Application of a recombinant Escherichia coli whole-cell catalyst expressing hydroxynitrile lyase and nitrilase activities in ionic liquids for the production of (S)-mandelic acid and (S)-mandeloamide

Article abstract of DOI:10.1002/adsc.201100391

The conversion of benzaldehyde and cyanide into mandelic acid and mandeloamide by a recombinant Escherichia coli strain which simultaneously expressed an (S)-hydroxynitrile lyase (oxynitrilase) from cassava (Manihot esculenta) and an aryl-acetonitrilase from Pseudomonas fluorescens EBC191 was studied. Benzaldehyde exhibited a pronounced inhibitory effect on the nitrilase activity in concentrations ≥25 mM. Therefore, it was tested if two-phase systems consisting of a buffered aqueous phase and the ionic liquid 1-butyl-1-pyrrolidinium bis(trifluoromethanesulfonyl)imide (BMpl NTf ₂) or 1-butyl-3-methylimidazolium hexafluorophosphate (BMim PF ₆) could be used for the intended biotransformation. The distribution coefficients of the substrates, intermediates and products of the reaction were determined and it was found that BMpl NTf₂ and BMim PF₆ were highly efficient as substrate reservoirs for benzaldehyde. The recombinant E. coli strain was active in the presence of BMpl NTf₂ or BMim PF₆ phases and converted benzaldehyde and cyanide into mandelic acid and mandeloamide. The two-phase systems allowed the conversion of benzaldehyde dissolved in the ionic liquids to a concentration of 700 mM with product yields (=sum of mandelic acid and mandeloamide) of 87-100%. The cells were slightly more effective in the presence of BMpl NTf₂ than in the presence of BMim PF₆. In both two-phase systems benzaldehyde and cyanide were converted into (S)-mandeloamide and (S)-mandelic acid with enantiomeric excesses ≥94%. The recombinant E. coli cells formed, in the two-phase systems with ionic liquids and increased substrate concentrations, higher relative amounts of mandelo-amide than in a purely aqueous system with lower substrate concentrations.

Full text of DOI:10.1002/adsc.201100391

FULL PAPERS
DOI: 10.1002/adsc.201100391
Application of a Recombinant Escherichia coli Whole-Cell
Catalyst Expressing Hydroxynitrile Lyase and Nitrilase Activities
in Ionic Liquids for the Production of (S)-Mandelic Acid and
(S)-Mandeloamide
Stefanie Baum,a Fred van Rantwijk,b and Andreas Stolza,*
a
Institut fꢀr Mikrobiologie, Universitꢁt Stuttgart, Allmandring 31, 70569 Stuttgart, Germany
Fax : (+49)-711-685-65725; e-mail: Andreas.Stolz@imb.Uni-Stuttgart.de
Laboratory of Biocatalysis and Organic Chemistry, Department of Biotechnology, Delft University of Technology,
b
Julianalaan 136, 2628 BL Delft, The Netherlands
Received: May 16, 2011; Revised: August 15, 2011; Published online: January 12, 2012
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/adsc.201100391.
Abstract: The conversion of benzaldehyde and cya- BMim PF6 phases and converted benzaldehyde and
nide into mandelic acid and mandeloamide by a re- cyanide into mandelic acid and mandeloamide. The
combinant Escherichia coli strain which simultane- two-phase systems allowed the conversion of benzal-
ously expressed an (S)-hydroxynitrile lyase (oxynitri- dehyde dissolved in the ionic liquids to a concentra-
lase) from cassava (Manihot esculenta) and an aryl- tion of 700 mM with product yields (=sum of man-
A
from
Pseudomonas
fluorescens delic acid and mandeloamide) of 87–100%. The cells
EBC191 was studied. Benzaldehyde exhibited a pro- were slightly more effective in the presence of BMpl
nounced inhibitory effect on the nitrilase activity in NTf2 than in the presence of BMim PF6. In both
concentrations ꢀ25 mM. Therefore, it was tested if two-phase systems benzaldehyde and cyanide were
two-phase systems consisting of a buffered aqueous converted into (S)-mandeloamide and (S)-mandelic
phase and the ionic liquid 1-butyl-1-pyrrolidinium acid with enantiomeric excesses ꢀ94%. The re-
bis(trifluoromethanesulfonyl)imide (BMpl NTf2) or combinant E. coli cells formed, in the two-phase sys-
1-butyl-3-methylimidazolium
(BMim PF6) could be used for the intended biotrans- centrations, higher relative amounts of mandelo-
formation. The distribution coefficients of the sub- amide than in a purely aqueous system with lower
hexafluorophosphate tems with ionic liquids and increased substrate con-
AHCTUNGTRENNUNG
strates, intermediates and products of the reaction substrate concentrations.
were determined and it was found that BMpl NTf2
and BMim PF6 were highly efficient as substrate res-
ervoirs for benzaldehyde. The recombinant E. coli Keywords: cascade reactions; hydroxynitrile lyase;
strain was active in the presence of BMpl NTf2 or ionic liquids; nitrilase; recombinant Escherichia coli
Introduction
EBC191 in
(Scheme 1).
a
bienzymatic cascade reaction
A
The combination of the two enzymes can be
and 2-hydroxycarboxylic acids are interesting prod- achieved basically in two different ways. Initially, an
ucts for the chemical industry and also can be used as in-vitro system was established in which the oxynitri-
building blocks for various follow-up reactions. We lase and the nitrilase were combined in an immobi-
are currently analysing the synthesis of chiral (S)-2- lised form (“combi-CLEAs”). This system could oper-
hydroxycarboxylic acids and (S)-2-hydroxycarbox- ate in a biphasic reaction system consisting largely of
A
(S)-selective hydroxynitrile lyase (oxynitrilase) from v/v) of sodium citrate buffer (pH 5). The immobilised
the cassava plant (Manihot esculenta) with a nitrilase enzymes converted benzaldehyde and HCN highly
from the bacterium Pseudomonas fluorescens enantioselectively into (S)-mandelic acid and (S)-
Adv. Synth. Catal. 2012, 354, 113 – 122
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
113

Stefanie Baum et al.
