Loop Swapping as a Potent Approach to Increase Ene Reductase Activity with Nicotinamide Adenine Dinucleotide (NADH)

Article abstract of DOI:10.1002/adsc.201900073

The asymmetric reduction of alkenes is a widely used transformation in industry. Ene reductases (ERs) are (βα)₈-barrel folded enzymes capable of catalyzing this hydrogenation reaction. At the expense of nicotinamide coenzymes, ERs can reduce a wide range of electron-deficient alkenes in an anti-specific manner and with high regio- and stereoselectivities. However, a cost-effective industrial use of these enzymes is hampered, since most ERs prefer nicotinamide adenine dinucleotide phosphate (NADPH) to the more stable and less expensive non-phosphorylated nicotinamide adenine dinucleotide (NADH) as coenzyme. Here, we demonstrate an approach to both modify the biocatalysts coenzyme selectivity and strongly increase the activity and affinity with NADH. By swapping loop regions of the cyanobacterial NostocER1 for the corresponding regions of two NADH-favoring ERs, a strong alteration of the biocatalyst's coenzyme binding was achieved. This made possible a transfer of the respective donor-ER kinetic parameters to NostocER1. Additionally, outperformance of both donors in terms of activity was achieved through combinatorial swapping of loops of both species. These findings demonstrate the high potential of loop swapping as protein engineering approach to selectively optimize the coenzyme binding of ERs. (Figure presented.).

Full text of DOI:10.1002/adsc.201900073

Title: Loop Swapping as a Potent Approach to Increase Ene
Reductase Activity with Nicotinamide Adenine Dinucleotide
(NADH)
Authors: Christoph Mähler, Franziska Kratzl, Melina Vogel, Stefan
Vinnenberg, Dirk Weuster-Botz, and Kathrin Castiglione
This manuscript has been accepted after peer review and appears as an
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 published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Adv. Synth. Catal. 10.1002/adsc.201900073
Link to VoR: http://dx.doi.org/10.1002/adsc.201900073

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
1
2
3
Loop Swapping as a Potent Approach to Increase
Ene Reductase Activity with Nicotinamide Adenine
Dinucleotide (NADH)
4
5
Christoph Mähler,a Franziska Kratzl,a Melina Vogel,a Stefan Vinnenberg,a
Dirk Weuster-Botz,a and Kathrin Castiglioneb,*
6
7
8
a Technical University of Munich, Institute of Biochemical Engineering, Boltzmannstr.
15, D-85748 Garching, Germany
9
b Friedrich-Alexander-University Erlangen-Nürnberg, Institute of Bioprocess
Engineering, Paul-Gordan-Str. 3, D-91052 Erlangen, Germany
10
11
12
13
14
15
16
17
Key words: Alkene reduction, Biocatalysis, Cofactors, Ene reductase, Loop
swapping, Protein engineering
18
19
20
21
22
* Corresponding author:
Telephone: +49 9131 85 23003
Fax: +49 9131 85 23002
Email: kathrin.castiglione@fau.de
Homepage: https://www.bvt.tf.fau.de
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
2
Abstract
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
The asymmetric reduction of alkenes is a widely used transformation in industry. Ene
reductases (ERs) are (βα)8-barrel folded enzymes capable of catalyzing this hydrogenation
reaction. At the expense of nicotinamide coenzymes, ERs can reduce a wide range of
electron-deficient alkenes in an anti-specific manner and with high regio- and
stereoselectivities. However, a cost-effective industrial use of these enzymes is hampered,
since most ERs prefer nicotinamide adenine dinucleotide phosphate (NADPH) to the more
stable and less expensive non-phosphorylated nicotinamide adenine dinucleotide (NADH) as
coenzyme. Here, we demonstrate an approach to both modify the biocatalysts coenzyme
selectivity and strongly increase the activity and affinity with NADH. By swapping loop
regions of the cyanobacterial NostocER1 for the corresponding regions of two NADH-
favoring ERs, a strong alteration of the biocatalyst’s coenzyme binding was achieved. This
made possible a transfer of the respective donor-ER kinetic parameters to NostocER1.
Additionally, outperformance of both donors in terms of activity was achieved through
combinatorial swapping of loops of both species. These findings demonstrate the high
potential of loop swapping as protein engineering approach to selectively optimize the
coenzyme binding of ERs.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
3
Introduction
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
The industrial demand for enantiopure products is steadily increasing.[1] The asymmetric
hydrogenation of alkenes represents one way to synthesize pure enantiomers or
diastereomers. Whereas syn-hydrogenations are mostly accomplished with expensive chiral
metal complexes,[2] anti-hydrogenations can be achieved using biocatalysts, i.e. ene
reductases (ERs). ERs are of potential interest for industrial processes not only due to their
high regio-, stereo- and enantioselectivity, but also because of a broad substrate
spectrum.[3,4] The substrates are alkenes bearing an electron-withdrawing group (EWG) such
as α, β-unsaturated ketones, aldehydes, carboxylic acids, esters, imides, nitriles, and
nitroalkenes.[4] In this context the most investigated class of ERs are flavin-dependent
oxidoreductases of the old yellow enzyme family (OYEs, EC 1.6.99.1). The reduction
proceeds via a ping-pong bi-bi mechanism, thus the catalytic cycle of ERs is separated in
two half reactions: a reductive and an oxidative half reaction (Scheme 1).[5,6] The reductive
half reaction comprises a hydride transfer from NAD(P)H to the enzyme-bound prosthetic
group, flavin mononucleotide (FMN), and the release of oxidized NAD(P)+.[7] During the
oxidative half reaction this hydride reduces the electron deficient double bond in an anti-
specific manner creating up to two stereogenic centers.[6] All currently known members of the
OYE family have essentially the same tertiary structure, an (βα)8-barrel structure or so-called
triose phosphate isomerase (TIM)-barrel fold.[4] This means that, eight alternating α-helices
and parallel β-sheets form a barrel-like structure, in which the sheets shape the inner wall
and the helices the outer wall of the barrel. The enzymes’ active site, comprising the FMN-
binding site, is located on top of this barrel, formed by the flexible loop regions between the
C-terminal ends of the β-sheets and the N-terminal ends of the α-helices.[4,8] These loops, as
well as the corresponding secondary structure elements, are numbered consecutively
beginning with 1 at the N-terminus and ending with 8 at the C-terminus.
