Structure and activity of NADPH-dependent reductase Q1EQE0 from streptomyces kanamyceticus, which catalyses the R-selective reduction of an imine substrate

Article abstract of DOI:10.1002/cbic.201300321

NADPH-dependent oxidoreductase Q1EQE0 from Streptomyces kanamyceticus catalyzes the asymmetric reduction of the prochiral monocyclic imine 2-methyl-1-pyrroline to the chiral amine (R)-2-methylpyrrolidine with >99% ee, and is thus of interest as a potential biocatalyst for the production of optically active amines. The structures of Q1EQE0 in native form, and in complex with the nicotinamide cofactor NADPH have been solved and refined to a resolution of 2.7 A. Q1EQE0 functions as a dimer in which the monomer consists of an N-terminal Rossman-fold motif attached to a helical C-terminal domain through a helix of 28 amino acids. The dimer is formed through reciprocal domain sharing in which the C-terminal domains are swapped, with a substrate-binding cleft formed between the N-terminal subunit of monomer A and the C-terminal subunit of monomer B. The structure is related to those of known β-hydroxyacid dehydrogenases, except that the essential lysine, which serves as an acid/base in the (de)protonation of the nascent alcohol in those enzymes, is replaced by an aspartate residue, Asp187 in Q1EQE0. Mutation of Asp187 to either asparagine or alanine resulted in an inactive enzyme.

Full text of DOI:10.1002/cbic.201300321

CHEMBIOCHEM
FULL PAPERS
DOI: 10.1002/cbic.201300321
Structure and Activity of NADPH-Dependent Reductase
Q1EQE0 from Streptomyces kanamyceticus, which
Catalyses the R-Selective Reduction of an Imine Substrate
Marꢀa Rodrꢀguez-Mata,[a, b] Annika Frank,[a] Elizabeth Wells,[a] Friedemann Leipold,[c]
Nicholas J. Turner,[c] Sam Hart,[a] Johan P. Turkenburg,[a] and Gideon Grogan*[a]
NADPH-dependent oxidoreductase Q1EQE0 from Streptomyces
kanamyceticus catalyzes the asymmetric reduction of the pro-
chiral monocyclic imine 2-methyl-1-pyrroline to the chiral
amine (R)-2-methylpyrrolidine with >99% ee, and is thus of in-
terest as a potential biocatalyst for the production of optically
active amines. The structures of Q1EQE0 in native form, and in
complex with the nicotinamide cofactor NADPH have been
solved and refined to a resolution of 2.7 ꢁ. Q1EQE0 functions
as a dimer in which the monomer consists of an N-terminal
Rossman-fold motif attached to a helical C-terminal domain
through a helix of 28 amino acids. The dimer is formed
through reciprocal domain sharing in which the C-terminal do-
mains are swapped, with a substrate-binding cleft formed be-
tween the N-terminal subunit of monomer A and the C-termi-
nal subunit of monomer B. The structure is related to those of
known b-hydroxyacid dehydrogenases, except that the essen-
tial lysine, which serves as an acid/base in the (de)protonation
of the nascent alcohol in those enzymes, is replaced by an
aspartate residue, Asp187 in Q1EQE0. Mutation of Asp187 to
either asparagine or alanine resulted in an inactive enzyme.
Introduction
Optically active amines constitute some of the most important
synthetic intermediates in pharmaceuticals synthesis. Interest
in developing sustainable chemical routes to optically active
amines has stimulated research in various enzymatic methods
for their production.[1] Many enzymes are capable of synthesiz-
ing chiral amines from prochiral or racemic precursors, with
examples ranging from the (dynamic) kinetic resolution of rac-
emic amines using N-acylases[2] or other hydrolases, transami-
nases for the quantitative conversion of prochiral ketones to
amines at the expense of an ammonia donor,[3] and the dera-
cemisation of racemic amines using flavin-dependent amine
oxidases as part of a chemoenzymatic oxidation–reduction
cycle with a chemical reductant.[4] In industry, optically active
amines are also derived by chemical reduction from prochiral
imines, themselves easily generated from precursor ketones
through their reaction with amines. Many recent examples
exist of abiotic catalysts for the asymmetric reduction of pro-
chiral imine substrates for this purpose.[5–8]
The enzymatic asymmetric reduction of imines has been less
well-explored in the area of preparative biocatalysis, largely be-
cause of the aqueous lability of the imine substrates, but also
because of the dearth of apparent “imine reductases” in the lit-
erature. The well-studied dihydrofolate reductase (DHFR) from
E. coli[9] and related enzymes perform asymmetric hydrogena-
tion of its imine substrate dihydrofolate, but the enzyme has
not been the focus of study as a biocatalyst for chiral amine
preparation, save for its application in the generation of its
native product, (S)-tetrahydrofolate.[10] However, asymmetric
imine reduction by whole-cell preparations of yeasts has re-
cently been reported[11,12] and imine reductase activity has also
been attributed to some strains of anaerobic bacteria.[13] An
NADPH-dependent reductase, PchG, from Pseudomonas aerugi-
nosa and acting as a thiazonolinyl imine reductase has also
been discovered in the biosynthetic pathway towards the side-
rophore pyochelin[14] and a related enzyme, Irp3 from Yersinia
enterocolitica, has recently been the subject of structural stud-
ies.[15] However, the most interesting candidate enzymes pos-
sessing imine reductase activity with potential for application
come from Streptomyces spp. Imine reductase activities depen-
dent on the unusual deazaflavin cofactor F-420 have been im-
plicated in the biosynthetic pathways of sibiromycin,[16] tomay-
mycin[17] and chlortetracycline,[18] amongst others, in various
Streptomyces species. Additionally, a wide-ranging screen of
Streptomyces organisms revealed asymmetric imine reductase
activity in two strains: Streptomyces sp. GF3587 catalysed the
reduction of 2-methyl-1-pyrroline (1, Scheme 1) to (R)-2-meth-
ylpyrrolidine (2) with 99.2% ee.[19] Another strain, Streptomyces
[a] Dr. M. Rodrꢀguez-Mata, A. Frank, E. Wells, S. Hart, Dr. J. P. Turkenburg,
Dr. G. Grogan
York Structural Biology Laboratory
Department of Chemistry, University of York
Heslington, York, YO10 5DD (UK)
E-mail: gideon.grogan@york.ac.uk
[b] Dr. M. Rodrꢀguez-Mata
Universidad de Oviedo
Facultad de Quꢀmica, Quꢀmica Orgꢁnica e Inorgꢁnica
Julian Claverꢀa 8, C.P. 33006 Oviedo, Asturias (Spain)
[c] Dr. F. Leipold, Prof. N. J. Turner
Manchester Institute of Biotechnology, University of Manchester
131 Princess Street, Manchester, M1 7DN (U.K.)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201300321.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1372

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
ble expression were achieved and purification of the enzyme
using nickel-affinity and size-exclusion chromatography yielded
a protein which, confirmed by SDS-PAGE analysis, had an
approximate molecular weight of 32 kDa (Figure S2). Using a
standard spectrophotometric assay for the measurement of
substrate-stimulated NADPH oxidation, a Km value of 8.21Æ
1.07 mm and a kcat of (0.018Æ1.4)ꢃ10À3 sÀ1 for substrate 1
were recorded (Figure S3, Table 1).[20] At concentrations of
Table 1. Kinetic constants for NADPH-dependent reductases Q1EQE0
acting on substrates 1, 3 and 5.
