Reaction engineering of biocatalytic enantioselective reduction: A case study for aliphatic ketones

Article abstract of DOI:10.1021/op400117t

Previously, it could be demonstrated, that the monophasic, enzymatic reduction of aliphatic 2-ketones into the corresponding (R)-2-alcohols is an adequate and viable method as carried out in a cascade of two enzyme-membrane reactors (Leuchs, S.; Na'amnieh, S. N.; Greiner, L. Green Chemistry 2013, 15, 167-176.). In the present work, the process metrics of the ketone reduction were calculated. A cost analysis revealed that the enzyme costs are negligible, but the cost for nicotinamide cofactor NADP⁺ is dominating the overall cost of the chemical raw material followed by the ionic liquid (TEGO IL K5) used as solubiliser and the buffer. The overall cost of chemicals was €148/kg_product. To assess the environmental impact of the process, the E-factor (kg_waste/kg_product) 132 and the process mass intensity 133 (PMI, kg_substrate/kg_product) were calculated. A process model based on initial rate experiments was elaborated and used to improve the process under cost and environmental aspects. Applying several measures to enhance the cofactor utilisation, the cost base could be reduced by 65% and the E-factor (PMI) to 17 (18).

Full text of DOI:10.1021/op400117t

Article
pubs.acs.org/OPRD
Reaction Engineering of Biocatalytic Enantioselective Reduction: A
Case Study for Aliphatic Ketones
Susanne Leuchs,† Joana Lima-Ramos,‡ Lasse Greiner,*,§ Naweed Al-Haque,‡ Par Tufvesson,‡
̈
and John M. Woodley*,‡
†DECHEMA Research Institute, Frankfurt am Main, Germany
‡Center for Process Engineering and Technology, Department of Chemical and Biochemical Engineering, Technical University of
Denmark, DK-2800 Lyngby, Denmark
§Mannheim University of Applied Sciences, Paul-Wittsack-Strasse 10, D-68163 Mannheim, Germany
ABSTRACT: Previously, it could be demonstrated, that the monophasic, enzymatic reduction of aliphatic 2-ketones into the
corresponding (R)-2-alcohols is an adequate and viable method as carried out in a cascade of two enzyme−membrane reactors
(Leuchs, S.; Na’amnieh, S. N.; Greiner, L. Green Chemistry 2013, 15, 167−176.). In the present work, the process metrics of the
ketone reduction were calculated. A cost analysis revealed that the enzyme costs are negligible, but the cost for nicotinamide
cofactor NADP+ is dominating the overall cost of the chemical raw material followed by the ionic liquid (TEGO IL K5) used as
solubiliser and the buffer. The overall cost of chemicals was €148/kgproduct. To assess the environmental impact of the process, the
E-factor (kgwaste/kgproduct) 132 and the process mass intensity 133 (PMI, kgsubstrate/kgproduct) were calculated. A process model
based on initial rate experiments was elaborated and used to improve the process under cost and environmental aspects. Applying
several measures to enhance the cofactor utilisation, the cost base could be reduced by 65% and the E-factor (PMI) to 17 (18).
of all parameters, which rapidly exceeds experimental capacity
and time. In such cases, process modeling can help to reduce
the experimental effort and to quickly assess process
alternatives or improvement strategies. In doing so, it is
important to carefully define the aim of the process
development. For example, this can be based on space-time-
yield (STY), enzyme productivity, a decrease in environmental
impact (in order to make the calculation quick and easy the E-
factor defined by Sheldon10,11 can be used for a primary
environmental assessment), conversion, reduced cost for raw
material, reduction of side-product formation, or ease of
downstream processing. According to the literature, we listed
key values to be obtained for an economically viable industrial
process. However, these values are guidelines that depend on
the desired product (Table 1). Defining maximum conversion
as the modelling goal will always lead to higher residence times
and catalyst concentrations. This will normally not lead to a
INTRODUCTION
■
Today, biocatalysis is not restricted to lab-scale use and
fundamental academic research but has made its way to
mainstream industrial production.1,2 Prominent examples of
large-scale biocatalytic processes yielding low-cost products are,
for example, the production of acrylamide, L-aspartate, and high
fructose corn syrup.2,3 Nevertheless, the unique stereo-
selectivity (enantio- as well as diastereoselectivity) and
regioselectivity of most enzymes makes biocatalysis most
suitable for the pharmaceutical and fine chemical industries.4,5
While hydrolases dominate process biocatalysis, several
oxidoreductases have also made it to the production level,3,5,6
mainly because reductive cofactor regeneration can be routinely
performed.7−9 Most of the biocatalytically produced substances
possess at least one chiral centre in which the stereo centre
originates from the biogenic starting material or alternatively is
created during the biocatalytic conversion.5
In view of limited patent lifetimes, especially in pharmaceut-
ical development, time to market is an important issue.
Typically, at the beginning subkilogram quantities are sufficient
to allow for primary assessment of the product. However, in
order to make the most of the patent lifetime, soon
multikilogram amounts of the respective product are required.4
Catalysts in general, and enzymes in particular, respond to
changes in reaction conditions sensitively. At a minimum, a
reaction comprises the enzymes, oxidising and reducing agents,
and their respective products. This alone imposes a multivariate
nonlinear optimisation problem for productivity and stability.
From a practical perspective, other substances, e.g. cofactors in
all forms, buffer, solubilisers, and other additives, as well as
temperature and pH will further add to the overall complexity.
