In vitro biocatalytic pathway design: Orthogonal network for the quantitative and stereospecific amination of alcohols

Article abstract of DOI:10.1039/c7ob01927k

The direct and efficient conversion of alcohols into amines is a pivotal transformation in chemistry. Here, we present an artificial, oxidation-reduction, biocatalytic network that employs five enzymes (alcohol dehydrogenase, NADP-oxidase, catalase, amine dehydrogenase and formate dehydrogenase) in two concurrent and orthogonal cycles. The NADP-dependent oxidative cycle converts a diverse range of aromatic and aliphatic alcohol substrates to the carbonyl compound intermediates, whereas the NAD-dependent reductive aminating cycle generates the related amine products with >99% enantiomeric excess (R) and up to >99% conversion. The elevated conversions stem from the favorable thermodynamic equilibrium (K′_eq = 1.88 × 10⁴² and 1.48 × 10⁴¹ for the amination of primary and secondary alcohols, respectively). This biocatalytic network possesses elevated atom efficiency, since the reaction buffer (ammonium formate) is both the aminating agent and the source of reducing equivalents. Additionally, only dioxygen is needed, whereas water and carbonate are the by-products. For the oxidative step, we have employed three variants of the NADP-dependent alcohol dehydrogenase from Thermoanaerobacter ethanolicus and we have elucidated the origin of the stereoselective properties of these variants with the aid of in silico computational models.

Full text of DOI:10.1039/c7ob01927k

View Article Online
View Journal
Organic &
Biomolecular
Chemistry
Accepted Manuscript
This article can be cited before page numbers have been issued, to do this please use: T. Knaus, L.
Cariati, M. F. Masman and F. G. Mutti, Org. Biomol. Chem., 2017, DOI: 10.1039/C7OB01927K.
This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
accepted for publication.
Volume 14 Number
1 7 January 2016 Pages 1–372
Organic &
Biomolecular
Chemistry
www.rsc.org/obc
Accepted Manuscripts are published online shortly after
acceptance, before technical editing, formatting and proof reading.
Using this free service, authors can make their results available
to the community, in citable form, before we publish the edited
article. We will replace this Accepted Manuscript with the edited
and formatted Advance Article as soon as it is available.
You can find more information about Accepted Manuscripts in the
author guidelines.
Please note that technical editing may introduce minor changes
to the text and/or graphics, which may alter content. The journal’s
standard Terms & Conditions and the ethical guidelines, outlined
in our author and reviewer resource centre, still apply. In no
event shall the Royal Society of Chemistry be held responsible
for any errors or omissions in this Accepted Manuscript or any
consequences arising from the use of any information it contains.
ISSN 1477-0520
COMMUNICATION
Takeharu Haino et al.
Solvent-induced emission of organogels based on tris(phenylisoxazolyl)
benzene
rsc.li/obc

Page 1 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
In vitro biocatalytic pathway design: orthogonal
network for the quantitative and stereospecific
amination of alcohols
Tanja Knaus‡, Luca Cariati‡, Marcelo F. Masman, Francesco G. Mutti*
Van ‘t Hoff Institute for Molecular Sciences, HIMSꢀBiocat, University of Amsterdam, Science
Park 904, 1098 XH, The Netherlands.
Corresponding Author: f.mutti@uva.nl
‡These authors contributed equally.
Supporting Information. Site directed Mutagenesis, Expression and purification conditions for
the recombinant enzymes, analytical methods as well as GC chromatograms and calculations of
Gibbs free energies.
1

Organic & Biomolecular Chemistry
Page 2 of 34
View Article Online
DOI: 10.1039/C7OB01927K
Abstract
The direct and efficient conversion of alcohols into amines is a pivotal transformation in
chemistry. Here, we present an artificial, oxidationꢀreduction, biocatalytic network that employs
five enzymes (alcohol dehydrogenase, NADPꢀoxidase, catalase, amine dehydrogenase and
formate dehydrogenase) in two concurrent and orthogonal cycles. The NADPꢀdependent
oxidative cycle converts a diverse range of aromatic and aliphatic alcohol substrates to the
carbonyl compound intermediates, whereas the NADꢀdependent reductive aminating cycle
generates the related amine products with >99% enantiomeric excess (R) and up to >99%
conversion. The elevated conversions stem from the favorable thermodynamic equilibrium (K’eq
= 1.88 × 1042 and 1.48 × 1041 for the amination of primary and secondary alcohols, respectively).
This biocatalytic network possesses elevated atom efficiency, since the reaction buffer
(ammonium formate) is both the aminating agent and the source of reducing equivalents.
Additionally, only dioxygen is needed, whereas water and carbonate are the byꢀproducts. For the
oxidative step, we have employed three variants of the NADPꢀdependent alcohol dehydrogenase
from Thermoanaerobacter ethanolicus and we have elucidated the origin of the stereoselective
properties of these variants with the aid of in silico computational models.
Introduction
The direct conversion of alcohol moieties into the corresponding amines is a pivotal
transformation in chemistry as a great number of chemical products and intermediates contain
amine functionalities (e.g. active pharmaceutical ingredients, agrochemicals, fine chemicals,
polymers, dyes, plasticizers, emulsifiers, detergents, etc).1 Chemical and enzymatic methods are
currently widely applied in industry for the conversion of prochiral ketones into optically active
2

Page 3 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
amines.2, 3 However, a future transition to a bioꢀbased economy requires the development of
efficient strategies for the direct conversion of alcohols into amines. A classical method for the
4
oneꢀpot synthesis of amines from alcohols relies on three sequential steps: i) the conversion of
alcohol into azide via Mitsunobu reaction,5 ii) the transformation of azide to iminophosphorane
via Staudinger reaction,6 followed by iii) hydrolysis to the amine product in diluted acid.
However, this method and related ones possess extremely low atom efficiency because of the
requirement of various and complex reagents in stoichiometric amounts. Organometalꢀcatalysis
using iridium, ruthenium, iron or copper complexes enables the oneꢀpot amination of alcohols
using a single catalyst in a process that is known as hydrogenꢀborrowing amination.7 8 This
method encompasses a first oxidation step of the alcohol into a carbonyl compound followed by
the in situ generation of an imine intermediate and its reduction to the amine. In this process, the
hydride liberated in the first step is quantitatively consumed in the second step. Nevertheless, the
chemocatalytic hydrogenꢀborrowing amination using ammonia as amine donor is plagued by low
chemoselectivity because complex mixtures of primary, secondary and tertiary amines are
obtained.9ꢀ12 Modifications of the chemocatalytic hydrogenꢀborrowing amination with improved
chemoselectivity require other amine donors such as aniline and its derivatives,13, 14 or aminoꢀ
pyridines,15 or an enantiopure auxiliary.16 Nonetheless, the moderate or total absence of
stereoselectivity, the elevated catalyst and cocatalyst loading, the requirement of an excess of the
alcohol starting material at a low concentration, the stringent reaction conditions and the
observed substrate inactivation of the catalyst during operation constitute current challenges that
thwart the widespread application of these methods.17
In this context, the exquisite properties of compatibility, chemoꢀ and stereoselectivity of enzymes
allowed our group and others to develop recently the asymmetric biocatalytic hydrogenꢀ
3

