 (Ar-C), 115.4 (Ar-
C), 44.8 (CH), 18.1 (CH3); IR (ATR,
cm−1): 3166.6, 1730.6, 1696.7,
˜
ν
1621.2, 1580.7, 1560.9, 1509.4, 1482.5, 1460.3, 1419.0, 1389.9,
1269.8, 1233.3, 1218.7, 1176.1, 1144.7, 1129.9, 1091.0, 1070.8,
1039.6, 1009.4, 924.2, 915.6, 884.4, 847.0, 794.7, 761.4, 732.3, 722.6,
696.8, 646.0; MS (ESI-, m/z): 243 [M−H]-; rp-HPLC (C18): tR((S)-
3a) = 0.81 min (> 99%); tR((R)-3a) = 0.81 min (> 99%).
Enzyme preparation
E. coli BL21 (DE3) cells bearing a pET28a vector with an N-terminal
His-tag sequence and the desired AMDase mutants were cultivated in
LB-medium (200 mL) containing 30 μg · mL−1 kanamycin at 37 °C in a
1 L shake flask. After reaching an OD600 of 0.5, IPTG (isopropyl-β-D-
thiogalactopyranoside,1 mM) was added and the cells were cultivated
for another 12 h at 30 °C. Cells were harvested by centrifugation
(20 min, 4 °C, 5000×g) and washed twice with Tris−HCl buffer
(20 mL, 50 mM, pH 7.5, 300 mM NaCl). Cells were resuspended in
Tris−HCl buffer (5 mL) containing 10 mM imidazole and were dis-
rupted by sonication. Cell debris was removed by centrifugation
(30 min, 4 °C, 8000×g). AMDase variants were, when required, pur-
ified by His-tag purification by using a Ni sepharose spin coloumn
(Thermo Fischer) according to the instructions of the manufacturer. The
protein was eluted with 250 mM imidazole solution. The enzyme so-
lution was prepared by washing the protein in centricon tubes (10 kDa
membrane, Thermo Fischer) with Tris−HCl buffer (50 mM, pH 7.5,
300 mM NaCl).
Acknowledgements
The Federal Ministry for Innovation, Science and Research of North
Rhine-Westphalia (grant number PtJ-TRI/1411ng006) is gratefully ac-
knowledged for financial support. R.K. and F.B. thank the European
Union’s Horizon 2020 MSCA ITN-EID program (BIOCASCADES under
Grant Agreement No. 634200) for financial support. Sina Schäkermann
and Julia Bandow (Ruhr-University Bochum) are gratefully acknowl-
edged for recording high-resolution mass spectra.
Appendix A. Supplementary data
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.mcat.2019.01.024.
References
[1] K. Faber (Ed.), Biotransformations in Organic Chemistry, Springer-Verlag Berlin
Heidelberg, Heidelberg, Dordrecht, London, New York, 2011.
Determination of specific activities for the enantioselective decarboxylation
of arylmalonates 2a-2d by AMDase mutants
[2] R. Kourist, P.D. de María, K. Miyamoto, Green Chem. 13 (2011) 2607–2618.
[3] S.K. Gaßmeyer, J. Wetzig, C. Mügge, M. Assmann, J. Enoki, L. Hilterhaus, R. Zuhse,
K. Miyamoto, A. Liese, R. Kourist, ChemCatChem 8 (2016) 916–921.
[4] K. Miyamoto, R. Kourist, Appl. Microbiol. Biotechnol. 100 (2016) 8621–8631.
