Comparing Different Strategies in Directed Evolution of Enzyme Stereoselectivity: Single- versus Double-Code Saturation Mutagenesis

Article abstract of DOI:10.1002/cbic.201600296

Saturation mutagenesis at sites lining the binding pockets of enzymes constitutes a viable protein engineering technique for enhancing or inverting stereoselectivity. Statistical analysis shows that oversampling in the screening step (the bottleneck) increases astronomically as the number of residues in the randomization site increases, which is the reason why reduced amino acid alphabets have been employed, in addition to splitting large sites into smaller ones. Limonene epoxide hydrolase (LEH) has previously served as the experimental platform in these methodological efforts, enabling comparisons between single-code saturation mutagenesis (SCSM) and triple-code saturation mutagenesis (TCSM); these employ either only one or three amino acids, respectively, as building blocks. In this study the comparative platform is extended by exploring the efficacy of double-code saturation mutagenesis (DCSM), in which the reduced amino acid alphabet consists of two members, chosen according to the principles of rational design on the basis of structural information. The hydrolytic desymmetrization of cyclohexene oxide is used as the model reaction, with formation of either (R,R)- or (S,S)-cyclohexane-1,2-diol. DCSM proves to be clearly superior to the likewise tested SCSM, affording both R,R- and S,S-selective mutants. These variants are also good catalysts in reactions of further substrates. Docking computations reveal the basis of enantioselectivity.

Full text of DOI:10.1002/cbic.201600296

Accepted Article
Title: Comparing Different Strategies in Directed Evolution of Enzyme Stereoselec-
tivity: Single versus Double Code Saturation Mutagenesis
Authors: Zhoutong Sun; Richard Lonsdale; Guangyue Li; Manfred T. Reetz
This manuscript has been accepted after peer review and the authors have elected
to post their Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by using the Digital
Object Identifier (DOI) given below. The VoR will be published online in Early View as
soon as possible and may be different to this Accepted Article as a result of editing.
Readers should obtain the VoR from the journal website shown below when it is pu-
blished to ensure accuracy of information. The authors are responsible for the content
of this Accepted Article.
To be cited as: ChemBioChem 10.1002/cbic.201600296
Link to VoR: http://dx.doi.org/10.1002/cbic.201600296
A Journal of
www.chembiochem.org

ChemBioChem
10.1002/cbic.201600296
FULL PAPERS
Comparing Different Strategies in Directed Evolution of Enzyme
Stereoselectivity: Single versus Double Code Saturation
Mutagenesis
Zhoutong Sun[a, b], Richard Lonsdale[a, b], Guangyue Li[a, b] and Manfred T. Reetz*[a, b]
Abstract: Saturation mutagenesis at sites lining the binding pocket of
enzymes constitutes a viable protein engineering technique for
enhancing or inverting stereoselectivity. Statistical analysis shows
that oversampling in the screening step (the bottleneck) increases
astronomically as the number of residues in the randomization site
increases, which is the reason why reduced amino acid alphabets
have been employed in addition to splitting large sites into smaller
ones. Limonene epoxide hydrolase (LEH) has previously served as
the experimental platform in these methodology efforts, enabling
comparisons between single code saturation mutagenesis (SCSM)
and triple code saturation mutagenesis (TCSM) which employ only
one versus three amino acids as building blocks, respectively. In the
present study the comparative platform is extended by exploring the
efficacy of double code saturation mutagenesis (DCSM) in which the
reduced amino acid alphabet comprises two members, chosen
rationally on the basis of structural information. The hydrolytic
desymmetrization of cyclohexene oxide is used as the model reaction
with formation of either (R,R)- or (S,S)-cyclohexane-1,2-diol. DCSM
proves to be clearly superior to the likewise tested SCSM, affording
both (R,R)- and (S,S)-selective mutants. The respective variants are
also good catalysts in the reaction of other further substrates. Docking
computations reveal the origin of enantioselectivity.
one residue to, e.g., 10. Therefore, two “tricks” have been
proposed which offer a way out of this dilemma: 1) Splitting up a
large site into smaller ones, followed by iterative saturation
mutagenesis (ISM) if necessary;[1g,4] and/or 2) Use of reduced
amino acid alphabets.[1g,4] Reduced amino acid alphabets have
been used previously for other purposes, e.g., studying whether
correct folds[5a] and enzyme activity[5b] are maintained when
employing less than the 20 canonical amino acids in protein
construction, or when applying a binary code in the production of
functional antibodies[6a] or monobodies.[6b] In the applications that
we focus on, consideration of oversampling is essential. The
necessary degree of oversampling for ensuring a certain degree
of library coverage that the operator stipulates, e.g., 95%, is
easily calculated using the CASTER computer aid
(www.kofo.mpg.de/en/research/biocatalysis),[7] which is based on
the Patrick/Firth algorithm.[8] Alternatively, the Nov-metric can be
used to compute the nth best mutant as a function of the number
of screened transformants.[9] One and the same reduced amino
acid alphabet as defined by the respective codon degeneracy can
be applied when randomizing combinatorially the whole multi-
residue site, or a different codon degeneracy can be chosen at
each individual position of a multi-residue site, likewise in a single
SM experiment.[10]
Introduction
In the present study we compare the virtues and drawbacks of
two different approaches to SM using appropriately chosen codon
degeneracies for controlling stereoselectivity: Double code
saturation mutagenesis (DCSM) based on two amino acids as
building blocks versus single code saturation mutagenesis
(SCSM) utilizing only one amino acid, in both cases for the same
entire randomization site. In these and other reduced amino acid
alphabets, wildtype (WT) amino acids are included.[10] In the
model system we employ limonene epoxide hydrolase (LEH)[11]
as the catalyst for the hydrolytic desymmetrization of cyclohexene
oxide (1) with formation of cyclohexane-1,2-diols (R,R)- or (S,S)-2
(Scheme 1). WT LEH shows minimal enantioselectivity in slight
favor of (S,S)-2 (4% ee). This experimental platform has already
been utilized in several previous directed evolution studies,
enabling further comparisons.[12] In one approach, 10 CAST
residues were grouped into a single large site which was
randomized using SCSM with valine, phenylalanine or tyrosine
being tested as the sole building block, respectively.[12b] Valine
proved to be the superior building block in this particular enzyme.