Organic & Biomolecular Chemistry
Page 4 of 34
View Article Online
DOI: 10.1039/C7OB01927K
borrowing amination of alcohols (Scheme 1A).18ꢀ20 In the first step an alcohol dehydrogenase
22
(ADH)21,
oxidizes the alcohol substrate to the carbonyl compound intermediate with
concomitant transfer of the hydride to the oxidized form of the nicotinamide coenzyme (NAD+).
In the second step, an amine dehydrogenase (AmDH)3, 23ꢀ28 performs the reductive amination of
the carbonyl moiety employing the NADH, which was generated in the first cycle, and ammonia.
Under the optimized reaction conditions,29 secondary alcohols were converted up to 95%
conversion within 48 h. Another formal enzymatic hydrogenꢀborrowing amination utilizes a
combination of an ADH, a ωꢀtransaminase and an alanine dehydrogenase.30ꢀ34 However, the
latter method has lower efficiency due to the requirement of 5 equivalents of alanine as
additional sacrificial amine donor and results in modest conversion and mediocre selectivity for
the amination of secondary alcohols.
In this work, we present a new and high efficient oneꢀpot biocatalytic strategy for the asymmetric
amination of secondary and primary alcohols. The new method relies on the different specificity
of some oxidoreductases for accepting either the phosphorylated form (NADP) or the nonꢀ
phosphorylated form (NAD) of the nicotinamide coenzyme (Scheme 1B). In the first oxidative
cycle, an ADH oxidizes the alcohol starting material to the carbonyl compound intermediate (e.g.
ketone or aldehyde) by the transfer of a hydride from the substrate to NAPD+. The highly
specific NADPꢀoxidase (YcnD) from Bacillus subtilis35 regenerates in situ NADP+, consuming
dioxygen and liberating H2O2. A third enzyme, a catalase, effects the disproportion of H2O2 to
water and dioxygen; the latter reꢀenters in the first cycle. In the second reductive cycle, an amine
dehydrogenase (AmDH) performs the reductive amination of the carbonyl compound
intermediate to afford the final amine product. A formate dehydrogenase from Candida boidinii
(CbꢀFDH)36 recycles the required NADH at the expense of formate. Notably, the reaction buffer
4

Page 5 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
(e.g. ammonium formate) is both the source of ammonium and electrons for the reductive
amination step, whereas the electron acceptor in the first step is the innocuous molecular oxygen
from air. As the two cycles are running concurrently in oneꢀpot without compartmentalization,
the biocatalytic network constitutes an elegant example of an oxidationꢀreduction orthogonal
chemical process.37 Only one other similar type of biocatalytic orthogonal twoꢀstep oxidationꢀ
reduction network (i.e. involving a combination of dehydrogenases with divergent cofactor
dependence, NAD or NADP) has been created previously for the deracemisation of racemic
alcohols.38, 39 Another outstanding application of the same concept for the deracemization of
alcohols required, instead, the compartmentalization of the enzymes in order to separate
physically the oxidative and the reductive steps.40
Therefore, the herein reported oxidationꢀreduction biocatalytic network for the amination of
alcohols is conceptually different from other already known systems. In fact, other types of
oxidationꢀreduction cascades for the deracemization of alcohol moieties were created through the
combination of an alcohol oxidase (e.g. Cu2+ or flavin dependent) for the oxidative step, with a
dehydrogenase for the reductive step.41, 42 A laccase/TEMPO system can also substitute the
alcohol oxidase in the oxidative step.43 Finally, the combination of an alcohol oxidase, a
transaminase along with an alanine dehydrogenase for recycling of the amine donor, allowed for
the oneꢀpot twoꢀstep amination of alcohols.44, 45 The variation of the latter cascade using a
laccase/TEMPO system with a transaminase has been recently published.46 The last example is
the orthogonal twoꢀstep amination of alcohols involving ωꢀtransaminases, which was limited by
a maximum of 32% conversion.31
5

Organic & Biomolecular Chemistry
Page 6 of 34
View Article Online
DOI: 10.1039/C7OB01927K
Scheme 1. Biocatalytic strategies for the oneꢀpot conversion of alcohols to enantiopure amines
employing alcohol dehydrogenases (ADHs) and amine dehydrogenases (AmDHs): (A)
biocatalytic hydrogenꢀborrowing amination (i.e. redoxꢀinterconnected steps); (B) orthogonal
biocatalytic network (i.e. concurrent redox separated steps, this work). ꢁrG’° is the calculated
free Gibbs energy for the reaction in standard conditions (T 298.15 K, P 1 bar) and in aqueous
buffer at pH 8.5, ionic strength 1 M.
Results and discussion
Thermodynamic calculations. As alcohol and amine are in the same oxidation state, no
additional reagents are required in the biocatalytic hydrogenꢀborrowing amination (Scheme 1A).
Therefore, the hydrogenꢀborrowing amination possesses the highest atom efficiency. On the
other hand, the biocatalytic hydrogenꢀborrowing amination is a reversible reaction; hence, the
final conversion is dictated by the thermodynamics of the reaction. In fact, the variation of the
standard Gibbs free energy for the biocatalytic hydrogenꢀborrowing amination of secondary
alcohols was estimated in this work as ꢁrG’° ꢀ0.6 KJ molꢀ1 (for details, see SI Table S1).47 This
6

Page 7 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
value was calculated for an aqueous system, at pH 8.5 and ionic strength 1 M. These parameters
are analogous to the experimental conditions applied in our biocatalytic hydrogen borrowing
amination.18 Using the same approach, the estimated ꢁrG’° for the amination of primary alcohols
is ꢀ6.8 KJ molꢀ1 (for details, see SI Table S2). These data are in agreement with our previous
experimental results that showed a more favorable thermodynamic equilibrium for the hydrogenꢀ
borrowing amination of primary alcohols (>99% conversion) compared to secondary alcohols
(max 95% conversion) using an excess of ammonium/ammonia as driving force for the reaction.
Thus, we thought that an oxidationꢀreduction orthogonal chemical process (Scheme 1B) might
lead, in general, to quantitative conversions because of a more favorable thermodynamic
equilibrium. Furthermore, the use of two separated redox steps may also lead to higher reaction
rates. Our calculations showed that the estimated overall ꢁrG’° for the amination of secondary
alcohols following the orthogonal pathway depicted in Scheme 1 is ꢀ235.0 KJ molꢀ1 (i.e. K’eq =
1.48 × 1041, for details see SI Table S5). Also in this case, the amination of primary alcohols is
more favored as the estimated overall ꢁrG’° is ꢀ241.3 KJ molꢀ1 (i.e. K’eq = 1.88 × 1042, for details
see SI Table S8). Interestingly, the thermodynamic driving force mainly stems from the first
oxidative step (for details, see SI paragraph 5).
Testing suitable alcohol dehydrogenases and rational understanding of their stereoselective
properties. Aiming at creating an atomꢀefficient and orthogonal biocatalytic network for the
amination of secondary alcohols with improved thermodynamic equilibrium, we initially
searched for suitable high selective NADPꢀdependent enzymes for the oxidative cycle.
The alcohol substrates selected for our study are reported in Figure 1.
7

Organic & Biomolecular Chemistry
Page 8 of 34
View Article Online
DOI: 10.1039/C7OB01927K
Figure 1. Structures of the secondary (1aꢀp) and primary (1qꢀt) alcohol substrates explored in
this study. For the detailed structural representation of the substrates, see SI Figures S1ꢀ3.
Besides the specificity for NADP as coenzyme, the amination of S or R configured secondary
alcohols as well as their racemic mixtures requires the availability of stereocomplementary
ADHs or a nonꢀenantioselective ADH. In this regard, a Prelog ADH from Thermoanaerobacter
ethanolicus (TeSꢀADH, also known as ADHꢀT) was previously characterized as NADPꢀ
dependent ADH; however, its substrate scope was limited to the reduction of short and mediumꢀ
48ꢀ52
length chain aliphatic ketones.
The mutation of tryptophan 110 into an alanine (TeSꢀADH
W110A variant) has shown to increase the substrate scope of the ADH to include a number of
53ꢀ55
aromatic ketones.
In another study of enzyme engineering on the same scaffold, the
mutation of isoleucine 86 into an alanine created a second variant (TeSꢀADH I86A) that showed
reversed stereoselectivity (e.g. antiꢀPrelog) for the reduction of a set of seven aromatic and
heteroaromatic ketones.56 Furthermore, other variants were created via: i) saturation mutagenesis
of the amino acid residue W110;57, 58 ii) combination of the mutations I86A, W110A and/or
8

