Mechanistic Insight into the Catalytic Promiscuity of Amine Dehydrogenases: Asymmetric Synthesis of Secondary and Primary Amines

Article abstract of DOI:10.1002/cbic.201800626

Biocatalytic asymmetric amination of ketones, by using amine dehydrogenases (AmDHs) or transaminases, is an efficient method for the synthesis of α-chiral primary amines. A major challenge is to extend amination to the synthesis of secondary and tertiary amines. Herein, for the first time, it is shown that AmDHs are capable of accepting other amine donors, thus giving access to enantioenriched secondary amines with conversions up to 43 %. Surprisingly, in several cases, the promiscuous formation of enantiopure primary amines, along with the expected secondary amines, was observed. By conducting practical laboratory experiments and computational experiments, it is proposed that the promiscuous formation of primary amines along with secondary amines is due to an unprecedented nicotinamide (NAD)-dependent formal transamination catalysed by AmDHs. In nature, this type of mechanism is commonly performed by pyridoxal 5′-phosphate aminotransferase and not by dehydrogenases. Finally, a catalytic pathway that rationalises the promiscuous NAD-dependent formal transamination activity and explains the formation of the observed mixture of products is proposed. This work increases the understanding of the catalytic mechanism of NAD-dependent aminating enzymes, such as AmDHs, and will aid further research into the rational engineering of oxidoreductases for the synthesis of α-chiral secondary and tertiary amines.

Full text of DOI:10.1002/cbic.201800626

Accepted Article
Title: Mechanistic Insight into the Catalytic Promiscuity of Amine
Dehydrogenases: Asymmetric Synthesis of Secondary and
Primary Amines
Authors: Vasilis Tseliou, Marcelo F. Masman, Wesley Böhmer, Tanja
Knaus, and Francesco Mutti
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: ChemBioChem 10.1002/cbic.201800626
Link to VoR: http://dx.doi.org/10.1002/cbic.201800626
A Journal of
www.chembiochem.org

10.1002/cbic.201800626
ChemBioChem
Mechanistic Insight into the Catalytic Promiscuity of
Amine Dehydrogenases: Asymmetric Synthesis of
Secondary and Primary Amines
Vasilis Tseliou,a Marcelo F. Masman,a Wesley Böhmer,a Tanja Knaus,a and
Francesco G. Mutti*,a
Van ‘t Hoff Institute for Molecular Sciences, HIMS-Biocat, University of Amsterdam, Science
Park 904, 1098 XH, The Netherlands.
Corresponding Author: f.mutti@uva.nl
1
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
Abstract
Biocatalytic asymmetric aminations of ketones using amine dehydrogenases (AmDHs) or
transaminases, are efficient methods for the synthesis of α-chiral primary amines. A major
challenge is the extension of the amination to the synthesis of secondary and tertiary amines.
Herein, we show for the first time that AmDHs are capable of accepting other amine donors, hence
giving access to enantioenriched secondary amines with conversions up to 43%. Surprisingly, in
several cases, we observed the promiscuous formation of enantiopure primary amines along with
the expected secondary amines. By conducting practical laboratory experiments and computational
experiments, we propose that the promiscuous formation of primary amines along with secondary
amines is due to an unprecedented nicotinamide (NAD)-dependent formal transamination
catalysed by AmDHs. In nature, this type of mechanism is commonly performed by pyridoxal 5’-
phosphate (PLP) aminotransferase and not by dehydrogenases. Finally, we propose a catalytic
pathway that rationalises the promiscuous NAD-dependent formal transamination activity and
explains the formation of the observed mixture of products. This work increases our understanding
on the catalytic mechanism of NAD-dependent aminating enzymes such as AmDHs and will aid
further research on the rational engineering of oxidoreductases for the synthesis of α-chiral
secondary and tertiary amines.
2
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
Introduction
2
3
4
5
6
7
Catalytic enzyme promiscuity is the ability of an enzyme to catalyse alternative chemical reactions
by often following catalytic mechanisms that are different from the natural one.[1] Although
catalytic enzyme promiscuity has been the object of intensive investigation for two decades, the
discovery of new promiscuous activities still proceeds at regular pace as witnessed by recent
publications.[2] Notably, these new biocatalytic activities have been frequently exploited in
chemical synthesis.
8
9
Amine dehydrogenases (AmDHs) catalyse the reductive amination of carbonyl compounds at the
expense of aqueous ammonium/ammonia buffer. All the known AmDHs so far were obtained
mainly by protein engineering starting from α-amino acid dehydrogenases such as the leucine
dehydrogenases from Bacillus stereothermophilus,[3] Exiguobacterium sibiricum,[4] Lysinibacillus
fusiformis[5] and Bacillus sphaericus,[5] as well as the phenylalanine dehydrogenases from Bacillus
badius,[6] Rhodococcus sp. M4 [7] and Caldalkalibacillus thermarum. [8] In all these cases, the
mutations of the highly conserved lysine and asparagine residues in the active site, which interact
with the two oxygen atoms of the carboxylic moiety of the natural substrate, were essential for
switching the substrate specificity from α-keto carboxylic acids to ketones.[3, 6-8] A few native
amine dehydrogenases have been also identified from Petrotoga mobilis, Fervidobacterium
nodosum, Clostridium sticklandii, Dethiosulfovibrio peptidovorans, Staphylothermus hellenicus,
and Thermosediminibacter oceani.[9] Several works on the asymmetric biocatalytic reductive
amination using the above mentioned amine dehydrogenases (AmDHs) have been published very
recently using isolated enzymes, immobilised enzymes or whole cells biocatalysts.[3-10] In
particular, our group has shown that AmDHs are efficient catalysts for the reductive amination of
prochiral ketones (i.e. TONs > 103) and their substrate scope covers already a respectable range of
10
11
12
13
14
15
16
17
18
19
20
21
22
23
3
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
structurally diverse substrates.[10d] Biocatalytic reductive amination is also possible with imine
reductases (IReds), which catalyse naturally the asymmetric reduction of cyclic imines.[11] In fact,
during the past few years, various groups have shown that IReds are also capable of catalysing the
reductive amination of non-cyclic imines although with modest efficiency.[12] In a recent study,
the reductive amination between ketones and small aliphatic and benzylic amines was carried out
with a novel dehydrogenase from Aspergillus o. (AspRedAm).[13] The enzyme was classified as a
“reductive aminase (RedAm)”. In a follow-up study two additional RedAms were characterised.[14]
Finally, a previous patent from Codexis reported a reductive aminase activity from a library of
variants originated from an opine dehydrogenase.[15]
10
11
12
13
14
15
16
17
Despite differences in names and apparent reactivity, the catalytic mechanisms of IReds, AmDHs
and RedAm are closely related as the actual mechanism is essentially based on the hydride transfer
from the nicotinamide coenzyme (NADH or NADPH) to the in-situ formed imine intermediate or
the already existing cyclic imine in solution. Thus, we hypothesised that the reactivity of AmDHs
may not be restricted only to ammonia as amine donor as previously believed and reviewed
elsewhere.[11b] That was also suggested by a number of oxidoreductases present in nature such as
the opine dehydrogenases, which are capable of generating secondary amines by the coupling of
an α-amino acid with an α-keto acid.[16]
18
19
20
21
22
23
Herein, we show for the first time that the reactivity of AmDHs is indeed not limited to ammonia
as the amine donor. The two AmDHs, considered in the present study, can accept other amine
donors to enable the formation of secondary and tertiary amines. Surprisingly, in some cases, the
reaction was accompanied by the promiscuous formation of primary amines along with the
expected secondary and tertiary amines. With the aim of explaining such unprecedented
promiscuous activity as well as the stereoselective outcome of the reaction, we performed a study
4
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
based on an accurate combination between practical laboratory experiments and computational
studies. Consequently, herein we also propose a catalytic cycle for amine dehydrogenases, which
explains the promiscuous formation of all products (i.e. secondary and primary amines) as well as
the enantiomeric composition of the reaction mixture.