When applying triple code saturation mutagenesis (TCSM)
employing three amino acids as building blocks in randomization,
even better results were obtained.[12c]
Methodology development in directed evolution of stereoselective
enzymes is an important current task, aiming to enhance the
speed, efficacy and reliability of this widely practiced protein
engineering technique.[1] Since screening for stereoselectivity,
activity and/or substrate scope is the labor-intensive bottleneck,
the goal is the creation of small and maximally “smart” mutant
libraries. In this endeavor, a stochastic process called saturation
mutagenesis (SM)[2] at sites lining the binding pocket has
emerged as the method of choice (dubbed CASTing) for many
enzymes.[1a,1c,3] When employing NNK codon degeneracy
encoding all 20 canonical amino acids as building blocks, protein
sequence space and concurrent oversampling increases
astronomically as the size of a randomization site increases from
[a]
Dr. Z. Sun, Dr. R. Lonsdale, Dr. G. Li, Prof. Dr. M. T. Reetz
Department of Synthetic Organic Chemistry
Max-Planck-Institut für Kohlenforschung
Kaiser-Wilhelm-Platz 1, 45470 Mülheim an der Ruhr (Germany)
Fax: (+49) 208-306-2985
E-mail: reetz@mpi-muelheim.mpg.de
[b]
Dr. Z. Sun, Dr. R. Lonsdale, Dr. G. Li, Prof. Dr. M. T. Reetz
Fachbereich Chemie
Philipps-Universität Marburg
Hans-Meerwein-Strasse, 35032 Marburg (Germany)
Supporting information for this article is available on the WWW
under http://www.chembiochem.org or from the author.
1

ChemBioChem
10.1002/cbic.201600296
The crystal structure of WT LEH[11b] shows that almost all of the
residues surrounding the binding pocket have hydrophobic
character. Therefore, when designing a single or triple code, it
was logical to choose amino acids with hydrophobic sidechains.
In the case of the previous SCSM study[12b], valine, phenylalanine
or tyrosine were chosen, valine leading to the best results in
terms of enantioselectivity and activity, which was improved
further by ISM using one of the other two amino acids as building
blocks. In the TCSM study,[12c] V-F-Y was chosen as the triple
code, which provided notably better results.
Scheme 1. Hydrolytic desymmetrization catalyzed by mutants of
limonene epoxide hydrolase (LEH) generated by different
saturation mutagenesis strategies.
Results and Discussion
Based on these data, we chose valine for SCSM at
randomization sites A5 and B5, respectively. In each case, 92
transformants corresponding to ~95% library coverage were
assayed for activity, the hits (55) then being screened for
enantioselectivity using chiral GC. The results are summarized in
Generation and screening of mutant libraries. As in previous
directed evolution studies of LEH,[12b-c] the crystal structure of
wildtype (WT) LEH[11b] served as a guide for choosing ten CAST
residues. This time they were grouped into two 5-residue
randomization sites, A5 (L74/F75/M78/I80/L103) and B5
(L114/I116/F134/F139/L147) (Fig. 1), thereby setting the platform
for systematic comparison of SCSM and DCSM.
Table
1.
The
best
(R,R)-selective
mutant
SZ539
(L74V/M78V/I80V) originated from library A5, showing 77% ee. It
was then used as a template for SCSM at site B5, but this failed
to provide any improved variants.
In case of site B5, mutant SZ544 (L114V/I116V/F139V) was
discovered showing 75% ee in favor of (S,S)-2. This mutant
formed the basis of SCSM at site A5, which provided variant
SZ634 (L103V/L114V/I116V/F139V) with enhanced (S,S)-
selectivity (89% ee). Further ISM was carried out using SZ639
(L74V/M78V/I80V) and SZ634 (L103V/L114V/I116V/F139V) as
templates to visit the rest of the residues of L103/L114/I116/L147
and L74/M78/I80/L147 with phenylalanine as building block,
respectively. The improved hits up to 92% ee were found in
library AB5-V-F-2. The best results are summarized in Table 1.
Figure 1. LEH binding pocket featuring substrate 1 surrounded
by 10 CAST residues which are grouped into two 5-residue
randomization sites A5 (blue) and B5 (yellow).
Table 1. Results of applying single code saturation mutagenesis (SCSM) to LEH-catalyzed hydrolytic desymmetrization of substrate 1.
Library
(Code)
%-
conv.
