Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy

Article abstract of DOI:10.1016/j.bcp.2016.07.015

The clinical stage anti-cancer agent PR-104 has potential utility as a cytotoxic prodrug for exogenous bacterial nitroreductases expressed from replicating vector platforms. However substrate selectivity is compromised due to metabolism by the human one- and two-electron oxidoreductases cytochrome P450 oxidoreductase (POR) and aldo-keto reductase 1C3 (AKR1C3). Using rational drug design we developed a novel mono-nitro analog of PR-104A that is essentially free of this off-target activity in vitro and in vivo. Unlike PR-104A, there was no biologically relevant cytotoxicity in cells engineered to express AKR1C3 or POR, under aerobic or anoxic conditions, respectively. We screened this inert prodrug analog, SN34507, against a type I bacterial nitroreductase library and identified E. coli NfsA as an efficient bioactivator using a DNA damage response assay and recombinant enzyme kinetics. Expression of E. coli NfsA in human colorectal cancer cells led to selective cytotoxicity to SN34507 that was associated with cell cycle arrest and generated a robust ‘bystander effect’ at tissue-like cell densities when only 3% of cells were NfsA positive. Anti-tumor activity of SN35539, the phosphate pre-prodrug of SN34507, was established in ‘mixed’ tumors harboring a minority of NfsA-positive cells and demonstrated marked tumor control following heterogeneous suicide gene expression. These experiments demonstrate that off-target metabolism of PR-104 can be avoided and identify the suicide gene/prodrug partnership of E. coli NfsA/SN35539 as a promising combination for development in armed vectors.

Full text of DOI:10.1016/j.bcp.2016.07.015

Biochemical Pharmacology xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Biochemical Pharmacology
journal homepage: www.elsevier.com/locate/biochempharm
Rational design of an AKR1C3-resistant analog of PR-104
for enzyme-prodrug therapy
Alexandra M. Mowday a, Amir Ashoorzadeh a, Elsie M. Williams b,1, Janine N. Copp b,2, Shevan Silva a,
Matthew R. Bull a, Maria R. Abbattista a, Robert F. Anderson a,c, Jack U. Flanagan a,c, Christopher P. Guise a,c
,
David F. Ackerley b,c, Jeff B. Smaill a,c, Adam V. Patterson a,c,
⇑
a Auckland Cancer Society Research Centre, School of Medical Sciences, The University of Auckland, Auckland 1023, New Zealand
b School of Biological Sciences, Victoria University of Wellington, Wellington 6140, New Zealand
c Maurice Wilkins Centre for Molecular Biodiscovery, School of Biological Sciences, The University of Auckland, Auckland 1023, New Zealand
a r t i c l e i n f o
a b s t r a c t
Article history:
The clinical stage anti-cancer agent PR-104 has potential utility as a cytotoxic prodrug for exogenous
bacterial nitroreductases expressed from replicating vector platforms. However substrate selectivity is
compromised due to metabolism by the human one- and two-electron oxidoreductases cytochrome
P450 oxidoreductase (POR) and aldo-keto reductase 1C3 (AKR1C3). Using rational drug design we devel-
oped a novel mono-nitro analog of PR-104A that is essentially free of this off-target activity in vitro and
in vivo. Unlike PR-104A, there was no biologically relevant cytotoxicity in cells engineered to express
AKR1C3 or POR, under aerobic or anoxic conditions, respectively. We screened this inert prodrug analog,
SN34507, against a type I bacterial nitroreductase library and identified E. coli NfsA as an efficient bioac-
tivator using a DNA damage response assay and recombinant enzyme kinetics. Expression of E. coli NfsA
in human colorectal cancer cells led to selective cytotoxicity to SN34507 that was associated with cell
cycle arrest and generated a robust ‘bystander effect’ at tissue-like cell densities when only 3% of cells
were NfsA positive. Anti-tumor activity of SN35539, the phosphate pre-prodrug of SN34507, was
established in ‘mixed’ tumors harboring a minority of NfsA-positive cells and demonstrated marked
tumor control following heterogeneous suicide gene expression. These experiments demonstrate that
off-target metabolism of PR-104 can be avoided and identify the suicide gene/prodrug partnership of
E. coli NfsA/SN35539 as a promising combination for development in armed vectors.
Received 25 May 2016
Accepted 20 July 2016
Available online xxxx
Chemical compounds studied in this article:
PR-104 (PubChem CID: 11455973)
PR-104A (PubChem CID: 9848786)
SN34037 (PubChem CID: 73671441)
SN34507 (PubChem CID: 90043967)
SN35539 (PubChem CID: 90043246)
Keywords:
Prodrug
Alkylation
Oxidoreductase
Aldo-keto reductase 1C3
Hypoxia
Ó 2016 Elsevier Inc. All rights reserved.