5
Results and discussion
6
7
8
9
Screening of carbonyl compounds and amine donors. The reactivities of Rs-AmDH
(originated by enzyme engineering from the phenylalanine dehydrogenases from
Rhodococcus sp. M4)[7, 10d] and of Ch1-AmDH[10b, 10d] (a chimeric enzyme obtained via
domain shuffling of first generation variants) were assayed with eleven different amine
10 donors, which also constituted the buffer system. Four different types of carbonyl
11 compounds were selected: aliphatic ketones, aromatic ketones, aliphatic aldehydes and
12 aromatic aldehydes (Scheme 1; for the complete list of carbonyl compounds and amine
13 donors, see Figure S1 and Figure S2). We conducted the initial biocatalytic reactions as
14 follows: carbonyl compound (10 mM), amine/aminium formate buffer (1 M, pH 8.5), N-
15 terminal His6-tagged Rs-AmDH (>99% purity, 103 µM, 1 mol%) or N-terminal His6-
16 tagged Ch1-AmDH (>99% purity, 92 µM, 0.9 mol%), NAD+ (1 mM) and N-terminal His6-
17 tagged Cb-FDH (>99% purity, 24 µM, 0.2 mol%; for recycling of NADH), at ambient
18 temperature, for 48 h.
19
5
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
Scheme 1 Biocatalytic reductive amination of carbonyl compounds (1-6, 10 mM) with different amine donors
b-h) catalysed by N-terminal His6-tagged amine dehydrogenases (AmDHs). The amine donor was also used
(
as buffer species in water for the reaction (1 M, pH 8.5, formate as counter-anion). The nicotinamide
coenzyme (NAD+, 1 mM) was recycled by reduction with formate, which is catalysed by N-terminal His6-
tagged formate dehydrogenase from Candida boidinii (Cb-FDH). The amination reaction generated the
following secondary or tertiary amines (1b-d, 2b, 3b, 4b, 5e, 6b-c, 6f-h) along with the unexpected formation
of the following structurally related primary amines (1-3a). For full experimental screening, see SI, Table S1.
9
10 Under the mentioned reactions conditions, both tested AmDHs produced secondary or
11 tertiary amines starting from ketones and aldehydes (for complete list of products, see
12 Figure S3). Selected results are reported in Table 1, whereas all screening results are
13 reported in SI section 4 (Table S1). Ch1-AmDH catalysed the reductive amination between
14 4’-fluorophenylacetone (
1
) and methylamine (
b
), whereas Rs-PhAmDH accepted ketone
1
15 in combination with ethylamine (
c
) and cyclopropylamine (
d) (Table 1, entries 1, 3, 4). For
16 these reactions, we were surprised to identify also the formation of the structurally related
6
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
primary amine, 4’-fluoroamphetamine (1a), as a second product. Notably, 1a was obtained
in enantiopure form in all the cases (ee > 99%, ). Commercial reagent grade solution of
methylamine contains only negligible traces of ammonia, whereas ethylamine and
cyclopropylamine do not contain ammonia at all. In addition, the amination of at 4-6 M
concentration of methylamine buffer ( ) resulted in lower conversion into 1a compared to
the same reaction performed at 3 M of (Figure 1B). Hence, we ruled out the possibility
R
1
b
b
that the primary amine could have been formed via the enzymatic reaction between the
ketone and the ammonia present, eventually, as an impurity. Due to the complexity and the
elevated polarity of the components in the reaction mixtures, we succeeded in measuring
10 the enantiomeric excess of the secondary amine products by chromatography in the case of
11 compounds 1b and 1d 1b was obtained in enantioenriched form (ee 72% , Table 1, entry
.
R
12 1), whereas 1d in nearly racemic form (Table 1, entry 4). In contrast, the other product -
13 primary amine 1a - was obtained in enantiopure form as previously observed (for
14 comparison, Table 1, entries 1 and 4).
15 Ch1-AmDH was the most active enzyme for the amination of 2-hexanone (
2) and 2-
16 heptanone ( ) with (Table 1, entries 5 and 6). Even in this case, the structurally related
3
b
17 primary amines 2-aminohexane (2a) and 2-aminoheptane (3a) were obtained with perfect
18 ee. These results confirm that the formation of primary amines occurs during the catalytic
19 cycle of the AmDHs. Conversely, the amination of aldehydes using the AmDHs and
20 different amine donors under our initial reaction conditions (1 M of amine/aminium buffer,
21 pH 8.5), proceeded with perfect chemoselectivity to afford the expected secondary amines
22 as the sole product up to 43% conversion (Table 1, entry 7-11). As a proof of principle, the
23 amination was run also with secondary amines as amine donor (e.g. dimethylamine, ethyl-
7
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
methylamine), leading to tertiary amines albeit with poor conversions (Table 1, entry 12
and 13).