63
64
65
Although many wild type ERs possess desirable properties like a broad substrate spectrum
and high stereoselectivities, protein engineering has been successfully applied to change
unwanted characteristics of these biocatalysts.[9] For instance, a small active site limiting the
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
4
Scheme 1. The catalytic cycle of OYEs is separated in two half reactions. The reductive half reaction
comprises a hydrid transfer from NAD(P)H to the enzyme-bound FMN, generating a reduced FMN-H2.
During the oxidative half reaction FMN-H2 reduces the alkene bearing an electron-withdrawing group
(EWG), transferring the OYE back to the starting point carrying an oxidized FMN.
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
substrate size was altered through changes of the active site entrance,[10] or rational
mutagenesis was applied to alter the enzymes stereoselectivity.[11] Another example is the
modulation of the substrate scope through exchanges of loop regions.[12] However, the
industrial use of ERs remains hampered since the majority of OYEs prefer NADPH to the
less expensive and more stable NADH as coenzyme.[4,8,13] Additionally, the variety of NADH-
regenerating enzymes is higher than for its phosphorylated derivative,[14] facilitating the
application of NADH-preferring biocatalysts in industrial bioprocesses.[15] Since an undesired
coenzyme selectivity is a common phenomenon among oxidoreductases, coenzyme
engineering has emerged as an important area of protein engineering.[14,16] All approaches to
alter enzymatic coenzyme binding can be categorized into three major groups: random,
rational and semi-rational.[14] The application of knowledge-free random approaches like
error-prone PCR or gene shuffling are rarely used for changing coenzyme binding since
beneficial changes are nearly always limited to the coenzyme binding site[17] and screening of
large libraries is labor, time and material intensive.[14] Rational approaches mostly focus on
the change of specific residues responsible for coenzyme binding.[18] This requires a deep
understanding of coenzyme binding, which is in most cases neither easily determinable nor
transferable through structural diversity within an enzyme family, let alone a common protein
fold.[19] Driven by computational tools and algorithms, semi-rational approaches have
become increasingly relevant for coenzyme engineering.[14,16] Through different techniques,
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
5
85
86
87
88
89
90
91
92
potential hot spots for coenzyme engineering can be identified and used to create ‘small but
smart’ libraries.[20] In this context, a promising, easy-to-use web tool called Cofactor
Specificity Reversal-Structural Analysis and Library Design (CSR-SALAD) was established
and successfully applied.[17,21] CSR-SALAD detects and classifies residues that determine
cofactor specificity and designs degenerated codon libraries. However, a reliable and
detailed knowledge about the interactions of coenzyme and protein is inevitable for the
application of these computational tools. Furthermore, natural evolution phenomena like
insertions or deletions are not considered.[16]
93
94
Another semi-rational approach to influence coenzyme binding that requires less pre-
knowledge and includes natural evolution is the modulation of whole protein parts. This can
span from swapping of short loops[22] up to entire binding pockets.[14] In this context, probably
the best-known example is the adjustment of the nucleoside ribose 2’ binding-site of
dehydrogenases possessing a Rossmann-fold. Thus, by exchanging two loops it was
possible to transfer the preference for NADP(H) from an α-keto acid reductase from
Sphingomonas sp. A1 named A1-R to the homologue A1-R’.[23] Similar results could be
achieved through exchange of one loop of a malate dehydrogenase[24] and an
isopropylmalate dehydrogenase.[25] Loop swapping was also applied to engineer a NADP-
dependent Rossmann-folded isocitrate dehydrogenase to NAD-dependency.[26] However, the
exchange of entire loops to influence coenzyme binding in other protein folds, particularly in
the widespread (βα)8-barrel fold, was rarely applied so far.[16] This is surprising since loop
swapping among enzymes possessing this fold is seen highly promising.[27] Additionally, it
was successfully applied to transfer substrate specifities,[12,28,29] whereby partly an
unexpected modification of coenzyme binding occurred.[12,29] For these reasons, we
demonstrate in this work the high potential of loop swapping as engineering strategy to
selectively control coenzyme binding of (βα)8-barrel folded ERs.
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
6
Results and Discussion
110
111
112
113
114
115
116
117
The aim of the work was to influence the coenzyme preference as well as to increase the
activity and affinity of an OYE towards NADH. This should be accomplished with a low
screening effort. The cyanobacterial ene reductase 1 from Nostoc sp. PCC7120
(NostocER1) was used as scaffold. This ‘acceptor’ enzyme is characterized by a very high
stereoselectivity with a broad substrate spectrum. Furthermore, NostocER1 exhibits high
activity with its natural coenzyme NADPH, revealing the enzyme’s fast oxidative half
reaction.[30] However, like most OYEs, NostocER1 has a strong preference towards NADPH,
as indicated by a more than 20-fold higher catalytic efficiency (keff) compared to NADH.
