Binding of the anticancer prodrug CB1954 to the activating enzyme NQO2 revealed by the crystal structure of their complex

Article abstract of DOI:10.1021/jm050730n

CB1954 is an attractive prodrug for directed-enzyme prodrug therapy (DEPT) and a conventional prodrug against tumors in which the enzyme NQO2 is highly expressed. We have determined the crystal structure of the NQO2-CB1954 complex to 2.0 ? resolution. The binding of the prodrug is governed by hydrophobic forces, while two key electrostatic contacts determine the specific orientation of the ligand. The structure also reveals an unfavorable interaction, therefore suggesting possible avenues for DEPT-tailored engineering studies.

Full text of DOI:10.1021/jm050730n

7
714
J. Med. Chem. 2005, 48, 7714-7719
Binding of the Anticancer Prodrug CB1954 to the Activating Enzyme NQO2
Revealed by the Crystal Structure of Their Complex
Majed AbuKhader,§,† John Heap,
§,‡,†
§,
Cristina De Matteis, Barrie Kellam, Stephen W. Doughty,
§
§
§
‡
Nigel Minton, and Massimo Paoli *
School of Pharmacy and The Institute of Infection, Immunity and Inflammation, Centre for Biomolecular Sciences,
University of Nottingham, University Park, Nottingham, NG7 2RD, UK
Received July 28, 2005
CB1954 is an attractive prodrug for directed-enzyme prodrug therapy (DEPT) and a
conventional prodrug against tumors in which the enzyme NQO2 is highly expressed. We have
determined the crystal structure of the NQO2-CB1954 complex to 2.0 Å resolution. The binding
of the prodrug is governed by hydrophobic forces, while two key electrostatic contacts determine
the specific orientation of the ligand. The structure also reveals an unfavorable interaction,
therefore suggesting possible avenues for DEPT-tailored engineering studies.
12
The use of prodrugs to treat cancer relies on the
enzymatic activation of a relatively harmless compound
into a cytotoxic molecule. The nitro-aromatic compound
CB1954 (5-(aziridin-1-yl)-2,4-dinitrobenzamide, Figure
GDEPT/VDEPT;10,11 and CDEPT ). These two-step
approaches involve the delivery of an enzyme to the
cancer tissue, followed by the administration of a
specific prodrug which is activated by the enzyme
causing tumor-specific cytotoxicity. To this end, the
E.coli NTR/CB1954 system has been implemented in
1
) has been extensively studied as a prodrug since it
was first synthesized over 30 years ago during a series
1
13,14
of drug design studies. CB1954 is relatively nontoxic
clinical trials.
to cells but is transformed to a potent cytotoxic agent
upon reduction of its 4-nitro group to a 4-hydroxylamine
Another enzyme with pronounced CB1954-nitrore-
ductase activity is NQO2;1
5-17
this is of special interest
(Figure 1). The latter acts as a bifunctional alkylating
since it is expressed at high levels in several human
3
agent capable of forming DNA-DNA interstrand cross-
links, via the hydroxylamine and aziridine groups.2
Originally, it was demonstrated to inhibit in vivo tumor
growth in rat Walker 256 carcinoma cell lines1,4 follow-
ing activation by the enzyme NQO1 (also known as
NAD(P)H:quinone oxidoreductase 1 or QR1 or DT-
cancer cell lines. Expression of NQO2 in cancer tissue
,3
means that the use of CB1954 in a conventional prodrug
therapy approach, without the complications of the
DEPT strategies, can be reevaluated, and this is cur-
1
rently under clinical investigation. NQO2 shares 49%
sequence identity with NQO1 and was therefore thought
to be an isozyme of NQO1, although it is 43 amino acids
shorter. Recent work in mice suggests that NQO2 acts
as an endogenous factor protecting against chemical
carcinogenesis.18 Interestingly, NQO2 is not active with
NAD(P)H, but receives electrons from NRH (dihydro-
nicotinamide riboside) and other reduced pyridinium
5
diaphorase).
NQO1, a mammalian detoxifying NAD(P)H-depend-
ent flavoenzyme expressed in several tissues, catalyzes
the two-electron reduction of a number of quinone
6
metabolites and also appears to function as a nitro-
reducing enzyme (involving a four-electron reduction).
NQO1 protects the body from oxidative damage of
semiquinones produced in the one-electron reduction of
quinones by several enzymes (e.g. cytochrome P450).7
On the basis of the experiments in rat, CB1954 was
tested on human cancer cell lines. The poor results
obtained have been attributed to the presence of Gln in
the human enzyme in place of Tyr104 in the rat enzyme.