3
4
5
Table 1. Reductive amination reactions performed by AmDHs for the synthesis of secondary and tertiary
amines. Standard reaction conditions: substrate (10 mM), N-terminal His6-tagged Ch1-AmDH (>99% purity,
6
92 µM, 0.9 mol%) or N-terminal His6-tagged Rs-AmDH (>99% purity, 103 μM, 1 mol%), N-terminal His6-
7
8
9
tagged Cb-FDH (>99% purity, 24 µM, 0.2 mol%) and buffer constituted by amine donor with formate as
counter-anion (1 M, pH 8.5). The reactions were performed at 30 °C, for 48 h using an orbital shaker at 170
rpm.
Secondary
Amine
Primary Amine
(1‐11a)
Conv.
ee (%) c
(%) b
Amine
Entry Substrate donor
AmDH
Conv.
ee (%) c
(%)b
1
2
1
1
1
1
2
3
4
5
6
6
6
6
6
b
b
c
Ch1
Ch1
Rs
15
72 (R)
64 (R)
N.m.
1 (R)
N.m.
N.m.
N.a.
8.
6
>99 (R)
>99 (R)
>99 (R)
>99 (R)
>92 (R)d
>99 (R)
N.a.
40 a
4
3
26
4
d
b
b
b
e
b
c
Rs
11
5
26
5
Ch1
Ch1
Rs
4
6
32
24
41
29
40
43
3 e
8 e
26
7
N.d.
N.d
N.d
N.d
N.d
N.d
N.d
8
Ch1
Rs
N.a.
N.a.
9
N.a.
N.a.
10
11
12
13
Rs
N.a.
N.a.
f
Rs
N.a.
N.a.
g
h
Rs
N.a.
N.a.
Rs
N.a.
N.a.
a
+
b,c
10
11
12
Concentration of MeNH2/MeNH3 buffer was increased to 6 M.
For analytical determination of
conversions and ees, see SI paragraphs 8-10. d Unable to determine with precision, due to experimental noise
of the GC traces. e Measured by GC-MS.
13
14
8
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
Influence of the enzyme and amine donor concentration. For further studies, we selected
the amination of 4’-fluorophenylacetone ( ) with methylamine ( ) as test reaction. In the
1
b
first set of experiments, we kept the buffer concentration constant (1 M, pH 8.5) and we
varied the concentration of N-terminal His6-tagged Ch1-AmDH (>99% purity, 4.5–91 μΜ,
equal to 0.045-0.9 mol%). As expected, increasing the enzyme concentration affected
positively the overall conversion (Figure 1A, Table S2). Interestingly, the ratio between the
formation of secondary (1b) and primary (1a) amines also varied depending on the enzyme
concentration from 1:1.2 (at AmDH 4.5 µM) to 1:1.9 (at AmDH 91 µM).
In a follow-up experiment (Figure 1B), we fixed the concentration of N-terminal His6-
10 tagged Ch1-AmDH (>99% purity, 91 µM, 0.9 mol%) and we increased the concentration of
+
11 the CH3NH2/CH3NH3 buffer (200 mM – 6 M). Increasing the buffer concentration
12 correlated with a consistent increase in the formation of 1b. At 6 M buffer concentration,
13 1b was finally obtained in 40% conversion (Figure 1B and Table 1 entry 2; for details see
14 SI, Table S3). The side-reaction that leads to the primary amine product showed a
15 continuous increment of conversion of 1a (11% conversion) up to a buffer concentration of
16 3 M. Surprisingly, more elevated concentration of buffer (4, 5, 6 M) reduced the formation
17 of 1a (6% at 6 M buffer concentration). Hence, the ratio between 1a and 1b further
18 increased to 1:6.6. We also monitored a possible variation in the stereoselective outcome
19 of the reaction. The enantiomeric excess of the secondary amine product 1b was partially
20 affected with a reduction from 72% ee (1 M buffer) to 64% ee (6 M buffer). In contrast, the
21 optical purity of the primary amine remained perfect (>99%, Table 1 entries 1 and 2). In all
22 cases, the
R
-configured amine (secondary or primary) was the favoured enantiomer.
9
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
Figure 1. (A) Dependence of enzyme concentration (>99% purity, 4.5-91 μM, equal to 0.045-0.9 mol%) at fixed
concentrations of methylamine buffer (1 M) for the conversion of 4’-fluorophenylacetone (1) into primary (1a, blue
columns) and secondary (1b, grey columns) amines catalysed by N-terminal His6-tagged Ch1-AmDH. Catalytic
amount of NAD+ (1 mM) was applied and recycled using N-terminal His6-tagged Cb-FDH (>99% purity, 24 μM, 0.2
mol%). Error bars represent the standard deviation calculated from three independent experiments. (B) Dependence
of CH3NH2 concentration (200 mM - 6M) at a fixed concentration of N-terminal His6-tagged Ch1-AmDH
4
5
6
7
8
(>99% purity, 91 μM, 0.9 mol%) for the conversion of 4’-fluorophenylacetone (1) into primary (1a, purple
9
columns) and secondary (1b, grey columns) amines catalysed by Ch1-AmDH. Catalytic amount of NAD+
(1mM) was applied and recycled using N-terminal His6-tagged Cb-FDH (>99% purity, 24 μM, 0.2 mol%).
Error bars represent the standard deviation from three independent experiments. Enantiomeric excess (ee) for
1a (blue) and 1b (grey) are depicted with circles connected with a solid line.
10
11
12
13
10
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
Initial biochemical and computational studies towards the understanding of the reaction
mechanism. Both Ch1-AmDH and Rs-AmDH showed to possess a general perfect
stereoselectivity (i.e. “R-selectivity”) for the reductive amination of prochiral ketones with
ammonia.[10d] However, different points remained unclear at this stage: 1) the molecular
discriminants that determine a productive combination between the ketone/aldehyde and the amine
donor; 2) the formation of secondary amines with moderate steroselectivity; 3) the unexpected
formation of primary amine in enantiopure form.