Favored
enantiomer
Template
WT
Code
WT
Mutations
ee%
A/T*
4
7
99
97
98
(S,S)
(S,S)
(R,R)
(R,R)
SZ537
SZ538
SZ539
SZ541
SZ543
SZ544
SZ546
L74V/L103V
L74V/I80V
A5-V
(Val)
61
92/96
L74V/M78V/I80V
I80V
77
75
54
75
63
97
97
95
98
95
(R,R)
(S,S)
(S,S)
(S,S)
L114V/I116V/L147V
L114V/I116V/F139V
L114V/I116V
B5-V
(Val)
WT
92/96
A5-V-B5-V
(Val)
SZ539
SZ544
SZ639
SZ633
SZ634
SZ635
L74V/M78V/I80V
F75V/L103V/L114V/I116V/F139V
L103V/L114V/I116V/F139V
L114V/I116V/F139V
77
88
89
88
97
97
99
99
(R,R)
(S,S)
(S,S)
(S,S)
92/96
92/96
B5-V-A5-V
(Val)
AB5-V-F-1
(Phe)
SZ639
SZ634
no improvement
46/48
46/48
(S,S)
(S,S)
(S,S)
(S,S)
SZ662
SZ663
L74F/M78F/L103V/L114V/I116V/F139V/L147F
L74F/M78F/L103V/L114V/I116V/F139V
L74F/I80F/L103V/L114V/I116V/F139V
L74F/M78F/I80F/L103V/L114V/I116V/F139V
90
90
92
92
89
95
94
96
AB5-V-F-2
(Phe)
SZ664
SZ665
*A/T: Colonies picked/ computed colonies number for 95% coverage
2

ChemBioChem
10.1002/cbic.201600296
Before presenting the data obtained by application of DCSM,
the present SCSM strategy can be compared to the original
SCSM approach in which the same 10 residues formed a single
randomization site with valine also serving as the single code
followed by phenylalanine-based ISM.[12b] The final results of the
two approaches are comparable, but the screening effort differs
considerably. The advantage of grouping 10 residues into two 5-
residue compared with the previous study in which all residues
formed a single randomization site using valine as the single code
is the fact that the number of screened transformants has been
reduced by a factor of eight (from 3220 to 368).
phenylalanine were chosen as the double code (V-F) for
randomizing sites A5 and B5. In the A5-VF library using WT as
template, mainly (R,R)-selective variants were found, the two best
hits being M78F/I80V (80% ee) and M78V/I80V (75% ee) (Table
2). Enantioselectivity was then boosted by performing ISM. These
two mutants were then used as template to visit the B5 site with
generation of two libraries A5-B5-VF-1 and A5-B5-VF-2,
respectively. The best hits SZ643 (M78F/I80V/L114F/L147V) and
SZ655 (M78V/I80V/L114F) reached ≈92% ee. On the other hand,
the initial B5-VF library contained the best (S,S)-selective variant
SZ597 (L114V/I116V/F139V/L147F) (85% ee). ISM based on
employing SZ597 as template and focusing randomization on site
A5 provided several improved mutants showing enantioselectivity
in the range 92-95% ee. The results are summarized in Table 2.
The main goal of the present study was to compare SCSM with
DCSM within the same experimental framework as defined by the
two 5-residue randomization sites. In this case valine and
Table 2. Results of applying double code saturation mutagenesis (DCSM) based on valine-phenylalanine as building blocks to LEH-
catalyzed hydrolytic desymmetrization of substrate 1.
%-
Favored
Library
Template
Code
WT
Mutations
ee%
4
conv.
enantiomer
A/T*
99
80
99
99
92
99
99
99
98
100
99
98
71
97
22
86
82
88
43
99
98
97
97
96
99
96
98
92
93
94
82
(S,S)
(S,S)
(R,R)
SZ548
SZ549
SZ550
SZ551
SZ553
SZ574
SZ575
SZ576
SZ577
SZ580
SZ583
SZ558
SZ562
SZ566
SZ569
SZ570
SZ597
SZ603
SZ605
SZ640
SZ641
SZ642
SZ643
SZ655
SZ646
SZ648
SZ649
SZ650
SZ651
SZ653
L74F
L74V/M78V/I80V
13
59
53
54
72
80
68
71
58
75
80
21
83
36
72
72
85
61
82
90
84
90
91
92
94
92
93
95
93
92
L74F/M78F/I80F
(S,S)
(R,R)
F75V/M78F
(R,R)
(R,R)
(R,R)
(R,R)
I80V
A5-VF
WT
M78F/I80V
552/486
M78F/I80V/L103V
F75V/M78F/I80V
L74F/M78V
(S,S)
(R,R)
M78V/I80V
(R,R)
(R,R)
L74V/M78F/I80V
L114F
L114V/I116V/F139V
L114F/L147V
(S,S)
(R,R)
(S,S)
B5-VF
WT
L114V/I116V
368/324
(S,S)
(S,S)
L114V/I116V/L147F
L114V/I116V/F139V/L147F
L114F/L147F
(R,R)
L114V/I116V/F139V/L147V
M78F/I80V/L114F
(S,S)
(R,R)
(R,R)
(R,R)
(R,R)
(R,R)
(S,S)
(S,S)
(S,S)
(S,S)
(S,S)
(S,S)
A5-B5-
VF-1
SZ574
SZ580
SZ597
M78F/I80V
368/324
368/324
552/486
M78F/I80V/L114F/L147F
M78F/I80V/L114F/L147V
M78V/I80V/L114F
A5-B5-VF-2
L74F/M78F/I80F/L114V/I116V/F139V/L147F
M78V/I80F/L114V/I116V/F139V/L147F
I80F/L114V/I116V/F139V/L147F
L74F/I80F/L114V/I116V/F139V/L147F
M78F/I80F/L114V/I116V/F139V/L147F
L74F/M78F/L114V/I116V/F139V/L147F
B5-A5-
VF-3
*A/T: Colonies picked/ computed colonies number for 95% coverage
Even though the particular version of SCSM in this study
involved the screening of only 460 transformants and provided
excellent (S,S)-variants, it failed to generate optimal (R,R)-
selective mutants. When comparing the present DCSM strategy
with the original SCSM approach described in the previous study
featuring best mutants (S,S)-selective (92%ee) and (R,R)-
selective mutants (up to 96% ee),[12b] it becomes clear that similar
enantioselectivities occur in both cases. However, the total
screening effort in the original SCSM approach (4508
transformants) is about twice as high as in the present DCSM
strategy (2208 transformants).
