High throughput screening of monoamine oxidase (MAO-N-D5) substrate selectivity and rapid kinetic model generation

Article abstract of DOI:10.1016/j.molcatb.2015.07.001

Full kinetic models provide insight into enzyme mechanism and kinetics and also support bioconversion process design and feasibility assessment. Previously we have established automated microwell methods for rapid data collection and hybrid kinetic modelling techniques for quantification of kinetic constants. In this work these methods are applied to explore the substrate selectivity and kinetics of monoamine oxidase, MAO-N-D5, from Aspergillus niger. In particular we examine the MAO-N-D5 variant Ile246Met/Asn336Ser/Met348Lys/Thr384Asn to allow the oxidation of secondary amines Initial screening showed that MAO-N-D5 enabled the selective oxidation of secondary amines in 8 and 9 carbon rings, as well as primary ethyl and propyl amines attached to secondary amines of indolines and pyrrolidines. Subsequently we developed a first kinetic model for the MAO-N-D5 enzyme based on the ping-pong bi-bi mechanism (similar to that for the human MAO-A enzyme). The full set of kinetic parameters were then established for three MAO-N-D5 substrates namely; 3-azabicyclo[3,3,0]octane, 1-(2 amino ethyl) pyrrolidine and 3-(2,3-dihydro-1H-indole-1-yl)propan-1-amine. The models for each amine substrate showed excellent agreement with experimentally determined progress curves over a range of operating conditions. They indicated that in each case amine inhibition was the main determinant of overall reaction rate rather than oxygen or imine (product) inhibition. From the perspective of larger scale bioconversion process design, the models indicated the need for fed-batch addition of the amine substrate and to increase the dissolved oxygen levels in order to maximize bioconversion process productivity.

Full text of DOI:10.1016/j.molcatb.2015.07.001

Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
Contents lists available at ScienceDirect
Journal of Molecular Catalysis B: Enzymatic
journal homepage: www.elsevier.com/locate/molcatb
High throughput screening of monoamine oxidase (MAO-N-D5)
substrate selectivity and rapid kinetic model generation
L. Rios-Solisa,b, B. Mothiaa, S. Yia, Y. Zhoua, M. Michelettia, G.J. Lyea,∗
a The Advanced Centre for Biochemical Engineering, Department of Biochemical Engineering, University College London, Gordon Street,
London WC1H 0AH, UK
b Faculty of Pharmacy, Autonomous University of the State of Morelos, Av. Universidad 1001, Col. Chamilpa, Cuernavaca, Morelos 62209, Mexico
a r t i c l e i n f o
a b s t r a c t
Article history:
Full kinetic models provide insight into enzyme mechanism and kinetics and also support bioconver-
sion process design and feasibility assessment. Previously we have established automated microwell
methods for rapid data collection and hybrid kinetic modelling techniques for quantification of kinetic
constants. In this work these methods are applied to explore the substrate selectivity and kinetics of
monoamine oxidase, MAO-N-D5, from Aspergillus niger. In particular we examine the MAO-N-D5 variant
Ile246Met/Asn336Ser/Met348Lys/Thr384Asn to allow the oxidation of secondary amines Initial screening
showed that MAO-N-D5 enabled the selective oxidation of secondary amines in 8 and 9 carbon rings, as
well as primary ethyl and propyl amines attached to secondary amines of indolines and pyrrolidines.
Subsequently we developed a first kinetic model for the MAO-N-D5 enzyme based on the ping-pong
bi-bi mechanism (similar to that for the human MAO-A enzyme). The full set of kinetic parameters were
then established for three MAO-N-D5 substrates namely; 3-azabicyclo[3,3,0]octane, 1-(2 amino ethyl)
pyrrolidine and 3-(2,3-dihydro-1H-indole-1-yl)propan-1-amine. The models for each amine substrate
showed excellent agreement with experimentally determined progress curves over a range of operat-
ing conditions. They indicated that in each case amine inhibition was the main determinant of overall
reaction rate rather than oxygen or imine (product) inhibition. From the perspective of larger scale bio-
conversion process design, the models indicated the need for fed-batch addition of the amine substrate
and to increase the dissolved oxygen levels in order to maximize bioconversion process productivity.
© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Received 13 April 2015
Received in revised form 17 June 2015
Accepted 3 July 2015
Available online 6 July 2015
Keywords:
Monoamine oxidase
Aspergillus niger
Microscale kinetic modelling
Substrate screening
1. Introduction
order to determine process feasibility and cost effectiveness [6,7].
Nevertheless, most reported bioconversion kinetic studies only
Biocatalytic processes are increasingly being considered in
industrial organic syntheses since in certain situations they are con-
sidered ‘greener’ and can bring considerable reductions to overall
process costs [1,2]. This is a consequence of the mild temperature,
pressure and pH conditions under which enzymes function. Their
high regio- and stereo selectivity facilitates conversions with high
atom efficiency and reduced environmental impact [3].
To develop a biocatalyst for commercial application, even after
several rounds of successful protein engineering or evolution, it is
still necessary to screen several enzyme variants with a variety of
substrates over a range of concentrations [4,5]. In addition, it is use-
ful to establish enzyme kinetic mechanisms and full kinetic models.
These give fundamental insights into bioconversion operation and
enable different reactor operating strategies to be simulated in
report the maximum reaction rate, Vmax, and the Michaelis con-
stant, Km, along with qualitative comments on the level of substrate
and/or product inhibition. This is because obtaining a full kinetic
model along with values of all the inhibition constants involves
multiple experiments and complex data processing, consuming
valuable time and resources [8].