FULL PAPERS
Scheme 1. Bienzymatic synthesis of (S)-mandelic acid and (S)-mandeloamide from benzaldehyde and HCN by combining
the (S)-selective oxynitrilase from Manihot esculenta (MeHnL) and the non-selective nitrilase from Pseudomonas fluorescens
EBC191 (NLase).
mandeloamide.[1] This system has the advantage that the presence of ionic liquids have used isolated en-
the oxynitrilase is rather active in the presence of di- zymes, there are also some first reports which de-
ACHTUNGTRENNUNGisopropyl ether. Furthermore, HCN is highly soluble scribe the application of whole-cell catalysts in the
in the diisopropyl ether phase and the consumption of presence of ionic liquids.[7,8,9] Therefore, it was at-
HCN by the biocatalyst in the aqueous phase does tempted in the present study to develop a cyanide
not significantly influence the pH of the reaction and aldehyde transfer system by using ionic liquids
system. The potential problem of this system is that which is compatible with the “two-enzyme whole-cell
nitrilases generally occur to be rather sensitive to or- catalyst”.
ganic solvents and that diisopropyl ether is rather
problematic for this type of biotransformation reac-
tion because of its low logP (logP=1.9).[2,3,4]
Results
Recently, a second system for the combination of
the oxynitrilase from cassava and the Pseudomonas Identification of the Enzymatic Activity that is Most
nitrilase was described. In this system, recombinant E. Sensitive against Increased Substrate Concentrations
coli strains were constructed which simultaneously ex-
pressed both enzymes and which can be used as We have previously reported a significant activity loss
whole-cell catalysts in aqueous media for the effective when the whole-cell biocatalyst was exposed to high
synthesis of (S)-carboxylic acids or (S)-carboxamides substrate concentrations.[5] Therefore, resting cells of
using slightly acidic pH values.[5,6] These reactions can E. coli JM109
ACTHNUTRGNEU(GN pIK9)(pJOE5361.1) – which encodes
be operated in a purely aqueous system which is ben- for a His-tagged variant of the nitrilase from P. fluo-
eficial for the enzymes used. The major drawback of rescens EBC191 and the oxynitrilase from Manihot es-
this system is that the cyanide is added in the form of culenta – were incubated in sodium citrate buffer
KCN (which can cause a significant alkalisation of the (100 mM, pH 5) with 175 mM benzaldehyde and
medium) and that the addition of larger substrate 350 mM cyanide in order to analyse if the oxynitrilase
concentrations resulted in significantly decreased re- or the nitrilase activities were primarily effected. An
action rates and enantioselectivities.[5]
aqueous solution of cyanide (3.6M, pH 5) was used in
A possible strategy to improve the existing systems order to prevent an alkalisation of the reaction mix-
could be the development of cyanide and benzalde- ture by the addition of high amounts of KCN. It was
hyde transfer systems, which enhance the bioavailabil- found that the resting cells rapidly transformed the
ity of the substrates and which are compatible with substrates into mandelonitrile, but only low concen-
the “two-enzyme whole-cell catalyst”. The solvent trations of mandelic acid and mandeloamide were
systems that are used in the established synthetic oxy- formed (Figure 1, A). In contrast, in a cell-free con-
nitrilase reactions with purified enzyme preparations trol only a much slower spontaneous formation of
(e.g., diisopropyl ether) are rather problematic for mandelonitrile was observed (and no mandelic acid
this type of biotransformation reactions as these sol- or mandeloamide) (Figure 1, B). This demonstrated
vents demonstrate pronounced toxic effects on whole that mainly the nitrilase activity was inhibited by in-
cell catalysts. Therefore, it was attempted in the pres- creased substrate concentrations.
ent study to identify solvent systems that might be
less problematic for the intended reactions. An inter-
esting alternative to classical organic solvents are Effects of Benzaldehyde and Cyanide on the Nitrilase
ionic liquids which have already been used as solvent Activity
systems for various enzymatic processes. The utiliza-
tion of ionic liquids as solvent systems for this type of In the following it was analysed if the inhibitory
reactions might offer significant advantages because effect on the nitrilase activity was mainly due to the
of their relative hydrophilicity combined with a high increase in the benzaldehyde or cyanide concentra-
degree of compatibility with cellular processes. Al- tions. Therefore, resting cells of E. coli JM109
ACHTUNGTRENNUNG(pIK9)
though most studies which focused on biocatalysis in – which only express the nitrilase activity – were incu-
114
asc.wiley-vch.de
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2012, 354, 113 – 122

Application of a Recombinant Escherichia coli Whole-Cell Catalyst
Figure 1. Conversion of benzaldehyde and cyanide in the presence of resting cells of E. coli JM109ACTHNUGRTNEUNG(pIK9)(pJOE5361.1) (A)
or in the absence of cells (B). The reaction mixtures contained in 1 mL 100 mM sodium citrate buffer (pH 5), 10% methanol,
175 mM benzaldehyde and 350 mM cyanide (from a 3.6M stock solution, brought to pH 5 by the addition of citric acid). In
(A) resting cells of E. coli JM109ACHTNUGTRNEN(UG pIK9)(pJOE5361.1) were added corresponding to an OD600nm =67. The reaction mixtures
were incubated in a thermomixer at 308C and 1100 rpm. At the indicated time intervals, aliquots (90 mL each) were taken,
mixed with 10 mL 1M HCl, the samples were centrifuged (10 min, 13500 rpm, 48C), the supernatants appropriately diluted
!