Finally, we were interested in the outcome of applying DCSM
to the entire 10-residue randomization site in the absence of
splitting it into smaller sites (Fig. 1). Here again the double code
V-F was used, which means that ≈5x104 transformants would
have to be screened for 95% coverage of the respective library
C10-VF. However, in this case no effort was made to reach high
library coverage. Instead, only 1840 transformants were screened
3

ChemBioChem
10.1002/cbic.201600296
(3.5% library coverage). This is equal to the screening in the case
of DCSM with split libraries, thereby allowing for comparison. The
results show that several excellent (S,S)-selective variants were
the same building block whenever the initial library fails to fulfill all
requirements that the operator has defined. The results can also
be compared to our original study in which an even smaller
reduced amino acid alphabet (valine) was used to randomize the
same 10-residue site in LEH.[12b] In that case full library coverage
was easily achieved, leading to the discovery of both (S,S)- and
(R,R)-selective variants (86% ee and 76% ee, respectively).
Further optimization up to 92% ee (S,S) and 96% ee (R,R) was
possible by ISM using a different amino acid as the single code
(phenylalanine).
evolved,
the
best
one
being
SZ724
(L74F/M78F/I80F/L114V/I116V/F139V) with 97% ee. Reversal of
enantioselectivity in favor of (R,R)-2 with 71% ee was also
achieved, but truly high levels were not detected (Table 3). When
aiming for both (S,S)- and (R,R)-selectivity, then splitting into two
smaller randomization sites each comprising five residues is the
superior strategy. Moreover, this approach allows for ISM using
Table 3. Results of applying double code saturation mutagenesis (DCSM) based on valine-phenylalanine for ten residues in one library
to LEH-catalyzed hydrolytic desymmetrization of substrate 1.
Favored
Library
Code
WT
Mutations
ee%
4
c%
99
99
99
94
81
94
56
65
97
97
96
92
98
97
98
97
98
98
enantiomer
(S,S)
(S,S)
(S,S)
(R,R)
(S,S)
(R,R)
(R,R)
(R,R)
(S,S)
(R,R)
(S,S)
(R,R)
(S,S)
(S,S)
(R,R)
(S,S)
(R,R)
(S,S)
A/T*
SZ669
SZ670
SZ672
SZ674
SZ675
SZ676
SZ677
SZ678
SZ679
SZ681
SZ682
SZ724
SZ806
SZ807
SZ808
SZ810
SZ811
L74F/M78F/I80V/L114V/I116V/F139V/L147V
L74F/L103V/L114V/I116V/F139V
L74V/M78F/I80V/L103V/I116V
L74F/I80F/L103V/L114V/I116V/F139V/L147F
L74F/M78V/I80V/L103V/L114V/L147V
L74F/L103F/L114V/F139V
88
91
16
86
19
42
49
95
38
86
40
97
91
71
82
43
83
C10-
VF
1840/52488
(~3.5%
L74F/L114F/I116V/L147F
L74F/I80F/L114V/I116V/F139V
L74V/M78F/I80V/L103V/I116V/F139V
L74F/M78V/L103V/L114V/I116V/F139V
L74F/L103V/L114F/L147F
coverage)
L74F/M78F/I80F/L114V/I116V/F139V
L74F/M78F/I80F/L114V/I116V/L147F
L74V/M78F/I80V/F139V
L114V/I116V/F139V/L147V
L74F/I80V
L74F/M78F/I80V/L114V/I116V
*A/T: Colonies picked/ computed colonies number for 95% coverage
The different strategies based on the new data originating from
the present study are summarized in Scheme S1. The results
show that DCSM is the superior strategy relative to SCSM. It is
also more efficient than the previous SCSM approach.[12b] (see
Table 4). As previously demonstrated,[12c] TCSM appears to be
even better than DCSM based on the data shown here. However,
when comparing TCSM directly with DCSM, it needs to be
pointed out that different sets of residues were selected and
different strategies for grouping single residues into larger
randomization sites were chosen in each case. The kinetic
parameters of the best mutants obtained from the DCSM strategy
were also compared with the kinetic profiles of the best variants
identified previously from application of SCSM and TCSM.[12b-c]
Significant differences in activity were not found, except that
variant SZ724 is four times as active as SZ348, both mutants
favoring (S,S)-2. In general, activity of the evolved variants
proved to be lower than that of WT LEH (Table 5).
Table 4. Comparative summary of different saturation mutagenesis strategies for the hydrolytic desymmetrization of substrate 1.
Strategy
Library
Grouping and residues included
Total
screening Best hits
Ref.