In order to assess the best conditions for production, hundreds
of experiments can easily be necessary to cover the whole range
Table 1. Key values to be obtained for an economically
viable (bio-)catalytic industrial process
bulk
104−106
fine-chemical
102−104
pharma
10−103
product amount
(t per year)10
typical cost12 /(€/kg)
E-factor10
1(−5)
<1−5
>15
>100
5−50
25−100
(kgwaste/kgproduct
)
STY5 /g L−1 h−1
>0.1
>0.001
>0.1
5
cProduct /g L−1
>1
catalyst productivity/
5000−20000
670−1700
100−250
12
(kgproduct/kgenzyme
)
Received: May 3, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 1. Proposed work flow for optimisation.
Figure 2. Reaction scheme of the synthesis of aliphatic, enantiopure alcohols.
more economic process. Thus, in this case, a conversion range
EXPERIMENTAL SECTION
■
should be used. Most probably, all aspects end up leading to
lower overall production costs, which may be the result of
reduced costs of the raw materials or a reduction in the waste
produced and thus lower costs for wastewater treatment and
waste disposal (see also ref 2). For this reason a scheme to
improve and develop a process based on pre-existing
knowledge was derived and has been applied throughout the
present paper (Figure 1).
All continuous and initial rate experiments as well as stability
tests were carried out as described by Leuchs et al.1 All
chemicals were purchased from Carl-Roth, Karlsruhe, Germany.
C8-alcohols and -ketones were from Sigma-Aldrich, Steinheim,
Germany. Enzymes were from X-Zyme GmbH/Johnson
Matthey Catalysts, Dusseldorf, Germany. The parameter
̈
estimation, process modelling, and process improvements
were performed using MATLAB2009a from Mathworks.
In this work we focused on the enantioselective reduction of
aliphatic ketones into their corresponding alcohols using
oxidoreductases, since there is no chemical alternative available
that yields similar enantiomeric excess (ee).13−16 Long-chain
enantiopure aliphatic alcohols are an interesting raw material in
the pharmaceutical industry. They have also been used for the
production of liquid crystals with interesting optical proper-
ties.17
RESULTS AND DISCUSSION
■
Process Pre-existing Knowledge. Alcohol dehydrogenase
from Lactobacillus brevis (LbADH, EC 1.1.1.2) exhibits
outstanding enantioselectivity, robustness, and flexibility in a
wide variety of reaction conditions.18 The substrate scope of
this enzyme is broad, and its ability to convert long-chain
aliphatic ketones into the desired (R)-2-alcohols has been
proven in many batch syntheses,19,20 as well as in mono-
phasic,1,19 and biphasic20−22 continuous experiments. Water-
B
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
solubility restrictions in single-phase approaches can be
overcome by applying an ionic liquid (IL) as a solubiliser.1,19,23
Kinetics21,23 and stability tests1,23 reveal the applicability of
LbADH in many processes. Among these possibilities, the
monophasic continuous synthesis with an ionic liquid as
solubiliser1 was chosen for further assessment and develop-
ment.
LbADH requires NADPH as a cofactor or redox
equivalent.18,24 For in situ cofactor regeneration, an enzyme-
coupled approach with glucose dehydrogenase from Bacillus
spp. (GDH, EC 1.1.1.47) was chosen (Figure 2). To enhance
the solubility of the aliphatic compounds (2-ketones and 2-
alcohols), a solubiliser was used, in this case the ionic liquid
(IL) TEGO IL K5.1,19 A cascade of two enzyme−membrane
reactors (EMRs) was employed for the synthesis of
enantiopure (R)-2-alcohols (2-octanol, 2-nonanol, 2-decanol)
(Figure 3). In the outlets of reactors 1 and 2, flow cells were
Table 2. Base case conditions for the continuous synthesis of
(R)-2-octanol
substance
concentration
a
ADA-buffer
MgCl2
150 mmol L−1
20 mmol L−1
100 g L−1
60 mmol L−1
200 mmol L−1
66.7 mg L−1
280 mg L−1
0.1 mmol L−1
values
TEGO IL K5
2-octanone
glucose
LbADH
GDH
NADP+
other conditions
pHinlet
7.5
4 mL h−1
3.75 h
25 °C
̇
V
τ
T
a
ADA = N-(2-acetamido)iminodiacetic acid.
Figure 3. Flow-scheme of the process to be improved according to
Leuchs et al.;1 S = substrate solution, SP = syringe pump, P = product,
B = NaOH solution, W = waste, EMR = enzyme−membrane reactor, f
= flow cell, SPE = solid-phase extraction.
integrated for automatic online GC measurements. Due to the
stoichiometric formation of gluconic acid and the resulting
decrease in pH, the pH had to be adjusted to recycle the
aqueous stream. This was achieved by dosing NaOH solution
with the aid of a pH-stat immediately after reactor 2. The
product and remaining substrate were removed by passing the
solution through a stainless steel column filled with solid-phase
extraction (SPE) material, HR-P (Macherey-Nagel). The
product was eluted from the column by washing with n-
heptane which was selected because of its lower toxicity
compared to that of n-hexane.25 Ninety percent of the aqueous
stream was recycled after recharging with 2-octanone, glucose
(via fresh solution), and NADP+. Previous work led to the
choice of the reaction conditions, such as the best buffer, pH,
type and concentration of the salt, and type and concentration
of the solubiliser.1 All these conditions elicited a very stable and
promising process with more than 1000 h of continuous
operation and very low deactivation. Thus, this process was
chosen for further development with 2-octanone as the model
substrate representing the class of poorly water-soluble aliphatic
ketones. The continuous synthesis of (R)-2-octanol was taken
as the base case; its conditions are detailed in Figure 21 and
Table 2. All alternatives were compared to the results obtained
for these conditions.
