Microfluidic multi-input reactor for biocatalytic synthesis using transketolase

Article abstract of DOI:10.1016/j.molcatb.2013.05.016

Biocatalytic synthesis in continuous-flow microreactors is of increasing interest for the production of specialty chemicals. However, the yield of production achievable in these reactors can be limited by the adverse effects of high substrate concentration on the biocatalyst, including inhibition and denaturation. Fed-batch reactors have been developed in order to overcome this problem, but no continuous-flow solution exists. We present the design of a novel multi-input microfluidic reactor, capable of substrate feeding at multiple points, as a first step towards overcoming these problems in a continuous-flow setting. Using the transketolase-(TK) catalysed reaction of lithium hydroxypyruvate (HPA) and glycolaldehyde (GA) to l-erythrulose (ERY), we demonstrate the transposition of a fed-batch substrate feeding strategy to our microfluidic reactor. We obtained a 4.5-fold increase in output concentration and a 5-fold increase in throughput compared with a single input reactor.

Full text of DOI:10.1016/j.molcatb.2013.05.016

Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
Contents lists available at SciVerse ScienceDirect
Journal of Molecular Catalysis B: Enzymatic
journal homepage: www.elsevier.com/locate/molcatb
Microfluidic multi-input reactor for biocatalytic synthesis using transketolaseଝ
James Lawrencea, Brian O’Sullivana, Gary J. Lyea, Roland Wohlgemuthb, Nicolas Szitaa,∗
a Department of Biochemical Engineering, University College London, Torrington Place, London WC1E 7JE, UK
b Sigma–Aldrich, Industriestrasse 25, 9470 Buchs, Switzerland
a r t i c l e i n f o
a b s t r a c t
Article history:
Biocatalytic synthesis in continuous-flow microreactors is of increasing interest for the production of
specialty chemicals. However, the yield of production achievable in these reactors can be limited by the
adverse effects of high substrate concentration on the biocatalyst, including inhibition and denaturation.
Fed-batch reactors have been developed in order to overcome this problem, but no continuous-flow
solution exists. We present the design of a novel multi-input microfluidic reactor, capable of substrate
feeding at multiple points, as a first step towards overcoming these problems in a continuous-flow setting.
Using the transketolase-(TK) catalysed reaction of lithium hydroxypyruvate (HPA) and glycolaldehyde
(GA) to l-erythrulose (ERY), we demonstrate the transposition of a fed-batch substrate feeding strategy
to our microfluidic reactor. We obtained a 4.5-fold increase in output concentration and a 5-fold increase
in throughput compared with a single input reactor.
Received 31 January 2013
Received in revised form 8 May 2013
Accepted 17 May 2013
Available online 28 May 2013
Keywords:
Micro reactor
Fed-batch
Transketolase
Laser fabrication
Microfluidic reactor
© 2013 The Authors. Published by Elsevier B.V. All rights reserved.
1. Introduction
of conversion. They therefore represent a highly attractive and
versatile biocatalytic platform technology with an ever-increasing
Microfluidic reactors are of increased interest for preparative
synthetic work due to a variety of advantages such as improved
heat and mass transfer, mixing, safety, process intensification, scal-
ability and reproducibility. As such they are being applied to a
growing number of reactions in research and industry [1–4]. The
use of microreactors for catalytic asymmetric synthesis has a high
potential among the many different methodologies for the effi-
cient and sustainable synthesis of valuable chiral compounds with
improved control of reaction conditions and fast optimisation [5].
As biocatalytic asymmetric synthesis has proven to be a sustain-
able and viable methodology in organic synthesis and industrial
manufacturing processes [6–9], efforts to combine the advantages
of biocatalysis and micro-reactor technology in a mutually benefi-
cial way have started to create numerous new application areas of
interest, including resolution of racemic mixtures, reactions using
hazardous substrates or two-phase systems [10–21].
range of applications [22–27]. Since the scale-up to viable manufac-
turing processes requires also work on the process design, reaction
engineering and the optimisation of reaction conditions [28,29], the
application of microreactor technology to transketolase-catalyzed
carbon-carbon bond formation reactions is of fundamental interest
[30].
In a previous publication, we have described the cascading of a
microfluidic reactor and a filtration system for biocatalytic asym-
metric synthesis using transketolase [31]. The system was able to
fully convert HPA to the chiral product l-erythrulose (ERY) with
high enantiomeric purity, as well as separating out the transke-
tolase (TK) from the product, synthesising a pure pharmaceutical
intermediate in a continuous-flow setting. However, the yield of
ERY was low relative to the typical output that would be expected
from chemical synthesis processes. Therefore it would be of great
value to investigate improvements of yield and throughput in such
devices.
