Enhancing the thermal robustness of an enzyme by directed evolution: Least favorable starting points and inferior mutants can map superior evolutionary Pathways

Article abstract of DOI:10.1002/cbic.201100412

In a previous directed evolution study, the B-FIT approach to increasing the thermal robustness of proteins was introduced and applied to the lipase from Bacillus subtilis. It is based on the general concept of iterative saturation mutagenesis (ISM), according to which sites in an enzyme are subjected to saturation mutagenesis, the best hit of a given library is then used as a template for randomization at other sites, and the process is continued until the desired catalyst improvement has been achieved. The appropriate choice of the ISM sites is crucial; in the B-FIT method the criterion is residues characterized by highest B factors available from X-ray crystallography data. In the present study, B-FIT was employed in order to increase the thermal robustness of the epoxide hydrolase from Aspergillus niger. Several rounds of ISM resulted in the best variant showing a 21 °C increase in the T ⁶⁰₅₀ value, an 80-fold improvement in half-life at 60 °C, and a 44 kcalmol^-1 improvement in inactivation energy. Seven other variants were also evolved with moderate yet significant improvements; these were characterized by 10-14°C increases in T ⁶⁰₅₀, 20-30-fold improvement in half-lives at 60°C and 15-20 kcalmol^-1 elevations in activation energy. Unexpectedly, in the ISM process the best variants were obtained from essentially neutral or even inferior mutant parents, that is, when a given library contains no improved mutants. This constitutes a practical way to escape from what appear to be local minima (" dead ends" ) in the fitness landscape- a finding of notable significance in directed evolution.

Full text of DOI:10.1002/cbic.201100412

DOI: 10.1002/cbic.201100412
Enhancing the Thermal Robustness of an Enzyme by
Directed Evolution: Least Favorable Starting Points and
Inferior Mutants Can Map Superior Evolutionary Pathways
[a]
[a, b]
Yosephine Gumulya and Manfred T. Reetz*
In a previous directed evolution study, the B-FIT approach to
increasing the thermal robustness of proteins was introduced
and applied to the lipase from Bacillus subtilis. It is based on
the general concept of iterative saturation mutagenesis (ISM),
according to which sites in an enzyme are subjected to satura-
tion mutagenesis, the best hit of a given library is then used as
a template for randomization at other sites, and the process is
continued until the desired catalyst improvement has been
achieved. The appropriate choice of the ISM sites is crucial; in
the B-FIT method the criterion is residues characterized by
highest B factors available from X-ray crystallography data. In
the present study, B-FIT was employed in order to increase the
thermal robustness of the epoxide hydrolase from Aspergillus
niger. Several rounds of ISM resulted in the best variant show-
60
ing a 218C increase in the T50 value, an 80-fold improvement
ꢀ
1
in half-life at 608C, and a 44 kcalmol improvement in inacti-
vation energy. Seven other variants were also evolved with
moderate yet significant improvements; these were character-
60
ized by 10–148C increases in T , 20–30-fold improvement in
50
ꢀ
1
half-lives at 608C and 15–20 kcalmol elevations in activation
energy. Unexpectedly, in the ISM process the best variants
were obtained from essentially neutral or even inferior mutant
parents, that is, when a given library contains no improved
mutants. This constitutes a practical way to escape from what
appear to be local minima (“dead ends”) in the fitness land-
scape—a finding of notable significance in directed evolution.
Introduction
Sufficiently high thermostability of enzymes is a prerequisite
ble and the number of chosen sites, N, determine the number
of pathways (N!). When applying ISM, it is not necessary to ex-
plore all pathways, although this can be of theoretical interest.
The reason for the pronounced efficacy of ISM, which has
been observed in numerous cases involving different enzymes,
is due to strong cooperative effects occurring between muta-
tions within a randomized site and between two or more sites
[1]
for their application in organic chemistry and biotechnology.
For this reason protein engineering techniques utilizing site-
[
2]
specific mutagenesis based on rational design or directed
[
3]
evolution by using error-prone polymerase chain reaction
epPCR), DNA shuffling and/or other mutagenesis methods
have been applied successfully in numerous cases. Increases in
(
6
50
0
[9]
melting temperature (T ) or T value (the temperature at
as the evolutionary process proceeds (more than additivity).
m
which half of the activity is maintained after a 60 min heat
The choice of the randomization sites is crucial, which in
turn depends on the protein property to be evolved. For each
catalytic property (e.g., thermostability, rate, stereoselectivity,
etc.) a different criterion for choosing appropriate randomiza-
tion sites needs to be considered. In the case of thermostabili-
zation, thermoresistance to undesired aggregation, or robust-
ness towards hostile organic solvents, we have developed the
[
4]
treatment) in the range of 5–158C proved to be typical in
[
2–5]
60
this important endeavor.
It should be noted that T50 values
are generally used to describe thermostability, but this might
in fact include thermoresistance to aggregation accompanied
by undesired precipitation (in extreme cases only this factor).
In order to enhance the efficacy of directed evolution in
general by producing smaller yet higher-quality mutant librar-
[10]
B-FIT method with the following protocol. Atomic displace-
ment parameters obtained from X-ray crystallography data are
utilized as a guide, namely B factors, which reflect smearing of
atomic electron densities with respect to their equilibrium po-
[
6]
ies requiring less screening effort (the bottleneck of laborato-
[
7]
ry evolution), we have developed the concept of iterative sat-
[
8,9]
uration mutagenesis (ISM).
It is a straightforward process
that begins by identifying sites in an enzyme appropriate for
saturation mutagenesis (amino acid randomization), a site
being defined as comprising one, two or more amino acid po-
sitions in the protein. Following the formation of initial mutant
libraries and a screening procedure specific for a given catalyt-
ic property, the best variants (“hits”) are identified. The gene of
a respective hit is then utilized as a template for further satura-
tion mutagenesis at the other sites, and the iterative process is
continued until the desired degree of catalyst improvement
has been achieved. Different evolutionary pathways are possi-
[
a] Dr. Y. Gumulya, Prof. Dr. M. T. Reetz
Department of Synthetic Organic Chemistry
Max-Planck-Institut fꢀr Kohlenforschung
Kaiser-Wilhelm-Platz 1, 45470 Mꢀlheim an der Ruhr (Germany)
E-mail: reetz@mpi-muelheim.mpg.de
[
b] Prof. Dr. M. T. Reetz
Fachbereich Chemie, Philipps-Universitꢁt Marburg
Hans-Meerwein-Strasse, 35032 Marburg (Germany)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201100412.
2
502
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 2502 – 2510

Enhancing the Thermal Robustness of an Enzyme
[
14]
sitions as a result of thermal motion and positional disorder.
Rather than considering only the B factor of the Ca atom of a
given amino acid in the protein, all atoms of a particular resi-
due are utilized (except the H atoms). The B-FITTER computer
aid, which is available free of charge on the homepage of the
identifying active hits, Reymond’s adrenaline-based assay
15
was used, followed by measurement of T50 values, the test re-
action being the hydrolysis of rac-1. The improved variants
were then isolated, purified and re-examined by subjecting
6
0
them to heat treatment for 1 h, resulting in more precise T5
0
corresponding
author
(http://www.kofo.mpg.de/manfred-
values. For this purpose medium-throughput HPLC analysis
was utilized in order to monitor the hydrolysis of rac-1. On the
practical side, we succeeded in evolving a significant increase
reetz.html), then provides the average B values of all residues.
