Rapid and Quantitative Profiling of Substrate Specificity of ω-Transaminases for Ketones

Article abstract of DOI:10.1002/cctc.201900399

ω-Transaminases (ω-TAs) have gained growing attention owing to their capability for asymmetric synthesis of chiral amines from ketones. Reliable high-throughput activity assay of ω-TAs is essential in carrying out extensive substrate profiling and establishing a robust screening platform. Here we report spectrophotometric and colorimetric methods enabling rapid quantitation of ω-TA activities toward ketones in a 96-well microplate format. The assay methods employ benzylamine, a reactive amino donor for ω-TAs, as a cosubstrate and exploit aldehyde dehydrogenase (ALDH) as a reporter enzyme, leading to formation of benzaldehyde detectable by ALDH owing to concomitant NADH generation. Spectrophotometric substrate profiling of two wild-type ω-TAs of opposite stereoselectivity was carried out at 340 nm with 22 ketones, revealing subtle differences in substrate specificities that were consistent with docking simulation results obtained with cognate amines. Colorimetric readout for naked eye detection of the ω-TA activity was also demonstrated by supplementing the assay mixture with color-developing reagents whose color reaction could be quantified at 580 nm. The colorimetric assay was applied to substrate profiling of an engineered ω-TA for 24 ketones, leading to rapid identification of reactive ketones. The ALDH-based assay is expected to be promising for high-throughput screening of enzyme collections and mutant libraries to fish out the best ω-TA candidate as well as to tailor enzyme properties for efficient amination of a target ketone.

Full text of DOI:10.1002/cctc.201900399

Accepted Article
Title: Rapid and Quantitative Profiling of Substrate Specificity of ω-
Transaminases for Ketones
Authors: Sang-Woo Han and Jong-Shik Shin
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: ChemCatChem 10.1002/cctc.201900399
Link to VoR: http://dx.doi.org/10.1002/cctc.201900399
A Journal of
www.chemcatchem.org

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
Rapid and Quantitative Profiling of Substrate Specificity of
-Transaminases for Ketones
Sang-Woo Han[a] and Jong-Shik Shin*[a]
-Transaminases (-TAs) have gained growing attention owing to
their capability for asymmetric synthesis of chiral amines from ketones.
Reliable high-throughput activity assay of -TAs is essential in
carrying out extensive substrate profiling and establishing a robust
screening platform. Here we report spectrophotometric and
colorimetric methods enabling rapid quantitation of -TA activities
toward ketones in a 96-well microplate format. The assay methods
employ benzylamine, a reactive amino donor for -TAs, as a
cosubstrate and exploit aldehyde dehydrogenase (ALDH) as a
reporter enzyme, leading to formation of benzaldehyde detectable by
ALDH owing to concomitant NADH generation. Spectrophotometric
substrate profiling of two wild-type -TAs of opposite stereoselectivity
was carried out at 340 nm with 22 ketones, revealing subtle
differences in substrate specificities that were consistent with docking
simulation results obtained with cognate amines. Colorimetric readout
for naked eye detection of the -TA activity was also demonstrated
by supplementing the assay mixture with color-developing reagents
whose color reaction could be quantified at 580 nm. The colorimetric
assay was applied to substrate profiling of an engineered -TA for 24
ketones, leading to rapid identification of reactive ketones The ALDH-
based assay is expected to be promising for high-throughput
screening of enzyme collections and mutant libraries to fish out the
best -TA candidate as well as to tailor enzyme properties for efficient
amination of a target ketone.
design, is highly dependent on fidelity of the assay method in
discriminating beneficial activity improvements against neutral
and detrimental changes. A number of assay methods have been
developed to assess -TA activities for amino donors[7b, 10] and
acceptors.[11] Depending on the mode of detection, these methods
can be classified into three types: 1) direct measurement of
differences in physicochemical properties between substrates
and products, including conductivity,[10a] UV absorbance[11a] and
fluorescence[11g] 2) enzymatic modulation of a coproduct into a
measurable readout using oxidase[10b-d] or dehydrogenase[7b, 11b,
Introduction
Chiral amines are important building blocks for a wide range of
pharmaceutical drugs,[1] spurring development of chemocatalytic
[2]
and biocatalytic[3] strategies aiming at cost-effective and eco-
friendly production of chiral amines with optical purities higher
than a pharmaceutical requirement (i.e., > 99.5 % ee).[4] In this
context, growing research interests for -transaminase (-TA)
witnessed in the last decade are ascribed to its unique catalytic
performance affording stereoselective transfer of an amino group
from cheap donors, such as isopropylamine and alanine, to
prochiral ketones.[5] Thermodynamic limitation in the -TA-
catalyzed amination of ketones has been identified as a crucial
obstacle to implementable process design, which attracted a
great deal of research efforts to invent physicochemical and
enzymatic methods for driving an equilibrium shift by removing a
coproduct.[1b, 3, 5a, 6]
Despite a number of successful examples focused on
overcoming the unfavorable reaction thermodynamics,[7] a paucity
of naturally occurring -TAs available for efficient amination of a
target ketone has remained a challenge for scalable process
development because of a high enzyme cost.[8] As well
exemplified with the sitagliptin production[6b] and reported later
elsewhere,[8b, 9] protein engineering of -TAs to broaden substrate
specificity and improve turnover rate for ketones has been one of
the important research goals for enriching product pipelines
obtainable by the -TA processes.
11c]
and 3) spontaneous chemical conversion of a coproduct into
a chromogenic form.[11d-f, 12]
Among the assay methods, the UV absorbance and oxidase-
based methods have been successfully applied to mutant library
screening to engineer -TA activites for bulky ketones by
assessing altered activites of mutants for cognate amines.[9b, 9e, 9f]
Besides, two methods affording activity assays directly for target
ketones have been applied to HTS of metagenomic collection.[12a]
These two methods, based on assessment of activities for
ketones, employ an amino donor unconventional for -TAs, i.e. o-
xylylenediamine[11d] and 2-(4-nitrophenyl)ethan-1-amine,[11e]
whose
deamination
products
undergo
spontaneous
polymerization or imine formation, respectively, leading to a
colored precipitate. A potential drawback of these methods is that
the amines used in the assays are not generally accepted amino
donors for -TAs and thereby a resulting enzyme candidate
displaying a strong colorimetric readout could be a false positive
showing an activity improvement for the amino donor rather than
the target ketone.