Therefore, we performed several computational analysis in the attempt of elucidating these
points. First, we generated several models of Rs-AmDH in complex with NADH and
10 different protonated imine intermediates (i.e. carbonyl compounds: 4’-fluorophenylacetone
11 ), hexanal ( ), heptanal ( ), phenylacetaldehyde ( ), 4-phenyl-butan-2-one ( ),
12 acetophenone (10); amine donors: ammonia ( ), methylamine ( ), ethylamine ( ) and
13 propylamine ( )). Details on the creation of these models are reported in the experimental
(
1
4
5
6
9
a
b
c
n-
f
14 section. Based on the reported catalytic mechanism of the parent wild-type phenylalanine
15 dehydrogenase from Rhodococcus sp. M4,[17] we analysed the models considering two
16 crucial parameters as depicted in Figure 2: 1) the distance between the departing hydride
17 of NADH and the prochiral carbon of the imine intermediate (herein referred as “distance
18 1”); 2) the distance between the negatively charged oxygen atom of the terminal carboxylic
19 group of Asp118 of Rs-AmDH and the hydrogen of the positively charged iminium group
20 of the intermediate (herein referred as “distance 2”).
21
11
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
Figure 2. Simplified schematic view of the productive binding mode for the iminium intermediate bound in
the active site of the AmDH. For clarity, only the nicotinamide coenzyme in its reduced form (NADH) and
the highly conserved amino acid residue Asp 118 (numbering of Rs-AmDH) of the enzyme are depicted. The
actual spatial positions and orientations of NADH, substrate and Asp 118 in the active site were retained
according to our calculated model. The iminium intermediate (1a*) is depicted. 1a* is generated by
interaction between 4’-fluorophenylacetone (1) and ammonia (a). The crucial distances 1 and 2 are depicted
with a solid black line.
9
10 Our analysis revealed that highest reactivity (i.e. conversion) for the reductive amination is
11 achieved when both above-mentioned distances have an optimal value. The sum of the van
12 der Waal’s radius [18] of carbon and hydrogen can be considered as the threshold distance,
13 which is ca. 2.9 Å (ꢀꢁꢂꢃꢄ ꢅ 1.2 Å, ꢀꢆꢂꢃꢄ ꢅ 1.7 Å; for details see experimental part). Thus,
14 an optimal distance must be around or moderately below 2.9 Å. For instance, the reductive
15 amination of
1 with ammonia (a) catalysed by Rs-AmDH was reported to proceed quickly
16 and in quantitative manner.[10d] Our model for this reaction with Rs-AmDH and the related
17 iminium intermediate (1a*) displays indeed an optimal “distance 1” of 2.7 Å and an optimal
18 “distance 2” of 1.9 Å (see SI, Figure S5A). Conversely, the same reaction between
1
and
12
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
ethylamine (
c
) affords only 4% conversion (Table 1, entry 3). In our models for the latter
reaction considering the related iminium intermediate 1c*, the “distance 1” is still ideal (2.9
- 3.0 Å), whereas the “distance 2” is significantly larger (4.8 - 6.0 Å), (see SI, Figure S5B
and S5C). Another interesting example is the reductive amination between acetophenone
(
10) and either ammonia (a) or ethylamine (b). The reaction with ammonia (a) as amine
donor afforded only a moderate conversion of 34% after 48 h;[10d] our model for this
reaction with the related iminium intermediate 10a* (see SI, Figure S5D) displays an
optimal “distance 1” of 3.0 Å, but “distance 2” is slightly elongated (3.3 Å). Conversely,
the reaction between acetophenone (10) and ethylamine (b) did not occur at all (see SI
10 Table S1); in fact, although one of our models for this reaction with iminium intermediate
11 10b* displays an optimal “distance 2” of 3.00 Å, “distance 1” is much larger (4.0 Å; see
12 SI, Figure S5E). Finally, 4-Phenyl-butan-2-one (
9) and ammonia (a) react quantitatively
13 within short reaction time.[7, 10d] Our model for this reaction with iminium intermediate 9a
*
14 shows both perfect “distance 1” and “distance 2” of 2.8 Å and 3 Å, respectively (see SI
15 Figure S5G). In contrast, the reaction between 4-phenyl-butan-2-one ( ) and ethylamine
16 ) did not occur. Coherently, our models for this reaction with iminium intermediate 9b*
9
(
b
17 show a longer “distance 1” ranging from 3.4 Å and 3.5 Å (see SI, Figure S5H and S5I). A
18 similar study conducted with aldimine as intermediates provided an analogous trend (see
19 SI, Figure S6). In conclusion, our analysis revealed that the distance between the prochiral
20 iminium carbon of the intermediate and the hydride of NADH (Figure 2, distance 1) is the
21 parameter of primary importance for the reaction to occur. Nevertheless, the residue
22 Asp118 (numbering of Rs-AmDH) appears to play a role for achieving relevant turnovers
23 (Figure 2, distance 2). However, this analysis permits only to rationalise which are the
13
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
crucial parameters that enable the reduction of the iminium intermediate bound in the active
site of an AmDH. Otherwise, the reductive amination of a prochiral ketone with different
amine donors (b-e) catalysed by AmDHs generates complex reaction mixtures containing
secondary amines (1b-d, 2b, 3b, 4b, 5e) with moderate optical purity and the unexpected
primary amines in enantiopure form (1-3a). This observation suggests the existence of, at
least, two stereocomplementary productive binding configurations for the reduction of
secondary iminium intermediates.
On the other hand, as the unexpected primary amines (1-3a) were always obtained in
optically pure form (>99% ee, R), the promiscuous formation of these products was clearly
10 catalysed by the AmDHs. At this stage, we envisioned three possible reaction mechanisms
11 that could explain the formation of primary amines (Scheme 2).