Recently, we have developed more rapid numerical techniques
for enzyme kinetics determination that combine traditional initial
rate experiments, to identify a solution in the vicinity of the global
minimum, with non-linear regression methods to determine the
exact location of the solution. These greatly reduce the number
of experiments required to obtain the full set of kinetic parame-
ters [9,10]. These numerical techniques have been combined with
the creation of automated, microwell scale methods that increase
the throughput and reduce the costs and timescales associated
with experimental studies [11,12]. In the biocatalysis field, these
microscale tools have been successfully applied to evolve, screen
and characterize a broad range of enzymes [13,14]. We have also
∗
Corresponding author.
E-mail address: g.lye@ucl.ac.uk (G.J. Lye).
http://dx.doi.org/10.1016/j.molcatb.2015.07.001
1381-1177/© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
101
shown how the results obtained at microwell scale can be used to
2.3. Shake flask MAO-N-D5 fermentations and lysate preparation
predict fermentation and bioconversion kinetics at laboratory and
pilot scale [15,16].
Competent E. coli BL21-Gold (DE3), E. coli BL21 Plyse and E. coli
BL21 Plysee cells were transformed with the plasmid pET16b MAO-
N-D5 using the heat shock technique described by the supplier
(Stratagene, Amsterdam, NL). An overnight culture of the trans-
formed cells was produced in a 100 ml shake flask (10 ml working
volume) of LB broth (10 g l−1 tryptone, 5 g l−1 yeast extract, 10 g l−1
NaCl) containing 150 mg ml−1 ampicillin. Growth was performed at
30 ◦C with orbital shaking at 250 rpm using an SI 50 orbital shaker
(Stuart Scientific, Redhill, UK). The total volume of this culture was
used to inoculate a 1 l shake flask (100 ml working volume) which
was left to grow for 8 h. After 4 h, IPTG was added to a concentration
of 0.5 mM for enzyme induction. Following the removal of broth
by centrifugation, the cells were resuspended in 1 M phosphate
buffer, pH 7.8 and used for whole cell bioconversions. When MAO-
N-D5 lysate was needed, the cells were sonicated with a Soniprep
150 sonicator (MSE, Sanyo, Japan). The lysate suspension was cen-
trifuged at 5000 rpm in Falcon tubes for 5 min. The clarified lysate
was either used immediately or stored at −20 ◦C.
In this work we applied these methods to study the kinet-
ics of the monoamine oxidase, MAO-N-D5, which is a homo
dimer flavin dependant enzyme that catalyses the oxidation
of amines [17]. Interest has arisen in the MAO-N-D5 enzyme
due to its 49.1 and 48.7% sequence homology to the human
monoamine oxidases MAO-A and MAO-B, which are responsi-
ble for oxidizing important neurotransmitters such as serotonin
and dopamine [18]. MAO-N-D5 has previously been found to
possess a high catalytic rate and selectivity with many aliphatic
and aromatic amines [19,20]. The particular enzyme variant stud-
ied here has been subject to in vitro evolution, and the mutant
Ile246Met/Asn336Ser/Met348Lys/Thr384Asn has been shown to
facilitate the selective oxidation of secondary and tertiary amines
[21–23]. The aim of this work is to further examine the substrate
selectivity of the MAO-N-D5 mutant and to establish the first full
kinetic models of the enzyme for conversion of a range of sub-
strates. These will provide important insights into bioconversion
process design and scale-up.
2.4. High throughput platform and microscale bioconversions
A Tecan Genesis platform (Tecan, Reading, UK) was used to
perform the bioconversions in polypropylene 96-deep square well
plates (DSW) (Becton Dickenson, NJ, USA) covered with a thermo
plastic elastomer cap designed to work with automated equip-
ment (Micronic, Lelystad, Netherlands). The automated platform
included an integrated Thermomixer Comfort Shaker (Eppendorf,
Cambridge, UK), with a shaking frequency and diameter of 400 rpm
and 6 mm respectively, which was used to ensure homogeneous
enzyme and temperature distribution. It also included a microplate
centrifuge (Hettitch Rotanta 46 RSC, Tuttlingen, Germany) and
an Ultraspec 1100 Pro UV/Visible microplate spectrophotome-
ter (Amersham Biosciences, Buckinghamshire, UK) for absorbance
measurements [26].
2. Materials and methods
2.1. Materials and plasmid
Molecular biology enzymes were obtained from New England
Bio-laboratories (NEB, Hitchin, UK). Competent Escherichia coli
BL21-Gold (DE3) cells were obtained from Stratagene (Amster-
dam, Netherlands), while One Shot® E. coli BL21 Plyss and One
Shot® E. coli BL21 Plyse cells were obtained from Invitrogen
(Paisly, UK). Primers and oligonucleotides were purchased from
Operon (Cologne, DE). All other reagents were obtained from
Sigma–Aldrich (Gillingham, UK) unless noted otherwise, and were
of the highest purity available.
In a standard microscale bioconversion the total volume was
600 ␮l which comprised 300 ␮l of the MAO-N-D5 resuspended cell
solution and 300 ␮l of a 40 mM solution of the selected substrate
in 1 M phosphate buffer at pH 7.8. The plates were incubated at
37 ◦C with all experiments performed in triplicate. In individual
experiments the biocatalysts form (whole cell or lysate), substrate
type and concentration and buffer concentration were varied as
indicated in the text. Samples (100 ␮l) were either withdrawn or
whole wells were sacrificed for analysis as required.
In order to investigate oxygen availably during bioconversions,
the dissolved oxygen tension (DOT) was measured in a specially
adapted 96-DSW plate. A fluorescent oxygen sensor spot (Pst3, Pre-
Sens – Precision Sensing GmbH, Germany) was mounted flush with
the inside wall of a single well. The sensor spot was connected to
a light emitting diode and photodetector (OXY-4, PreSens) via an
optical fibre (PreSens) attached through the outside wall inline with
the sensor spot. DOT measurements were taken every 15 s. Before
each run, the sensor was calibrated at 0% air saturation at 30 ◦C
after blanketing the plate with nitrogen for a period of 10 min. The
calibration at 100% air saturation was similarly performed at 30 ◦C
after pumping air over the well.