Site-directed mutagenesis of Gln104 to Tyr confers to
the human enzyme similar CB1954-nitroreductase ac-
tivity to the rat enzyme.8,9
1
derivatives. Coadministration of an appropriate elec-
tron donor is essential for efficacy of any therapy using
NQO2-mediated bioactivation of CB1954. Both NQO1
and NQO2 contain the redox cofactor FAD which
receives electrons from a donor and then transfers these
onto the substrate. These homodimeric proteins contain
two FAD molecules bound at the extensive interface
formed between the monomers.17 The “head” of the FAD,
the three-ring isoalloxazine system, is held in the active
site by noncovalent interactions. A similar arrangement
is observed in the bacterial NTR, which contains the
cofactor FMN (FMN has an identical headgroup but a
shorter tail with respect to FAD). It has been estab-
lished that transfer of two electrons, in the form of a
hydride anion, takes place from the reduced nitrogen
atom at position 5 (N5) in the center of the isoalloxazine
ring.1
Efficient turnover of CB1954 is also carried out by
the E.coli nitroreductase (NTR), and this observation
has been key to the development of treatments to target
cells using directed enzyme prodrug therapy (DEPT)
(
e.g.: antibody-directed, gene/virus-directed and clostrid-
ia-directed enzyme prodrug therapy, ADEPT and
9,20
*
To whom correspondence should be addressed. Tel. (0115)8468015,
Even though NTR catalyzes the reduction of the nitro
group at either position 2 or 4 of the CB1954 aromatic
ring, and therefore half of the product is the less
cytotoxic 2-hydroxylamine, its rates of turnover are
fax (0115)9513412, e-mail max.paoli@nottingham.ac.uk.
§
School of Pharmacy.
‡
Institute of Infection, Immunity and Inflammation.
†
These authors contributed equally to the work.
1
0.1021/jm050730n CCC: $30.25 © 2005 American Chemical Society
Published on Web 11/02/2005

Binding of the Anticancer Prodrug CB1954
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7715
Figure 1. The enzyme NQO2 activates the prodrug CB1954 (1) through a four-electron reduction of the nitro group at the 4
position to the hydroxylamine derivative (3), presumably via the nitroso derivative (2). This will require two electron donor molecules
with two successive two-electron reduction steps in which an intermediary product dissociates from the active site to allow access
to the electron donor necessary to restore the cofactor FAD to its reduced state.
sufficiently high to yield a therapeutic effect. Human
NQO2 catalyzes the reduction of CB1954 at the 4
were consistent with the characteristics of CB1954. The
pseudo-symmetry of the ligand’s electron density how-
ever made it difficult to interpret the data unambigu-
ously for the assignment of the prodrug’s orientation.
Visual inspection indicates that there are four possible
orientations of CB1954 which might fit the data; in each
of these the ligand is parallel to and stacked onto the
isoalloxazine ring of the FAD, but with differing sub-
stituents facing Trp105, a residue that distinctively
packs beside the cofactor perpendicular to it. To resolve
this issue, our approach has been to carry out four
separate model refinement procedures, with the ligand
in each of the four possible orientations. In two orienta-
tions, both with the nitro groups both pointing away
-1
position only, with rates comparable to NTR (Kcat 6 s
)
and a better Km (NQO2: 263 µM and NTR: 862 µM).1
Both rat NQO1 and human NQO2 have a tyrosine that
packs underneath the isoalloxazine ring, but NQO2 has
better kinetic parameters for the activation of the
prodrug. Efforts have been made to gain insights into
the binding of the CB1954 compound to the active site
of these enzymes. Molecular modeling studies based on
the structures of rat21 and human NQO1 suggested
two possible binding modes for the association of the
22
1
7,22,23
prodrug in the active site.
The modeling work on
the rat NQO1, which also encompassed data from
mutagenesis results, places the aziridine group near
Tyr155 with carbon atoms 2 and 3 of the benzene ring
positioned above N5 of the FAD. In the model based on
the structure of the human enzyme, the aziridine is
thought to engage in hydrophobic interactions with
Trp105, thus putting the aziridine nitrogen and the
benzene carbon 5 close to N5.
from Trp105, peaks of negative density (in F - Fc
o
difference maps) resulted over the atoms of the nitro
groups, thus demonstrating that these models were
incorrect. On the other hand, the arrangements with
either the 4-NO or the 2-NO in front of Trp105 gave
2
2
satisfactory and comparable refinement results. The
former orientation puts the 4-NO group in proximity
2
The only structural data reported in the literature to
to the flavin’s N5, while the latter results in a nonpro-
date on a CB1954-protein complex are those with the
ductive binding mode. As shown in Figure 2, the shape
of the electron density map provides a good fit for both
arrangements. At 2.0 Å resolution it is not possible to
discriminate between the density of the aziridine and
carboxamide groups and thus the possible coexistence
of two distinct binding modes cannot be confirmed (see
Experimental Section). Previously, it was shown in
another flavoprotein that the averaging effect of alter-
nate orientations in the active site could not be resolved
at 1.5 Å but only with data to 0.9 Å.26 In both our refined
models, the B factors for the ligand atoms range
2
4
NTR enzyme. The authors interpreted the electron
density as indicating that the prodrug is bound in a
given orientation in one monomer and a different
orientation in the other. In one of these arrangements,
the ligand has its 4-nitro oxygen atoms in proximity of
the N5 of the FAD which implies reduction to the
4
-hydroxylamine derivative, the most cytotoxic product,
while in the other monomer the binding of the com-
pound is such that reduction would occur at the 2
position.