The success rate of high-throughput screening (HTS) of
mutant libraries, constructed by either random mutation or rational
Here we aimed at developing a generally applicable assay
method for quantitative activity measurements of both S- and R-
selective -TAs toward ketones using a universal amino donor
whose deamination product is detectable spectrophotometrically
or colorimetrically after enzymatic modulation. We explored a
possibility that the new assay method can be exploited for rapid
characterization of substrate specificity of a given -TA for
ketones and eventually for application to HTS of mutant libraries.
[a]
Dr. S.-W. Han, Prof. Dr. J.-S. Shin
Department of Biotechnology
Yonsei University
Yonsei-Ro 50, Seodaemun-Gu, Seoul 03722 (South Korea)
Fax: (+82) 2-362-7265
E-mail: enzymo@yonsei.ac.kr
Supporting information for this article is given via a link at the end
of the document.((Please delete this text if not appropriate))
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
O
O
O
O
Results and Discussion
Feasibility test of the enzyme assay
1a
1b
1c
1d
O
O
O
O
To develop an activity assay method for both S- and R-selective
-TAs toward ketones, the amino donor substrate used in the
assay reactions should be achiral and signal amplification of the
deamination product should be possible without interference by
the ketone substrate. We posited that benzylamine (3) would fulfil
the requirements to be considered as an ideal amino donor for the
enzyme assay we intended to develop. First, 3 is a reactive amino
donor for -TAs known to date.[13] Based on literature surveys,
activity data for 3 are available with sixteen -TAs, i.e. 12 S-
selective and 4 R-selective ones, and they display 10 to 260 %
activities for 3 relative to -methylbenzylamine (-MBA), a typical
amino donor for -TAs (Table S1). The only exception is the R-
selective-TA from Arthrobacter sp., i.e. 3 % relative activity,
although its engineered variant shows 28 % activity.[14] Second,
the deamination product of 3, i.e. benzaldehyde (4), is chemically
different from ketone and thereby can be specifically quantified by
an aldehyde dehydrogenase (ALDH) as a reporter in colorimetric
as well as spectrophotometric manners (Scheme 1). ALDH
oxidizes 4, generated from the -TA reaction, into benzoic acid
(5) at the expense of reduction of NAD+, leading to formation of
NADH detectable spectrophotometrically as an increase in UV
absorbance at 340 nm with a molar extinction coefficient,  = 6220
M-1 cm-1 (method 1). Colorimetric visualization of the -TA activity
can be also achieved by supplementing the assay mixture with
nitroblue tetrazolium (NBT) and phenazine methosulfate (PMS)
that lead to formation of blue-purple NBT-formazan, detectable at
580 nm (=12300 M-1cm-1)[15] as shown in method 2.
Feasibility of the proposed assay method is contingent upon the
reaction conditions where ALDH is active only for 4 and totally
inactive for ketones used as substrates for -TA as listed in
Scheme 2. We used ALDH cloned from Azospirillum brasilense
(GenBankTM accession number: AB241137)[16] and found that the
ALDH efficiently catalysed oxidation of 4 in the presence of NAD+
with a specific activity of 4.8 ± 0.3 U mL-1 M-enzyme-1 (Fig. S1).
In contrast, the ALDH showed a non-detectable activity for
acetophenone (1n), a typical ketone substrate for -TAs (Fig. S1).
In addition to 1n, ALDH showed negligible oxidative activities for
all the 24 ketones shown in Scheme 2 (i.e., < 0.2 % activity
relative to that for 4).
HO
O
HO
1e
1f
1g
1h
O
O
O
O
O
OH
O
OH
1i
1j
1k
1m
1l
O
O
O
1o
1n
1p
1q
O
O
O
O
O
1r
1t
1u
1s
O
O
OH
O
1w
1x
1v
Scheme 2. Alkyl (1a-m) and arylalkyl (1n-x) ketones used in this study.
The fidelity of the coupled enzyme assay for precise
quantitative measurement of -TA activity in terms of readout
from the ALDH reaction requires that the reporter enzyme
reaction should be irreversible and thereby 4 is completely
converted to 5. Otherwise, the coupled enzyme assay would lead
to underestimation of the -TA activity. Thermodynamic
equilibrium of the ALDH reaction between acetaldehyde and
acetic acid is known to be strongly shifted to acetic acid.[17]
Likewise, ALDH was found to be unable to catalyse reduction of
5 in the presence of NADH (Fig. S2), indicating that conversion of
5 to 4 by a reducing power of NADH is thermodynamically
unfavorable. It is notable that ALDH was also incapable of
catalysing reduction of 22 ketones tested (1a-1v, data not shown),
indicating ALDH is totally inactive for ketones in both oxidative
and reductive reactions. Besides the ALDH cloned from A.
brasilense, the ALDH from Escherichia coli (i.e. AldH)[18] was also
observed to be devoid of any activities for ketones in Scheme 2
(data not shown).
Another important requirement for the quantitative real-time
monitoring of -TA activity is that 4 should be instantly consumed
by the reporter enzyme as soon as it is formed, so the generation
rate of 4 (V4) becomes the same as that of NADH (VNADH). To
achieve this condition, the reporter enzyme activity should be in
excess over the -TA activity. To determine such a threshold
ALDH level required to overwhelm the -TA activity, VNADH was
measured at varying concentrations of ALDH under a fixed
concentration of-TA (Fig. 1). Time-course monitoring of optical
density at 340 nm (OD340) was done for 10 min, which showed
linear increases in OD340 during the assay time (r2 > 0.97; Fig. S3).
The VNADH values, obtained by linear regressions of the OD340
data sets, showed a hyperbolic dependence on the ALDH
Scheme 1. ALDH-based assay of -TA activity for ketone through
spectrophotometric (method 1) or colorimetric (method 2) detection of 4.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
Figure 2. Spectrophotometric measurement of the -TA activity in the assay
mixtures containing different doses of OATAW58L. Assay conditions: 1n (10 mM),
3 (100 mM), NAD+ (2 mM), OATAW58L (0-1 M) and ALDH (10 U mL-1).