Plasmid pET16b MAO-N-D5 contained the Aspergillus niger
MAO-N-D5 gene with the following mutations: Ile246Met/
Asn336Ser/Met348Lys/Thr384Asn [24], and was kindly provided
by Professor Nick Turner (University of Manchester, UK). DNA
sequencing was performed by the UCL Wolfson Institute for
Biomedical Research. 12 ␮l of DNA sample were provided at a con-
centration of 100 ng ␮l−1 in 1.5 ml Eppendorf tubes. The primers
used are shown in Table 1. Primers were commercially syn-
thetized (Eurofins MWG Operon, London, UK) and prepared at a
concentration of 5 pmoles l−1. Sequencing results were routinely
analyzed using Bioedit software (www.mbio.ncsu.edu/BioEdit/
bioedit.html).
2.2. Synthesis of product imine 3-azabicyclo-[3,3,0]oct-2-ene
3-Azabicyclo[3,3,0]octane hydrochloride (7) (Table 2) (3.07 g,
21.0 mmol) was dissolved in 100 ml dichloromethane (DCM) and
N-chlorosuccinimide (6.45 g, 48.3 mmol, 2.3 equiv) was added in
one portion. The mixture was stirred for 1.5 h at room temper-
ature and subsequently filtered. The filtrate was reduced to a
volume of 20 ml, where precipitation occurred. The filtration was
repeated, the precipitate was extracted with a small amount of DCM
(5–7 ml) and the combined filtrates were added to a solution of KOH
(2.41 g, 43.0 mmol, 2.04 mmol) in 50 ml methanol (MeOH). The tur-
bid mixture was stirred overnight and filtered. The crude product
was purified by flash column chromatography (hexane: EtOAc 9:1
followed by CH2Cl2: MeOH 99:1) to give the product 3-azabicyclo-
[3,3,0]oct-2-ene (17) as yellow crystals (0.25 g, 2.29 mmol, 10%),
identical by NMR with the literature [25].
2.5. Analytical methods
Gas chromatography (GC) analysis was performed on a Trace
130 Gas Chromatograph (Thermo Scientific, UK) with a Thermo
Scientific AS 1310 Autosampler, an instant Connect-FID flame ion-
ization detector on a CAM column (30 m × 0.32 mm × 0.25 ␮m, J&W
Scientific, Agilent Technologies). In order to carry out GC analysis,
aliquots of 100 ␮l were withdrawn at various time intervals from

102
L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
Table 1
Primer sequences used for sequencing the MAO-N-D5 gene.
Primer
Sequence
MAO forward:
MAO reverse:
5ꢀ-CCAGGGACCAGCAATGACCTCCCGAGACGGATACCAG-3ꢀ
5ꢀ-GAGGAGAAGGCGCGTTATCACAAACGAGCCTTCACCTCC-3ꢀ
the bioconversion solution and quenched with 10 ␮l of a 1 M NaOH
solution. 200 ␮l tert-butyl methyl ether was added to the samples
which were incubated at 37 ◦C and 900 rpm for 30 min in a Ther-
momixer Comfort shaker. The samples were then centrifuged for
5 min at 5000 rpm and the organic phase was collected for anal-
ysis. The injector temperature was set at 250 ◦C with triple times
of pre-wash and post-wash setting. The split injection was set as
50:1. Measurements were carried out using the FID detector at con-
stant flows 1.6 ml min−1 of He, 45 ml min−1 of H2 and 350 ml min−1
of air at 250 ◦C. Measurements were controlled and recorded by
Chromeleon® software (version: 7.1.2.1478, Dionex Cor., Sunny-
vale, CA, 2013).
Substrate and product concentrations were quantified by GC
based on their peak areas and calibration curves prepared for
the individual compounds. For the standard bioconversion the
retention times for substrate 7 and its corresponding imine
3-azabicyclo-[3,3,0]oct-2-ene (17) were 5.85 min and 6.87 min
respectively. Protein and enzyme concentrations were determined
using the Bradford assay and SDS-PAGE as described previously
[27]. Enzyme specific activities were determined from the GC sam-
ples as the amount of imine product formed per unit of time
normalized by the dry cell weight of E. coli biomass (or enzyme
units) used in the reaction. One unit of enzyme activity was defined
as the amount of enzyme which catalyzed the formation of 1 ␮mol
of 3-azabicyclo-[3,3,0]oct-2-ene (17) for 1 min in a forward reac-
tion at a 20 mM concentration at 37 ◦C, 400 rpm and a pH of 7.8,
using a 1 M phosphate buffer.
Fig. 1. Specific activity of MAO-N-D5 expressed in different E. coli strains. Growth
of the various strains was performed in shake flasks as described in Materials and
Methods section. Activities based on conversion of the substrate 3-azabicyclo [3.3.0]
octane (7). Error bars represent one standard deviation about the mean (n = 3).
not been previously investigated, hence the enzyme was initially
expressed in three different hosts: E. coli BL21 DE3 gold, E. coli
BL21 DE3 plyss and E. coli BL21 DE3 plyse. These were selected due
to their potentially different expression levels and degree of con-
trol of recombinant enzyme expression [30]. Enzyme activity was
quantified using the standard substrate (7), previously shown to be
successfully oxidized by the selected MAO-N-D5 variant to produce
imine 3-azabicyclo-[3,3,0]oct-2-ene (17), which is an intermediate
in the production of telaprevir drug for treatment of Hepatitis C
[25].