2
Herein we report the crystal structure of the NQO2-
CB1954 complex refined against data to 2.0 Å resolution
between 50 and 54 Å while those for the FAD and the
2
protein atoms are in the 20-50 and 20-30 Å range,
(
the cloning, expression, and purification procedures are
respectively. It is possible that the densities shown in
Figure 2 result from an average between a productive
and a nonproductive binding mode of the prodrug.
Nonproductive binding has been observed in the struc-
ture of the antibiotic nitrofurazone bound to E.coli
presented in the Supporting Information). The structure
was solved using crystals soaked in solutions of artificial
mother liquor containing the ligand. The starting model
was derived from the coordinates of the native (apo)
enzyme (PDB code: 1QR2, ). The refined model of the
NQO2-CB1954 complex has excellent crystallographic
statistics and very good geometry; the only Ramachan-
dran outlier is Tyr132, but its conformation is supported
by an indisputable electron density map in both mono-
mers, as also seen in the original structure of NQO2
2
5
27
NTR; in addition, “flipped binding geometries” of the
substrate have been observed in different redox states
in the flavoenzyme PETN.28 This would imply that the
active site conformation of the oxidized form is some-
what different from that of the reduced form. However,
at least in the case of the E.coli NTR, only slight atomic
2
5
(
PDB code 1QR2 ).
shifts in protein atoms were detected when the reduced
structure was compared to the oxidized form.2
4,29
It is
Initial maps showed large peaks of density in the
active sites of both monomers, indicating the presence
of a bound ligand; the size and shape of these peaks
known that the isoalloxazine ring changes from a very
slightly bent “butterfly” conformation in the oxidized

7
716 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
AbuKhader et al.
Figure 2. Electron density maps for the final 2F
o
- F
c
syntheses contoured at 0.8 σ are shown over two refined models of the
above
CB1954 ligand, one with the 4-NO above N5 (panels A and B for chains A and B, respectively) and the other with 2-NO
2
2
N5 (panels C and D for chains A and B, respectively). Density for Trp105 and Asp161, as well as for the cluster of ordered water
molecules in the active site, is also shown. Difference maps and omit maps were also carefully examined; the refinement gives
good results with the ligand in either of these orientations.
state to a more bent form in the reduced state of many
flavoenzymes.20 The extent of this effect may differ in
different flavoenzymes. In addition, in the oxidized
structures of most flavoproteins, the N5 atom of the
isoalloxazine ring is engaged in an interaction with a
respectively, as well as residues 51-59 of a disordered
loop in the B chain). All the large differences between
the two monomers occur at surface residues and do not
involve atoms within the active site. The position of the
ligand and its interactions with the protein are the
same, within experimental error, in both monomers. A
minor difference, in the A chain but not in the B chain,
is given by the presence of a water molecule that
provides bridging contacts between the 2-NO2 and the
carbonyl oxygen of Gly174. With respect to the modeling
2
0
hydrogen bonding donor. Upon reduction of the FAD
N5, this interaction is likely to be affected, and this
might be associated with structural changes. It is
therefore conceivable that even small structural changes
could, in concert, affect the binding of the ligand. In our
crystallographic analysis, refinement of the ligand with
its 2-NO2 group placed toward Trp105 forces the car-
boxamide group in plane with the nitro group, resulting
in a highly unfavorable steric clash (from the oxygen
or nitrogen atom of the carboxamide to one of the
oxygens of the nitro group: 2.0-2.1 Å). With the 4-NO2
in front of Trp105, the electron density map clearly
indicates that the carboxamide is not coplanar with the
benzene ring and the nitro group. The latter conforma-
tion appears more consistent with the high-resolution
X-ray crystallographic structure of CB195430 in which
the carboxamide is offset at 64.7° from the plane of the
aromatic ring, and the neighboring nitro group at
position 2 is offset at 26.3° (compared to the 4-NO2 at
work carried out on the binding of CB1954 to the NQO1
(DT-diaphorase) enzyme,2
2,23
the orientation of the
prodrug in the crystal structure in the NQO2 complex
is different to what would perhaps have been expected
based on the similarity between the active sites of the
two enzymes. It is unclear whether the amino acid
differences between rat NQO1, human NQO1 and
human NQO2, might account for the different predicted
and observed orientations of CB1954.
