7.

5 Reaction Engineering of Biotransformations

M Bechtold and S Panke, ETH Zurich, Basel, Switzerland
r 2012 Elsevier Ltd. All rights reserved.

7.5.1 Introduction 72
7.5.2 Important Process Properties of Enzymes 72
7.5.2.1 Important Aspects of Enzyme Kinetics 73
7.5.2.2 Kinetics and Optical Activity 76
7.5.2.3 Model-Based Experimental Analysis for Parameter Estimation 77
7.5.2.4 Kinetic versus Thermodynamic Control 78
7.5.2.5 Enzymes in Nonconventional Media and Implications for Processes 78
7.5.2.6 pH and Kinetics 80
7.5.3 Enzyme Stabilization 80
7.5.3.1 Inactivation Mechanisms 80
7.5.3.2 Single and Comprehensive Stability Parameters 81
7.5.3.2.1 Single stability parameters 83
7.5.3.2.2 Comprehensive operational stability characterization 83
7.5.3.3 Stabilization Techniques 83
7.5.3.3.1 Immobilization 83
7.5.3.3.1.1 Immobilization on a carrier 86
7.5.3.3.1.2 Entrapment/encapsulation 86
7.5.3.3.1.3 Self-immobilization 87
7.5.3.3.1.4 Some remarks on immobilization of lipases 87
7.5.3.3.1.5 Mass transfer and diffusion limitations of immobilized biocatalysts 87
7.5.3.3.2 Catalyst stabilization and protein engineering 88
7.5.4 Typical Reactors 89
7.5.4.1 Selection and Design of Bioreactors 89
7.5.4.1.1 Batch reactor 90
7.5.4.1.2 CSTR 92
7.5.4.1.3 PFR 92
7.5.4.1.4 Comparison CSTR–PFR 93
7.5.4.1.5 CSTRs in sequence 93
7.5.4.2 Reactor Concepts 94
7.5.4.2.1 Membrane bioreactors 94
7.5.4.2.2 Immobilized enzyme reactors 95
7.5.4.2.3 Reactors for enzyme reactions in unconventional media 95
7.5.5 Current and Future Developments in Enzyme Technology 96
7.5.5.1 Synthetic Biology 96
7.5.5.2 Miniaturization 96
7.5.5.3 Model-Based Process Design 96
7.5.5.4 Integrated Operation of Reaction and Separation 97
7.5.6 Summary 98
References 98

Glossary Immobilization An enzyme that is attached to an inert

Bioreactors Reactors used for the industrial-scale insoluble carrier, predominantly polymers, is referred to as
implementation of biotransformations. Typical examples are immobilized enzyme. Immobilization often provides
batch reactors (stirred tanks), coupled setups of continuous stabilization and enables easy reuse of the biocatalyst.
stirred tank reactors and ultrafiltration modules for retention Protein engineering The use of genetic modification such
of solubilized enzymes and packed bed reactor operation as directed evolution for improvement of important
when the enzyme is supplied in immobilized form. properties of the biocatalyst such as activity or stability
Biotransformation A single or cascade of biochemical under process conditions.
reaction(s) catalyzed by a single or a number of enzyme(s), Reaction engineering ‘Reaction engineering is that
which can be realized by the use of purified enzyme, cell- engineering activity concerned with the exploitation of
free extract or whole cells. In contrast to fermentations the chemical reactions on a commercial scale. Its goal is the
product is structurally similar to the substrate. successful design of chemical reactors, and probably more

Comprehensive Chirality, Volume 7 http://dx.doi.org/10.1016/B978-0-08-095167-6.00705-9 71

72 Reaction Engineering of Biotransformations

than any other activity its sets chemical engineering apart as bioreaction engineering, however the methods need to be
a distinct branch of the engineering profession.’ This adapted to the specific properties of the catalyst, the
definition from Octave Levenspiel can be equally applied to enzyme.

7.5.1 Introduction

Enzymes – the catalysts of biotransformations – have acquired a prominent place in the manufacturing of chemicals, from
research1 to industrial-scale manufacturing.2 This development is a result of their fascinating selectivities – reaction selectivity,
regioselectivity, and stereoselectivity – and their ability to conduct chemistry under nonintensive conditions, combined with the
improvement of methods of producing them cheaply and in large scale, of finding enzymes for the catalysis of a specific reaction,
and of improving their performance. With the venturing of biotransformations to ever larger scales, the proper design of the
corresponding processes becomes more and more important. Here, the implementation of biotransformations can draw on the
large repertoire of theories and methods known from chemical engineering, and adapt them to the peculiarities of the catalyst
central to the process, which is the enzyme, and its requirements. This focus on a catalyst which (usually, see Section 7.5.2) prefers
aqueous environments and neutral pH (exceptions recruited from extremophilic microorganisms notwithstanding), hardly
concerns itself with gas-phase reactions, typically does not lead to the release of large amounts of heat, but displays excellent
selectivity limits the set of engineering methods that are regularly applied to the design of enzyme processes. This led to the
formation of the field of enzyme technology, which covers the most important methods for the application of enzymes for
preparative chemical purposes.
The core of the methods in enzyme technology has remained constant over the past three decades. However,
notable developments repeatedly occur in particular in a few fields, enzyme immobilization, separation-integrated processes, and
the ambition to use the broadly available hydrolases, which constitute one of the two major enzyme pillars of biotransformations
(the other being redox enzymes), for synthetic purposes. The latter aspect involves typically working under conditions of low
water activity, such as the use of enzymes in organic solvents. Therefore, this text will cover what we consider the most important
aspects in enzyme technology while taking into account the major developments in the mentioned two fields.

7.5.2 Important Process Properties of Enzymes

Enzymes are organized into six major classes according to a bird’s eye view of the chemistry they catalyze (Table 1), and from
there on hierarchically in further subgroups, so that a 4-digit ‘EC number’ (EC stands for enzyme commission) is sufficient to
identify any given enzyme and to inform at the same time about the applied catalytic strategy. Of the six classes recognized, two
stand out for their importance2,3: Redoxenzymes (class 1), for example, for the enantioselective synthesis of alcohols from
prochiral ketones, and hydrolases (class 3), for example, for the kinetic resolution of racemates of amino acid derivatives.
However, also enzymes from classes 4 (lyases) and 5 (isomerases) play an ever increasing role in the synthesis of optically pure
chemicals.4

Table 1 Enzyme classification and relative importance

First digit Second digit, subclass, example Third digit subclass, example Comment 3

1. Oxidoreductases 1.1. Oxidoreductases acting on the 1.1.1. Oxidoreductases acting on the One of the two most frequently used
CH–OH group of donors CH–OH group of donors with enzyme classes
(22 subclasses) NAD þ or NADP þ as acceptor
2. Transferases 2.1. Transferring one-carbon groups 2.1.1. Transferring methyl groups
(9 subclasses)
3. Hydrolases 3.1. Acting on ester bonds 3.1.1. Carboxylic ester hydrolases One of the two most frequently used
(13 subclasses) enzyme classes
4. Lyases 4.1. Carbon–carbon bond lyases 4.1.1. Carboxy lyases Prominent examples, for example,
(7 subclasses) hydratation of acrylonitrile to
acrylamide
5. Isomerases 5.1. Racemases and epimerases 5.1.1. Acting on amino acids and Prominent examples, for example,
(6 subclasses) derivatives epimerization of glucose to glucose/
fructose mixture
6. Ligases 6.1. Forming CO bonds (6 subclasses) 6.1.1. Ligases forming aminoacyl-
tRNA and related compounds

A number of specific properties deserve mentioning here, to set the scene and also to introduce the main themes of the
following text. First, many enzymes require cofactors. For example, the prominent alcohol dehydrogenases for the asymmetric
 Reaction Engineering of Biotransformations 73

synthesis of chiral alcohols require either reduced nicotinamidedinucleotide (NADH) or nicotinamidedinucleotidephosphate

(NADPH) as a hydrogen donor. These compounds are too expensive to be used stoichiometrically, so they have to be regenerated
after they donated their hydrogen. The same principle applies to other important cofactors, such as adenosinetriphosphate (ATP)
or acetyl-coenzyme A. Regeneration is typically achieved by a second reaction, in which another, preferably rather cheap, substrate
is consumed to achieve the regeneration. The classic example here is the oxidation of formate to carbon dioxide by formate
dehydrogenase, during which hydrogen is transferred to NAD þ (the oxidized form of NADH), leaving the resulting NADH ready
for a new round of catalysis. Thus, many biotransformations are effectively two-step processes, making process design slightly
more complex, but clearly not to a prohibitive extent.5
Next, many enzymes are – viewed through the eyes of an organic chemist who is used to have a whole battery of organic
solvents and reaction conditions concerning temperature and pH at his or her disposal – sensitive to environmental conditions.
Even if ever larger numbers of thermostable enzymes are either recruited from thermophilic organisms or engineered starting from
mesophilic templates, a still larger number of available enzymes does not effectively work above 50 1C. Similar things can be said
about enzymes and extreme pH (including the part on exceptions). These effects can be counteracted to some extent by enzyme
immobilization, which frequently contributes greatly to stabilizing an enzyme (see Section 7.5.3.3.1).
Immobilization also contributes to the reuse of enzymes. Enzyme production costs are frequently nonnegligible in the overall
process scheme, prompting the development of schemes for facile retention. The diverse chemical functionalities on the outside of
proteins presented by the various amino acid residues provide ample opportunity to design covalent or ionic strategies. Also, the
large size of proteins (as a rule of thumb, on the order of 50 kDa) can be readily exploited for retention, as will be discussed in
Section 7.5.4.2.
Another important point for process design is that enzymes need a minimum of water to function. This minimum of water – the
so-called ‘structural water’ – is required for maintaining the critical structure of the enzyme and potentially also as part of the reaction
mechanism. Adding more water typically leads to an increase in specific activity, but it is – in sensu strictu – not necessary. This has
important implications for the process design with enzymes in nonconventional media, as will be discussed in Section 7.5.2.5.

7.5.2.1 Important Aspects of Enzyme Kinetics

In order to design suitable processes to harvest the catalytic power of an enzyme, its kinetics must be known. Important
parameters to characterize an enzyme in view of its utility for a process are, for example, the turnover number (TON) and the
turnover frequency (TOF). The former number is dimensionless and informs about the average number of conversions a single
catalyst molecule can do before it becomes inactivated. The latter denotes the number of conversions per active site of the enzyme
per time, thus acknowledging the fact that some enzymes are available as multimers.
Number of molecules converted
TOF ¼ ðs1 Þ
Time  number of active sites available

Enzyme activity is frequently reported strictly in terms of conversion power. The corresponding SI-unit is ‘katal,’ which is
defined as the amount of enzyme that converts 1 mol of substrate in one second under a given set of conditions. As this is a rather
impractically large conversion, an alternative definition persists in enzyme technology, which is the ‘international unit,’ referring
to the amount of product produced or substrate converted per time:
mmol
1 unit ¼ 1
min

Please note that despite the definition being based on, for example, produced product, in the end what is described is a certain
amount of enzyme that is required to achieve this conversion, without requiring any specific information as to how the structure
of that enzyme is or whether it is immobilized or not. Care needs to be taken to give a careful definition of the environmental
conditions under which the enzyme activity was determined (including temperature, pH, medium, etc.).
Frequently, this absolute activity is referenced to a suitable variable to yield the specific activity. This can be the mass of purified
protein giving the specific activity of the pure enzyme. Alternatively, this can also be the mass of total protein in the sample (if cell-
free extracts (CFEs) are used instead of purified protein), the total mass of (wet) catalyst (if the enzyme is immobilized and it is
not clear how much of the catalyst is support material, how much is active enzyme and how much is other proteins, and what the
percentage of water is in the catalyst preparation), or any other useful reference variable.
The progress in time of an enzyme-catalyzed conversion can be easily obtained from the model of Michaelis and Menten,6
where an enzyme E encounters a substrate S and reversibly forms an enzyme–substrate complex (ES), which dissociates irre-
versibly into enzyme and product P with the corresponding rate constants ki:
k1 k2
E þ S ! ðESÞ ! E þ P

k1

Of course this model is too simple to capture all the intricacies of enzyme operation (and it was never intended to accurately
reflect them), but it serves remarkably well to rationalize all sorts of properties of enzyme reactions and is extremely useful to
74 Reaction Engineering of Biotransformations

design enzyme reactions in the initial stages. If this model is translated into the mass balances of the participating species, these
are typically solved by making one of two additional assumptions: either, the first step – the reversible formation of the ES – is
expected to happen much faster than the second step (‘rapid equilibrium assumption’), which allows exploiting the equilibrium
between enzyme, substrate, and the ES as an additional constraint (see Table 2). Alternatively, this assumption cannot be made
and the Bodenstein steady-state assumption is invoked, which claims that in a time relevant for the observation, the concentration
of the reactive intermediate – in this case (ES) – does not change.

Table 2 Summary of the effects of the two assumptions in deriving Michaelis–Menten kinetics

Rapid equilibrium Bodenstein steady state

k1 ,k1 ook2 dðESÞ

¼0
k1 ðESÞ dt
Keq ¼ ¼
k1 ES
vmax ¼ k2 E0 vmax ¼ k2 E0
1 ES k1 k1 þ k2 k2
KM ¼ KS ¼ ¼ ¼ KM ¼ ¼ KS þ
Keq ðES Þ k1 k1 k1

E0, total concentration of all enzyme molecules; KS, saturation constant.

Both assumptions lead to the same expression for the description of the reaction rate v, with only slight differences in the
definition of one parameter (see Table 2). Note that the model presented above stipulates a reaction in which one molecule of the
substrate is converted into only one molecule of the product:

dP vmax S
¼ v ¼ k2 ðESÞ ¼
dt KM þ S

which is of course the famous ‘Michaelis–Menten’ equation. A few further comments are in order in view of some prominent
properties (see also Figure 1): First, the affinity of the enzyme to its substrate is reflected in the Michaelis–Menten constant KM,
which can be interpreted as the substrate concentration at which the enzyme operates with half the maximally possible speed.

vmax
Slope vmax/KM
Reaction rate (M s−1)

vmax/2

KM 2KM 4KM 6KM 8KM 10KM

Initial substrate concentration S0 (M)

Figure 1 Michaelis–Menten plot.