Figure 1. A Homology model of NostocER1 (green) based on morphinone reductase crystal structure
(PDB: 2r14). The coenzymes FMN (yellow) and NADH (blue) are depicted bound in the active site, as well as all
loop regions (red) that have been exchanged. B Sequence alignment of the exchanged loop regions.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
7
118
Swapping of single loop regions
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
In order to alter the biocatalyst’s dependency on the phosphorylated coenzyme, we identified
and systematically exchanged all loop regions that might have a direct interaction with the
coenzyme (Figure 1A). Due to their size, Loop 2 and 3 have been divided into two possible
contact areas (a and b), and the contact area after Loop 8 between Helix 8 and the C–
terminus is labeled Loop C. Loops 7 and 8, as well as Loop 4, have not been exchanged
because of their presumed long distance to the coenzyme binding site or high primary
structure identity, respectively. The eight dedicated protein regions have been exchanged for
the corresponding regions originating from two different ‘donor’ enzymes. On the one hand,
the cyanobacterial ene reductase 1 from Acaryochloris marina (AcaryoER1) was used due to
its high activity with NADH.[30] On the other hand, the more distantly related proteobacterial
Old Yellow Enzyme 4 from Achromobacter sp. JA81 (AchrOYE4) was employed because of
its high affinity towards NADH.[31] This resulted in various modifications of the loops, ranging
from single point mutations to the exchange of up to 17 amino acids, including insertions and
deletions (Figure 1B). Thereby, our goal was to swap these NADH-binding properties without
transferring undesired donor properties, e.g. a slow substrate reduction during the oxidative
half reaction.
135
136
137
138
139
140
141
142
143
144
145
Figure 2 shows the specific activities using 200 µM NADH (v200) and 10 mM maleimide as
substrate of the eight NostocER1-AcaryoER1 hybrid enzymes (blue columns) and the eight
NostocER1-AchrOYE4 hybrid enzymes (red columns) in comparison to the wild type
enzymes (WTs). The activities with 200 µM NADH were chosen for an initial comparison,
since most industrial biotransformations are operated with catalytical amounts of coenzyme
in order to reduce costs.[32] Already five out of sixteen of these single loop exchanges
showed the intended increased activity compared to NostocER1. The exchange of Loop 2a
and 3b with the corresponding regions of AcaryoER1 led to a 2.2- or 1.5-fold increased
activity. The exchange of the NostocER1 Loop 1 with the complementary region of
AchrOYE4 resulted in a remarkable 6.6-fold increased activity of 14.5 ± 1.5 U mg-1. This
hybrid enzyme was the only one of these single loop swaps outperforming its ‘donor’ enzyme
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
8
Figure 2. Specific activities using 200 µM NADH (v200) of the ‘acceptor’ enzyme NostocER1 (white), the ‘donor’
enzymes AcaryoER1 and AchrOYE4 (gray) and the corresponding hybrid enzymes of NostocER1 and
AcaryoER1 loop regions (blue, A) or NostocER1 and AchrOYE4 (red, B). The values are means of at least two
biological replicates (m) with five technical replicates each. Divergent biological replications were applied for
NostocER1 (m = 6), the NostocER1-AcaryoER1 Loop swaps 1 (m = 3), 2a (m = 4), 5 and 8 (m = 3), AcaryoER1
(m = 4), the AchrOYE4 Loop swap 1 (m = 5), and AchrOYE4 (m = 5).
146
147
AchrOYE4 in terms of activity under these conditions. Furthermore, the alteration of Loop 2b
enhanced the v200 1.9-fold, the one of Loop 5 1.4-fold.
148
149
150
151
152
153
154
155
156
157
Interestingly, the insertion of AcaryoER1 and AchrOYE4 loops didn’t reveal a common
NostocER1 ‘target loop’ to augment its NADH activity. Four of the five identified loop
changes that increase the activity of NostocER1, i.e. Loop 1, 2a, 2b, 3b, are positioned close
to the adenine-moiety of the coenzyme. This is hardly surprising, since the phosphate group
(which is the only feature that distinguishes NADPH from NADH) is located close to these
loops, making them to a common target for coenzyme engineering.[17,21] Moreover, a larger
number of loops close to this area has been swapped, thus increasing the probability of a
positive change. Nonetheless, the exchange of Loop 5 represents one target area close to
the nicotinamide moiety which is spatially more distant to the adenine moiety. By applying a
more rational method, this target area likely could have been lost.
158
159
160
In addition, the v200 values with NADPH were determined (Supporting Information,
Figure S2). All hybrids with an increased NADH activity showed either no significant change
or only a slightly decreased NADPH activity compared to NostocER1, with the exception of
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
9
Table 1. Kinetic parameters catalytic constant (kcat), Michaelis-Menten constant (Km) and the catalytic efficiency
(keff) of the wild type enzymes and hybrid enzymes showing an increased specific activity with NADH. Additionally
shown is the relative catalytic efficiency (RCE), which describes the ratio of the enzyme’s keff with NADH and
NostocER1’s keff with NADPH. The kinetic parameters are the result of at least one nonlinear regression.
Divergent biological replication was applied for the WTs NostocER1 (m = 3), AcaryoER1 (m = 2), AchrOYE4
(m = 2), the ArchOYE4 Loop 1 exchange (m = 2) and the Loop 1,5 exchange (m = 2). The depicted means and
standard deviations were calculated according to equation 1 and 3.