Scheme 1. Biotransformation of imine substrates by imine reductase Q1E1E0
from S. kanamyceticus.
Substrate
kcat [sÀ1
]
Km [mm]
kcat/Km [sÀ1 mmÀ1
]
1
3
5
0.018Æ1.4ꢃ10À3
8.21Æ1.07
1.16Æ0.15
0.724Æ0.09
2.2ꢃ10À3 Æ3.0ꢃ10À4
7.4ꢃ10À4 Æ1.0ꢃ10À4
5.2ꢃ10À3 Æ1.0ꢃ10À4
sp. GF3546, catalysed asymmetric reduction of 1 to the S enan-
tiomer, with 92.3% ee. Further research by the Mitsukura
group was performed in which the R-selective enzyme (termed
RIR) was purified and characterised.[20] The enzyme appeared
to be a dimer composed of monomer subunits of 32 kDa, and
Km and Vmax values of 3.5 mm and 10.2 mmolminÀ1 mgÀ1 respec-
tively for substrate 1 were recorded. Cofactor requirement was
limited to NADPH. The acquisition of N-terminal amino acid
sequence data led to the cloning of the gene encoding RIR,
which allowed the identification of a homologue, Uniprot ac-
cession number Q1EQE0, in the genome of Streptomyces kana-
myceticus,[21] which was also shown to convert 1 to (R)-2 with
99.6% ee.[22] A sequence alignment of RIR with Q1EQE0 re-
vealed significant homology throughout the length of the
protein chain, with 50% sequence identity and a further 19%
strongly similar residues (Figure S1). Q1EQE0 is slightly longer
(309 residues versus 295 for RIR), due to an insertion of
14 amino acids in the N-terminal region that is absent in RIR.
Genes encoding RIR and Q1EQE0,[22] and also the S-selective
imine reductase (SIR) from Streptomyces sp. GF3546,[23] were
cloned and expressed heterologously in E. coli, creating re-
combinant biocatalysts for the reduction of the imine sub-
strate. No details concerning the purification and characteriza-
tion of Q1EQE0 were reported.
9ꢃ10À4 Æ2.0ꢃ10À5
3.8ꢃ10À3 Æ1.0ꢃ10À4
above 30 mm 1, significant substrate inhibition was observed,
with a calculated Ki value of 56.5Æ10.7 mm. No activity was
detected using NADH as the cofactor. Biotransformations of
5 mm 1 using the pure enzyme and NADPH, in the presence of
glucose-6-phosphate and glucose-6-phosphate dehydrogenase
for NADPH recycling, gave the R product in >99% ee with
23% conversion (Figure 1). No significant reaction was ob-
served in the absence of Q1EQE0.
As part of a screen of enzymes suitable for the reduction of
synthetically relevant imines, we synthesised the gene encod-
ing Q1EQE0 with the sequence codon-optimised for expression
in E. coli. In this study, we report the structure of Q1EQE0 from
S. kanamyceticus to 2.7 ꢁ, and its complex with the cofactor
NADPH, and describe structural homology with hydroxyisobu-
tyrate dehydrogenases that may shed light on mechanism. It is
envisaged that the structure may be used as the basis for
rational engineering of enzymes catalysing the asymmetric
reduction of imines for improved activity and wider substrate
specificity.
Figure 1. Chiral HPLC chromatograms of GITC-derivatised amine 2. The top
trace shows the GITC-derivatised racemic amine 2, with the S and R amines
eluting at 49.1 and 52.0 min respectively; the bottom trace shows the GITC-
derivatised (R)-2 recovered from biotransformation of 2-methyl-1-pyrroline 1
(5 mm) by NADPH-dependent Q1EQE0.
Low conversions (<5%) of imine substrates 3,4-dihydroiso-
quinoline (3, Scheme 1, Figure S4A and B) and 2-methyl-3,4-di-
hydroisoquinolin-2-ium triflate (5, Scheme 1, Figure S5A and B)
were also observed over a period of 24 h, during which negli-
gible degradation of the substrates was observed in buffered
solution.
The kinetic constants for the transformations of these sub-
strates are presented in Table 1. Substrates 3 and 5 were not
reported to be substrates for RIR or Q1EQE0 previously.