Second, there are two regimes in which the nonlinear full Michaelis–Menten equation can be substituted for simpler for-
mulations: When the substrate concentration is much higher than the KM value, the expression becomes zeroth order with respect
to the substrate concentration:

KM ooS
vmax S
v¼ Evmax
KM þ S
In contrast, when the substrate concentration is much lower than KM, then the expression becomes first order in S:

KM 44S
vmax S vmax
v¼ E S
KM þ S KM
 Reaction Engineering of Biotransformations 75

The reaction rate constant k2 in the simple physical model discussed above is identical to TOF for a monomeric enzyme. In the
context of looking at enzyme mechanism, this variable is usually called the catalytic constant, kcat, which for an enzyme with n
subunits is defined as:

vmax k2
kcat ¼ ¼
nE0 n

The ratio of kcat and KM, an important parameter in evaluating directed evolution experiments, is called the specificity constant,
for reasons which will become clearer later when we discuss inhibitions.
The reaction model formulated above is actually not very relevant for biocatalysis – in fact, taken at face-value, it is only valid
for an implausible hypothetical irreversible isomerization. However, in fact, the model is very useful in biocatalysis for a number
of reasons. First, a major class of biocatalytically important reactions is the hydrolysis reactions, for example, the hydrolysis of an
ester in water. Although this is a two-substrates/two-products reaction, one of the two substrates is water, which hardly changes its
concentration during the reaction and is available in large excess, making the reaction effectively irreversible. Therefore, treating
this as a one-substrate reaction is physically wrong, but without effect on the mathematical description. Furthermore, the two
products are formed at the same rate, and thus it is sufficient to look at one of the two. Similar arguments apply to many
biotransformations and explain to a large extent why this simple equation is still so much in use today.
In addition, the simple model also serves nicely in rationalizing reversible inhibitory effects on enzymes, which are unfor-
tunately rather frequent. Two simple modifications of the model allow capturing a large part of the observed inhibitory phe-
nomena (Figure 2). First, an inhibitor I is allowed to interact with the enzyme at the active site, and when I is present, substrate S
can no longer bind. I is not converted. Effectively, I competes with S for the active site, which has given this type of inhibition its
name: competitive inhibition. Second, I might not interact with the active site, but at a different site of the enzyme. In this case, the
enzyme can hold S when I interacts. Depending on how large the fraction of free enzyme E is in the system, this can amount to the
situation where I interacts effectively only with the complex (ES). This situation is called uncompetitive inhibition. Finally,
combinations of the two situations discussed above can exist, where the inhibitor attaches outside the active site to either E or
(ES), and interconversion between (EI) and (EIS) can occur. This situation is then termed ‘mixed’ inhibition. All these cases can be
easily treated mathematically on the basis of the simple model shown in Figure 2 and the Bodenstein steady-state assumption.
The results can be summarized as shown in Table 3.

k1 k2
E+S E+S E+P
k −1
−I +I −I +I
k−3 k3 k−5 k5
k4
EI+S EIS
k−4
Competitive Uncompetitive
Figure 2 Systematic of enzyme inhibitions.

Table 3 Systematic of formulas for enzyme inhibitions

vmax Sapp
vmax app KM app
v¼
KM app þ S

Inhibition type:
 
Competitive vmax I
KM 1 þ
Kic
 
Mixed vmax
KM 1 þ KIic
1 þ KIiu
1 þ KIiu
Uncompetitive vmax KM
1 þ KIiu 1 þ KIiu

Kic, competitive inhibitory constant; Kiu, uncompetitive inhibitory constant.

The formalism of competitive inhibition is useful to also rationalize certain other frequent phenomena. The first is product
inhibition. When the product is structurally still very similar to the substrate, it might as well have a nonnegligible affinity to the
active site and occupy it for a certain period, blocking access for the substrate. Effectively, the product behaves as a competitive
inhibitor. Furthermore, according to the reaction scheme that corresponds to the Michaelis Menten reaction mechanism substrate
76 Reaction Engineering of Biotransformations

and product are stoichiometrically dependent. Then, after consulting Table 3 for the formalism of competitive inhibition and
substituting product P for inhibitor I:

P ¼ S0  S

v S
v¼  max 
KM 1 þ KPic þ S
vmax S
v¼  
KM þ KKMic S0 þ S 1  KKMic
If the concept of the inhibitor is opened up and I is allowed to be transformed by the enzyme, then we have the situation of the
enzyme facing two substrates and the question which substrate is used preferably. While substrate S1 is in the active site, S2 cannot
react and vice versa. Assuming furthermore that S1 and S2 rapidly equilibrate with the enzyme, then the competitive inhibition
constants Kic2 (characterizing the competition for the conversion of S1) and Kic1 (characterizing the competition for the con-
version of S2) are nothing else than the KM-values of the respective compounds. Then:
vmax1 S1
v1 ¼  
KM1 1 þ KSM2
2
þ S1

v S
v2 ¼  max2 2
KM2 1 þ KSM11
þ S2
kcat1
S1
v1 K
¼ M1
v2 kcat2
S2
KM2
In other words, the specificity constant kcat/KM determines in competition situations how strongly an enzyme discriminates
between two substrates.
The formalism of uncompetitive inhibition is useful to rationalize another frequent phenomenon in enzyme catalysis: sub-
strate inhibition. If substrate concentrations become very high, they frequently lower the reaction rate again rather than allowing
the enzyme to operate at vmax. The model behind this is that the substrate has some affinity for a site on the enzyme outside the
active site, and when the substrate docks there the enzyme is no longer active. Consultation of Table 3 with the substrate as the
uncompetitive inhibitor leads quickly to

vmax ½S
v¼ 2
KM þ ½S þ ½S
Kiu

Of course, even though the Michaelis–Menten equation is very useful in rationalizing effects and planning early process stages,
it is of course no longer enough when reversible reactions need to be considered or the concentrations of multiple substrates are in
the range of their respective KM. Rather than giving a comprehensive treatment of multisubstrate kinetics here (for which the
reader is referred to excellent literature, such as,7) we will include only the equation for reversible one-substrate reactions in this
brief summary of the most important phenomena in enzyme kinetics, because this reaction is important in using enzymes as
racemization catalysts, for example, in dynamic kinetic resolutions. The corresponding reaction model is easily formulated as
k1 k2
E þ S ! ðESÞ ! E þ P
 
k1 k2

and with the help of, for example, the Bodenstein steady-state assumption, one arrives quickly at

d½P vmax ½S  KKM,P

M
vmax,r ½P 
¼
dt KM þ ½S þ KKM,PM
½P 

with

k1 þ k2 k1 þ k2
vmax ¼ k2 E0 KM ¼ vmax,r ¼ k1 E0 KM,P ¼
k1 k2

7.5.2.2 Kinetics and Optical Activity

The time course of an enzyme reaction can also have considerable impact on the optical activity in certain situations. For this
discussion, it is useful to define one measure for optical activity, such as the enantiomeric excess, ee, with R as the concentration of
 Reaction Engineering of Biotransformations 77

the R-enantiomer and S as that of the S-enantiomer:

j R  Sj
ee ¼ 100 ð%Þ
RþS

Although the ee that can be obtained in simple one-step desymmetrization reactions (e.g., mesotricks) or asymmetric syntheses
is not dependent on the progression of the reaction, the situation is different in kinetic resolutions, in particular if the used
enzyme does not discriminate sufficiently between the two presented enantiomers. Here, one of two enantiomers in a racemate is
preferentially converted by an enzyme, and the concentration of the preferred substrate decreases correspondingly. This leads to
the depletion of the enantiomer mix of one enantiomer, producing a certain ee for the remaining substrate. At the same time, the
produced product has also an ee. As soon as the mix is sufficiently depleted of the preferred enantiomer, the corresponding
reaction rate is reduced and the conversion of the unwanted enantiomer becomes more important, compromising as well the ee of
the remaining substrate as of the produced product.
In order to free the corresponding analyses from the limitations of reporting results in ees, the enantiomeric ratio E was
coined,8 which is independent of conversion (usually, this ratio is termed E, but to differentiate it from the enzyme E in this text, it
is called E here):
k1S k2S
E þ S ! ðESÞ ! E þ S#

k1S
k1R k2R
E þ R ! ðERÞ ! E þ R#

k1R
E0 kcatS S
 
R kcatS
KMS 1 þ þS S
¼ MS ¼ E
vS KMR K S
¼
vR E0 kcatR R kcatR R
  R
S KMR
KMR 1 þ þR
KMS
kcatS
E ¼ MS
K
kcatR
KMR

In the scheme above, S# and R# refer to the products that are produced from the S- and the R-form of the substrate respectively,
and S and R in the subscripts refer to the parameters of the Michaelis–Menten equation consuming either the S or the R
enantiomer. Furthermore, it is assumed that the two substrates adhere to the competitive inhibition formalism as discussed above.
The enantiomeric ratio describes then simply the ratio of the two specificity constants of the enzyme for the two substrates. In
order to determine it, formulas for deriving it from regular measurements of enantiomeric excesses of either substrate (eeS) or
product (eeP) and the extent of conversion X have been devised:8

E ¼
ln ½1  Xð1 þ eeP Þ
ln ½1  Xð1  eeP Þ

E ¼
ln ½ð1  XÞð1  eeS Þ
ln ½ð1  XÞð1 þ eeS Þ
For extreme values of X, which can no longer be reliably determined, the following equation can be applied9:

eeP ð1  eeS Þ
ln
E ¼
eeP þ eeS
eeP ð1 þ eeS Þ
ln
eeP þ eeS

7.5.2.3 Model-Based Experimental Analysis for Parameter Estimation

After having established the basic equations of enzyme kinetics, a related topic of great practical importance constitutes the
determination of the corresponding kinetic parameters. Traditionally, graphical methods based on initial reaction velocities (e.g.,
Lineweaver–Burk plots, etc.) were applied for estimation of kinetic parameters such as vmax and KM. For a comprehensive overview
on the different procedures, the reader is referred to the seminal work of Cornish–Bowden.10 Alternatively, sophisticated
numerical and statistical in silico procedures have been developed that provide information not only on the parameterization but
also on the reliability of the estimated parameters and whether the applied mechanistic model can be used for the description of
78 Reaction Engineering of Biotransformations

the obtained experimental data in the first place (model discrimination). Such an approach is generally referred to as model-based
experimental analysis.
Principally, there are three distinct components to a model-based analysis: (1) the experimental data; (2) a model describing
the experiment from which the experimental data was obtained; and (3) fitting of the data by minimizing a function that reflects
the deviation between model prediction and experimental data.11 By comparing residuals between model description and
experiment, model discrimination can be conducted. Next, application of statistical methods allows for the calculation of
confidence intervals for the parameterization and for the goodness of the fit. For statistical analysis, the variance in the experi-
mental data is required which requires producing the data (at least) in triplicates. Next, the experimental variance should be low
so that a reasonable comparison between residuals originating from the lack of fit and the experimental scatter can be conducted,
for example, by application of a F-test12 (see Figure 3).

Experimental data
Set of model equations
Progress curves with initial and boundary conditions
Conversion

dS S
= c E kcat
dt S + KM
IC :S (t = 0) = S0

Po
or
ties in parameters

fit:
Modify experiment

Nonlinear regression

mo d
Time statistical analysis

ify mechanistic
Fit model parameters to experimental data
rtain

Conversion

Evaluation of obtained parameter set

nce

Best fit

m
• Confidence intervals

od
u
gh

el
kcat = k ± k KM = K ± K
Hi

• Comparison lack of fit with experimental

variance: F-test, significance levels

Time
Figure 3 Work flow scheme of model-based experimental analysis.

Estimation of kinetic parameters is typically based on progress curves that cover different initial substrate concentrations.
Statistical analysis suggests to vary the initial substrate concentrations in specific steps relative to the expected KM value in case of
simple Michaelis–Menten kinetics.13 Model-based experimental analysis has proven useful for estimating kinetic parameters of
even complex kinetics such as the bi-bi mechanisms,14 which are typically beyond the scope of graphical interpolation methods.

7.5.2.4 Kinetic versus Thermodynamic Control

Of course, enzyme-catalyzed reactions are also subject to undesirable reactions such as side- and consecutive reactions, which
decrease the yield. One particular application of enzymes deserves special mention in the present context: the kinetically con-
trolled synthesis with hydrolases, as it finds application, for example, in the manufacturing of semisynthetic antibiotics15 and in
the field of glycoside production by transglycosylation.16 The process is explained in more detail in Figure 4 for the production of
the b-lactam antibiotic ampicillin. Essentially, the idea is to use hydrolases, in contrast to their natural roles, for synthesis in
aqueous media. For this, the fact that the enzyme forms a covalent reaction intermediate with a substrate is exploited, as this
intermediate can be attacked by different nucleophiles, one of which gives rise to the desired product. To increase the chance of
the formation of the covalent enzyme–substrate intermediate, the substrate is activated (e.g., phenylglycinamide instead of
phenylglycine (PG)). However, the intermediate can also be resolved by, for example, water, leading to the inactivation of the
substrate. Clearly, in this situation it is essential to understand the various kinetic parameters optimally (synthesis vs. hydrolysis
ratio and influence of water activity) and to time the end of the reaction optimally.