[5] R. Lewin, M. Goodall, M.L. Thompson, J. Leigh, M. Breuer, K. Baldenius,
J. Micklefield, Chem. Eur. J. 21 (2015) 6557–6563.
Reaction buffer solutions (50 mM Tris−HCl, 300 mM NaCl) were
prepared freshly and the pH adjusted to pH 8.5 with NaOH at the re-
action temperature (30 °C) to ensure the proper conditions throughout
the reactions. Substrate stock solution (250 μL, 20 mM substrate in re-
action buffer) and pure buffer (240 μL) were mixed in a small glass vial
and pre-incubated at 30 °C and 600 rpm for 5 min. Afterwards, enzyme
solution (10 μL, 12.5 μg · mL−1, 50 mM Tris−HCl/NaOH pH 7.5 at
25 °C, cend = 0.25 μg · ml-1 for AMDase wt and IPLL; cend = 0.50 μg·mL-1
[6] K. Okrasa, C. Levy, M. Wilding, M. Goodall, N. Baudendistel, B. Hauer, D. Leys,
J. Micklefield, Angew. Chem. Int. Ed. 48 (2009) 7691–7694.
[7] K. Miyamoto, Y. Yatake, K. Tamura, Y. Terao, H. Ohta, J. Biosci. Bioeng. 104 (2007)
263–267.
[8] Y. Yatake, K. Miyamoto, H. Ohta, Appl. Microbiol. Biotechnol. 78 (2008) 793–799.
[9] Y. Ijima, K. Matoishi, Y. Terao, N. Doi, H. Yanagawa, H. Ohta, Chem. Commun.
(2005) 877–879.
141

J. Enoki et al.
MolecularCatalysis467(2019)135–142
[10] T.A. Hylton, J.A. Walker, Process for Preparing Therapeutic 2-Arylpropionic Acids
and Esters, and Novel Arylmethylmalonate Esters, (1981) EP0032620 (A1).
[11] G. Lu, R. Franzén, X.J. Yu, Y. jun Xu, Chin. Chem. Lett. 17 (2006) 461–464.
[12] Y. Terao, Y. Ijima, K. Miyamoto, H. Ohta, J. Mol. Catal. B Enzym. 45 (2007) 15–20.
[13] R. Obata, M. Nakasako, Biochemistry (Mosc.) 49 (2010) 1963–1969.
[14] S. Yoshida, J. Enoki, R. Kourist, K. Miyamoto, Biosci. Biotechnol. Biochem. 79
(2015) 1965–1971.
[15] Y. Miyauchi, R. Kourist, D. Uemura, K. Miyamoto, Chem. Commun. 47 (2011)
7503–7505.
[16] Y. Terao, K. Miyamoto, H. Ohta, Appl. Microbiol. Biotechnol. 73 (2006) 647–653.
[17] M. Aßmann, C. Mügge, S.K. Gaßmeyer, J. Enoki, L. Hilterhaus, R. Kourist, A. Liese,
S. Kara, Front. Microbiol. 8 (2017) 448.
2395–2397.
[20] (a) R. Morrone, G. Nicolosi, A. Patti, M. Piattelli, Tetrahedron Asymmetry 6 (1995)
1773–1778;
(b) L. Qian, S. Chen, B. Shi, Biocatal. Biotransform. 25 (1) (2007) 29–34.
[21] R. Zuhse, J. Wetzig, R. Kourist, S. Gassmeyer, Verfahren Zur Herstellung von 2-Aryl-
2-Alkyl-Malonsäuren, (2015) DE102014006224 (A1).
[22] G. Lu, R. Franzén, Q. Zhang, Y. Xu, Tetrahedron Lett. 46 (2005) 4255–4259.
[23] S.F. Yip, H.Y. Cheung, Z. Zhou, F.Y. Kwong, Org. Lett. 9 (2007) 3469–3472.
[24] R. Kourist, Y. Miyauchi, D. Uemura, K. Miyamoto, Chem. Eur. J. 17 (2011)
557–563.
[25] M.E.S. Lind, F. Himo, ACS Catal. 4 (2014) 4153–4160.
[26] K. Okrasa, C. Levy, B. Hauer, N. Baudendistel, D. Leys, J. Micklefield, Chem. Eur. J.
14 (2008) 6609–6613.
[18] T. Mizushima, M. Otsuka, Y. Okamoto, N. Yamakawa, 2-Fluorophenylpropionic
Acid Derivative, (2014) EP2799424 (A1).
[27] K. Matoishi, M. Ueda, K. Miyamoto, H. Ohta, J. Mol. Catal. B Enzym. 27 (2004)
[19] Y. Terao, Y. Ijima, H. Kakidani, H. Ohta, Bull. Chem. Soc. Jpn. 76 (2003)
161–168.
142