~
*
*
and the concentrations of benzaldehyde ( ), mandeloamide ( ), mandelic acid ( ), and mandelonitrile ( ) analysed by
HPLC (the missing amount of benzaldehyde in Figure 1, B at t=0 min is due to the incomplete mixing of benzaldehyde,
which initially formed a second liquid phase in the reaction vessel).
bated with different concentrations of benzaldehyde 25 mM a pronounced inhibitory effect on the nitrilase
(0–75 mM) or cyanide (0–340 mM) and the residual activity of the whole-cell catalyst (Figure 2, A). Fur-
nitrilase activities determined after 30 min. These ex- thermore, it was found that benzaldehyde demonstrat-
periments demonstrated that benzaldehyde caused in ed a much stronger inhibitory effect on the nitrilase
purely aqueous systems already in concentrations of activity than cyanide (Figure 2, B).
Figure 2. Effects of different concentrations of benzaldehyde (BA) or cyanide on the nitrilase activity of E. coli
JM109ACHTUNGTRENNUNG(pIK9). E. coli JM109ACHTUNGTNER(NUGN pIK9) was grown in LB/Amp medium, the cells harvested by centrifugation, washed in Tris/
HCl buffer (50 mM, pH 7,5), and finally resuspended in sodium citrate buffer (100 mM, pH 5). The reaction mixtures con-
tained (in 1 mL) sodium citrate buffer (100 mM, pH 5), resting cells (OD600nm =0.4), and the indicated concentrations of ben-
zaldehyde (from a 2M stock solution in methanol) or HCN (from a 3M stock solution, brought to pH 5 by the addition of
citric acid). The reaction mixtures were incubated in a thermomixer at 308C and 1100 rpm. After 30 min, 50 mL of a mande-
lonitrile stock solution (210 mM in methanol) were added to the vials. After 10 min 90 mL were removed and mixed with
10 mL 1M HCl. The aliquots were centrifuged (10 min, 13500 rpm, 48C) and the supernatants analysed by HPLC.
Adv. Synth. Catal. 2012, 354, 113 – 122
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
115

Stefanie Baum et al.
FULL PAPERS
Determination of the Distribution Coefficients of the
Substrates and Products of the Reaction
The quantification of HCN in the remaining sodium
citrate buffer indicated that fractions of HCN were
also soluble in the ILs (Table 1).
The experiments described above suggested that
Mandelonitrile is produced as intermediate in the
“whole-cell-compatible” solvent systems with the abil- bienzymatic cascade reaction but is prone to decom-
ity to serve as substrate reservoir for benzaldehyde position reactions in aqueous media. Therefore, it was
could be beneficial for the intended biotransforma- important to estimate also the partition of mandeloni-
tion reaction. The water-immiscible ionic liquids 1- trile into the ionic liquid phase (see Experimental
butyl-1-methylpyrrolidinium bis(trifluoromethanesul- Section). The distribution coefficients (logD) of man-
fonyl)imide (BMpl NTf2) and 1-butyl-3-methylimida- delonitrile between the ionic liquids BMim PF6 or
zolium hexafluorophosphate (BMim PF6) (Scheme 2) BMpl NTf2 and sodium citrate buffer were almost
have been previously identified as suitable for syn- identical (1.5, corresponding KBMimPF /W =31.2 and 1.5,
thetic purposes using whole cells of E. coli.[9]
corresponding KBMplNTf2/W =31.6, re6spectively). This
It was therefore attempted to determine the parti- indicated that the majority of mandelonitrile parti-
tion coefficients of the substrates, intermediates, and tioned into the ionic liquid phase.
products of the intended biotransformation reaction
in two-phase systems consisting of an aqueous phase and mandelic acid demonstrated that both reaction
and the respective ionic liquids. products preferentially moved into the aqueous
The respective experiments with mandeloamide
The distribution coefficients were experimentally phase, although mandelic acid was significantly more
determined after intensive mixing of the respective hydrophilic compared to mandeloamide (Table 1).
phases for 1 h. Thus, it was found that the ionic liq- The partitioning of mandelic acid was reanalysed
uids BMim PF6 and BMpl NTf2 could function as effi- after 2, 3 and 18 h of stirring to assure that the equi-
cient substrate reservoirs for benzaldehyde (Table 1). librium was reached after 1 h. These experiments re-
vealed that 1 h of stirring was sufficient to reach equi-
librium conditions (results not shown).
For comparison, the respective values were also de-
termined for the organic solvents heptane and octane,
which would be suitable for whole cell biotransforma-
tion reactions, because of their relatively high logP
values (4–4.5). These experiments demonstrated that
heptane and octane were compared to the ionic liq-
uids less efficient as solvents for benzaldehyde and
HCN (Table 1).
Conversion of Benzaldehyde and Cyanide in the
Presence of the Ionic Liquid BMpl NTf2
E. coli JM109 (pIK9)(pJOE5361.1) was cultivated on
Scheme 2. Ionic liquids used in this study.
a complex medium, the oxynitrilase and nitrilase ac-
Table 1. Distribution coefficients and the corresponding logd values of the substrates and products in different two-phase
systems between ILs or organic solvents and sodium citrate buffer (100 mM, pH 5).
Mandelic acid
Mandeloamide
Benzaldehyde
HCN
KBMim PF /W
logD
0.088Æ0.02
À1.07Æ0.12
0.072Æ0.03
À1.18Æ0.21
0.059Æ0.03
À1.30Æ0.27
0.03Æ0.008
À1.54Æ0.12
3.41Æ0.25[a]
0.96Æ0.02
112.24Æ2.15
2.05Æ0.01
94.53Æ1.93
1.98Æ0.01
12.09Æ0.17
1.08Æ0.01
11.36Æ0.40
1.06Æ0.02
9.8[b]
2.57Æ0.12
0.41Æ0.02
2.03Æ0.12
0.31Æ0.03
0.21Æ0.05
À0.68Æ0.11
0.1Æ0.04
À1.5Æ0.86
9.2
6
À0.02Æ0.01
0.61Æ0.01
KBMpl NTf /W
2
logD
Kheptane/W
logD
Koctane/W
logD
pKa
À0.21Æ0.01
0.011Æ0.002
À1.96Æ0.10
0.003Æ0.0027
À2.34Æ0.08
12.46Æ0.20[a]
[a]
At 258C (these pKa values were obtained from the SciFinder database).