Those with high average B values indicate the highest flexibili-
ty; such data then serve as the criterion for choosing appropri-
60
in robustness, the DT50 value of the best ANEH mutant
amounted to 218C.
[
10]
ate sites for saturation mutagenesis. The expectation is that
proper mutations thereat will increase rigidity, and therefore,
enhance thermostability or thermal robustness in the more
general sense. Since these residues generally occur on the pro-
tein surface, other influences, such as solvent effects and/or
protein–protein interactions, might also play a role, among
others. In an initial study we showed the success of B-FIT by
subjecting the lipase from Bacillus subtilis, which is a small 181
Results and Discussion
The X-ray structure of wild-type (WT) ANEH (PDB ID: 1Q07), a
396 residue enzyme, has been determined at 1.8 ꢁ resolution,
[15]
and has revealed a typical a/b-hydrolase fold. It is a dimer
composed of two 44 kDa subunits in the asymmetric unit. The
data provide a means to identify residues with the highest
average B factors as calculated by the B-FITTER computer
[
10]
amino acid protein, to this method. Preliminary biophysical
results show that the apparent increase in robustness is due to
the suppression of undesired aggregation, which causes detri-
[10b]
aid.
The top 50 were ranked as summarized in Table S1 in
[
11]
mental precipitation of the enzyme upon heating, an unusu-
al effect, which Rao et al. had previously discovered in their ap-
proach to stabilization of the same lipase using epPCR as the
the Supporting Information. Not all of them were considered
in an ISM scheme, because such extensive exploration of the
respective protein sequence space was not the goal of this
study. From these 50 residues, ten sites each composed of two
amino acid positions, which are scattered over the surface of
the protein, were initially considered: (Glu26/Gln27), (Ser70/
Glu71), (Arg105/Glu138), (Ala217/Cys350), (Met218/Leu330),
(Ala220/Ser226), (Glu223/Gly224), (Ser229/Leu230), (Met329/
Lys332) and (Asp368/Glu371). The distance between two resi-
dues within a given site is substantially less than 10 ꢁ, some of
them are direct neighbors. Cooperative effects acting between
two such residues can be expected as the evolutionary process
of B-FIT proceeds, in addition to cooperativity between the
[
12]
gene mutagenesis method. Recently, Visser and co-workers
have applied B-FIT successfully in the quest to enhance the
[
13a]
thermostability of E. coli uridine phosphorylase,
and in the
case of a lipase the Bornscheuer group has likewise reported
[
13b]
positive results using this method.
In these studies exten-
sive exploration of the respective ISM scheme was not
necessary.
Here, we report the result of applying B-FIT to the epoxide
hydrolase from Aspergillus niger (ANEH). In a previous study
this enzyme was subjected to ISM in the form of the combina-
torial active site-saturation test (CAST) as part of an effort to in-
crease the enantioselectivity of the test reaction involving the
hydrolytic kinetic resolution of racemic phenyl glycidyl ether,
the selectivity factor increased from E=4.6 (WT ANEH) to E=
[8b,9,10]
sites themselves.
It should be noted that a different
grouping of the 20 residues would also be possible.
Two additional sites in an unresolved region in the crystal
structure of WT ANEH were also chosen, for which no B factors
are accessible, but which can also be expected to constitute
flexible regions: (Ala321/Ser322) and (Gly326/Ala327). The con-
ventional form of saturation mutagenesis as defined by the
[
8a]
115 (best mutant in favor of (S)-2).
[
16a]
QuikChange protocol,
which is based on earlier studies re-
[16b–f]
garding saturation mutagenesis,
utilizes NNK codon de-
generacy (N: adenine/cytosine/guanine/thymine; K: guanine/
thymine) encoding all 20 canonical amino acids. In contrast,
we discovered in previous studies that reduced amino acid al-
phabets lead to high-quality libraries, which entail drastically
The purpose of the present study was not only to obtain no-
tably more robust ANEH mutants for practical purposes, an im-
portant goal was also methodology development for more effi-
cient directed evolution. This entailed the exploration of sever-
al evolutionary pathways in the ISM scheme—not just one—as
well as the development of a strategy for escaping from what
can be termed, from a practical point of view, as a “local mini-
mum” or “dead end”; that is, what to do if a given mutant li-
brary contains no improved variants (hits). In order to measure
thermal robustness of the generated mutants in a high-
[8b,9,10b]
lower screening efforts.
For example, for 95% library
coverage, randomization of a two-residue site by using NNK re-
quires the screening of about 3000 transformants (assuming
the absence of amino acid bias), while the use of, for example,
NDT codon degeneracy (D: adenine/guanine/thymine; T: thy-
mine) encoding only 12 amino acids (Phe, Leu, Ile, Val, Tyr, His,
Asn, Asp, Cys, Arg, Ser, Gly) needs only 430 transfor-
[8b,9,10b]
mants.
Due to the expected amino acid bias present in
15
throughput manner, we employed T50 values, the temperature
required to reduce the initial enzymatic activity at room tem-
perature following a 15 min heat treatment. As a pretest for
most mutagenesis methods in directed evolution, including
saturation mutagenesis, such calculations constitute rough but
useful guides (see also the Experimental Section).
ChemBioChem 2011, 12, 2502 – 2510
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
2503

M. T. Reetz and Y. Gumulya
Therefore, in the present study we generally chose NDT
codon degeneracy, the encoded 12 amino acids forming a
well-balanced set of polar/nonpolar, aromatic/nonaromatic and
charged/uncharged amino acids as building blocks. In one
case NNG codon degeneracy (G: guanine) was also tested; this
encodes the eight amino acids that are not covered by NDT,
namely Ala, Pro, Trp, Thr, Met, Gln, Glu and Lys. In the initial
round of saturation mutagenesis at all 12 sites by using NDT
codon degeneracy, the most positive results originated from
six sites, namely A, B, C, D, E and F (Figure 1; the letter desig-
nations A–F are arbitrary). A quality check of the libraries was
performed, and revealed, not surprisingly, amino acid bias in
some cases (Table S2 in the Supporting Information). Neverthe-
In the first round of saturation mutagenesis at 12 sites with
the creation of 12 mutant libraries, a total of 5495 transfor-
mants were assayed, which in each case guaranteed 95% li-
brary coverage (again, due to the bias phenomenon, this is
only a rough guide). Table 1 lists the respective hits. The best
mutants, GUY-003 and GUY-004, both showing the greatest im-
60
provement (DT50 =88C), originate from the library obtained by
randomization at site B, which is in an unresolved part of the
X-ray structure. Since B-FIT targets residues with the highest
degree of flexibility as measured by the respective B factor, it is
not surprising that most of the randomization sites are located
on the surface of the protein, generally in loops. Sites D and E
are located between the end of a helix and the beginning of a
loop, respectively. We also note that another excellent mutant,
60
less, the evolved increases in DT50 values proved to be sub-
stantial, despite the possibility that even better hits were per-
haps missed. The other six libraries contained mutants display-
60
GUY-005, with an increase in T50 of 78C, originates from site C,
which has the highest average B factor. Thus, the hypothesis
that mutagenesis should be focused on positions of highest
flexibility constitutes a practical guide.