Figure 1. Determination of a threshold concentration of ALDH for accurate
measurement of -TA activity. Assay conditions: 1n (10 mM), 3 (100 mM), NAD+
(2 mM), -TA (OATAW58L, 1 M) and ALDH (0.05 - 20 U mL-1).
concentration, yielding 32.5 M min-1 as an upper limit which can
be regarded as an actual V4. Since VNADH levelled off above 5 U
mL-1 of ALDH, we set 10 U mL-1 as a working concentration for
the ALDH excess condition in further enzyme assays unless V4
exceeded 32.5 M min-1. Note that the -TA used in the enzyme
assay was an engineered S-selective enzyme from
more striking when a target substrate is non-chromogenic alkyl
ketones (e. g. 1a-m) because UV detection of the resulting
amines requires laborious derivatization with a chromogenic
reagent such as a Marfey’s reagent.[9c, 19]
To corroborate the reliability of the proposed assay, we
compared the assay results between HPLC and ALDH-based
methods. To this end, we employed two wild-type -TAs, i.e. S-
selective one from Paracoccus denitrificans (PDTA)[13d] and R-
selective one from Arthrobacter sp. (ARTA).[20] For both enzymes,
real-time spectrophotometric measurement of NADH generation
led to reaction rates close to those obtained by end-point HPLC
analysis of the produced amine (Fig. 3). It is notable that the
Ochrobactrum anthropi carrying a W58L substitution (OATAW58L
)
which was rationally introduced to impart activity improvements
for ketones in the previous study.[9c]
Verification of the ALDH-based assay
ALDH-based
method
can
be
applied
to
both
To verify the ALDH excess condition for precise measurement of
-TA activity, we examined how VNADH responded to varying
concentrations of OATAW58L at 10 U mL-1 ALDH. As observed in
Fig. S3, linear increases in OD340 were also observed during the
10-min assay time irrespective of the -TA concentrations (r2 >
0.99; Fig. S4). We expected a linear correlation between the
concentration of -TA and VNADH when the assay conditions
enantiocomplementary -TAs under the same reaction conditions
because the amino donor, 3, is achiral.
In addition to the spectrophotometric measurement,
colorimetric detection of the -TA activity could be also carried
out when the assay mixture was supplemented with NBT and
PMS. In contrast to the linear increase of OD340 within the entire
assay time as shown in Fig. S4, onset of such a linear increase in
optical density at 580 nm (OD580) became retarded up to 1.5 min
fulfilled V4
≈ VNADH. Indeed, the VNADH values showed a strong
linear dependence on the -TA concentration (r2 = 0.99),
illustrating reliability of the assay method for quantitative
measurement of -TA activity (Fig. 2). Limit of detection (LOD) for
the concentration of OATAW58L, calculated from the regression
results, was 20 pM which corresponds to a reaction rate of 0.6 M
min-1. Therefore, this method provides reliable measurement of
-TA activity as long as VNADH is higher than 0.6 M min-1 under
the ALDH excess condition.
In the conventional HPLC analysis to measure -TA activity for
ketones (e.g., amination of 1n using alanine or isopropylamine as
an amino donor), UV detection of the produced amine (e.g., -
MBA (2n)) usually permits reliable quantitation of the product
formation higher than 10 M.[9c, 19] In this case, an experimental
LOD in terms of a reaction rate is over 10 M min-1 when the
reaction is allowed for a minute which is usually regarded as a
minimal time required to keep enzyme reactions running reliably.
Therefore, the ALDH-based assay is at least more than 15-fold
sensitive than the HPLC method. In addition to the higher
detection sensitivity, benefit of the ALDH-based method becomes
Figure 3. Comparison of the ALDH-based assay with the HPLC method.
Specific reaction rate represents an initial reaction rate normalized by a -TA
concentration. Reaction conditions: 1n (5 mM), 3 (50 mM), NAD+ (2 mM), -TA
(5 and 15 M for PDTA and ARTA, respectively) and ALDH (10 U mL-1). For the
HPLC method, produced amines after a 10-min reaction were analyzed by chiral
HPLC.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
Table 1. Substrate specificities of PDTA and ARTA for ketones examined
by the spectrophotometric assay.[a]
Specific reaction rate (V) (M h-1 M-enzyme-1)
Ketone
log2 ቀVVPDTAቁ
ARTA
PDTA
ARTA
1a
1b
1c
1d
1e
1f
4.7 ± 0.6
4.9 ± 0.7
28 ± 1
-0.3
-0.5
-2.3
1.7
3.2
0.7
-0.5
0.7
0.6
-1.0
-3.3
-0.3
1.2
1.1
2.3
2.0
0.7
-0.2
2.3
3.7
2.6
-1.7
3.5 ± 0.4
3.2 ± 0.4
6.0 ± 0.8
40 ± 10
120 ± 20
110 ± 20
3.7 ± 0.1
5.2 ± 0.3
8.9 ± 0.2
5.0 ± 0.7
4.5 ± 0.2
3.5 ± 0.7
8.8 ± 0.4
63 ± 4
12 ± 2
13 ± 2
72 ± 5
1g
1h
1i
5.0 ± 1.3
3.2 ± 1.3
5.8 ± 1.0
9.2 ± 1.0
34 ± 1
Figure 4. Colorimetric measurements of the -TA activity. Assay conditions
were the same as those for Figure 2 except that the assay mixtures were
supplemented with NBT (122 M) and PMS (16 M). The inset figure shows
visual images of each microplate well after 10-min reaction.
1j
1k
1l
4.3 ± 0.5
3.8 ± 0.5
28 ± 0.5
2.7 ± 0.5
2.6 ± 0.6
10 ± 2
1m
1n
1o
1p
1q
1r
(Fig. S5). This is presumably because the color-developing
reaction (i.e., reduction of NBT to NBT-formazan by PMS and
NADH) is slower than the ALDH reaction and/or follows complex
kinetics consisting of multiple reaction steps.[21] We did not
optimize the assay conditions to speed up the color reaction.