161
the already mentioned Loop 5 swap. This hybrid has a strongly decreased NADPH activity,
represented by a 13.6-fold lower v200 compared to NostocER1. Thus, the loop seems to be
associated with the low NADPH-activity of AchrOYE4 (v200 of 0.07 ± 0.01 U mg-1) and the
enzymes distinct coenzyme specificity.
162
163
164
165
166
167
168
169
The improved enzyme variants were characterized in detail. The catalytic constants (kcat) and
Michaelis-Menten constants (Km) of the five identified hybrids with an increased NADH
activity and of the three wild type enzymes are listed in Table 1. A graphical comparison of
these values is shown in the Supporting Information, Figure S3. In Figure S4, exemplary
plots of the data used for the parameter estimation by non-linear regression analysis are
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
10
170
171
172
173
174
175
176
shown. It can be seen that the two NostocER1-AcaryoER1 hybrid enzymes only have a
slightly higher affinity towards NADH than NostocER1, with Km values around 800 µM. A
comparable picture emerges from the NostocER1-AchrOYE4 hybridizations Loop 2b and 5,
showing a reduced Km of around 600 µM. The Loop 1 alteration once again reveals the
biggest improvement, reducing the Km by a factor of more than 5 from 1050 ± 220 µM to
190 ± 14 µM. However, none of the single swaps could transfer the corresponding wildtype
Km to NostocER1.
177
178
179
180
181
182
183
184
To exclude negative influences caused by the loop swaps, expression levels and thermal
stabilities of the hybrids were checked. All hybrids with an increased NADH activity could be
isolated in quantities ranging from 21.3 to 67.4 mg Lculture-1. These values are comparable to
the results obtained for the wildtype since nine independent expressions and purifications of
NostocER1 resulted in an average isolated protein amount of 40.8 ± 11.2 mg per liter culture.
Additionally, the hybrids with the exchanged Loops 1, 2b and 5 were randomly chosen for a
comparative analysis of the thermal stability at 40 °C. These experiments revealed no
decrease in thermal stability compared to the NostocER1 wildtype.
185
Swapping of multiple loop regions
186
187
188
189
190
191
192
193
194
195
196
Combinations of the five identified loops were applied in order to evaluate whether a further
increase in activity with or affinity for NADH is accessible. Therefore, separate combinations
of both ‘donor’ enzymes were conducted initially, leading to one NostocER1-AcaryoER1 and
four NostocER1-AchrOYE4 hybridizations. As a first comparison, the v200 values of these
mutants were compared to the values of the wildtypes (Supporting Information, Figure S5A).
Three out of five multiple loop swapped enzymes showed a specific activity comparable to
the one of the ‘donors’. In particular, the combination of the two AcaryoER1 Loops 2a and 3b
led to a further positive effect, resulting in a 5.5-fold higher specific activity of
11.9 ± 0.7 U mg-1. This is also reflected by its kinetic parameters. The hybrid has a similar kcat
(19.2 ± 0.4 s-1) and a slightly decreased Km (237 ± 17 µM) compared to the ‘donor’
AcaryoER1 (Table 1). Although the four combinations of the exchanged AchrOYE4 Loops 1,
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
11
197
198
199
200
201
202
2b and 5 resulted in no further increase of activity, a remarkable result could be achieved in
terms of the Km (Table 1 and Supporting Figure S6). The Km of the NostocER1-AchrOYE4
Loop 1,5 hybridization was further decreased to 73 ± 13 µM, which means that no significant
difference between the Km of the hybrid enzyme and the excellent Km of its ‘donor’ AchrOYE4
is observable. Consequently, the swapping of two loops of each of the ‘donors’ enabled the
transfer of both kinetic parameters kcat and Km to the ‘acceptor’ enzyme NostocER1.
203
204
205
206
207
208
209
210
211
212
213
214
Furthermore, we combined the most promising loops, i.e. AcaryoER1 Loop 2a and 3b as well
as AchrOYE4 Loop 1 and 5, to determine whether outperforming of both donors in terms of
activity and affinity is possible. Thus, five additional multiple loop swaps were generated.
Remarkably, all combinations with this set of loops showed at least a similar kcat and a by a
factor of 2.5 reduced Km compared to NostocER1, indicating the excellent combinability of
the determined protein loops (Table 1 and Supporting Figure S7). One loop combination with
a strongly increased activity could be identified, namely AcaryoER1 Loop 2a merged with
AchrOYE4 Loop 1. The Loop 1,2a-hybrid even resulted in the highest detected activity with a
kcat of 29.1 ± 0.4 s-1, which corresponds to a 1.9-fold increased value compared to the
‘acceptor’ enzyme NostocER1. The Michaelis-Menten constants of these hybrid enzymes
containing loops of both donors were between 161 µM and 430 µM, which means that no
further improved Km could be detected.