1. Introduction
combination for GDEPT is the nitroreductase (NTR) from
Escherichia coli, NfsB, in combination with the prodrug CB1954
Gene-directed enzyme prodrug therapy (GDEPT) is an approach
whereby cancer tropic vectors such as replication-competent
viruses or bacteria are employed to deliver therapeutic genes to
the tumor microenvironment. The exogenous gene typically intro-
duces a new catalytic function into the tumor microenvironment
that can confer conditional sensitivity to otherwise inert prodrugs
[1]. The most widely studied nitroaromatic enzyme/prodrug
(5-(aziridin-1-yl)-2,4-dinitrobenzamide); see review [2] and
references therein. Although efficacy was demonstrated with this
combination in cell culture models [3,4] and tumor xenografts
using replication-defective viruses [5,6], to date this combination
has demonstrated limited utility in human clinical trials. Poor
aqueous solubility of CB1954 [7], modest kinetics of CB1954 reduc-
tion by NfsB [8], and dose-limiting hepatotoxicity in humans [9]
are thought to be possible reasons for the lack of efficacy observed.
The prodrug PR-104 (Fig. 1A), (2-((2-bromoethyl)(2-((2-hydro
xyethyl)carbamoyl)-4,6-dinitrophenyl)amino) ethyl methanesul-
fonate phosphate ester) represents an alternative nitroaromatic
substrate for NTRs. PR-104 is a water-soluble phosphate ester
‘pre-prodrug’ which undergoes facile conversion to the corre-
sponding lipophilic alcohol PR-104A in plasma, and was initially
designed and optimized as a hypoxia-activated prodrug (HAP) with
⇑
Corresponding author at: Auckland Cancer Society Research Centre, School of
Medical Sciences, The University of Auckland, Auckland 1023, New Zealand.
E-mail address: a.patterson@auckland.ac.nz (A.V. Patterson).
Current address: Department of Chemistry, Emory University, 1515 Dickey Drive,
1
Atlanta, Georgia, USA.
2
Current address: Michael Smith Laboratories, University of British Columbia,
Vancouver, BC, Canada.
http://dx.doi.org/10.1016/j.bcp.2016.07.015
0006-2952/Ó 2016 Elsevier Inc. All rights reserved.
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

2
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
A
B
Tyr-24
Tyr-55
Three Reduction Paths
NHOH(NH2)
NO2
NO2
N
1. Human - O2 dependent
H
N
H
N
H
N
Systemic
phosphatases
Trp-227
Phe-311
X
X
X
N
O
O
N
O
OP(O)(OH)2
OSO2Me
OH
OSO2Me
OH
OSO2Me
2. Human - O2 independent
3. Bacterial - O2 independent
Br
Br
Br
X = NO2 PR-104
X = H SN35539
X = NO2 PR-104A
X = H SN34507
X = NO2; X = H
His-117
Hydroxylamine / Amine
(cytotoxins)
X = NO2 Path 1
3
Pre-prodrug
Prodrug
X = H Path 3 only
Fig. 1. Rationale for design of the prodrug SN34507: (A) Schematic of pre-prodrug PR-104 or SN35539 conversion to prodrug PR-104A and SN34507, respectively, with
subsequent reduction to cytotoxic hydroxylamine or amine metabolites by up to three independent metabolic pathways; (B) Binding model of PR-104A in the active site of
human aldo-keto reductase 1C3 indicating potential for hydrogen-bond formation between the ortho nitro group and tyrosine-24 residue (Figure generated with PyMol
(Schrodinger)).
a bystander effect (diffusion of active metabolites from the cell of
origin [10]). One-electron reduction of the para nitro group on
PR-104A (relative to the nitrogen mustard) by human one-
electron reductases such as cytochrome P450 oxidoreductase
(POR) and methionine synthase reductase (MTRR) [11,12] creates
a nitro radical anion intermediate. The radical anion is back-
oxidized in the presence of molecular oxygen or, in the absence
of oxygen (anoxia), will undergo further electron addition to form
the DNA alkylating hydroxylamine (PR-104H) or amine (PR-104M).