Figure 4. Cost distribution for the base case as presented in Leuchs et
al.;1 base case: without recycling of the aqueous phase; recycling: with
recycling of the aqueous phase; recycling: with recycling of the
aqueous phase; improved: improved synthesis with reduced cofactor
concentration.
NADP+, accounting for 36.7% of the entire expenditure, the
ionic liquid TEGO IL K5 (20.4%), and the ADA-buffer26
(26.3%). The overall cost was €149/kgproduct. In contrast to
many other biocatalytic processes, the cost analysis revealed
that the enzyme costs were relatively small. Their contribution
to the overall cost was less than 1% . Depending on the case, for
a fine chemical, enzyme, or catalyst the costs can be up to 5% of
the selling costs.12 Thus, the enzyme productivity is not subject
to possible improvements.
Environmental Assessment. In general, a process should be
judged not only by its costs but also by its impact on the
environment. An easy method to get a first idea about a
process’s environmental impact is the E-factor (kgwaste/kgproduct
)
introduced by Sheldon.10,11,27 In contrast to a full assessment of
the environmental impact of a process, the E-factor only
considers the mass of waste. The American Chemical Society
Green Chemistry Institute Pharmaceutical Roundtable favors
another measure, the process mass intensity (PMI).28 To allow
for a comparison with other processes, we calculated both the
E-factor and PMI of the respective experiments. The E-factor
for the base case was 132 (including water), and the PMI, 133.
Amongst others,10,11 benchmarks for the E-factor were given.
Process Evaluation. Economic Assessment. In order to
identify possible bottlenecks in the base process,1 a cost analysis
was carried out, taking into account the costs of all the
chemicals (substrates, buffer, enzymes, and cofactor) (Figure
4). The substances which dominated cost were the cofactor
C
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 5. Effect of confidence intervals on the conversion (grey lines) in comparison to the predicted conversion using the estimated parameters
−1
̇
(black lines) and measured conversion (black dots) in batch (left) and continuous synthesis (right, see also ref 1). V = 4 mL h ; c2‑octanone = 60
mmol L−1; cLbADH = 50 mg L−1; cGDH = 250 mg L−1 (batch); cGDH = 250 mg L−1 (conti) ; ttotal = 10 h (batch); ttotal = 1000 h (conti); cNADP = 0.05
+
mmol L−1 (batch); cNADP = 0.1 (conti).
+
For a bulk chemical it should not exceed 5; for fine chemicals a
value between 5 and 50 is reasonable; for a pharmaceutical,
values up to 100 are standard. Thus, even for a fine chemical, an
E-factor of 132 and thus a PMI of 133 are too high and should
be decreased. Sheldon11 suggested not to include water into the
calculation of the E-factor because exceptionally high E-factors
may be obtained with biocatalytic processes. Whether water can
be left out or should be included certainly depends on the
process considered and on the available separation techniques.
As we only assessed and compared improvements in a single
process scheme, the inclusion or exclusion of water did not
change the results qualitatively. However, as the solubiliser is
considered toxic to aquatic organisms, water was included into
the E-factors given throughout.
Model-Aided Process Optimisation. When optimising a
process, modeling can help to reduce experimental effort by
avoiding time-consuming experiments and thus save money
and resources. On the basis of initial rate experiments and a few
batch experiments, a process model was developed (see also
Leuchs et al.1). The quality of the model was tested by
comparing experimental results with a prediction using the
estimated parameters (Figure 5 and Table 3). Given that the
model was based on initial rates and contained no further
correction factors,33 the predictions were very good. Upper and
lower bounds were determined by numerical inspection of all
131072 possible combinations of errors, yielding the confidence
interval for the batch and the continuous synthesis shown in
Bottleneck Analysis. In order to reduce the cost, the E-
factor, and the consumption and release of the IL, which is
possibly harmful to the environment,29 recycling of 90% of the
aqueous product stream was carried out. The effect of this
recycling step on the amount of waste produced and chemicals
consumed to produce 1 kg of product was enormous. The E-
factor could be reduced from 132 to 27, while the PMI was
reduced from 133 to 28. More benefits of this step are
demonstrated in Figure 4.
The overall cost was reduced by 36.5% to €95/kgproduct. Due
to the recycling, the contribution of the ADA buffer was
lowered to 8.8% and that of the IL to 7.0%. Although the
absolute contribution of the cofactor was not changed, its
contribution to the overall cost was increased to 63.5% due to
the fact that the IL and the buffer were recycled but not the
cofactor. Less costly than NADP(H) is NAD(H),30 and the
preference of LbADH for NADP(H) is regarded as its major
drawback. So far, attempts to change the cofactor preference to
NAD(H) have not been successful, and the resulting enzyme
variants are not commercially available.31,32 Instead of relying
on time-consuming molecular biology to change the cofactor
preference of the enzyme, we decided to use chemical
engineering tools to ameliorate the productivity of the cofactor.
In that way, we also wanted to demonstrate that an enzyme’s
preference for NADP(H) instead of NAD(H) is not a general
knockout criterion. Thus, the cofactor-consumption per
kilogram of product was further optimised.