(code usage)
transformants (library
coverage)
[12b]
SCSM
SCSM
Val
L74/F75/M78/I80/L103/
L114/I116/F134/F139/L147
A5: L74/F75/M78/I80/L103
3220 (>95%)
86% (S,S)
76% (R,R)
89% (S,S)
77% (R,R)
A5, B5,
A5
B5
(Val)
A5, B5,
A5
368 (~95%)
This study
B5
A5
B5: L114/I116/F134/F139/ L147
A5: L74/F75/M78/I80/L103
DCSM
1840 (>95%)
95% (S,S)
92% (R,R)
This study
B5
A5
B5
B5: L114/I116/F134/F139/ L147
(Val/Phe)
DCSM
TCSM
Val/Phe
L74/F75/M78/I80/L103/
L114/I116/F134/F139/L147
A: I80/V83/L114/I116
B: L74/M78/L147
1840 (~3.5%)
1344 (95%)
97% (S,S)
71% (R,R)
99% (S,S)
97% (R,R)
This study
[12c]
A, B, C
A
B
C
4

ChemBioChem
10.1002/cbic.201600296
A
C
B
C: M32/L35/L103
(Val/Phe/Tyr)
Table 5. Kinetic parameters of best mutants from SCSM, DCSM and TCSM for hydrolytic desymmetrization of substrate 1.
Strategy
SCSM
DCSM
TCSM
Best
mutants
WT
Mutations
ee%
c%
Km (mM)
kcat (s−1)
kcat/Km
(s−1 M−1)
122.39
Ref.
4(S,S)
99
99
6.70±0.45
9.79±0.53
0.82±0.017
[12b]
SZ92
L74F/M78F/L103V/L114V/I116
V/F139V/L147V
L74F/M78V/I80V/L114F
92(S,S)
0.257±0.004
26.00
SZ338
SZ655
SZ724
96(R,R)
92(R,R)
83
99
4.96±0.58
34.07±1.21
0.241±0.008
0.64±0.01
49.00
18.78
M78V/I80V/L114F
This
study
L74F/M78F/I80F/L114V/I116V/
F139V
I80Y/L114V/I116V
20.32±2.64
0.51±0.02
25.09
97(S,S)
99(S,S)
98
97
[12c]
SZ348
SZ529
19.11±1.93
10.12±0.81
0.11±0.005
0.18±0.005
5.75
M32V/M78V/I80V/L114F
17.78
97(R,R)
99
Exploring substrate scope of best variants. Some of the best
hits obtained in the libraries described above were tested as
catalysts in the reaction of other substrates, beginning with
compounds 3, 5 and 7 (Scheme 2). The results are summarized
in Table 6. The most striking feature concerns the reaction of
cyclopentene oxide (3), which reacts sluggishly if at all. This
surprising phenomenon was also observed in an earlier study, in
which case a crystal structure of the respective LEH mutant
harboring 3 was obtained showing the substrate to be in a pose
not amenable to smooth ring-opening attack by activated
water.[12b] A similar explanation is likely to apply in the present
case. In the case of the cyclic seven-membered epoxide 5,
excellent (S,S)-selectivity (>92-97% ee) at good conversion is
possible, the best (R,R)-selective variant leading to 85% ee at
moderate conversion.
Scheme
2.
Further
meso-substrates
for
hydrolytic
desymmetrization catalyzed by LEH variants evolved for
substrate 1.
Table 6. Results of testing the best LEH mutants evolved for epoxide 1 as catalysts in the hydrolytic desymmetrization of further meso-
substrates.
1
3
5
7
Code
WT
ee%
%-conv.
>99
98
ee%
13(R,R)
nd
%-conv.
84
<5
<5
<5
6
ee%
%-conv.
97
ee%
%-conv.
>99
55
4(S,S)
17(S,S)
nd
93 (R,R)
>99(R,R)
>99(R,R)
>99(R,R)
94(R,R)
71(R,R)
76(R,R)
>99(R,R)
74(R,R)
75(R,R)
>99(R,R)
44(S,S)
8(R,R)
90(R,R)
90(R,R)
91(R,R)
96(S,S)
93(S,S)
94(S,S)
95(S,S)
94(S,S)
95(S,S)
92(R,R)
91(S,S)
93(S,S)
93(S,S)
90(S,S)
96(S,S)
97(S,S)
<5
27
SZ640
SZ642
SZ643
SZ646
SZ648
SZ649
SZ650
SZ651
SZ653
SZ655
SZ663
SZ664
SZ665
SZ670
SZ678
SZ724
97
nd
79(R,R)
75(R,R)
95(S,S)
91(S,S)
90(S,S)
85(S,S)
92(S,S)
93(S,S)
85(R,R)
93(S,S)
95(S,S)
94(S,S)
93(S,S)
96(S,S)
97(S,S)
41
96
nd
13
33
97
79(S,S)
81(S,S)
nd
49
13
94
8
72
74
96
<5
<5
10
10
<5
<5
<5
<5
<5
8
39
72
93
nd
25
15
94
81(S,S)
76(S,S)
nd
83
91
98
73
19
99
40
75
96
nd
74
43
99
nd
60
40
99
nd
56
6(S,S)
16
93
nd
85
23(S,S)
67(R,R)
90(R,R)
75
97
85(S,S)
85(S,S)
62
36
98
13
73
69
nd: not determined.
5

ChemBioChem
10.1002/cbic.201600296
Finally, we studied the hydrolytic kinetic resolution of styrene
oxide (rac-9) (Scheme 3). It can be seen that good (S)-selectivity
can be achieved (E = 40.2), as opposed to (R)-selectivity which
reaches a maximum of only E = 9.5 (Table 7).