The prodrug fills at least half of the space in the
binding pocket, aligning itself over the pyrimidine dione
ring of the FAD headgroup (Figure 3); the remaining
space, mainly contained within a pocket beyond the
benzo-fused architecture of the FAD head, is occupied
in both monomers by a cluster of well-ordered water
molecules, clearly defined by the electron density map
1
3.6°); the authors comment on the role of steric
30
hindrance in determining this conformation. Semiem-
pirical and ab initio molecular orbital calculations of the
energies of these two conformations of isolated CB1954
showed substantial differences, with the conformation
arising from the ligand with 2-NO2 pointed toward
Trp105 showing substantially higher energy, of the
order of 300 to 400 kJ mol (see Supporting Informa-
tion). In light of the likely enhanced stability, together
with the productivity of its binding orientation, the
remainder of this discussion will focus on the model with
the 4-NO2 positioned toward Trp105.
(
Figure 2). It is striking to note that six of the eight
water molecules in the NQO2-CB1954 complex are in
the same position (within experimental error) as in the
structure of the native (apo) enzyme (PDB code 1QR2; );
25
only two waters have moved (1.5-2.0 Å) upon arrival
of the ligand. Binding of CB1954 induces little pertur-
bation in this oxidized active site, the structure of which
is presumably rather rigid, as indicated by the con-
served positions of the solvent molecules. Hydrophobic
side chains that engage in interactions with CB1954
show only very slight shifts but preserve their confor-
mation, relative to the native NQO2 structure. Further
evidence of structural rigidity is provided by the complex
of NQO2 with the inhibitor resveratrol:31 this larger
-
1
The conformations of the two monomers are very
similar, with an rms deviation of 0.8 Å for all well-
ordered main chain atoms (excluding the first three and
last three residues of the N- and C-terminal ends

Binding of the Anticancer Prodrug CB1954
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7717
Figure 3. Surface representation of the active site of NQO2 in the apo-form (A), in complex with the prodrug CB1954 (B) and
in complex with the inhibitor resveratrol (C). Both the FAD cofactor and the ligands are shown in a stick model while water
molecules are drawn green in a cpk model.
2
4
ligand has to displace the waters in order to bind but
does not cause any significant shifts in atomic positions
within the active site. A comparison of the binding of
resveratrol and CB1954 is highlighted in Figure 3.
interaction types (PDB code 1IDT ), the binding of
CB1954 to NQO2 is governed by hydrophobic contacts.
There are, however, two important hydrogen-bonding
interactions (Figure 4b). The first involves one of the
nitro groups and Asn161; either nitro group could in
theory interact with Asn161, as is apparent from Figure
CB1954 is bound at an angle of 35° in the A chain
and 37° in the B chain (relative to the FAD); the plane
of the benzene ring cannot adopt a perfectly parallel
orientation with respect to the plane of the isoalloxazine
2
. In the second interaction, the carboxamide group of
the prodrug is engaged in contacts with the main chain
carbonyl oxygen of Gly149 (Figure 4b). This interaction
cannot take place if the 2-NO2 group is placed above
the N5 atom of the FAD; it is therefore likely to provide
a key anchoring point for the binding of the prodrug in
its correct orientation for catalysis and therefore deter-
mines the specificity of the enzyme for the hydroxylation
at the 4 position.