12 The first option (Scheme 2a) foresees an oxidative deamination of the amine donor (b-d; d
13 in the example), which is catalysed by the AmDH. The free ammonia generated in the first
14 step may serve as the amino donor in a subsequent reductive amination that is always
15 catalysed by the AmDH. The second option (Scheme 2b) for explaining the formation of
16 the enantiopure primary amine as side-product foresees an unlikely promiscuous hydrolytic
17 activity of AmDHs. In fact, a polarised water molecule in the active site of AmDH and
18 possessing a proper orientation, may attack the sp3-carbon in α-position to the nitrogen of
19 the ketiminium intermediate (1d* in the example). Although such a hydrolytic step is
20 chemically unlikely, it cannot be excluded beforehand because of the particular catalytic
21 environment in the active site of the AmDH in which a polarised water molecule is normally
22 involved in catalysis.[17] If such a hydrolytic step occurred, a subsequent hydride transfer
23 from NADH would generate a primary amine. Furthermore, only in the case of cyclic
14
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
intermediates such as 1d*, the same nucleophilic attack of a water molecule may also
provoke the opening of the cyclopropyl ring to give an aminol upon reduction (not depicted
in Scheme 2b). However, this route is also incompatible with the observed mixture of
products. The third option (Scheme 2c) is an unprecedented formal transamination reaction.
In nature, transamination reactions are catalysed by pyridoxal 5’-phosphate (PLP)
dependent aminotransferase through a ping-pong mechanism.[19] However, an alternative
NAD+/NADH redox-mediated mechanisms is conceivable. According to this hypothesis,
the key catalytic step would be the isomerisation of the ketiminium intermediate (1d* in
the scheme 2c) by the action of NAD+/NADH, to give the other ketiminium intermediate
10 1d**
.
1st option - in situ NH3 formation followed by reductive amination of ketone substrate
NH3
H2O
NAD+
NAD+
NADH
NADH
a
+
O
NH2
NH2
then
+
NH3
A)
+
H2O
O
NH2
F
F
a
1
(R)-1a
d
2nd option - C(sp3)-N-iminium water attack followed by reductive amination
NH2
OH
NAD+
NADH
d
H2O
H2O
B)
+
O
NH+
NH2
NH2
F
F
F
F
water molecule
attack
1
(R)-1a
1a*
1d*
3rd option - NAD+/NADH redox-mediated formal transamination
NH2
O
NADH NAD+
NAD+ NADH
H2O
H2O
d
C)
NH+
O
NH+
NH2
(R)-1a
NH
F
F
F
F
F
1
1d**
1d
1d*
11
12
13
Scheme 2. Possible mechanistic explanations for the promiscuous formation of enantiopure primary amine
as by-product.
15
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
One could already exclude the first option (Scheme 2a) by considering the analysis of the
composition of the reaction mixture after the biocatalytic reductive amination. In fact, the
concentration of amine donor (b-d) remained constant at nearly 1 M concentration from
the beginning until the end of the reaction. This observation indicates a negligible, if any at
all, formation of ammonia during the course of the reaction. On the other hand, our research
group as well as other groups have shown that a large excess of ammonia (ca. 0.2-1 M) is
required in order to drive the reductive amination of ketone substrates (15-50 mM) in
aqueous buffer to a significant extent.[6-8, 10d, 10i] Moreover, AmDHs such as Ch1-AmDH
show a KM value of ca. 1 M for ammonia.[10b] Similarly, with the only exception of the
10 glutamine dehydrogenases from bovine liver, frog liver and Clostridium SB4, the parent
11 wild-type amino acid dehydrogenases are also characterised by elevated KM values for
12 ammonia (20-500 mM).[20] Consequently, the biocatalytic reductive amination with
13 ammonia is kinetically disfavoured already at significant concentration of ammonia as
14 amine donor. Nonetheless, we investigated this possibility by incubating Rs-AmDH (102.8
15 µM) with cyclopropylamine (
16 at 30 °C (SI, section 6.1). The possible formation of ammonia was determined indirectly
17 by analytical quantification of the consumption of . As expected, the concentration of
18 remained constant during 48 h hence excluding any detectable formation of ammonia by
19 oxidative deamination of catalysed by Rs-AmDH.
d
, 50 mM) and NAD+ (60 mM) in phosphate buffer at pH 8.5,
d
d
d
20 The second option foresees the elimination of an alcohol (Scheme 2b; cyclopropanol in the
21 example) during the possible catalytic cycle of the enzymatic reductive amination between
22 ketones (1 in the example) and amine donor (d, cyclopropylamine in the example). By
23 careful monitoring of the reactions, the formation of cyclopropanol as by-product was never
16
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
observed. Analysis was accomplished by comparing the GC-MS chromatogram (i.e.
retention times and fragmentation patterns) of authentical cyclopropanol as reference
compound against the GC-MS chromatograms for the enzymatic reductive amination (for
details, see SI section 6.2). Crucially, control experiments also revealed that cyclopropanol
(used as reference compound) is stable in the reaction buffer (1M, pH 8.5) at 30 °C and
within the 48 h reaction time (SI, section 6.2).
The last option foresees a promiscuous formal transamination via NAD+/NADH redox-
mediated iminium isomerisation (Scheme 2c). Firstly, we ascertained that the presence of
the nicotinamide coenzyme was crucial for a possible formal transamination. Hence, Rs-
10 AmDH (102.6 µM) and ketone
1 (10 mM) were incubated in aqueous buffer of amine d
11 (1M, pH 8.5), but in absence of NADH. As expected, we did not observe any formation of
12 products (for details, see SI section 6.3). Then, in order to exclude any possible classical
13 transamination reaction that is enabled by pyridoxal 5’-phosphate (PLP) as cofactor, we
14 also repeated the same experiment but in presence of exogenous PLP (0.5 mM). Even in
15 this case, the formation of any product was not observed. Finally, we undertook further
16 experiments for proving the promiscuous enzymatic activity as depicted in Scheme 3c. In
17 the first set of experiments, we incubated chemically synthesised racemic
N-(1-(4'-
18 fluorophenyl)propan-2-yl)cyclopropanamine (1d) in aqueous buffer in presence of NAD+
19 (varied concentration from 2 mM to 20 mM) and Rs-AmDH. In this way, we aimed at
20 creating a dynamic equilibrium between all possible oxidative deamination pathways and
21 reductive amination pathways (for schematic details, see SI section 6.4). Crucially, control
22 experiments (i.e. without AmDH) showed that 1d is stable under these reaction conditions
23 and no reaction was observed. In contrast, the incubation of 1d and NAD+ with the AmDH
17
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
produced measurable amounts of primary amine 1a (for details see SI section 6.4, Table
S6). The concentration of 1a detected was typically low (ca. 1% conversion) because 1a is
also in equilibrium with
1 (ca. 7% conversion), the latter species being favoured because
of the aqueous environment (schematic details in SI section 6.4). Thus, with these
experiments, we proved that the Rs-AmDH converts the secondary amine rac-1d into the
primary amine (
R
)-1a, which is the second part of the mechanism depicted in Scheme 3c.