The maximum product yield achievable with a T-channel reac-
tor design, such as that used in our previous publication, is heavily
dependent upon the effects of increased substrate concentration on
the biocatalyst. These effects include inhibition and denaturation
of the biocatalyst, with the latter causing clogging in continuous-
flow reactors. In batch reactors, such effects are overcome by
controlled feeding of the substrate solution at defined time points
(fed-batch mode). Miniaturised fed-batch reactors with opera-
tion volumes of 12 mL [32] and 0.5 mL [33] exist, and could be
employed to maintain the concentration of the substrate below
Transketolases have been shown to be useful biocatalysts for
creating new carbon–carbon bonds between aldehydes and the
irreversible two-carbon donor lithium hydroxypyruvate (HPA)
with high selectivity, broad substrate specificity and a high degree
ଝ
This is an open-access article distributed under the terms of the Creative
Commons Attribution License, which permits unrestricted use, distribution and
reproduction in any medium, provided the original author and source are credited.
∗
Corresponding author. Tel.: +44 0207 679 4418; fax: +44 0207 916 3943.
E-mail address: n.szita@ucl.ac.uk (N. Szita).
1381-1177/$ – see front matter © 2013 The Authors. Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.molcatb.2013.05.016

112
J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
The interconnect blocks of the MIR were fabricated in 5 mm
polycarbonate (PC) with a micromilling machine (Folken IND, Glen-
dale, USA), using a 2 mm end mill (Kyocera, Kyoto, Japan) with a
spindle speed of 10,000 rpm and feed rate of 80 mm min−1. M3
and M6 taps were used to prepare the interconnect blocks for use.
Standard connection fittings were used to attach tubing (P-221,
Upchurch Scientific, Oak Harbour, WA, USA).
Scheme 1. Reaction scheme. The transketolase-catalysed reaction of lithium
hydroxypyruvate (HPA) and glycolaldehyde (GA) to l-erythrulose (ERY).
Plugs used to seal unused auxiliary inputs were fabricated in
poly(dimethylsiloxane) (PDMS; Sylgard 184, Dow Corning, Mid-
land, USA). A mould was milled from 5 mm PMMA with a 2 mm
inhibitory levels, allowing a higher product output. However these
reactors cannot be operated in continuous-flow fashion, which
limits their applicability for multi-step process integration. Addi-
tionally, the significant space-time-yield increases, characteristic
of continuous-flow microreactors, cannot be exploited [34].
Reactor systems allowing injection of substrates at multiple
points have been demonstrated for the purpose of controlling
exothermic chemical reactions. These systems were designed for
the continuous synthesis of allylcarbinol and of organometallic
compounds, using multi-point feeding to control the formation
of impurities and the generation of heat respectively [35–37].
However, such systems have not been applied to the problem of
substrate inhibition in biocatalytic reactions.
Continuous-flow ‘loop-type’ microreactors designed to allow
the recycling of unconverted substrates have previously been
demonstrated [38,39]. It is possible to use these reactors to gradu-
ally feed substrate, however the continuous injection of substrate
into the recycling loop at a single point means that a pure prod-
uct stream is fundamentally difficult to achieve. The continuous
removal of product also requires that the biocatalyst is either
immobilised or removed by an in-line separation system, neces-
sitating a more complex reactor design.
tool, a spindle speed of 7000 rpm and feed rate of 40 mm min−1
.
The liquid polymer was prepared in a ratio of 10:1 (monomer to
curing agent), cast, degassed and then cured at 90 ◦C for 2 h.
2.3. Preparation of transketolase lysate
Transketolase lysates were prepared according to the method
of Matosevic et al. [40]. Overnight cultures of E. coli BL21gold (DE3)
(with transketolase-producing plasmid pQR791) were grown in 2 L
shake-flasks from inoculation of 400 mL Lysogeny Broth (LB) with
1 mL of concentrated cell suspension in LB-glycerol stock solu-
tion (25%, v/v, glycerol, stored at −80 ◦C until inoculation). This
was incubated for 20–24 h at 37 ◦C, until the bacterial growth had
reached stationary phase as confirmed by optical density measure-
ments. The contents of the flask were transferred to 50 mL falcon
tubes and centrifuged at 5000 rpm for 10 min. The supernatant was
discarded and the cell paste was frozen at −80 ◦C until needed for
lysis and purification.
For lysis, the cell paste was resuspended in 2 mL 50 mM Tris–HCl
buffer, cooled on ice and sonicated (10 cycles of 10 s on, 10 s off)
with a sonication probe (Soniprep 150, Sanyo, Japan). The sus-
pension was then centrifuged at 5000 rpm for 10 min and the
supernatant containing the enzyme was stored at −20 ◦C until
required.