215
216
217
218
219
220
221
222
The v200 values with NADPH of all multiple loop swaps are shown in Supporting Information,
Figure S8. A result comparable to the single loop swaps was obtained. The hybrids v200
values were in the region of the v200 of NostocER1, ranging between a maximal increase of
50 % and decrease of 30 % in comparison to the wildtype. The sole exceptions are again the
hybrids comprising the Loop 5 originating from AchrOYE4. These mutants showed a
drastically reduced activity of up to 90 %. This is a further indication for the low NADPH
acceptance associated with this loop. Thus, besides increasing the affinity towards NADH,
AchrOYE4 Loop 5 can additionally be used to strongly increase the selectivity towards
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
12
223
224
NADH. Depending on the application, it can be desirable that the biocatalyst uses only one
coenzyme.[16]
225
Valuation of NostocER1 engineering
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
As additional comparison between the ‘best performers’ and the wild type enzymes, the
catalytic efficiency (keff) for NADH was calculated (Table 1). The keff of the best single swap
(AchrOYE4 Loop 1) was already more than 6-fold higher than the one of NostocER1. The
efficiencies of the best multiple loop swaps, i.e. the various combinations of AchrOYE4
Loop 1, Loop 5 and AcaryoER1 Loop 2a, were 8- to 9-fold higher. Six hybrid enzymes could
be created that had a higher keff than the cyanobacterial ‘donor’ AcaryoER1 and three of the
multiple loop exchanges showed no significant difference to the proteobacterial ‘donor’
AchrOYE4. To assess the engineering success, the kinetic parameters of NostocER1 using
its preferred coenzyme NADPH are additionally listed. With a kcat of 29.1 ± 0.4 s-1, the
Loop 1,2a hybrid outperformed the wildtype’s NADPH catalytic constant (26.7 ± 0.9 s-1) and
the Loop 1,5 hybrid yielded in a similar Km (73 ± 13 µM) compared to the one of NostocER1
with NADPH (80 ± 19 µM). As a further comparison, the relative catalytic efficiency (RCE),
i.e. the ratio of the enzyme’s keff with NADH and the NostocER1 WT with NADPH, was used.
Three multiple loop swaps with the highest determined keff values between 118 ± 8.7 s-1 µM-1
and 130 ± 6.7 s-1 µM-1 yielded an RCE from 0.36 to 0.39. Up to now, there have been 51
oxidoreductase mutants with an altered coenzyme specificity towards NAD(H) described in
the literature. Only 11 of them possess a higher RCE.[16] Furthermore, 9 out of these 11
engineered enzymes have a Rossmann-like cosubstrate binding domain, for which
cosubstrate engineering is already well-understood. On the contrary, successful changes of
coenzyme specificity of enzymes belonging to EC 1.6 are very rare and coenzyme
engineering of oxidoreductases possessing a prosthetic group, e.g. a flavo-group, tend to be
more challenging.[16] These facts underpin the high potential of the presented optimization
approach.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
13
Figure 3. Specific activities with 500 µM NADH (v500) and 10 mM (R)-Carvone of the NostocER1 (white), the
‘donor’ enzymes AcaryoER1 and AchrOYE4 (gray), the hybrid enzymes of NostocER1 with AcaryoER1 loop
regions (blue), NostocER1 with AchrOYE4 loop regions (red) and NostocER1 with loop regions of both ‘donors’
(blue/red). Depicted in addition is the specific activity of NostocER1 with 500 µM NADPH and 10 mM (R)-carvone
(white/black). The v500 values are means of three technical replicates. Additional biological replication was applied
for NostocER1 (m = 3 (NADH) / m = 2 (NADPH)), AcaryoER1 (m = 2), and the Loop combination 1,2a (m = 3).
249
Enhanced dihydrocarvone synthesis
250
251
252
253
254
255
256
257
258
259
260
261
Finally, we evaluated the specific activities of the reduction of the substrate (R)-carvone to
(2R,5R)-dihydrocarvone (Supporting Information, Figure S9A). Dihydrocarvone has industrial
relevance due to its use in the synthesis of natural products, e.g. tetraoxane derivatives as
compounds with antimalarial activity,[33] or keto decalin derivatives as insect antifeedants.[34]
Due to the higher detection limits of the gaschromatographic analysis, an assay with a higher
NADH concentration (500 µM, v500) was applied. Like a variety of organic substrates,[35] the
water solubility of (R)-carvone is low.[36] Thus, the applied substrate concentration (10 mM)
was close to the applicable maximum. Furthermore, in order to reduce toxicity on the
biocatalyst and to overcome these low substrate solubilities, biocatalytic transformations on a
laboratory or industrial scale are often conducted in aqueous biphasic systems.[37] Therefore,
only catalytic activities using low substrate concentrations are of interest. The specific
activities for the WT enzymes and all hybrids with an increased NADH activity are shown in
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
14
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
Figure 3. All identified hybrid enzymes possessing an increased activity with the coenzyme
showed a faster conversion of (R)-carvone using NADH. Similar to the previous results, the
biggest increase of activity was achieved by the enzymes possessing the highest keff values
with NADH, i.e. the swapping of AchrOYE4 Loop 1, and additional combinations with
AchrOYE4 Loop 5 and/or AcaryoER1 Loop 2a. Interestingly, one additional hybrid with a
lower keff possessed an activity comparable to these ‘best performers’, the combination of
AcaryoER1 Loop 3b with AchrOYE4 Loop 1. These five enzymes had an activity between
2.27 ± 0.03 U mg-1 (Loop 1 swap) and 2.59 ± 0.08 U mg-1 (Loop 1,2a swap), which
represented an up to 3.4-fold increased activity compared to NostocER1 (0.77 ± 0.06
U mg-1). Special emphasis is placed on comparing the activity of these five hybrid enzymes
and the NostocER1 activity using 500 µM NADPH. The hybrids only possess a marginal
lower or not significantly different specific activity, revealing a shift of the rate-limiting step
from the reductive half reaction to the oxidative half reaction. The fact that the extremely low
activity of AchrOYE4 with (R)-carvone (0.02 ± 0.01 U mg-1) was not transferred to any of the
hybrid enzymes comprising AchrOYE4 loops deserves additional highlighting. Furthermore,
no negative influence on NostocER1’s excellent stereoselectivity could be observed. None of
the hybrids showed a significant increase in (2S,5R)-dihydrocarvone byproduct formation,
represented by a preservation of the wildtypes desirable diastereomeric ratio of 99.4 ± 0.5 %
(Supporting Information, Figure S9B). The fact that none of the hybrids revealed a change in
(R)-carvone binding points towards a preservation of NostocER1’s excellent substrate
spectrum and stereoselectivity. However, since the main focus of this work was on
engineering the reductive half-reaction, a detailed characterization of the oxidative half
reaction of the loop swapped hybrids remains part of further studies.