25
ring, as in the case of menadione (PDB code 2QR2; ),
since this would cause steric hindrance between the
carboxamide of the prodrug and the atoms of the peptide
segment 149-150. Despite this mode of binding, the
association is clearly driven by extensive π-stacking and
hydrophobic interactions. Over 40 contacts with inter-
atomic distances less than 4 Å exist between the ligand
and the FAD. In addition, the snug fit with the roof of
the active site, lined with a hydrophobic triad formed
by Phe126, Ile128, and Phe178, results in numerous
nonpolar contacts, as illustrated in Figure 4a. In fact,
formation of the complex is accompanied by the burial
of a significant amount of solvent accessible surface
In all flavoproteins the substrate atoms at the site of
reduction are in proximity to the FAD N5 (within 3.8
Å) to facilitate direct hydride transfer.20 In the structure
of the NQO2-CB1954 complex, the 4-NO2 group is
placed with one oxygen atom at 3.0 Å (3.1 Å in the B
chain) and the nitrogen at 3.2 Å (3.3 Å in the B chain)
from the N5 (Figure 4b). These distances are closer than
the average distance observed across a range of flavoen-
zymes20 and are also shorter than in the E.coli NTR-
CB1954 complex (oxygen at 3.8 Å and nitrogen at 3.9 Å
from the N5 of the isoalloxazine ring). If in NQO2 the
hydride anion is transferred to the closest atom, then
reduction will take place at one of the oxygen atoms of
the nitro group of CB1954. This would imply that the
reaction mechanism is via a nitroso intermediary prod-
uct, an idea previously alluded to32 and supported by
experimental work that revealed an extremely fast
reduction of the CB1954 nitroso derivative by NQO133
and of a nitrofurazone nitroso derivative by E.coli
NTR.27 Such a mechanism requires two successive two-
electron reduction steps and hence, since the enzyme
active site can be occupied by only one substrate at any
one time, it is presumed that the intermediary product
has to dissociate and leave the enzyme for the FAD to
be reduced again by an electron donor. In stoichiometri-
cal terms the overall reaction from the nitro to the
hydroxylamine derivative is a four-electron reduction
as indicated in Figure 1, in contrast to many diagrams
in the literature in which the stoichiometry is not
referred to. It is worth noting that the reduction of the
nitroso derivative of CB1954 can be achieved nonenzy-
matically by NADH and other weak biological reducing
agents.33
2
2
area, 228 Å in the A chain and 212 Å in the B chain,
of which about 80% is nonpolar. While the complex with
E. coli NTR is characterized by a range of different
Figure 4. Interactions governing binding and molecular
recognition of CB1954 in the NQO2 active site. (A) The roof of
the active site is lined with a hydrophobic triad formed by
Phe126, Ile128 and Phe178 that make extensive nonpolar
contacts with the bound prodrug. Both the residues (cyan) and
the ligand (red) are displayed in a space-filling model which
highlights the extent of the interactions and the complemen-
tary fit of the atomic surfaces. (B) CB1954 (red) makes two
electrostatic interactions with the protein (green), one with
Asn161, the other with the main chain of Gly149. The latter
is likely to be a key anchor to favor the binding of the prodrug
in a productive orientation. There is, however, a set of
unfavorable contacts between the nitro group and Trp105.
Finally, the close approach of the nitro oxygen atom to the
FAD N5 is emphasized.
Our structure provides crucial insights into the mo-
lecular interactions governing the recognition of the

7
718 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
AbuKhader et al.
prodrug by the enzyme. One unusual feature is the
arrangement of the nitro group in front of Trp105, with
some interatomic distances down to 3.3 Å (Figure 4b).
Given the chemical properties of the indole ring and the
nitro group, their close proximity is not expected to be
energetically favorable and is presumably compensated
by the numerous other favorable interactions between
the prodrug and the protein. Interestingly, in E.coli
NTR, despite its very different active site structure, an
aromatic group is also relatively close to the nitro group,
with a geometry reminiscent of that reported here in
terms of the distances between the N5/nitro group to
the aromatic side chain (in NTR distances from N5 to
Phe124 atoms are in the 6-7 Å range and from 4-NO2
to Phe124 in the 3.6-4.0 Å range; in NQO2, N5 to
Trp105: 4-5 Å range and 4-NO2 to Trp105: 3.2-3.8 Å
range). The residue Phe124 was extensively mutated
in engineering experiments on NTR and several changes
were found to improve the activity with CB1954, in
Trp105, the other with the 2-NO
2
group placed toward Trp105.
Refinement procedures, carried out for two models with the
ligand in either of the above orientations, gave the same
results: 19.3% and an Rfree of 24.8%.
Refinement carried out with coordinated in which models
for both orientation were included, each with the occupancy
of the ligand atoms set at 0.5, gave comparable results with
the previous refinement procedures (R-factor 19.3% and Rfree
24.7%).
The final model was evaluated with PROCHECK, and
structural analyses were carried out using COOT, DeepView
3
9
7
40
41
(
www.expasy.org/spdbv), and CCP4 programs. Figures were
drawn using either the software PyMOL (http://pymol.source-
42
forge.net) or VMD.
Coordinates have been deposited in the Protein Data Bank
with accession code 2BZS.
Acknowledgment. We are grateful to P. McEwan
for help with data collection. We thank P. Moody for
helpful discussions and R. Eastham and S. Schneider
for critical reading of the manuscript. This work was
supported by funds provided by the University of
Nottingham and the BBSRC.
3
4
particular mutations to Lys and Asn. On the basis of
the structure of NQO2 in complex with CB1954 pre-
sented here, we hypothesise that mutagenesis of Trp105
may produce better engineered variants as potential
candidates for DEPT.
Note Added in Proof
Upon submission of our atomic coordinates, 2BZS, to
the Protein Data Bank, it was noted that two files for
the same complex were recently deposited by other
groups with accession codes 1ZX1 and 1XI2. The latter
was described in a paper that has just been published
by Z. Zhang and colleagues.