Source of data: http://ecb.jrc.it/IUCLID-Data-Sheet/100527.pdf.
[b]
116
asc.wiley-vch.de
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2012, 354, 113 – 122

Application of a Recombinant Escherichia coli Whole-Cell Catalyst
Figure 3. Conversion of benzaldehyde and cyanide by resting cells of E. coli JM109ACTHUNTGRNEUNG(pIK9)(pJOE5361.1) in an aqueous phase
(A) or a biphasic system with the ionic liquid BMpl NTf2 (B). (A) The reaction mixture contained in a total volume of
0.8 mL 100 mM sodium citrate buffer (pH 5), 10 mM benzaldehyde, 20 mM KCN and an appropriate amount of resting cells
(OD600nm =9). (B) The reaction mixture contained 0.8 mL of resting cells (OD600nm =9.3) suspended in 100 mM sodium cit-
rate buffer (pH 5) and 0.2 mL BMpl NTf2 containing 40 mM benzaldehyde and 80 mM HCN. Samples (90 or 180 mL each)
were taken at the indicated time-intervals and the reactions terminated by the addition of 10 mL or 20 mL 1M HCl. The cells
were removed by centrifugation in an Eppendorf centrifuge (10 min, 13,200 rpm) and the concentrations of benzaldehyde
!
~
*
*
( ), mandeloamide ( ), mandelic acid ( ), and mandelonitrile ( ) were analysed in the supernatants by reversed phase
HPLC.
tivities induced, the cells harvested by centrifugation, loamide and mandelic acid were produced (Figure 3,
washed in Tris/HCl-buffer and subsequently resus- B).
pended in sodium citrate buffer (100 mM, pH 5; see
In the following experiments it was tested if the
Experimental Section). Initially, the reaction rates in system was also functional in the presence of higher
a purely aqueous reaction system were compared substrate concentrations. In addition, in these experi-
with those in the presence of the ionic liquid BMpl ments the substrate and product concentrations were
NTf2. In a purely aqueous phase, resting cells of E. analysed in both reaction phases. Benzaldehyde and
coli JM109ACHTUNGTRENNUNG(pIK9)(pJOE5361.1) converted 10 mM HCN were added to BMpl NTf2 in concentrations of
benzaldehyde and 20 mM KCN completely within 0.7M and 1.4M, respectively. The IL phase (200 mL)
5 min into mandeloamide and mandelic acid with an containing the substrates was mixed with 800 mL of
amide/acid ratio of about 0.6 (Figure 3, A).
the aqueous phase (sodium citrate buffer, 100 mM,
The effect of ionic liquids on the reaction was ini- pH 5) containing the whole-cell catalyst (OD600nm
=
tially analysed by adding benzaldehyde and HCN to 67). At different time intervals samples from both
BMpl NTf2 in concentrations of 40 mM and 80 mM, phases were taken and subsequently analysed by
respectively. The IL phase (200 mL) containing the HPLC (see Experimental Section). The analysis of
enzyme substrates was subsequently mixed with the ionic liquid phase clearly demonstrated that a
800 mL of the aqueous phase (sodium citrate buffer, large proportion of benzaldehyde was converted. The
100 mM, pH 5) containing the whole-cell catalyst. The analysis of the aqueous phase confirmed a significant
formation of the hydrophilic products mandelic acid substrate conversion because high concentrations of
and mandeloamide was quantified by analysing the mandelic acid and mandeloamide were detected
aqueous phase using HPLC. Thus, it was found that (Table 2).
the reaction also occurred in the presence of the ionic
The analysis of the reaction products by using a
liquid but that the observed reaction rates were signif- chiral stationary phase demonstrated that the forma-
icantly lower than those observed in the purely aque- tion of mandelic acid and mandeloamide occurred
ous system (Figure 3, B).
highly enantioselectively and that the respective (S)-
Surprisingly, it was observed that in the two-phase enantiomers were formed consistently with ee values
system the relative amounts of mandeloamide in- ꢀ94% (Table 2).
creased in comparison to the purely aqueous system
and that almost equimolar concentrations of mande-
Adv. Synth. Catal. 2012, 354, 113 – 122
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
117

Stefanie Baum et al.
FULL PAPERS
Table 2. Conversion of benzaldehyde and HCN by resting cells of E. coli JM109ACHTGNUTRENNUNG
with the ionic liquid BMpl NTf2.[a]
t
Concentrations Concentrations in the “Average concentra- MAA
Conversion ee (S)-MA ee (S)-MAA
[min]
in the H2O
phase [mM]
IL phase [mM]
tion” [mM][b]
[%][c]
[%][d]
[%][e]
[%][e]
MA
MAA
MA MAA BA MN MA MAA BA MN
30
60
150
40.0
45.2
46.7
84.3
99.4
102.0
3.0 67.5
2.6 66.1
1.8 72.0
9.2 98.4 32.6 81.0
0.1 5.4 36.7 92.7
1.8 19.7 71
85
97
100
94
95
95
>99
>99
99
0
0
1.1 72
72
0
0
37.7 96.0
0
[a]
The reaction mixture contained dissolved in 200 mL BMpl NTf2, 1.4M of HCN and 0.7M benzaldehyde. The ionic liquid
phase with the substrates was mixed with 800 mL of the aqueous phase (sodium citrate buffer, 100 mM, pH 5) containing
resting cells of E. coli JM109ACHTUNRGTNEUNG(pIK9)(pJOE5361.1) (OD600nm =67). The reactions were performed under a fume hood in
tightly closed glass vials which were placed in a water bath at 308C and vigorously stirred. Samples of the two phases
were removed at the indicated time intervals and the two phases analysed separately by HPLC as described in the Exper-
imental Section. MA mandelic acid, MAA mandeloamide, MN mandelonitrile, BA benzaldehyde.