1H NMR spectra for the imine product 3-azabicyclo-[3,3,0]oct-
2-ene (17) was recorded on a Bruker AVANCE III 600 (600 MHz)
using CDCl3 solvent. The chemical shifts (ı) were given in ppm units
relative to tetramethylsilane, and coupling constants (J) are mea-
sured in Hertz. NMR: (600 MHz, CDCl3), 7.30 (1 H, m), 4.05 (1
Fig. 1 shows the specific MAO-N-D5 activity quantified using
the different strains. No significant statistical difference was found
␦H
H, dddd, J = 7.8, 4.2, 3.0, 1.8 Hz), 3.52 (1 H, dq, not fully resolved,
J = 10.8, 2.4 Hz), 3.29 (1 H, m), 2.66 (1 H, m), 1.70–1.20 (7 H, 4 m).
Chemical ionization (CI) mass spectra were recorded on a Thermo
Finnegan MAT 900XP spectrometer. MS: (CI+), C7H11N, m/z: 110.1
[M+H]+.
between the different E. coli strains with all specific activities being
−1
in the range 1.05–1.2 mM min
g
DCW
. Final biomass concentra-
tions during the initial fermentations were also similar at 2.1,
1.8 and 1.7 gDCW l−1 for E. coli DE3 gold, plyss and plyse respec-
tively. These levels of MAO-N-D5 specific activity are in agreement
with literature results [22,25]. Therefore E. coli BL21 DE3 Gold was
selected as the host strain for all further work due to the slightly
higher cell concentrations produced during fermentation.
2.6. Kinetic parameter determination
We have previously described a series of methodologies and
numerical techniques for rapid kinetic parameter estimation
[10,28]. These were implemented here using a programme devel-
oped in Matlab® software (MathWorks, Natick, MA, USA) in order
to automatically perform all the non-linear regressions and sta-
tistical analyses. Non-linear regressions were performed using the
mesh adaptive pattern search algorithm in Matlab® known as: “The
Genetic Algorithm and Direct Search Toolbox”. This method was
previously shown to be more likely to achieve global optimization
than gradient-based methods [28].
3.2. High throughput substrate and kinetics screening
Previous studies on the substrate selectivity of the MAO-N-D5
have reported that increasing the substrate bulk and lipophilic-
ity led to higher rates of oxidation by the enzyme [25,29].
Consequently 16 different compounds were selected for activity
screening based on structural variations of compound 7 used above.
Modifications were considered that mainly related to the num-
ber and size of the carbon rings and various substitutions. Table 2
shows the structure of the selected amines and the measured activ-
ity levels. Compounds 5 and 11 were used as negative controls, as
they were an amide and a fully oxidized heterocycle respectively,
which theoretically could not be oxidized by MAO-N. Activity val-
ues are based on initial rates of product formation using a starting
concentration of 20 mM of each substrate.
3. Results and discussion
3.1. Biocatalyst production and choice of host strain
Leaky expression of the selected MAO-N-D5 enzyme variant
had the potential for intracellular toxicity effects due to produc-
tion of H2O2 while reacting with cell metabolites [29]. To the best
of our knowledge the choice of host to avoid leaky expression has
From the 16 compounds screened, compounds 3, 4, 7, 8, 13
and 14 resulted in a measurable rate of substrate conversion. The

L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
103
Table 2
High throughput screening of the MAO-N-D5 mediated oxidation of various amines. All substrates screened at an initial concentration of 20 mM using 1.6 units of MAO-N-D5
in lysate form. Initial rates determined using the high throughput techniques described in the Materials and Methods Section. Dash indicates no measurable conversion after
40 h.
Substrate number
Name
Structure
Initial rate (mM h−1
)
1
2
Pyrrolidine
–
–
1 Methyl pyrrolidine
3
Hexamethyleneimine
0.52
4
5
Heptamethyleneimine
0.15
2-Azabicyclo(2.2.1)hept-5-ene-3-one
–
6
Decahydroquinoline
–
7
8
3-Azabicyclo[3,3,0]octane
0.1
cis-8-azabicyclo[4.3.0] nonane
0.28
9
Indoline
–
HCl
10
11
3-hydroxy piperidine
5-Aminobenzimidazole
–
–

104
L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
Table 2 (Continued )
Substrate number
Name
Structure
Initial rate (mM h−1
)
12
13
2,3-Dihydro-7-azaindole
–
1-(2 amino ethyl) pyrrolidine
1.6
14
3-(2,3-dihydro-1H-indole-1-yl)propan-1-amine
0.2
15
16
Benzyl-2-azabicyclo[3.3.0]octane-3 carboxylate
–
–
5-Methylindoline
highest activity was found for compound 13 (1–2 amino ethyl
pyrrolidine) while the lowest activity was for compound 7. Com-
pounds 1, 2, 7 and 8 have previously shown no activity with
MAO-N-D5 in conventional labscale experiments and these results
are confirmed here using the microwell platform [25].
Fig. 3 where, for ease of illustration, results are presented for the
substrates demonstrating ‘low’ and ‘high’ activities respectively.
Based on this data initial estimates of the maximum initial rate,
Vmax, and Michaelis constant, Kamine, were obtained for the six pri-
ority substrates as shown in Table 3. These results suggest that
increasing the lipophilicity of the substrates results in higher MAO-
N-D5 activity, which is in agreement with previous work [25].
The ability to collect this additional kinetic data on multiple sub-
strates in parallel gives further insight to the screening results and
also to issues of inhibition that will inform strategies for bioconver-
sion process design [31]. The kinetic profiles in Fig. 3 indicate that
the substrate concentration of 20 mM used in the initial screening
experiments (Table 2) was already above inhibitory levels, e.g. for
compounds 3 and 4 which was severely inhibited at 20 mM, or well
below them e.g. for compound 14 where the activity at 100 mM
was five times higher than in the screening experiments. These
results reemphasise that quantification of enzyme activity at single
substrate concentrations can often be misleading.