Page 9 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
inclusion of a further mutation such as C295A.58ꢀ60 A number of these variants showed relaxed
stereoselectivity for the reduction of prochiral ketones and, therefore, they were applied for the
deracemization of alcohols or dynamic kinetic resolution.58, 61 Nevertheless, comprehensive data
on the activity and stereoselectivity of TeSꢀADH W110A and TeSꢀADH I86A towards our
selection of alcohol substrates were not available. Thus, we investigated the reactivity and the
stereoselectivity of TeSꢀADH W110A and TeSꢀADH I86A for the asymmetric reduction of
structurally diverse ketones (enantiomeric excesses are reported in SI, Tables S9 and S10). We
expressed the ADHs with a StrepꢀIIꢀtag at the Cꢀterminus that allows for an easier purification.
As expected, TeSꢀADH W110A afforded always the Prelog alcohol (Sꢀconfigured alcohol, for
the selected substrates) as the major enantiomer. Nonetheless, the enzyme was stereospecific
(enantiomeric excess, ee ≥99%) only in four cases out of thirteen. In particular, the reduction of
2ꢀhexanone (2n) afforded a nearly racemic mixture of alcohols (5% ee) whereas the reduction of
2ꢀheptanone (1m) gave the (S)ꢀenantiomer in 39% ee. These findings are in agreement with
previous publications that showed the imperfect stereoselectivity of TeSꢀADH W110A.53, 54, 62
Conversely, our biocatalytic orthogonal aminating network utilizes the ADH to accomplish the
oxidation of an alcohol substrate to a ketone intermediate. Thus, TeSꢀADH W110A turned out to
be a useful catalyst because perfect stereoselectivity is not required for our purpose: the
important property is instead the elevated selectivity of TeSꢀADH W110A for the coenzyme (i.e.
NADP compared to NAD).
Then, we investigated whether TeSꢀADH I86A is indeed a general antiꢀPrelog variant because
the literature reports the reductions of only few ketones as substrates.56 Surprisingly, we
discovered that TeSꢀADH I86A is a nonꢀstereospecific enzyme when tested with our selection of
substrates (see SI, Table S10). TeSꢀADH I86A always afforded mixtures of enantiomers that
9

Organic & Biomolecular Chemistry
Page 10 of 34
View Article Online
DOI: 10.1039/C7OB01927K
were enriched either of the Prelog (Sꢀconfigured) or antiꢀPrelog (Rꢀconfigured) alcohol. In
particular, reverse stereoselectivity compared to TeSꢀADH W110A (i.e. main formation of the Rꢀ
enantiomer) was observed only for the reduction of paraꢀfluoro phenylacetone (2b), 2ꢀhexanone
(2n) and 4ꢀmethyl butanꢀ2ꢀone (2p). In all the other cases investigated, the Sꢀenantiomer (i.e.
Prelog) was still the main product with an enantiomeric excess ranging from 27% to >99%.
Hence, TeSꢀADH I86A cannot be applied as a general antiꢀPrelog ADH in our orthogonal
biocatalytic network. Therefore, we turned our attention to the antiꢀPrelog ADH from
Lactobacillus brevis (LbꢀADH).63 A wealth of data regarding the substrate scope and the
stereoselectivity of LbꢀADH for the reduction of prochiral ketones in analytical and preparative
scale is available from literature.64ꢀ67 In all the cases reported, it was shown that LbꢀADH affords
the antiꢀPrelog alcohol as the major enantiomer with excellent enantiomeric excess. Thus, it was
not required to investigate the reduction of ketones with LbꢀADH any further.
The final extension of our biocatalytic network to racemic alcohols as starting material requires a
nonꢀenantioselective ADH in the first oxidative step. It has been reported that simultaneous
mutations of WT TeSꢀADH at the positions I86, W110 and C295 with alanine residues generates
variants with relaxed enantioselectivity.59 However, C295 might be involved in the binding of
the catalytic Zn2+ ion of the TeSꢀADH. Therefore, mutation of C295 might decrease the stability
of TeSꢀADH as previously observed by other authors.59 This finding is corroborated further by
our computational models (see below), which show a distance of 4 Å between the thiol group of
C295 and the catalytic Zn2+. As robust enzymes are generally preferable for a concurrent multiꢀ
step artificial biocatalytic network, we focused our interest to TeSꢀADH I86A W110A double
variant.58, 68 We expressed the enzyme in our lab as StrepꢀIIꢀtagged recombinant enzyme and we
evaluated its stereoselectivity as for TeSꢀADH W110A and TeSꢀADH I86A. TeSꢀADH I86A
10

Page 11 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
W110A was completely nonꢀstereoselective (ee 0%) for the reduction of 2ꢀhexanone (2n).
Additionally, an extremely low level of stereoselectivity (i.e. from 10% to 24%) was observed
for the reduction of arylꢀaliphatic ketones 2a, 2b, 2d and all the other aliphatic ketones 2mꢀp
(see SI, Table S15). The origin of the stereoselective properties of the three variants (TeSꢀADH
W110A, TeSꢀADH I86A and TeSꢀADH I86A W110A) and the related intricacies of the amino
acid residues I86 and W110 were elucidated via computational studies. Local docking studies on
the mutant TeSꢀADH W110G with 2ꢀtetralone and derivatives were published concomitantly to
our study.68 First, we created a series of models for TeSꢀADH W110A, TeSꢀADH I86A and TeSꢀ
ADH I86A W110A based on the structure of the thermophilic alcohol dehydrogenase from
Thermoanaerobacter brockii as the template (PDB code 1YKF).69 Structural and computational
studies for TbꢀADH are also available.70 Notably, the wildꢀtype ADH from Thermoanaerobacter
ethanolicus (WT TeSꢀADH) and the ADH from Thermoanaerobacter brockii (TbꢀADH) possess
both 352 amino acid residues and they differ only in three of them (i.e. 99% structural identity,
SI paragraph 13). Furthermore, the three different amino acid residues are located further from
the active site (for more information see Experimental part). Herein, the 3D model structures of
TeSꢀADHs (WT, W110A, I86A and I86A W110A) were created containing acetone as model
substrate in the active site. Afterwards, acetone was utilized as a proꢀchiral control in all the
model structures of the TeSꢀADH variants because all the substrates were in silico created from
acetone. We performed this control in order to ensure that the enzymatic system under study
does not show any chirality biases (i.e. pseudoꢀprochirality preferences for a totally symmetric
substrate). In practice, we also submitted the enzymes models studied herein to the MD protocol
using acetone as ligand. Observing the simulation, such chirality biases did not occur. Hence,
starting from the reactive pose of acetone in the active site of all TeSꢀADH variants, models with
11