7.5.2.5 Enzymes in Nonconventional Media and Implications for Processes

A central theme in the use of biotransformations in organic synthesis is enzyme behavior under conditions of low water activity, as
for example, in the presence of organic solvents. This behavior is ruled by multiple influences. Most important is the availability of
 Reaction Engineering of Biotransformations 79

O Ser
O Ser
N H O NH2
HN N NH2 EnzI
HN O
O− NH2
O−
NH3
O
O
(a) PGA

Ser
O
O NH2
O Ser HN N
H
HN N HO O−
H
O− NH2 N
Ampicillin
H O O
N
S
N S
O 6-APA
N
O HOOC
COOH
NH2

PG
OH H2O

(b) O

Concentration

6-APA

PGA
PG

Ampicillin
Time
Reaction Reaction under
under thermodynamic control
kinetic
(c) control
Figure 4 Kinetic and thermodynamic control. (a) The activated substrate, phenylglycinamide (PGA), forms with PenG acylase, a serine-type
hydrolase (here shown with the ‘catalytic triade’ including the aspartate- and histidine-residues required to increase the nucleophilic character of
the serine-hydroxyl group), a covalent intermediate (EnzI), which in a subsequent step is resolved by 6-aminopenicillanic acid (6-APA) to yield
the reaction product, ampicillin. (b) The resolution of the covalent intermediate can occur, however, also by a variety of other nucleophiles,
including water, which leads to the formation of phenylglycine (PG). (c) Qualitative time course of the reaction if allowed to run beyond the
optimal point. First, the desired reaction product ampicillin is preferably produced, as the resolution of EnzI by 6-APA is the fastest reaction. As
the concentration of 6-APA decreases, this reaction becomes less important. At the same time, ampicillin accumulates, which can be hydrolyzed
by the enzyme as well, accelerating product degradation. The reaction continues until the product distribution dictated by thermodynamics is
reached, which in this case means that all amide bonds that could be hydrolyzed (PGA, ampicillin), are in fact hydrolyzed, and the corresponding
reaction products (PG, 6-APA) accumulate.

sufficient water to guarantee the structural and functional integrity of the enzyme. This minimum requirement is in some cases
met already by very few water molecules,17 so that enzymes can function in essentially neat organic solvents as long as a certain
(low) water activity is maintained. Of course, the enzymes then form a second phase, as they do not dissolve in the organic solvent
unless their overall hydrophobicity is altered by chemical modifications of the enzyme.18 The enzymes also tend to be less active,
as the absence of much water tends to rigidify the structure.19 However, the absence of large amounts of water also frequently has
80 Reaction Engineering of Biotransformations

a stabilizing effect.20 Such schemes are extremely helpful when using hydrolases in the synthesis, such as ester or peptide synthesis,
when the traditional irreversibility of the hydrolysis can be reversed due to the absence of water.
Clearly, a careful control of the reaction is required for optimal performance. However, when enzymes are provided to
essentially neat organic solvents, the performance of the reaction is often dominated by poorly controlled experimental details,
such as the manner of enzyme immobilization, stirring, the type of stirrer, etc., which complicates a chemical engineering analysis.
A second factor that influences the performance is of course the maintenance of the optimal water activity. This has spurred a
few process-oriented activities, in particular in the field of chromatographic reactors,21 where the water produced during the
reversed hydrolysis is immediately separated from the reaction liquid and thus the low water activity retained. Other methods
applied for continuous water removal include evaporation, pervaporation, perstraction into a saturated salt solution, use of
molecular sieves,22 and operation of a packed bed reactor with supercritical CO2.23
Alternatively, the enzyme can be used in a two-phase set-up, in which the enzyme is part of the aqueous phase and the organic
phase serves as a reservoir for the substrate and as a sink for the product. In this situation, the process is mostly governed by
classical mass transfer considerations.24

7.5.2.6 pH and Kinetics

The influence of pH can be reversible or become irreversible if too extreme. In the reversible situation, some amino acid residues
can change their charge status as a function of pH. If there is a charged amino acid in the active center of the enzyme that is
important for the function of the enzyme and the charge of this amino acid residue changes, the enzyme becomes inactive. This
concept leads to the following simple but useful model. Assuming that the species E is the active form, and E and E2 are not
active, then:
Hþ Hþ
E ! E ! E2

þ
 
þ

þH þH

H þ E
K1 ¼
E

Hþ E2
K2 ¼
E
E0 ¼ E þ E þ E2

E 1
¼ Hþ
E0 K1 þ 1 þ HK2þ
As the ratio of concentrations of active enzyme species over total enzyme concentration is a measure for the vmax that is
possible at this pH, distinctly bell-shaped curves with a relative maximum over the optimal pH emerge when plotting vmax over
pH.7

7.5.3 Enzyme Stabilization

Enzyme catalysis frequently provides excellent reaction characteristics in terms of chemo-, regio-, and stereoselectivity while
operating under benign reaction conditions. However, for industrial applications the biocatalyst’s long-term stability under
operational conditions and the ability to run the process efficiently with standard equipment contribute equally to the economic
feasibility of a specific biotransformation route.25 These objectives (if not fulfilled by the wild-type free enzyme) in principle
require modification of two properties of the biocatalyst: (1) making it more stable while retaining activity in order to utilize it for
an extended, economically attractive period of time and (2) providing the biocatalyst in a mechanically stable matrix with
sufficient specific activity that enables application in an efficient reactor setup (e.g., packed bed, see Section 7.5.4.2).
Although the latter point is predominantly tackled by applying one of the many different forms of immobilization, enzyme
stabilization can be achieved by a number of fundamentally different principles.26 Depending on the source of inactivation, some
immobilization methods are more suitable than others. In addition, modification of the primary sequence of the protein by
mutagenesis methods or chemical modifications and supplying additives acting on the solvation of proteins can provide a
remarkable improvement in enzyme stability. Consequently, a rational approach to enzyme stabilization requires knowledge of
the most prominent inactivation mechanism in order to select an appropriate method.27

7.5.3.1 Inactivation Mechanisms

Enzyme inactivation can principally be attributed to mechanisms related to the reactor, the medium components, or the protein.
Enzyme inactivation is often induced by phase interfaces resulting, for example, from dispersed air bubbles or biphasic liquid/
 Reaction Engineering of Biotransformations 81

liquid systems. In the case of air bubbles and droplets of nonpolar organic solvents, interfacial interaction with the enzyme results
in hydrophobic forces that disturb the secondary structure of the enzyme. Other irreversible mechanisms include cleavage by
proteases present, for example, in a CFE, precipitation, or aggregation,28 or chemical processes such as oxidation, racemization,
condensation, hinge region or tryptophan hydrolysis, or aspartate isomerization.29 However, there are also more subtle
mechanisms that render enzymes inactive. As proteins are active only in a specific complex structural configuration, unfolding of
some domains of the protein or subunit dissociation typically results in a drastically less active or completely inactive enzyme.
Consequently, mechanistic models of enzyme deactivation consider at least three fundamental protein states (1) the correctly
folded active state E; (2) an inactive, at least partially unfolded state U (this state is often considered to be randomly coiled,
although many recent studies indicate that it also includes partially folded structured variants); and (3) a completely denatured
state D30 that originates from the physical and chemical irreversible inactivation mechanisms28 introduced above. Stabilization of
the active state E with respect to unfolding can principally be attributed to a high activation energy barrier in case of a rate-
controlled unfolding process, or to the position of the unfolding equilibrium favoring the active state in case of thermo-
dynamically controlled unfolding. Experimental studies suggest that both mechanisms are ubiquitous in nature.28a,30a Obviously,
in both situations operational conditions such as temperature, pH, or amount of denaturing agents determine the extent of
unfolding and hence stability.

7.5.3.2 Single and Comprehensive Stability Parameters

Many detailed studies have been directed at elucidating the folding and unfolding mechanism of proteins induced by a shift in
temperature or solvent composition using a number of methods including differential scanning calorimetry, circular dichroism,
and nuclear magnetic resonance.30b,31 The obtained data are typically analyzed using a physical unfolding model that recognizes
the three different protein states E, U, and D introduced above and assumes rapid equilibrium between E and U, although this
assumption is not universally valid.28a Thus, in its simplest form deactivation can be represented by a sequential three state model,
with irreversible deactivation occurring only from the inactivated state U. This is also referred to as Lumry–Eyring model:

Keq kD
E"U ! D

Recently, a similar expression was derived in the context of operational stability investigations that differs only formally in the
description of the equilibrium partner and was termed ‘equilibrium model.’32
By introduction of temperature-dependent expressions – Van’t Hoff’s law for the equilibrium constant Keq and Arrhenius’ law
for the inactivation rate constant kD – enzyme deactivation can be qualitatively displayed in the time and temperature domain
(see Figure 5). In detail, equilibrium unfolding is represented by the enthalpy change DHU associated with unfolding of the active
state E to U and a reference state that is typically given by the temperature Teq at which half of the total enzyme is in the active
0
state, hence Keq ¼ 1.32
  
0 DHU 1 1
Keq ¼ Keq exp 
R T Teq

Applying the transition state theory of chemical reactions, the temperature dependence of the rate constant kD or kcat can be
expressed by physical constants (Boltzmann constant kB, gas constant R, and Planck’s constant h) and the free energy of activation
DG.33
   
kB T DGi kB T DGa
kD ¼ exp kcat ¼ exp
h RT h RT

Based on these four parameters – the activation energy of the biochemical reaction DGa and thermal inactivation DGi, DHU,
and Teq – a comprehensive description of active enzyme concentration, specific activity Aspec, and conversion in any reactor setting
(in case all other relevant kinetic parameters are known) can be derived.34

dE kD ðTÞKeq ðTÞE Ueq

ðTÞ ¼ Keq ¼
dt Keq ðTÞ þ 1 Eeq

EðTÞkcat ðTÞ kcat ðTÞ

Aspec ðTÞ ¼ ¼
EðTÞ þ UðTÞ 1 þ Keq ðTÞ

In this case, the specific activity is defined as the ratio of observed initial rate under saturated conditions (initial substrate
concentration cKm) and total enzyme concentration that, depending on the temperature, consists of different fractions of active
 82
Reaction Engineering of Biotransformations
0

ln [A(t)/A(t = 0)] (−)

dE/dt = −kDE

dE/dt(T) = −K(T)kD(T)/(K(T)+1)E −0.5 T = Tx 1/2 = −ln (0.5)/kD

100
ln 0.5

Specific activity A(t,T)/Amax (%)

100
Active enzyme fraction (%)

−1
Time dep. 75
75 inactivation
1/2
(c) Time (t)
50
50 100 A(tx,T)/A(t = 0,T)

Residual activity (%)

t = tx
25 25
lnKeq = ΔHeq/R(1/T−1/Teq)
50
Equilibrium thermodynamics
0
0 T = Tx
Temperature (T) Teq t = tx
(a) (b) Temperature (T)
Topt Time (t)
0
(d) Temperature T1/2
Figure 5 Time- and temperature-dependency of the active enzyme (a) and the evolution of specific activity (b) including the most prominent biocatalyst stability parameters. (c) Isothermal inactivation
curve with the fraction of residual activity on a logarithmic scale in order to easily determine the observed inactivation rate kD and half-life period t1/2. (d) Residual activity observed in temperature variant
assays after a specific time period.
 Reaction Engineering of Biotransformations 83

and inactive states of the enzyme. Although deactivation mechanisms can be significantly more complex, this model is widely
applied in practice and arguably serves well to illustrate some key properties and to introduce prominent stability parameters.34,35
At higher temperatures, two effects diminish the active enzyme pool – an increasingly unfavorable position of the thermo-
dynamic equilibrium resulting in a lower fraction of active enzyme and faster degradation due to an increasingly activated inacti-
vation rate. However, the biochemical reaction also becomes activated with increasing temperature resulting in a distinct temperature
optimum with respect to specific activity as a function of the operation time, a feature that has often been observed experimentally32
but is lacking from the classical model (see Figure 5) that recognizes only direct irreversible inactivation according to:
kD,obs
E ! U

7.5.3.2.1 Single stability parameters

The temperature at which highest specific activity is observed is commonly referred to as optimal temperature Topt. This tem-
perature can in practice be equated with Teq as the difference was shown to be less than 0.01 K for typical enzyme reactions.36
Next, under isothermal conditions the observed loss in residual activity can often be approximated by a first-order decay and the
corresponding rate constant kD,obs can be estimated by model-based analysis or simple semilogarithmic representation of the
residual activity fraction over time (see Figure 5). Alternatively, the half-life period t1/2 – the time period at which the observed
activity is half of the initial activity – can be used. It is inversely proportional to the deactivation rate.

dE E ln2
¼  kD,obs E ) ln ¼  kD,obs t t1=2 ¼
dt E0 kD,obs

Recording residual activities after a certain incubation time and for different temperatures constitutes another approach. Here,
the ratio of residual activity to initial activity at the respective temperature is formed and the temperature T1/2 is estimated that
yields half of the initial activity observed (see Figure 5).
In case the approximate operating temperature is known, calculation of the biocatalyst’s total TON constitutes an interesting
shortcut approach (compare Section 7.5.2.1). Based on the observed specific activity and an assumed first-order inactivation of the
enzyme, the total product yield over the biocatalyst’s lifetime can be estimated:35a
Z N Z N
vmax ðtÞdt E0 kcat,obs expðkD,obs tÞdt
kcat,obs
TON ¼ t ¼ 0 ¼ t¼0
¼
E0 E0 kD,obs

The TON is particularly useful for comparisons between different variants derived from mutagenesis experiments or a first
estimate on biocatalyst costs. However, it is hardly applicable to process design.35a

7.5.3.2.2 Comprehensive operational stability characterization

The parameters Topt, T1/2, t1/2, and kD,obs are valid only for a specific temperature or time point and hence can be used in a
qualitative manner or for comparison between different enzyme variants. Extrapolation to long-term operation or other tem-
peratures can hardly be performed relying on those parameters, which renders them inadequate for process design purposes. For
an exhaustive characterization, the parameterization of a sufficiently representative mechanistic model needs to be estimated.
Next, multimeric enzymes usually dissociate during inactivation,37 which requires a characterization with respect to enzyme
concentration in addition to temperature and time.38 Reliable characterization of the operational stability of a given biocatalyst
requires efficient experimental setups that allow for monitoring the time-dependent behavior of enzyme activity in the presence of
relevant concentrations of substrate, for example, by monitoring the conversion in continuous reactors. In order to avoid excessive
experimentation times, in particular when using rather stable enzymes, operation at higher temperature is required.39 Next, lab-
scale reactors that sufficiently mimic conditions at large-scale operation should be applied such as packed bed reactors in case of
immobilized enzymes or enzyme membrane reactors. There are in principle two approaches to grasp the temperature dependence
of enzyme performance (1) performing multiple experiments at different temperatures35a,40 and (2) applying a temperature
gradient. The latter has attracted some interest over the past decade and in principle should allow for reliable process design.41
However, a systematic evaluation has not been performed yet. Obviously, parameter estimation based on such experiments
requires sophisticated model-based experimental analysis procedures and statistical evaluation for model identification/
discrimination11 (see Section 7.5.2.3) (Figure 6).