Structurally similar compounds 7 and 8 exhibited substrate
inhibition above concentrations of 100 mM and while the Vmax val-
ues were ofthe same order ofmagnitude, the Km ofcompound 7 was
five times smaller. Likewise, compounds 3 and 4 both showed sig-
nificant substrate inhibition above concentrations of 10 mM. Both
these substrates presented similar Vmax values that were one order
of magnitude higher than those of compounds 7 and 8, while the
Km values were similar (with the exception of compound 8). The
Vmax values of compounds 13 and 14 were 15 and 33 times higher
than the reference substrate 7 although the Km values were 2.5
As expected, negative control substrates 5 and 11 did not
present any oxidation activity, and substrates 6, 9, 10, 12, 15
and 16 also showed that compounds containing a phenyl group
gave no detectable activity. This could be due to phenyl groups
being highly stable substituents, displaying steric hindrance and
preventing any conversion taking place. The activity of MAO-
N-D5 towards compounds 3, 4, 13 and 14 has not previously
been reported in the literature. The results here demonstrate
the capacity of the enzyme to oxidize secondary amines in rings
with 8 and 9 carbons (compounds 3 and 4), as well as primary
ethyl and propyl amines attached to secondary amines of indo-
lines and pyrrolidines (compounds 13 and 14). These last two
compounds are of particular interest since when the amine gets
oxidized by MAO-N-D5, there is the potential to form new 5
and 6 membered heterocyclic compounds yielding the imines
3,5,6,7-tetrahydro-2H-pyrrolo[1,2-a]imidazole (17) and 2,3,4,10-
tetrahydropyrimido-[1,2-a]indole (18) respectively as shown in
Fig. 2. The identity of products 17 and 18 were confirmed by mass
spectrometry and NMR.
Using the automated microwell platform further kinetic data
was obtained in parallel on compounds 3, 4, 7, 8, 13 and 14 where
initial rates of conversion were quantified as a function of initial
substrate concentration. Michaelis–Menten profiles are shown in

L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
105
Fig. 2. Reaction scheme of the MAO-N-D5 mediated oxidation of compounds 7, 13 and 14, forming imine 3-azabicyclo-[3,3,0]oct-2-ene (17) and new imines with 5- and
6-membered heterocyclic rings: 3,5,6,7-tetrahydro-2H-pyrrolo[1,2-a]imidazole (18) and 2,3,4,10-tetrahydropyrimido-[1,2-a]indole respectively (19).
and 15 times higher respectively indicating lower binding affinity
(Table 3).
bioconversions in order to avoid any build-up of H2O2 in the reac-
tion mixture since it is known to be potentially damaging to the
enzyme thus complicating kinetic measurements with issues of
enzyme stability [23]. The in situ removal of H2O2 also has the ben-
efit of removing the by-product making the reaction irreversible as
shown in Fig. 4(A).
Because pH and dissolved oxygen fluctuations can affect the
quality of the experimental data, and lead to determination of
incorrect kinetic parameter values, these were first investigated in
the microscale format used here for high throughput kinetic data
collection. Fig. 5 shows the imine production curves for the MAO-
N-D5 mediated oxidation of substrate 7 at an initial concentration
of 20 mM. This figure also shows the pH and DOT measured in
the microwell during the course of bioconversions at two different
buffer concentrations.
In terms of progressing to a first full kinetic model for the MAO-
N-D5 enzyme compound 7 was selected since this model substrate
has been used in previous works where partial kinetic data has been
gathered and is available for comparison [22,23]. Given the nov-
elty of the products obtained from compounds 13 and 14, kinetic
models were also established for these two additional substrates.
Derivates from compound 17 have been successfully used as cata-
lysts for the synthesis of polyurethanes [32], while compound 18 is
of special interest acting as a potential intermediate in the synthesis
of heterocyclic amides used as anti-tumoral agents [33].
3.3. MAO-N-D5 mechanism and kinetic parameter determination
Full conversion of substrate 7 is seen after 8 h in the biocon-
version using 1 M phosphate buffer while the conversion using
100 mM buffer only reached 75% conversion in the same period.
This difference is attributed to the significant decrease in reac-
tion pH, from 7.8 to 6.0, observed when using the lower buffer
concentration. Consequently all subsequent bioconversions were
performed in 1 M phosphate buffer to ensure reaction kinetics
were determined at a constant pH. Fig. 5 also shows that the mea-
sured DOT levels remained constant, at close to 100% saturation, for
both bioconversions, demonstrating that the combination of liquid
fill volume, well geometry and shaking diameter/frequency used
MAO-N-D5 belongs to a family of enzymes that catalyze the oxi-
dation of monoamines using FAD as a cofactor [20]. Previous studies
have shown that the MAO-N-D5 kinetic profile has more resem-
blance to the human MAO-A, which follows a ping–pong bi–bi
mechanism, rather than to the human MAO-B which follows a bi–bi
ordered reaction mechanism [17]. Based on this literature prece-
dent a proposed King-Altman scheme for the MAO-N-D5 mediated
oxidation of amines is shown in Fig. 4. The corresponding rate
model is derived in Eq. (1).
Vmax([Amine][O2])
v =
(1)
den
ꢀ
ꢁ
ꢀ
ꢁ
KamineKiO
2 [Imine]
[Amine]
Ki,amine
[O2]
KAmine
den = KO2 [Amine] 1 +
+ Kamine[O2] 1 +
+ [Amine][O2] +
[O2][Imine] +
Ki,imine
Ki,O
Ki,imine
2
Here Vmax is the maximum reaction rate, KO and Kamine are the
corresponding Michaelis constants for O2 and2the amine substrate
[34] were sufficient to ensure that oxygen mass transfer into the
microwell never became rate limiting.