15
ing essentially no improvements in the T50 screen, and they
were not utilized in further mutagenesis experiments. The re-
spective sites were likewise not considered in iterative satura-
tion mutagenesis (although past experience has shown that
In the case of saturation mutagenesis at site D, no improved
mutants were detected, although several showed thermal ro-
bustness comparable to that of WT ANEH. This could have sev-
eral reasons; for example, all combinations of residues as en-
coded by NDT fail to provide superior mutants. Therefore, NNG
codon degeneracy, encoding the eight amino acids not includ-
ed in NDT, was tested. In this case, screening of 765 transfor-
mants, which ensured more than 95% coverage, led to the
[
9]
this might in fact be successful). Such decisions need not
cause unease, because past and present experience shows
that a given ISM study can lead to many different hits, depend-
ing on how much time is invested in exploring different sites
and pathways. Since the libraries are small, requiring a mini-
mum of screening, such mini-explorations of protein sequence
[
9]
15
space generally prove to be rewarding.
identification of mutant GUY-007 with a slightly improved T
5
0
*
60
value of 1 C, but upon isolation, the same T50 value as WT
ANEH was found (Table 1). NNK codon degeneracy was not
tested in this case.
In protein engineering of thermostabilization, an increase in
60
[2–5]
T5 of 5–88C is generally considered to be respectable.
Nev-
0
ertheless, we wanted to continue optimization by testing ISM
using some, but for practical reasons, not all of the six sites A–
F. In curiosity-driven experiments, we chose two extreme mu-
tants as points for continuation, the best one and the worst
one: GUY-003 originating from randomization at site B showing
an improvement of 88C (same as mutant GUY004, which could
also have been chosen), and mutant GUY-007 (site D) lacking
any improvement. Two different strategies were followed. In
Table 1. Best mutants resulting from the first round of saturation muta-
genesis by using NDT codon degeneracy in the randomization at sites A,
B, C, E and F, and NNG codon degeneracy in the case of site D.
1
0
5
15
60
60
Site
Mutant
Mutations
T5
DT50
T50
DT50
[
8C]
[8C]
[8C]
[8C]
WT ANEH
GUY-001
GUY-002
GUY-003
GUY-004
GUY-005
GUY-006
GUY-007
GUY-008
GUY-009
GUY-010
–
50
56
56
58
58
58
56
51
53
55
54
–
6
6
8
8
8
6
1
3
5
4
46
51
51
54
54
52
50
46
50
51
50
–
6
6
8
8
7
5
0
4
6
5
A
B
C
A321H/S322C
S321C/S322G
G326F/A327Y
G326Y/A327Y
A220C
A220V/S226H
S229E
E26N/Q27N
A217V/C350F
A217F
Figure 1. Location of saturation mutagenesis sites A–F in the X-ray structure
[
15]
of WT ANEH by using NDT codon degeneracy, leading to improved mu-
tants with enhanced thermostability. Sites displaying high average B factors
are, C (residues 220/226), D (residues 229/230), E (residues 26/27), and F (res-
idues 217/350). Sites A (residues 321/322) and B (residues 326/327) are in an
unresolved region of the crystal structure. The catalytically active residues,
Tyr251 and Tyr314, bind and activate the epoxide substrate by forming
H bonds to the epoxy O atom, while Asp192 acts as the nucleophile in the
rate-determining step.
D
E
F
2
504
www.chembiochem.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 2502 – 2510

Enhancing the Thermal Robustness of an Enzyme
the case of the mutant derived
from site B as the template for
further saturation mutagenesis,
the conventional procedure was
adhered to by considering only
improved mutants, that is, “hits”
in the traditional meaning. If no
such mutant was found in a
given library, the respective
pathway was viewed as ending
there and was no longer consid-
ered. In the case of the mutant
identified by saturation muta-
genesis at site D, improved mu-
tants were again chosen for a
subsequent mutagenesis round,
but if none was identified, then
an unimproved or inferior var-
iant was chosen as a starting
point for the next ISM round.
1
5
Figure 2. T50 values of variants found in the first, second and third rounds of ISM. Light gray bars: mutants of the
Complete systematization was
first generation; dark gray bars: mutants of the second generation; black bars: mutants of the third generation.
maintained in the build-up
phase of the two options by
testing five branches in each
and/or to Eigen’s concept of quasi-species in natural evolu-
[18]
case, namely steps WT!B!A, WT!B!C, WT!B!D, WT!
B!E and WT!B!F on the one hand, and WT!D!A, WT!
D!B, WT!D!C, WT!D!E and WT!D!F on the other
hand. Thereafter, only a few pathways were continued, the ad-
ditional total screening effort amounting to 4300 trans-
formants.
tion, which Mannervik has invoked previously in several im-
[19]
portant directed evolution studies.
Likewise, our previous
study regarding the enhancement of activity and enantioselec-
tivity of an enoate reductase, in which it was discovered that it
is not always optimal to pick the very best variant for the next
ISM round, becomes relevant, specifically when two different
catalytic parameters are being optimized (rate and
This limited exploration led to some remarkable results as
summarized in Figure 2 (Table S3 in the Supporting Informa-
tion). At this point we note that many, but not all, pathways
provide improved variants. Surprisingly, positive results also
originate in the pathway that begins with the least thermosta-
bilized mutant originally obtained in the library formed by ran-
domization at site D. While the step from WT to D indicates a
very slight improvement (Figure 2), isolation of the mutant and
[20]
stereoselectivity).
Inspired by these results, we decided to explore additional
pathways defined by mutants obtained by saturation muta-
genesis at sites B and D, this time following the second strat-
egy in both cases, that is, just continuing with an inferior
mutant if a given library failed to contain hits in the conven-
tional sense. Again, unusual observations were made. As al-
ready noted, pathway WT!B!C!A provides a mutant with
60
characterization by T50 shows that no improvement was in fact
achieved (Table 1). Nevertheless, pathways WT!D!A, WT!
D!B, WT!D!C and WT!D!F provided improved mutants.
Interestingly, upon exploring pathway WT!D!E, protein ro-
bustness (or “thermoresistance” as a referee has suggested)
failed to be improved in the early stages of the evolutionary
pathway. This means that the respective epistatic effects are
either antagonistic or non-additive. Normally, one would aban-
don such a pathway, that is, as soon as a library lacking im-
proved mutants is encountered. It is a “dead end” or “local
minimum” as judged from a practical point of view, whatever
the reason(s) might be. These could include insufficient screen-
ing, inherent amino acid bias, imperfect PCR conditions or an
actual local minimum on the theoretical fitness landscape.
Contrary to conventional wisdom in the field of directed evolu-
60
a respectable T5 value of 598C, but further ISM failed to im-
0
prove thermal robustness (Figure 3 and the Supporting Infor-
mation). In the case of WT!B!F!A!C!E!D, the situation
is different, because it was possible to escape from two local
minima simply by continuing with non-improved mutants, and
6
0
0
the final randomization at site D led to a better mutant (T5
=
608C).
Importantly, when starting from the least improved original
mutant GUY-007 (site D), pathway WT!D!E!C failed to ach-
ieve any positive results as already pointed out, but continuing
on this seemingly hopeless pathway according to WT!D!
E!C!F!A!B, proved to be highly successful, and provided
the best mutant of the present study. An increase in thermal
60
robustness amounting to DT50 =218C was achieved, which is a
high score in directed evolution of thermally robust proteins.