PR-104H and PR-104M can cause inter-strand DNA crosslinks with
disruption of the replication fork upon mitosis, the most likely
mechanism of toxicity [13,14]. Hypoxia selectivity was demon-
strated in a range of cancer cell lines, with hypoxia cytotoxicity
ratios (HCR; aerobic IC50 value divided by anoxic IC50 value) rang-
ing from 5 to 100-fold [10]. However despite clear evidence for
hypoxia-selective activation, atypical aerobic sensitivity of some
cell lines to PR-104A suggested the presence of an aerobic reduc-
tase. The human oxidoreductase AKR1C3 (aldo-keto reductase
member 1C3; P42330) was subsequently shown to reduce
PR-104A in an oxygen independent (two-electron) manner by
bypassing the formation of the oxygen-sensitive nitro radical inter-
mediate, and is now accepted as the major determinant of aerobic
PR-104A sensitivity [15].
based GDEPT. Consistent with this rationale, PR-104 has been eval-
uated in pre-clinical models with considerable success [24].
Ideally a GDEPT prodrug should be an exclusive substrate for a
vector-encoded enzyme. PR-104A is subject to aerobic metabolism
by human AKR1C3 and hypoxic metabolism by various one-
electron oxidoreductases, which may limit the therapeutic speci-
ficity of the prodrug for NTR-armed vectors. In this study we have
used rational drug design to develop more selective GDEPT
prodrugs. We hypothesized that elimination of the ortho nitro of
PR-104 would reduce the ability of AKR1C3 to recognize the dinitro
motif. We also predicted that, in the absence of the electron-
withdrawing effect of the ortho nitro group, the one-electron affin-
ity (E(1)) of the remaining (para) nitro group would be lowered
below that required for metabolism by human oxidoreductase
enzymes [25], while retaining capacity for two-electron reduction
by bacterial NTR enzymes [26]. The use of a single nitro group as
the ‘‘bioreductive switch” should also provide more reactive/cyto-
toxic hydroxylamine and amine metabolites than PR-104H and PR-
104M, as the residual deactivating influence (electron withdrawing
nature) of the ortho nitro group would be absent. Based on the
above hypotheses, we reasoned that replacement of the ortho nitro
group of PR-104A with hydrogen would provide a suitable prodrug
for selective activation by bacterial NTR enzymes. Here we report
the novel PR-104A analog SN34507, and its water-soluble ‘pre-
prodrug’ SN35539. We also identify the major E. coli NTR NfsA as
an effective activating enzyme of SN34507 and demonstrate signif-
icantly improved therapeutic activity of this enzyme/prodrug
combination over that of NfsA/PR-104 in an HCT116 colorectal
cancer xenograft model.
AKR1C3 is a type 2 3
a-HSD (hydroxysteroid dehydrogenase)
and type 17b-HSD/prostaglandin
5
F
synthase [16], and is
expressed at high levels in CD34+ myeloid progenitor cells [17].
Expression of AKR1C3 in these cells is a likely explanation for the
underlying mechanism of the myelosuppression observed in
humans after treatment with high doses of PR-104. Initial Phase I
studies of PR-104 in patients with solid tumors determined that
the primary dose-limiting toxicity was myelosuppression, predom-
inantly neutropenia and thrombocytopenia, with delayed onset
that was sometimes prolonged or irreversible [18,19]. This toxicity
restricted the plasma exposure of PR-104A in humans to approxi-
mately one-third of the levels achievable in mice; mean plasma
2. Methods and materials
2.1. Chemicals
PR-104A was synthesized, purified and stored as previously
reported [27,28]. PR-104 was supplied by Proacta, Inc. SN34037
was synthesized as previously reported [29].
AUC of 39
compared to 120
l
M-h is recorded in humans at 1100 mg/m2 [18,19],
l
M-h in mice at 2144 mg/m2 [20]. The inability
of PR-104 to achieve sufficiently high plasma exposures in humans
has limited its clinical utility in solid tumors.