Organic & Biomolecular Chemistry
Page 12 of 34
View Article Online
DOI: 10.1039/C7OB01927K
bound substrates 2b (paraꢀfluorophenylacetone), 2p (4ꢀmethylꢀpentanꢀ2ꢀone), 2u
(acetophenone), 2g (phenoxyꢀacetone), and 2n (2ꢀhexanone) were obtained (see Experimental
part for detailed procedure). This selection of the substrates was made to study all the possible
cases: i) low ee values for TeSꢀADH I86A W110A, but high ee values for TeSꢀADH W110A
and TeSꢀADH I86A W110A (2b and 2p); ii) high ee values for all the TeSꢀADH variants (2u
and 2g); iii) low or moderate ee values for all the TeSꢀADH variants (2n). We point out that the
model of WT TeSꢀADH was used as a starting point for the generation of the models of the TeSꢀ
ADH variants, but it was not considered for simulations any further.
Following the in silico protocol (see Experimental part), we were able to score the
enantioselective preferences of these TeSꢀADH variants as well as to understand the structural
consequences of inducing mutations at the amino acid positions 86 and 110. One can observe in
Figure 1a that acetone was located in its reactive conformation in the active site of the model of
WT TeSꢀADH. In Figure 2a, it can also be observed that the active site of WT TeSꢀADH is
rather constricted in space, hence limiting the substrate acceptance to rather nonꢀvoluminous
substrates such as acetone. This finding is in agreement with the early reported substrate scope of
the WT enzyme that was limited to mediumꢀlength chain linear aliphatic ketones.49 As
previously reported, I86 and W110 are located in opposite positions in the active site of the
55, 56
enzyme.
Mutation of one of these positions to a rather small side chain amino acid such as
alanine, would allow for the creation of space in the active site that is required to accommodate
the substrate either in the proꢀS or proꢀR orientation. In the present work, these structural
intricacies were experimentally determined and computationally confirmed by inducing such
mutations. In relation to the stereoselective properties of the TeSꢀADH variants, the level of
agreement between the experimental data and the computational models was evaluated using a
12

Page 13 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
scoring function (ꢀꢁꢆꢂ/ꢃꢇꢄꢅ, see the Experimental part for more details). A score value higher than 1
indicates a high preference for the proꢀS binding conformation of the given substrate, a score
value lower than 1 indicates higher preference for the proꢀR binding conformation of the given
substrate, while a score equal or close to 1 indicates no stereoselective preferences of the variant
for the given substrate. The average score values for each variant and each selected substrate is
reported in Figure 2b. It is noteworthy that these theoretical results showed a good agreement
with the experimental data, hence validating the model structures presented in this work.
One can observe that the creation of more space into the “proꢀS enzyme binding pocket”
(W110A mutation in TeSꢀADH W110A) allows a more voluminous substrates like 2b to be
accommodated in the active site in a reactive conformation (see Figure 2c). On the other hand,
the generation of space in the opposite side (I86A mutation in TeSꢀADH I86A) allows the same
substrate to bind assuming its proꢀR conformation (see Figure 2d). Inducing both mutations at
the same time (I86A and W110A mutations in TeSꢀADH I86A W110A) generates space at both
sides of the hydride transfers region of the enzyme, thus enabling the observed nonꢀ
stereoselective preference of this variant towards most of the substrates from this study (SI,
Table S15). It is interesting to note that TeSꢀADH I86A W110A is still high proꢀS selective for
substrates 2u and 2g. This feature can be in part explained by the high tendency of these
substrates to form a ꢈꢀꢈ interaction between their aromatic group and the residue Y267, which is
located on the proꢀS side of the hydride transfer center. Residue M285 seems to play a role as
well.
13

Organic & Biomolecular Chemistry
Page 14 of 34
View Article Online
DOI: 10.1039/C7OB01927K
Figure 2. Structural intricacies of enantioselective preferences of TeSꢀADH. Panel a shows the
active site of WT TeSꢀADH with acetone in its reactive conformation. Panel b shows the scoring
values of all the TeSꢀADH variants for substrates 2b, 2p, 2u, 2g, 2n and acetone, the latter used
as a control substrate (ctrl). A score value higher than 1 means high preference for the proꢀS
binding conformation of the given substrate, a score value lower than 1 indicates higher
preference for the proꢀR binding conformation of the given substrate, while a score equal or close
to 1 indicates no stereoselective preferences of the variant for the given substrate. Panel c and d
14

Page 15 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
show the active site of TeSꢀADH W110A and TeSꢀADH I86A W110A, respectively, containing
the substrate 2b in its proꢀS conformation (panel c) and its proꢀR conformation (panel d).
Optimization of the oxidative step. We investigated the first oxidative cycle at different pH
values. Hence, we combined the NADPꢀdependent W110A variant of the ADH from
Thermoanaerobacter ethanolicus (TeSꢀADH W110A) with a high coenzyme selective NADPHꢀ
oxidase from Bacillus subtilis (YcnD)35 to carry out the oxidation of (S)ꢀphenyl 2ꢀpropanol ((S)ꢀ
1a) to phenylacetone (2a). The following reaction conditions were applied: (S)ꢀ1a (20 mM),
TeSꢀADH W110A (13 ꢂM), His6ꢀtagged YcnD (5 ꢂM), catalase (0.2 ꢂM) and NADP+ (0.5 mM)
in TrisꢀHCl buffer (50 mM) at varied pH (7 ꢀ 8.5). The most elevated conversion was 91% at pH
7.5 after 24 h reaction time, whereas performing the reaction at pH 8.5 led to 85% conversion
(see SI, Table S11, entry 2 and 4, respectively). In a following set of experiments (see SI, Table
S12) we tried to improve the conversion for the oxidation of (S)ꢀ1a in TrisꢀHCl buffer (50 mM)
at pH 8.5 as it is the optimal pH for the reductive amination with AmDHs (i.e. second cycle in
Scheme 1).18, 24 In particular, at this stage, we had to apply a double concentration of TeSꢀADH
W110A (26 ꢂM) and His6ꢀYcnD (10 ꢂM) and extend the reaction time up to 2 days in order to
obtain 98% conversion. However, the use of HCOONH4 (the required buffer for the subsequent
reductive amination) instead of TrisꢀHCl, under the same reaction conditions, reduced again the
conversion (85% after 2 days, data not shown in SI).
In our previous study on the biocatalytic hydrogenꢀborrowing amination, we have shown that
some ADHs (as purified enzymes in ammonium buffer) suffer of mediocre stability when
combined in oneꢀpot reactions with other enzymes that bear a polyꢀhistidine tag.18 We proposed
that the imidazole groups of the polyꢀhistidine chain can coordinate to the divalent cations (e.g.
Ca2+, Mg2+, Zn2+) that are present in the structure of most microbial ADHs. These cations are
15

Organic & Biomolecular Chemistry
Page 16 of 34
View Article Online
DOI: 10.1039/C7OB01927K
essential either for retaining the correct oligomeric state of the ADHs (i.e. stability) or for the
catalytic activity. These properties apply to the ADHs used in the present study.55, 63 In fact,
when we used Cꢀterminal His6ꢀtagged YcnD for the dualꢀenzyme oxidation of (S)ꢀ1a, the
formation of a white enzyme precipitate during the course of the biocatalytic oxidation was
observed, hence corroborating our hypothesis. Consequently, we prepared an alternative
recombinant YcnD that was devoid of a Hisꢀtag. Indeed, the substitution of Cꢀterminal His6ꢀ
tagged YcnD with the YcnD devoid of Hisꢀtag was beneficial for the oxidation as >99%
conversion was obtained in HCOONH4 (1 M, pH 8.5) in 12 h under the optimized reaction
conditions: (S)ꢀ1a (20 mM), TeSꢀADH W110A (26 ꢂM), YcnD (10 ꢂM), catalase (0.2 ꢂM) and
NADP+ (0.5 mM), (data not shown in SI). This result confirms that one must consider and select
very carefully enzyme terminal tags when planning multiꢀenzyme processes in vitro.18 For this
reason, in this work, we expressed and purified the TeSꢀADH variants with a StrepꢀIIꢀtag linked
at their Cꢀterminus.
Orthogonal biocatalytic network for the amination of (S)-configured alcohols with
inversion of configuration. The amine dehydrogenases (AmDHs) used in this study were
obtained by protein engineering of wildꢀtype amino acid dehydrogenases (AADHs) that are
dependent on NAD as coenzyme. As none of the mutations was located in the binding site of the
coenzyme, the AmDH variants retain the same specificity for NADH as the parent wildꢀtype
25, 28, 71
enzymes.
This property is important in order to assure the orthogonality between the
oxidative and the reductive cycle of the biocatalytic network.72 The AmDHs used in this study
were: i) BbꢀPhAmDH that was originated from the phenylalanine dehydrogenase from Bacillus
badius;71 ii) RsꢀPhAmDH that was originated from the phenylalanine dehydrogenase from
28
Rhodococcus sp.;25 iii) Ch1ꢀAmDH that is a chimeric enzyme obtained via domain shuffling
16