In this contribution, we combine a continuous-flow reactor and
the principles of a fed-batch substrate feeding strategy for the first
time at the microfluidic scale. We present the design of a novel
microfluidic reactor capable of substrate feeding at multiple points.
We demonstrate the application of the reactor to the TK-catalysed
reaction of lithium hydroxypyruvate (HPA) and glycolaldehyde
(GA) to l-erythrulose (ERY; Scheme 1).
2.4. Continuous-flow microfluidic reaction of HPA to ERY (with
multiple GA inputs)
2. Experimental
2.1. Reagents and analysis
Two separate solutions were used to perform the reac-
tion. The main reaction mixture (solution A) consisted of
0.069 mg mL−1 clarified transketolase lysate, different HPA concen-
trations (211/316/421/526 mM HPA, depending on experiment),
2.53 mM thiamine diphosphate (ThDP) and 10.3 mM MgCl2 in
50 mM Tris–HCl buffer, pH 7. The concentrations of the solutes
were chosen such that they would be diluted to the following con-
centrations once combined with the first GA input: 0.066 mg mL−1
clarified transketolase lysate, 200/300/400/500 mM HPA, 2.4 mM
thiamine diphosphate (TDP) and 9.8 mM MgCl2. The supplemen-
tary GA solution (solution B) consisted of 1 M GA in 50 mM Tris–HCl
buffer, pH 7 (Scheme 2).
The MIR was primed with Tris–HCl buffer (50 mM, pH 7). Solu-
tion A was pumped into the first primary input of the reactor with
a single-drive syringe pump. Solution B was pumped into the sec-
ond primary input, along with a number of auxiliary inputs, using a
dual-drive syringe pump adapted to fit ten 1 mL syringes. The flow
rates and the number of auxiliary inputs used were dependent upon
the desired residence time and input HPA concentration (Table 1).
The reactor was allowed to equilibrate for 2.5 residence times
before sampling began. Samples were collected in pre-weighed
vials containing 270 ␮L 0.1% (v/v) aqueous trifluoroacetic acid
(TFA). The quenched samples were weighed, centrifuged and the
supernatant was diluted 1:1 with 0.1% TFA before being analysed
by HPLC as described in Section 2.1.
Unless otherwise stated, all chemicals and reagents
(Sigma–Aldrich, Gillingham, UK) were used without further
purification.
Transketolase concentrations were measured by SDS-PAGE
electrophoresis with 12% Tris–glycine resolving gel, using bovine
serum albumin standards. 20 ␮g of total protein was applied to each
lane and the samples were stained with Coomassie Blue R-250.
HPLC quantification of lithium hydroxypyruvate (HPA) and l-
erythrulose (ERY) was performed on Aminex (Biorad, Hemel Hemp-
stead, UK) ion-exchange column (HPX-87H, 300 mm × 7.8 mm),
mobile phase: 0.1% (v/v) aqueous trifluoroacetic acid (TFA) at
0.6 mL min−1. HPA and ERY were detected by UV absorption at
210 nm.
2.2. Fabrication of the microfluidic multi-input reactor
The channels and cut-outs of the multi-input reactor (MIR) were
fabricated in three layers of 1.5 mm poly(methylmethacrylate)
(PMMA) using a CO2 laser marking head (Laserlines, Banbury, UK)
with a maximum power of 25 W. The features were ablated with a
power of 50% and a mark speed (laser tracking speed) of 200 mm s−1
and 10 mm s−1 for channels and cut-outs, respectively. The three
layers were cleaned and thermally bonded (1.5 h, 105 ◦C, 90 min).

J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
113
Table 1
Reactor configuration. Configuration of reactor inputs, residence times and flow rates used to perform reactions with different input HPA concentrations and residence times.
HPA input concentration (mM)
200
120
300
165
400
264
500
286
Residence time (mins)
Position
Primary 1
Primary 2
Aux 1
Aux 2
Aux 3
Aux 4
Aux 5
Aux 6
Aux 7
TK/HPA
TK/HPA
TK/HPA
GA
GA
GA
GA
GA
GA
GA
-
TK/HPA
GA
GA
GA
GA
GA
GA
GA
GA
GA
-
GA
GA
-
-
GA
-
GA
-
-
-
-
GA
GA
-
-
GA
GA
-
-
GA
-
Aux 8
Aux 9
Aux 10
-
GA
-
-
-
Flow rate (␮L min−1
)
Primary 1
Primary 2 & Aux
4.49
0.24
4.94
0.26
8.24
0.43
11.90
0.63
2.5. Fed-batch microplate reactions
2.6. 12-h continuous production of ERY
Fed-batch reactions were performed in 96-well microplates.