Nevertheless the improved potency and bystander effect of
PR-104A relative to CB1954 make it a potential candidate for
2.2. Generation of a PR-104A binding model for AKR1C3
A three dimensional model of PR-104A was generated using the
SKETCHER module followed by CONCORD(v.6.1.3) implemented
within the SYBYL(v8.0.3) molecular modeling package (TRIPOS, St
Louis, MO). A single low energy conformer was then generated
by OMEGA(v2.2.1; OpenEye Scientific Software, Sante Fe, NM)
using the MMFF94s forcefield, with the dielectric set at 80, the
maximum number of conformers generated set to 30,000 and
RMS set at 1. The low energy PR-104A conformer was then docked
into the active site of the AKR1C3 enzyme with flufenamic acid
GDEPT. PR-104A demonstrated
a
9-fold improvement in
anti-proliferative activity when compared to CB1954 in
NfsB-expressing cells [21], and there is strong evidence for larger
bystander effects from the dinitrobenzamide mustard class of
compounds at tissue-like cell densities [22,23]. This evidence,
taken together with the large therapeutic ratio demonstrated for
PR-104A in vitro (WT:NfsB IC50 ratio of 2300-fold [21]), suggests
that PR-104A remains a viable candidate for nitroreductase-
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
3
bound (PDB code 1S2C) using GOLD(v4.1.2) [30]. The AKR1C3 flufe-
namic acid structure was prepared for molecular docking by
removing all water atoms and adding hydrogen atoms using
SYBYLv8.0.3. Molecular docking was performed using the Gold-
score fitness function. The docking site was defined as a 20 Å cavity
centered on the carbon atom of the hydride transfer site in the
nicotinamide ring of the NADP co-factor. The search efficiency
was set at 200% and 20 Genetic Algorithm runs were performed
with all poses kept. Both protein and ligand atom types were auto-
matically defined, the option to use a ring template was enabled,
and pyramidal nitrogen atoms and amide bonds were allowed to
flip. The soft potentials 1 option was applied to phenylalanine
306, tryptophan-227 and phenylalanine-311. Only poses that pre-
dicted an interaction with the oxyanion hole were considered
further.
(4.40 mL, 45.36 mmol). The reaction mixture was warm to room
temperature for 30 min, then washed with water (3ꢁ), dried with
Na2SO4 and concentrated under reduced pressure. The residue was
purified by chromatography on silica eluting with CH2Cl2/MeOH
(20:1) (Macron Fine Chemicals, USA) to give bis-mesylate 4
(8.04 g, 96%) as
a d 8.73
yellow gum. 1H NMR [(CD3)2SO]
(t, J = 5.5 Hz, 1H), 8.15–8.10 (m, 2H), 7.27 (d, J = 9.2 Hz, 1H), 4.32
(t, J = 5.3 Hz, 4H), 3.82–3.74 (m, 6H), 3.56–3.40 (m, 5H), 3.13
(s, 6H), 1.76–1.71 (m, 1H), 1.68–1.61 (m, 1H), 1.50–1.47 (m, 4H).
2.3.4. ((2-((2-Hydroxyethyl)carbamoyl)-4-nitrophenyl)azanediyl)bis
(ethane-2,1-diyl) dimethanesulfonate (5)
A solution of bis-mesylate 4 (3.03 g, 5.47 mmol) in dry MeOH
(Macron Fine Chemicals, USA) (100 mL) was treated with methane-
sulfonic acid (Sigma Aldrich, Germany) (17.8 mL, 20.37 mmol). The
reaction mixture was stirred at room temperature for 20 min and
the solvent was evaporated. The residue was dissolved in EtOAc,
washed with water (3ꢁ), dried with Na2SO4 and concentrated
under reduced pressure. The crude product was purified by chro-
matography on silica eluting with CH2Cl2/MeOH (19:1) to give
alcohol 5 (1.92 g, 75%) as a yellow gum. 1H NMR [(CD3)2SO] d
8.73 (t, J = 5.5 Hz, 1H), 8.15–8.10 (m, 2H), 7.27 (d, J = 9.2 Hz, 1H),
4.32 (t, J = 5.3 Hz, 4H), 3.82–3.74 (m, 6H), 3.56–3.40 (m, 5H), 3.13
(s, 6H), 1.76–1.71 (m, 1H), 1.68–1.61 (m, 1H), 1.50–1.47 (m, 4H).
HRMS(ESI) calcd for C15H23N3NaO10S2 [M+Na]+ m/z 492.0717:
found 492.0720.
2.3. Synthesis of SN34507 and SN35539
The prodrug SN34507 was prepared from commercially
available 2-fluoro-5-nitrobenzoic acid in 5 steps. A further 2 steps
provided phosphate pre-prodrug SN35539.
2.3.1. 2-Fluoro-5-nitro-N-(2-((tetrahydro-2H-pyran-2-yl)oxy)ethyl)
benzamide (2)
A stirred solution of 2-fluoro-5-nitrobenzoic acid 1 (Matrix
Scientific, USA) (4.56 g, 24.63 mmol) in SOCl2 (Scharlau Chemicals,
Spain) (50 mL) and DMF (Acros Organics, USA) (3 drops) was
heated under reflux for 4 h. Excess SOCl2 was removed by distilla-
tion under reduced pressure and the residue was dissolved in THF
(Acros Organics, USA) (30 mL), cooled to ꢀ10 °C and treated with
2-((tetrahydro-2H-pyran-2-yl)oxy)ethanamine (Sigma Aldrich,
Germany) (3.93 g, 27.10 mmol) then warmed to room temperature
for 30 min. The solvent was evaporated and the residue dissolved
in EtOAc (Macron Fine Chemicals, USA), washed with water (3ꢁ)
and brine, dried with Na2SO4 (Scharlau Chemicals, Spain) and con-
centrated under reduced pressure. The crude product was purified
by chromatography on silica eluting with EtOAc/Hexane (1:1)
(Macron Fine Chemicals, USA) to give the amide 2 (3.17 g, 41%)
as a colorless oil. 1H NMR [(CD3)2SO] d 8.67 (br, 1H), 8.42–8.38
(m, 2H), 7.64–7.59 (m, 1H), 4.62 (t, J = 3.5 Hz, 1H), 3.55–3.41 (m,
5H), 1.79–1.70 (m, 1H), 1.66–1.60 (m, 1H), 1.53–1.41 (m, 5H).