1d catalysed by Rs-
A further experiments of kinetic resolution (SI, section 6.4) on rac
-
AmDH and using NAD-oxidase (NOx) for NAD+ recycling was also performed. Detailed
analysis of the composition of the reaction mixture revealed that the enzyme is indeed
10 capable of distinguishing between the two enantiomeric forms of 1d. In fact, the
11 enantiomeric excess of remaining substrate 1d increased during the time. After 48 h
12 reaction time, the remaining 1d was ca. 70% whereas its ee was ca. 16% (SI, Table S7). If
13 perfect kinetic resolution had occurred, the remaining ee should have been instead 30% (i.e.
14 70% remaining 1d equal to 20% of 1st enantiomer and 50% of 2nd enantiomer; hence
15 theoretical ee = (50-20) = 30%). Thus, these data clearly demonstrate that, albeit one
16 enantiomer of the secondary amine 1d is preferred, both enantiomers can be accepted by
17 Rs-AmDH.
18 In conclusion, considering all the results, the side-product formation of enantiopure primary
19 amines is originated by a non-classical promiscuous transamination activity that is
20 mediated by the nicotinamide coenzyme. The expected ketone by-products such as
21 cyclopropanone, or formaldehyde, or acetaldehyde could not be observed because of their
22 known elevated instability and reactivity in solution.
23
18
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
Elucidation of the stereoselective properties of AmDHs with the aid of computational
studies. The models of Rs-AmDH, previously discussed, served as starting point towards
a deeper understanding on how a given substrate interacts with the active-site residues of
the enzyme while adapting its putative reactive pose(s). This information was used in the
subsequent molecular modelling of the Ch1-AmDH described in this paragraph. In fact, for
the in-depth computational analysis aimed at elucidating the experimental observations
regarding the stereoselectivity of the AmDHs, we selected the reaction between ketone (
and methylamine ( ), which is catalysed by Ch1-AmDH. This choice was taken because
data of conversion and enantiomeric excess for the reaction between and were available
1)
b
1
b
10 (Table 1, entry 1 and 2). Therefore, direct comparison between computational data and
11 experimental laboratory data was possible.
12 Since the crystal structure of Ch1-AmDH is not publically available, the initial step was to
13 generate a high-quality homology model of this enzyme. The model of Ch1-AmDH was
14 generated in two steps. Firstly, an “exploratory homology modelling run” was executed in
15 order to determine what was (were) the best suitable template(s) for this enzyme. Secondly,
16 a “productive homology modelling run” was executed only considering the “best”
17 template(s) as candidate(s), (for details, see experimental part and SI section 5.2). It is
18 noteworthy that our homology model was created with the enzyme in its reactive
19 conformation (i.e. “closed conformation”) in which the nicotinamide coenzyme is bound
20 in the active site. That is an important prerequisite for performing molecular docking
21 simulation with these enzymes as described in this work. In contrast, the available crystal
22 structure of opine dehydrogenases (ODHs), which also catalyse the formation of secondary
23 amine functionalities following a formally similar mechanism, are reported in the non-
19
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
reactive conformation (“open conformation”) in which the nicotinamide coenzyme is
unfortunately absent.[16c]
3
We initially performed Molecular Docking simulations using the obtained model of Ch1-AmDH
as target with the iminium intermediates 1a* and 1b* as ligands (i.e. 1a* and 1b* are generated
by the interaction between the ketone 1 with ammonia (a) or methylamine (b), respectively). The
aim was to obtain models of the possible reactive conformations that explain the different
stereoselectivity of the reaction, considering the formation of the R-configured enantiopure
primary amine by reaction of 1 with a [10b, 10d] or both secondary amine enantiomers by reaction of
1 with b. Through the analysis of the structures obtained with molecular docking simulations, we
were able to create the putative pro-R and pro-S binding modes for both iminium intermediate:
1a* and 1b*. Comparing the models of Ch1-AmDH possessing 1a* bound either in pro-S or in
pro-R reactive conformation, the calculated binding energies as well as the calculated distances
between the departing hydride of NADH and the prochiral carbon of the iminium intermediate
(distance 1 as defined in Figure 2) were similar. A similar scenario was also observed in the models
of Ch1-AmDH possessing 1b* bound either in pro-S or in pro-R reactive conformation. For this
reason, we executed a set of Molecular Dynamics simulations with the aim of “relaxing” the
docked enzyme-substrate complexes and, therefore, allowing for a more accurate re-evaluation of
the reactive conformations for 1a* and 1b* (Figure 3a-b and SI section 5.2 for details). After MD
relaxation, it was evident that Ch1-AmDH does not tolerate well 1a* in the pro-S binding mode
(see SI section 5.2, Figure S6b and S7b); hence, the pro-R binding mode is highly preferred for
1a*. After MD conformational relaxation, the following scenario was observed. Figure 3c and 3d
depict: 1) the average relative binding energies for the binding of the imine intermediates (pro-S
1a*, pro-R 1a*, pro-S 1b* and pro-R 1b*) in the active site of Ch1-AmDH; 2) the average distances
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
20
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
between the departing hydride of NADH and the pro-chiral carbon of the imine intermediates
(distance 1 as previously defined). The pro-S binding conformation of 1a* is the less favoured
compared to all the other binding conformations (Figure 3c) that appears energetically similar.
More instructive is the analysis of the behaviour of the distance 1 over the time of the simulations.