Table 3. Kinetic constants for LbADH and GDH measured
with fluorescence spectroscopy at 25 °C
a
constant
constant value
confidence interval
LbADH
17.50 μmol min−1 mg−1
0.206 mmol L−1
0.03704 mmol L−1
0.208 mmol L−1
0.212 mmol L−1
163 mmol L−1
̇
Vmax,forw
0.13
0.017
0.00065
0.040
0.028
10
KM2‑octanone
KP2‑octanol
KMNADPH
+
KPNADP
KS2‑octanone
9.96 μmol min−1 mg−1
0.0332 mmol L−1
0.835 mmol L−1
0.0122 mmol L−1
0.359 mmol L−1
6065 mmol L−1
GDH
0.18
̇
Vmax,back
KM2‑octanol
KP2‑octanone
0.0014
0.049
0.0021
0.030
4040
+
KMNADP
KPNADPH
KS2‑octanol
Vmax,forw
5.82 μmol min−1 mg−1
2.73 mmol L−1
0.0258 mmol L−1
0.028 mmol L−1
3780 mmol L−1
62260 mmol L−1
0.053
0.23
̇
KMGlucose
KPGDL
0.00067
0.0033
279
11.1 × 108
+
KMNADP
KPNADPH
KSGlucose
a
cLbADH = 1.25 mg L−1; cGDH = 5 mg L−1; cADA‑buffer = 100; cMgCl = 10
2
mmol L−1; cTEGO IL K5 = 100 g L−1; pH = 7.0.
D
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 5. In the batch experiment, a deviation between the
experimental data and the model-based predictions was
observed at higher conversions, which was most likely caused
by a pH shift that was not included in the model. The
beginning of the experiment was well predicted. In the
continuous synthesis, the conversion and the overall trend
were well represented. Thus, the model could be used to obtain
reliable predictions and for further process optimisation.
The main aim of the optimisation procedure is to enhance
the cofactor utilisation. In order to enhance the amount of
product per amount of cofactor consumed, there are basically
two promising strategies. The first one is to increase the
substrate/cofactor ratio, because the stoichiometry limits the
maximum possible turnover number for the cofactor
and a lower pH will enhance the stability of the oxidised
cofactors.29 In order to quantify these trends, we investigated
the half-lives of the oxidised and reduced cofactors under
conditions that were relevant for the process (Figure 7). The
+
(TONNADP ). The second strategy is to recycle the cofactor
together with the aqueous stream.
The substrate/cofactor ratio can be increased by increasing
the substrate concentration or by decreasing the cofactor
concentration. As 2-octanone is a poorly water-soluble
substrate, its concentration cannot be further increased without
increasing IL concentration beyond a practical viscosity limit.
Therefore, the cofactor concentration has to be reduced. As the
reduction reaction rate strongly depends on the cofactor
concentration, a decrease of the latter may lead to a reduced
reaction rate. In order to estimate the effect of a decreased
cofactor concentration, the process model was consulted to
predict the conversion in reactors 1 and 2 as a function of the
cofactor inlet concentration (Figure 6). The calculations
indicated that reducing the cofactor concentration from 100
μmol L−1 to 50 μmol L−1 affected only marginally the
conversion in both reactors.
Figure 7. Half life of NADP+ (circles) and NADPH (squares) as a
function of pH, T = 25 °C; cADA = 150 mmol L−1; cMgCl = 20 mmol
2
L−1; pH = 6.0−7.5; values for NADP+ partly extrapolated, measuring
time was 1950 h.
half-life of the oxidised cofactor NADP+ was longer than 1000 h
in the pH range of 6.0−7.5. Thus, the stability of NADP+ was
not limiting in this case. The stability of NADPH varies from
less than 10 h at pH 6.0 to almost 150 h at pH 7.5. Hence, in
order to ameliorate the cofactor stability and thereby its
recyclability, the pH has to be increased to the inlet pH of 7.5
immediately after the product stream leaves the reactor. The
huge difference in the stability of NADPH and NADP+ can be
utilised as well to improve recyclability. By varying the
LbADH/GDH ratio in reactor 2, the steady-state ratio of
NADP+/NADPH can be shifted to higher NADP+ and lower
NADPH concentrations. Thus, the overall stability of the
cofactor is increased and allows for a decreased NADP(H)
replacement in the recycled solution.
The effects of varying LbADH-to-GDH ratios on the
NADP+/NADPH ratio and the conversion in reactor 2 were
predicted using the aforementioned process model (Figure 8).
A low LbADH/GDH ratio in reactor 1 in combination with a
high LbADH/GDH ratio in reactor 2 turned out to be
beneficial for the NADP+/NADPH ratio in reactor 2, allowing
for high recyclability, as well as for the conversion in reactor 2,
which is mandatory to conduct a reasonable synthesis and
facilitate downstream processing. Unfortunately, the regime
with high conversion (Figure 8 right) did not overlap with the
regime of a high NADP+/NADPH fraction (Figure 8 left).
Considering the results from modelling and the inves-
tigations on cofactor stability, another experiment was
designed, aiming for a higher cofactor utilisation. Measures to
improve cofactor productivity included a lower cofactor inlet
concentration of 50 μmol L−1 instead of 100 μmol L−1, and
enzyme concentrations changed from 66 mg L−1 for LbADH
and 280 mg L−1 for GDH in both reactors to 50 mg L−1 for
LbADH and 250 mg L−1 for GDH in reactor 1 and 250 mg L−1
for LbADH and 100 mg L−1 for GDH in reactor 2. To make
use of the higher NADPH stability at elevated pH values, the
pH was adjusted directly after reactor 2 instead of fixing it after
Figure 6. Estimated mean conversion in reactor 1 and reactor 2 as a
̇
function of the cofactor concentration in the feed stream. V = 4 mL
h−1; c2‑octanone = 60 mmol L−1; cLbADH = 50 mg L−1; cGDH = 250 mg L−1;
ttotal = 1000 h.