HRMS(ESI) calcd for C14H17FN2NaO5 [M+Na]+ m/z 335.1014: found
335.1014.
2.3.5. 2-((2-Bromoethyl)(2-((2-hydroxyethyl)carbamoyl)-4-nitroph-
enyl)amino)ethyl methanesulfonate (SN34507)
To a solution of alcohol 5 (8.00 g, 17.04 mmol) in acetone
(240 mL) was added LiBr (Sigma Aldrich, Germany) (1.48 g,
17.04 mmol). The reaction mixture was stirred overnight at room
temperature before the solvent was removed. The residue was dis-
solved in EtOAc, washed with water (2ꢁ), dried with Na2SO4 and
concentrated under reduced pressure. The crude product was puri-
fied by flash column chromatography on silica gel eluting with
CH2Cl2/MeOH (20:1) to give SN34507 (3.74 g, 48%) as a yellow
gum. 1H NMR [(CD3)2SO] d 8.65 (t, J = 5.6 Hz, 1H), 8.14–8.09 (m,
2H), 7.22 (d, J = 9.1 Hz, 1H), 4.75 (br, 1H), 4.31 (t, J = 5.4 Hz, 2H),
3.80–3.74 (m, 5H), 3.65–3.53 (m, 5H), 3.13 (s, 3H). HRMS(ESI) calcd
for C14H21BrN3O7S [M+H]+ m/z 454.0278: found 454.0273.
2.3.2. 2-(Bis(2-hydroxyethyl)amino)-5-nitro-N-(2-((tetrahydro-2H-
pyran-2-yl)oxy)ethyl) benzamide (3)
2.3.6. 2-((2-Bromoethyl)(2-((2-((di-tert-butoxyphosphoryl)oxy)
ethyl)carbamoyl)-4-nitrophenyl)amino)ethyl methanesulfonate (6)
To a stirred solution of SN34507 (2.61 g, 5.76 mmol) in DMF
(2 mL) at 10 °C, was added 1H-tetrazole (Sigma Aldrich, Germany)
(61.9 mL, 26.50 mmol; 3% w/w solution in MeCN), followed by
di-tert-butyl diisopropylphosphoramidite (Chem-Impex Interna-
tional, USA) (95%, 7.68 mL, 23.05 mmol) dropwise. The mixture
was stirred at room temperature for 4 h, then cooled to 5 °C and
treated with m-CPBA (Sigma Aldrich, Germany) (70%, 7.46 g,
43.21 mmol) portion wise, before being warmed to room temper-
ature for 2 h. The reaction was then concentrated and the residue
dissolved in EtOAc, washed with 10% aqueous Na2S2O5 (Scharlau
Chemicals, Spain), 5% aqueous NaHCO3 (Scharlau Chemicals, Spain)
and water before being dried with Na2SO4 and concentrated in
vacuo. The crude product was purified by chromatography on silica
eluting with EtOAc/Hexane (1:1) to give phosphate ester 6 (1.45 g,
39%) as a yellow gum. 1H NMR [(CD3)2SO] d 8.84 (t, J = 5.6 Hz, 1H),
8.13 (2d, J = 2.8 Hz, 1H), 8.09 (d, J = 2.8 Hz, 1H), 7.23 (d, J = 9.3 Hz,
1H), 4.32 (t, J = 5.4 Hz, 2H), 4.00 (q, J = 6.3 Hz, 2H), 3.80–3.74 (m,
4H), 3.64 (t, J = 6.7 Hz, 2H), 3.51 (q, J = 5.6 Hz, 2H), 3.14 (s, 3H),
1.42 (s, 18H). HRMS(ESI) calcd for C22H37BrKN3O10PS [M+K]+ m/z
684.0752: found 684.0740.