Results and Discussion
Cloning and expression of the gene encoding Q1EQE0 and
purification and assay of the enzyme
The gene encoding Q1EQE0 was synthesised with codons opti-
mised for expression in E. coli BL21 (DE3). Good levels of solu-
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1373

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
Structure of Q1EQE0
ment; the tartronic semialdehyde reductase from Salmonella
typhimurium (1YB4, 19%).[25]
The structure of Q1EQE0 was solved by molecular replacement.
Data collection and refinement statistics are given in Table 2.
For structure solution, the highest sequence homologue in the
Protein Data Bank was an oxidoreductase from Pseudomonas
putida KT2440 (PDB ID: 3L6D, 26% sequence identity to
Q1EQE0), although the molecular replacement pipeline
BALBES[24] selected an alternative model for molecular replace-
The solution contained four monomers (A–D), arranged as
two pairs of dimers (A–B and C–D). The monomer, represented
by subunit A, of Q1EQE0 consists of an N-terminal Rossman-
fold domain of approximately 181 amino acids (1–181) and
a C-terminal helical domain (210–306) connected by a long in-
terdomain helix from residues Leu182–Gly209 (Figure 2, left).
The N-terminal domain consists of a b-sheet of a sheet of six
parallel strands [b1 (22–25), b2 (45–48), b3 (64–65), b4 (78–81),
b5 (105–108), b6 (130–136)] adjacent to two antiparallel
strands [b7 (149–153), b8 (172–175)] interspersed in sequence
and encapsulated by six well-defined a-helices: a1 (29–42), a2
(56–60), a3 (69–74), a4 (86–93), a5 (114–127) and a6 (156–
167). The C-terminal domain is comprised of the end of the
interconnecting helix a7, and a further four helices: a8 (216–
241), a9 (252–269), a10 (274–289) and a11 (297–303). In the
native enzyme structure, the model covered residues 18–306
in subunits A–C, but it was also possible to model N-terminal
residues 10–17 in subunit D. This was also the case for the
NADP complex structure.
Table 2. Data collection and refinement statistics for oxidoreductase
Q1EQE0 in native form and in complex with NADPH. Numbers in brackets
refer to data for highest resolution shells.
Q1EQE0 native
Q1EQE0 NADPH
beamline
wavelength [ꢁ]
resolution [ꢁ]
Diamond I03
0.97630
108.68–2.71
(2.79–2.71)
C121
Diamond I04-1
0.91999
74.27–2.73
(2.80–2.73)
C121
space group
unit cell
a, b, c [ꢁ]
203.41, 131.11, 77.48 205.80, 130.30, 77.70
In the native structure, each subunit featured poorer density
in the NADPH phosphate recognition loop between residues
51 and 65, meaning that residues 53 to 56 were not modelled
in subunits A and D, and residues 51–65 were not modelled in
subunits B and C. Near these regions of poorer density, Asn49
(B) and Ala52 (A) were noted as the only outliers in the Rama-
chandran plot. In the NADPH complex structure, the density in
this loop was much improved, with only residues 55 in chains
A–D and 54 in subunit B not modelled. Only two residues,
Ala52 (C) and Ala53(D) were Ramachandran outliers in the
NADPH complex structure.
a, b, g [8]
90.00, 107.22, 90.00
4
90.00, 107.10, 90.00
4
no. of molecules in the
asymmetric unit
unique reflections
completeness [%]
Rmerge [%]
Rp.i.m.
multiplicity
<I/s(I)>
CC1/2
49759
51705
99.3 (99.6)
0.05 (0.52)
0.05 (0.49)
3.4 (3.5)
13.9 (2.3)
0.99 (0.80)
55
99.4 (99.1)
0.08 (0.60)
0.07 (0.50)
4.3 (4.3)
12.2 (2.0)
1.00 (0.78)
47
overall B factor from
Wilson plot [ꢁ2]
Rcryst/Rfree [%]
18.5/21.1
7915
106
0.014
1.55
66
67
52
–
18.5/21.6
8094
111
0.015
1.79
57
58
43
64
no. protein atoms
no. water molecules
RMSD 1–2 bonds [ꢁ]
RMSD 1–3 angles [8]
average main chain B [ꢁ2]
average side chain B [ꢁ2]
average water B [ꢁ2]
average ligand B [ꢁ2]
An electrostatic surface of the monomer revealed the inter-
connecting helix a7 to be somewhat hydrophobic, which
would prove to be significant in dimer formation (Figure S6).
Analysis of the monomer structure using the DALI server,[26]
showed that, of structures of enzymes for which an activity
has been determined, Q1EQE0 is most similar to members of
the gamma-hydroxybutyrate dehydrogenase (GHBDH) family
Figure 2. Left: Structure of Q1EQE0 monomer illustrating N-terminal Rossman-fold domain made up of a-helices a1–6 and b-strands b1–8 and C-terminal
helical domain containing a-helices a8–11 connected by inter-domain helix a7. Helix a7 stretches from Leu182 to Gly209. Right: Structure of Q1EQE0 dimer,
formed by reciprocal domain swapping between subunits A (dark grey) and B (light grey); NADPH is shown in cylinder format with the carbon atoms in grey.
NADPH is bound at the interface formed between, in one case, the N-terminal domain of A and the C-terminal helical domain of B.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1374

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
such as PDB structures 3PEF and 3PDU, or of the hydroxyiso-
butyrate dehydrogenase (HIBDH) family, such as 2CVZ, but
there are significant differences in orientation of the C-terminal
domain in Q1EQE0 that result in extensive domain swapping
not observed in characterised HIBDHs as described below.
perimposition of these structures with Q1EQE0 can be found
in the Supporting Information (Figure S9A and S9B).
NADPH binding in Q1EQE0 and the active site
Datasets resulting from crystals of Q1EQE0 that had been
soaked in a solution of NADPH (10 mm) resulted in electron
density maps in which the omit map at the subunit interface
could be modelled unambiguously as the cofactor, with excel-
lent occupancy in each of the four active sites found in the
pair of dimers in the structure solution. The electron density of
residues and their side chains in the region binding the
NADPH ribose-2’-phosphate (Asn49–Lys54) was much im-
proved in the NADPH complex than in the native structure.