In view of recycling, it is known that nicotinamide cofactors
possess only restricted stability in aqueous solutions, in which
the phosphorylated cofactors (NADP(H)) are less stable than
the nonphosphorylated ones (NAD(H)) and the reduced ones
(NAD(P)H) are less stable than the oxidised ones (NAD(P)+)
under typical, near-neutral pH.8,34 Parameters strongly
influencing the stability of cofactors are pH and temperature.
While decreasing the temperature will increase stability, a
higher pH will increase the stability of the reduced cofactors,
E
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 8. Estimated NADP+/NADPH fraction in reactor 2 (left) and mean conversion in reactor 2 (right) as a function of the LbADH/(LbADH +
−1
GDH) fraction in reactors 1 and 2. V = 4 mL h ; c2‑octanone = 60 mmol L−1; cLbADH = 10−280 mg L−1; cGDH = 300 − cLbADH mg L−1; ttotal = 1000 h;
̇
cNADP = 0.05 mmol L−1; black star: conditions base case; white star: conditions improved synthesis.
+
the adsorption unit. The pH-stat unit was kept at 5 °C, and the
experimental setup was optimised to reduce void volume in
order to reduce cofactor residence time in the recycling loop.
Evaluation of the Improved Process. A continuous
experiment was conducted, applying the conditions elaborated
in the previous section. The enzyme ratio in the second reactor
and the cofactor inlet concentration were changed (Figure 9).
compared to base case) (Figure 4). The costs per kilogram of
product in our case exceed €15/kgproduct given in Tufvesson et
al.12 as a minimum selling price for a fine chemical. However, as
the costs of the substrates 2-octanone and glucose already lie in
the same range (Figure 10), €15/kgproduct is not achievable in
this case. Typically, enantiopure alcohols achieve higher prices,
judging from related examples.
Future Perspectives. The overall system for the
continuous synthesis of enantiomeric pure alcohols has a high
level of complexity and leads to a multiparameter optimisation
problem. In order to guide the process towards an economically
viable one, a sensitivity analysis was carried out (see Table 4).35
In the study of Degenring et al.35 conversion and enantiomeric
ratio were the evaluation targets. In the present investigation,
the enantioselectivity of the enzyme was not an issue. In all
cases, the measured ee was always ≥99.5% for long-chain
aliphatic alcohols, which was in line with previous find-
ings.1,19−21,33
The cost analysis revealed that cofactor utilisation was the
bottleneck of this process. Thus, the target key figures were XR2,
+
gNADP consumed/g2‑octanol, and STY, as well as the overall cost.
Modifications of the process that needed to be addressed were
improved enzyme stability or activity, different enzyme and
cofactor concentrations, as well as higher flow rates and higher
solubility of the substrates. These modifications fit into three
groups: Group 1 involved changes on a reaction engineering
basis, such as increased flow rate, increased enzyme
concentrations, or elevated cofactor concentration. Group 2
consisted of changes on a molecular biology basis such as
improved enzyme activity or stability, whereas group 3
contained changes that would require a reconsideration of the
process itself. An increased substrate concentration fell in group
3 due to the restricted solubility of the substrate. A higher IL
concentration would improve substrate solubility, but at the
same time, the viscosity of the solution would increase (data
not shown) which leads to practical limitations in the
ultrafiltration step. Moreover, a higher IL concentration also
renders the downstream processing more challenging due to
different partition behavior (higher solubility in the aqueous IL-
containing phase). Thus, another solubiliser should be
considered when a higher substrate concentration is required.
The effect of improving or increasing all these parameters by a
Figure 9. Improved continuous production of (R)-2-octanol; x2‑octanone
= conversion of 2-octanone in reactor 1 (grey) and reactor 2 (black);
grey line = model prediction reactor 1; black line = model prediction
−1
reactor 2; conditions: V = 4 mL h ; c2‑octanone = 60 mmol L−1; cLbADH
=
̇
50 mg L−1 (R1) ; cGDH = 250 mg L−1 (R1); cLbADH = 250 mg L−1 (R2)
; cGDH = 100 mg L−1 (R2); ttotal = 350 h; cNADP = 0.05 mmol L−1.
+
The conversion was stable for more than 350 h. Compared to
the base case synthesis and recycling, a further reduction of the
E-factor to 17 (PMI = 18) was achieved. Although the run time
of the new experiment was shorter, the TONs for both
enzymes were still in the range of several million, and the
enzyme contribution to the overall cost was within the above-
mentioned range of 5%.12 The cofactor turnover number per
pass through the cascade was improved from 539 to 918. In
addition to this improvement, roughly 25% of the cofactor
could be recycled. This was also reflected by the much lower
overall cost of €52/kg for the improved synthesis (−65%
F
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Figure 10. (Left) Cost analysis based on 250 mL of substrate solution without recycling, afterwards 90% recycling for the base case, the recycling
case, and the improved synthesis as well as for selected hypothetical improvements; grey line = stoichiometric minimum €/kgproduct based on
−1
−1
−1
substrate costs; conditions here: cLbADH = 50 mg L−1; cGDH = 250 mg L ; V = 4 mL h ; cNADP = 0.1 mmol L ; c2‑octanone = 60 mmol L−1; cTEGO IL K5
+
̇
= 100 g L−1; cADA = 100 mmol L−1; cMgCl = 20 mmol L−1; pH = 7.0. (Right) Sensitivity analysis of the process (mean values based on the prediction
2
+
of a 1000 h experiment). Improvements and aggravations compared to the base case. black bars = conversion in reactor 2; grey bars = gNADP
/
consumed
g2‑octanol produced; white bars = STYtotal
.