Having verified the experimental system, the methodology
established in our previous work was applied in order to rapidly
respectively, while Ki,amine, Ki,O and Ki,imine are the corresponding
2
inhibition constants for the amine substrate, O2 and the result-
ing imine product. In the experiments required to determine
those kinetic parameters, catalase (2.5 mg ml−1) was added to the

106
L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
Fig. 5. Kinetics of whole cell MAO-N-D5 mediated oxidation of 3-
azabicyclo[3,3,0]octane (7) in either 100 mM phosphate buffer (open symbols)
or 1 M phosphate buffer (solid symbols): imine product concentration (ꢁ,ꢀ), pH
(ꢂ,᭹) and dissolved oxygen tension, DOT (ᮀ,᭿). Bioconversion conditions: 20 mM
−1
MAO-N-D5 whole cells, initial pH 7.8, 37 ◦C, 400 rpm, total
substrate, 15 g
l
DCW
volume of 0.6 ml in 96 well plates. Error bars represent one standard deviation
about the mean (n = 3).
determine the kinetic constants for each of the three selected
substrates [9,10]. Briefly this involves first performing traditional
initial rate experiments at low concentrations to determine the
Michaelis–Menten and rate constants, which represents values
close to ‘real solution’. This is combined with nonlinear regression
methods applied to full progress curves at high substrate concen-
trations to determine the inhibition constants, as well as helping
refine the exact location of the solution, improving the quality of the
parameters while reducing the number of experiments required.
Here then, the appropriate experimental Michaelis–Menten
parameters obtained in Table 3 were used as initial values to per-
form non-linear regression analysis with 15 progress curves for
each substrate, at initial substrate concentrations between 50 and
−300 mM, in order to determine the full set of kinetic parameters.
At initial substrate concentrations higher than 100 mM, reactions
with each of the 3 substrates did not reach full conversion, and no
Fig. 3. Variation of MAO-N-D5 activity in lysate form with substrate concentration:
(a) substrates with ‘low’ activity and (b) substrates with ‘high’ activity. Kinetics
quantified using the high throughput techniques described in Section 2. Substrates
in 1 M phosphate buffer pH 7.8, 37 ◦C and 400 rpm, 1.6 units of MAO-N-D5 were
used in all reactions.
Fig. 4. Proposed King-Altman scheme of the MAO-N-D5 mediated oxidation of amines following a ping-pong bi-bi mechanism. EFADox: enzyme bound to cofactor FAD in
the oxidized form, EFADred: enzyme bound to cofactor FAD in the reduced form.

L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
107
Table 3
Apparent Vmax and Km values determined for the MAO-N-D5 mediated oxidation of the selected amines shown in Fig. 3. Concentration of MAO-N-D5 was 1.6 units as defined
in the Materials and Methods Section.
Substrate number
Structure
Vmax (mM h−1
)
Km (mM)
3
4
1.2
1.4
3
4
7
8
0.16
0.42
4
18
13
2.5
10
14
5.3
60
increase in imine production was detected after 20 h bioconver-
sion. Therefore the maximum duration of the progress curves used
to determine the kinetic parameters was set to 20 h.
respective ␦-1-pyrroline imine, which is followed by an attack of
the primary amine side chain, resulting in the bicyclic derivative.
Subsequently, another oxidation is required to obtain products 18
and 19. Further work needs to be performed in order to determine
the exact reaction mechanism of this second oxidation. In this work,
however, we have considered it to be oxygen mediated, and due
to the slow rate of the first MAO-N oxidation and the excess of
molecular oxygen, we have not included this step in the kinetic
model.
Fig. 6 shows representative examples of the experimental and
predicted progress curves for oxidation of the three selected com-
pounds at different initial substrate and enzyme concentrations.
Very good agreement is seen between the predicted and experi-
mental data under all reaction conditions. Equation 1 does not take
into account enzyme deactivation, and therefore predicted data
after 20 h reaction time over predicted the experimental values for
substrate concentrations higher than 100 mM in the case of all three
compounds (data not shown).
Because the dissolved O2 concentration could only be monitored
and not controlled in the standard microwells, Michaelis–Menten
plots varying O2 could not be experimentally obtained. The initial
value for the Michaelis constant of O2 (KO2) was taken from pre-
vious studies of MAO-N-D5 that reported a value of 0.7 mM [17].
Using air as the O2 source, the maximum concentration that could
be reached in the microwell reactions at 37 ◦C was 0.21 mM [35],
which is approximately four times lower than the literature KO2
value. The Michaelis–Menten profiles (Fig. 3) and kinetic parame-
ters obtained from them (Table 3) will therefore be oxygen limited
due to solubility limitations in the aqueous reaction media. In the
current experimental set-up it was not possible to use pure oxy-
gen instead of atmospheric air in the robotic containment cabinet,
therefore the kinetic parameter values determined here will be
apparent kinetic parameters suitable when using air as the oxygen
source.
The final values determined for each of the apparent kinetic
parameters in Fig. 4 are summarized in Table 4. Results are shown
for each of the three selected substrates. For substrates 13 and
14, initial oxidation of the pyrrolidine ring by MAO-N leads to the
3.4. Implications for bioconversion process design
Considering the kinetic parameter values in Table 4, the kcat
of compound 14 was 4 and 25 times higher than compounds 13

108
L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
and 7 respectively, but the Michaelis–Menten constant Kamine was
15 and 10 times higher respectively, which drastically affected
the catalytic effectiveness of compound 14. In terms of inhibition,
the amine substrate was found to be the most inhibitory com-
pound, greatly affecting compound 7 with an inhibition constant
of 0.8 mM, while the Ki,amine for compounds 13 and 14 were 4 and
6 times greater (less inhibitory) respectively. In this case substrate
fed-batch operating would be recommended for compound 7 when
attempting to achieve high space-time yields at larger scales.
Imine product inhibition was also found to be present, but
to a much lower degree than for the substrates, with inhibition
constants, Ki,imine, 1–2 orders of magnitude higher than the cor-
responding substrate inhibition constants (Table 4). For all the
selected substrates, the oxygen inhibition constant, Ki,O2 , was
found to be between 60 and 90 mM, which is 2 orders of magni-
tude above the aqueous O2 saturation level indicating that oxygen
inhibition was negligible under the reaction conditions examined.