200 ␮L of solution A was made up in each well. 10 ␮L of solution
B was manually added to each well to start the reactions. Fur-
ther 10 ␮L doses of solution B were then manually added to the
reaction mixtures at specific time points, corresponding with the
feeding strategy used in the continuous-flow reactions. Endpoint
samples were taken after the reactions had run for the equivalent
residence times, diluting 30 ␮L of reaction mixture in 270 ␮L 0.1%
TFA. Samples were centrifuged before being analysed by HPLC.
The experiment was run at the conditions of the 500 mM HPA
conversion (Table 1), using the same solutions and reactor setup
as described in Section 2.5. The output tubing from the MIR was
connected to a RotAXYS auto-sampling system (Cetoni GmbH, Kor-
bussen, Germany), programmed to take 30 ␮L samples every hour
from the start of the reaction. 12 wells of a 96-well plate were each
filled with 270 ␮L of 0.1% TFA to quench the samples. Samples were
centrifuged before being analysed by HPLC.
3. Results and discussion
3.1. Reactor design
For a typical microreactor consisting of a microchannel with
two (or occasionally more) inputs at one end of the channel, the
substrate concentration is defined at a single point (the input) and
cannot be supplemented further. In such reactor designs, a higher
concentration of product is obtained by increasing the concentra-
tion of substrate at the input. However, in biocatalysis, the effects
of increased substrate concentrations, such as substrate inhibition
and protein denaturation, set limits on the amount of substrate that
can be added at a single point.
In the case of the TK-catalysed synthesis of ERY, we have
observed an increase in reaction rate and yield with increased HPA
and GA input, up to concentrations of 300 mM GA. It has previously
been shown that GA concentrations of 500 mM GA cause a signifi-
cant reduction in the activity half-life of TK, to 200 min as compared
with 400 min for a concentration of 250 mM [41].
From our own observations when performing microfluidic reac-
tions with GA concentrations of 500 mM, it becomes clear from
visual inspection that a build-up of solid material is occurring,
causing the reactor channel to clog. This severely impairs the func-
tionality of the reactor, usually halting the flow before it has time
to equilibrate. We therefore decided to explore whether feeding
strategies, analogous to those used in fed-batch processes, could be
used in a continuous-flow microreactor to overcome these limita-
tions. It would thus be worthwhile investigating how to overcome
these problems in order to remove the limitations on the yield of
product that is achievable in biocatalytic microreactors.
To achieve this, we designed a microfluidic reactor with multi-
ple ports, capable of feeding the total amount of substrate across
multiple inputs, and through this enhance the flexibility of microre-
actors for biocatalytic synthesis. Fig. 1 illustrates the design of the
reactor. The design consisted of one main channel with a total
Scheme 2. Reactor input layout and in-channel concentrations. Layout of reactor
inputs referred to in Table 1 and concentrations found at inputs and in main reaction
channel.

114
J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
Fig. 1. Design of the microfluidic multi-input reactor (MIR). (A) Top-down layout of channels in the reactor; shaded areas indicate ablated channels, lines indicate cuts
through the substrate, dimensions in mm; (B) photo of the assembled MIR, showing interconnect bars with (back) and without (front) fittings attached; colored dyes were
used to visualise the channels (dyes with different colors illustrate the several inputs of the MIR, but are not related to the color scheme of figures A, C, and D); (C) exploded
view of reactor design also showing interconnect ports (fabricated in PC); (D) section view of the microfluidic reactor showing the ‘through-holes’ which connect the channels
on different layers, dimensions in mm. (For interpretation of the references to color in text, the reader is referred to the web version of the article.)
volume of 1650 ␮L fabricated in poly(methylmethacrylate)
(PMMA), with two primary inputs at the front end of the chan-
nel where the biocatalyst and substrate were combined to start
the reaction (Fig. 1A). Ten further auxiliary inputs were spaced
evenly along the channel, where further substrate could be added.
The inputs were designed for standard fittings, facilitating the easy
reconfiguration of the reactor. PDMS plugs were also designed that
were compatible with the standard inputs, allowing any unused
inputs to be reliably and reversibly sealed (Fig. 1B). In this way the
reactor could easily be configured to test a range of different feeding
strategies.
To enable the conversion of large amounts of substrate, the
reactor was designed to be operated with long residence times,
and hence required a relatively large total volume. At the same
time, to allow it to be fabricated on most available laser abla-
tion systems, the outer dimensions of the reactor needed to be
limited to 85 mm × 85 mm. We therefore opted for an elegant
design which divided the channel over two layers, with ‘through-
holes’ ablated into the middle layer to allow fluid to ascend and
descend (Fig. 1C and D). All channels in the reactor had a cross
section of 1 mm × 0.5 mm, and the length of individual sections of
the main channel was 60 mm. Sections of channel on the middle
(yellow in Fig. 1) layer were aligned 91◦ from the normal and chan-
nels on the base (red) layer were aligned 269◦ from normal. When
bonded together, the channels on the different layers aligned to
form a single channel.