7.5.3.3 Stabilization Techniques

7.5.3.3.1 Immobilization
The term ‘immobilized enzyme’ was coined in the early 1970s by the late Ephraim Katchalski-Katzir as ‘enzymes physically
confined or localized in a certain defined region of space with retention of their catalytic activities, and which can be used
repeatedly and continuously.’42 The first industrial use of immobilized catalyst has been reported in 1967 by Tanabe Seiyaku for
84 Reaction Engineering of Biotransformations

Isothermal reactor (IR) T-ramping (TR)

Experimental setup: Continuous X = f(cE,Sin,T,)

Membrane
reactor, e.g.,
lab-scale
enzyme T T
Time-dependent membrane
conversion in the reactor Jacketed
S in CSTR
respective experimental
settings

Conversion
at reactor
Conversion (−)

T-profile
outlet

T-profile

Time (h) Time (h)

Characteristics Operation of a continuous

Isothermal operation of a reactor with a linear temperature
continuous reactor gradient
Duration up to several days Duration 6−12 h
Multiple experiments at different Moderate substrate consumption,
temperatures moderate experimental
High substrate consumption, effort, and high complexity
high experimental effort, and
moderate complexity

Figure 6 Overview on experimental approaches for biocatalyst stability estimation and the corresponding typical conversion over time-diagrams.

the enantioselective hydrolysis of racemic N-acetyl amino acids by a DEAE-Sephadex immobilized aminoacylase.43 Over the years,
many different methods were developed and literally thousands of papers have been published on that topic. Principally,
immobilization techniques can be classified into carrier-based immobilization, entrapment or encapsulation of enzymes by a
polymeric network, and cross-linking of enzymes44 (see Figure 7). However, such a classification is by no means rigid in particular
as methods are often combined, for example, the adsorption of cross-linked enzyme aggregates (CLEAs) to mesoporous silica.45
Obviously, the properties of the support superstructure are crucial for successful immobilization as they should yield stabi-
lization of the enzyme without major leakage of enzyme or diffusion limitation of substrate(s) and product(s) and should provide
mechanical stability. Materials used as carrier or for entrapment are diverse including natural and synthetic polymers as well as
inorganic materials (see Table 4).46
Immobilization remains by and large a trial and error procedure. Still, understanding the fundamental mechanism of
immobilization facilitates rational selection of an appropriate procedure. For example, enzymes with potentially reactive groups
close to the catalytic center are likely to become inactivated by cross-linking, in particular when using a small molecule with high
mobility like glutaraldehyde.52 However, multimeric enzymes that predominantly inactivate by subunit dissociation can often be
greatly stabilized by cross-linking the subunits.27 Next, process economics must be considered. As a rule of thumb, enzyme costs
should not contribute more than a few percent to the total production costs requiring either very cheap immobilized enzymes or
highly stabilized enzymes. Consequently, the amount of product obtained per amount of supplied enzyme (often expressed as kg
product per kg enzyme) should be rather large. In the production of the semisynthetic antibiotic precursor 6-aminopenicillanic
acid (6-APA) by hydrolysis of penicillin G using an immobilized penicillin G amidase, one gram of immobilized enzyme yields
600 g of 6-APA.25 Similarly, a few grams of immobilized lipase yield hundreds of grams of the active drug ingredient Lotrafiban in
a final kinetic resolution step.53 Overall enzyme costs in documented processes have been calculated to 7–14 h per kg of active
pharmaceutical ingredient.53 Further, 11 tons of high fructose corn syrup (HFCS) can be obtained per kg of immobilized glucose
isomerase (D-xylose isomerase54) at temperatures of approximately 60 1C indicating the unsurpassed stability of this immobilized
 Reaction Engineering of Biotransformations 85

Carrier Entrapment/encapsulation
O
Covalent
OH NH linkage

− Ionic
− −− + interaction
−++
−+
− −− −
+ +−
Cross-linking
ic

Hydrophobic
Hydrophob

interaction

CLE, CLEC, CLEA

or van der Waals interactions,

hydrogen bonds
Figure 7 Overview of immobilization techniques. CLE, cross-linked enzyme; CLEC, cross-linked enzyme crystal; CLEA, cross-linked enzyme
aggregate.

Table 4 Selection of materials applied for entrapment and carrier-based immobilization

Method Material Comments

Carriers: (Covalent) attachment Biopolymers Cellulose, starch Requires functionalization

of the biocatalyst Synthetic Polymers Epoxy-functionalized acrylic resins Commercialized as Sepabeads/Eupergit47
Inorganic Mesoporous silica Easy functionalization, control of
structural parameters48
Clays, for example, bentonite Low costs46
Hydrogels: In situ entrapment Biopolymers Alginate Agarose Carrageenan Benign and easy synthesis49
of biocatalyst Synthetic Polymers Acrylates Benign synthesis50
Inorganic Sol–gels from metal alkoxides Control of structural parameters51

catalyst.25 Today, many large-scale processes such as the formation of HFCS, semisynthetic antibiotics, or acrylamide rely on
immobilized catalysts and are widely applied in resolutions and asymmetric and other syntheses (see Table 5).

Table 5 Some examples of industrial chiral synthesis using immobilized catalysts (extracted from Reference 2)

Enzyme (Class) Organism Immobilization Product Manufacturer, Scale

D-amino acid oxidase (1) Trijonopsis variabilis na Precursor for 7- Sandoz 4100 t a  1
aminocephalosporanic
acid
Lipase (3) Burkholderia plantarii na S-Phenylethylamine BASF4100 t a  1
Lipase (3) Candida antarctica Adsorbed to acrylic resin Lotrafiban GSK4100 kg
(Novozyme 435)
Penicillin amidase (3) Escherichia coli Covalently bound to (2R,3S)-Azetidinone Eli Lilly
carrier (Eupergit) Loracarbef precursor
Sialic acid aldolase (4) Escherichia coli Covalently bound to Neuraminic acid GSK (multiton)
carrier (Eupergit)
Fumarase (4) Brevibacterium flavum Entrapment of whole cells L-Malic acid Tanabe Seiyaku
in k  Carragenan 4500 t a  1

na, information not available.

In the following, a brief overview on the principal immobilization techniques is given. For a more comprehensive discussion,
the interested reader is referred to some very readable recent review articles26a,27,44,46 and the references therein.
86 Reaction Engineering of Biotransformations

7.5.3.3.1.1 Immobilization on a carrier

Carrier-based immobilization relies on covalent binding or physisorption of enzymes to the surface of the carrier. Highly porous
materials are predominantly applied in order to achieve sufficiently high enzyme loadings by providing a high specific surface
area. Next, protection from the external environment, for example, in case of air bubbles or organic solvents, often constitutes an
important stabilization mechanism that requires dispersion of the enzyme within the carrier matrix. The dispersion of macro-
molecules within the immobilization matrix prevents undesired interactions, for examples, with proteases present in CFE or other
components. In contrast, after attachment to nonporous materials such as nanoparticles, enzymes are in principle openly exposed
to the external environment and no particular stabilization with respect to above mentioned deactivation sources can be
expected.26a Covalent linkage of the enzyme to the carrier is irreversible and provides a much stronger binding than physical
interactions that are typically reversible. Physisorption can be achieved, for example, (1) via ionic interaction of the polycharged
protein with a functionalized resin, (2) via hydrogen bridges acting, for example, on sugar moieties of glycosylated proteins, or (3)
via entropy-driven hydrophobic interaction between a hydrophobic carrier and hydrophobic patches of proteins in an aqueous
medium. Obviously, appropriate conditions in view of pH, salt concentration, or solvent composition need to be applied in order
to promote such interactions. However, conditions can be deliberately changed after use in order to free the adsorbed (wasted)
enzymes and hence enabling regeneration of the carrier. Noncovalent adsorption is particular useful for reactions carried out in
organic solvent as leakage is prevented by the insolubility of enzymes in such solvents.
Covalent binding is typically accomplished by generating electrophilic groups on the support that can react with nucleophiles
on the protein surface such as an e-amino group of lysine, a thiol from cystein, or the carboxyl groups from aspartic or glutamic
acid. Establishing multiple linkages between enzyme and support can result in rigidification of the protein structure rendering it
less susceptible to thermal motion and hence increasing thermostability. A direct correlation between number of covalent bonds
and thermostability was found for some protease immobilized on glyoxyl-agarose supports.55 Next, multipoint attachment is
especially useful for multimeric enzymes in case multisubunit immobilization can be achieved, which depends on the support
and might require the use of specifically tailored linkers (Figure 8).56

Single linkage Rigidification by

no restraints on Subunit stabilization
multiple linkages by multiple linkages
thermal motion

Figure 8 Concept of multipoint immobilization for rigidification.

7.5.3.3.1.2 Entrapment/encapsulation
The terms entrapment and encapsulation are often used interchangeably in the literature. Both methods rely on confining the
enzyme to a defined space by establishing diffusion barriers. The location of the predominant diffusion barrier can in principle be
used to distinguish between supports with regularly distributed diffusion barriers (entrapment) and supports with a reinforced
diffusion barrier at the surface (encapsulation)57 (Figure 7). As no direct interactions of the support with the enzyme are present,
the enzyme is in principle preserved in its native state and retains its intrinsic properties. Entrapment is typically achieved by
benign hydrogel formation or sol-gel synthesis in the presence of the enzyme forming a defined polymeric network in which large
molecules like proteins become entrapped. Hydrogels can be obtained by chemical polymerization (chemotropic gels) such as
polyacrylamide or cross-linked dextran gels,50 by chelating reactions such as addition of calcium to an alginate solution (che-
latotropic gels)49a or by cooling (psychrotropic gels, e.g., carrageenan and agarose).49b,50 Sol–gels are porous silica materials that
can be shaped in any desired way and are formed by hydrolytic polymerization of metal alkoxides such as tetraethoxysilane. By
appropriate selection of precursors and conditions, structural parameters such as pore size as well as the hydrophobicity of the
formed polymer can be adjusted and hence the material can be tailored to the specific requirements of the respective enzyme.27,44
For example, using RSi(OMe)3 nonhydrolyzable alkyl moieties and Si(OMe)4 as precursors yielded a rather hydrophobic polymer
that results in hyperactivation of a Pseudomonas putida lipase, whereas sol–gels produced from Si(OMe)4 provided disappointingly
low activity.58 In a similar manner, cryogels can be obtained by freezing of the initial system. Such gels have been successfully
applied for the entrapment of whole cells.50 Next, bio-inspired mineralization constitutes another inexpensive, robust, and
particularly mild approach for biocatalyst entrapment that mimics biosilification processes as, for example, observed with diatoms
or marine sponges.51b
 Reaction Engineering of Biotransformations 87

Adsorption of large protamine molecules on the surface of hydrogel beads followed by silica precipitation or using water-
insoluble crosslinkers that can only act on the interface of a hydrogel bead/oil emulsion results in beads with a tailored structure
on the bead’s surface.57 In this way, the main diffusion barrier is established at the outer surface resulting in encapsulation of
enzymes in particular if the inner matrix is subsequently chemically dissolved.

7.5.3.3.1.3 Self-immobilization
Cross-linking aggregates of precipitated enzymes into CLEAs has emerged as an efficient immobilization technique over
the past decade. Similar older methods such as cross-linking solubilized enzymes (leading to CLEs) or enzyme crystals (CLECs)
have proven either ineffective or suffer from the laborious preparation of enzyme crystals. In contrast, CLEAs are
prepared by precipitation of the enzyme from aqueous buffer and a CFE and subsequent cross-linking of the physical aggregates.
Obviously, reaction conditions and selection of cross-linker bear a large optimization potential that can be covered, for
example, by high-throughput experimentation. A large variety of enzymes have been successfully immobilized including
many hydrolases and lyases demonstrating the broad applicability of this method.44,52,59 The resulting matrix consist
predominantly of the protein mix from the CFE (the small cross-linker molecules can be neglected) in contrast to carrier
immobilized enzymes where 90–99% of total mass consists of noncatalytic ballast.25 Therefore, the obtained productivities
(with respect to the mass or volume of the immobilized catalyst) are typically much higher for CLEA catalysts. However, in
particular for large clusters diffusion limitations can occur, and small CLEA aggregates are often not amenable to packed bed
reactor operation. By adsorption of CLEAs in larger superstructures like mesoporous foams60 mechanically stable catalyst can be
obtained.
Next, cross-linking has been applied for fixing the quaternary structure of multimers by establishing covalent bonds between
the subunits. Consequently, enzyme dissociation is prevented and indeed increased thermostability was observed for such
constructs. This strategy was predominantly applied to solubilized enzymes, but prevention of subunit dissociation was also
observed for CLEAs of lactase.56

7.5.3.3.1.4 Some remarks on immobilization of lipases

It is commonly presumed that most lipases exist in two different structural conformations – one with a polypeptide chain
blocking the active center from the environment (referred to as closed lid) and another with the lid open. Computer simulations
suggest that the closed lid configuration provides a less hydrophobic surface and hence is preferred in hydrophilic environments.
However, in the presence of hydrophobic interfaces, the open configuration (generally deemed the more active configuration)
should be preferred.27 This effect has been exploited in the immobilization of lipases using hydrophobic supports26a or addi-
tives61 resulting in significant activation of the lipase in particular for small hydrophobic substrates. Similarly, cross-linking in the
presence of a detergent resulted in a hyperactivated CLEA prepared from a lipase.62
Interestingly, enantioselectivity of lipases can be modulated by immobilization and the selection of the support. Covalent
linkage as well as physical interactions with the support can, in principle, alter the conformation of an enzyme and its flexibility
during reaction with unpredictable results – including changes in enantioselectivity. Such an effect has indeed been observed for
many lipases, in some cases even an inversion in enantiopreference could be induced by conditions applied during the immo-
bilization procedure.26a

7.5.3.3.1.5 Mass transfer and diffusion limitations of immobilized biocatalysts

The migration of a substrate molecule to the active center of the enzyme becomes more complex when supplying the enzyme in
immobilized form. Both product and substrate have to transfer between the bulk phase and the pores of the immobilization
matrix, a process that is driven by the concentration gradient between the bulk and the interface. Please note that in case the fluid
phase inside the pores of the carriers has a different composition compared to the bulk phase (e.g., an organic bulk phase and
aqueous phase in the pores), the partitioning of solute needs to be accounted for. Inside the pores motion of the solute is
controlled by diffusion along the concentration gradient as stagnant flow is typically assumed. While diffusing through the pore,
the substrate can interact with the enzyme linked to support matrix and hence be converted into product. For spherical
immobilization beads with a bed porosity ep and a radius rp the mass balances (in this case expressed by concentrations c) of a
given solute i in the bulk and pore are

dcbulk 3kL bulk

Bulk : ep i
¼ ð1  ep Þ ðc  cpore ðr ¼ rp ÞÞ
dt rp i i

 2 pore pore 
q cpore q ci 2 q ci
Pore :i
¼ Dpore
i þ þ vi
qt q r2 r qr
pore 
pore pore q ci 
Boundary conditions : kl ðcbulk  ci ðr ¼ rp ÞÞ ¼ Di
i
q t r ¼ rp

with kL denoting the mass transfer coefficient and Dpore the diffusion coefficient in the pore. In order to determine
whether the immobilized catalyst suffers from diffusion or mass transfer limitations, an efficiency coefficient Z is
88 Reaction Engineering of Biotransformations

introduced that relates the observed reaction rate to the expected rate obtained with free enzyme of the same
concentration.