+
increase STY and lower the ratio of gNADP
/
Table 4. Comparison between the base case (including
recycling) and the improved synthesis
consumed
g2‑octanol produced compared to the base case. This could be
achieved by increasing the enzyme concentration, the enzyme
Vmax, and the enzymes’ half-life. Among these, the enzyme half-
life has the least effect on the target key figures. The effects of
the increased Vmax and increased enzyme concentration are the
same.
Assuming the same price for the native enzymes and the
improved enzymes, nearly no differences were observed in the
overall cost. Thus, to improve the process a higher enzyme
concentration is advisable, whereas expensive, time-consuming,
and economically risky molecular biological improvements are
not warranted in this case.
base case
improved
reactor 1
reactor 2
reactor 1
reactor 2
X
2‑octanone/% *
75.8
45.6
7.7
96.9
12.8
2.9
50.5
12.3
0.945
588
83.5
1.62
3.12
918
TONLbADH/106 a
TONGDH/106 a
+
TONNADP (per
411
539
pass)/− /+
STY/mmol L−1 d−1
291 (37)
124 (16)
>99.5
194 (25)
124 (16)
>99.5
(g L−1 d−1)*
ee/%
A higher cofactor concentration increased the conversion and
STY slightly, particularly elevating the amount of cofactor
needed per kg of product. Hence, increasing the cofactor
concentration is not advisable, since better results are obtained
with reduced cofactor concentrations.
factor of 2 is demonstrated in Figure 10. Only one parameter
was considered at a time to avoid excessive extrapolation of the
model.
When judging the effects of all the modifications, processes
should be favored that increase the conversion in reactor 2,
Figure 11. Space-time-yield (average R1 and R2, left) and conversion (after R2, right) as a function of total enzyme concentration and flow rate
(mean values based on the prediction of a 1000 h experiment). Conditions: cLbADH, R1 = 50 mg L−1; cGDH, R1 = 250 mg L−1; cLbADH, R2 = 250 mg L−1;
−1
cGDH, R2 = 100 mg L−1 and multiples of the enzyme concentrations; V = 4−20 mL h ; cNADP = 0.01−0.1 mmol L−1; c2‑octanone = 60 mmol L−1;
+
̇
cTEGO IL K5 = 100 g L−1; cADA = 100 mmol L−1; cMgCl = 20 mmol L−1; pH = 7.0.
2
G
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
Doubling the flow or the substrate concentration generates
the same results when considering the three target key figures
(XR2, gNADP consumed/g2‑octanol produced, and STY). The cost
CH3COC6H13 + CH3CHOHCH3
→ CH3CHOHC6H13 + CH3COCH3
+
analysis confirmed that a higher flow rate made the product
more expensive due to the lower cofactor utilisation. Therefore,
a higher flow rate is only advisable if the enzyme concentration
is adjusted at the same time to maintain a high level of
conversion. The effect of applying a higher flow rate together
with higher enzyme concentrations on STY (average R1 and
R2) and conversion (after R2) is predicted using the process
model (Figure 11).
Before these conditions can be used for a larger-scale
synthesis, some research should focus on the filtration step. We
used polyethersulfone (PES) membranes which normally allow
for higher flow rates compared to a regenerated cellulose
membrane. Ionic surfactants like TEGO IL K5 can interact with
such PES-membranes (see manufacturers’ data sheet) which
can lead to an accelerated fouling of the membrane.
Experimentally, at a constant flow of 8 mL h−1 a constant
increase in the reactor pressure was observed, rendering
continuous operation impossible. Hence, to increase the
synthesis further in view of STY, some effort has to be put
into the improvement of the filtration step.
All the cost analyses and modeling did not include
downstream processing. Thus, solvent consumption during
product elution was not included in the calculations. As the
performance of the adsorber depends on the solubility of the
substrates/products in the aqueous solution, comparing
syntheses with the same IL concentrations will not falsify the
results. However, preliminary results showed, that the n-
heptane/ethanol mixture could be reused after distillation for
eluting the substrates/products from the column. When
solvents are recycled by distillation, generally a loss of 10%
can be assumed.11 Thus, even though n-heptane is a relatively
ecotoxic solvent, its release to the environment can largely be
avoided.
Apart from the changes in the existing process, a totally new
synthesis route could be considered, for example the biphasic
synthesis as described in Leuchs et al.21 Methyl-tert-butyl-ether
(MTBE) is very suitable for biphasic synthesis in combination
with LbADH.36 In this context, it is mandatory to use 2-
propanol for cofactor regeneration in continuous synthesis due
to the low solubility of glucose in MTBE.
Assuming high turnover numbers of the catalysts (enzymes)
and cofactors in both cases, the atom efficiency is 42% for the
monophasic approach (40% if hydrolysis of GDL to gluconic
acid is taken into account) and 69% for the biphasic approach.