Instead, linear regression to obtain the formation rate of NBT-
formazan (VNBT-formazan) was carried out at a 1.5-10 min time
range where a time-dependent linear increase in OD580 was
observed (r2 > 0.99; Fig. S5). The VNBT-formazan values were
plotted against -TA concentrations (Fig. 4), resulting in a strong
linear relationship (r2 = 0.99). In line with the delay in onset of the
14 ± 1
11 ± 0.2
16 ± 1
10 ± 4
8.9 ± 0.8
24 ± 2
1s
1t
4.9 ± 0.6
1.3 ± 0.7
16 ± 1
18 ± 3
1u
1v
96 ± 3
67 ± 5
18 ± 1
[a] Assay conditions: ketone (5 mM), 3 (50 mM), DMSO (10 % (v/v)), NAD+
(2 mM), -TA (5 or 15 M) and ALDH (10 U mL-1).
linear increase in OD580, the assay conditions of V4
≈ VNADH >
ketone (i.e. Vrel = VPDTA/VARTA) on a log scale with a base of 2
(Table 1). In the case of 13 alkyl ketones (1a-m), Vrel did not show
a general trend but was dependent on a specific substrate
structure (i.e., 7 negative and 6 positive log2Vrel values). But when
the alkylamines were divided into two groups depending on the
presence of a non-alkyl substituent flanking the carbonyl group, a
rule-of-thumb difference in the substrate specificities could be
captured. For ketones carrying both alkyl substituents (i.e., 1a-d,
1g and 1k-m), ARTA showed a higher activity than PDTA did
except the smallest and the largest ketones (i.e., 1m and 1d,
respectively). In constrast, the two enzymes exhibited an opposite
trend (i.e., log2Vrel > 0) for ketones carrying a non-alkyl substituent
(i.e. 1e-f and 1h-j) except 1j.
Similarly, positive log2Vrel values were observed for arylalkyl
ketones (i.e., 1n-v) except 1r and 1v. Besides the negative
log2Vrel value, 1v attracted our attention because of an opposite
trend of the two -TAs in the activities relative to 1n (i.e. V1v/V1n).
Note that the V1v/V1n values are 0.29 and 2.4 for PDTA and ARTA,
respectively. Consistent with the spectrophotometric assay
results, HPLC analysis of the produced amines showed that
PDTA accepted 1n as an amino acceptor better than 1v while
ARTA did the other way around (Table 2). Note that the
cosubstrate used for these -TA reactions was isopropylamine
which is a popular amino donor for amination of ketones owing to
easy removal of a coproduct by vacuum evaporation for
equilibrium shift. Enantiomeric excesses (ee) of the produced
amines were > 99 % in all the chiral analyses. Due to the inversed
VNBT-formazan led the slope in Fig. 4 to become 2.54-fold lower than
that in Fig. 2. This result indicates that the VNBT-formazan value
obtained from the colorimetric assay should be multiplied by a
scaling factor of 2.54 to evaluate an actual V4. Visual inspection
of the color intensity in each well is qualitatively consistent with
the OD580 measurements.
Spectrophotometric substrate profiling of -TA
Based on the reliability of the ALDH-based assay, we carried out
substrate profiling of PDTA and ARTA for 22 ketones (1a-v) using
the spectrophotometric detection of NADH (Table 1). Compared
to 1n, three ketones (i.e., 1e, 1f and 1u) showed higher
reactivities with the S-selective PDTA. The R-selective ARTA
showed such higher activities for 1f, 1k and 1v than it did for 1n.
These results indicate different active site structures that define
the mode of substrate binding and the efficiency of catalytic
turnover, which can be taken for granted considering distinct
phylogenetic evolution of the two enzymes. Note that S- and R-
selective -TAs belong to a subgroup II of -family and a
D-
alanine transaminase family, respectively, based on
a
classification of PLP-dependent enzymes by Christen and
Mehta.[22] Likewise, the two enzymes belong to different
subclasses based on another classification system,[23] i.e. fold-
type I and IV for S- and R-selective -TAs, respectively.
To gain more insight into how substrate specificities of the two
enzymes are different, we scored the relative activity for each
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
Table 2. HPLC measurements of -TA activities of PDTA and ARTA for amination of 1n and 1v as well as deamination of their cognate amines.
Amination reaction rate[a] (M h-1 M-enzyme-1) Deamination reaction rate[b] (mM h-1 M-enzyme-1)
-TA
V1v/V1n
V2v/V2n
1n
1v
2n
2v
PDTA
ARTA
32
9.7
15
0.30
3.2
75
0.52
6.2
6.9  10-3
4.7
7.8
0.79
[a] Reaction conditions: ketone (5 mM), isopropylamine (50 mM) and -TA (5 M). After 90-min reaction, produced amines were analyzed by chiral HPLC.
[b] Reaction conditions: (S) or (R)-amine (10 mM), pyruvate (10 mM) and -TA (0.1 or 1 M for 2n or 2v, respectively). After 10-min reaction, produced ketones
were analyzed by reverse-phase HPLC.
assigning priority in the Cahn-Ingold-Prelog rule for 2n and 2v,
amines produced by PDTA were (S)-2n and (R)-2v while those
by ARTA were (R)-2n and (S)-2v. The V1v/V1n values determined
by the HPLC analysis are in good agreements with those obtained
by the spectrophotometric assay, indicating that the choice of an
amino donor, at least either 3 or isopropylamine, does not
significantly affect the measurement of relative activities for
ketones. This result broadens applicability of the ALDH-based
assay to consideration of process scale-up for amine synthesis
using isopropylamine.
The opposite preferences of the two -TAs for 1n and 1v
intrigued us to measure enzyme activities for cognate amines of
the two ketones (i.e., 2n and 2v). The preference of PDTA for 1n
over 1v was found to be more striking when the activities for their
cognate amines were compared (Table 2). PDTA showed only
0.7 % activity for (R)-2v relative to (S)-2n, which is a drastic
activity reduction for 30 % relative activity of 1v over 1n. This is in
line with our previous observation with PDTA that the steric
constraint in the small substrate binding pocket (i.e., S pocket)
becomes stronger for amino donors compared with amino
acceptors.[24] In contrast, such an abrupt activity drop did not
occur with ARTA and instead activities for (R)-2n and (S)-2v were
not very different.