Observed rate
Z¼
Rate at bulk concentration

It is possible to estimate the efficiency coefficient a priori from characteristic numbers such as the Thiele or the Weisz modulus
and the Sherwood number.25

7.5.3.3.2 Catalyst stabilization and protein engineering

Next to immobilization, a number of other principal approaches have been developed for obtaining stabilized proteins: Mining
the genomes of extremophiles populating extreme environments with conditions similar to those of the projected reaction
conditions such as, for example, high temperature or harsh pH typically provide enzymes that also show high resistance to the
respective stress in isolated form. However, naturally occurring enzymes show a distinct trade-off between (thermo)stability and
activity. Homologous enzymes from hosts adapted to different temperature regimes often show comparable reaction rates at their
respective optimal temperature.63 Therefore, a thermophilic enzyme selected because of its superior thermostability might only
display low reaction rates at reduced temperature. However, psychrophilic enzymes will – according to Arrhenius’ law – be in
orders of magnitude more active at high temperatures than their meso- or thermophilic counterparts but inactivate rapidly due to
their very limited stability. In many protein engineering studies, it was demonstrated that biocatalysts could be obtained that were
highly active and stable at the same time. Consequently, the required level of flexibility for efficient catalysis around the active
center and the structural rigidity required for stability can be addressed in the same protein molecule as they presumably address
different positions in a protein. Hence, this trade-off observed for natural enzymes is not a result of fundamental laws of
thermodynamics but probably rather an effect of natural evolution. Introduction of mutations that are neutral – in the sense that
they do not affect the fitness of an organism – act with a high probability deleterious toward stability and hence there is a constant
tendency to maintain stability only at the required temperature. Next, enzymes with high stability and activity are potentially
impaired with respect to control of the enzyme concentration level and might therefore constitute an evolutionary disadvantage.63
Consequently, in order to obtain enzymes with high activity and stability protein engineering methods are required. Princi-
pally protein redesign can be performed by (1) rational introduction of stabilizing mutations based on structural or sequence
information and knowledge of the principal deactivation mechanism; (2) by a combinatorial approach based on large
libraries with randomly introduced mutations (directed evolution); or (3) an in silico de novo construction of
thermostable structures.26b,35c,63,64 All these methods face certain limitations and are therefore often used in combination.65
Destabilization typically originates from a locally defined segment of the protein and can act via different mechanisms (see
Section 7.5.3.1), and hence successful stabilization requires redesign of the respective area using adequate substitutions. Due to
the heterogeneity of deactivation mechanisms and protein structures, no generically applicable rational stabilization strategy is
available. However, some rather successful methods can be extracted from recent studies:64

• ‘Entropic stabilization’ or rigidification by Gly-Ala and Xxx-Pro mutations or introduction of disulfide bridges
• Improvement of molecular packing, for example, by shortening loops
• Modification of surface charge networks
• Reinforcement of higher oligomerization states

Identification of hot spots in the protein sequence constitutes an efficient first step on which a limited set of combinatorial
variants obtained by site-directed saturation mutagenesis or specific mutations derived by computational design can be tested.66
Such potentially destabilizing residues can be estimated by sequence alignment of different homologous enzymes and selection of
those with the highest variability (and allegedly highest potential for stabilization)67 or by properties derived from crystal
structure analysis such as the B-factor that correlates with flexibility.68 Substitutions that result in a reduction of these properties
are expected to lead to more stable proteins.68 For many proteins, structural information and deactivation mechanisms have
already been reported that enable hypothesizing about potential stabilizing substitutions and hence a rational identification of
hot spots. Alternatively, unprejudiced directed evolution allows for an unbiased screening of sequence space for positions or
regions inducing stabilization, which after identification can be comprehensively studied by site-directed mutagenesis.69 However,
multiple rounds of directed evolution have yielded similar impressive results,63 despite the fact that only a small part of sequence
space can actually be covered this way. Even to cover a small part of sequence space, large numbers of clones have to be processed
requiring high-throughput methods. As assaying for stability often represents the throughput-limiting step, preselection of
functional variants, for example, in a growth-based in vivo assay and subsequently performing the stability assay with a smaller set
of functional clones constitutes a very elegant solution.70 Next, sophisticated bioinformatic tools have been developed that in
principle allow for a de novo generation of large parts of the protein structures with presumably much enhanced stability. However,
so far such tools are predominantly applied for identification of a limited number of potentially beneficial substitutions that are
then tested experimentally and evolved further by experimental techniques.71
Impressive stabilization can already be obtained by a single mutation,72 but is also often obtained by the cumulative effect of a
larger number (5–10) of mutations, each of which has only a minor beneficial effect. Some examples of successful stabilization by
protein engineering are listed in Table 6.
 Reaction Engineering of Biotransformations 89

Table 6 Selected examples for successful enzyme stability engineering

Enzyme Method Stabilization References

Prolyl endopeptidase Computational analysis of variability of homologues 200-fold increase resistance to 66

Sphingomonas capsulate suggests beneficial mutations. From those 47 variants pepsin digestion
were constructed that were experimentally evaluated and
the results used in a sensitivity analysis for obtaining
suitable variants for the next round. 91 variants tested,
five amino acid (aa) substitutions
o-transaminase In silico docking studies combined with site-directed Operation in 50% DMSO at 45 1C 73
Arthrobacter sp. mutagenesis enabled engineering activity for the possible starting from an enzyme
nonnatural substrate. Directed evolution (10 rounds) with no activity and little stability
gradually improved thermostability and resistance for
cosolvent DMSO. 27 aa mutations
Subtilisin E Bacillus subtilis Five rounds of directed evolution based on error prone 200-fold increase in t1/2 at 60 1C 63
PCR and recombination using residual activity after
incubation at elevated temperatures for screening. 8 aa
substitutions
Thermolysine-like protease Comparison with more stable thermolysine variants and Boilysin: t1/2 (100 1C)¼170 min 74
Bacillus rationally selected stabilizing substitutions (e.g., wt: t1/2 (90 1C)¼1.5 min
stearothermophilus Ala-Pro and introduction of disulfide bridges) yielded
hyperthermophile eightfold mutant
a-amylase Bacillus Analysis of previous studies and systematic replacement of 50-fold increase in t1/2 at 85 1C 75
lichenformis Asn/Gln sites prone to deamidation resulted in fivefold
mutant
Xylanase Screening for increased thermostability from a DTeq ¼35 1C 76
comprehensive library of point mutations introduced by
site saturation mutagenesis technology identified nine
beneficial mutations. Combinatorial assembly of
mutation yielded a number of much improved variants

Yeast cytosine deaminase In silico redesign of a predefined structural backbone (with DTeq ¼10 1C 71
exception of the active site region) using a random
iterative procedure and optimization toward minimized
free energy. Inserts are selected according to the
Dumbrack rotamer library. Manual identification of most
relevant substitutions (3aa) in comparison with wild type

wt, wild type.

7.5.4 Typical Reactors

7.5.4.1 Selection and Design of Bioreactors

Principally, bioreactors for biotransformations can be classified into three fundamental operation modes: (1) the batch reactor or
ideal stirred tank reactor (BR), (2) the continuously operated stirred tank reactor (CSTR), and (3) the plug flow reactor (PFR) (see
Figure 9). Briefly, operation of biotransformations in a CSTR requires the installation of a membrane in order to retain the
enzyme within the reaction space. Typically, ultrafiltration membranes are applied when the enzyme is supplied in solubilized
form, for whole cell biotransformations or immobilized enzymes larger cutoffs in the range of microfiltration membranes can be
accepted. PFRs are percolated tubular reactors that for operation of enzyme reactions are equipped with a packed bed of
immobilized catalyst since it is impossible to obtain a homogeneous and sufficiently permeable bed with solubilized enzymes.
Due to the assumed plug flow regime, no concentration gradients occur in radial direction. Historically, the BR was selected for
most processes due to its simplicity and the use of standard equipment. However, continuous operation (CSTR and PFR) is
increasingly considered as a key technology in biotechnology, which in fact has been realized for some biotechnological processes
(e.g., lactic acid or human insulin production.77)
For the present discussion, ideal mixing is assumed for stirred tank reactors in both batch and continuous operation resulting
in a homogeneous distribution of all involved reaction species over the whole reaction space. Although in batch mode a steady
state is only obtained when the reaction equilibrium is reached (at which point the process will be stopped anyway), continuous
operation usually results in a steady state different from the equilibrium, the precise conditions of which are a function of
the residence time (provided that the substrate feed concentration and the flow rate are constant over time). However, when the
substrate is supplied to the percolated column of the PFR, the substrate becomes gradually depleted while traveling through the
90 Reaction Engineering of Biotransformations

S in S

v(S)
r(S) r(S)
u Sz Sz+Δz

Batch reactor Continuous stirred 0 z z + Δz L

BR tank reactor Plug flow reactor
CSTR PFR
Figure 9 Ideal reactor schemes.

column. In this case, concentration gradients evolve in axial direction, but this concentration gradient usually becomes inde-
pendent of time (steady state) and constant conversion is obtained at the reactor outlet.
Although real reactors differ from these ideal models, the models are frequently used in process design as they are often
sufficient to capture the central elements of the reaction design, and more accurate prediction of the flow regimes typically
requires disproportionately more elaborate modeling and simulations. In general, bioreactor design can be rationally conducted
using mass and energy balances of the reactor based on a comprehensive characterization of the biochemical reaction that
includes (1) thermodynamic data, (2) a parameterized reaction kinetics model including mass transfer (in case of immobilized
catalysts), and (3) the stability of the enzyme under reaction conditions. Energy balances are of practical importance for calcu-
lating the heat transfer rates in order to achieve temperature control in the reactor which is often neglected for isothermal
operation due to the typical low heat of reaction of biotransformations.78 Fundamental process performance parameters such as
the conversion X and the space-time-yield STY are obtained by mass balances that combine reaction kinetics with a representative
reactor model.
In order to characterize the different reactor configurations here, their performance is evaluated for biotransformations that can
be described by simple one-substrate/one product Michaelis–Menten kinetics using generic dimensionless expressions derived
from the mass balance. In this way, the conversion X can be simply expressed as a function of a dimensionless time t and a
reduced Michaelis–Menten constant KM . In order to derive such an expression the reaction rate v, here Michaelis–Menten
kinetics, needs to be expressed as a function of conversion:

X¼
P  ¼ KM
KM
S0=in S0=in
vðXÞ 1X
v0 ðXÞ ¼ ¼ X
vmax 1 þ KM

The conversion is defined as ratio of product concentration P in the reactor or reactor outlet to initially supplied substrate
concentration S0 or continuously fed substrate concentration Sin for batch and continuous operation, respectively. Similarly, the
dimensionless Michaelis–Menten constant KM denotes the ratio of actual Michaelis–Menten constant KM to initially supplied
substrate concentration S0 or continuously fed substrate concentration Sin.

7.5.4.1.1 Batch reactor

In case of the batch reactor, the mass balance constitutes an ordinary differential equation that requires initial conditions (IC)
such as the initially supplied substrate concentration S0:
dS
¼  vðSÞ IC : Sðt ¼ 0Þ ¼ S0
dt

By substituting the operation time t with the dimensionless time t defined as the product of limiting rate vmax and operation
time t divided by the initially supplied substrate concentration, the mass balance can be expressed as a function of conversion in
dimensionless form to calculate the time required to achieve a certain conversion:
dX vmax t
¼ v0 ðXÞ tBR ¼
dtBR S0

Z IC : XðtBR ¼ 0Þ ¼ 0
X
tBR ¼
1
¼ X  KM ln ð1  XÞ
0
0 v ðXÞ

Using this dimensionless expression, a number of principle considerations can be illustrated. Typically, reactor performance is
characterized by a volumetric productivity (amount of product per reactor volume per time) which can be expressed in
 Reaction Engineering of Biotransformations 91

dimensionless form STY by division with the limiting rate vmax:

X
STY ¼
tBR

With increasing dimensionless time t the productivity gradually decreases whereas yield (which in the absence of side reactions
coincides with the conversion) increases until full conversion is reached (see Figure 10). Hence, the two objectives on the process
engineer’s wish list, maximization of productivity and yield, cannot be aligned but need to be balanced in the face of costs such as
the yield on substrate (substrate costs), product purity (typically influences downstream processing costs), and equipment costs.
Please note that similar relations are obtained for the CSTR and the PFR case.

0.9
1
Productivity

0.8 0.8
STY (−)

0.6

X (−)
0.7

0.4

0.6
0.2
Yield/conversion

0.5 1 1.5
BR (−)

Figure 10 Productivity and yield/conversion as a function of residence time t calculated for KM ¼ 0:2.