An ISM scheme based on six sites, A, B, C, D, E, and F, as in
the present study, has so many different pathways (720) that
exploring all of them, requires excessive experimental work. In
[
3,6,9]
tion,
application of the second strategy based on the use
of non-improved or even inferior mutants in ISM led to superi-
or mutants in the third generation (Figure 2). This interesting
[17]
phenomenon might be related to the theory of neutral drift
ChemBioChem 2011, 12, 2502 – 2510
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
2505

M. T. Reetz and Y. Gumulya
the formal estimation of half-lives and the determination of in-
activation energies (E ). These data were collected for WT and
a
two mutants, GUY-060 and GUY-066. The improvements are
clearly evident (Table 3). For example, the thermal E improves
a
ꢀ
1
ꢀ1
from 39.7 kcalmol (WT) to 71 kcalmol (best mutant GUY-
66). Further data corroborating the increase in thermal ro-
0
bustness are collected in the Supporting Information, namely
specific activity of WT ANEH and of mutants GUY-060 and GUY-
066 at different temperatures (Figure S1 in the Supporting In-
formation) as well as residual activity (%) of these two mutants
as a function of temperature (Figure S2 in the Supporting
Information).
In the absence of detailed biophysical studies, it is difficult
to interpret the results on a molecular level. One of several
possibilities, although not proven by current data, concerns
the introduction of cysteine residues in all of the improved
mutants, these perhaps lead to stabilizing disulfide bridges.
Other strategies for improving thermostability
Numerous studies of structural comparisons between meso-
philic and thermophilic proteins have revealed the importance
of salt bridges, sometimes in networks, in promoting thermo-
[3–5,21]
stability.
Therefore, in addition to directed evolution
based on iterative B-FIT libraries, as described above, we creat-
ed several libraries for introducing salt bridges or removing un-
favorable electrostatic interactions. The residues chosen for li-
brary creation were Arg36, Ala41, Gln49, Phe54, Ser58, Leu61,
Asn82, Glu104, Gln129, Gln175, and Glu292. The codon degen-
eracy used for these libraries was RRK, which encodes for
acidic (Asp, Glu) and basic (Arg, Lys) amino acids. However, no
Figure 3. Results of limited ISM exploration starting from the best mutant,
GUY-003 (site B), and the worst mutant, GUY-007, in the initial round of satu-
ration mutagenesis at sites A–F. In all cases NDT codon degeneracy was
used except when performing saturation mutagenesis at site D, in which
case NNG codon degeneracy was applied.
[22]
the present study only four complete upward pathways were
studied, namely WT!B!F!A!C!E!D, WT!B!F!A!
C!D!E, WT!B!C!A!F!E!D and WT!D!E!C!F!
A!B (Figure 3), the final mutants were GUY-055, GUY-057,
improved mutants were found in these libraries (data not
shown).
The N and C termini of protein chains are normally known
to adopt flexible conformations, and some examples in the lit-
erature show that shortening the termini and/or anchoring
60
GUY-060, and GUY-066, and showed T5 values of 60, 59, 58
0
and 658C, respectively. The sequences and thermostability
data are summarized in Table 2. It can be seen that all four
pathways lead to notably improved mutants as measured by
a
Table 3. Estimated half-life (t1/2) and thermal inactivation energy (E ) of
WT ANEH and mutants at different temperatures.
60
DT5 values ranging between 14 and 218C. It is also evident
0
that the point mutations are quite different in nature, which
means that the effects on a molecular level must also be
diverse.
Mutant
t
1/2 [min]
658C
E
a
ꢀ
1
608C
628C
688C
708C
[kcalmol ]
WT
GUY-060
GUY-066
ꢁ1
24
80
ꢁ1
21
77
–
7
16
–
3
6
–
2
4
39.7
62.9
71.0
Analysis of thermal robustness of WT ANEH and the four se-
lected mutants at temperatures 55, 60, 65 and 708C enable
Table 2. Characterization of the final isolated and purified ANEH mutants arising from the four fully transversed pathways.
6
0
0
Pathway
Mutant
Mutations
A220C
DT5 [8C]
WT!B!F!A!C!D!E
GUY-055
A321G
S322C
A321G
S322C
A321N
S322C
A321L
S322C
G326F
A327Y
G326F
A327Y
G326F
A327Y
G326C
A327G
S229M
S229E
S229E
S229E
E26L
Q27R
E26N
Q27H
E26G
Q27F
E26Y
Q27L
A217V
16
15
14
21
S226N
A220C
S226N
A220C
S226L
A220C
S226G
C350F
A217V
C350F
A217V
C350F
A217C
WT!B!F!A!C!E!D
WT!B!C!A!F!E!D
WT!D!E!C!F!A!B
GUY-057
GUY-060
GUY-066
2
506
www.chembiochem.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 2502 – 2510

Enhancing the Thermal Robustness of an Enzyme
[
23]
them increases thermostability. We, therefore, performed sat-
uration mutagenesis at residues Lys3 and Glu138 (N termini) as
well as Asp107, Trp396, Val109 and Val392 (C termini) using
NDT codon degeneracy. However, again positive results were
proved mutants. Thus, such a strategy constitutes the simplest
technique for escaping from a “local minimum”, whatever the
reasons for the occurrence of such a “dead-end” might be. It
opens up new perspectives for directed evolution based on
ISM, but also for protein engineering in general by using other
techniques, such as error-prone PCR or DNA shuffling. Indeed,
in preliminary work we have observed that this strategy is also
successful when optimizing the stereoselectivity of enzymes.
These unusual observations could be related to the phenom-
[
22]
not achieved (data not shown).
Conclusion
Iterative saturation mutagenesis (ISM) has emerged as a pow-
erful method in protein engineering, and makes directed evo-
[17]
enon of neutral drift and/or to Eigen’s hypothesis of quasi-
[9]
[18]
lution more efficient, and therefore, faster than in the past.
We, and others, have applied this concept to a number of dif-
species in natural evolution, as recently invoked by Manner-
vik in a series of seminal papers regarding the directed evolu-
[
9,24]
[19]
ferent proteins.
Efficacy is increased even more when using
tion of other enzymes (glutathione transferases).
[
8b,9]
reduced amino alphabets.
Smaller, and at the same time,
While several arbitrarily chosen pathways as studied in the
present investigation proved to be successful, more systematic
work is necessary in order to fully illuminate the nature of ISM
as a general method in directed evolution. The exploration of
a complete ISM scheme would shed additional light on this im-
portant issue, be it for increasing thermostabilization, substrate
acceptance or stereoselectivity. Nevertheless, for practical pur-
poses when attempting to solve a specific biocatalyst problem,
this is not necessary. We also note that many of the mutations
in the evolved variants described herein differ considerably in
nature, which means that future theoretical and biophysical
work needs to focus on the source of increased thermal ro-
bustness of the ANEH mutants on a molecular level. The effect
of amino acid bias and of imperfect PCR conditions in general
as well as the question of how residues are optimally grouped
into sites when performing ISM, likewise constitute current
issues of interest in our laboratory. Success is likely even if
such parameters are ignored, but optimization can be expect-
ed to increase efficacy even more.
higher-quality libraries result from these techniques, thereby
drastically reducing the screening effort, which is traditionally
the bottleneck of directed evolution. So far, the two embodi-
[10]
ments of ISM are B-FIT for enhancing thermal robustness,
[
8,9]
and CASTing
for manipulating substrate scope (rate) and
stereoselectivity. However, it is not always clear which pathway
in an ISM scheme should be chosen, and what to do if a local
minimum in the respective fitness landscape is encountered,
that is, the absence of any improved variants in a given library.
In the case of CASTing, several previous studies have shown
[
8,9]
that many pathways lead to improved mutants.
Neverthe-
less, a complete ISM scheme has yet to be explored in the
quest to answer the question as to how many of the theoreti-
cally possible upward pathways do in fact provide notably im-
proved mutants in terms of activity and/stereoselectivity.