1. Given the number of channels and other features involved, fab-
ricating the same design using a micro milling machine, another
commonly used tool for rapid fabrication, would have taken a
few days. As a further advantage, the low cost of PMMA makes
the reactors very inexpensive to produce, meaning that they can
easily be disposed of and replaced at the end of their useful
life.
The increasing availability of laser ablation systems means
that the reactor could be fabricated by groups without specialist
microfluidics knowledge, and the modular design facilitates easy
adaptation to suit the needs of different biocatalytic reactions. The
offset of the auxiliary input ports from the main reactor chan-
nel provides an added benefit, allowing the structure of the main
channel to be changed without modification of the port layout.
Decoupling the inputs from the channel in this way means that
the residence volume between individual ports can be adjusted,
for example to allow for shorter residence times between inputs
earlier in the reaction. Combined with the rapidity of the fabri-
cation method, this would allow the production of reactors that
were tailored to the exact substrate feeding schedule required by
the reaction. Additionally, the reactor could be scaled up by sim-
ple modifications to the design and by use of a larger laser ablation
system. This would then allow bench-top or even pilot scale pro-
duction in continuous-flow, which is of interest to the chemical and
biochemical industries [4].
The flow regime within the main reactor channel was deter-
mined to be laminar, with a Reynolds number of 2.2 × 10−4 at a
flow rate of 10 ␮L min−1 (calculated using the viscosity and density
of water). This was confirmed by confocal imaging of diffusion at a
junction between an auxiliary input and the main channel, where
the fluorescent tracer dye fed though the auxiliary was shown to
form a lamina over the bulk fluid in the channel (see Electronic
Supplementary Information 2, Fig. S2).
Laser ablation was chosen as the fabrication method for its
rapidity. Using the laser, it was possible to fabricate the compar-
atively complex middle layer in less than 6 min and to complete all
three layers of the reactor in around 15 min. This fits well with the
‘rapid prototyping’ paradigm that is a common feature of microflu-
idics. A video showing the rapid laser fabrication of the middle
layer can be found in the Electronic Supplementary Information

J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
115
Fig. 2. End point concentrations of HPA and ERY from TK-catalysed reactions. (A)
Fed-batch 96-well plates; (B) microfluidic multi-input microfluidic reactor. Error
bars are standard deviation (N = 3). The residence times used for the conversion of
different input HPA concentrations are given in Table 1.
Despite the laminar regime and the absence of mixing struc-
tures, the tracer appeared to be evenly distributed throughout the
channel within a residence time of 1 min, equivalent to less than
half the length of a single channel section at the highest flow rates
used in the synthesis experiments. This is in agreement with the
calculation of the diffusional length of the tracer in the channel,
which predicts that it will have propagated 0.39 mm across the
channel depth in 1 min (where the diffusion coefficient of fluores-
cein in water is 0.64 × 10−9 m2 s−1 [42]).
3.2. Performance of the multi-input reactor
Fig. 3. Throughput, volumetric productivity and conversion data for different reac-
tion modes.(A) Throughput of ERY from each system; (B) volumetric productivity
(C) percentage conversion of HPA to ERY. Squares (ꢀ) = fed-batch 96-well plates, tri-
angles (ꢁ) = microfluidic multi-input reactor. Error bars refer to standard deviation
(N = 3).
The conversion of four different input concentrations of HPA
was carried out in the MIR and in 96-well microplates, using four
different feeding strategies. In each strategy the configuration or
timing of GA feeding points was altered in order to avoid inhibitory
effects on the enzyme whilst aiming for complete conversion to
ERY. The process of selecting GA feeding points is explained further
in Electronic Supplementary Information 3.
The configurations used for each HPA input concentration are
shown in Table 1. The flexibility of the reactor was such that each
of these strategies could be tested with a minimal amount of effort
in setting up, simply by changing the auxiliary connections and flow
rates as required.
Fig. 2 shows the endpoint concentrations of HPA and ERY from
the MIR and fed-batch microplate well reactions. Compared with
our own single substrate input microfluidic reactor [31], the multi-
point feeding strategy increased the yield of product from 50 mM
to a maximum of 224 mM with the MIR (Fig. 2B). Compensating
for the dilution caused by the nine GA inputs used in the feeding
strategy, the maximum concentration of ERY generated is 313 mM.