Average concentration. The results of the concentrations determined for the aqueous and the IL phase were combined
by calculating an “average concentration” of the single compounds in correlation to the complete reaction volume, which
contained the aqueous phase and the ionic liquid phase in a ratio of 4:1 throughout the complete reaction.
[MAA]/[MA]+[MAA].
[MA]+[MAA]/initial “average” benzaldehyde concentration (the initial “average” benzaldehyde concentration was
about 133 mM because the commercially available benzaldehyde contained benzoate, which was not converted during
the reaction).
[b]
[c]
[d]
[e]
The ee values were determined from samples obtained from the aqueous phase.
Biotransformations in the Presence of the Ionic
Liquid BMim PF6
as a third IL. Thus, it was found that in the presence
of BMim NTf2 a similar high degree of substrate con-
version and amide formation was found as with BMpl
In the following experiments the conversion of benz- NTf2 (Table 4). This indicated that the nature of the
ACHTUNGTRENNUNGaldehyde and HCN was performed in the presence of anion mainly influenced the relevant reaction charac-
the second ionic liquid, BMim PF6. The whole-cell teristics.
catalyst performed the biotransformation in the pres-
ence of BMim PF6 as second phase with a similar
degree of enantioselectivity as observed before with Comparison of the Toxic Effects of the Solvents
BMpl NTf2, but the degree of substrate conversion BMpl NTf2, BMim PF6 and Diisopropyl Ether on
was slightly lower than with BMpl NTf2 (Table 3). Whole Cells of E. coli
Furthermore, in comparison to the system with BMpl
NTf2 a slightly lower degree of amide formation was The results described above demonstrated that the
found.
whole cells were efficient catalysts in the presence of
In order to analyse if the differences between BMpl ionic liquids. Therefore, the effects of ionic liquids
NTf2 and BMim PF6 were due to the anions or cations and diisopropyl ether on the survival of the cells were
constituting the ionic liquids, BMim NTf2 was tested compared. Thus, resting cells of E. coli JM109 were
Table 3. Conversion of benzaldehyde and HCN by resting cells of E. coli JM109ACTHNUTRGNEUG(N pIK9)(pJOE5361.1) in a two-phase system
[a]
with the ionic liquid BMim PF6.
t
Concentrations Concentrations in the
“Average concentra-
tion” [mM][b]
MAA
[%][c]
Conversion ee (S)-MA ee (S)-MAA
(min)
in the H2O
phase [mM]
IL phase [mM]
[%][d]
[%][e]
[%][e]
MA
MAA MA MAA BA MN MA MAA BA MN
30
60
120
270
52.4
52.2
53.3
53.4
61.0
65.6
68.0
69.8
3.6 76.2 10.7 125.4 42.6 64.0 2.1 25.1
2.6 70.2 5.3 77.8 42.3 66.5 1.1 15.6
2.4 80.2 3.7 55.7 43.1 70.4 0.8 11.1
3.9 82.6 3.5 51.7 43.5 72.4 0.7 10.3
61
61
62
63
80
82
85
87
95
96
94
94
>99
>99
99
98
[a–e] See the legend of Table 2. The reaction was performed as described in the legend of Table 2 – only BMpl NTf2 was re-
placed by BMim PF6.
118
asc.wiley-vch.de
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2012, 354, 113 – 122

Application of a Recombinant Escherichia coli Whole-Cell Catalyst
Table 4. Conversion of benzaldehyde and HCN by resting cells of E. coli JM109ACTHNUTRGNEUG(N pIK9)(pJOE5361.1) in a two-phase system
[a]
with the ionic liquid BMim NTf2.
t
Concentrations Concentrations in the
“Average concentra-
tion” [mM][b]
MAA
Conversion ee (S)-MA ee (S)-MAA
(min)
in the H2O
phase [mM]
IL phase [mM]
[%][c]
[%][d]
[%][e]
[%][e]
MA
MAA MA MAA BA MN MA MAA BA MN
30
60
120
240
35.2
42.3
42.7
44.4
73.9
92.8
91.7
92.8
1.9 88.6 16.7 154.6 28.5 76.8 3.3 30.9
2.6 91.1 1.3 10.3 34.4 92.5 0.3 2.1
73
73
72
72
79
95
95
97
93
93
94
94
>99
>99
99
1.9 90.7
2.2 91.8
0
0
2.8 34.5 92.4
36.0 92.6
0
0
0.6
0
0
99
[a–e] See the legend of Table 2. The reaction was performed as described in the legend of Table 2 – only BMpl NTf2 was re-
placed by BMim NTf2.
incubated in the presence of 20% (v/v) of BMim PF6, the cells than the ionic liquids. In addition it was
BMpl NTf2 or diisopropyl ether and the number of found that BMim PF6 was slightly more biocompat-
colony forming units (CFU) determined after differ- ible than BMpl NTf2 (Figure 4).