[
9]
Relative to the widely used iterative CASTing, thus far B-FIT
[
10,13]
has not been applied as often.
In the present study, we re-
sorted to ISM in the form of B-FIT in an effort to increase the
thermal robustness (thermoresistance) of the epoxide hydro-
lase from Aspergillus niger (ANEH), a 396 amino acid protein.
Ten residues characterized by the highest average B factors Experimental Section
and two unresolved residues were grouped into six sites for
Materials: Glycidyl phenyl ether (rac-1) was purchased from Acros
Geel, Belgium).
saturation mutagenesis, each comprising two amino acid posi-
tions. The initial six libraries were created by using convention-
al NNK codon degeneracy encoding all 20 canonical amino
acids. In order to reduce the screening effort while still main-
taining 95% library coverage, reduced amino alphabets were
used in subsequent randomization experiments, in most cases
NDT encoding 12 amino acids, and in one instance NNG en-
coding the eight amino acids not covered by NDT.
(
Calculation of oversampling and %coverage of libraries: All
numbers regarding the degree of oversampling as a function of
%
coverage of a mutant library and codon usage were calculated
by using the computer aids B-FIT or CASTER, available free of
charge on the homepage of the corresponding author (http://
www.kofo.mpg.de/manfred-reetz.html). These programs are based
on statistical algorithms published previously by Patrick and Firth
[
25a]
A limited amount of ISM exploration was performed, and led
assuming the absence of amino acid bias;
rithms, see the papers by Ostermeier
for related algo-
60
[25b]
[25c]
and by Pelletier.
to several notably improved mutants characterized by DT50
values in the range of 14 to 218C. This underscores the viabili-
ty of B-FIT. In four cases the complete six-step ISM pathways
were constructed and evaluated. In several instances local
minima were encountered, meaning that in a given library no
improved hits were detected. Normally, one would terminate
the exploration of such a pathway, the only possibilities being
Mutant library generation: ANEH mutant libraries were created
TM
[16a]
by QuikChange saturation mutagenesis method
Each PCR reaction contained (50 mL final volume): 10ꢂKOD buffer
(Novagen, 5 mL), dNTPs (2 mm each, 5 mL), MgCl (25 mm, 2 mL), de-
generate primers (forward and reverse 2.5 mm, 2.5 mL), template
(Stratagene).
2
ꢀ
1
plasmid (20–50 ngmL , 0.5 mL), KOD Hot start polymerase (0.5 U).
The thermal cycler program consisted of 1ꢂ958C 3 min, 20ꢂ958C
[
9]
backtracking or probing a different pathway. In the present
study we developed an unconventional strategy by testing un-
improved and even inferior mutants in such “dead-end” librar-
ies as templates in subsequent mutagenesis steps. Surprisingly,
continuation on a pathway of this type did in fact deliver im-
3
0 s, gradient 1 min, 728C 7 min, and 1ꢂ728C 16 min. The PCR
product was transformed into E. coli DH5a after DpnI treatment
New England Biolabs, 1 U, 378C, 2 h, 2ꢂdigestion in buffer provid-
(
ed by the supplier). The qualities of the libraries generated and the
ANEH mutants were analyzed by sequencing with primers pQE-for
ChemBioChem 2011, 12, 2502 – 2510
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
2507

M. T. Reetz and Y. Gumulya
(
GTATC ACGAG GCCCT TTCGT CT) or pQE-rev (CATTA CTGGA TCTAT
E. coli DH5a. An aliquot (20 mL) of overnight culture was inoculat-
CAACA GGAG; Eurofins MWG Operon, Ebersberg, Germany).
ed into expression culture (200 mL, TB medium supplemented
ꢀ1
with 100 mgmL carbenicillin). The culture was grown for 24 h at
08C (200 rpm). Cells were harvested by centrifugation at 1650g
Expression of gene libraries: Colonies were picked with colony
picker QPIX (Genetix, New Milton, UK), inoculated into 96-deep-
well plates containing LB medium (800 mL per well; supplemented
3
for 10 min at 48C. The cell pellet (about 4 g wet mass) was resus-
pended in lysis buffer (40 mL, with 0.5m NaCl, 20 mm imidazole).
The cell suspension was disrupted by sonication (Bandelin, 60 s, six
times, 40% pulse, on ice). The cell debris was removed by centrifu-
gation at 9500g for 1 h at 108C. The clarified lysate was firstly fil-
tered (0.22 mm filter) before being loaded onto 5 mL HisTrap FF af-
finity column (GE Healthcare). His-tagged enzymes were eluted
with an imidazole linear gradient by using ꢃKTA purifier. The frac-
tion containing the ANEH (10 mL) was concentrated by using ultra-
filtration centrifugal filter (10 kDa cut-off membrane, Amicon) to a
final volume of 3 mL, and was applied to a PD-10 desalting column
in order to remove excess imidazole. The eluted protein was dilut-
ed with PBS buffer (9 mL; 10 mm, pH 7.2), filter-sterilized, and
stored at 48C. Enzyme purity was monitored by SDS-PAGE. The
concentration of the ANEH mutants was measured spectrophoto-
ꢀ
1
with 100 mgmL carbenicillin), and incubated, overnight, at 378C
with shaking (800 rpm). An aliquot (100 mL) of this overnight cul-
ture was transferred into a glycerol plate (96-well plate filled with
1
00 mL glycerol solution 70% each well) and stored at ꢀ808C. An-
other aliquot (100 mL) was used to inoculate fresh medium and
grown for 24 h at 308C with shaking at 800 rpm. The cells were
harvested by centrifugation (1055g, 10 min), the supernatant was
removed and the cells were resuspended by adding lysis buffer
(
400 mL). The cell lysate was shaken at 378C for 1 h before being
stored at ꢀ208C. The cells were thawed the following day by
being shaken at 378C for 1 h and centrifuged at 1055g for 1 h.
Screening for thermal robustness: The screening for improved
thermal robustness was performed in 96-well PCR plates by heat-
ing enzyme solutions (cell free lysate) at 508C for 15 min in a PCR
thermocycler (Whatman Biometra, Gçttingen, Germany). After heat
treatment, the residual enzyme activity was measured by using
glycidyl phenyl ether (rac-1; GPE) as substrate. Heat treated cell
free lysates (22 mL) were transferred to a 96-well UV/Vis plate
reader; phosphate buffer saline (145 mL; pH 7.2, 10 mm) and sub-
strate solution (13 mL; 125 mg solubilized in 20 mL acetonitrile)
were added. After being shaken for 1 h at 308C (800 rpm), the re-
ꢀ1
metrically (the absorbance of a 1 mgmL solution at 280 nm is
.428 based on a molar extinction coefficient of e280nm =64400).