-
SZ663
SZ664
SZ665
SZ670
SZ678
SZ724
nd
nd
nd
<5
-
<5
<5
-
-
11.6
-
nd
82 (S)
<5
15
nd
<5
Docking analysis for explaining the source of evolved
stereoselectivity. In order to shed some light on the possible
origin of the observed evolved stereoselectivity, we analyzed the
results of docking substrate 1 into homology models of the (R,R)-
selective mutant SZ655 (92% ee) and the (S,S)-selective mutant
SZ724 (97% ee) (Fig. 2). The mechanism of WT LEH is known to
involve substrate binding and activation by hydrogen bonding
between Asp101 and the epoxide O-atom, followed by SN2
reaction initiated by the attack of a correctly positioned and
activated water molecule.[11] In the case of the (R,R)-selective
mutant, the binding mode of 1 is expected to position C1 closest
to the activated water molecule, while the reshaped binding
pocket of the (S,S)-selective mutant should position C2 closest to
the activated water molecule. Indeed, the highest ranked docking
pose for the SZ724 mutant places C2 closest to the nucleophilic
water oxygen, C1 being further away by 0.5 Å. (Fig. 2a and Table
S1). Out of the 15 docking poses found for this mutant, the pose
shown in Figure 1a is the only one in which both the activating
hydrogen bond to D101 is present, and the epoxide ring carbon
atoms are close enough to the water molecule in order for
nucleophilic attack to occur. For the SZ655 mutant, two poses
were found in which both of these criteria were met,
corresponding to the two highest ranked poses. Both poses,
including the one displayed in Figure 2b, place the C1 atom
closest to the water oxygen (Table S1), consistent with the
experimentally observed preference for the formation of the
(R,R)-2.
Scheme 3. Hydrolytic kinetic resolution catalyzed by LEH
variants evolved for substrate 1.
Table 7. Results of testing the best LEH mutants evolved for
epoxide 1 as catalysts in the hydrolytic kinetic resolution of
styrene oxide (9).
9
E
Code
WT
eep(%)
21 (R)
c%
46
1.8
-
SZ640
SZ642
SZ643
SZ646
SZ648
SZ649
SZ650
SZ651
SZ653
SZ655
nd
nd
nd
<5
<5
<5
-
-
-
nd
92 (S)
<5
36
40.2
9.5
-
79 (R)
12
nd
91 (S)
<5
36
35.3
-
nd
<5
7
3.8
57 (S)
Figure 2. Highest ranked docking poses for cyclohexene oxide (1) in the (a) SZ724 and (b) SZ655 mutant models of LEH. The catalytic
residues are shown in white stick representation. The mutated residues are shown in green and blue for the SZ724 and SZ655 mutants,
respectively. Geometric parameters for the displayed docking poses are provided in Table S1.
Conclusions
6

ChemBioChem
10.1002/cbic.201600296
In previous studies,[12] the hydrolytic desymmetrization of
cyclohexene oxide (1) catalyzed by limonene epoxide hydrolase
(LEH) with formation of (R,R)- and (S,S)-cyclohexane-1,2-diols (2)
served as a model system for testing different directed evolution
strategies. In the present study we have extended these efforts
by applying and comparing single code saturation mutagenesis
(SCSM) with double code saturation mutagenesis (DCSM). Both
approaches focus combinatorial randomization on sites lining the
binding pocket (CASTing),[1g,3] and both proved to be successful.
However, the experimental data indicates that DCSM is the
superior approach. It should be mentioned that in directed
evolution of LEH as catalyst in the desymmetrization of 1, triple
code saturation mutagenesis (TCSM) using three appropriately
chosen amino acids is also more efficient than SCSM and
surpasses all alternative approaches.[12c]
The best (R,R)- and (S,S)-selective variants are characterized
by sequences that have not been observed in previous CAST-
based directed evolution studies of LEH,[12] but some of the point
mutations do in fact occur in earlier mutants. On the basis of
docking computations it was possible to develop simple models
for explaining the source of enhanced and inverted
stereoselectivity of the new variants. Some variants are also
viable catalysts in asymmetric reactions of several other
substrates. When focusing only on synthetic aspects, other
epoxide hydrolases should also be considered.[13] Finally, we
emphasize once more that the appropriate choice of highly
reduced amino acid alphabets for beating the numbers problem in
directed evolution of stereoselective enzymes needs to be guided
by X-ray structures, consensus data or exploratory NNK-based
saturation mutagenesis at individual CAST sites.[7,10,12] The
optimal strategy will depend upon the particular enzyme under
study.
coliBL21(DE3) to create the final library for Quick Quality Control
and screening.
Primer design and Library creation
Primer design and library construction depend upon different
grouping strategies and the particular amino acid chosen, and in
the case of LEH this involves ten residues which were divided
into two groups (Scheme 2): 1) Amplification of the short
fragments of LEH using mixed primers F1/R1 for A5, F2/R2 for
B5, F1/R1 and F2/R2 for C10, respectively; 2) using mixed
primers F1/R2 by over-lap PCR[14] to create the full fragment for
C10, using the purified fragment from step 1 as template; 3)
Amplification of the whole plasmid pET22bLEHwt[12b-c] using the
products of step 1 (A5, B5) or step 2 (C10) as megaprimers,
leading to the final variety plasmids for library generation. Primers
are listed in Tables S2-S4. The PCR products were digested by
Dpn I and transformed into electro-competent E. coli BL21(DE3)
to create the library for screening.