Conclusions
285
286
287
We could demonstrate that the systematic exchange of loop regions involved in coenzyme
binding is a potent approach to alter coenzyme binding. Thus, we could strongly increase the
activity and affinity with NADH of a cyanobacterial ER. By using a homology model, we
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
15
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
determined eight target loop regions of NostocER1 that might interact with the coenzyme and
swapped these with the corresponding regions of two donor ERs possessing a high activity
with NADH (AcaryoER1) or a high affinity (AchrOYE4) towards NADH. We could identify five
engineered ERs showing the desired higher activity with and affinity towards the coenzyme
through the generation of only 16 loop swapped proteins. These five hybrids included one
‘unexpected’ target area more distant to the adenine moiety of NAD(P)H, where an enzyme’s
specificity to one of the coenzymes typically originates from. This hit likely could have been
missed by applying a more rational engineering approach. Combination of the five identified
target loops enabled not only the transfer of the donor enzymes reductive half reaction
characteristics to NostocER1, but also the outperformance of both donors in terms of activity.
Additionally, some hybrids showed increased or unchanged activity with NADPH, some
showed a strongly decreased NADPH activity. Depending on the application, it can be
desirable that the biocatalysts use selectively only one or both coenzymes.[16] Moreover, no
undesired characteristics were transferred through this engineering approach. The
NostocER1 hybrids showed no changes in the wildtypes’ excellent stereoselectivity. These
findings demonstrate the high potential of loop swapping as enzyme engineering technique
with a low screening effort to selectively influence coenzyme binding in a (βα)8-folded ERs.
Experimental Section
305
Recombinant enzyme expression and purification
306
307
308
309
310
311
312
313
Chemically competent E. coli BL21 (DE3) cells (Novagen, San Diego, CA) were transformed
for protein expression and cultivated on Lysogeny broth (LB) agar plates supplemented with
35 mg L-1 kanamycin. Preculture preparation was performed in 4 mL Terrific Broth
supplemented with 35 mg L-1 kanamycin (TB-Kan), inoculated with a single colony and
incubated overnight (13 mL polypropylene tubes, 30°C, 250 rpm, 5 cm excentricity). 1 L
shaking flasks without baffles supplemented with 200 mL TB-Kan were inoculated with 1 mL
preculture and cultivated (37°C, 200 rpm, 5 cm excentricity) until an attenuance at 600 nm
(OD600) of 0.6 was reached. Cultivation media was cooled to 20 °C on ice and expression
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
16
314
315
316
was induced using by adding 200 µM isopropyl β-D-thiogalactoside (IPTG). Thereafter, cells
were cultivated (20 °C, 250 rpm, 3.5 cm excentricity) for exactly 20 h and collected by
centrifugation (3260 g, 4°C, 5 min).
317
318
319
320
321
322
323
324
325
The harvested cells containing ERs with an N-terminal His6-tag were purified via affinity
chromatography using HisPur™ Ni-NTA Spin Columns (Thermo Fisher Scientific, Waltham,
MA). After cell lysis in 5-times the cell wet weight EQ-buffer (20 mM sodium phosphate, 300
mM sodium chloride with 20 mM imidazole, pH 7.4) via sonication, purification was
performed at 4°C according to the manufacturer’s protocol, with the exception that an
imidazole concentration of 55 mM in the wash buffer was applied. Buffer exchange to
100 mM sodium phosphate buffer, using a minimal protein solution to buffer ratio of 1:200,
was applied via dialysis (4°C, overnight) using ZelluTrans membranes (Carl Roth, Karlsruhe,
Germany; cutoff 14 kDa).
326
Protein analysis
327
328
329
330
331
Protein concentration was determined using the Pierce™ BCA Protein Assay Kit (Thermo
Fisher Scientific, Waltham, MA). Enzymatic purity and size were determined via SDS-PAGE
using 3% and 12.5% Bis-Tris gels in Tris-glycine running buffer with JustBlue Prestained
Protein Marker (NIPPON Genetics EUROPE, Düren, Germany). Gels were stained according
to Fairbanks and coworkers.[38]
332
Enzyme assay
333
334
335
336
337
338
339
Enzyme activity was determined by a photometric assay measuring the decrease of
NAD(P)H at 340 nm using a molar absorption coefficient of 6.22 mM-1 cm-1. Reactions were
performed using 1.5 - 500 µg mL-1 purified enzyme in sodium phosphate buffer (100 mM,
pH 7.0). Assays were conducted at 30 °C using F96 microwell plates (Nunc, Roskilde,
Denmark) and a Multiscan™ FC Photometer (Thermo Fisher Scientific, Waltham, MA).