1
Analysis of the thermal robustness of purified ANEHs: The heat
treatment of purified ANEHs was carried out in thin-walled PCR
tubes (0.2 mL) by using a PCR thermocycler to allow precise tem-
ꢀ1
perature control. The purified ANEHs (70 mL, 0.012 mgmL
in
20 mm sodium phosphate buffer pH 7.2) was incubated at various
temperatures (e.g., 50, 55, 60, 65, and 708C) for a defined period
of time (e.g., 60 or 120 min). The samples were cooled on ice and
equilibrated at room temperature before the residual activity was
measured. Purified ANEH (56 mL), with or without heat treatment,
was transferred into 96-well plates and reaction mixture (124 mL;
26 mL glycidyl phenyl ether (41.62 mm solubilized in acetonitrile
HPLC grade)+98 mL sodium phosphate buffer, 20 mm, pH 7.2) was
added into the protein solution. The reaction plate was shaken at
308C for 1 h and then centrifuged at 1055g for 20 min. Then the
reaction mixture (150 mL) was transferred into new reaction plates
and mixed with the same volume of internal standard solution (R)-
(+)-1-phenyl-1-butanol (6.66 mm in HPLC-grade methanol). This
mixture was further centrifuged for another 30 min at 1055g. This
solution (200 mL) was finally transferred to 96-well microtiter plates
and the conversion was measured by HPLC. The chiral analysis of
the hydrolytic kinetic resolution of rac-gylcidyl phenyl ether (rac-1)
were performed by using Kromasil 3-Cellucoat RP 100 chiral
column (4.6 mm i.d, Akzo Nobel, Sweden) and Kromasil 5-Cellucoat
action was stopped by adding NaIO (20 mL, 25 mm). The reaction
4
was further developed for another 30 min. A blank measurement
was first carried out (A490cells =absorbance of cell free lysate+buf-
fer+solvent+substrate+NaIO ). The adenochrome absorption was
4
then measured after adding l-adrenaline (20 mL, 27.5 mm;
A490adren =absorbance of cell free lysate+buffer+solvent+sub
strate+NaIO +l-adrenaline). The same reaction was also performed
4
for the untreated samples (without heat treatment). The thermo-
stability was assessed by measuring the residual activity after expo-
sure to high temperatures. For each assay (12 different tempera-
tures), temperature was plotted against residual activity (see the
Supporting Information). Using this graph, the temperature, T, at
which 50% residual activity pertains was derived.
Screening scale reproduction: Positive clones were confirmed by
regrowing them in 96-deep-well plates, overnight, at 378C,
8
00 rpm. Then an aliquot (500 mL) of this overnight culture was
transferred into a culture tube (12 mL) filled with LB medium
1
0 (4.6 mm i.d.) precolumn. The conditions of analysis were as fol-
ꢀ
1
(
5 mL), supplemented with carbenicillin (100 mgmL ), and incubat-
ꢀ1
lows: methanol/H O 70:30, 1.0 mLmin , UV 220 nm, 298 K. The
2
ed for 24 h at 308C at 800 rpm. The cells were harvested by centri-
fugation at 1055g for 10 min; lysis buffer (2.5 mL) was then added
and the cells were shaken for 1 h at 378C before being stored at
corresponding retention times were, (R)-2: 1.9 min; (S)-2: 2.1 min;
(R)-1: 4.1 min; (S)-1: 4.7 min.
The inactivation energy (E ) of the mutants were measured by
ꢀ
208C. Cells were thawed the following day by being shaken at
78C (800 rpm) for 1 h and then centrifuged at 1055g for 1 h. The
cell free lysate was heat-shocked in a gradient PCR thermocycler
45–65 or 708C). The enzyme residual activity was measured, as de-
a
using Arrhenius plots of the rate of thermal inactivation (lnK ,
3
d
ꢀ
1
min ) at different temperatures versus the reciprocal of the abso-
ꢀ1
lute temperature (1000/T, K ).
(
scribed above.
Acknowledgements
Expression and purification of thermostable mutants: The vector
carrying the ANEH gene (pQE-60, Qiagen) has His6 coding se-
quence at the gene C terminus. The stop codons of pQEEH, WT
ANEH and mutants were deleted to allow transcription to be con-
tinued into the polyhistidine tagged coding sequence. Site-direct-
ed mutagenesis was performed according to the improved PCR
method by using the mutagenic primers: GAGCA GGTGT GGCAG
AAGAG ATCTC ATCAC C and CCTCG CTCTG CTAAT CCTGT TACCA
GTGGC (Invitrogen). WT ANEH and mutants were over-expressed in
Support from the Arthur C. Cope Foundation (USA) is gratefully
acknowledged.
Keywords: directed evolution
hydrolases · quasi-species · neutral drift
·
enzymes
·
epoxide
2
508
www.chembiochem.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 2502 – 2510

Enhancing the Thermal Robustness of an Enzyme
[
1] a) K. Faber, Biotransformations in Organic Chemistry, 5th ed., Springer,
Berlin, 2004; b) K. Drauz, H. Waldmann, Enzyme Catalysis in Organic Syn-
thesis: A Comprehensive Handbook, Vols. I–III, 2nd ed., Wiley-VCH, Wein-
heim, 2002; c) A. Liese, K. Seelbach, C. Wandrey, Industrial Biotransfor-
mations, 2nd ed., Wiley-VCH, Weinheim, 2006; d) J. Tao, G.-Q. Lin, A.
Liese, Biocatalysis for the Pharmaceutical Industry, Wiley-VCH, Weinheim,
(Eds.: V. Gotor, I. Alfonso, E. Garcꢄa-Urdiales), Wiley-VCH, Weinheim,
2008, pp. 21–63; l) L. G. Otten, F. Hollmann, I. W. C. E. Arends, Trends
Biotechnol. 2010, 28, 46–54; m) A. V. Shivange, J. Marienhagen, H.
Mundhada, A. Schenk, U. Schwaneberg, Curr. Opin. Chem. Biol. 2009, 13,
19–25.
[7] a) J.-L. Reymond, Enzyme Assays: High-Throughput Screening, Genetic Se-
lection and Fingerprinting, Wiley-VCH, Weinheim, 2006; b) Y. L. Boersma,
M. J. Drçge, W. J. Quax, FEBS J. 2007, 274, 2181–2195; c) J.-L. Reymond,
V. S. Fluxꢊ, N. Maillard, Chem. Commun. 2009, 34–46.
[8] a) M. T. Reetz, L.-W. Wang, M. Bocola, Angew. Chem. 2006, 118, 1258–
1263; Angew. Chem. Int. Ed. 2006, 45, 1236–1241; b) M. T. Reetz, D. Ka-
hakeaw, J. Sanchis, Mol. BioSyst. 2009, 5, 115–122.
2009; e) V. Gotor, I. Alfonso, E. Garcꢄa-Urdiales, Asymmetric Organic Syn-
thesis with Enzymes, Wiley-VCH Weinheim, 2008.
[
2] a) T. Oshima, Curr. Opin. Struct. Biol. 1994, 4, 623–628; b) C. ꢅ’Fꢆgꢆin,
Enzyme Microb. Technol. 2003, 33, 137–149; c) V. G. H. Eijsink, A. Bjørk,
S. Gꢇseidnes, R. Sirevꢇg, B. Synstad, B. van den Burg, G. Vriend, J. Bio-
technol. 2004, 113, 105–120.
[
3] Reviews of directed evolution of thermal robustness of proteins: a) P. L.
Wintrode, F. H. Arnold, Adv. Protein Chem. 2001, 55, 161–225; b) V. G. H.
Eijsink, S. Gaseidnes, T. V. Borchert, B. van den Burg, Biomol. Eng. 2005,
[
9] Review of iterative saturation mutagenesis for evolving stereoselective
enzymes: M. T. Reetz, Angew. Chem. 2011, 123, 144–182; Angew. Chem.
Int. Ed. 2011, 50, 138–174.