The substrate intermediate 1a* in its pro-R conformation showed an average distance 1 of
2.90±0.20 Å, while in its pro-S conformation moved considerably above the threshold distance
with an average of 3.74±0.64 Å. A slightly similar behaviour was observed for the substrate
intermediate 1b*. 1b* showed an average distance 1 value of 2.69±0.11 Å in its pro-R
conformation, while it showed a higher average distance value of 3.29±0.36 Å in its pro-S
conformation. Nevertheless, it is important to note that distance 1 for pro-S 1b* (3.29±0.36 Å) is
relatively shorter than the same distance for pro-S 1a* (3.74±0.64 Å). This finding indicates that
Ch1-AmDH must tolerate better the pro-S conformation of 1b* than the pro-S conformation of
1a*. On the other hand, 1b* in the pro-R orientation is still the preferred binding mode by Ch1-
AmDH. That can be observed further in Figure 3d, in which approximately 94% of the MD
snapshots (within the simulation time) represent the pro-R 1b* intermediate in the active site of
Ch1-AmDH with a distance between its prochiral carbon atom and the attacking hydride of NADH
(distance 1 as previously defined) below the threshold of 3 Å (required for hydride shift). Thus,
there is an elevated probability that the hydride shift from NADH to 1b* occurs if 1b* is bound in
its pro-R binding conformation. Conversely, the pro-S binding conformation of 1b* is less
favourable for hydride shift from NADH because only approximately 32% of the MD snapshots
showed a distance 1 below the threshold of 3 Å.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
On the other hand, the relative binding energies of 1b* on pro-R and pro-S conformations
respectively seem to be approximately the same (Figure 3c). We can therefore conclude that,
21
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
although tolerated, the pro-S configuration of 1b* has a lower probability to react than the pro-R
contra-part. The pro-S conformation of 1b* seems to be stabilised by hydrophobic interactions of
the methyl-group with residues M75, A298 and C310. The pro-R binding mode of 1b* is stabilised
as already known from literature for wild-type AADHs.[17] This stabilisation involves the
previously discussed highly conserved Asp residue (numbering 125 for Ch1-AmDH) in a similar
way as depicted in Figure 2 for Rs-AmDH. The described enzyme/substrate contacts for Ch1-
AmDH are depicted in Figure 3e-f for both pro-R and pro-S binding modes. It is interesting to note
that the intermediate 1b* bound in its reactive pro-R binding mode was mainly observed assuming
the E configuration at the C=N double bond (Figure 3a and 3e). In contrast, the intermediate 1b*
bound in its reactive pro-S binding mode was mainly observed assuming the Z configuration
(Figure 3b and 3f). It is known for literature that N-alkylimines can isomerise in solution at ambient
temperature by tautomerisation and rotation.[21] In the case of intermediate 1b*, which bears non-
bulky substituents such as hydrogen or methyl groups, the E isomer is known to be favoured in
equilibrium in solution at room temperature.[22] Nonetheless, in our case, the enzyme selects the
preferred E or Z configuration of the intermediate in the active site.
3
4
5
6
7
8
9
10
11
12
13
14
15
22
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
Figure 3. Pro-chiral preferences of Ch1-AmDH. Representative relaxed molecular dynamics snapshots of the 1b* on
its pro-R (a) and pro-S (b) conformations. The substrate is shown in yellow, while the NADH is shown in green. For
clarity, all non-polar hydrogen atoms were not shown. Panel (c) depicts the relative binding energy for 1a* and 1b*
on their pro-R and pro-S configuration. All the binding energies are given as relative to the average binding energy of
1a* on its pro-R conformation. A lower value of relative binding energy means a less favourable binding. Only
those snapshots that showed the substrate on its “reactive conformation” were considered for binding energy
determination. Panel (d) depicts the distance from the departing hydride of the coenzyme (NADH) to the pro-chiral
23
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
carbon of the ligand (iminium) for 1a* and 1b* (in both pro-R and pro-S configurations). The number in between
brackets indicate the percentage of snapshots that contributed with substrates located under the distance threshold (3
Å). Panel (e) and (f) depict the observed enzyme/substrate contacts for 1b* on its pro-R and pro-S conformation,
respectively. Lines in grey indicate hydrophobic interactions, while the dashed green line indicate hydrophilic
interactions. The NADH cofactor is depicted behind the substrate in light-grey.
6
7
8
9
Additionally, a non-bonding dihedral angle (χ) was defined in order to describe the stereo-
binding mode of the substrate in the active site (Figure 4a). This dihedral angle was defined
following the Cahn–Ingold–Prelog priority rules between the three atoms bonded to the
10 pro-chiral carbon of the substrate and the hydride atom of the NADH cofactor. This
11 dihedral angle was determined for the
12 of the reaction, indicating that the
R
- and
S- configured products (both for 1a and 1b)
R
-configured product shows an average value of
13 63.85±0.20°, while the
14 Figure 4b shows the behaviour of the χ angle in function of the time for substrate
15 intermediates 1a* and 1b* and starting from both pro- and pro- binding conformations.
16 Notably, in the case where the intermediate substrate 1a* is bound in a pro- binding mode,
17 a switch to the pro- conformation occurred within the simulation time (Figure 5b; for
18 details see SI section 5.2 and Figure S7b; see also SI movie file “switch_pro-S_to_pro-
19 R_primary_imine.avi”). In particular, 50% of the simulation runs for pro- 1a* (3 out of 6)
20 showed this switch from pro- to pro- (Figure 5b, black line). An additional simulation
S-configured product shows an average value of -63.95±0.27°.
R
S
S
R
S
S
R
21 (run 5) also showed a switch in the conformation but, during this simulation, the substrate
22 moved out of the active-site. Therefore, we did not consider this simulation into the
23 calculation of the average values. Only in 2 cases out of 6, substrate 1a* maintained the
24 pro-
S
binding conformation (Figure 5b, green line). In contrast to the simulation with pro-
24
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
6
7
8
9
S
1a*, a conformational switch was never observed for simulations starting from either pro-
1a* or pro- 1b*or pro- 1b* (Figure 4; for details see SI section 5.2 and Figure S7a,c,d).
R
R
S
It is important to remark that the average value of the herein defined dihedral angle for pro-
1a* amongst simulation runs has only illustrative value, thus indicating great
S
conformational changes in the active site of Ch1-AmDH.
In summary, the molecular dynamic simulations provide an insight at molecular level on
the different stereoselective behaviour of the AmDHs in the reductive amination of
prochiral ketones with either ammonia or more complex amines. When ammonia is the
amine donor, the reactive pro-R conformation of the iminium intermediate is much more
10 favoured than the pro-
S
conformation explaining why the primary amine product is always
-configured form.[10d] In the event that a pro-
S
11 experimentally obtained in enantiopure
R
12 binding mode for the primary iminium intermediate is generated in the active site, a
13 conformational switch to pro- binding mode (Figure 4B; see also SI movie file
R
14 “switch_pro-S_to_pro-R_primary_imine.avi”) or even the release of the intermediate from
15 the active site is likely to occur. Conversely, when a primary amine is the amine donor for
16 the reductive amination, both pro-R and pro-S binding modes can be generated as reactive
17 conformations so that the secondary amine product is obtained in both enantiomeric forms.