Fig. 3 shows the conversion, throughput and volumetric pro-
ductivity data from the MIR and equivalent fed-batch microplate
well reactions. Comparing the output from the respective systems,
we see a very large improvement in the throughput of ERY from
the MIR, regardless of the input HPA concentration used, with the
highest observed increase being 8-fold at the lowest concentration
(Fig. 3A). By comparison with our single substrate input reactor
we also see an increase in throughput from 3 mg h−1 to 15 mg h−1
This clearly shows the improvements in throughput and space-time
.

116
J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
12 h (depending on the input HPA concentration); Fig. 4 suggests
that the TK activity may have decayed significantly by the time
reactions with longer residence times were sampled.
A further modification that could be made to overcome this
problem is the addition of a cooling system to maintain the stabil-
ity of the solubilised enzyme. The enzyme could be thermo-stated
to the desired temperature shortly before it is introduced into the
reactor. A mutant of the TK enzyme with greater temperature sta-
bility could also be used. If the initial concentration of ERY observed
in Fig. 4 could be maintained, this would allow production at a rate
of 19 mg h−1, and thus would have led to a total of 171 mg over the
12 h.
4. Conclusion
We have successfully demonstrated the application of ‘fed-
batch’ like substrate feeding strategies to continuous-flow
microreactor technology to address the limitations of single-
point feeding. Our novel flexible microfluidic reactor design offers
numerous input configurations and we tested different feeding
strategies using the TK-catalysed condensation of HPA and GA as
the test reaction. Furthermore, the design is easily fabricated with
laser ablation, with the potential for modification to develop cus-
tomised reactions for different biocatalysts and feeding strategies.
The modularity of the design would also allow the incorporation of
the MIR into a wider network of reactors and separation technolo-
gies, with the potential to develop multi-step synthetic processes
for complex molecular products.
Using our substrate feeding strategies and the flexibility of the
MIR, we have greatly improved on the yield of product achieved
with our previous microfluidic reactor [31]. The output concentra-
tion of ERY has been increased from 50 mM to 224 mM ERY, the
equivalent of 313 mM if the dilution by GA feeding is compensated
for, without exposing the enzyme to destructive concentrations
of substrate. In addition, we have shown a 5-fold increase in the
throughput of the MIR over our previous reactor, to a maximum of
15 mg h−1, with a maximum 8-fold improvement over fed-batch
microplate reactions. This highlights the primary advantage of
continuous-flow reactors in terms of synthetic productivity.
We have shown that the reactor is capable of very high con-
version for short periods of time, but in order to achieve and
maintain full conversion for higher input concentrations further
improvements are necessary, addressing issues of CO2 evolution
and enzyme stability. This would allow the synthesis of ERY at a rate
which would represent a significant step towards the high through-
put and output concentrations that are commonly associated with
chemical microreactors.
Fig. 4. Profile of output concentrations from a 12-h continuous reaction run. Output
concentrations of HPA (diamonds, ꢂ) and ERY (squares, ꢀ) from reaction run for
12 h using 500 mM input HPA concentration, 9 GA inputs and 286-min residence
time (N = 1). Time refers to the number of hours elapsed after flow was started. No
HPA/ERY was recorded for the first 3 h as the reaction mixture had not yet reached
the output. The line indicates an estimate of the cumulative weight of ERY produced
over the course of the run.
yield that are the foremost advantage of continuous flow reac-
tors over batch or fed-batch reactors, which is in agreement with
similar comparisons from the literature [34]. A continuous flow
system is also of great interest for use in multi-step synthesis pro-
cesses, where a cascade of separate reactors can be created to
perform a complex conversion. The MIR would be well-suited to
this, having standard interconnect ports that would allow connec-
tion to other reactors or intermediate purification steps.
The volumetric productivity and conversion achieved in the MIR
was higher than that of the fed-batch microplate at lower input
HPA concentrations, reaching a maximum of 9.4 (g h−1) L−1 and
86% respectively (Fig. 3B and C). At higher HPA concentrations the
conversion and productivity of the fed-batch microplate reactions
surpassed that of the MIR. This may have been caused by decreasing
pH as the reaction progressed, due to the evolution of CO2 as a by-
product, which was released to the headspace in the microplates
but remained in solution in the MIR.
A reaction system designed for synthesis should aim to reach
100% conversion in order to minimise the downstream require-
ment for purification, or unintended side reactions when used as
part of a multi-step process. Any change in pH could be mitigated
in a manner analogous to that of a pH-stat reactor. This could be
accomplished by using some of the auxiliary inputs on the MIR
to feed a base solution, thereby counteracting the pH drop and
allowing complete conversion of substrate to product. However,
to investigate this further would require integration of online pH
monitoring into the MIR; a technology that has thus far only been
used in cellular microreactors [43].