Experimental Section
The procedures for cloning, expression, and purification are
available as Supporting Information. Below details of the
crystallographic work are provided.
Crystallization and Data Collection. Crystals of NQO2
were grown in sitting-drops using the vapor diffusion method.
A volume of 10 µL of protein solution was mixed with an equal
volume of reservoir solution (1.66 M ammonium sulfate in 100
mM Na-HEPES pH 7, 12 µM FAD, 1 mM dithiothreitol) and
allowed to equilibrate. Crystals grew in about one week; they
were then transferred into a solution of artificial mother liquor
containing the CB1954 prodrug at a concentration of 2 mM
Supporting Information Available: Procedures for clon-
ing, expression, and purification; table with statistics for the
crystallographic data. This material is available free of charge
via the Internet at http://pubs.acs.org.
References
(
1) Knox, J. R.; Burke, P. J.; Chen, S.; Kerr, D. J. CB 1954: from
the Walker tumor to NQO2 and VDEPT. Curr. Pharm. Des.
2003, 9, 2091-2104.
(
from a stock solution prepared in DMSO). Crystals were left
soaking for 3 days. For the X-ray diffraction experiments,
crystals were frozen using a mixture of 50% glycerol, PEG 400,
and MPD (all at 20% final concentration) as cryoprotectant in
a solution of artificial mother liquor. Data were collected at
(2) Knox, R. J.; Friedlos, F.; Boland, M. P. The bioactivation of CB
1
954 and its use as a prodrug in antibody-directed enzyme
prodrug therapy (ADEPT). Cancer Metastasis Rev. 1993, 12,
95-212.
1
-
180 °C on a Rigaku R-AXIS IV2+ detector using Cu KR
(
3) Knox, R. J.; Jenkins, T. C.; Hobbs, S. M.; Chen, S.; Melton, R.
G.; Burke, P. J. Bioactivation of 5-(Aziridine-1-yl)-2,4-dinitroben-
zamide (CB1954) by human NADPH Quinone Oxidoreductase
2: A Novel Cosubstrate-mediated Antitumor Prodrug Therapy.
Cancer Res. 2000, 60, 4179-4186.
radiation from a Rigaku Micromax-007 rotating anode. Reflec-
tions were indexed, integrated, and scaled using the programs
35
36
Mosflm and SCALA. Details of the data processing statistics
are reported in Table 1. Efforts to collect synchrotron higher
resolution data were made but were not successful.
(
4) Knox, R. J.; Boland, M. P.; Friedlos, F.; Coles, B.; Southan, C.;
Roberts, J. The nitroreductase enzyme in Walker cells that
activates 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954) to
5-(aziridin-1yl)-4-hydroxylamino-2-nitrobenzamide is a form of
NAD(P)H dehydrogenase (quinone). Biochem. Pharmacol. 1998,
37, 4671-4677.
5) Boland, M. P.; Knox, R. J.; Roberts, J. J. The differences in
kinetics of rat amd human DT diaphorase result in a differential
sensitivity of derived cell lines to CB1954 (5-(aziridin-1-yl)-2,4-
dinotrobenzamide). Biochem. Pharmacol. 1991, 41, 867-875.
6) Chen, S.; Wu, K.; Knox, R. Structure-function studies of DT-
diaphorase (NQO1) and NRH: quinone oxidoreductase (NQO2).
Free Radical Biol. Med. 2000, 29, 276-284.
7) Monks, T. J.; Hanzlik, R. P.; Cohen, G. M.; Ross, D.; Graham,
D. G. Quinone chemistry and toxicity. Toxicol. Appl. Pharm.
1992, 112, 2-16.
Model Refinement and Structural Analysis. Crystals
were isomorphous with those previously obtained and de-
2
5
scribed, apart from some small changes in the cell dimen-
3
7
sions. Refmac was employed to carry out rigid body refine-
ment using the native (apo) human NQO2 as a starting model
(
2
5
(
PDB code 1QR2 ), after all waters were removed from this
file. This procedure, using only data to 3.0 Å, decreased the
R-factor from values above 50% to 34.2% and the Rfree from
above 50% to 34.9%. Visualization of electron density map,
rebuilding, and model fitting were carried out on a Linux
(
(
3
8
system using the program COOT. Positional and B-factor
refinement using Refmac reduced the R-factor down to 23.8%
and the Rfree to 27.0%, which was followed by limited manual
rebuilding and addition of water molecules. Further refine-
ment gave an R-factor of 20.0% and an Rfree of 25.3%.
(8) Chen, S.; Knox, R.; Wu, K.; Deng, P. S.-K.; Zhou, D.; Bianchet,
M. A.; Amzel, L. Mario. Molecular basis of the catalytic differ-
ences among DT-diaphorase of human, rat, and mouse. J. Biol.