Screening Procedures
The same screening procedures described previously were used
to assay all libraries.[12b-c]
Homology modelling
Models of the SZ724 and SZ655 mutants of LEH were
constructed by homology modeling, using the previously
determined
crystal
structures
of
the
SZ92
(L74F/M78F/L103V/L114V/I116V/F139V/L147V)
and
SZ338
(L74F/M78V/I80V/L114F) mutants as templates, respectively.[12b]
Homology modeling was performed using the Structure Prediction
Wizard in Prime,[15] using the knowledge-based approach.[12b] The
resulting structures were refined and prepared for docking using
the Protein Preparation Wizard.[16] Hydrogen atoms were added
according to the protonation states determined by PROPKA.[17]
The protonation state of D101 was manually selected to be in the
protonated (neutral) form. The coordinates of the nucleophilic
water oxygen atom were copied from the WT LEH crystal
structure (PDB 1NU3)[11b] following structural alignment with the
mutant structures, ensuring that the known activating hydrogen
bonds to Y53, N55 and D132 were formed. The positions of all
hydrogen atoms were energy minimized using the Impact
program[18] and the OPLS2005 forcefield.
Experimental Section
Materials
KOD Hot Start DNA Polymerase was obtained from Novagen.
Restriction enzyme Dpn
I
was bought from NEB. The
oligonucleotides were synthesized by Life Technologies. Plasmid
preparation kit was ordered from Zymo Research, and PCR gel
extraction kit was bought from QIAGEN. DNA sequencing was
conducted by GATC Biotech. All commercial chemicals were
purchased from Sigma-Aldrich, Tokyo Chemical Industry (TCI) or
Alfa Aesar. Lysozyme and DNase I were purchased from
AppliChem.
Docking of cyclohexene oxide.
The structure of cyclohexene oxide was built in the Maestro
program and prepared for docking using LigPrep.[19] Docking was
performed to the prepared models of the SZ724 and SZ655
mutants using Glide[20] with standard precision (SP) settings. A
maximum of ten docking poses was requested for each model. A
total of 15 and 9 docking poses was obtained for the SZ724 and
SZ655 models, respectively. Activation of the epoxide ring
oxygen by D101 is a prerequisite for ring opening in LEH;
therefore, only the docking poses that contained this interaction
were considered during analysis.
PCR based methods for library construction
Libraries were constructed using the Over-lap PCR and
megaprimer approach with KOD Hot Start polymerase. 50 µL
reaction mixtures typically contained 30 µL water, 5 µL KOD hot
start polymerase buffer (10×), 3 µL 25 mM MgSO4, 5 µL 2 mM
dNTPs, 2.5 µL DMSO, 0.5 µL (50~100 ng) template DNA, 100 µM
primers mix 0.5 µL each and 1 µL KOD hot start polymerase. The
PCR conditions for short fragment: 95 °C 3 min, (95 °C 30 sec, 56
°C 30 sec, 68 °C 40 sec) × 32 cycles, 68 °C 120 sec, 16 °C 30
min. For mega-PCR: 95 °C 3 min, (95 °C 30 sec, 60 °C 30 sec,
68 °C 5 min 30 sec) × 24 cycles, 68 °C 10 min, 16 °C 30 min. The
PCR products were analyzed on agarose gel by electrophoresis
and purified using a Qiagen PCR gel extraction kit. 2 µL NEB
CutSmart™ Buffer and 2 µL Dpn I were added in 50 µL PCR
reaction mixture and the digestion was carried out at 37 °C for
more than 3 h. After Dpn I digestion, the PCR products (1 µL)
Acknowledgements
Support from the Max-Planck-Society and the LOEWE Research
cluster SynChemBio is gratefully acknowledged.
Keywords: reduced amino acid alphabet • directed evolution •
epoxide hydrolases • saturation mutagenesis • stereoselectivity
were
directly
transformed
into
electro-competent
E.
7

ChemBioChem
10.1002/cbic.201600296
[1]
Recent reviews of directed evolution: a) A. S. Bommarius, Annu. Rev.
Chem. Biomol. Eng. 2015, 6, 319-345; b) A. Currin, N. Swainston, P. J.
Day, D. B. Kell, Chem. Soc. Rev. 2015, 44, 1172-1239; c) C. A. Denard,
H. Ren, H. Zhao, Curr. Opin. Chem. Biol. 2015, 25, 55-64; d) E. M. J.
Gillam, J. N. Copp, D. F. Ackerley, in Methods in Molecular Biology., Vol
1179, Humana Press, Totowa, NJ, 2014; e) M. Goldsmith, D. S. Tawfik,
Methods Enzymol. 2013, 523, 257-283; f) E. M. Brustad, F. H. Arnold,
Curr. Opin. Chem. Biol. 2011, 15, 201-210; g) M. T. Reetz, Angew.
Chem. Int. Ed. 2011, 50, 138-174; h) C. Jäckel, D. Hilvert, Curr. Opin.
Biotechnol. 2010, 21, 753-759; i) N. J. Turner, Nat. Chem. Biol. 2009, 5,
568-574; j) S. Lutz, U. T. Bornscheuer, Protein Engineering Handbook,
Wiley−VCH, Weinheim, 2009.
[2]
Recent reviews of saturation mutagenesis: a) E. Williams, J. N. Copp, D.
F. Ackerley, in Directed Evolution Library Creation, (Eds.: E. M. J. Gillam,
J. N. Copp, F. Ackerley), Methods in Molecular Biology, Vol. 1179,
Humana Press, Totowa, NJ, 2014, pp. 83-101; b) R. M. P. Siloto, R. J.