ent incubation times. These experiments clearly estab-
lished that diisopropyl ether was much more toxic to
Discussion
Ionic liquids are receiving an increasing attention as
solvent systems for biocatalytical processes. This is
based on their non-volatile character, thermal stabili-
ty, their ability to efficiently dissolve various com-
pounds and the almost indefinite combinations and
modifications which in principle allows optimising
them as “designer solvents” for specific processes. It
is therefore not surprising that during the last years
several examples have been described for the utilisa-
tion of ionic liquids as solvent systems for enzyme-
catalysed reactions and even an in-vitro application of
the hydroxynitrile lyases from Prunus amygdalus and
Hevea brasiliensis for the synthesis of (R)- or (S)-
mandelonitrile has been described.[10,11] Nevertheless,
there are still only very few examples for the utilisa-
tion of ionic liquids as solvent systems for whole-cell
Figure 4. Comparison of the toxic effects of the solvents
BMpl NTf2, BMim PF6 and diisopropyl ether on whole cells
of E. coli. E. coli JM109 was cultivated on a shaker
biotranformation reactions. Most of these studies de-
scribed the (often asymmetric) reduction of ketones
by different yeasts.[7]
(100 rpm, 308C) in LB medium for 17 h to the late exponen-
Surprisingly, there are only very few examples
tial growth phase. The cells were harvested by centrifuga-
found in the literature which studied the effects of
tion, washed in 1 mL sodium phosphate buffer (50 mM,
pH 7) and finally resuspended in sodium citrate buffer
(100 mM, pH 5) to an optical density (OD600nm) of 1. The
number of viable cells was determined by plating on LB-
ionic liquids on cells of E. coli. Several reports de-
scribed that growth of E. coli is severely impaired in
the presence of even low concentrations of ionic liq-
agar and the cell suspension divided into four aliquots uids and that many ionic liquids even possess an anti-
microbial activity.[8,12,13] On the other hand, it was
(1 mL each). Subsequently, diisopropyl ether, BMim PF6 or
BMpl NTf2 were added (20%, v/v each) to three of the vials.
shown by Pfruender et al. (2006) that in a resting cell
The control experiment did not receive any addition. The
cups were incubated in a water bath at 308C and the reac-
tion mixtures intensively stirred. Aliquots from the aqueous
the cells and that this (and other) ionic liquid(s) ex-
phase (100 mL each) were removed after 1 h and 4 h, the
situation even the addition of 20% (v/v) of Bmim PF6
for 5 h had no detrimental effect on the survival of
hibited complete biocompatibility.[9] There are also
samples diluted and the remaining colony forming units
only very few publications which described the utilisa-
(CFU) determined by plating (LB-agar). The following sym-
tion of recombinant E. coli strains for biotransforma-
tion reactions in the presence of ionic liquids.[8,14,15]
Thus, Brꢁutigam et al. (2007) demonstrated the asym-
bols for the bars were used: black: control; black and white
striped: +20% diisoproyl ether; dark grey: +20% BMim
PF6, light grey: +20% BMpl NTf2.
Adv. Synth. Catal. 2012, 354, 113 – 122
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
119

Stefanie Baum et al.
FULL PAPERS
metric reductions of 4-chloroacetophenone, ethyl 4- their high logP values were less efficient as substrate
chloroacetoacetate, and phenacyl chloride to the cor- reservoirs for benzaldehyde (and HCN). In summary,
responding alcohols by recombinant E. coli strains co- the ionic liquids combined two beneficial effects on
expressing an alcohol dehydrogenase from Lactobacil- the whole-cell catalyst. First, they served as reservoir
lus brevis and a formate dehydrogenase from Myco- for large concentrations of benzaldehyde but only al-
bacterium vaccae.[14] From the distribution coefficients lowed very low concentrations of this toxic substrate
given by the authors it can be deduced that in these in the aqueous phase. Second, the ILs were much
experiments the substrate (ketones) and the products more biocompatible than diisopropyl ether, which is
(alcohols) will preferentially partition into the ionic the “traditional solvent” for oxynitrilase-catalysed re-
liquid phases. Subsequently, it was demonstrated that actions. In combination, these two beneficial effects
ketone concentrations of 600 mM (in the ionic liquid of the ILs allowed the production of synthetically rel-
phases) could be converted by the cells with yields of evant amounts of mandelic acid derivatives in short
about 40–60%.[14] In addition, Cornmell et al. (2008) reaction times with rather high enantiomeric excesses.
utilised a recombinant E. coli strain expressing ring-
hydroxylating dioxygenases for the conversion of tolu-
ene to toluene cis-glycol in the presence of ionic liq- Experimental Section
uids and Schroer et al. (2007) studied the reduction of
1-phenyl-2-propanone to the chiral alcohol in the Chemicals
presence of an ionic liquid by a recombinantly ex-
The ionic liquids 1-butyl-1-methylpyrrolidinium- bis(trifluor-
pressed alcohol dehydrogenase using 2-propanol as
omethanesulfonyl)imide (BMpl NTf2), 1-butyl-3-methylimi-
ultimate source of reduction equivalents.[8,15]
dazolium bis(trifluoromethanesulfonyl)imide (BMim NTf2)
The reaction system described in the present manu-
script differs significantly from most of the previously
decribed two-phase systems applying ILs because in
these preceding studies the organic substrates and the
products usually accumulated in the IL phase. In con-
trast, in the present system one of the reaction sub-
strates (HCN) and both products (mandelic acid and
mandeloamide) partitioned to significant concentra-
tions into the aqueous phase. Thus, the ionic liquid
phase mainly served as substrate reservoir for the
second substrate benzaldehyde which exhibited in
purely aqueous systems a pronounced inhibitory
effect on the whole-cell catalyst. Furthermore, the re-
sults obtained from the logD determination and from
the experiments concerning the lethal effects of the
ILs on whole cells of E. coli JM109 suggested that the
ionic liquid Bmim PF6 should be best suited for the
intended biotransformation reaction. However, the
biocatalyst was slightly more effective in the presence
of the ionic liquid BMpl NTf2. Therefore, these pa-
rameters (log D and biocompatability) were not suffi-
cient for an estimation of the best suited IL for the
entire reaction system which confirmed results previ-
ously obtained by Brꢁutigam et al. (2007).[14]
and
1-butyl-3-methylimidazolium
hexafluorophosphate
(BMim PF6) were purchased from Merck (Darmstadt, Ger-
many). The sources of all other chemicals have been de-
scribed before.[2,3,16]
Bacterial Strains and Culture Conditions
Escherichia coli JM109
(pJOE5361.1) were used for all experiments.[5,16] E. coli
JM109(pIK9) and E. coli JM109(pIK9)(pJOE5361.1) were
ACHTUGTNREN(UNG pIK9) or E. coli JM109ACHTUNGTRENNUNG(pIK9)
G
ACHTUNGTRENNUNG
routinely grown at 308C in LB medium or 198C in dYT
medium, respectively. The media were supplemented with
100 mgmLÀ1 ampicillin (plus 20 mgmLÀ1 chloramphenicol if
appropriate).[17] The induction of the nitrilase and oxynitri-
lase activities was achieved by the addition of rhamnose
(0.2% w/v) to the growth medium as described previously.[5]
The cells were harvested by centrifugation in the early sta-
tionary growth phase, washed once with Tris/HCl buffer
(50 mM, pH 7.5) and frozen at À708C.