Chem. 1997, 272, 1437-1439.
No ligand atoms were included in the model till this point;
omit maps were carefully examined and refinement procedures
with the CB1954 model in four different orientations were
carried out. Analysis of difference maps allowed us to rule out
two possible orientations. It is not possible, however, to
discriminate between the two other orientations of the
(9) Wu, K.; Eng, E.; Knox, R.; Chen, S. Demonstration of the
Activation of Prodrug CB 1954 Using Human DT-Diaphorase
Mutant Q104Y-Transfected MDA-MB-231 Cells and Mouse
Xenograft Model. Arc. Biochem. Biophys. 2001, 385, 203-208.
10) Niculescu-Duvaz, I.; Cooper, R. G.; Stribbling, S. M.; Heyes, J.
A.; Metcalfe, J. A.; Springer, C. J. Recent developments in gene-
directed enzyme prodrug therapy (GDEPT) for cancer. Curr.
Opin. Mol. Ther. 1999, 4, 480-6.
(
2
ligand: one orientation with the 4-NO group placed toward

Binding of the Anticancer Prodrug CB1954
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7719
(
11) Jung, M. Antibody Directed Enzyme Prodrug Therapy (ADEPT)
and Related Approaches for Anticancer Therapy. Mini Rev. Med.
Chem. 2001, 1, 399-407.
(27) Race P. R.; Lovering A. L.; Green R. M.; Ossor A.; White S. A.;
Searle P. F.; Wrighton C. J.; Hyde E. I. Structural and mecha-
nistic studies of Escherichia coli nitroreductase with the anti-
biotic nitrofurazone - Reversed binding orientations in different
redox states of the enzyme. J. Biol. Chem. 2005, 280, 13256-
13264.
(28) Barna, T. M.; Khan, H.; Bruce, N. C.; Barsukov, I.; Scrutton, N.
S.; Moody, P. C. E. Crystal structure of pentaerythriol tetrani-
trate reductase: “flipped” binding geometries for steroid sub-
strates in different redox states of the enzyme. J. Mol. Biol. 2001,
(
12) Minton, N. P. Clostridia in cancer therapy. Nature Rev. Micro-
biol. 2003, 1, 1-6.
(
13) Chung-Faye, G.; Palmer, C.; Anderson, D.; Clark, J.; Downes,
M.; Baddeley, J.; Hussain, J.; Murray, P. I.; Searle, P.; Seymour,
L.; Harries, P. A.; Ferry, D.; Kerr, D. J. Virus-directed, enzyme-
prodrug therapy with nitroimidazole reductase: a phase I and
pharmacokinetic study of its prodrug, CB1954. Clin. Cancer Res.
2
001, 7, 2662-2668.
14) Denny, W. A. Nitroreductase-based GDEPT. Curr. Pharm. Des.
002, 8, 1349-1361.
15) Long, D. J., II; Jaiswal, A. K. NRH: quinone oxidoreductase2
NQO2) Chemino-Biol. Interact. 2000, 129, 99-112.
3
10, 433-447.
(
(
(
(
(
(
(
(
29) Haynes, C. A.; Koder, R. L.; Miller, A. F.; Rodgers, D. W.
2
Structures of nitroreductase in three states. J. Biol. Chem. 2002,
2
30) Iball, J.; Scrimgeour, S. N.; Williams, B. C. The crystal and
molecular structure of 2,4-dinitro-5-ethyleneiminobenzamide.
Acta Crystallogr. 1975, B31, 1121-1123.
77, 11513-11520.
(
16) Skelly, J. V.; Knox, R. J.; Jenkins, T. C. Aerobic nitroreduction
by flavoproteins: enzyme structure, mechanism and role in
cancer chemotherapy. Mini Rev. Med. Chem. 2001, 1, 293-306.
17) Bianchet, M. A.; Faig, M.; Amzel, L. M. Structure and mecha-
nism of NAD(P)H: quinone accpetor oxidoreductases (NQO).
Methods Enzymol. 2004, 382, 144-174.
31) Buryanovskyy, L.; Fu, Y.; Ma, Y.; Hsieh, T.; Wu, J. M.; Zhang,
Z. Crystal structure of quinone reductase 2 in complex with
resveratrol. Biochemistry 2004, 43, 11417-11426.
(
(
32) Knox, R. J.; Chen, S. Quinone reductase-mediated nitro-reduc-
tion: clinical applications. Methods Enzymol. 2004, 382, 194-
18) Iskander, K.; Paquet, M.; Brayton, C.; Jaiswal, A. K. Deficiency
of NRH: quinone oxidoreductase 2 increases susceptibility to
2
21.