2
2, 21–30; c) A. Martin, F. X. Schmid, V. Sieber in Directed Enzyme Evolu-
[
10] a) M. T. Reetz, J. D. Carballeira, A. Vogel, Angew. Chem. 2006, 118, 7909–
915; Angew. Chem. Int. Ed. 2006, 45, 7745–7751; b) M. T. Reetz, J. D.
tion: Screening and Selection Methods, Vol. 230, (Eds.: F. H. Arnold, G.
Georgiou), Humana, Totowa, New Jersey, 2003, pp. 57–70; d) N. Amin,
A. D. Liu, S. Ramer, W. Aehle, D. Meijer, M. Metin, S. Wong, P. Gualfetti, V.
Schellenberger, Protein Eng. Des. Sel. 2004, 17, 787–793; e) R. A. Chica,
N. Doucet, J. N. Pelletier, Curr. Opin. Biotechnol. 2005, 16, 378–384.
4] Examples of the use of T50 values or simply of data pertaining to residu-
al activity after heat treatment at a given temperature: a) L. Sun, I. P.
Petrounia, M. Yagasaki, G. Bandara, F. H. Arnold, Protein Eng. 2001, 14,
7
Carballeira, Nat. Protoc. 2007, 2, 891–903.
[
[
11] W. Augustyniak, A. A. Brzezinska, T. Pijning, H. Wienk, R. Boelens, B. Dijk-
stra, M. T. Reetz, unpublished results.
12] a) P. Acharya, E. Rajakumara, R. Sankaranarayanan, N. M. Rao, J. Mol. Biol.
[
2
004, 341, 1271–1281; b) S. Ahmad, B. M. Rao, Protein Sci. 2009, 18,
1183–1196.
[
[
13] a) D. F. Visser, F. Hennessy, J. Rashamuse, B. Pletschke, D. Brady, J. Mol.
Catal. B 2011, 68, 279–285; b) H. Jochens, D. Aerts, U. T. Bornscheuer,
Protein Eng. Des. Sel. 2010, 23, 903–909.
14] a) D. Wahler, J.-L. Reymond, Angew. Chem. 2002, 114, 1277–1280;
Angew. Chem. Int. Ed. 2002, 41, 1229–1232; b) D. Wahler, O. Boujard, F.
Lefꢋvre, J.-L. Reymond, Tetrahedron 2004, 60, 703–710; c) F. Cedrone, T.
Bhatnagar, J. C. Baratti, Biotechnol. Lett. 2005, 27, 1921–1927; d) D. Ka-
hakeaw, M. T. Reetz, Chem. Asian J. 2008, 3, 233–238.
6
2
6
99–704; b) B. Morawski, S. Quan, F. H. Arnold, Biotechnol. Bioeng.
001, 76, 99–107; c) J. Hao, A. Berry, Protein Eng. Des. Sel. 2004, 17,
89–697; d) N. Declerck, M. Machius, P. Joyet, G. Wiegand, R. Huber, C.
Gaillardin, Protein Eng. 2003, 16, 287–293; e) S. Akanuma, A. Yamagishi,
N. Tanaka, T. Oshima, Protein Sci. 1998, 7, 698–705; f) H. Zhao, F. H.
Arnold, Protein Eng. 1999, 12, 47–53; g) B. Van den Burg, A. de Kreij, P.
Van der Veek, J. Mansfeld, G. Venema, Biotechnol. Appl. Biochem. 1999,
3
6
0, 35–40; h) K.-H. Oh, S.-H. Nam, H.-S. Kim, Protein Eng. 2002, 15, 689–
95; i) M. G. Pikkemaat, A. B. M. Linssen, H. J. C. Berendsen, D. B. Janssen,
[
[
15] a) J. Y. Zou, B. M. Hallberg, T. Bergfors, F. Oesch, M. Arand, S. L. Mowbray,
T. A. Jones, Structure 2000, 8, 111–122; b) http://www.pdb.ccdc.cam.
ac.uk/pdb.
Protein Eng. 2002, 15, 185–192; j) L.-J. Wang, X.-D. Kong, H.-Y. Zhang,
X.-P. Wang, J. Zhang, Biochem. Biophys. Res. Commun. 2000, 276, 346–
16] a) H. H. Hogrefe, J. Cline, G. L. Youngblood, R. M. Allen, BioTechniques
3
49; k) J. Mansfeld, G. Vriend, B. W. Dijkstra, O. R. Veltman, B. Van den
Burg, G. Venema, R. Ulbrich-Hofmann, V. G. H. Eijsink, J. Biol. Chem.
997, 272, 11152–11156; l) E. H. Muslin, S. E. Clark, C. A. Henson, Protein
2
002, 33, 1158–1165; selected early developments of saturation muta-
genesis methods and various applications: b) J. A. Wells, M. Vasser, D. B.
Powers, Gene 1985, 34, 315–323; c) M. S. Z. Horwitz, L. A. Loeb, Proc.
Natl. Acad. Sci. USA 1986, 83, 7405–7409; d) A. R. Oliphant, A. L. Nuss-
baum, K. Struhl, Gene 1986, 44, 177–183; e) J. F. Reidhaar-Olson, R. T.
Sauer, Science 1988, 241, 53–57; f) L. B. Evnin, J. R. Vasquez, C. S. Craik,
Proc. Natl. Acad. Sci. USA 1990, 87, 6659–6663; g) J. C. Hu, E. K. O’Shea,
P. S. Kim, R. T. Sauer, Science 1990, 250, 1400–1403; h) S. Kamtekar, J. M.
Schiffer, H. Xiong, J. M. Babik, M. H. Hecht, Science 1993, 262, 1680–
1
Eng. 2002, 15, 29–33; m) O. Turunen, M. Vuorio, F. Fenel, M. Leisola,
Protein Eng. 2002, 15, 141–145; n) T. W. Johannes, R. D. Woodyer, H.
Zhao, Appl. Environ. Microbiol. 2005, 71, 5728–5734; o) J.-H. Kim, G.-S.
Choi, S.-B. Kim, W.-H. Kim, J.-Y. Lee, Y.-W. Ryu, G.-J. Kim, J. Mol. Catal. B
2
004, 27, 169–175; p) D. E. Stephens, K. Rumbold, K. Permaul, B. A.
Prior, S. Singh, J. Biotechnol. 2007, 127, 348–354.
5] Recent examples of protein engineered enzyme thermostabilization:
a) M. J. McLachlan, T. W. Johannes, H. Zhao, Biotechnol. Bioeng. 2008, 99,
2
[
1685; i) D. M. Kegler-Ebo, C. M. Docktor, D. DiMaio, Nucleic Acids Res.
1
994, 22, 1593–1599; j) G. Chen, I. Dubrawsky, P. Mendez, G. Georgiou,
68–274; b) A. Nakamura, Y. Takakura, N. Sugimoto, N. Takaya, K. Shira-
B. L. Iverson, Protein Eng. 1999, 12, 349–356; first example of saturation
mutagenesis in the quest to enhance enantioselectivity: k) M. T. Reetz,
S. Wilensek, D. Zha, K.-E. Jaeger, Angew. Chem. 2001, 113, 3701–3703;
Angew. Chem. Int. Ed. 2001, 40, 3589–3591.
ki, T. Hoshino, Biosci. Biotechnol. Biochem. 2008, 72, 2467–2471; c) K.
Hirokawa, A. Ichiyanagi, N. Kajiyama, Appl. Microbiol. Biotechnol. 2008,
78, 775–781; d) L. Foit, G. J. Morgan, M. J. Kern, L. R. Steimer, A. A. von
Hacht, J. Titchmarsh, S. L. Warriner, S. E. Radford, J. C. A. Bardwell, Mol.