Weselake, Biocatal. Agric. Biotechnol. 2012, 1, 181-189.
[3]
[4]
M. T. Reetz, M. Bocola, J. D. Carballeira, D. Zha, A. Vogel, Angew. Chem.
2005, 117, 4264-4268; Angew. Chem. Int. Ed. 2005, 44, 4192-4196.
a) M. T. Reetz, L.-W. Wang, in part M. Bocola, Angew. Chem. 2006, 118,
1258-1263; Erratum, 2556; Angew. Chem. Int. Ed. 2006, 45, 1236-1241;
Erratum, 2494; b) C. G. Acevedo-Rocha, S. Kille, M. T. Reetz, In
Directed Evolution Library Creation: Methods and Protocols, 2nd Edition,
Vol 1179 (Eds.: D. Ackerley, J. Copp, E. Gillam) Methods in Molecular
Biology, Humana Press, Totowa, 2014, pp 103-128.
[5]
[6]
a) A. R. Davidson, K. J. Lumb, R. T. Sauer, Nat. Struct. Biol. 1995, 2,
856-864; b) K. U. Walter, K. Vamvaca, D. Hilvert, J. Biol. Chem. 2005,
280, 37742-37746.
a) F. A. Fellouse, B. Li, D. M. Compaan, A. A. Peden, S. G. Hymowitz, S.
S. Sidhu, J. Mol. Biol. 2005, 348, 1153-1162; b) A. Koide, R. N. Gilbreth,
K. Esaki, V. Tereshko, S. Koide, Proc. Natl. Acad. Sci. U S A. 2007, 104,
6632-6637.
[7]
[8]
[9]
M. T. Reetz, J. D. Carballeira, Nat. Protoc. 2007, 2, 891-903.
W. M. Patrick, A. E. Firth, Biomol. Eng. 2005, 22, 105−112.
Y. Nov, Appl. Environ. Microbiol. 2012, 78, 258−262.
[10] Z. Sun, Y. Wikmark, J.-E. Bäckvall, M. T. Reetz, Chem. Eur. J. 2016, 22,
5046-5054.
[11]
a) M. J. van der Werf, K. M. Overkamp, J. A. M. de Bont, J. Bacteriol.
1998, 180, 5052–5057; b) M. Arand, B. M. Hallberg, J. Zou, T. Bergfors,
F. Oesch, M. J. van der Werf, J. A. M. de Bont, T. A. Jones, S. L.
Mowbray, EMBO J. 2003, 22, 2583–2592.
[12] a) H. Zheng, M. T. Reetz, J. Am. Chem. Soc. 2010, 132, 15744-15751; b)
Z. Sun, R. Lonsdale, X.-D. Kong, J.-H. Xu, J. Zhou, M. T. Reetz, Angew.
Chem. 2015, 127, 12587-12592. Angew. Chem. Int. Ed. 2015, 54,
12410-12415; c) Z. Sun, R. Lonsdale, L. Wu, G. Li, A. Li, J. Wang, J.
Zhou, M. T. Reetz, ACS Catal. 2016, 6, 1590-1597.
[13] Reviews of epoxide hydrolases: a) M. Kotik, A. Archelas, R. Wohlgemuth,
Curr. Org. Chem. 2012, 16, 451-482; b) J. H. L. Spelberg, E. J. de Vries,
In Enzyme Catalysis in Organic Synthesis, 3rd Edition. (Eds.: K. Drauz, H.
Gröger, O. May) Wiley-VCH, Weinheim, 2012, pp 363-416; c) S. Hwang,
C. Y. Choi, E. Y. Lee, J. Ind. Eng. Chem. 2010, 16, 1-6; d) M. Widersten,
A. Gurell, D. Lindberg, Biochim. Biophys. Acta. 2010, 1800, 316–326.
[14] A. V. Bryksin, I. Matsumura, Biotechniques. 2010, 48, 463-465.
[15] Prime, version 3.9, Schrödinger, LLC, New York, NY, 2015.
[16] Schrödinger Suite 2015-1 Protein Preparation Wizard; Epik version 3.1,
Schrödinger, LLC, New York, NY, 2015; Impact version 6.6, Schrödinger,
LLC, New York, NY, 2015; Prime version 3.9, Schrödinger, LLC, New
York, NY, 2015.
[17]
M. H. M. Olsson, C. R. Søndergaard, M. Rostkowski, J. H. Jensen, J.
Chem. Theory Comput. 2011, 7, 525.
[18] Impact, version 6.5, Schrödinger, LLC, New York, NY, 2014.
[19] LigPrep, version 3.2, Schrödinger, LLC, New York, NY, 2014.
[20] Glide, version 6.5; Schrödinger, LLC: New York, NY, 2014.
8

ChemBioChem
10.1002/cbic.201600296
FULL PAPERS
Two is better than one: Double code
Z. Sun, R. Lonsdale, G. Li, M. T. Reetz*
saturation
mutagenesis
(DCSM)
based on two amino acids as building
blocks is more efficient than single
code saturation mutagenesis (SCSM)
in the directed evolution of enzyme
stereoselectivity, as demonstrated by
the hydrolytic desymmetrization of
cyclohexene oxide catalyzed by
limonene epoxide hydrolase.
Page No. 1– Page No. 8
Comparing Different Strategies in
Directed Evolution of
Stereoselectivity: Single
Double Code Saturation Mutagenesis
Enzyme
versus
9