5-phosphate form of the enzyme (E-PLP) and a pyridoxamine 5-
phosphate form (E-PMP), respectively.[25] Therefore, reliable
docking simulations of the donor-acceptor substrate pairs require
precise active site structures of both enzyme forms. However, X-
ray structures of only the E-PLP form of the two -TAs are
available yet.[26] This led us to carry out molecular docking of
donor substrates only, i.e. 2n and 2v (Fig. 5). The docking models
of 2n clearly show that spatial orientations of the reacting amine
enantiomer and the PLP moiety in the two enantiocomplementary
-TAs are mirror images of each other. The distance between the
C4 of PLP and the nitrogen of the amine substrate (C4-N) was
3.0 and 3.2 Å for (R)-2n in ARTA and (S)-2n in PDTA, respectively.
In the previous report, the C4-N distance was 2.9 Å in a docking
model of (S)-2n in OATA.[19] Considering that 2n is a typical amino
donor for -TAs, these results suggest that the optimal length of
C4-N in a productive Michaelis complex is around 3 Å. Note that
the optimal distance between the electrophilic carbon center and
the N of the nucleophile in a transition state for a nucleophilic
attack is known to be 2.5 Å.[27] We examined whether the
differential V2v/V2n values, depending on the -TA, could be
explained by how much the C4-N distance for 2v deviated from
the optimal distance observed with 2n. The docking pose of (S)-
2v in ARTA indicates that C4-N is marginally influenced by the
additional steric burden in the S pocket imposed by the hydroxyl
group of 2v (Fig. 5A). The C4-N distance in the Michaelis
These results of the V1v/V1n and V2v/V2n values in Table 2
prompted us to carry out substrate docking simulations, which we
expected to provide structural insights into how the two -TAs
ended up with differential relative activities for the cognate
substrates. The whole catalytic cycle of transaminase reactions
consists of oxidative deamination of an amino donor and reductive
amination of an amino acceptor which are mediated by a pyridoxal
complex with (S)-2v is only 0.2 Å longer than that with (R)-2n and
the bound (S)-2v is found to be stabilized by a H-bond between
the hydroxyl hydrogen of (S)-2v and the nitrogen in the peptide
bond between T283 and A234. In contrast, the S pocket of PDTA
is not spacious enough to allow such an accommodation of the
Figure 5. Molecular docking of 2n and 2v to the E-PLP forms. (A) Docking poses of (R)-2n and (S)-2v in ARTA. (B) Docking poses of (S)-2n and (R)-2v in PDTA.
The internal aldimine moiety from PLP to a Schiff base nitrogen is shown in a ball-and-stick model. For visual clarity, hydrogens are omitted in the PLP where the
C4 carbon is colored in yellow. Thick and thin sticks represent 2n and 2v, respectively. Active sites are shown as a Connolly surface. Yellow and red circles
represent the small (S) and large (L) binding pockets, respectively. Green lines represent a C4-N distance. Color code for atoms: dark grey, light grey, blue and
red for C, H, N and O, respectively.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
Figure 7. Time-course monitoring of amination of 1n and 1u. Reaction
conditions: ketone (5 mM), isopropylamine (100 mM) and OATAL57A/W58A (40 M).
Figure 6. Colorimetric assay for substrate profiling of OATAL57A/W58A with
ketones. (A) Photo images of the 96-well plate taken after 20 and 5-min
reactions with purified enzyme and cell-free extract, respectively. (B)
Measurement of the reaction rate of purified enzyme by monitoring increase in
OD580. Reaction conditions: 1a-x (5 mM), 3 (50 mM), NAD+ (2 mM), DMSO (10 %
(v/v)), purified enzyme or cell-free extract (7.5 M or 0.95 mg-protein mL-1,
respectively), ALDH (15 U mL-1), NBT (122 M) and PMS (16 M).
fitting for the reaction rate evaluation (Fig. 6B). Consistent with
the visual inspection, 1d and 1u were ranked top 2 reactive
ketones. Reactivity improvements for 1d and 1u by the
L57A/W58A substitutions were remarkable and the fold-increases
in the activities of the engineered variant for the two ketones,
compared with those of the wild-type OATA,[8b] were 1020 and
270, respectively. It is notable that the activities of OATAL57A/W58A
for 1d and 1u were 280 and 90-fold, respectively, higher than
those of PDTA displaying the same S-stereoselectivity.
Taken together, the substrate profiling results indicate that
OATAL57A/W58A is promising for efficient amination of 1d and 1u.
To examine whether the colorimetric assays provide reliable
enzyme-ketone matching information available for practical amine
synthesis, we carried out amination of 1u, in comparison with 1n,
using isopropylamine (20 molar eq. to the ketone substrate) as an
amino donor (Fig. 7). Conversion of 1u to (S)-2u reached 67 % at
4.5 h even without coproduct removal. In contrast, conversion of
1n to (S)-2n was only 25 % at the same reaction time.
hydroxymethyl group of (R)-2v and the resulting C4-N is 1.2 Å
longer compared with that for (S)-2n (Fig. 5B). The docking
models are in good accordance with the relative enzyme activities
of both -TAs and illustrate well how the subtle difference in the
S pocket geometry leads to such a dramatic change in the
substrate preference.
Colorimetric substrate profiling of -TA
In addition to the spectrophotometric method, the ALDH-based
assay also provides colorimetric detection of -TA activities for
ketones (method 2 in Scheme 1). We applied the colorimetric
method to rapid visualization of substrate preference of an OATA
Conclusions
mutant carrying L57A and W58A substitutions (OATAL57A/W58A
)
which was engineered to accept bulky ketones such as 1o and 1p
in a previous study.[8b] Substrate specificity of OATAL57A/W58A has
not yet been characterized using a wide range of ketones besides
1n-p. We run the colorimetric assay of purified OATAL57A/W58A with
all the 24 ketones shown in Scheme 2 and the image of the 96-
well plate was taken after 20-min reaction (Fig. 6A, left image).
Visual inspection of the color intensity led to rapid evaluation of
substrate specificity for ketones in a qualitative manner; 1) strong
-TA activities for 1c, 1d, 1f and 1u, 2) low activities for 1a, 1g,
1h, 1j, 1l, 1m, 1s, 1t and 1x and 3) moderate activities for the rest
11 ketones. For rapid characterization of a number of enzymes,
cell-free extract is a favored formulation over purified enzyme.