The cofactors (two per dimer) are bound in clefts formed by, in
one instance, the N-terminal domain of subunit A, helices a7
and a8 of the C-terminal helical bundle of subunit B and the
loop that connects them (Figure 2, right). NADPH only forms
close interactions with the N-terminal domain however. The
active site of Q1EQE0 is a large channel that traverses the
enzyme structure at the surface of the dimer interface. Apart
from residues that bind the phosphate residues of NADPH, the
entrance to the channel and the channel itself are strikingly
negatively charged, perhaps commensurate with its activity in
recognising, binding, and perhaps stabilising, imine substrates
or their iminium ions (Figure S10).
Dimer formation
The active form of Q1EQE0 is a dimer, which is formed by re-
ciprocal domain swapping between two subunits (Figure 2,
right) and signified in the description below by the notation A
and B. Analysis by PISA[27] indicated a total contact area of
3761 ꢁ2 between the subunits. The N-terminal Rossman-fold
domain of subunit A makes contacts with the C-terminal heli-
cal domain of subunit B forming a cleft which constitutes the
active site of the enzyme (vide infra); this interface is stabilised
by inter-subunit interactions including that between Ser114 (A)
at the beginning of helix a4 and Glu265 (B) in helix a8. The N-
terminal domains are at the periphery of the dimer. At the end
of N-terminal domain A, the interdomain helix a7 protrudes
and inserts through a hydrophobic channel in the C-terminal
domain of B (Figure S7) to then emerge and continue as the C-
terminal helical domain of A, which contacts the N-terminal
domain of subunit B. Reciprocal salt bridges between Arg285
(A or B) and Glu274 (A or B) assist in stabilization of the dimer.
As described above, DALI-based analysis of the structure of
Q1EQE0 revealed that the structures of known activity that
most resemble Q1EQE0 are NADPH-dependent dehydrogenas-
es acting on isobutyrate substrates of the GHBDH or HIBDH
families. However, in representatives of each of those cases, al-
though the structures are dimeric, extensive domain swapping
is not observed. Superimposition of the Q1EQE0 monomer
with a monomer of 2CVZ for example, a HIBDH from Thermus
thermophilus HB8[28] (Figure S8A and S8B) shows that there is
good fold conservation of the N-terminal domain and the in-
terdomain helix a7 (RMSD 1.7 ꢁ for residues 18–209 in Q1EQE0
and 2–190 for 2CVZ). There is also good fold conservation be-
tween the isolated C-terminal domains (2.0 ꢁ for residues 232–
303 in Q1EQE0 superimposed with residues 212–282 for
2CVZ). However a sharp b-turn at residue Ser204 in 2CVZ in
the helix equivalent to a8 in Q1EQE0 results in the chain re-
turning in the direction of the N-terminal domain of A, to con-
tinue and form the C-terminal helical bundle required to com-
plete the active site cleft within the same monomer. In
Q1EQE0, helix a8 continues at the equivalent point (Gln224)
and travels away from the N-terminal domain to form the C-
terminal helical bundle that will form the active site cleft with
the N-terminal domain of subunit B.
The characteristic GXGXXG consensus sequence for NADPH
binding in Q1EQE0 is present as G(26) LGMLG (31). The ribose-
2’-phosphate of NADPH is secured by interactions with the fol-
lowing residues: Arg50, which also makes p-stacking interac-
tions with the adenine ring of the ADP moiety; Lys54; Asn49;
and the side-chain and peptidic NÀH of Thr51. The 2’ and 3’-
ribose hydroxyls interact with the side chain of Ser111 and the
backbone of Ser111, Val83 and Ser84. The re-face of the nicoti-
namide ring stacks against the side chain of Met30; the si-face
is presented to the active-site cavity (Figure 3). Protic residue
side chains, which may have a role in mechanism (vide infra)
within a 9 ꢁ sphere of the nicotinamide ring include Thr254
from subunit B and Ser111 and Asp187 from subunit A.
In NAD(P)H-dependent ketoreductases such as the HIBDH
2CVZ, a protic residue within the vicinity of the active site nico-
tinamide ring will be required to provide a proton to the nas-
cent hydroxyl group in the reductive direction. In 2CVZ, this
function is fulfilled by Lys165.[28] Superimposition of the active
sites of Q1EQE0 and 2CVZ reveal that the lysine residue is
replaced by the aspartate Asp187 (Figure S11). Mutation of this
residue to either alanine (Asp187Ala) or asparagine
(Asp187Asn) resulted in inactive enzyme variants. Additionally,
the wild-type Q1EQE0 displayed no HIBDH activity towards
commercially available 3-hydroxyisobutyrate when supplied
with NADP+.
Domain swapping of this kind in the dehydrogenases has
not been reported extensively in the literature, although one
further outcome of the DALI-based analysis was the identifica-
tion of two structures that share more significant structural ho-
mology with Q1EQE0 throughout the length of the chain, and
which also participate in domain swapping. These are 3QHA
(from Mycobacterium avium 104) and 3L6D (from P. putida
KT2440), although in each case annotation is restricted to
putative oxidoreductases of as yet undetermined activity. A su-
The biological reduction of imines by possible imine reduc-
tases has been the focus of some recent interest owing to the
role of these activities in natural product biosynthesis, but also,
increasingly because of the potential for enzymes capable of
asymmetric imine reduction for preparative biocatalysis. Prior
to either of those areas being explored, however, the best-
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1375

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
mide ring is bound within the core of the Rossman fold, within
which it is brought into contact with the DHFR substrate as
observed in PDB structure 7DFR.[30] In Irp3, the nicotinamide
ring is presented at the surface of the N-terminal domain to
the active site (Figure S12B), as a result of rotation around the
bond connecting the ribose to the diphosphate, relative to the
case with DHFR. In this respect, Q1EQE0 is more similar to Irp3
except that the active site is formed within one subunit in the
latter enzyme.