At first glance, the biphasic approach seems to be favorable,
but it is known that, for this kind of transfer hydrogenation,
thermodynamics play an important role.21,37−39 With a molar
ratio of 1:1 = 2-octanone/2-propanol, a conversion of 38% is
expected.21 With a 3.3-fold excess, as used in the present study
(glucose/2-octanone), a conversion of 68% is expected when
using 2-propanol as the hydride donor compared to full
conversion when using glucose/GDH as the hydride donor. To
reach an industrially relevant conversion, a large excess of 2-
propanol would be required.40 In view of atom economy and
efficiency, the selective addition of a hydrogen molecule per
molecule of substrate would be the most efficient method.41
Here, the use of suitable hydrogenases for cofactor regeneration
is a biocatalytically favorable alternative.42 In chemical catalysis
the asymmetric hydrogenation using ligand-modified noble
metal catalysts is very successful.43−45 A major drawback of this
synthesis route is the low enantioselectivity when it comes to
the reduction of aliphatic nonfunctionalised ketones.43
CONCLUSION
■
The continuous synthesis of (R)-2-octanol could be improved
to decrease the chemical cost basis by 65%. Modelling and
simulation allows the optimisation of this complex process.
Further targets of optimisation can be revealed by careful
extrapolation and sensitivity analysis. We demonstrated that an
enzyme’s affinity to NADP(H) instead of NAD(H) is not a
knockout criterion for an industrial process. This is important,
given the fact that molecular biological methods that change an
enzymes’ affinity to NAD(H) instead of NAD(P)H are not
always applicable, as is the case for the LbADH, our enzyme of
choice. Improvements by reaction engineering may be more
cost and time effective.
AUTHOR INFORMATION
■
Corresponding Author
*E-mail: l.greiner@hs-mannheim.de (L.G.); jw@kt.dtu.dk
(J.M.W.).
Sheldon10 proposed the use of atom efficiency for a quick
assessment of the environmental impact of an alternative
process when compared to an existing benchmark process,
which has been used frequently.11,27 In contrast to the E-factor,
which includes conversion, byproducts, and coupled products,
the atom efficiency can easily be calculated even before the
process is tried at the laboratory scale. Atom efficiency is the
molar mass of the desired product divided by the sum of the
molar masses of all products. It is based on stoichiometry only.
The overall reaction equation for the existing monophasic
approach is:
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
The work was financially supported by the Federal Ministry of
Economics and Technology (BMWi) through the AiF (ZIM
KF2609402) and by the Deutsche Forschungsgemeinschaft
(DFG GR 2282). We thank Prof. Marcel Liauw (RWTH
Aachen University) for discussion and support, Dr. Shukrallah
Na’amnieh (X-Zyme/Johnson Matthey Catalysts) for providing
enzymes, and Kerstin Hell and Dr. Schwab (Evonik Gold-
schmidt) for providing TEGO IL K5.
CH3COC6H13 + C6H12O6
→ CH3CHOHC6H13 + C6H10O6
REFERENCES
■
(1) Leuchs, S.; Na’amnieh, S. N.; Greiner, L. Green Chem. 2013, 15,
167−176.
(2) Rozzell, J. Bioorg. Med. Chem. 1999, 7, 2253−2261.
The overall reaction equation for the alternative biphasic
approach is:
H
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX

Organic Process Research & Development
Article
(3) Liese, A.; Seelbach, K.; Wandrey, C. Industrial Biotransformations;
Wiley-VCH: Weinheim, Germany, 2006.
(37) Peters, M.; Eckstein, M. F.; Hartjen, G.; Spiess, A. C.; Leitner,
W.; Greiner, L. Ind. Eng. Chem. Res. 2007, 46, 7073−7078.
(38) Eckstein, M. F.; Lembrecht, J.; Schumacher, J.; Eberhard, W.;
Spiess, A. C.; Peters, M.; Roosen, C.; Greiner, L.; Leitner, W.; Kragl,
U. Adv. Synth. Catal. 2006, 348, 1597−1604.
(4) Pollard, D. J.; Woodley, J. M. Trends Biotechnol. 2007, 25, 66−73.
(5) Straathof, A. J.; Panke, S.; Schmid, A. Curr. Opin. Biotechnol.
2002, 13, 548−556.
(39) Eckstein, M. F.; Peters, M.; Lembrecht, J.; Spiess, A. C.; Greiner,
L. Adv. Synth. Catal. 2006, 348, 1591−1596.
(6) Rosen, T. C.; Daussmann, T.; Stohrer, J. Speciality Chemicals
Magazine 2004, 4, 39−40.
(40) As outlined above, 2-propanol would be needed in excess to
increase equilibrium conversion to a viable value. Typically, an excess
of 5−20-fold is used. Therefore, even if prices per mol reduction
equivalent are in the same range, 2-propanol will effectively contribute
more.
(7) Eckstein, M.; Daussmann, T.; Kragl, U. Biocatal. Biotransform.
2004, 2, 89−96.
(8) Chenault, H. K.; Whitesides, G. M. Appl. Biochem. Biotechnol.
1987, 14, 147−197.
(9) Wichmann, R.; Vasic-
2005, 225−260.
́ ̌
Racki, D. Adv. Biochem. Eng./Biotechnol.
(41) Trost, B. M. Science 1991, 254, 1471−1477.