Amide 2 (3.17 g, 10.15 mmol) was dissolved in dioxane (Acros
Organics, USA) (150 mL) and treated with Et3N (J.T.Baker Chemi-
cals, The Netherlands) (4.24 mL, 30.45 mmol) and diethanolamine
(Sigma Aldrich, Germany) (3.89 mL, 40.60 mmol). The reaction
mixture was heated to 55 °C overnight then cooled to room tem-
perature, and the solvent was evaporated. The residue was dis-
solved in EtOAc, washed with water (3ꢁ) and brine, dried with
Na2SO4 and concentrated under reduced pressure. The crude pro-
duct was purified by chromatography on silica eluting with
EtOAc/Hexane (3:1) to give diol 3 (3.51 g, 87%) as a yellow gum.
1H NMR [(CD3)2SO] d 8.71 (t, J = 5.5 Hz, 1H), 8.10–8.06 (m, 2H),
7.17 (d, J = 9.0 Hz, 1H), 4.73 (t, J = 5.3 Hz, 2H), 4.62–4.60 (m, 1H),
3.80–3.72 (m, 2H), 3.57–3.49 (m, 5H), 3.47–3.40 (m, 7H),
1.77–1.70 (m, 1H), 1.66–1.61 (m, 1H), 1.50–1.46 (m, 4H).
2.3.3. ((4-Nitro-2-((2-((tetrahydro-2H-pyran-2-yl)oxy)ethyl)carbamoyl)
phenyl)azanediyl)bis(ethane-2,1-diyl) dimethanesulfonate (4)
A solution of diol 3 (6.01 g, 15.12 mmol) in CH2Cl2 (180 mL) was
cooled to 0 °C and treated with Et3N (7.38 mL, 52.93 mmol) fol-
lowed by methanesulfonyl chloride (Acros Organics, USA)
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

4
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
2.3.7. 2-((2-Bromoethyl)(4-nitro-2-((2-(phosphonooxy)
(C10) was calculated by interpolation. Bystander effect efficiency
(BEE) was calculated using the previously reported equation;
[logC10(T) ꢀ logC10(TA)/logC10(T) ꢀ logC10(A)], where C10(T) is the
C10 for target cells in MCLs grown from targets alone, C10(TA) is
the C10 for targets grown in co-culture with activators and the
ethyl)carbamoyl)phenyl)amino)ethyl methanesulfonate (SN35539)
A stirred solution of phosphate ester 6 (1.45 g, 2.24 mmol) in
CH2Cl2 (30 mL) was treated with TFA (10 mL) at room temperature
with stirring for 1 h, then concentrated under reduced pressure to
remove excess TFA. The resulted yellow residue was dissolved in
EtOAc and evaporated to dryness to give SN35539 (1.20 g, 100%)
as a yellow gum. 1H NMR [(CD3)2SO] d 8.86 (t, J = 5.6 Hz, 1H),
8.15–8.11 (m, 2H), 7.24 (d, J = 9 Hz, 1H), 4.32 (t, J = 5.4 Hz, 2H),
4.07–4.02 (m, 2H), 3.79-3.74 (m, 4H), 3.65 (t, J = 6.7 Hz, 2H),
3.51 (q, J = 5.5 Hz, 2H), 3.13 (s, 3H). HRMS(ESI) calcd for
C10(A) is the C10 for activators in the co-cultures [32].
2.9. Statistical analysis 3D
All cell viability data were tested for significance by unpaired
Student’s t-test following confirmation each of the two samples
being compared followed a normal distribution (Shapiro-Wilk nor-
mality test) (SigmaStat version 4.0 integrated with SigmaPlot ver-
sion 12; Systat software, CA, USA). Significance of tumor growth
delay (time for untreated control versus treated tumor population
to exceed four-times intial treatment volume; RTV4) was evalu-
ated by Log rank P-test (SigmaStat version 4.0 integrated with Sig-
maPlot version 12; Systat software, CA, USA).
C
14H21BrKN3O10PS [M+K]+ m/z 571.9500: found 571.9451.
2.4. Preparation and storage of stock solutions
For in vitro studies, prodrug stocks and AKR1C3 inhibitor
(SN34037) were dissolved in DMSO (Acros Organics, USA) (0.1 M)
and stored at ꢀ80 °C. For in vivo studies, PR-104 (as sodium salt
lyophilized with mannitol) (Proacta Inc, USA) was reconstituted
in 2 mL water and diluted in PBS. SN35539 free acid was dissolved
in phosphate-buffered saline containing two equivalents of
NaHCO3. Dosing solutions were prepared fresh, held at room tem-
perature in amber vials, and used within three hours. cLogP values
of the prodrugs were calculated using ChemBioDraw Ultra Version
13.0.