Cell 2009, 36, 861–871; e) H. Jochens, D. Aerts, U. T. Bornscheuer, Pro-
tein Eng. Des. Sel. 2010, 23, 903–909; f) A. S. Bommarius, J. K. Blum,
M. J. Abrahamson, Curr. Opin. Chem. Biol. 2011, 15, 194–200; g) E.
Garcꢄa-Ruiz, D. Matꢈ, A. Ballesteros, A. T. Martinez, M. Alcade, Microb.
Cell Fact. 2010, 9, 17; h) J. C. Joo, S. P. Pack, Y. H. Kim, Y. J. Yoo, J. Biotech-
nol. 2011, 151, 56–65; i) D. Gessmann, F. Mager, H. Naveed, T. Arnold, S.
Weirich, D. Linke, J. Liang, S. Nussberger, J. Mol. Biol. 2011; DOI:
[
17] a) S. G. Peisajovich, D. S. Tawfik, Nat. Methods 2007, 4, 991–994; b) M. A.
DePristo, HFSP J. 2007, 1, 94–98; c) J. D. Bloom, F. H. Arnold, Proc. Natl.
Acad. Sci. USA 2009, 106, 9995–10000.
18] M. Eigen, J. McCaskill, P. Schuster, J. Phys. Chem. 1988, 92, 6881–6891.
19] a) L. O. Emren, S. Kurtovic, A. Runarsdottir, A.-K. Larsson, B. Mannervik,
Proc. Natl. Acad. Sci. USA 2006, 103, 10866–10870; b) S. Kurtovic, B.
Mannervik, Biochemistry 2009, 48, 9330–9339; c) B. Mannervik, A. Run-
arsdottir, FEBS Lett. 2010, 584, 2565–2571; d) A. Runarsdottir, B. Man-
nervik, J. Mol. Biol. 2010, 401, 451–464.
[
[
10.1016/j.jmb.2011.07.054.
[
6] a) S. Lutz, W. M. Patrick, Curr. Opin. Biotechnol. 2004, 15, 291–297;
b) M. R. Parikh, I. Matsumura, J. Mol. Biol. 2005, 352, 621–628; c) J. M.
Joern, P. Meinhold, F. H. Arnold, J. Mol. Biol. 2002, 316, 643–656; d) T.-W.
Wang, H. Zhu, X.-Y. Ma, T. Zhang, Y.-S. Ma, D.-Z. Wei, Mol. Biotechnol.
[20] D. J. Bougioukou, S. Kille, A. Taglieber, M. T. Reetz, Adv. Synth. Catal.
2009, 351, 3287–3305.
[21] a) S. Kumar, C. Tsai, B. Ma, R. Nussinov, J. Biol. Struct. Dynam. 2000, 11,
79–85; b) A. H. Elcock, J. Mol. Biol. 1998, 284, 489–502; c) C. Vetriani,
D. L. Maeder, N. Tolliday, K. S.-P. Yip, T. J. Stillman, K. L. Britton, D. W.
Rice, H. H. Klump, F. T. Robb, Proc. Natl. Acad. Sci. USA 1998, 95, 12300–
12305; d) J. H. G. Lebbink, S. Knapp, J. van der Oost, D. Rice, R. Laden-
stein, W. M. de Vos, J. Mol. Biol. 1999, 289, 357–369; e) L. Xiao, B. Honig,
J. Mol. Biol. 1999, 289, 1435–1444; f) G. I. Makhatadze, V.-V. Loladze,
D. N. Ermolenko, X. Chen, S. T. Thomas, J. Mol. Biol. 2003, 327, 1135–
1148; g) B. N. Dominy, H. Minoux, C. L. Brooks, Proteins Struct. Funct.
2006, 34, 55–68; Recent reviews of directed evolution: e) T. W. Jo-
hannes, H. Zhao, Curr. Opin. Microbiol. 2006, 9, 261–267; f) S. Lutz, U. T.
Bornscheuer, Protein Engineering Handbook, Vols. 1–2, Wiley-VCH, Wein-
heim, 2009; g) N. J. Turner, Nat. Chem. Biol. 2009, 5, 567–573; h) C.
Jꢉckel, P. Kast, D. Hilvert, Annu. Rev. Biophys. Biomol. Struct. 2008, 37,
153–173; i) S. Bershtein, D. S. Tawfik, Curr. Opin. Chem. Biol. 2008, 12,
151–158; j) P. A. Romero, F. H. Arnold, Nat. Rev. Mol. Cell Biol. 2009, 10,
866–876; k) M. T. Reetz in Asymmetric Organic Synthesis with Enzymes
ChemBioChem 2011, 12, 2502 – 2510
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
2509

M. T. Reetz and Y. Gumulya
Bioinf. 2004, 57, 128–141; h) M. Robinson-Rechavi, A. Alibꢈs, A. Godzik,
J. Mol. Biol. 2006, 356, 547–557.
22] Y. Gumulya, Doctoral Thesis, Ruhr-Universitꢉt Bochum (Germany), 2010.
23] a) A. Grinberg, R. Bernhardt, Biochem. Biophys. Res. Commun. 1998, 249,
dendistel, B. Hauer, D. Leys, J. Micklefield, Angew. Chem. 2009, 121,
7827–7830; Angew. Chem. Int. Ed. 2009, 48, 7691–7694; h) M. R. M. De
Groeve, M. De Baere, L. Hoflack, T. Desmet, E. J. Vandamme, W. Soetaert,
Protein Eng. Des. Sel. 2009, 22, 393–399.
[
[
9
33–937; b) O. Almog, D. T. Gallagher, J. E. Ladner, S. Strausberg, P.
Alexander, P. Bryan, G. L. Gilliland, J. Biol. Chem. 2002, 277, 27553–
7558.
[25] a) W. M. Patrick, A. E. Firth, Biomol. Eng. 2005, 22, 105–112; b) A. D.
Bosley, M. Ostermeier, Biomol. Eng. 2005, 22, 57–61; c) M. Denault, J. N.
Pelletier in Protein Engineering Protocols, Vol. 352 (Eds.: K. M. Arndt, K. M.
Mꢌller), Humana, Totowa, 2007, pp. 127–154; see also earlier ap-
proaches regarding this type of statistical analysis, for example: d) M. S.
Warren, S. J. Benkovic, Protein Eng. 1997, 10, 63–68.
2
[
24] Recent examples of ISM (refs. [9,13]): a) S. Kille, F. E. Zilly, J. P. Acevedo,
M. T. Reetz, Nat. Chem. 2011, 3, 738–743; b) M. A. Norrgard, B. Manner-
vik, J. Mol. Biol. 2011, 412, 111–120; c) W. L. Tang, Z. Li, H. Zhao, Chem.
Commun. 2010, 46, 5461–5463; d) K. Engstrçm, J. Nyhlen, A. G. Sand-
strçm, J.-E. Bꢉckvall, J. Am. Chem. Soc. 2010, 132, 7038–7042; e) M.-W.
Bhuiya, C.-J. Liu, J. Biol. Chem. 2010, 285, 277–285; f) M. T. Reetz, S.
Prasad, J. D. Carballeira, Y. Gumulya, M. Bocola, J. Am. Chem. Soc. 2010,
Received: June 27, 2011
132, 9144–9152; g) K. Okrasa, C. Levy, M. Wilding, M. Goodall, N. Bau-
Published online on September 13, 2011
2
510
www.chembiochem.org
ꢀ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 2502 – 2510