Page 17 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
of BbꢀAmDH and another variant originated from the leucine dehydrogenase from Geobacillus
stereothermopilus.27 In a very recent extensive study we have determined the optimal conditions
for the reductive amination in HCOONH4 buffer (1 M, pH 8.5) of carbonyl compounds (20–50
mM) catalyzed by AmDHs in combination with the formate dehydrogenase from Candida
boidinii (CbꢀFDH).24 The high specificity of CbꢀFDH for the reduction of NAD+ to NADH is
also very important for the present study.36
In a preliminary set of experiments (see SI, Table S13), we proved unequivocally that the
presence of any terminal Hisꢀtagged enzyme in the biocatalytic network is detrimental. The
amination of (S)ꢀ1b (20 mM) was attempted with StrepꢀIIꢀtagged TeSꢀADH W110A (26 – 52
ꢂM), Hisꢀtagged YcnD (5 – 10 ꢂM), Hisꢀtagged Ch1ꢀAmDH (126 ꢂM), CbꢀFDH (20 ꢂM) and
catalase (0.2 ꢂM) in HCOONH4 buffer (1 M, pH 8.5) at 30 °C. The concentration of both
coenzymes NAD+ and NADP+ was 0.5 mM. The maximum conversion to the amine (R)ꢀ3b was
limited to 64% after 48 h (SI, Table S13, entry 1); the remaining component of the reaction
mixture was exclusively the alcohol (S)ꢀ1b (i.e. the concentration of ketone 2b was ca. 1%).
Thus, the moderate conversion stems from a catalytic inefficiency of the ADH employed in the
oxidation cycle. Other data (SI, Table S13, entries 1ꢀ4) revealed that a decrease in the conversion
of (S)ꢀ1b into (R)ꢀ3b correlated with an increase of the concentration of Hisꢀtagged YcnD,
independently from the concentration of StrepꢀIIꢀtagged TeSꢀADH W110A. Finally, the use of
Ch1ꢀAmDH devoid of Hisꢀtag in combination with Hisꢀtagged YcnD, slightly increased the
conversion up to a maximum of 78% (SI, Table S13, comparing entries 1 and 5).
Finally, we repeated the biocatalytic reaction with Hisꢀtagged YcnD (5 µM) and Hisꢀtagged
Ch1ꢀAmDH (126 µM) by adding the StrepꢀIIꢀtagged TeSꢀADH W110A in small aliquots every
three hours (4.5 ꢂL per aliquot, 5 aliquots, total concentration 65 ꢂM). In this case, the
17

Organic & Biomolecular Chemistry
Page 18 of 34
View Article Online
DOI: 10.1039/C7OB01927K
conversion raised up to 98% in 48 h. These data all together indicated that the limitation of the
orthogonal biocatalytic network was due to the poor stability of the StrepꢀIIꢀtagged TeSꢀADH
W110A in the concomitant presence of the any other Hisꢀtagged enzyme in solution. Further
experiments demonstrated that the activity of the StrepꢀIIꢀtagged TeSꢀADH W110A drops to
undetectable levels within 1 h of incubation in presence of any other Hisꢀtagged enzyme from
this study (data not shown).
When we prepared all the five enzymes involved in the biocatalytic network as variants devoid
of Hisꢀtags, the compatibility of the system was finally assured. Under the reaction conditions of
(S)ꢀ1b (20 mM), StrepꢀIIꢀtagged TeSꢀADH W110A (45 ꢂM), YcnD (10 ꢂM), Ch1ꢀAmDH (126
ꢂM), CbꢀFDH (20 ꢂM), catalase (0.2 ꢂM), NAD+ (0.5 mM) and NADP+ (0.5 mM) in HCOONH4
buffer (1 M, pH 8.5) at 30 °C, the concentration of the amine (R)ꢀ3b surpassed 99% after 12 h.
Figure 3 depicts the progress of the reaction over the time (SI, table S14). The enantiomeric
excess of the amine (R)ꢀ3b remained constant during the course of the reaction and it was more
than 99%.
Figure 3. Time study for the amination of (S)ꢀ1b (20 mM) with inversion of configuration using
the fiveꢀenzyme orthogonal biocatalytic network.
18

Page 19 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
The broad applicability of the simultaneous twoꢀstep biocatalytic network was evaluated for the
amination of a panel of (S)ꢀconfigured secondary alcohols (Figure 1). Since all the AmDHs from
this study perform the amination of prochiral ketones with (R)ꢀselectivity, the overall biocatalytic
network proceeded with inversion of configuration.
(S)ꢀphenylꢀpropanꢀ2ꢀol and derivatives ((S)-1a-d) were quantitatively converted into the related
enantiopure amines (Table 1, entries 1ꢀ4) as predicted from our thermodynamic calculations. In
addition, the other arylꢀaliphatic alcohols ((S)-1e-g) were aminated with elevated conversion
(95% or higher, Table 1, entries 5ꢀ7). In contrast, 1ꢀphenylethanol and derivatives ((S)-1h-k)
proved to be more challenging substrates as the conversion ranged from 17% to 22% (Table 1,
entries 8ꢀ11). Depending on the substrate, we observed either a relevant amount of the alcohol
starting material (2h, 2j) or an accumulation of the ketone intermediate (2i, 2k). In general, a
favorable thermodynamic equilibrium is not the only requirement for achieving a quantitative
amination within a reasonable time (e.g. ca. 24 h). In addition, both ADH and AmDH must
possess relevant activity towards the alcohol substrate and ketone intermediate, respectively.
Hence, the system is currently mainly limited by the low activity of the available AmDHs
towards acetophenone and its derivatives as already reported in our previous publication.24
Additionally, the activity of the employed ADH is not ideal, yet, for the oxidation of (S)ꢀ2h and
(S)ꢀ2j. Nonetheless, the final amine product was always obtained in enantiopure form. Finally,
aliphatic secondary alcohols ((S)-1l-p) were in general well accepted affording up to >99%
conversion (Table 1, entries 12ꢀ16). Notably, the orthogonal biocatalytic network proceeded with
perfect stereoselectivity in all the cases investigated.
19