2.10. Animal husbandry
Specific pathogen-free homozygous NIH-III (NIH-Lystbg Foxn1nu
Btkxid) nude mice were obtained from Charles River Laboratories
(Wilmington, MA, USA), bred in the Vernon Jansen Unit (University
of Auckland), and supplied at 7–9 weeks of age. Mice were housed
in groups of 66 in Techniplast microisolator cages with a 12 h
light/dark cycle, and were fed a standard rodent diet (Harlan
Teklad diet 2018i) and water ad libitum. All animals were uniquely
identifiable by ear tag number and weighed 18–25 g at the time of
the experiment. All animal protocols were approved by the Univer-
sity of Auckland Animal Ethics Committee.
2.5. Cell lines and candidate gene expression
HCT116 WT cells were purchased from the ATCC (Manassas, VA,
USA). HCT116 cell lines over-expressing AKR1C3 [15], POR [12] and
E. coli NfsA [21] had been previously generated and validated for
candidate gene expression as described. Frozen stocks were con-
firmed to be mycoplasma free by PCR enzyme-linked immunosor-
bent assay (ELISA) (Roche Diagnostics Corp, Basel, Switzerland).
2.11. Tumor growth delay
Neoplastic cell lines were cultured in
a
-minimal essential medium
The maximum tolerated dose of SN35539 was determined
using a fixed 8-step logarithmic scale with 1.33-fold dose incre-
ments as required. A conservative starting dose was determined
from related compounds. Animals were promptly culled if body
weight loss exceeded 20% or if there were clinical signs of severe
morbidity. The MTD was defined as the highest dose at which all
animals survived with no unacceptable morbidity and mean body
weight loss did not exceed 15%.
using humidified incubator (37 °C, 5% CO2) as previously
a
described [10,23] for a maximum of 12 weeks. Harvested cells
were counted using an electronic particle counter (Z2 Coulter
Particle Analyzer, Beckman Coulter, Florida, USA).
2.6. Measurement of one-electron reduction potential
The E(1) value of SN34507 was determined by measuring the
fast formation of the redox equilibrium (650 ls) between it and
methyl viologen (Acros Organics, USA) as the redox indicator (E
(1) = ꢀ447 7 mV), following pulse radiolysis (2.5 Gy in 200 ns)
of deaerated solutions containing tert-butanol (0.2 M) and phos-
phate buffer (2.5 mM, pH 7).
Tumors were inoculated onto the flanks of female NIH-III nude
mice by subcutaneous injection of 107 cells. Tumor-bearing mice
were randomized to treatment groups when tumors reached treat-
ment size (300–350 mm3) and injected with a single intraperi-
toneal dose of prodrug or vehicle. Tumor size and body weights
were measured 2–3 per week. Tumor volume was calculated as
p
(Lxw2)/6 where L is the major axis and w is the perpendicular
minor axis. Animals were culled when the tumor volume had
increased four-fold relative to pre-treatment volume (RTV4, sur-
vival endpoint) or if body weight loss exceeded 20% of the pre-
treatment value. Kaplan Meier plots were constructed to calculate
median time to endpoint. Treatment efficacy was assessed by com-
paring the median survival time with untreated animals using the
Log-rank test P-test (SigmaStat version 4.0 integrated with Sigma-
Plot version 12; Systat software, CA, USA).
2.7. Anti-proliferative (IC50) assay
The anti-proliferative IC50 was determined as the concentration
of prodrug required for 50% inhibition of cell growth, following a
four hour drug exposure and five days regrowth in the absence
of drug. Assays were performed under oxic or anoxic conditions
as described previously [10,31], the latter using a 5% H2/palladium
catalyst scrubbed Bactron anaerobic chamber (Sheldon Manufac-
turing, OR, USA) to achieve severe anoxia (<10 ppm O2 gas phase)
during prodrug exposure. Total exposure of cells to anoxia did
not exceed 6 h.
2.12. Core NTR over-expression library
The 58-membered candidate NTR gene over-expression library
in E. coli strain SOS-R2 was as described in [21], with the addition
of 11 additional NTR candidates. The additional NTRs included
eight NfsA homologs from Bacillus coagulans (ATCC 7050), Bacillus
thuringiensis serovar konkukian (str. 97-27), Erwinia carotovora
subsp. atrosepticum (SCRI1043), Lactobacillus sakei subsp. sakei
2.8. 3D multicellular bystander assay
Multicellular layers (MCLs) were grown and bystander effect
assays were performed as previously reported [23]. The drug con-
centration required for 10% survival relative to untreated controls
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
5
Table 1
(23K), Listeria welshimeri (ATCC 35897), Mycobacterium smegmatis
(str. MC2 155), Nostoc punctiforme (PCC 73102), and Frp cloned
from a different Vibrio harveyi (KCTC 2720) strain than that previ-
ously published (and herein referred to as Frp2_Vh). The remaining
additional NTRs were KefF and YcaK from E. coli W3110, previously
published in [33], and a previously unpublished YcaK homolog
from Pseudomonas aeruginosa PAO1. Gene specific primers for all
previously unpublished NTR candidates are listed in Table 1.