The putative mechanisms of C=X (where X=N or O) bond
reduction in DHFR, Irp3 and the ketoreductase 2CVZ, which
Q1EQE0 resembles structurally, appear to be distinct. DHFRs
from E. coli and other bacteria are thought to first protonate
the imine through a water molecule, to form an iminium ion
that is reduced by hydride.[32] In Irp3, hydride transfer is
thought to be followed by proton transfer from either a histi-
dine (His101) or a tyrosine (Tyr128) residue, both of which may
be expected to be in the protonated form at physiological
pH.[15] In 2CVZ, a proton is thought to be delivered to the nas-
cent alcohol product by the protonated lysine residue
Lys165,[28] which superimposes well with Asp187 in Q1EQE0.
Mutation of Asp187 in Q1EQE0 to either asparagine or alanine
yields an inactive enzyme. In the absence of a structure of
Q1EQE0 in complex with both NADPH and 1 or other sub-
strate, it is difficult to precisely determine the role of this resi-
due in catalysis, although its role as a proton donor to the
imine may be facilitated by an increase in pKa as a result of
being sandwiched between the hydrophobic residues Leu191
and Leu137; an environment not provided in 2CVZ, in which
a serine (Ser117) and an asparagine (Asn169) interact with
Lys165. Protonation would result in an iminium ion that would
be the substrate for reduction by hydride (Scheme 2). The
stabilisation of such an iminium ion may be fostered by the
predominantly negatively charged active-site channel.
Figure 3. Active site of Q1EQE0. The peptide backbone is shown in ribbon
format with subunits A in dark grey and B in light grey. Side chains of sub-
units A and B are shown in cylinder format with carbon atoms in dark grey
and light grey respectively. NADPH is shown in cylinder format with carbon
atoms in grey. The dashed black line represents hydrogen bonding interac-
tion between Ser111 and NADPH. Electron density map represents the FoÀFc
omit map refined in the absence of NADPH, contoured at a level of 3s. The
NADPH atoms from the refined complex have then been added for clarity.
studied imine reductase is undoubtedly DHFR, such as the
enzyme from E. coli, and which catalyses the asymmetric re-
duction of dihydrofolate to tetrahydrofolate at the expense of
NADPH.[9] Many structures of this enzyme and its mutants have
been reported, in complex with both the NADPH cofactor,[29]
substrate dihydrofolate,[30] methotrexate[30] and other inhibi-
tors.[31] Recently, the structure of an imine reductase (Irp3)
acting on a thiazoline imine attached to an acyl-carrier protein
from Yersinia enterocolitica, and with homology to the PchG
enzyme from the pyochelin biosynthetic pathway in P. putida,
has also been presented.[15] None of these enzymes has yet
been shown to possess asymmetric imine reductase activity to-
wards imine substrates of synthetic interest however. Q1EQE0
and the RIR and SIR from Streptomyces spp. described by Mit-
sukura and co-workers represent the first examples of isolated
enzymes capable of such reductions, and the structure of
Q1EQE0 allows these reductive reactions to be put into an
enzyme structural context for the first time. The low values of
catalytic efficiency of Q1EQE0 towards 1, 3 and 5 (Table 1)
clearly suggest that these are not the natural substrates for the
enzyme. The identities of genes surrounding that which enco-
des Q1EQE0 in the genome of S. kanamyceticus are also not
informative in this context. The active site of Q1EQE0 is large
however, hinting perhaps at a larger natural substrate than
those studied herein. It is also possible that the reduction of
imines represents a promiscuous activity for these oxidoreduc-
tases of otherwise uncharacterised activity.
Conclusions
The structure of Q1EQE0 represents the first opportunity to
study the active-site determinants of mechanism and selectivi-
ty in NADPH-reductases that might be of use in industrial bio-
catalytic processes for the reduction of imine substrates. The
exact mechanism of the imine reduction remains to be eluci-
dated. Whilst the substrate specificity of this and related wild-
type enzymes appear to be narrow, the structures will provide
a basis for protein engineering experiments targeted at im-
proving enzyme activity and altering characteristics such as
stability, process suitability, substrate range and enantioselec-
tivity.
It is interesting to compare the structure and proposed
mechanism of DHFR and Irp3 with Q1EQE0. While each of
these reductase enzymes shares the Rossman fold associated
with binding the ADP moiety of the NADP cofactor, both
Q1EQE0 and Irp3 possess a separate C-terminal domain that is
absent in DHFR. DHFR binds both cofactor and substrate
within a discrete single domain that shares some structural ho-
mology with both other imine reductases, notably of the core
b-sheet region (Figure S12A). In DHFR, however, the nicotina-
Experimental Section
Gene synthesis, cloning, expression and protein purification:
The gene encoding Q1EQE0 was synthesised by GeneArt (Invitro-
gen), with codons optimised for expression in E. coli. The full gene
sequence can be found in the Supporting Information (Figure S13).
The gene was amplified by PCR from the commercial template
using the following primers: 5’-CCAGGG ACCAGC AATGCC GGATAA
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1376

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
with 1 min intervals and the solu-
ble and insoluble fractions were
separated
by
centrifugation
(26892g, 30 min) in a Sorvall SS34
rotor. The clear supernatant was
loaded onto a 5 mL His-Trap che-
lating HP nickel column. After
washing with Tris·HCl buffer (ten
column
volumes)
containing
imidazole (30 mm), the Q1EQE0
protein was eluted with a gradient
of imidazole (30–500 mm) over
20 column volumes. Column frac-
tions containing Q1EQE0 (as deter-
mined by SDS-PAGE analysis) were
pooled and then concentrated
using a 10 kDa cut-off Centricon
filter membrane. The concentrated
enzyme was then loaded onto
a
pre-equilibrated S75 Superdex
16/60 gel-filtration column, and
eluted with the Tris·HCl buffer
(120 mL) at
1 mLminÀ1
pure Q1EQE0, as determined by
SDS-PAGE analysis were pooled
and stored at 48C.
a
flow rate of
.