Determination of the Distribution Coefficients of
Substrates and Products
The distribution coefficients (logd values) between ionic liq-
uids or organic solvents and sodium citrate buffer were de-
termined in mixtures consisting of 80% (v/v) sodium citrate
buffer (100 mM, pH 5) containing 10 mM of the compound
under investigation and 20% (v/v) of the ionic liquids or or-
ganic solvents in a gas-tight vial with stir bar. The samples
were placed in an oilbath at 308C and vigorously stirred for
1 h. After centrifugation (10 min, 13,200 rpm), the solute
concentrations in the aqueous phase were determined using
HPLC. Subsequently the logd values were calculated.
The determination of the distribution coefficients of man-
delonitrile were hampered as a-hydroxynitriles are unstable
at neutral pH values and are subject to chemical decomposi-
tion even at pH 5. Therefore, the reaction mixtures contain-
ing 10 mM mandelonitrile (dissolved in 100 mM sodium cit-
Conclusions
It was demonstrated that benzaldehyde exerted a
much stronger inhibitory effect than HCN on the ni-
trilase activity of recombinant E. coli cells. The toxic
effects of benzaldehyde could be significantly reduced
in two-phase systems with ionic liquids (BMpl NTf2,
BMim PF6, BMpl NTf2). In contrast, the organic sol-
vents heptane and octane which should be suitable
for whole-cell biotransformation reactions because of rate buffer, pH 5) and 20% (v/v) ionic liquid (BMpl NTf2 or
120
asc.wiley-vch.de
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2012, 354, 113 – 122

Application of a Recombinant Escherichia coli Whole-Cell Catalyst
BMim PF6) were vigorously shaken by hand for a few sec-
For the determination of the partition coefficients a Chro-
onds, and the distribution coefficients determined by direct-
ly injecting an aliquot (10 mL) of the aqueous phase after
centrifugation (13,200 rpm, 10 min, 48C). All samples were
kept on ice prior to injection.
molith SpeedRod RP-18 (50ꢃ4.6 mm) column in combina-
tion with a solvent system consisting of water plus 0.1%
AHCTUNGTRENNUNG
Cyanide concentrations in the aqueous phase were deter-
mined in a total volume of 1 mL by using the Spectroquant
cyanide test kit from Merck (Darmstadt, Germany).
were separated on a Chiral-HSA column (ChromTech AB,
Hꢁgersten, Sweden). The mobile phase consisted of sodium
phosphate buffer (50 mM, pH 7) plus 3.0% (v/v) acetoni-
trile. The detection was performed at 210 nm.
Extraction of HCN into the Ionic Liquids
Calculation of Substrate Turnover and Product
Formation
In order to reach a defined concentration of HCN in the
ionic liquid phase a modification of the procedure described
by van Langen et al. (2003) was used.[18] Thus, 130 mg of po-
tassium cyanide and 220 mg of citric acid·H2O were dis-
solved in 0.36 mL of water. The solution was cooled in an
ice-water bath and stirred with 0.4 mL of the ionic liquid
used. The IL phase with the dissolved HCN was removed
and the concentration of HCN in the ionic liquid deter-
mined by using a commercial cyanide test system (Spectro-
quant, Merck, Darmstadt). The HCN solution was subse-
quently brought to the desired concentration by the addition
of the ionic liquid under investigation.
The concentrations of the individual compounds in the two
phases were separately determined by HPLC as described
above. The results of these quantifications were combined
by calculating an “average concentration” of the single com-
pounds in correlation to the complete reaction volume,
which contained the aqueous phase and the ionic liquid
phase in a ratio of 4:1 throughout the complete reaction.
Thus, if at a certain time point the mandelic acid concentra-
tion in the aqueous phase was determined as 50 mM and in
the IL phase as 5 mM an “average concentration” of 41 mM
was calculated.
Benzaldehyde was directly added to the ionic liquids to
the desired concentration.
Biotransformation Reactions
Acknowledgements
The reactions were performed in a fume-hood which was
monitored by a cyanide detector (Draeger, Lꢀbeck, Germa-
ny). For the biotransformation reactions cell pellets of E.
This work was supported by the Deutsche Forschungsgemein-
schaft (DFG) and the COST Action CM0701.
coli JM109ACHTUNGTRENNUNG(pIK9)(pJOE5361.1) were suspended in sodium
citrate buffer (100 mM, pH 5) and diluted to appropriate op-
tical densities. Subsequently, 800 mL of the cell suspension
were combined with 200 mL of the ionic liquid which con-
tained the substrates (benzaldehyde and HCN). The reac-
tion mixtures were incubated in 2-mL tightly-closed glass
vials and intensively stirred in a water bath at 308C. At the
indicated time intervals aliquots were taken (108 mL aque-
ous phase and 27 mL IL phase). The reaction was stopped in
the aqueous phase by the addition of 12 mL of 1M HCl. The
samples (aqueous and IL phases separately) were then cen-
trifuged (10 min, 48C, 13,500 rpm). The supernatants ob-
tained from the centrifugation of the aqueous phases were
then diluted and analysed by HPLC. The samples taken
from the IL phases consistently demonstrated after the cen-
trifugation the presence of a residual fraction of the aqueous
phase, which was removed. Aliquots of the IL phases were
subsequently diluted with methanol and analysed by HPLC.