Fig. 4 shows the output concentration profiles of HPA and ERY
from a reaction performed in the MIR using the feeding strategy
developed for a 500 mM HPA input concentration. The reaction was
sampled every hour with an automated sampling arm. The reaction
equilibrated between hour 4 and 5 (as might be expected given that
the residence time was 286 min), with a maximum output con-
centration of around 400 mM ERY. An estimated 90 mg of ERY was
produced over the course of the experiment. The output concentra-
tion drops continuously between hours 5 and 12, suggesting decay
in TK activity due to exposure to room temperature conditions. This
may provide a further explanation of the decrease in conversion
observed at higher input HPA concentrations (Fig. 3C). The reac-
tions were each sampled at 2.5 residence times, at between 5 and
Acknowledgements
The authors would like to acknowledge the BBSRC for providing
funding for a studentship to James Lawrence. The authors would
also like to thank Pat Morris for her invaluable advice on the wild-
type transketolase enzyme used in this work.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.molcatb.
2013.05.016.
References
[1] K. Jähnisch, V. Hessel, H. Löwe, M. Baerns, Angew. Chem. Int. Ed. 43 (2004)
406–446.

J. Lawrence et al. / Journal of Molecular Catalysis B: Enzymatic 95 (2013) 111–117
117
[2] T. Wirth (Ed.), Microreactors in Organic Synthesis and Catalysis, first ed., Wiley-
VCH Verlag, Weinheim, Germany, 2008.
[24] P. Payongsri, D. Steadman, J. Strafford, A. MacMurray, H.C. Hailes, P.A. Dalby,
Org. Biomol. Chem. 10 (2012) 9021–9029.
[3] A.J. de Mello, Nature 442 (2006) 394–402.
[4] D. Roberge, B. Zimmermann, F. Rainone, M. Gottsponer, M. Eyholzer, N. Kock-
mann, Org. Proc. Res. Dev. 12 (2008) 905–910.
[5] X.Y. Mak, P. Laurino, P.H. Seeberger, Beilstein J. Org. Chem. 5 (2009) 19.
[6] C.M. Clouthier, J.N. Pelletier, Chem. Soc. Rev. 41 (2012) 1585–1605.
[7] K.H. Drauz, O. May, H. Gröger (Eds.), Enzyme Catalysis in Organic Synthesis,
third ed., Wiley-VCH Verlag, Weinheim, Germany, 2012.
[8] H.P. Meyer, E. Eichhorn, S. Hanlon, S. Lütz, M. Schürmann, R. Wohlgemuth, R.
Coppolecchia, Catal. Sci. Technol. 3 (2013) 29–40.
[25] A. Ranoux, S.K. Karmee, J. Jin, A. Bhaduri, A. Caiazzo, I.W.C.E. Arends, U. Hanefeld,
ChemBioChem 13 (2012) 1921–1931.
[26] D. Yi, T. Devamani, J. Abdoul-Zabar, F. Charmantray, V. Helaine, L. Hecquet, W.D.
Fessner, ChemBioChem 13 (2012) 2290–2300.
[27] F. Charmantray, V. Hélaine, B. Legeret, L. Hecquet, J. Mol. Catal. B: Enzym. 57
(2009) 6–9.
[28] J. Shaeri, R. Wohlgemuth, J.M. Woodley, Org. Proc. Res. Dev. 10 (2006) 605–610.
[29] J. Shaeri, I. Wright, E.B. Rathbone, R. Wohlgemuth, J.M. Woodley, Biotechnol.
Bioeng. 101 (2008) 761–767.
[9] R. Wohlgemuth, Curr. Opin. Biotechnol. 21 (2010) 713–724.
[10] S. Nagl, P. Schulze, M. Ludwig, D. Belder, Electrophoresis 30 (2009) 2765–2772.
[11] M.P.C. Marques, P. Fernandes, J.M.S. Cabral, P. Znidarsˇicˇ-Plazl, I. Plazl, Chem.
Eng. J. 160 (2010) 708–714.
[12] A.T. Lewandowski, W.E. Bentley, H. Yi, G.W. Rubloff, G.F. Payne, R. Ghodssi,
Biotechnol. Progr. 24 (2008) 1042–1051.
[13] S. Kataoka, Y. Takeuchi, A. Harada, M. Yamada, A. Endo, Green Chem. 12 (2010)
331–337.
[14] A. Schwarz, M.S. Thomsen, B. Nidetzky, Biotechnol. Bioeng. 103 (2009)
865–872.