The colorimetric assay using cell-free extract led to considerable
background development (Fig. 6A, right image). However, the
background was low enough to pinpoint highly reactive ketones
such as 1d and 1u.
Rapid and precise identification of an optimal -TA, displaying
desirable kinetics and stability for a target ketone, out of
commercially available enzyme collections or lab-made mutant
libraries is a crucial step for successful process development of
asymmetric synthesis of chiral amines using -TAs. The ALDH-
based assay is generally applicable to ketones regardless of
chemical structures. Moreover, the assay method can be
expanded to activity measurement for keto acids as well as a
limited range of aldehydes for which the ALDH used in the activity
assay is not active. For example, benzaldehyde dehydrogenase
from Pseudomonas putida ATCC 12633 is known to exhibit
narrowly defined substrate specificity for aldehydes.[28] The
proposed method offers flexbility in selecting a detection mode,
i.e. spectrophotometric and colorimetric, whose choice is
dependent on trade-off between precise activity quantitation and
rapid target enzyme identification. It is worth noting that the
colorimetric assay affords rapid naked-eye inspection of a number
of reaction samples, eliminating a need for any expensive
analytical equipment to capture an invisible product readout.
The colorimetric method enables quantitative activity
measurements in terms of a reaction rate through time-course
monitoring of the color development at 580 nm. To this end, OD580
data obtained with purified enzyme were subjected to linear curve
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
simulations were carried out using CDOCKER module under a default
setting (2000 steps at 700 K for a heating step; 5000 steps at 300 K for a
cooling step; 8 Å grid extension) within the active site defined by the
Binding-Site module.
Experimental Section
Chemicals
1k and 1p were purchased from Alfa Aesar (Ward Hill, MA, USA). 4, 1m
and DMSO were obtained from Duksan Pure Chemicals Co. (Ansan,
South Korea). Isopropylamine was obtained from Junsei Chemical Co.
(Tokyo, Japan). All other chemicals were purchased from Sigma-Aldrich
Co. (St. Louis, MO, USA) and of the highest purity available.
Asymmetric synthesis of chiral amines
Reactions were carried out in a 1.8-mL glass vial under magnetic stirring
at 37 °C. Total reaction volume was 1.5 mL. Reaction conditions were 1n
or 1u (5 mM), isopropylamine (100 mM), DMSO (10 % (v/v)) and
OATAL57A/W58A (40 M) in 50 mM Tris buffer (pH 7). Aliquot of the reaction
mixture (typically 100 L) was taken at predetermined reaction times and
mixed with 37.5 L of 16 % (v/v) perchloric acid to stop the reaction.
Enzyme precipitate was removed by centrifugation (13000 rpm, 10 min)
and then the supernatant was subjected to HPLC analysis. Conversion
was measured by chiral analysis of the produced (S)-amine.
Protein expression and purification
The -TAs, i.e. PDTA,[13d] ARTA,[29] OATAW58L
and OATAL57A/W58A,
[9c]
[8b]
were cloned in a pET28(+) expression vector (Novagen) as described in
our previous studies. Cultivation of E. coli BL21(DE3) cells transformed
with the plasmids, overexpression of the His6-tagged -TAs and protein
purification were carried out as described elsewhere with minor
modifications.[8b] Standard reaction conditions for activity assay were 10
mM (S) or (R)-2n and 10 mM pyruvate in 50 mM Tris buffer (pH 7). The
enzyme reaction was stopped after 10 min by adding acetonitrile and then
1n was analyzed by HPLC. Preparation of purified ALDH was performed
as described elsewhere.[18] One unit of ALDH was defined as the enzyme
amount required to oxidize 1 mol of 4 to 5 in 1 min at 10 mM 4, 2 mM
NAD+ and 50 mM Tris buffer (pH 7). Reaction progress was monitored by
a UV spectrophotometer (UV-1650PC, Shimadzu Co.) at 340 nm and the
initial rate was obtained by a linear regression of the UV absorbance data
within the time interval showing a linear increase. Protein concentrations
were determined by measuring UV absorbance at 280 nm. Molar extinction
coefficients of the purified proteins were obtained by a protein extinction
coefficient calculator (http://www.biomol.net/en/tools/proteinextinction.
htm). The values of homodimeric -TAs and were 101885 M-1 cm-1 for
ARTA, 122576 M-1 cm -1 for PDTA and 67076 M-1 cm-1 for both OATAW58L
and OATAL57A/W58A. The value of the homotetrameric ALDH was 174220
M-1 cm-1.
HPLC analysis
All the HPLC analyses were carried out using an Alliance HPLC system
(Waters Co.) or a 1260 Infinity HPLC system (Agilent Technologies).
Quantitative analyses of 1n and 1v were performed using a Symmetry C18
column (Waters Co.). 2v was analyzed on the same column after
derivatization with 1-fluoro-2,4-dinitrophenyl-5-L-alanine amide (Marfey’s
reagent) as described elsewhere.[8b] Chiral analyses of 2n and 2u were
carried out using Crownpak CR(-) and Crownpak CR-I(+) columns (Daicel
Co.), respectively. Details of the HPLC analyses are described in the
Supporting Information.
Acknowledgements
This work was funded by the National Research Foundation of
Korea under the Basic Science Research Program
(2016R1A2B4008470). Dr. S.-W. Han was financially supported
by Initiative for Biological Function & Systems under the BK21
PLUS program of Korean Ministry of Education.
Coupled enzyme assay
All the coupled enzyme assays were conducted in 50 mM Tris buffer (pH
7) at 37 °C on a 96-well microplate and a typical reaction volume was 200
L. Concentrations of NADH or NBT-formazan were determined by
measuring OD340 or OD580, respectively, using a microplate reader (BioTek
Instruments, Inc). Initial rates were obtained from at least three
independent measurements by linear regression of the optical density data
within the time range showing a linear increase. LOD of the -TA
concentration was calculated by LOD = 3.3/S where  and S are the
standard deviation of y-intercept and the slope of the regression curve.[30]
Keywords: -transaminase • chiral amines • asymmetric
synthesis • enzyme assay • substrate profiling
[1]
a) T. C. Nugent, Chiral Amine Synthesis: Methods, Developments and
Applications, Wiley-VCH, Weinheim, 2010; b) U. T. Bornscheuer, G. W.