References
[1] C. Mateo, A. Chmura, S. Rustler, F. van Rantwijk, A.
Stolz, R. A. Sheldon, Tetrahedron: Asymmetry 2006,
17, 320–323.
[2] C. Laane, S. Boeren, K. Vos, C. Veeger, Biotechnol.
Bioeng. 1987, 30, 81–87.
[3] D. Costes, E. Wehtje, P. Adlercreutz, Enzyme Microb.
Technol. 1999, 25, 384–391.
[4] a) N. Layh, A. Willetts, Biotechnol. Lett. 1998, 20, 329–
331; b) U. Heinemann, D. Engels, S. Bꢀrger, C. Kiziak,
R. Mattes, A. Stolz, Appl. Environ. Microbiol. 2003, 69,
4359–4366; c) P. Kaul, U. C. Banerjee, J. Ind. Microbiol.
Biotechnol. 2008, 35, 713–720.
[5] O. Sosedov, K. Matzer, S. Bꢀrger, C. Kiziak, S. Baum,
J. Altenbuchner, A. Chmura, F. van Rantwijk, A. Stolz,
Adv. Synth. Catal. 2009, 351, 1531–1538.
Analytical Methods
[6] O. Sosedov, S. Baum, S. Bꢀrger, K. Matzer, C. Kiziak,
A. Stolz, Appl. Environ. Microbiol. 2010, 76, 3668–
3674.
Benzaldehyde, mandelonitrile, mandelic acid and mandeloa-
mide were analysed by HPLC. The achiral analysis was rou-
tinely performed using a Pro-C18 Aq-column (Trentec Ana-
lysentechnik, Rutesheim, Germany) with methanol (40%
v/v) and acetic acid (0.25% v/v) in H2O as mobile phase. In
addition, for certain experiments a Lichrospher 100 RP-18
(250ꢃ4 mm, 5 mm) column was used together with a solvent
system consisting of water/methanol/phosphoric acid
(50:50:0.3 v/v/v). In most cases a flow rate of 0.6 mLminÀ1
and a detection wavelength of 254 nm were applied.
[7] a) J.-Y. He, L.-M. Zhou, P. Wang, L. Zu, Proc. Bio-
chem. 2009, 44, 316–321; b) J. Howarth, P. James, J.
Dai, Tetrahedron Lett. 2001, 42, 7517–7519; c) W.-Y.
Lou, L. Chen, B.-B. Zhang, T. J. Smith, M. H. Zong,
BMC Biotechnol. 2009, 9, 90; d) W.-Y. Lou, M. H.
Zong, T. J. Smith, Green Chem. 2006, 8, 147–155; e) T.
Matsuda, Y. Yamagishi, S. Koguchi, N. Iwai, T. Kita-
zume, Tetrahedron Lett. 2006, 47, 4619–4622; f) W.
Adv. Synth. Catal. 2012, 354, 113 – 122
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
asc.wiley-vch.de
121

Stefanie Baum et al.
FULL PAPERS
Wang, M. H. Zong, W.-Y. Lou, J. Mol. Catal. B:
Enzym. 2009, 56, 70–76.
[11] R. P. Gaisberger, M. H. Fechter, H. Griengl, Tetrahe-
dron: Asymmetry 2004, 15, 2959–2963.
[8] a) R. J. Cornmell, C. L. Winder, S. Schuler, R. Good-
[12] S.-M. Lee, W.-J. Chang, A.-R. Choi, Y.-M. Koo, Korean
ACHTUNGTRENNUNGacre, G. Stephens, Green Chem. 2008, 10, 685–691;
J. Chem. Eng. 2005, 22, 687–690.
b) R. J. Cornmell, C. L. Winder, C. J. T. Tiddy, R.
Goodacre, G. Stephens, Green Chem. 2008, 10, 836–
841.
[13] F. Ganske, U. Bornscheuer, Biotechnol. Lett. 2006, 28,
465–469.
[14] S. Brꢁutigam, S. Bringer-Meyer, D. Weuster-Botz, Tet-
rahedron: Asymmetry 2007, 18, 1883–1887.
[15] K. Schroer, E. Tacha, S. Lꢀtz, Org. Process Res. Dev.
2007, 11, 836–841.
[9] a) H. Pfruender, R. Jones, D. Weuster-Botz, J. Biotech-
nol. 2006, 124, 182–190; b) D. Weuster-Botz, Chem.
Rec. 2007, 7, 334–340.
[10] a) S. A. Gangu, L. R. Weatherley, A. M. Scurto, Curr.
Org. Chem. 2009, 13, 1242–1258; b) S. Oppermann, F.
Stein, U. Kragl, Appl. Microbiol. Biotechnol. 2011, 89,
493–499; c) R. A. Sheldon, R. M. Lau, M. J. Sorge-
[16] C. Kiziak, D. Conradt, A. Stolz, R. Mattes, J. Klein, Mi-
crobiology 2005, 151, 3639–3648.
[17] J. Sambrook, E. F. Fritsch, T. Maniatis, Molecular clon-
ing: a laboratory manual, 2nd edn., Cold Spring
Harbor Press, Cold Spring Harbor, N.Y, 1989.
[18] L. M. van Langen, F. van Rantwijk, R. Sheldon, Org.
Process Res. Dev. 2003, 7, 828–831.
ACHTUNGTRENNUNGdrager, F. van Rantwijk, K. R. Seddon, Green Chem.
2002, 4, 147–151; d) F. van Rantwijk, R. M. Lau, R. A.
Sheldon, Trends Biotechnol. 2003, 21, 131–138; e) F.
van Rantwijk, R. A. Sheldon, Chem. Rev. 2007, 107,
2757–2785.
122
asc.wiley-vch.de
ꢂ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Adv. Synth. Catal. 2012, 354, 113 – 122