[15] J.W. Swarts, P. Vossenberg, M.M. Meerman, A.E.M. Janssen, R.M. Boom, Bio-
technol. Bioeng. 99 (2008) 855–861.
[30] J. Sukumaran, U. Hanefeld, Chem. Soc. Rev. 34 (2005) 530–542.
[31] B. O’Sullivan, H. Al-Bahrani, J. Lawrence, M. Campos, A. Cázares, F. Baganz, R.
Wohlgemuth, H.C. Hailes, N. Szita, J. Mol. Catal. B: Enzym. 77 (2012) 1–8.
[32] D. Weuster-Botz, R. Puskeiler, A. Kusterer, K. Kaufmann, G.T. John, M. Arnold,
Bioproc. Biosyst. Eng. 28 (2005) 109–119.
[33] M. Funke, A. Buchenauer, U. Schnakenberg, W. Mokwa, S. Diederichs, A.
Mertens, C. Müller, F. Kensy, J. Büchs, Biotechnol. Bioeng. 107 (2010) 497–505.
[34] I. Itabaiana Jr., L.S. de Mariz e Miranda, R.O.M.A. de Souza, J. Mol. Catal. B: Enzym.
85 (86) (2013) 1–9.
ˇ
[35] A.E. Cervera-Padrell, J.P. Nielsen, M.J. Pedersen, K.M. Christensen, A.R.
Mortensen, T. Skovby, K. Dam-Johansen, S. Kiil, K.V. Gernaey, Org. Proc. Res.
Dev. 16 (2012) 901–914.
[16] K. Koch, R.J.F. van den Berg, P.J. Nieuwland, H.E. Shoemaker, J.C.M. van Hest,
F.P.T.J. Rutjes, Biotechnol. Bioeng. 99 (2008) 1028–1033.
[17] M.S. Thomsen, B. Nidetzky, Biotech. J. 4 (2009) 98–107.
[18] S. Kundu, A.J. Bhangale, W.E. Wallace, K.M. Flynn, C.M. Guttmann, R.A. Gross,
K.L. Beers, J. Am. Chem. Soc. 133 (2011) 6006–6011.
[36] P. Barthe, C. Guermeur, O. Lobet, M. Moreno, P. Woehl, D.M. Roberge, N. Bieler,
B. Zimmermann, Chem. Eng. Technol. 31 (2008) 1146–1154.
[37] D.M. Roberge, N. Bieler, M. Mathier, M. Eyholzer, B. Zimmermann, P. Barthe,
C. Guermeur, O. Lobet, M. Moreno, P. Woehl, Chem. Eng. Technol. 31 (2008)
1155–1161.
[19] R. Moser, T. Chappuis, E. Vanoli, S. Crelier, J. Naef, Chimia 64 (2010)
799–802.
[38] M. Tudorache, D. Mahalu, C. Teodorescu, R. Stan, C. Bala, V.I. Parvulescu, J. Mol.
Cat. B 69 (2011) 133–139.
[20] K. Koch, R.J.F. van den Berg, P.J. Nieuwland, R. Wijtmans, M.G. Wubbolts, H.E.
Schoemaker, F.P.J.T. Rutjes, J.C.M. van Hest, Chem. Eng. J. 135 (S1) (2008)
S89–S92.
[21] B. Ngamsom, A.M. Hickey, G.M. Greenway, J.A. Littlechild, P. Watts, C. Wiles, J.
Mol. Catal. B: Enzym. 63 (2010) 81–86.
[22] R. Wohlgemuth, M.E.B. Smith, P.A. Dalby, J.M. Woodley, Transketolases,
Encyclopedia of Industrial Biotechnology, Bioprocess, Bioseparation, and Cell
Technology, John Wiley & Sons, Inc., Hoboken, NJ, 2009, pp. 1–7.
[23] R. Wohlgemuth, J. Mol. Catal. B: Enzym. 61 (2009) 23–29.
[39] N.M.Z.H. Alam, M. Pinelo, K. Samanta, G. Jonsson, A.K.V. Meyer, Gernaey Chem.
Eng. J. 167 (2011) 418–426.
[40] S. Matosevic, M. Micheletti, J. Woodley, G. Lye, F. Baganz, Biotechnol. Lett. 30
(2008) 995–1000.
[41] B.H. Chen, F. Baganz, J.M. Woodley, Chem. Eng. Sci. 62 (2007) 3178–3184.
[42] P. Galambos, F.K. Forster, Proceedings of the uTAS’ 98 Workshop, 1998, pp.
189–192.
[43] D. Schapper, M.N.H.Z. Alam, N. Szita, A.E. Lantz, K.V. Gernaey, Anal. Bioanal.
Chem. 395 (2009) 679–695.