Huisman, R. J. Kazlauskas, S. Lutz, J. C. Moore, K. Robins, Nature 2012,
484, 185-194; c) M. Breuer, K. Ditrich, T. Habicher, B. Hauer, M.
Kesseler, R. Sturmer, T. Zelinski, Angew. Chem. Int. Ed. 2004, 43, 788-
824; d) V. Gotor-Fernández, V. Gotor, Curr. Opin. Drug Discov. Devel.
2009, 12, 784-797.
Molecular modeling
Molecular modeling was carried out using a Discovery Studio package
(version 3.5.0, Accelrys) as described before with minor modifications.[8b]
Crystal structures of PDTA (PDB ID: 4GRX)[26a] and ARTA (PDB ID:
3WWH)[26b] were used for docking simulations of 2n and 2v. Both -TAs
assume a homodimeric structure. In 4GRX, two active site arginines (R415)
form different conformations as previously reported.[9c] One of the two
active sites, harboring R415 in an outward conformation (i.e., pointing
away from PLP), was chosen for docking simulation. In the case of 3WWH,
both active site arginines (R138) assume an inward conformation (i.e.,
pointing toward PLP). Therefore, the active site was modified by a R138A
substitution and then used for docking simulation to allow hydrophobic
substituents of the arylalkylamines to be accommodated in the L pocket.
In both crystal structures, the C4-N linkage in the internal aldimine is
assigned to a single bond instead of a double bond. To make the C4-N
bond close to an actual Schiff base linkage, hybridization states of the C4
and N atoms in the internal aldimine were changed to sp2. Docking
[2]
[3]
[4]
T. C. Nugent, M. El-Shazly, Adv. Synth. Catal. 2010, 352, 753-819.
D. Ghislieri, N. J. Turner, Top. Catal. 2014, 57, 284-300.
a) L. D. Tran, O. Daugulis, Angew. Chem. Int. Ed. 2012, 51, 5188-5191;
b) S. J. Zuend, M. P. Coughlin, M. P. Lalonde, E. N. Jacobsen, Nature
2009, 461, 968-970.
[5]
[6]
a) M. S. Malik, E. S. Park, J. S. Shin, Appl. Microbiol. Biotechnol. 2012,
94, 1163-1171; b) R. C. Simon, N. Richter, E. Busto, W. Kroutil, ACS
Catal. 2014, 4, 129-143; c) F. Guo, P. Berglund, Green Chem. 2017, 19,
333-360; d) S. A. Kelly, S. Pohle, S. Wharry, S. Mix, C. C. R. Allen, T. S.
Moody, B. F. Gilmore, Chem. Rev. 2018, 118, 349-367.
a) D. Koszelewski, K. Tauber, K. Faber, W. Kroutil, Trends Biotechnol.
2010, 28, 324-332; b) C. K. Savile, J. M. Janey, E. C. Mundorff, J. C.
Moore, S. Tam, W. R. Jarvis, J. C. Colbeck, A. Krebber, F. J. Fleitz, J.
Brands, P. N. Devine, G. W. Huisman, G. J. Hughes, Science 2010, 329,
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
305-309; c) M. Fuchs, J. E. Farnberger, W. Kroutil, Eur. J. Org. Chem.
M. N. Muraleedharan, G. Larsson, Appl. Environ. Microbiol. 2014, 80,
2293-2298.
2015, 2015, 6965-6982.
[7]
a) S. Heintz, T. Börner, R. H. Ringborg, G. Rehn, C. Grey, M. Nordblad,
U. Krühne, K. V. Gernaey, P. Adlercreutz, J. M. Woodley, Biotechnol.
Bioeng. 2017, 114, 600-609; b) J. L. Galman, I. Slabu, N. J. Weise, C.
Iglesias, F. Parmeggiani, R. C. Lloyd, N. J. Turner, Green Chem. 2017,
19, 361-366.
[13] a) F. Steffen-Munsberg, C. Vickers, A. Thontowi, S. Schätzle, T.
Tumlirsch, M. S. Humble, H. Land, P. Berglund, U. T. Bornscheuer, M.
Höhne, ChemCatChem 2013, 5, 150-153; b) E. S. Park, J. Y. Dong, J. S.
Shin, Appl. Microbiol. Biotechnol. 2014, 98, 651-660; c) U. Kaulmann, K.
Smithies, M. E. B. Smith, H. C. Hailes, J. M. Ward, Enzyme Microb.
Technol. 2007, 41, 628-637; d) E.-S. Park, M. Kim, J.-S. Shin, Appl.
Microbiol. Biotechnol. 2012, 93, 2425-2435; e) J. S. Shin, B. G. Kim, J
Org. Chem. 2002, 67, 2848-2853.
[8]
a) D. F. A. R. Dourado, S. Pohle, A. T. P. Carvalho, D. S. Dheeman, J.
M. Caswell, T. Skvortsov, I. Miskelly, R. T. Brown, D. J. Quinn, C. C. R.
Allen, L. Kulakov, M. Huang, T. S. Moody, ACS Catal. 2016, 6, 7749-
7759; b) S. W. Han, J. Kim, H. S. Cho, J. S. Shin, ACS Catal. 2017, 7,
3752-3762; c) I. V. Pavlidis, M. S. Weiß, M. Genz, P. Spurr, S. P. Hanlon,
B. Wirz, H. Iding, U. T. Bornscheuer, Nat. Chem. 2016, 8, 1076-1082.
a) K. S. Midelfort, R. Kumar, S. Han, M. J. Karmilowicz, K. McConnell, D.
K. Gehlhaar, A. Mistry, J. S. Chang, M. Anderson, A. Villalobos, J.
Minshull, S. Govindarajan, J. W. Wong, Protein Eng. Des. Sel. 2013, 26,
25-33; b) A. Nobili, F. Steffen‐ Munsberg, H. Kohls, I. Trentin, C.