25
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
Figure 4. Dihedral angle χ versus time for the Molecular Dynamics Simulations starting from pro-R 1a*, pro-S 1a*,
pro-R 1b* and pro-S 1b*. (a) Illustrative depiction of the χ angle for the R- and S-configured product of the reaction.
These values were used as reference for describing the hydride shift from NADH that would afford any of these
enantiomers. According to this definition, a positive value of the dihedral angle in the intermediate will lead
5
6
7
to the R-configured amine upon reduction. A negative value will lead to the S-configured amine upon
8
reduction. (b) The average variation of χ angle over the time is shown for the simulations of pro-R 1a* (purple), pro-
S 1a* (black or green), pro-R 1b* (orange) and pro-S 1b* (blue). For each system, the depicted line is the average of
six independent simulation runs (all simulation runs are reported in the SI, Figure S7). Only the simulation run number
5 for substrate 1a* in pro-S binding conformation was not considered for the average calculation due to fact that the
substrate moved away from the active site (for details, see SI, Figure S7b). For the sake of clarity of the depiction, the
error-bars of these average simulations have been omitted (for this detailed information see SI Figure S7). The average
χ values for the R-configured (blue) and S-configured products (purple) are also shown as dashed lines.
9
10
11
12
13
14
15
26
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
5
Proposed catalytic mechanism. Considering all the results obtained from practical experimental
laboratory and computational experiments, we postulated a biocatalytic cycle that illustrates the
formation of both secondary and primary amines (Scheme 3). The proposed cycle is adapted from
the catalytic mechanism of the phenylalanine dehydrogenases from Rhodococcus sp. M4 (the wild-
type parent of the variant Rs-AmDH).[17]
6
7
In the first step, the carbonyl compound and the amine donor generate the geminal amino alcohol
(I or I’) assisted by a protonated Lys and a deprotonated Asp residues from the AmDH active site.
Intermediate I and I’ can interconvert through a rotation around the carbon-carbon bond that
connects the prochiral C-sp2 of the ketone and the C-sp3 in its α-position (the latter connected to
substituent R1 in Scheme 3). This rotation appears to be the most probable one since our model
structures clearly show that the R1 substituent (e.g. in this study phenyl, phenylmethyl, alkyl) is
tightly accommodated in a hydrophobic cavity in the active site of the AmDH (Figure 5).
8
9
10
11
12
13
14
27
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
Figure 5. (a) Hydrophobic binding pocket of RsAmDH (b) Amino acid residues constituting the hydrophobic binding
pocket in the active site of Rs AmDH
3
4
Then, assisted by the same Lys residues, a water molecule is released from intermediate I or I’ to
generate the iminium intermediate II (pro-R binding mode) and II’ (pro-S binding mode),
respectively. From the analysis of the snapshots of our molecular docking simulations, it appears
that the iminium moiety of the intermediate II (pro-R) is stabilised further by a hydrogen bond
with the deprotonated aspartate residue (Figures 3a and 3e; Asp 118 in Rs-AmDH or Asp 125 in
Ch1-AmDH). In contrast, such a hydrogen bond was not observed with intermediate II’ since the
aspartate residue is too distant from the hydrogen of the iminium moiety. Intermediate II’ appears
to be stabilised by hydrophobic interactions of the methyl group with residues M75, A298, C310
(Figure 3b and 3f). As observed before, intermediate II assume the preferred E configuration
whereas intermediate II’ assumes the preferred Z configuration (in Scheme 3, C.I.P. priority: R1 >
R2). Although a conformational switch from pro-S to pro-R binding mode (or vice versa) of the
secondary imine intermediate was never observed within the time of our simulations, this event
can not be completely ruled out. Hence, this theoretically possible conformational binding switch
was depicted as dashed line in Scheme 3. In contrast, as previously described, such a
conformational binding switch from pro-S to pro-R (but not vice versa) is very likely to occur with
5
6
7
8
9
10
11
12
13
14
15
16
17
18 primary iminium intermediates (Figure 4b; see also SI movie file “switch_pro-S_to_pro-
19 R_primary_imine.avi”)
.
20
21
22
23
At this stage, hydride shift from NADH to intermediate II or II’ can occur to furnish the secondary
amine product in R-configuration or S-configuration, respectively. Nonetheless, the amine (R)-1b
obtained by reduction of intermediate II can be also subjected to a further promiscuous re-
oxidation in the active site by abstraction of the hydride from the other alkyl chain of the amine
28
This article is protected by copyright. All rights reserved.

10.1002/cbic.201800626
ChemBioChem
1
2
3
4
moiety. This promiscuous formal imine isomerisation generates intermediate III. Hydrolysis of
intermediate III by the same water molecule coordinated to the catalytic Lys residue or by another
polarised water molecule at proper distance in the active site forms the primary amine product (R)-
1b as single enantiomer.
5
6
7
Scheme 3. Proposed catalytic cycle for the reductive amination catalysed by Rs-AmDH and Ch1-AmDH with amine
donors different from ammonia (adapted from reference 17). Two stereocomplementary binding modes are the most
probable: either intermediate II (E configuration and pro-R binding mode) or intermediate II’ (Z configuration and
pro-S binding mode). Interconversion between II and II’ (and vice versa) was never observed for secondary iminium
intermediates during the time of our simulations. In contrast, interconversion from II’ to II was observed in 50% of
the cases in our simulations with primary iminium intermediates. The opposite interconversion (from II to II’) was
never observed. Further reduction of intermediate II’ by hydride shift from NADH forms the secondary amine product
in S-configuration. Reduction of intermediate II by hydride shift from NADH forms the secondary amine product in
R-configuration. However, intermediate II can also be re-oxidised to the iminium isomer III. After this formal imine
isomerisation step, the hydrolysis of intermediate III furnishes the primary amine product in R-configuration.
8
9
10
11
12
13
14
15
16
17 Conclusions
29
This article is protected by copyright. All rights reserved.