Fractions containing
Scheme 2. Contrasting mechanisms proposed for carbonyl and imine group reduction in HIBDH and Q1EQE0,
respectively. A) Lys165 protonates the nascent alcohol in the reduction of hydroxyisobutyric acid by 2CVZ;[28]
B) Asp187 of Q1EQE0 provides stabilization and protonation to a substrate iminium form of 1, which is reduced
by hydride from NADPH to yield the R product.
Protein crystallisation: Crystalliza-
tion conditions for Q1EQE0 were determined using a range of
commercially available trial screens in sitting-drop format 96-well
plates with 300 nL drops. The best initial crystals were obtained in
conditions containing 35% (v/v) tascimate at pH 7.0 and protein at
a concentration of 10 mgmLÀ1 (using a 1:1 ratio of protein and
precipitant solution). Larger crystals for diffraction analysis using
optimised conditions were prepared using the hanging-drop
method in 24-well plate Linbro dishes with 2 mL drops consisting
of 1 mL of protein at a concentration of 10 mgmLÀ1 and 1 mL of
the reservoir. The best crystals were obtained in 35% (v/v) tasci-
mate with 2.5% (v/v) ethylene glycol. Cocrystallisation with NADPH
(10 mm) did not yield crystals under the same buffer conditions. In
order to obtain the NADP(H) complex, native crystals were trans-
ferred from the growth drop into a cryogenic solution which con-
sisted of the mother liquor containing 10% (v/v) glycerol and
NADPH (10 mm), and incubated for 1 min, after which they were
immediately flash-cooled in liquid nitrogen. Native crystals were
flash-cooled in a cryogenic solution containing the only the
mother liquor and 10% (v/v) glycerol. All crystals were tested for
diffraction in-house using a Rigaku Micromax-007HF fitted with
Osmic multilayer optics and a Marresearch MAR345 imaging plate
detector. Those crystals that diffracted to greater than 3 ꢁ resolu-
tion were retained for full dataset collection at the synchrotron.
TCCGAG CACCAA AGGTC-3’ (forward) and 5’-GAGGAG AAGGCG
CGTTAT TATTTA CCGCTA TGGGTA CGAAAC TGTTC-3’ (reverse). After
gel analysis of the PCR product, the relevant band was eluted from
the gel using a PCR cleanup kit (Qiagen). The gene was then sub-
cloned into the pET-YSBL-LIC-3C vector using a published ligation-
independent cloning procedure.[33] The resultant plasmid MRM1
was then used to transform E. coli XL1-Blue cells (Novagen), yield-
ing colonies which in turn gave plasmids using standard miniprep
procedures that were sequenced to confirm the identity and se-
quence of the gene. The native gene was used as a template for
the creation of mutants Asp187Ala and Asp187Asn using a Quick-
change site-directed mutagenesis kit from Stratagene, using the
manufacturer’s protocol. The primers for the Asp187Ala were:
5’-CAAGCC TGTATG CGGCCG CAGGTC TGG-3’ (forward) and
5’-CCAGAC CTGCGG CCGCAT ACAGGC TTG-3’ (reverse). For
Asp187Ala, the primers used were: 5’-GGCAAG CCTGTA TAATGC
CGCAGG TC-3’ (forward) and 5’-GACCTG CGGCAT TATACA GGCTTG
CC-3’ (reverse). Following cloning, the presence of the designed
mutation sites was confirmed by sequencing.
The recombinant vector(s) containing the Q1EQE0 gene and its
Asp187Ala and Asp187Asn mutants were used to transform E. coli
BL21 (DE3) cells using kanamycin (30 mgmLÀ1) as antibiotic marker
on Luria–Bertani (LB) agar. A single colony from an agar plate
grown overnight was used to inoculate LB broth (5 mL), which was
then grown overnight at 378C with shaking at 180 rpm. The starter
culture served as an inoculum for a culture of LB broth (500 mL) in
which cells were grown until the optical density (OD600) had
reached a value of 0.6. Expression of Q1EQE0 was then induced by
the addition of isopropyl b-d-1-thiogalactopyranoside (final con-
centration of 1 mm). The culture was then incubated at 188C in an
orbital shaker overnight at 180 rpm for approximately 18 h. The
cells were harvested by centrifugation (4225g, 15 min) in a Sorvall
RC5B Plus centrifuge with a Sorvall GS3 rotor and then resuspend-
ed in Tris·HCl buffer (50 mm, pH 7.5, 50 mL) containing NaCl
(300 mm). The cells were then sonicated for 3ꢃ30 s bursts at 48C
Data collection, structure solution, model building and refine-
ment of Q1EQE0: Complete datasets for native Q1EQE0 and the
NADPH complex were collected on beam lines I03 and I04-1 respec-
tively at the Diamond Light Source (Didcot, UK) The data were pro-
cessed and integrated using XDS[34] and scaled using SCALA[35] as
included within the Xia2 processing system.[36] Data collection sta-
tistics are given in Table 1. The crystals of each complex were iso-
morphous and were in space group C121. The structure of the
enzyme was solved using BALBES,[24] which selected a truncated
dimer model of the tartronic semialdehyde reductase from S. typhi-
murium LT2 (PDB ID: 1YB4; 19% amino acid sequence identity to
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1377

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
QE1QE0 as a search model). The solutions each contained four
molecules in the asymmetric unit, representing two dimers. The
solvent content in each case was 69%. The structures were built
using Autobuild in the Phenix suite of programs[37] and Coot[38] and
refined using REFMAC[39] employing local NCS restraints. For the
NADPH complex, following building and refinement of the protein
and water molecules, the omit maps were observed to contain
clear residual density at the subunit–dimer interface, which was
modelled and refined as NADPH. The final structures exhibited Rcryst
and Rfree values of 18.5 and 21.1% (native) and 18.5 and 21.6%
(NADPH). Each structure was finally validated using PROCHECK.[40]
Refinement statistics are presented in Table 2. The Ramachandran
plot for the native enzyme showed 97.6% of residues to be situat-
ed in the most favoured regions, 1.6% in additional allowed and
0.8% outlier residues. For the NADPH complex, the corresponding
values were 96.6, 2.7 and 0.7% respectively. The coordinates and
structure factors for the native imine reductase and the NADPH
complex have been deposited in the Protein Data Bank with the
PDB IDs 3ZGY and 3ZHB, respectively.
grams were monitored at 254 nm. The eluant was potassium phos-
phate buffer (10 mm, pH 2.5) and methanol in a 55:45 ratio. The
elution times of the S and R enantiomers of 1 were 49.1 and
52.0 min respectively.