7
, 12-dimethyal(a)athracene and benzo[a]pyrene-induced skin
33) Knox, R. J., Friedlos, F., Biggs, P. J., Flitter, W. D., Gaskell,
M., Goddard, P., Davies, L., Jarman, M. Identification, synthesis
and properties of 5-(aziridin-1-yl)-2-nitro-4-nitrosobenzamide, a
novel DNA cross-linking agent derived from CB1954. Biochem.
Pharmacol. 1993, 46, 797-803.
carcinogenesis. Cancer Res. 2004, 64, 5925-5928.
(
(
(
19) Ghisla, S.; Massey, V. Mechanism of flavoprotein-catalysed
reactions. Eur. J. Biochem. 1989, 181, 1-17.
20) Fraaije, M. W.; Mattevi, A. Flavoenzymes: diverse catalysis with
recurrent features. Trends Biol. Sci. 2000, 25, 126-132.
21) Li, R.; Bianchet, M. A.; Talalay, P.; Amzel, L. M. The three-
dimensional structure of NAD(P)H: quinine reductase, a fla-
voprotein involved in cancer chemoprotein and chemotherapy:
mechanism of the two-electron reaction. Proc. Natl. Acad. Sci.
U.S.A. 1995, 92, 8846-8850.
(
34) Grove, J. I.; Lovering, A. L.; Guise, C.; Race, P. R.; Wrighton, C.
J.; White, S. A.; Hyde, E. I.; Searle, P. F. Generation of
Escherichia coli nitroreductase mutants conferring improved cell
sensitization to the prodrug CB1954. Cancer Res. 2003, 63,
5
532-5537.
(
(
(
35) Leslie, A. G. W. Joint CCP4 and ESF-EACMB Newsletter on
(
(
(
22) Skelly, J.; Sanderson, M. R.; Suter, D. A.; Baumann, U.; Read,
M. A.; Gregory, D. S. J.; Bennett, M.; Hobbs, S. M.; Neidle, S.
Crystal structure of human DT-diaphorase: a model for interac-
tion with the cytotoxic prodrug 5-(aziridin-1-yl)-2,4-dinitroben-
zamide (CB1954). J. Med. Chem. 1999, 42, 4325-4330
23) Chen, S.; Wu, K.; Zhang, D.; Sherman, M.; Knox, R.; Yang, C.
S. Molecular characterization of binding of substrates and
inhibitors to DT-diaphorase: combined approach involving site-
directed mutagenesis, inhibitor-binding analysis and computer
modeling. Mol. Pharm. 1999, 56, 272-278.
Protein Crystallography 1992, 26.
36) Evans, P. Joint CCP4 and ESF-EACMB Newsletter on Protein
Crystallography 1997, 33, 22-24.
37) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
macromolecular structures by the maximum-likelihood method.
Acta Crystallogr. Sect. D , Biol. Crystallogr. 1997, 53, 240-255.
(38) Emsley P.; Cowtan K. Coot: model-building tools for molecular
graphics Acta Crystallogr. Sect. D, Biol. Crystallogr. 2004, 60,
2126-2132.
(39) Laskowski, R. A.; MacArthur, M. W.; Moss, D. S.; Thornton, J.
M. PROCHECK: a program to check the stereochemical quality
of protein structures. J. Appl. Crystallogr. 1993, 26, 283-291.
(40) Guex, N.; Peitsch, M. C. SWISS-MODEL and the Swiss-
PdbViewer: an environment for comparative modeling. Elec-
trophoresis 1997, 18, 2714-23.
(41) Collaborative Computational Project Number 4. The CCP4
Suite: Programs for protein crystallography. Acta Crystallogr.
1994, D50, 760-763.
(42) Humphrey, W.; Dalke, A.; Schulten, K. “VMD - Visual Molecular
Dynamics”, J. Mol. Graphics 1996 14, 33-38.
24) Johnasson, E.; Parkinson, G. N.; Denny, W. A.; Neidle, S. Studies
on the nitroreductase prodrug-activating system. Crystal struc-
ture of complexes with the inhibitor dicoumarol and dinitroben-
zamide prodrugs and of the enzyme active form. J. Med. Chem.
2
003, 46, 4009-4020.
(
25) Foster, C. E.; Bianchet, M. A.; Talalay, P.; Zhao, Q.; Amzel, L.
M. Crystal structure of human quinine reductase type 2,
metalloflavoprotein. Biochemistry 1999, 38, 9881-9886.
26) Khan, H.; Barna, T.; Harris, R. J.; Bruce, N. C.; Barsukov, I.;
Munro, A. W.; Moody, P. C. E.; Scrutton, N. S. Atomic resolution
structures and solution behaviour of enzyme-substrate com-
plexes of Enterobacter cloacae PB2 pentaerythriol tetranitrate
reductase. J. Biol. Chem. 2004, 279, 30563-30572.
(
JM050730N