The kinetic models predict that increasing the dissolved oxygen
Fig. 6. Comparison of experimental and modelled bioconversion kinetics for the whole cell MAO-N-D5 mediated oxidation as a function of substrate and enzyme concentration
showing amine (•) depletion and imine (♦) production: (a1) 20 mM compound 7, 66 units of MAO-N-D5, (a2) 10 mM compound 7, 49 units of MAO-N-D5, (b1) 20 mM compound
13, 6.5 units of MAO-N-D5, (b2) 10 mM compound 13, 3.3 units of MAO-N-D5, (c1) 10 mM compound 14, 6.5 units of MAO-N-D5, (c2) 30 mM compound 14, 3.3 units of
MAO-N-D5. Experiments performed using the high throughput techniques described in Materials and Methods at 37 ◦C and pH 7.8 in 1 M phosphate buffer. Solid lines fitted
based on Eq. (1) using the final kinetic parameters in Table 4. Error bars represent one standard deviation about the mean (n = 3).

L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
109
Table 4
Final values of the apparent kinetic parameters determined for the selected substrates. 7: 3-azabicyclo[3,3,0]octane, 13: 1-(2 amino ethyl) pyrrolidine, 14: 3-(2,3-dihydro-
1H-indole-1-yl)propan-1-amine.
Substrate number
7
13
14
Chemical structure
Rate constant: kcat (mM h−1 units−1
Amine Michaelis constant: Kamine (mM)
)
0.22
6.4
1.2
1.5
4.2
0.9
5.5
63.2
0.7
O2 Michaelis constant: KO (mM)
2
Amine inhibition constant: Ki,amine (mM)
0.8
3.5
5.1
O2 inhibition constant: Ki,O (mM)
Imine inhibition constant: Ki,imine (mM)
57.1
54.5
72.6
170.0
89.5
300.3
2
concentration up to a saturation level of 1.25 mM would represent
an approximately 4–5 fold reduction in the time required to reach
full conversion. Hence it would be recommended to either sparge
the bioreactor with pure oxygen, or an air-oxygen mixture (for
non-flammable substrates), or to pressurize the vessel in order to
achieve enzyme saturation with O2 and maximize the reaction rate.
This predicted effect due to increased oxygen levels has been pre-
viously observed using MAO-N-D5 with primary amine substrates
[17].
4. Conclusions
The MAO-N-D5 variant (Ile246Met/Asn336Ser/Met348Lys/
Thr384Asn) studied here enabled the selective oxidation of sec-
ondary amines in 8 and 9 carbon rings, as well as primary ethyl
and propyl amines attached to secondary amines of indolines and
pyrrolidines. Those last two types of compounds were of special
interest, due to the formation of new heterocyclic rings when oxi-
dized by MAO-N-D5.
Compared to kinetic constants reported in the literature for
MAO-N-D5, the values of the Michaelis constant, Kamine, of com-
pounds 7 and 13 are of the same order of magnitude as ones
reported previously for other secondary amines, while that for
compound 14 is one order of magnitude higher [24]. Compared
to other constants found in the literature for primary amines using
the wild type MAO-N-D5, the Kamine values for compounds 7 and
13 are at least 1 order of magnitude higher and 2 orders of mag-
nitude higher than for compound 14 [17]. This could be explained
because secondary amines are not natural substrates for the MAO-
N-D5, and even with 4 mutations towards improved secondary
amines activity [25], further improvement is needed achieve the
Michaelis–Menten values and rate constants of MAO-N-D5 variants
with primary amines.
Our previously established high throughput microwell exper-
imental platform and hybrid kinetic modelling methodologies
enabled the rapid generation of full kinetic models for three MAO-
N-D5 substrates based on the ping–pong bi–bi mechanism. The
kinetic models for each amine substrate showed excellent agree-
ment with experimentally determined progress curves over a range
of operating conditions. The models indicated amine inhibition to
be the main factor limiting reactor productivity, suggesting the
need for fed-batch operation and enhanced dissolved oxygen con-
centrations in order to achieve the highest possible space-time
yields.
Acknowledgements
Extensive studies can be found in the literature to identify
inhibitors of MAO-A and MAO-B due to their importance for the
treatment of depression, Parkinson’s disease and several other
disorders [36]. The identified inhibitors have also been found to
severely inhibit MAO-N-D5, such as ˇ-carboline derivatives and
1-methyl-4-phenylpyridinium where inhibition constants values
ranged from 3 to 200 ␮M [17], which is closer to the value of
Ki,amine determined in this work for compound 7 (790 ␮M), while
the inhibition constants for compounds 13 and 14 were one order
of magnitude higher (Table 4).
Overall, the insights gained from the kinetic model generation
suggest that amine substrate inhibition is a critical factor when
considering MAO-N-D5 catalyzed bioconversions, which was in
agreement with previous works focused on the MAO-N mediated
oxidation of compound (7) [23] and of bicyclic amines [37]. Dis-
solved substrate concentrations above 100 mM severely affect the
bioconversion rate and yield. At larger scale fed-batch operation
and/or use of reactors in series will be necessary to make the con-
versions economically viable. The high throughput experimental
methods and kinetic models described here are thus important
tools to identify bottlenecks early and guide the selection of reac-
tion engineering or biocatalyst engineering strategies to enhance
overall productivity.
The research leading to these results has received fund-
ing from the European Union’s Seventh Framework Programme
FP7/2007–2013 under grant agreement n◦ 266025 (BIONEXGEN).
References
[1] T. Narancic, R. Davis, J.N.K.E.O. Connor, Biotechnology (2015), Early published
online.
[2] J.-M. Choi, S.-S. Han, H.-S. Kim, Biotechnol. Adv. (2015), Early published online.