Next, calculating the conversion over a large range of the parameter space spanned by the dimensionless time t and reduced
Michaelis–Menten constant KM (Figure 11) further indicates the deleterious effect of increasing KM values on conversion. For
larger values of KM the reaction rate is well below the limiting rate vmax and hence conversion becomes limited by the high KM
relative to the initial substrate concentration S0. Roughly, significant conversion can only be obtained for tBR40.5 with KM o0:1.

Change vmax, t
X (−)

Change S0
0.5

0
10−2 100
10−1 10−1
100
KM* (−) BR (−)

Figure 11 Conversion as a function of dimensionless KM and dimensionless time t for batch operation. The diverging behavior of changes in
the operational parameters S0 (black lines) and vmax and operation time t (blue lines) is illustrated. Equivalently this graph represents PFR
operation by substituting tBR with tPFR. For the PFR case, the blue line represents changes in the residence time given by the ratio of reactor
length L to interstitial flow velocity u (L/u).
92 Reaction Engineering of Biotransformations

Further, a number of basic principles can be deducted from the dimensionless parameters. Enzyme concentration and thus
vmax, the operation time t, and initial substrate concentration S0 constitute operational parameters amenable to process design.
Obviously, the former two parameters are proportional to the dimensionless residence time t and hence equally affect conversion.
As can be deduced from the graphical representation, increasing vmax or t can result in a significant increase in conversion
(Figure 11, blue lines). Changing the third parameter of tBR, the substrate concentration S0 at the same time yields a change in
KM . Hence, a change in S0 results in a different path in the X  KM –t space (Figure 11, black lines) compared to changes in
enzyme concentration or operation period. In particular when reducing the substrate concentration to the same order as KM, the
increase in conversion due to the increase in tBR is partially offset by the increase in KM . Next, decreasing the substrate
concentration also reduces the obtainable product concentration which reduces productivity and potentially aggravates sub-
sequent downstream processing.
From a protein engineering standpoint, the incentive to develop novel variants with a high kcat to KM ratio can be easily
illustrated by the expected increase in conversion that results from an increased vmax and a reduced KM .

7.5.4.1.2 CSTR
The mass balance for the CSTR in the general case involves an ordinary differential equation that can be reduced to an algebraic
equation when assuming steady-state operation. Again, in order to derive generally valid results dimensionless parameters are
introduced:

dS QðSin  SÞ steady state

¼  vðSÞ ¼ 0
dt VCSTR

P
X¼
Sin
vmax VCSTR X  Xin
tCSTR ¼ ¼ 0
Q Sin v ðXÞ
Again, the conversion in the reactor depends on KM and a dimensionless time constant tCSTR which in this case depends on
the feed concentration Sin, the limiting reaction rate vmax, and the residence time expressed by the ratio of reactor volume VCSTR to
flow rate Q.
The obtained results for conversion in a CSTR (see Figure 12) suggest a very similar behavior compared to the batch case and
hence the same conclusions apply.

Change vmax, t
X (−)

0.5
Change S0

0
10−2
10−1 100
100 10−1 CSTR (−)
KM* (−)

Figure 12 Conversion as a function of dimensionless KM and dimensionless time t for CSTR operation. The diverging behavior of changes in
the operational parameters Sin (black lines) and vmax and CSTR residence time (ratio of reactor volume VCSTR to volumetric flow rate V) (blue
lines) is illustrated.

7.5.4.1.3 PFR
The mass balance for the PFR in its general form is based on a partial differential equation consisting of derivatives of substrate
concentration in time and space. However, for steady-state operation the mass balance is reduced to an ordinary differential
equation that can be solved analytically. Interestingly, the derived expression is identical to the batch reactor case differing only in
 Reaction Engineering of Biotransformations 93

the definition of the dimensionless time parameter tPFR, which is a function of the residence time in the PFR (calculated from the
axial reactor coordinate z and interstitial flow velocity u), the feed concentration Sin, and the limiting reaction rate vmax. Con-
sequently, the results obtained for the batch case can be transferred directly to the PFR case (see Figure 11).

dS dS steady state
¼  u  vðSÞ ¼ 0
dt dz
vmax z
tPFR ¼
u S0
dX
¼ v0 ðXÞ
dtPFR

7.5.4.1.4 Comparison CSTR–PFR

Based on the introduced mass balances, performance comparison of CSTR and PFR can be easily conducted. Due to the ideal
mixing in the CSTR, the substrate concentration in the reactor is instantly reduced with respect to the feed concentration, whereas
in the PFR the substrate concentration decreases gradually from the feed concentration in axial direction providing a stronger
driving force of the reaction for reaction kinetics with reaction orders n40. Depending on the ratio of KM to Sin the reaction order
n of Michaelis–Menten kinetics varies between n ¼ 0 for KM ooSin with vEvmax and n ¼ 1 for KM 44Sin with vEvmaxS/KM (compare
Section 7.5.2.2). Consequently, for the same conversion CSTR operation principally requires larger residence times compared to
PFR operation for all scenarios where KM ooSin is not fulfilled (Figure 13).

300
ΔCSTR−PFR (%)

200

100

0
0.9

0.7 100
X (−) 10−1
0.5 10−2
KM* (−)

Figure 13 Percentage increase of residence time required to obtain the same conversion in a single CSTR compared to a PFR as function of
conversion X and reduced Michaelis–Menten constant KM .

Please note that in practice PFR operation requires immobilized biocatalyst which is frequently accompanied by a dilution of
catalytic activity and hence the achievable overall enzyme concentration in the reactor can be smaller when compared to
solubilized enzymes in membrane reactors (CSTR). Hence, the achievable reaction rates expressed by vmax are potentially higher
for CSTR offsetting at least to some degree the intrinsic disadvantage of the stirred tank operation mode, which anyways becomes
negligible when operating with high feed concentrations with respect to the KM. Consequently, the selection between CSTR and
PFR must be made on a case by case basis that includes the reaction analysis but should also consider additional cost for enzyme
immobilization, potential stabilization due to immobilization, and ease of reuse.

7.5.4.1.5 CSTRs in sequence

Implementing cascades of CSTRs constitutes a feasible technical solution for increasing the performance. When comparing a
single CSTR to a cascade of two CSTRs with the same overall residence time (reactors in the cascade each have half the volume of
the single CSTR and the absolute flow rate is the same for single and cascade operation), the substrate concentration in the first
reactor of the cascade is higher than the concentration in the reactor in the single-reactor case. This provides a higher driving force
for the reaction. Therefore, higher conversion can be achieved with CSTR cascades. Again, this is obviously only valid for reaction
orders n40. With increasing number of reactors, CSTR performance approaches PFR performance which constitutes the upper
performance boundary for CSTR cascades (see Figure 14).
94 Reaction Engineering of Biotransformations

PFR

X (−)
0.5

0
10
100 5
 (−) 10−1
1 # CSTRs (−)

Figure 14 Conversion as a function of the cumulative dimensionless residence time t and the number n of reactors applied calculated for
KM ¼ 0:2 and uniform distribution of residence time for all reactors (tCSTR ¼t/n).

7.5.4.2 Reactor Concepts

7.5.4.2.1 Membrane bioreactors
Membrane technology is instrumental to confine soluble enzymes to the reaction space of continuous tank reactors. The com-
bination of a continuous reactor with a membrane for enzyme retention is often termed as enzyme membrane reactor. On small
scale, the membrane is typically directly integrated in the reactor segregating the reactor space into a reaction area and smaller
noncatalytic space at the reactor effluent with an undefined flow regime (Figure 15(a)). On a larger scale, such a direct integration
is realized by submerging a membrane module in the reactor (Figure 15(b)), however at production scale, spatially separated
setups of tank reactor and membrane module are predominantly applied (Figure 15(c)).79 The availability of large-scale
equipment and more importantly the application of cross-flow filtration units constitute a major advantage of decoupled setups.80
In comparison to dead-end filtration (Figure 15(a)), cross-flow filtration typically provides higher permeate flow rates at com-
parable pressure drop and less fouling due to a significant reduction in concentration polarization resulting from the high shear
rates induced by the cross flow.81

E E E E E
E E E
E E E

E E E E E
E E E E
E E
E
(a) (b) (c)

Figure 15 Membrane reactor schemes: (a) in situ integration of reactor and membrane. (b) Immersed membrane module in reactor tank.
(c) External ultrafiltration module with recycle.

In an integrated reactor filtration process, the effluent of the reactor is directed to a membrane module that divides the
incoming mass flow of solvent, enzyme, product, and unconverted substrate into an enzyme-free product-rich permeate stream
and an enzyme-rich product-deprived retentate stream which is recycled to the reactor. Therefore, membranes are required that are
not permeable for protein-sized molecules, such as ultrafiltration membranes. Typically, more hydrophobic proteins are best
retained by hydrophilic membrane materials such as regenerated cellulose and hydrophilized polyethersulfone. In turn, hydro-
philic proteins are best retained with hydrophobic materials such as polysulfone-based membranes.80b Such membranes allow
high membrane flux and high rejections for proteins and are typically very stable at enzyme operation compatible pHs of 4–9.82
Integration of biotransformation and enzyme retention by membrane is applied for the industrial production of amino acids
on the several hundred tons per year scale. For example, L-methionine can be produced via kinetic resolution of racemic N-acetyl-
methionine by free aminoacylase from Aspergillus oryzae (Evonik, Germany) utilizing an ultrafiltration unit for retention of the
 Reaction Engineering of Biotransformations 95

enzymes.53,80b,83 Membrane technology is further enabling selective removal of substances as demonstrated by the production of
L-phenylalanine via resolution from racemic phenylalanine-isopropylester by free subtilisin Carlsberg from Bacillus lichenformis
that utilizes a liquid membrane for selective extraction of the unconverted substrate enantiomer whereas the charged product
remains in the shell phase (Coca Cola, USA).53,83 In this way, L-phenylalanine can be obtained in high purity and the isolated D-
ester can be racemized and recycled to the reaction. Membrane reactors can also be applied to cofactor dependent reactions. As
ultrafiltration membranes do not allow rejection of molecules in the typical size range of cofactors such as NADH or ATP,
cofactors need to be covalently bound to a water-soluble polymer such as polyethyleneglycol (PEG) in order to enlarge their size.
In this way, L-tert leucine can be produced by reductive amination of trimethylpyruvate using the NADH-dependent leucine
dehydrogenase with cofactor regeneration (Evonik, Germany).53,80b,83

7.5.4.2.2 Immobilized enzyme reactors

Immobilization provides the means to apply enzymes in packed bed reactors that yield a number of operational advantages
compared to CSTR reactors. As demonstrated in the section above, higher space-time yields can be obtained for typical enzyme
kinetics. Next, immobilized catalysts can be easily reused, and often immobilization provides stabilization of the biocatalyst
(compare Section 7.5.3.3.1).
However, it should be noted that immobilization usually yields a dilution of the catalytic activity (with the exception of CLEAs,
see Section 7.5.3.3.1.3) and that mass transfer limitation (compare Section 7.5.3.3.1.5) can occur which can potentially offset
some of the advantages mentioned above.
Typically packed bed reactors are operated with upward flow direction in order to avoid compression of the bed and for easier
release of gas bubbles (Figure 16(a)). In case the immobilization matrix does not provide the necessary mechanical strength for
packed bed operation or yields uneconomically high pressure drops (e.g., for small bead sizes that ensure high mass transfer),
fluidized bed operation represents a feasible alternative (Figure 16(b)). Obviously, the net concentration of the biocatalyst is
reduced in such a setting and hence lower space-time yields are obtained. By installing nets at different heights the hydrodynamics
can be influenced toward plug flow-like behavior. For even smaller particle sizes, slurry reactors that utilize a membrane for
immobilized catalyst retention are applied (Figure 16(c)). Due to the larger size compared to solubilized enzymes, microfiltration
membranes or filter sieves can be applied. It is worth noting that the energy input required for fluidization of the particles is much
higher in slurry reactors compared to fluidized bed reactors making the slurry reactor a viable solution only for very small
particles.80a

(a) (b) (c)

Figure 16 Schematic representation of bioreactors amenable to operation with immobilized enzymes. (a) Packed bed reactor. (b) Fluidized bed
reactor. (c) Slurry reactor.

7.5.4.2.3 Reactors for enzyme reactions in unconventional media

Principally, enzyme membrane reactors or the reactors introduced for immobilized biocatalysts can be applied to enzyme
reactions conducted in an organic solvent, provided that appropriate solvent-tolerant immobilization supports or membrane
materials are available. Membrane materials have been specifically developed for application in polar and apolar organic solvents
and are commercially available. Next, a number of immobilized lipases have been applied in organic solvent demonstrating the
principle compatibility of supports with unconventional media.
When an aqueous/organic two phase system is considered more advantageous (that is, an aqueous phase of a noticeable phase
ratio is applied), reactor operation poses some special challenges such as sufficient interfacial contact between the two phases. For
operation in biphasic liquid media packed beds with various fluid regimes, mixer settler cascades, emulsion cascades, and
membrane reactors have been suggested.84 Frequently, reaction and transfer to/from the applied auxiliary phase are compart-
mentalized by coupling an extraction unit or adsorption bed to a bioreactor.80a,85 Such integrated processes are discussed in detail
in the next chapter.
96 Reaction Engineering of Biotransformations

7.5.5 Current and Future Developments in Enzyme Technology

Biotransformation processes gradually enter the realm of traditional fossil raw materials-based bulk chemicals, which will only
succeed on a larger scale if the new processes indeed prove to be competitive compared to the conventional processes. Therefore,
biotechnology companies need to start to embrace the various chemical engineering tools available for process optimization. In
particular, the integration of development efforts from the molecular to reaction process and plant scale enables for a global
optimization and a true minimization of costs or other objectives such as the environmental footprint that certainly requires the
concerted efforts of a multidisciplinary team.86 Current trends in biotechnology process development include molecular-scale
optimization such as the engineering of dedicated production systems by synthetic biology principles87 or the identification of
biocatalyst with novel properties fueled by the ability to rapidly screen vast libraries, which enables to cover the large sequence
space that becomes available from random mutagenesis methods. Next, investigations on the reaction scale become increasingly
efficient due to the ongoing advances in miniaturized reactors and the application of design of experiment and model-based
analysis tools. Combining comprehensive reaction models with reactor and process models allows for in silico evaluation of
different process configurations over a range of operational conditions and identification of the most economic process variant
obviating the need for extensive experimentation. Further novel process variants are considered. In particular, integration of
reaction and product removal has attracted considerable interest as many of the most notorious constraints of biotransformation
can be addressed with this concept. In the following, these points are discussed in more detail.