Organic & Biomolecular Chemistry
Page 20 of 34
View Article Online
DOI: 10.1039/C7OB01927K
Table 1. Orthogonal, oxidationꢀreduction, biocatalytic network for the amination of enantiopure
secondary alcohols with inversion of configuration.
Enzymes
ADH
Conversion [%]
Amine Ketone Alcohol
Entry
Substrate
ee [%]
AmDH
Ch1
Ch1
Rs
1
2
(S)ꢀ1a
(S)ꢀ1b
(S)ꢀ1c
(S)ꢀ1d
(S)ꢀ1e
(S)ꢀ1f
(S)ꢀ1g
(S)ꢀ1h
(S)ꢀ1i
(S)ꢀ1j
(S)ꢀ1k
(S)ꢀ1l
(S)ꢀ1m
(S)ꢀ1n
(S)ꢀ1o
(S)ꢀ1p
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
TeS W110A
99
99
<1
<1
<1
<1
1
<1
<1
<1
<1
4
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
3
>99
>99
95
4
Rs
5
Rs
6
Rs
>99
95
<1
<1
25
78
25
73
3
<1
5
7
Bb
8
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
18
57
<1
58
5
9
22
10
11
12
13
14
15
16
17
22
97
<1
<1
<1
8
>99
>99
72
<1
<1
21
3
97
<1
Reaction conditions: Sꢀconfigured alcohol (20 mM), StrepꢀIIꢀtagged TeSꢀADH W110A (45
ꢂM), YcnD (10 ꢂM), Ch1ꢀAmDH (126 ꢂM) or BbꢀPhAmDH (117 ꢂM) or RsꢀPhAmDH (100
ꢂM), CbꢀFDH (20 ꢂM), catalase (0.2 ꢂM), NAD+ (0.5 mM) and NADP+ (0.5 mM) in HCOONH4
buffer (1 M, pH 8.5) at 30 °C and orbital agitation (170 rpm).
Orthogonal biocatalytic network for the asymmetric amination of racemic alcohols. Finally,
we applied the variant StrepꢀIIꢀtagged TeSꢀADH I86A W110A (52 ꢂM) in the orthogonal
biocatalytic network for the asymmetric amination of racemic alcohols. Starting from racemic
1a-b, 1l-n, 1p, enantiopure amines (>99% ee (R)) were obtained with high conversion (from
91% to >99%, Table 2, entries 1, 2, 13ꢀ15 and 17). Additionally, the amination of racꢀ1d and
racꢀ1o afforded the optically pure amine product with 75% and 70% conversion, respectively
20

Page 21 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
(Table 2, entries 4, 16). For the substrates racꢀ1c and racꢀ1e-k, it was required to combine two
stereocomplementary NADPꢀdependent ADHs in oneꢀpot: the Prelog TeSꢀADH W110A and the
antiꢀPrelog ADH from Lactobacillus brevis (LbꢀADH).63 The level of conversions were similar
to the results obtained for the amination with inversion of configuration (Table 1). Arylꢀaliphatic
substrates were converted better (from 63% to >99%) than 1ꢀphenylethanol derivatives (from
17% to 29%), (Table 2, entries 6ꢀ8 and entries 9ꢀ12, respectively). Finally, the amination of racꢀ
1cꢀd afforded 53% and 91% conversion, respectively (Table 1, entry 3 and 5). The
stereoselectivity was retained in all the cases (ee >99% (R)).
Furthermore, we performed the orthogonal biocatalytic network for the asymmetric amination of
a racemic alcohol in preparative scale. Substrate 1m (50 mg) was aminated to afford the amine
(R)ꢀ3m (>99% ee) with 97% conversion in 12 h. After a simple workꢀup, that is the extraction of
the unreacted alcohol and ketone under acid conditions followed by the extraction of the amine
under basic conditions, (R)ꢀ3m was isolated in pure form (45 mg, 90% yield, for details and
product characterization see Experimental and SI). Notably, no purification (e.g. column
chromatography) was required.
Table 2. Orthogonal, oxidationꢀreduction, biocatalytic network for the asymmetric amination of
racemic secondary alcohols.
Enzymes
ADH
Conversion [%]
AmDH Amine Ketone Alcohol
Entry Substrate
ee [%]
1
2
3
4
5
6
Rac-1a
Rac-1b
Rac-1c
Rac-1d
Rac-1d
Rac-1e
TeS I86A W110A
TeS I86A W110A
TeS W110A/Lb
TeS I86A W110A
TeS W110A/Lb
TeS W110A/Lb
Ch1
Ch1
Rs
97
91
53
75
91
86
3
<1
9
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
<1
<1
<1
<1
3
46
25
9
Rs
Rs
Rs
11
21

Organic & Biomolecular Chemistry
Page 22 of 34
View Article Online
DOI: 10.1039/C7OB01927K
7
Rac-1f
Rac-1g
Rac-1h
Rac-1i
Rac-1j
Rac-1k
Rac-1l
Rac-1m
Rac-1n
Rac-1o
Rac-1p
TeS W110A/Lb
TeS W110A/Lb
TeS W110A/Lb
TeS W110A/Lb
TeS W110A/Lb
TeS W110A/Lb
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
Rs
>99
63
<1
<1
30
77
23
81
2
<1
37
41
<1
56
2
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
8
Bb
9
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
29
10
11
12
13
14
15
16
17
23
21
17
94
4
>99
>99
70
<1
<1
26
5
<1
<1
4
92
3
Reaction conditions: racemic alcohol (20 mM), StrepꢀIIꢀtagged TeSꢀADH I86A W110A (52
ꢂM) or StrepꢀIIꢀtagged TeSꢀADH W110A (23 ꢂM) plus LbꢀADH (23 ꢂM), YcnD (10 ꢂM), Ch1ꢀ
AmDH (126 ꢂM) or BbꢀPhAmDH (117 ꢂM) or RsꢀPhAmDH (100 ꢂM), CbꢀFDH (20 ꢂM),
catalase (0.2 ꢂM), NAD+ (0.5 mM) and NADP+ (0.5 mM) in HCOONH4 buffer (1 M, pH 8.5) at
30 °C and orbital agitation (170 rpm).
Orthogonal biocatalytic network for the amination of (R)-configured alcohols with
retention of configuration. For the sake of completeness, we investigated the amination of
enantiopure (R)ꢀconfigured alcohols employing LbꢀADH. The results confirmed the general
trend for the amination of arylꢀaliphatic and aliphatic alcohols. However, the aminations of 1ꢀ
phenylethanol derivatives afforded up to 43% conversion indicating that LbꢀADH is a more
efficient ADH for the oxidation of these alcohols (Table 3).
Table 3. Orthogonal, oxidationꢀreduction, biocatalytic network for the amination of enantiopure
secondary alcohols with retention of configuration.
Enzymes
ADH
Conversion [%]
AmDH Amine Ketone Alcohol
Entry Substrate
ee [%]
1
2
(R)ꢀ1a
(R)ꢀ1b
TeS I86A W110A
TeS I86A W110A
Rh
Rh
96
96
1
3
4
>99 (R)
>99 (R)
<1
22