Acknowledgements
We thank the Spanish Ministry of Science and Innovation for
a FPU fellowship to M.R.M.; the industrial affiliates of the Centre
of Excellence for Biocatalysis, Biotransformations and Biomanu-
facture (CoEBio3) for awarding a studentship to E.W.; the Marie-
Curie ITN programme BIOTRAINS (FP7-PEOPLE-ITN-2008-238531)
for awarding a studentship to A.F.; F.L. received funding from the
European Union’s Seventh Framework Programme FP7/2007-
2013 under grant agreement n8 266025 (BIONEXGEN).
Keywords: amines
·
biocatalysis
·
imines
·
NADPH
·
Enzyme assays: The NADPH-dependent imine reductase activity of
recombinant QE1QE0 and its Asp187Ala and Asp187Asn mutants
was assessed using UV spectrophotometry.[20] Activity was deter-
mined on a Spectramax M2 spectrophotometer/plate reader (Mo-
lecular Devices, Sunnyvale, CA) by monitoring the decrease of
NADPH either at 340 nm (e=6.22 mmÀ1 cmÀ1) for substrate 1 or at
370 nm (e=2.216 mmÀ1 cmÀ1 for 3 and 5 as these substrates ab-
sorbed strongly at 340 nm. Activity was determined using microt-
est plates (Sterilin, Newport, UK). Reaction mixtures (0.2 mL) con-
tained pure enzyme (0.3 mg), NADPH (750 mm) and the appropriate
amount of substrate (2 mL) from a 100-fold concentrated stock in
DMSO. The reaction was started by adding the enzyme to the mix-
ture and the reaction followed for 5 min. One unit of imine reduc-
tase is defined as the amount of protein that oxidised 1 mmol
NADPH per minute. Kinetic parameters were determined using pu-
rified enzyme and analysed by nonlinear regression analysis based
on Michaelis–Menten kinetics using the program QtiPlot.
oxidoreductases
[1] M. Hçhne, U. T. Bornscheuer, ChemCatChem 2009, 1, 42–51.
[2] S. Buchholz, H. Grçger in Biocatalysis in the Pharmaceutical and Biotech-
nology Industries (Ed.: R. N. Patel), CRC Press, Boca Raton, 2006,
pp. 829–847.
[3] S. Mathew, H. Yun, ACS Catal. 2012, 2, 993–1001.
[4] R. Carr, M. Alexeeva, M. J. Dawson, V. Gotor-Fernꢄndez, C. E. Humphrey,
N. J. Turner, ChemBioChem 2005, 6, 637–639.
[5] K. A. Nolin, R. W. Ahn, F. D. Toste, J. Am. Chem. Soc. 2005, 127, 12462–
12463.
[6] M. Rueping, E. Sugiono, C. Azap, T. Theissmann, M. Bolte, Org. Lett.
2005, 7, 3781–3783.
[7] Z. Wang, X. Ye, S. Wei, P. Wu, A. Zhang, J. Sun, Org. Lett. 2006, 8, 999–
1001.
ˇ
ˇ
´
[8] A. V. Malkov, K. Vrankovꢄ, S. Stoncius, P. Kocovsky, J. Org. Chem. 2009,
74, 5839–5849.
[9] D. R. Smith, J. M. Calvo, Nucleic Acids Res. 1980, 8, 2255–2274.
[10] T. Eguchi, T. Oshiro, Y. Kuge, K. Mochida, T. Uwajima (Kyowa Hakko
Kogyo Co., Ltd.), EP0356934A2, 1990.
[11] T. Vaijayanthi, A. Chadha, Tetrahedron: Asymmetry 2008, 19, 93–96.
[12] M. Espinoza-Moraga, T. Petta, M. Vasquez-Vasquez, V. F. Laurie, L. A. B.
Moraes, L. S. Santos, Tetrahedron: Asymmetry 2010, 21, 1988–1992.
[13] H. Li, P. Williams, J. Micklefield, J. M. Gardiner, G. Stephens, Tetrahedron
2004, 60, 753–758.
Biotransformations: In order to confirm the enantioselectivity of
the reduction of 1 by Q1EQE0, biotransformations were performed
using the purified enzyme with NADPH-recycling by using the
method of Faber and co-workers.[41] Reactions were performed in
a
12 mL total reaction volume containing pure Q1EQE0
(0.25 mgmLÀ1) and substrate 1 (5 mm) from a stock solution
(0.25m) in DMF. The final concentration of dimethylformamide in
the reaction was 2% (v/v). For cofactor recycling NADPH (5 mg,
0.006 mmol), glucose-6-phosphate (3.5 mg, 0.013 mmol) and glu-
cose-6-phosphate dehydrogenase suspension (12 mL; G8404 from
Sigma–Aldrich) was added per mL of total reaction volume. After
overnight reaction at 308C, samples from biotransformations with
2-methyl-1-pyrroline (1), an aqueous sample of 100 mL was mixed
with triethylamine (100 mL of a 0.8% (w/v) solution, equivalent to
11.00 mLmLÀ1) in acetonitrile and 2,3,4,6-tetra-O-acetyl-b-d-gluco-
pyranosyl isothiocyanate (GITC, 200 mL of a 0.8% (w/v) solution
equivalent to 8 mgmLÀ1). The mixture was incubated at 378C for
30 min. Insoluble fractions were removed by centrifugation and
the supernatant analysed by HPLC.