7.5.5.1 Synthetic Biology

One important future research direction will be the rational design of catalytic systems to conduct ever more complex chemistry in
single reactors, either in living cells or in cell-free systems. Although the construction of living cells, typically bacterial or yeast
cells, to carry out complex natural product chemistry based on cheap carbon sources is making rapid progress,88 the focus here is
more on cell-free systems, as such systems suffer less from toxicity and mass transfer issues of nonnatural substrates and
products.89 However, such systems can still provide amazing catalytic complexity from cell-free protein synthesis to hydrogen and
monosaccharide production from glucose.90 Still, the crucial element in designing such systems remains the complexity of
assembling and then accurately describing the properties of multiple functional elements. Based on recent major increases in the
efficiency of de novo DNA synthesis, a number of efforts are on-going to systematize the construction of catalytic systems of
multiple elements, including standardization of functional elements,91 computer-aided design tools,92 and improved analytical
tools to study the system.93 Even though it remains to be seen in how far these efforts will reach in making the design of biological
multistep catalysis truly rational, these steps will definitely contribute to facilitating the set-up of multistep reaction systems that
are relevant for biotransformations.

7.5.5.2 Miniaturization
Miniature reactor technology constitutes an important tool for intensive screening of biocatalysts, be it the screening for enzyme
variants with improved properties or assessing enzyme performance over a range of process variables such as temperature, pH, or
solvent composition. The use of small volumes significantly reduces the amount of biocatalyst and substrate required for the
whole screening procedure and hence enables to probe large numbers of variants or variables. By now, different systems with
different volume ranges have been successfully developed, reaching from the milliliter94 to the picoliter scale.95 Microscale
reactors can be realized simply in microwell plates with stirring induced by shaking.96 Alternatively, encapsulating monoclonal
populations in nanoliter hydrogel beads constitutes another proven miniaturization concept that obviates the use of expensive
well plates and further reduces the processed volumes.97 Coupled to a powerful assay, for example, plate readers if a fluorogenic or
chromogenic substance is involved in the reaction, a coupled colorimetric assay in case of well plates, or particle-analyzer-based
sorting of nanoliter reactors based on fluorescence, such technologies can be operated with very high throughput. Next, current
microfluidic reactors realize even smaller volumes in the range of 1 nl to 1 pl and are compatible with continuous flow mode
operation. These often contain a network of flow channels and are often combined with other downstream unit operations (‘lab
on a chip’),96a,98 which is particularly useful for analytical and sensor applications. Together, these developments will play an
important part in accelerating the identification of suitable enzyme variants and of optimal reaction conditions.

7.5.5.3 Model-Based Process Design

In order to exploit a given biotransformation route to its full extent, in silico optimization based on validated reaction kinetics and
stability descriptions embedded in an adequate reactor model provides a time- and resource-efficient tool.11a,14a The benefit of
model-based process design based on comprehensive biocatalyst characterization can be easily demonstrated by considering the
selection of reaction temperature. An increase in temperature typically has a beneficial effect on the reaction kinetics and a
deleterious effect on the stability. Hence, for a specified reactor operation (yield, process time, supplied enzyme, and substrate
amount), a distinct temperature optimum exists with respect to the volumetric productivity (see Figure 17) or – if cost functions
are assigned to the various contributions – to the production costs. Next, having a reliable process model at hand provides
 Reaction Engineering of Biotransformations 97

educated estimates on productivity or process costs in a relatively early stage of the development without the need for excessive
testing on the pilot scale that comes with poorly understood and characterized reaction systems.

100

80

Norm. productivity (%)

60

40

20

0
300 310 320 330
Temperature (K)
Figure 17 Typical productivity–temperature correlation for optimal isocratic operation of enzyme reactors (simulated here for a continuous
enzyme membrane reactor operation with specified substrate feed, supplied enzyme amount, operation period with flow rates adjusted to provide
a specified conversion using a representative Michaelis–Menten kinetics and an extended equilibrium model for description of the operational
stability).

This example clearly serves as an incentive to probe the whole operational parameter space for identification of optimal
conditions. Please note that the usefulness of such process simulations directly correlates with the accuracy of the estimated
parameters that can be evaluated by applying statistical analysis14a,38 to the estimation procedure.

7.5.5.4 Integrated Operation of Reaction and Separation

Integration of product formation and separation constitutes an attractive solution for overcoming frequently encountered con-
straints in biotransformations such as (1) an unfavorable position of the thermodynamic equilibrium, (2) product inhibition, (3)
rapidly degrading products, or (4) reactions under kinetic control.99 By selective removal of product from the reaction space, the
residence time of product as well as the product concentration can be adjusted which are important parameters to minimize some
of the above mentioned constraints (2–4). Next, by selective removal of product the yield on substrate can be extended well
beyond the equilibrium for thermodynamically limited reactions. Such in situ product removal (ISPR) concepts obviously require
the use of an auxiliary phase and can be realized in different process configuration that can be schematically summarized by the
basic cases in Figure 18. In order to isolate the product selectively, differences in hydrophobicity (adsorption, extraction, and
perstraction), volatility (distillation, evaporation, and pervaporation), charge (reactive extraction and electrodialysis), size
(nanofiltration and molecular sieves), or specific binding properties (complexation) can be exploited.85a

Direct Indirect Batch Continuous

Contact Operation

Internal External
Product removal

Figure 18 Basic ISPR reactor configuration.

Attractive candidates for such an approach are, for example, aldolase- and isomerase-catalyzed reactions that provide direct
access to interesting molecule classes such as rare or unnatural saccharides and saccharide-like compounds. Classical chemical
synthesis of these compounds requires frequently intensive protection group chemistry,100 motivating the development of
98 Reaction Engineering of Biotransformations

alternative and in particular biocatalytic routes, where enzyme regioselectivity can facilitate the synthesis. However, the thermo-
dynamic equilibrium of those direct biocatalytic reactions frequently favors the substrate,101 and the resulting limitation in
product yields clearly offsets the advantages of regioselectivity and also of sustainable reaction conditions (mild temperature and
pH, aqueous solvent). For such a system, the production of commercially attractive D-psicose by epimerization of cheap
D-fructose, an integrated process consisting of an enzyme membrane reactor and a simulated moving bed (SMB) unit (continuous
chromatography), was implemented and high yield and robust operation were demonstrated.102 Chromatography typically
provides the separation power for difficult-to separate mixture and should therefore constitute a generic process solution for such
reactions (see Figure 19).

Reaction-separation integration

EMR SMB
Biotrans-
formation
S+M P

Separation
S P

Product
S M P
Substrate

Concentration

Recycle

NF

Figure 19 Integration of biotransformation, SMB, and nanofiltration concentration for the production of fine chemicals. The nanofiltration is
required in order to account for the inherent dilution in the SMB.

Physical separations are increasingly applied for the retrieval of enantiomers from racemic mixtures, most prominently SMB103
and crystallization.104 However, like all physical separation methods the yield from racemates is limited to 50%, which can be
addressed by integration of physical separation with, for example, a mild enzymatic racemization.105

7.5.6 Summary

The utilization of enzymes in the manufacturing of chemical compounds, including optically active compounds, has become an
indispensable tool of the fine chemical industry. Correspondingly, more and more focus is directed to the transfer of such
reactions from the laboratory to the process scale. The required methods for that are in part broadly established and in part still in
development. However, as demonstrated by a broad variety of already implemented enzyme-based processes,53,106 the integration
of enzymes into the production schemes of the chemical industry does not present major hurdles.

References

1. Faber, K. Biotransformations in Organic Chemistry, 5 ed.; Springer: Berlin, 2004.

2. Liese, A.; Seelbach, K.; Wandrey, C. Industrial Biotransformations; Wiley-VCH: Weinheim, 2000.
3. Straathof, A. J. J.; Panke, S.; Schmid, A. Curr. Opin. Biotechnol. 2002, 13(6), 548–556.
4. (a) Panke, S.; Wubbolts, M. G. Curr. Opin. Chem. Biol. 2005, 9, 188–194. (b) Schnell, B.; Faber, K.; Kroutil, W. Adv. Synth. Catal. 2003, 345(6–7), 653–666.
5. Bommarius, A. S.; Schwarm, M.; Drauz, K. Chimia 2001, 55, 50–59.
6. Michaelis, L.; Menten, M. L. Biochem. Z. 1913, 49, 333–369.
7. Segel, I. H. Enzyme Kinetics: Behavior and Analysis of Rapid Equilibrium and Steady-State Enzyme Systems; Wiley-Interscience: New York, 1975.
8. Chen, C.-S.; Fujimoto, Y.; Girdaukas, G.; Sih, C. J. J. Am. Chem. Soc. 1982, 104, 7294–7299.
9. Rakels, J. L. L.; Straathof, A. J. J.; Heijnen, J. J. Enzyme Microb. Technol. 1993, 15(12), 1051–1056.
10. Cornish-Bowden, A. Fundamentals of Enzyme Kinetics; Portland Press Ltd.: London, 1995.
11. (a) Gernaey, K. V.; Lantz, A. E.; Tufvesson, P.; Woodley, J. M.; Sin, G. Trends Biotechnol. 2010, 28(7), 346–354. (b) Marquardt, W. Chem. Eng. Res. Des. 2005,
83(A6), 561–573.
 Reaction Engineering of Biotransformations 99