Page 23 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
3
4
(R)ꢀ1c
(R)ꢀ1d
(R)ꢀ1e
(R)ꢀ1f
(R)ꢀ1g
(R)ꢀ1h
(R)ꢀ1i
(R)ꢀ1j
(R)ꢀ1k
(R)ꢀ1l
(R)ꢀ1m
(R)ꢀ1n
(R)ꢀ1o
(R)ꢀ1p
Lb
Rs
Rs
30
6
<1
<1
1
70
94
2
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>99 (R)
Lb
5
Lb
Rs
97
6
Lb
Rs
>99
82
<1
<1
10
72
40
73
2
<1
18
47
<1
24
7
7
Lb
Lb
Bb
8
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
Ch1
43
9
Lb
28
10
11
12
13
14
15
16
Lb
36
Lb
20
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
79
19
<1
<1
2
>99
>99
80
<1
<1
18
2
89
9
Reaction conditions: Rꢀconfigured alcohol (20 mM), StrepꢀIIꢀtagged TeSꢀADH I86A W110A
(52 ꢂM) or LbꢀADH (45 ꢂM), YcnD (10 ꢂM), Ch1ꢀAmDH (126 ꢂM) or BbꢀPhAmDH (117 ꢂM)
or RsꢀPhAmDH (100 ꢂM), CbꢀFDH (20 ꢂM), catalase (0.2 µM), NAD+ (0.5 mM) and NADP+
(0.5 mM) in HCOONH4 buffer (1 M, pH 8.5) at 30 °C and orbital agitation (170 rpm).
Orthogonal biocatalytic network for the amination of primary alcohols. Terminal amines are
also important compounds for the fine and bulk chemical industry.1 Hence, new sustainable
routes for the conversion of primary alcohols to amines are of high interest. During tests for the
activity of TeSꢀADH I86A W110A, we found out that the variant is able to oxidize some
primary alcohols. Consequently, we tested our orthogonal biocatalytic network for the amination
of a small set of primary aliphatic alcohols applying the TeSꢀADH I86A W110A in the first
oxidative cycle. Mediumꢀlength chain alcohols as nꢀhexanol (1s) and nꢀheptanol (1r) were the
most reactive substrates yielding the terminal amine with 88% and 75% conversion (Table 4,
entries 2 and 3). Conversions dropped when we tested longer or shorterꢀlength chain aliphatic
alcohols (ca. 20%, Table 4, entries 1 and 4). In these latter cases, the aldehyde intermediate was
23

Organic & Biomolecular Chemistry
Page 24 of 34
View Article Online
DOI: 10.1039/C7OB01927K
not detected at the end of the reaction. This finding demonstrates that the oxidation of the alcohol
substrate is currently the limiting step for the amination of 1q and 1t.
Table 4. Orthogonal, oxidationꢀreduction, biocatalytic network for the amination of primary
alcohols.
Enzymes
ADH
Conversion [%]
Entry
Substrate
AmDH Amine
Aldehyde
Alcohol
80
1
2
3
4
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
TeS I86A W110A
Ch1
Ch1
Ch1
Ch1
20
75
88
19
<1
<1
<1
<1
1q
1r
1s
1t
25
13
81
Reaction conditions: primary alcohol (20 mM), StrepꢀII tagged TeSꢀADH I86A W110A (52
ꢂM), YcnD (10 ꢂM), Ch1ꢀAmDH (126 ꢂM), CbꢀFDH (20 ꢂM), catalase (0.2 ꢂM), NAD+ (0.5
mM) and NADP+ (0.5 mM) in HCOONH4 buffer (1 M, pH 8.5) at 30 °C and orbital agitation
(170 rpm).
Experimental
General optimized procedure for the orthogonal biocatalytic amination of enantiopure
alcohols with inversion of configuration on analytical scale: The reactions were conducted in
ammonium formate buffer (1 M, pH 8.5, final volume 0.5 mL) containing NADP+ and NAD+
(final concentration 0.5 mM each). Enzymes StrepꢀIIꢀtagged TeSꢀADH W110A (45 µM), YcnD
(10 µM), AmDH (either Ch1ꢀAmDH 126 µM or BbꢀAmDH 117 µM or RsꢀAmDH 100 µM),
CbꢀFDH (20 µM) and catalase (0.2 ꢂM) and the substrate (50 mM) were added. Finally (S)ꢀ
configured alcohol substrate (20 mM) was added. The reactions were run at 30 °C in an
incubator for 24 h (170 rpm). Workꢀup was performed by the addition of KOH (100 µL, 10 M)
followed by the extraction with dichloromethane (600 µL). The water layer was removed after
centrifugation and the organic layer was dried with MgSO4. Conversion was determined by GC
with an Agilent DBꢀ1701 column. The enantiomeric excess of the amine product was determined
24

Page 25 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
after derivatization to acetamido. Derivatization of the samples was performed by adding 4ꢀ
dimethylaminopyridine in acetic anhydride (40 µL of stock solution 50 mg mLꢀ1). The samples
were shaken in an incubator at RT for 30 minutes. Afterwards water (150 µL) was added and the
samples were shaken for additional 30 minutes. After centrifugation, the organic layer was dried
with MgSO4. Enantiomeric excess was determined by GC with a Variant Chiracel DEXꢀCB
column. Details on the GC analysis and methods are reported in the SI paragraph S14 and S15.
General optimized procedure for the orthogonal biocatalytic asymmetric amination
of racemic alcohols on analytical scale: The procedure was the same as reported above with
two differences. StrepꢀIIꢀtagged TeSꢀADH I86A W110A (45 µM) or a combination of StrepꢀIIꢀ
tagged TeSꢀADH W110A (23 µM) plus LbꢀADH (23 µM) were used. Racemic alcohols (20
mM) were used as substrates.
Preparative scale procedure for the orthogonal biocatalytic amination of 1m to give (R)-
3m: NAD+ and NADP+ (final concentration 0.5 mM each) were dissolved in ammonium formate
buffer (87 mL, 1 M, pH 8.5) in a 250 mL bottle with cap. Enzymes StrepꢀIIꢀtagged TeSꢀADH
I86A W110A (45 µM), YcnD (10 µM), Ch1ꢀAmDH (126 µM), CbꢀFDH (20 µM) and catalase
(0.2 ꢂM) were added. Finally, the substrate racꢀ1m (50 mg) was added. The reaction mixture was
incubated in an orbital shaker (30 °C, 170 rpm) and the progress of the reaction was monitored
by TLC and GC. When quantitative conversion was achieved (ca. 12 h), the reaction mixture was
acidified to pH 2ꢀ4 via addition of HCl (1M) while cooling in an ice bath. The water layer was
washed with methyl tertꢀbutyl ether (15 mL) to remove any possible remaining alcohol starting
material and ketone intermediate. The pH of the water phase was increased to basic pH via KOH
(10 M) while cooling in an ice bath. The water layer was extracted with methyl tertꢀbutyl ether
(2 x 15 mL). The organic fractions containing the amine product were combined and dried over
25

Organic & Biomolecular Chemistry
Page 26 of 34
View Article Online
DOI: 10.1039/C7OB01927K
MgSO4. After filtration and evaporation of the solvent, the product was obtained in pure form
(45 mg, 90% yield). Column chromatography was not required. The authenticity of the product
1
was confirmed by 1HꢀNMR (SI, paragraph S13). H NMR (400 MHz, CDCl3, δ ppm): 2.87 (m,
1H, CH), 1.48 (s 2H, NH2), 1.28 (m, 8H, CH2), 1.05 (d, J = 6.9 Hz, 3H, CH3), 0.89 (t, J = 6.9 Hz,
3H, CH3).
In silico design of the 3D-structural models of WT TeS-ADH, TeS-ADH W110A, TeS-ADH
I86A and TeS-ADH I86A W110A: The crystal structure of the thermophilic alcohol
dehydrogenase from Thermoanaerobacter brockii (TbꢀADH, PDB code 1YKF)69 was used as the
structural scaffold for the design of the model of the WT TeSꢀADH. This scaffold was selected
after a careful comparison between the sequence of the WT TeSꢀADH and all the available 3Dꢀ
structures and related sequences of the ADHs in the protein data bank (PDB). The sequences of
TbꢀADH and WTꢀTeSꢀADH differ only for 3 amino acids positions, namely: R91W, R313P, and
R325Q (for sequence alignment, SI paragraph 12). These 3 mutations were induced in silico in
order to generate the model of the WT TeSꢀADH.
All in silico modifications were done with the Yasara software,73 utilizing the AMBER 03
force field.74 The protonation state of all atoms was automatically adjusted, with the exception of
those atoms involved in the hydride transfer between coꢀfactor and the substrate. The protonation
state of the latter atoms was adjusted manually accordingly to their biochemical properties. Each
mutation in the template structure was introduced singularly, one after the other. Furthermore, a
threeꢀstep energy minimization procedure was executed for each mutation. The steps were as
follows: i) step one ꢀ only the mutated amino acid residue was energetically minimized; ii) step
two ꢀ the mutated residue plus all the residues located within a radius of 6 Å from the mutated
residue were subjected to energy minimization; iii) step three ꢀ the complete enzyme scaffold
26