[14] C. Reimmann, H. M. Patel, L. Serino, M. Barone, C. T. Walsh, D. Haas, J.
Bacteriol. 2001, 183, 813–820.
[15] K. M. Meneely, A. L. Lamb, Biochemistry 2012, 51, 9002–9013.
[16] W. Li, A. Khullar, S. Chou, A. Sacramo, B. Gerratana, Appl. Environ. Micro-
biol. 2009, 75, 2869–2878.
[17] W. Li, S. Chou, A. Khullar, B. Gerratana, Appl. Environ. Microbiol. 2009, 75,
2958–2963.
[18] T. Nakano, K. Miyake, H. Endo, T. Dairi, T. Mizukami, R. Katsumata, Biosci.
Biotechnol. Biochem. 2004, 68, 1345–1352.
[19] K. Mitsukura, M. Suzuki, K. Tada, T. Yoshida, T. Nagasawa, Org. Biomol.
Chem. 2010, 8, 4533–4535.
[20] K. Mitsukura, M. Suzuki, S. Shinoda, T. Kuramoto, T. Yoshida, T. Nagasa-
wa, Biosci. Biotechnol. Biochem. 2011, 75, 1778–1782.
[21] K. Yanai, T. Murakami, M. Bibb, Proc. Natl. Acad. Sci. USA 2006, 103,
9661–9666.
[22] T. Nagasawa, T. Yoshida, K. Ishida, H. Yamamoto, N. Kimoto (Daicel
Chemical Industries, Ltd.), US 20110287494, 2011.
[23] T. Nagasawa, T. Yoshida, K. Ishida, H. Yamamoto, N. Kimoto (Daicel
Chemical Industries, Ltd.), US 20110262977, 2011.
Reversed-phase HPLC for samples of GITC-derivatised 2 was per-
formed on an Agilent system equipped with a G1322A degasser,
G1311A binary pump, a G1329A well-plate autosampler unit, a
[24] F. Long, A. A. Vagin, P. Young, G. N. Murshudov, Acta Crystallogr. Sect. D
Biol. Crystallogr. 2008, 64, 125–132.
G1316A temperature-controlled column compartment and
a
G1315B diode array detector. The column Luna 3 mm C18(2) 100 ꢁ
(Phenomenex 150 mm length, 4.6 mm diameter, 3 mm particle size)
was used. A typical injection volume was 10 mL and chromato-
[25] J. Osipiuk, M. Zhou, A. Moy, F. Collart, A. Joachimiak, J. Struct. Funct. Ge-
nomics 2009, 10, 249–253.
[26] L. Holm, C. Sander, Science 1996, 273, 595–560.
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1378

CHEMBIOCHEM
FULL PAPERS
www.chembiochem.org
[27] E. Krissinel, K. Henrick, J. Mol. Biol. 2007, 372, 774–797.
[28] N. K. Lokanath, N. Ohshima, K. Takio, I. Shiromizu, C. Kuroishi, N. Okaza-
ki, S. Kuramitsu, S. Yokoyama, M. Miyano, N. Kunishima, J. Mol. Biol.
2005, 352, 905–917.
[29] M. R. Sawaya, J. Kraut, Biochemistry 1997, 36, 586–603.
[30] C. Bystroff, S. J. Oatley, J. Kraut, Biochemistry 1990, 29, 3263–3277.
[31] R. L. Summerfield, D. M. Daigle, S. Mayer, D. Mallik, D. W. Hughes, S. G.
Jackson, M. Sulek, M. G. Organ, E. D. Brown, M. S. Junop, J. Med. Chem.
2006, 49, 6977–6986.
[37] P. D. Adams, P. V. Afonine, G. Bunkꢅczi, V. B. Chen, I. W. Davis, N.
Echoo Is, J. J. Headd, L.-W. Hung, G. J. Kapral, R. W. Grosse-Kunstleve,
A. J. McCoy, N. W. Moriarty, R. Oeffner, R. J. Read, D. C. Richardson, J. S.
Richardson, T. C. Terwilliger, P. H. Zwart, Acta Crystallogr. Sect. D Biol.
Crystallogr. 2010, 66, 213–221.
[38] P. Emsley, K. Cowtan, Acta Crystallogr. Sect. D Biol. Crystallogr. 2004, 60,
2126–2132.
[39] G. N. Murshudov, A. A. Vagin, E. J. Dodson, Acta. Crystallogr. Sect. D Biol.
Crystallogr. 1997, 53, 240–255.
[40] R. A. Laskowski, M. W. Macarthur, D. S. Moss, J. M. Thornton, J. Appl.
Crystallogr. 1993, 26, 283–291.
[32] C. M. Czekster, A. Vandemeulebroucke, J. S. Blanchard, Biochemistry
2011, 50, 367–375.
[41] M. Hall, C. Stueckler, W. Kroutil, P. Macheroux, K. Faber, Angew. Chem.
2007, 119, 4008–4011; Angew. Chem. Int. Ed. 2007, 46, 3934–3937.
[33] K. E. Atkin, R. Reiss, N. J. Turner, A. M. Brzozowski, G. Grogan, Acta Crys-
tallogr. Sect. F Struct. Biol. Cryst. Commun. 2008, 64, 182–185.
[34] W. Kabsch, Acta Crystallogr. Sect. D Biol. Crystallogr. 2010, 66, 125–132.
[35] P. Evans, Acta Crystallogr. Sect. D Biol. Crystallogr. 2006, 62, 72–82.
[36] G. Winter, J. Appl. Crystallogr. 2010, 43, 186–190.
Received: May 20, 2013
Published online on June 28, 2013
ꢂ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2013, 14, 1372 – 1379 1379