12. Constantinides, A.; Mostoufi, N. Numerical Methods for Chemical Engineers with MATLAB Applications; Prentice Hall: Upper Saddle River, NJ, 1999.
13. Murphy, E. F.; Gilmour, S. G.; Crabbe, M. J. C. J. Biochem. Biophys. Methods 2003, 55(2), 155–178.
14. (a) Sin, G.; Woodley, J. M.; Gernaey, K. V. Biotechnol. Progr. 2009, 25(6), 1529–1538. (b) Straathof, A. J. J. J. Mol. Catal. B: Enzym. 2001, 11(4–6), 991–998.
15. Wegman, M. A.; Janssen, M. H. A.; van Rantwijk, F.; Sheldon, R. A. Adv. Synth. Catal. 2001, 343(6–7), 559–576.
16. van Rantwijk, F.; Oosterom, M. W. V.; Sheldon, R. A. J. Mol. Catal. B: Enzym. 1999, 6(6), 511–532.
17. Bell, G.; Halling, P. J.; Moore, B. D.; Partridge, J.; Rees, D. G. Trends Biotechnol. 1995, 13, 468–473.
18. Walker, A. J.; Bruce, N. C. Tetrahedron 2004, 60, 561–568.
19. Rupley, J. A.; Gratton, E.; Careri, G. Trends Biochem. Sci. 1983, 8(1), 18–22.
20. Zaks, A.; Klibanov, A. M. Science 1984, 224(4654), 1249–1251.
21. (a) Mensah, P.; Carta, G. Biotechnol. Bioeng. 1999, 66(3), 137–146. (b) Migliorini, C.; Meissner, J. P.; Mazzotti, M.; Carta, G. Biotechnol. Progr. 2000, 16, 600–609.
22. Villeneuve, P. Biotechnol. Adv. 2007, 25(6), 515–536.
23. Barreiros, S.; Couto, R.; Vidinha, P.; et al. Ind. Eng. Chem. Res. 2011, 50(4), 1938–1946.
24. Lilly, M. D. J. Chem. Technol. Biotechnol. 1982, 32, 162–169.
25. Tischer, W.; Kasche, V. Trends Biotechnol. 1999, 17(8), 326–335.
26. (a) Mateo, C.; Palomo, J. M.; Fernandez-Lorente, G.; Guisan, J. M.; Fernandez-Lafuente, R. Enzyme Microb. Technol. 2007, 40(6), 1451–1463. (b) O’Fagain, C. Enzyme
Microb. Technol. 2003, 33(2–3), 137–149.
27. Hanefeld, U.; Gardossi, L.; Magner, E. Chem. Soc. Rev. 2009, 38(2), 453–468.
28. (a) Sanchez-Ruiz, J. M. Biophys. Chem. 2010, 148(1–3), 1–15. (b) Ahern, T. J.; Klibanov, A. M. Methods Biochem. Anal. 1988, 33, 91–127.
29. Manning, M. C.; Chou, D. K.; Murphy, B. M.; Payne, R. W.; Katayama, D. S. Pharm. Res. 2010, 27(4), 544–575.
30. (a) Lumry, R.; Eyring, H. J. Phys. Chem. 1954, 58(2), 110–120. (b) Tanford, C. Protein Denaturation. In Advances in Protein Chemistry; Anfinsen, C. B. J., Edsall, J.
T., Richards, F. M., Eds.; Academic Press: New York, 1970, Vol. 24; pp 2–95. (c) Klibanov, A. M. Adv. Appl. Microbiol. 1983, 29, 1–28.
31. Schellman, J. A. Biophys. Chem. 2002, 96(2–3), 91–101.
32. Daniel, R. M.; Danson, M. J.; Eisenthal, R. Trends Biochem. Sci. 2001, 26(4), 223–225.
33. Evans, M. G.; Polanyi, M. Trans. Faraday Soc. 1935, 31, 875–894.
34. Sanchez-Ruiz, J. M. Biophys. J. 1992, 61(4), 921–935.
35. (a) Rogers, T. A.; Bommarius, A. S. Chem. Eng. Sci. 2010, 65(6), 2118–2124. (b) Shortle, D. Unfolded Proteins 2002, 62, 1–23. (c) Polizzi, K. M.; Bommarius, A. S.;
Broering, J. M.; Chaparro-Riggers, J. F. Curr. Opin. Chem. Biol. 2007, 11(2), 220–225.
36. Peterson, M. E.; Eisenthal, R.; Danson, M. J.; Spence, A.; Daniel, R. M. J. Biol. Chem. 2004, 279(20), 20717–20722.
37. Shriver, J. W.; Edmondson, S. P. Methods Mol. Biol. 2009, 490, 57–82.
38. Bechtold, M.; Panke, S. Chem. Eng. Sci. submitted
39. Bommarius, A.; Broering, J. Biocatal. Biotransform. 2005, 23(3/4), 125–139.
40. (a) Illeova, V.; Polakovic, M.; Stefuca, V.; Acai, P.; Juma, M. J. Biotechnol. 2003, 105(3), 235–243. (b) Vrabel, P.; Polakovic, M.; Stefuca, V.; Bales, V. Enzyme Microb.
Technol. 1997, 20(5), 348–354.
41. (a) Bechtold, M.; Makart, S.; Reiss, R.; Alder, P.; Panke, S. Biotechnol. Bioeng. 2007, 98(4), 812–824. (b) Boy, M.; Bohm, D.; Voss, H. Biocatal. Biotransform. 2001,
19(5–6), 413–425. (c) Boy, M.; Dominik, A.; Voss, H. Chem. Eng. Technol. 1998, 21(7), 570–575. (d) Boy, M.; Dominik, A.; Voss, H. Process Biochem. 1999,
34(6–7), 535–547. (e) Gibbs, P. R.; Uehara, C. S.; Neunert, U.; Bommarius, A. S. Biotechnol. Progr. 2005, 21(3), 762–774. (f) Rogers, T. A.; Daniel, R. M.;
Bommarius, A. S. Chemcatchem 2009, 1(1), 131–137.
42. Katchalski-Katzir, E. Trends Biotechnol. 1993, 11(11), 471–478.
43. Tosa, T.; Mori, T.; Fuse, N.; Chibata, I. Biotechnol. Bioeng. 1967, 9(4), 603–615.
44. Sheldon, R. A. Adv. Synth. Catal. 2007, 349(8–9), 1289–1307.
45. Kim, M. I.; Kim, J.; Lee, J.; et al. Biotechnol. Bioeng. 2007, 96(2), 210–218.
46. Brena, B. M.; Batista-Viera, F. Methods Biotechnol. 2006, 22, 15–30.
47. (a) Katchalski-Katzir, E.; Kraemer, D. M. J. Mol. Catal. B: Enzym. 2000, 10(1–3), 157–176. (b) Mateo, C.; Grazu, V.; Palomo, J. M.; et al. Nat. Protoc. 2007, 2(5),
1022–1033. (c) Mateo, C.; Grazu, V.; Pessela, B. C. C.; et al. Biochem. Soc. Trans. 2007, 35, 1593–1601.
48. (a) Hartmann, M.; Jung, D. J. Mater. Chem. 2010, 20(5), 844–857. (b) Hudson, S.; Cooney, J.; Magner, E. Angew. Chem. Int. Ed. 2008, 47(45), 8582–8594.
49. (a) Fraser, J. E.; Bickerstaff, G. F. Methods Biotechnol. 1997, 1, 61–65. (b) Iborra, J. L.; Manjon, A.; Canovas, M. Methods Biotechnol. 1997, 1, 53–60.
50. Lozinsky, V. I.; Galaev, I. Y.; Plieva, F. M.; et al. Trends Biotechnol. 2003, 21(10), 445–451.
51. (a) Avnir, D.; Coradin, T.; Lev, O.; Livage, J. J. Mater. Chem. 2006, 16(11), 1013–1030. (b) Betancor, L.; Luckarift, H. R. Trends Biotechnol. 2008, 26(10), 566–572.
(c) Gill, I.; Ballesteros, A. Trends Biotechnol. 2000, 18, 282–296.
52. Sheldon, R. A. Biochem. Soc. Trans. 2007, 35, 1583–1587.
53. Liese, A.; Seelbach, K.; Wandrey, C. Industrial Biotransformations; Wiley-VCH: Weinheim, Germany, 2006.
54. Häusler, H.; Stütz, A. E. Top. Curr. Chem. 2001, 215, 77–114.
55. Pedroche, J.; Yust, M. D.; Mateo, C.; et al. Enzyme Microb. Technol. 2007, 40(5), 1160–1166.
56. Fernandez-Lafuente, R. Enzyme Microb. Technol. 2009, 45(6–7), 405–418.
57. Brady, D.; Jordaan, J. Biotechnol. Lett. 2009, 31(11), 1639–1650.
58. Reetz, M. T. Adv. Mater. 1997, 9(12), 943–954.
59. Sheldon, R. A.; Schoevaart, R.; Van Langen, L. M. Biocatal. Biotransform. 2005, 23(3–4), 141–147.
60. Jung, D.; Paradiso, M.; Hartmann, M. J. Mater. Sci. 2009, 44(24), 6747–6753.
61. Theil, F. Tetrahedron 2000, 56(19), 2905–2919.
62. Sheldon, R. A.; Lopez-Serrano, P.; Cao, L.; van Rantwijk, F. Biotechnol. Lett. 2002, 24(16), 1379–1383.
63. Wintrode, P. L.; Arnold, P. H. Adv. Protein Chem. 2001, 55(55), 161–225.
64. (a) Eijsink, V. G. H.; Bjork, A.; Gaseidnes, S.; et al. J. Biotechnol. 2004, 113(1–3), 105–120. (b) Eijsink, V. G. H.; Gaseidnes, S.; Borchert, T. V.; van den Burg, B.
Biomol. Eng. 2005, 22(1–3), 21–30.
65. Cherry, J. R.; Lamsa, M. H.; Schneider, P.; et al. Nat. Biotechnol. 1999, 17(4), 379–384.
66. Ehren, J.; Govindarajan, S.; Moron, B.; Minshull, J.; Khosla, C. Protein Eng. Des. Sel. 2008, 21(12), 699–707.
67. (a) Lehmann, M.; Pasamontes, L.; Lassen, S. F.; Wyss, M. Biochim. Biophys. Acta-Protein Struct. Mol. Enzymol. 2000, 1543(2), 408–415. (b) Lehmann, M.; Wyss, M.
Curr. Opin. Biotechnol. 2001, 12(4), 371–375. (c) Amin, N.; Liu, A. D.; Ramer, S.; et al. Protein Eng. Des. Sel. 2004, 17(11), 787–793. (d) Bae, E.; Bannen, R. M.;
Phillips, G. N. Proc. Natl. Acad. Sci. USA 2008, 105(28), 9594–9597.
68. (a) Reetz, M. T.; D Carballeira, J.; Vogel, A. Angew. Chem. Int. Ed. 2006, 45(46), 7745–7751. (b) Kim, S. J.; Lee, J. A.; Joo, J. C.; et al. Biotechnol. Progr. 2010,
26(4), 1038–1046.
69. (a) Gaseidnes, S.; Synstad, B.; Jia, X. H.; et al. Protein Eng. 2003, 16(11), 841–846. (b) Chakiath, C.; Lyons, M. J.; Kozak, R. E.; Laufer, C. S. Appl. Environ.
Microbiol. 2009, 75(23), 7343–7349.
100 Reaction Engineering of Biotransformations

70. (a) Emond, S.; Andre, I.; Jaziri, K.; et al. Protein Sci. 2008, 17(6), 967–976. (b) Emond, S.; Potocki-Veronese, G.; Mondon, P.; et al. J. Biomol. Screening 2007,
12(5), 715–723.
71. Korkegian, A.; Black, M. E.; Baker, D.; Stoddard, B. L. Science 2005, 308(5723), 857–860.
72. Williams, J. C.; Zeelen, J. P.; Neubauer, G.; et al. Protein Eng. 1999, 12(3), 243–250.
73. Savile, C. K.; Janey, J. M.; Mundorff, E. C.; et al. Science 2010, 329(5989), 305–309.
74. Van den Burg, B.; Vriend, G.; Veltman, O. R.; Eijsink, V. G. H. Proc. Natl. Acad. Sci. USA 1998, 95(5), 2056–2060.
75. (a) Declerck, N.; Machius, M.; Joyet, P.; et al. Protein Eng. 2003, 16(4), 287–293. (b) Machius, M.; Declerck, N.; Huber, R.; Wiegand, G. J. Biol. Chem. 2003,
278(13), 11546–11553.
76. Palackal, N.; Brennan, Y.; Callen, W. N.; et al. Protein Sci. 2004, 13(2), 494–503.
77. (a) Rao, N. N.; Lutz, S.; Wurges, K.; Minor, D. Org. Process Res. Dev. 2009, 13(3), 607–616. (b) Villadsen, J. Chem. Eng. Sci. 2007, 62(24), 6957–6968.
78. Berendsen, W. R.; Lapin, A.; Reuss, M. Chem. Eng. Sci. 2007, 62(9), 2375–2385.
79. Mazzei, R.; Drioli, E.; Giorno, L. Biocatalytic Membranes and Membrane Bioreactors. In Comprehensive Membrane Science and Engineering; Drioli, E., Giorno, L., Eds.;
Elsevier: Amsterdam, 2010, pp 195–211.
80. (a) Biselli, M.; Kragl, U.; Wandrey, C. Reaction Engineering for Enzyme-Catalyzed Biotransformations. In Enzyme Catalysis in Organic Synthesis; Drauz, K., Waldmann,
H., Eds.; Wiley-VCH: Weinheim, 2002, pp 185–257. (b) Woltinger, J.; Karau, A.; Leuchtenberger, W.; Drauz, K. Membrane Reactors at Degussa. In Advances in
Biochemical Engineering/Biotechnology, 2005, Vol. 92, pp 289–316.
81. van Reis, R.; Zydney, A. J. Membr. Sci. 2007, 297(1–2), 16–50.
82. Susantu, H.; Ulbricht, M. Polymeric Membranes for Molecular Separations. In Membrane Operations; Drioli, E., Giorno, L., Eds.; Wiley-VCH: Weinheim, 2009,
pp 19–43.
83. Hilterhaus, L.; Liese, A. Applications of reaction engineering to industrial biotransformations. In Biocatalysis for the Pharmaceutical Industry: Discovery, Development and
Manufacturing; Tao, J., Lin, G. Q., Eds.; John Wiley: Singapore, 2009, pp 65–88.
84. Vaidya, A. M.; Halling, P. J.; Bell, G. Biotechnol. Tech. 1993, 7(7), 441–446.
85. (a) Bechtold, M.; Panke, S. Chimia 2009, 63(6), 345. (b) de Roode, B. M.; Franssen, A. C. R.; van der Padt, A.; Boom, R. M. Biotechnol. Progr. 2003, 19(5),
1391–1402.
86. Charpentier, J. C. Chem. Eng. Res. Des. 2010, 88(3A), 248–254.
87. Heinemann, M.; Panke, S. Bioinformatics 2006, 22(22), 2790–2799.
88. Ro, D. K.; Paradise, E. M.; Ouellet, M.; et al. Nature 2006, 440, 940–943.
89. Santacoloma, P. A.; Sin, G.; Gernaey, K. V.; Woodley, J. M. Org. Process Res. Dev. 2011, 15, 203–212.
90. (a) Jewett, M. C.; Calhoun, K. A.; Voloshin, A.; Wuu, J. J.; Swartz, J. R. Mol. Syst. Biol. 2008, 4, 220. (b) Zhang, Y.-H. P.; Evans, B. E.; Mielenz, J. R.; Hopkins, R. C.;
Adams, M. W. W. PLOS One 2007, 2007, e456. (c) Bujara, M.; Schümperli, M.; Billerbeck, S.; Heinemann, M.; Panke, S. Biotechnol. Bioeng. 2010, 106, 376–389.
91. Endy, D. Nature 2005, 438(7067), 449–453.
92. Marchisio, M. A.; Stelling, J. Bioinformatics 2008, 24(17), 1903–1910.
93. Bujara, M.; Schümperli, M.; Pellaux, R.; Heinemann, M.; Panke, S. Nat. Chem. Biol. 2011, 7, 271–277.
94. Weuster-Botz, D.; Gebhardt, G.; Hortsch, R.; Kaufmann, K.; Arnold, M. Biotechnol. Progr. 2011, 27(3), 684–690.
95. Huebner, A.; Olguin, L. F.; Bratton, D.; et al. Anal. Chem. 2008, 80, 3890–3896.
96. (a) Fernandes, P. Int. J. Mol. Sci. 2010, 11(3), 858–879. (b) Buchs, J.; Huber, R. H. R.; Wulfhorst, H.; et al. Biotechnol. Progr. 2011, 27(2), 555–561.
97. Walser, M.; Leibundgut, R. M.; Pellaux, R.; Panke, S.; Held, M. Cytometry Part A 2008, 73A(9), 788–798.
98. Miyazaki, M.; Honda, T.; Yamaguchi, H.; Briones, M. P. P.; Maeda, H. Biotechnol. Genet. Eng. Rev. 2008, 25, 405–428.
99. Woodley, J. M.; Kim, P. Y.; Pollard, D. J. Biotechnol. Progr. 2007, 23(1), 74–82.
100. Fessner, W. D. Aldolases: Enzymes for Making and Breacking C–C Bonds. In Asymmetric Organic Synthesis with Enzymes; Gotor, V., Alfonso, I., Garcia-Urdiales, E.,
Eds.; Wiley-VCH: Weinheim, 2008, pp 275–318.
101. Goldberg, R. N.; Tewari, Y. B.; Bhat, T. N. Bioinformatics 2004, 20(16), 2874–2877.
102. Wagner, N.; Füreder, M.; Bosshart, A.; Panke, S.; Bechtold, M. Org. Process Res. Dev. accepted doi: 10.1021/op200160e.
103. Rajendran, A.; Paredes, G.; Mazzotti, M. J. Chromatogr. A 2009, 1216(4), 709–738.
104. Coquerel, G. Novel Opt. Resolut. Technol. 2007, 269, 1–51.
105. (a) Bechtold, M.; Makart, S.; Heinemann, M.; Panke, S. J. Biotechnol. 2006, 124, 146–162. (b) Palacios, J. G.; Kaspereit, M.; Kienle, A. Chem. Eng. Technol. 2011,
34(5), 688–698.
106. Straathof, A. J. J.; Saric, M.; van der Wielen, L. A. M. Chem. Eng. Sci. 2011, 66(3), 510–518.