[3] M.T. Reetz, J. Am. Chem. Soc. 135 (2013) 12480.
[4] A. Cázares, J.L. Galman, L.G. Crago, M.E.B. Smith, J. Strafford, L. Ríos-Solís, G.J.
Lye, P.A. Dalby, H.C. Hailes, Org. Biomol. Chem. 8 (2010) 1301.
[5] N.J. Turner, A. Wells, ChemCatChem 6 (2014) 900.
[6] G. Sin, J.M. Woodley, K.V. Gernaey, Biotechnol. Progr. 25 (2009) 1529.
[7] M. Stitt, Y. Gibon, Trends Plant Sci. 19 (2014) 256.
[8] B. Chen, F. Baganz, J. Woodley, Chem. Eng. Sci. 62 (2007) 3178.
[9] B. Chen, M. Micheletti, F. Baganz, J. Woodley, G. Lye, Chem. Eng. Sci. 64 (2009)
403.
[10] L. Rios-Solis, N. Bayir, M. Halim, C. Du, J.M. Ward, F. Baganz, G.J. Lye, Biochem.
Eng. J. 73 (2013) 38.
[11] M. Micheletti, G.J. Lye, Curr. Opin. Biotechnol. 17 (2006) 611.
[12] M. Halim, L. Rios-Solis, M. Micheletti, J.M. Ward, G.J. Lye, Bioprocess Biosyst.
Eng. 37 (2014) 931.
[13] L. Rios-Solis, M. Halim, A. Cázares, P. Morris, J.M. Ward, H.C. Hailes, P.A. Dalby,
F. Baganz, G.J. Lye, Biocatal. Biotransform. (2011).
[14] C.J. Du, L. Rios-Solis, J.M. Ward, P.A. Dalby, G.J. Lye, Biocatal. Biotransform. 32
(2014) 302.
[15] R.S. Islam, D. Tisi, M.S. Levy, G.J. Lye, Biotechnol. Bioeng. 99 (2008) 1128.

110
L. Rios-Solis et al. / Journal of Molecular Catalysis B: Enzymatic 120 (2015) 100–110
[16] L. Rios-Solis, P. Morris, C. Grant, A.O.O. Odeleye, H.C. Hailes, J.M. Ward, P.A.
Dalby, F. Baganz, G.J. Dalby, Chem. Eng. Sci. 122 (2015) 360.
[17] S.O. Sablin, V. Yankovskaya, S. Bernard, C.N. Cronin, T.P. Singer, Eur. J.
Biochem. 253 (1998) 270.
[18] K.E. Atkin, R. Reiss, N.J. Turner, A.M. Brzozowski, G. Grogan, Acta, Sect.
Crystallogr, F. Struct, Biol. Cryst. Commun. 64 (2008) 182.
[19] R. Carr, M. Alexeeva, A. Enright, T.S.C. Eve, M.J. Dawson, N.J. Turner, Angew.
Chem. Int. Ed. Engl. 42 (2003) 4807.
[26] J.Z. Baboo, J.L. Galman, G.J. Lye, J.M. Ward, H.C. Hailes, M. Micheletti,
Biotechnol. Progr. 28 (2012) 392.
[27] M.M. Bradford, Anal. Biochem. 72 (1976) 248.
[28] B.H. Chen, E.G. Hibbert, P.A. Dalby, J.M. Woodley, AIChE J. 54 (2008) 2155.
[29] N.J. Turner, Chem. Rev. 111 (2011) 4073.
[30] J.W. Dubendorff, F.W. Studier, J. Mol. Biol. 219 (1991) 45.
[31] K.V. Gernaey, A.E. Lantz, P. Tufvesson, J.M. Woodley, G. Sin, Trends Biotechnol.
28 (2010) 346.
[20] K.E. Atkin, R. Reiss, V. Koehler, K.R. Bailey, S. Hart, J.P. Turkenburg, N.J. Turner,
A.M. Brzozowski, G. Grogan, J. Mol. Biol. 384 (2008) 1218.
[21] C.J. Dunsmore, R. Carr, T. Fleming, N.J. Turner, J. Am. Chem. Soc. 128 (2006)
2224.
[32] O. Werbitzky, U. Daum, R. Bregy, US Patent US6476175 B2, 2002.
[33] Z. Wang, L. Li, US Patent US2014032229, 2014.
[34] S.D. Doig, S.C.R. Pickering, G.J. Lye, J.M. Woodley, Biotechnol. Bioeng. 80
(2002) 42.
[22] P. Zajkoska, M. Rosenberg, R. Heath, K.J. Malone, R. Stloukal, N.J. Turner, M.
Rebrosˇ, Appl. Microbiol. Biotechnol. (2014).
[35] R.R. Deshpande, Y. Koch-Kirsch, R. Maas, G.T. John, C. Krause, E. Heinzle,
ASSAY Drug. Dev. Technol. 3 (2005) 299.
[23] H. Ramesh, J.M. Woodley, J. Mol. Catal. B: Enzym. 106 (2014) 124.
[24] R. Carr, M. Alexeeva, M.J. Dawson, V. Gotor-Fernández, C.E. Humphrey, N.J.
Turner, ChemBioChem 6 (2005) 637.
[25] V. Köhler, K.R. Bailey, A. Znabet, J. Raftery, M. Helliwell, N.J. Turner, Angew.
Chem. Int. Ed. Engl. 49 (2010) 2182.
[36] S.K. Al-Nuaimi, E.M. Mackenzie, G.B. Baker, Am. J. Ther. 19 (2012) 436.
[37] T. Li, J. Liang, A. Ambrogelly, T. Brennan, G. Gloor, G. Huisman, J. Lalonde, A.
Lekhal, B. Mijts, S. Muley, L. Newman, M. Tobin, G. Wong, A. Zaks, X. Zhang, J.
Am. Chem. Soc. 134 (2012) 6467.