Measurements were recorded for 10 min and 100 data points. Reactions were initiated by
the addition of 10 mM maleimide as substrate. Reaction rates were determined by
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
17
340
341
342
343
344
345
346
347
348
automated linear regression using MATLAB R2015b (The MathWorks, Natick, MA). The
maximal rate of decrease was calculated for at least 2 min (20 data points) and a minimal R2
of 0.995. One unit of enzyme activity (U) was defined as the amount which will catalyze the
transformation of 1 µmol substrate per minute. The v200 values were determined using
200 µM NAD(P)H on a 200 µL scale. For the determination of kinetic parameters, 10 mM
maleimide was kept constant, while the concentration of NAD(P)H was varied between
20 µM and 1200 µM, whereby above 900 µM the assay volume was scaled to 150 µL. Kinetic
parameters were estimated according to the Michaelis–Menten equation via nonlinear
regression analysis (Sigma Plot 12.3, Systat Software, San Jose, CA).[39]
349
350
351
352
353
354
355
356
357
In order to determine the specific activities with (R)-carvone as substrate the assays were
performed on a 10 mL scale using 500 µM NAD(P)H, 30 - 500 µg mL-1 purified enzyme,
10 mM (R)-carvone (added as an EtOH solution, < 5% final EtOH concentration) in sodium
phosphate buffer (100 mM, pH 7.0). The reaction was initiated through the addition of
cosubstrate, and data points were gathered for 5 min in 30 s intervals. Samples were
immediately put on ice, in order to stop the reaction. Afterwards, the samples were extracted
with 25% (v/v) EtOAc containing 7.2 mM (R)-limonene as internal standard. The more
convenient sample storage on ice was preferred over direct extraction, since both strategies
yielded in the same results.
358
Statistical information
359
360
361
362
363
Assays were performed in technical quintuplicates (n = 5) resulting in means µT and standard
deviation σT. Additionally, between two (m = 2) and six (m = 6) biological replicates of these
technical replicates were conducted, starting from individual single colonies. Combined
biological means µ were calculated (Equation 1) using a relative weighting βi for every
technical mean µT,i (Equation 2).
푚
µ = ∑ µ푇,푖  훽푖
(1)
푖=1
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
18
1
σ푖2
훽푖 =
(2)
1
푚
∑
푗=1
휎 ²
푇,푗
364
The corresponding standard deviation (σ) was calculated according to Equation 3.
휎 = ∑(µ2푇,푖 + 휎푇,푖²) 훽푖 − µ²
(3)
√
푖
Additional information
365
366
367
Supporting information is available for this paper, including further descriptions about
applied materials and enzymes, cloning strategies, analytical procedures or homology
modelling.
368
Correspondence and requests for materials should be addressed to K.C.
Acknowledgements
369
370
371
372
373
374
375
This work was supported by the ‘BayBiotech’ project association (www.baybiotech.de)
financed by the Bavarian State Ministry of the Environment and Consumer Protection (grant
number TLK01U-69036). The authors would like to thank Sophie v. Schönberg, Maurits
Brinkman, Felix Schöpf, Fereshteh Sadeqi, David Auber, Konstantina Xypolia, Stephanie
Grümbel or their help with the laboratory work as well as Sabine Wagner and Ingmar Polte
for the fruitful discussions. The support of Christoph Mähler from the TUM Graduate School
is acknowledged as well.
References
376
377
378
379
380
[1] T. Matsuda, R. Yamanaka, K. Nakamura, Tetrahedron: Asymmetry 2009, 20, 513–557.
[2] a) W. S. Knowles, Adv. Synth. Catal. 2003, 345, 3–13; b) R. Noyori, Angew. Chem. Int. Ed. 2002,
41, 2008–2022.
[3] a) K. Durchschein, M. Hall, K. Faber, Green Chem. 2013, 15, 1764; b) R. Stuermer, B. Hauer, M.
Hall, K. Faber, Curr. Opin. Chem. Biol. 2007, 11, 203–213; c) H. S. Toogood, N. S. Scrutton, Curr.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
19
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
Opin. Chem. Biol. 2014, 19, 107–115; d) A. Tosstorff, C. Kroner, D. J. Opperman, F. Hollmann, D.
Holtmann, Eng. Life Sci. 2017, 17, 71–76.
[4] H. S. Toogood, J. M. Gardiner, N. S. Scrutton, ChemCatChem 2010, 2, 892–914.
[5] C. Breithaupt, J. Strassner, U. Breitinger, R. Huber, P. Macheroux, A. Schaller, T. Clausen,
Structure 2001, 9, 419–429.
[6] R. M. Kohli, V. Massey, J. Biol. Chem. 1998, 273, 32763–32770.
[7] C. R. Pudney, S. Hay, N. S. Scrutton, FEBS J. 2009, 276, 4780–4789.
[8] A. Scholtissek, D. Tischler, A. Westphal, W. van Berkel, C. Paul, Catalysts 2017, 7, 130.
[9] E. D. Amato, J. D. Stewart, Biotechnol. Adv. 2015, 33, 624–631.
[10] C. K. Winkler, D. Clay, S. Davies, P. O’Neill, P. McDaid, S. Debarge, J. Steflik, M. Karmilowicz, J.
W. Wong, K. Faber, J. Org. Chem. 2013, 78, 1525–1533.
[11] S. K. Padhi, D. J. Bougioukou, J. D. Stewart, J. Am. Chem. Soc. 2009, 131, 3271–3280.