Page 27 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
was subjected to energy minimization. The use of this protocol of energy minimization assures a
gradual adjustment of the complete structure to the new mutation, thus avoiding the production
of undesired secondary structure deformations.
Utilizing the same protocol as described above, the TeSꢀADH variants (i.e. W110A, I86A, and
I86A W110A) were generated from the model of WT TeSꢀADH. We point out that the model of
WT TeSꢀADH was created containing acetone as substrate and Zn2+ as the metal cofactor in its
active site. The position of the substrate and of Zn2+ was obtained from another available 3D
structure of thermophilic alcohol dehydrogenase from Thermoanaerobacter brockii (PDB code
1BXZ) that also contains Zn2+ ion and 2ꢀbutanol as substrate in its active site.75 This structural
information was utilized to introduce the substrate acetone in its reactive conformation into the
active site of all the models of TeSꢀADHs (WT, variant W110A, variant I86A and variant I86A
W110A). The standardized reactive conformation of the ketone substrate was stabilized
considering that the oxygen atom of the carbonyl group is in direct interaction with the Zn2+ ion
(1.5ꢀ2.5 Å C=OꢀꢀꢀꢀZn2+) and the hydride atom of the cofactor NADPH is in direct interaction
with the C atom of the same carbonyl group (1,5ꢀ3.5 Å O=CꢀꢀꢀꢀH). The Zn2+ ion is kept in place
by the interaction with the residues C37, H60, E60, and D150. For each TeSꢀADH variant, a set
of models was generated through the in situ introduction of each selected substrates (2b, 2p, 2u,
2g, and 2n). Each substrate was introduced starting from the spatial position of the acetone.
Notably, pairs of enzyme plus substrate models were generated, wherein the substrate is either in
the proꢀS or proꢀR conformation. For each enzyme plus substrate model, the threeꢀstep energy
minimization protocol was applied as explained before. After energy minimization, aiming at
further structural relaxation, a short molecular dynamic simulation (50 ps) was executed. Each
final relaxed structure was evaluated using the Autodock Vina scoring function in order to assess
27

Organic & Biomolecular Chemistry
Page 28 of 34
View Article Online
DOI: 10.1039/C7OB01927K
its binding energy at the respective reactive conformation. Thus, two main parameters for the
scoring function were used to rank the enzymatic complexes: i) the distance between the hydride
atom of the cofactor NADPH and the C atom of carbonyl group of the ketone substrate; ii) the
binding energy of the substrate in its reactive conformation. The following scoring function was
utilized:
ꢎ
ꢎ
ꢄ
ꢄ
ꢍ
ꢚ
ꢋꢌ
ꢋꢌ
ꢍ
ꢏꢐꢑ
ꢇꢛ
ꢉ ꢊ
ꢉ ꢊ
ꢗ + ꢘ ꢙ
ꢗ + ꢘ ꢙ
ꢒ ꢜ
ꢚ
ꢇꢛ
ꢄ
ꢒ
ꢌ
ꢄ
ꢌ
ꢓ
ꢓ
ꢀꢁꢆꢂꢃꢄꢅ
ꢀꢁꢇꢂꢃꢄꢅ
ꢏꢐꢑ
∑
∑
ꢔꢕꢖ ꢋ
∑
ꢔꢕꢖ ꢋ
ꢆ/ꢇ
ꢁꢂꢃꢄꢅ
ꢀ
=
=
ꢝ
ꢄ
ꢄ
ꢝ
ꢍ
ꢚ
ꢍ
ꢋꢌ
ꢋꢌ
ꢏꢐꢑ
ꢇꢛ
ꢒ ꢜ
ꢄ
ꢚ
ꢒ
ꢌ
ꢄ
ꢌ
ꢓ
ꢓ
ꢇꢛ
ꢏꢐꢑ
ꢔꢕꢖ ꢋ
∑
ꢔꢕꢖ ꢋ
ꢆ/ꢇ
ꢁꢂꢃꢄꢅ
where ꢀ
is the relative score of enzymatic preference of the proꢀS substrate (ꢀꢁꢆꢂꢃꢄꢅ) over the
proꢀR one (ꢀꢁꢇꢂꢃꢄꢅ); ꢞꢆ and ꢞꢇ are the distance HꢀC of the proꢀS and proꢀR substrate respectively;
ꢟ
ꢟ
ꢠꢟꢆ and ꢠꢟꢇ are the binding energy of the proꢀS and proꢀR substrate respectively; ꢞꢄꢅꢡ is the
reference HꢀC distance that was considered as the sum of the van der Waal radius of H and C
(2.9 Å); R is the gas constant and T is the temperature of the simulations (298.15 K). The
weighting parameter a and b were arbitrarily chosen to be a equal to 0.75 and b equal to 0.25.
All models herein were generated as homotetramers and each monomer possesses its own active
site. Furthermore, each monomer was considered as an independent simulation from each other;
thus each result is reported as the average observation over four active sites.
Conclusions
In this work, we present an atomꢀefficient artificial biocatalytic network that enables the oneꢀpot
stereoselective and up to quantitative conversion of alcohol functionalities into amines. The
overall artificial biocatalytic network proceeds in
a
concurrent fashion without
28

Page 29 of 34
Organic & Biomolecular Chemistry
View Article Online
DOI: 10.1039/C7OB01927K
compartmentalization. This network constitutes therefore an elegant example of orthogonal
tandem catalysis.72 We point out that our biocatalytic network exploits a high level of
orthogonality because the biocatalysts possess ꢀ at the same time ꢀ high specificity for the type of
compound and coenzyme without detectable crossꢀreactivity. On the other hand, the biocatalytic
network is substrate promiscuous, as it possesses the capability of converting a range of
structurally diverse alcohols.
The advantage of the herein described orthogonal biocatalytic network compared with the
biocatalytic hydrogenꢀborrowing amination,18 previously published by our group, is the more
favorable thermodynamic equilibrium (Scheme 1) that allows for obtaining >99% conversion for
the amination of secondary alcohols. In fact, the estimated K’eq for the amination of secondary
alcohol via the orthogonal network is 1.48 × 1041. In comparison the K’eq for the hydrogenꢀ
borrowing amination on the same substrate is 1.27.
Furthermore, the kinetics of the amination is also improved as the reaction times are reduced
from 48 h (hydrogenꢀborrowing process) to 6 h (orthogonal network) as shown in Figure 3. The
orthogonal biocatalytic network shows an elevated atom efficiency as the reaction buffer,
ammonium formate, is both the aminating agent and the source of reducing equivalents in the
second step. Overall, the amination of 1 mole of alcohol requires 1 mole of ammonium formate
and generates 1 mole of carbonate. Dioxygen is the only other reagent. The only drawback of
herein described orthogonal biocatalytic network compared with the biocatalytic hydrogenꢀ
borrowing amination is the requirement for two additional enzymes (YcnD and FDH) for the
recycling of the nicotinamide coenzymes. Nevertheless, this drawback can be alleviated if the
enzymes involved in the cascade are coexpressed altogether in E. coli as host organism and
whole cells are used as catalysts. Recent publications have shown that it is possible to produce
29