(42) Greiner, L.; Muller, D. H.; van den Ban, E. C. D.; Woltinger, J.;
̈
̈
(10) Sheldon, R. A. J. Chem. Technol. Biotechnol. 1997, 68, 381−388.
(11) Sheldon, R. A. Green Chem. 2007, 9, 1273−1283.
(12) Tufvesson, P.; Lima-Ramos, J.; Nordblad, M.; Woodley, J. M.
Org. Process Res. Dev. 2011, 15, 266−274.
Wandrey, C.; Liese, A. Adv. Synth. Catal. 2003, 345, 679−683.
(43) Noyori, R.; Ohkuma, T. Angew. Chem., Int. Ed. 2001, 40, 40−73.
(44) Noyori, R. Angew. Chem., Int. Ed. 2002, 41, 2008−2022.
(45) Ohkuma, T.; Koizumi, M.; Doucet, H.; Pham, T.; Kozawa, M.;
Murata, K.; Katayama, E.; Yokozawa, T.; Ikariya, T.; Noyori, R. J. Am.
Chem. Soc. 1998, 120, 13529−13530.
(13) Matsuo, J.-I.; Hattori, Y.; Hashizume, M.; Ishibashi, H.
Tetrahedron 2010, 66, 6062−6069.
(14) Kim, J.; Singaram, B. Tetrahedron Lett. 2006, 47, 3901−3903.
(15) Kawanami, Y.; Mikami, Y.; Hoshino, K.; Suzue, M.; Kajihara, I.
Chem. Lett. 2009, 38, 722−723.
(16) Boukachabia, M.; Zeror, S.; Collin, J.; Fiaud, J.-C.; Zouioueche,
L. A. Tetrahedron Lett. 2011, 52, 1485−1489.
(17) Mathews, M.; Zola, R. S.; ke Yang, D.; Li, Q. J. Mater. Chem.
2011, 21, 2098−2103.
(18) Leuchs, S.; Greiner, L. Chem. Biochem. Eng. Q. 2011, 25, 267−
281.
(19) Kohlmann, C.; Robertz, N.; Leuchs, S.; Dogan, Z.; Lutz, S.;
̈
Na’amnieh, S.; Greiner, L. J. Mol. Catal. B: Enzym. 2011, 68, 147−153.
(20) Kohlmann, C.; Robertz, N.; Leuchs, S.; Greiner, L. Green Chem.
2011, 13, 3093−3095.
(21) Leuchs, S.; Nonnen, T.; Dechambre, D.; Na’amnieh, S. N.;
Greiner, L. J. Mol. Catal. B: Enzym. 2013, 74, 52−59.
(22) Muller, P.; Bangasser, B.; Greiner, L.; Na’amnieh, S.; Bauerlein,
̈
̈
P. S.; Vogt, D.; Muller, C. Open Catal. J. 2011, 4, 113−116.
̈
(23) Kohlmann, C.; Leuchs, S.; Greiner, L.; Leitner, W. Green Chem.
2011, 13, 1430−1436.
(24) Niefind, K.; Muller, J.; Riebel, B.; Hummel, W.; Schomburg, D.
̈
J. Mol. Biol. 2003, 317−328.
(25) Alfonsi, K.; Colberg, J.; Dunn, P. J.; Fevig, T.; Jennings, S.;
Johnson, T. A.; Kleine, H. P.; Knight, C.; Nagy, M. A.; Perry, D. A.;
Stefaniak, M. Green Chem. 2008, 10, 31−36.
(26) The buffer results from previous optimisation.1 Especially,
cofactor stability and enzyme stability are lower in phosphate buffer
making the cheap buffer an expensive choice. Moreover, phosphate is
not compatible with Mg ions, leading to precipitation at pH > 6. As
the buffer is largely recycled (see below) the overall cost contribution
is bearable.
(27) Sheldon, R. A. Chem. Commun. 2008, 3352−3365.
(28) Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.;
Manley, J. B. Org. Process Res. Dev. 2011, 15, 912−917.
(29) Pretti, C.; Chiappe, C.; Pieraccini, D.; Gregori, M.; Abram, F.;
Monnia, G.; Intorre, L. Green Chem. 2006, 8, 238−240.
(30) http://www.Carl-Roth.de.
(31) Schlieben, N. H.; Niefind, K.; Muller, J.; Riebel, B.; Hummel,
̈
W.; Schomburg, D. J. Mol. Biol. 2005, 349, 801−813.
(32) Machielsen, R.; Looger, L. L.; Reaedts, J.; Dijkhuizen, S.;
Hummel, W.; Hennemann, H.-G.; Daussmann, T.; van der Oost, J.
Eng. Life. Sci. 2009, 9, 38−44.
(33) Greiner, L.; Laue, S.; Liese, A.; Wandrey, C. Chem.Eur. J.
2006, 12, 1818−1823.
(34) Hofmann, D.; Wirtz, A.; Santiago-Schubel, B.; Disko, U.; Pohl,
̈
M. Anal. Bioanal. Chem. 2010, 398, 2803−2811.
(35) Degenring, D.; Schroder, I.; Wandrey, C.; Liese, A.; Greiner, L.
Org. Process Res. Dev. 2004, 8, 213−218.
(36) Villela-Filho, M.; Stillger, T.; Muller, M.; Liese, A.; Wandrey, C.
̈
̈
Angew. Chem., Int. Ed. 2003, 42, 2993−2996.
I
dx.doi.org/10.1021/op400117t | Org. Process Res. Dev. XXXX, XXX, XXX−XXX