[12] S. Reich, B. M. Nestl, B. Hauer, ChemBioChem 2016, 17, 561–565.
[13] J. T. Wu, L. H. Wu, J. A. Knight, Clin. Chem. 1986, 32, 314–319.
[14] C. You, R. Huang, X. Wei, Z. Zhu, Y.-H. P. Zhang, Synthetic Syst. Biotechnol. 2017, 2, 208–218.
[15] V. Uppada, S. Bhaduri, S. B. Noronha, Curr. Sci. 2014, 106, 946–957.
[16] A. M. Chánique, L. P. Parra, Front. Microbiol. 2018, 9, 1–14.
[17] J. K. B. Cahn, C. A. Werlang, A. Baumschlager, S. Brinkmann-Chen, S. L. Mayo, F. H. Arnold,
ACS Synth. Biol. 2017, 6, 326–333.
[18] a) S. P. Miller, M. Lunzer, A. M. Dean, Science 2006, 314, 458–461; b) J. A. Rollin, T. K. Tam, Y.-
H. P. Zhang, Green Chem. 2013, 15, 1708.
[19] J. K. B. Cahn, S. Brinkmann-Chen, T. Spatzal, J. A. Wiig, A. R. Buller, O. Einsle, Y. Hu, M. W.
Ribbe, F. H. Arnold, Biochem. J. 2015, 468, 475–484.
[20] a) D. Cui, L. Zhang, S. Jiang, Z. Yao, B. Gao, J. Lin, Y. A. Yuan, D. Wei, FEBS J. 2015, 282,
2339–2351; b) S. Brinkmann-Chen, J. K. B. Cahn, F. H. Arnold, Metab. Eng. 2014, 26, 17–22.
[21] N. Borlinghaus, B. M. Nestl, ChemCatChem 2018, 10, 183–187.
[22] B. M. Nestl, B. Hauer, ACS Catal. 2014, 4, 3201–3211.
[23] R. Takase, B. Mikami, S. Kawai, K. Murata, W. Hashimoto, J. Biol. Chem. 2014, 289, 33198–
33214.
[24] M. Nishiyama, J. J. Birktoft, T. Beppu, J. Biol. Chem. 1993, 268.
[25] R. Chen, A. Greer, A. M. Dean, Proc. Natl. Acad. Sci. 1996, 93, 12171–12176.
[26] T. Yaoi, K. Miyazaki, T. Oshima, Y. Komukai, M. Go, J. Biochem. 1996, 119, 1014–1018.
[27] a) A. Ochoa-Leyva, X. Soberón, F. Sánchez, M. Argüello, G. Montero-Morán, G. Saab-Rincón, J.
Mol. Biol. 2009, 387, 949–964; b) A. Ochoa-Leyva, F. Barona-Gómez, G. Saab-Rincón, K. Verdel-
Aranda, F. Sánchez, X. Soberón, J. Mol. Biol. 2011, 411, 143–157; c) M. V. Golynskiy, J. C.
Haugner, B. Seelig, ChemBioChem 2013, 14, 1553–1563.
[28] H. Ma, T. M. Penning, Proc. Natl. Acad. Sci. 1999, 96, 11161–11166.
[29] E. Campbell, S. Chuang, S. Banta, Protein Eng. Des. Sel. 2013, 26, 181–186.
[30] Y. Fu, K. Castiglione, D. Weuster-Botz, Biotechnol. Bioeng. 2013, 110, 1293–1301.
[31] H.-B. Wang, X.-Q. Pei, Z.-L. Wu, Appl. Microbiol. Biot. 2014, 98, 705–715.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
20
421
422
423
424
425
426
427
428
429
430
431
432
433
[32] a) J. Tao, K. McGee, Org. Process Res. Dev. 2002, 6, 520–524; b) R. L. Hanson, S. L. Goldberg,
D. B. Brzozowski, T. P. Tully, D. Cazzulino, W. L. Parker, O. K. Lyngberg, T. C. Vu, M. K. Wong,
R. N. Patel, Adv. Synth. Catal. 2007, 349, 1369–1378; c) W. L. Parker, R. L. Hanson, S. L.
Goldberg, T. P. Tully, A. Goswami, Org. Process Res. Dev. 2012, 16, 464–469.
[33] Y. Dong, K. J. McCullough, S. Wittlin, J. Chollet, J. L. Vennerstrom, Bioorg. Med. Chem. Lett.
2010, 20, 6359–6361.
[34] B. J. M. Jansen, J. A. Kreuger, A. d. Groot, Tetrahedron 1989, 45, 1447–1452.
[35] R. A. Sheldon, J. M. Woodley, Chem. Rev. 2018, 118, 801–838.
[36] I. Fichan, C. Larroche, C. B. Gros, J. Chem. Eng. Data 1999, 44, 56–62.
[37] a) R. A. Sheldon, P. C. Pereira, Chem. Soc. Rev. 2017, 46, 2678–2691; b) K. Castiglione, Y. Fu,
I. Polte, S. Leupold, A. Meo, D. Weuster-Botz, Biochem. Eng. J. 2017, 117, 102–111.
[38] G. Fairbanks, T. L. Steck, D. F. H. Wallach, Biochemistry 1971, 10, 2606–2617.
[39] R. J. Ritchie, T. Prvan, Biochem. Educ. 1996, 24, 196–206.
This article is protected by copyright. All rights reserved.

10.1002/adsc.201900073
Advanced Synthesis & Catalysis
21
Graphical Abstract
434
This article is protected by copyright. All rights reserved.