Schulzke, M. Höhne, U. T. Bornscheuer, ChemCatChem 2015, 7, 757-
760; c) S. W. Han, E. S. Park, J. Y. Dong, J. S. Shin, Adv. Synth. Catal.
2015, 357, 1732-1740; d) K. E. Cassimjee, M. S. Humble, H. Land, V.
Abedi, P. Berglund, Org. Biomol. Chem. 2012, 10, 5466-5470; e) M. S.
Weiß, I. V. Pavlidis, P. Spurr, S. P. Hanlon, B. Wirz, H. Iding, U. T.
Bornscheuer, ChemBioChem 2017, 18, 1022-1026; f) M. S. Weiß, I. V.
Pavlidis, P. Spurr, S. P. Hanlon, B. Wirz, H. Iding, U. T. Bornscheuer,
Org. Biomol. Chem. 2016, 14, 10249-10254.
[14] E. S. Park, J. Y. Dong, J. S. Shin, Org. Biomol. Chem. 2013, 11, 6929-
6933.
[15] a) K. M. Mayer, F. H. Arnold, J. Biomol. Screen. 2002, 7, 135-140; b) C.
V. Agu, S. M. Lai, V. Ujor, P. K. Biswas, A. Jones, V. Gopalan, T. C. Ezeji,
Sci. Rep. 2018, 8, 3379.
[9]
[16] S. Watanabe, T. Kodaki, K. Makino, J. Biol. Chem. 2006, 281, 28876-
28888.
[17] R. A. Alberty, A. Cornish-Bowden, Trends Biochem. Sci. 1993, 18, 288-
291.
[18] E. S. Park, S. Park, J. S. Shin, J. Biosci. Bioeng. 2017, 123, 528-533.
[19] S. W. Han, E. S. Park, J. Y. Dong, J. S. Shin, Adv. Synth. Catal. 2015,
357, 2712-2720.
[20] A. Iwasaki, K. Matsumoto, J. Hasegawa, Y. Yasohara, Appl. Microbiol.
Biotechnol. 2012, 93, 1563-1573.
[21] V. Ponti, M. U. Dianzani, K. Cheeseman, T. F. Slater, Chem. Biol.
Interact. 1978, 23, 281-291.
[10] a) S. Schätzle, M. Höhne, K. Robins, U. T. Bornscheuer, Anal. Chem.
2010, 82, 2082-2086; b) M. S. Weiß, I. V. Pavlidis, C. Vickers, M. Hohne,
U. T. Bornscheuer, Anal. Chem. 2014, 86, 11847-11853; c) J. Hopwood,
M. D. Truppo, N. J. Turner, R. C. Lloyd, ChemComm 2011, 47, 773-775;
d) S. C. Willies, J. L. Galman, I. Slabu, N. J. Turner, Phil. Trans. R. Soc.
A 2016, 374, 20150084.
[22] P. Christen, P. K. Mehta, Chem. Rec. 2001, 1, 436-447.
[23] R. Percudani, A. Peracchi, BMC Bioinform. 2009, 10, 273.
[24] E. S. Park, S. R. Park, S. W. Han, J. Y. Dong, J. S. Shin, Adv. Synth.
Catal. 2014, 356, 212-220.
[25] a) M. D. Toney, Arch. Biochem. Biophys. 2014, 544, 119-127; b) K.
Hirotsu, M. Goto, A. Okamoto, I. Miyahara, Chem. Rec. 2005, 5, 160-172.
[26] a) C. Rausch, A. Lerchner, A. Schiefner, A. Skerra, Proteins 2013, 81,
774-787; b) L. J. Guan, J. Ohtsuka, M. Okai, T. Miyakawa, T. Mase, Y.
Zhi, F. Hou, N. Ito, A. Iwasaki, Y. Yasohara, M. Tanokura, Sci. Rep. 2015,
5, 10753.
[11] a) S. Schätzle, M. Höhne, E. Redestad, K. Robins, U. T. Bornscheuer,
Anal. Chem. 2009, 81, 8244-8248; b) M. D. Truppo, J. D. Rozzell, J. C.
Moore, N. J. Turner, Org. Biomol. Chem. 2009, 7, 395-398; c) M. D.
Truppo, N. J. Turner, Org. Biomol. Chem. 2010, 8, 1280-1283; d) A. P.
Green, N. J. Turner, E. O'Reilly, Angew. Chem. Int. Ed. 2014, 53, 10714-
10717; e) D. Baud, N. Ladkau, T. S. Moody, J. M. Ward, H. C. Hailes,
ChemComm 2015, 51, 17225-17228; f) A. Dunbabin, F. Subrizi, J. M.
Ward, T. D. Sheppard, H. C. Hailes, Green Chem. 2017, 19, 397-404; g)
T. Scheidt, H. Land, M. Anderson, Y. Chen, P. Berglund, D. Yi, W. D.
Fessner, Adv. Synth. Catal. 2015, 357, 1721-1731.
[27] E. F. Oliveira, N. M. Cerqueira, P. A. Fernandes, M. J. Ramos, J. Am.
Chem. Soc. 2011, 133, 15496-15505.
[28] C. K. Yeung, A. Yep, G. L. Kenyon, M. J. McLeish, Biochim. Biophys.
Acta 2008, 1784, 1248-1255.
[29] E. S. Park, M. S. Malik, J. Y. Dong, J. S. Shin, ChemCatChem 2013, 5,
1734-1738.
[12] a) L. Leipold, D. Dobrijevic, J. W. E. Jeffries, M. Bawn, T. S. Moody, J.
M. Ward, H. C. Hailes, Green Chem. 2019, 21, 75-86; b) M. Gustavsson,
[30] T. A. Little, BioPharm Int. 2015, 28, 48-51.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201900399
ChemCatChem
FULL PAPER
FULL PAPER
Naked-eye detection:
Sang-Woo Han and Jong-Shik Shin*
Spectrophotometric and colorimetric
assay of -transaminase activities has
been developed using aldehyde
dehydrogenase as a reporter,
enabling rapid substrate profiling for
ketones.
Page No. – Page No.
Rapid and Quantitative Profiling of
Substrate Specificity of -
Transaminases for Ketones
This article is protected by copyright. All rights reserved.