Primers for amplification and cloning of previously unpublished NTR candidate genes.
Primer
Sequence (50 ? 30)
Nitroreductase specific primer for amplification and cloning into pUCX and
pET28a+
NfsA_Eca_Fw
GGGGCATATGATACCAACTATTGATTTGCTACA
NfsA_Eca_Rv
GGGGCTCGAGCTAGCGTATCGCCCATCCTTGTTG
NfsA_Bc_Fw
GGGGCATATGAATACTATCATTGAAACGATTCTC
NfsA_Bc_Rv
NfsA_Np_Fw
NfsA_Np_Rv
NfsA_Bt_Fw
NfsA_Bt_Rv
NfsA_Ls_Fw
NfsA_Ls_Rv
NfsA_Ms_Fw
NfsA_Ms_Rv
NfsA_Lw_Fw
NfsA_Lw_Rv
Frp2_Vh_Fw
Frp2_Vh_Rv
Ycak_Pa_Fw
Ycak_Pa_Rv
GGGGGTCGACTTACTTTTTATCAAACCCTTGCGT
GGCATATGCCTTTACAGATGGAA
2.13. Bacterial SOS and IC50 assays
GGGTCGACTTACAGTAGCTTGAA
GGGGCATATGAATGAAATGATACATAAAATGGAG
GGGGGTCGACTCATTTCTTATTTAATCCTCTCTCAT
GGGGCATATGTCTGATTTAATCGCACAAATGCAAC
GGGGGTCGACTTATGCTAATGTAAACCCCTGTTTCTT
GGGGCATATGACGGTCATCGCGCGCTACGCAGACGTCGAT
GGGGCTCGAGTCAGCGGATTCCCAGGCCCAACCGCTCG
GGGGCATATGAATCAGGCGATAGATGCCATTTT
GGGGGTCGACTTATTTTTGATTTAAATGTTGC
GGGCATATGAACAATACGATTGAAA
Screening of the NTR library for enzymes able to reduce
SN34507 was carried out in E. coli strain SOS-R2 as previously
described in full [21]. SOS-R2 is a specialized E. coli screening strain
that contains gene knockouts of the NTRs nfsA, nfsB, nemA, azoR and
the export pump tolC. It contains a genomic copy of the lacZ repor-
ter gene under control of a SOS promoter sfiA meaning the DNA
damage experienced by this reporter strain leads to a quantifiable
increase in b-galactosidase expression. The SOS assay was carried
out as previously described [21]. Briefly, in microtiter plate format
NTR over-expression was induced and exponential growth phase
GGGGTCGACTTAGCGTTTTGCTAGCC
GGGGCATATGGTTGAAACCGCCAAGACC
bacteria were exposed to 10 lM of SN34507 for 4 h. Following
GGGGGTCGACCTAGACCTGGCTGCCGAGCT
this an aliquot of cell culture was incubated with reaction
buffer containing ortho nitrophenyl-b-galactosidase to measure
b-galactosidase levels and thus infer the DNA damage present in
each individual NTR over-expression strain as a consequence of
reduction of SN34507 to its DNA damaging metabolites.
Underlined text indicates restriction sites.
media for 24 h. Cells were then fixed in ice cold pure methanol for
30 min at ꢀ20 °C and stained with propidium iodide solution
For the bacterial IC50 assays, overnight cultures of SOS-R2 were
inoculated from glycerol stocks into individual microtiter plate
(20 lg/mL) containing RNAse (100 lg/mL). Analysis of cells was
performed on a flow cytometer (Becton Dickinson FACscan flow
cytometer) and curves were fitted using ModFit LT software (Verity
Software House).
wells containing 200
100
g mLꢀ1 ampicillin and 0.4% glucose, and grown at 30 °C shak-
ing at 200 rpm for ꢂ16 h. The next morning 50 L of overnight cul-
tures was used to inoculate 1 mL of assay media (LB supplemented
with 100 M IPTG) in indi-
g mLꢀ1 ampicillin, 0.2% glucose and 50
vidual 15 mL Falcon tubes. These were incubated at 30 °C, 200 rpm
for 2.5 h, after which 30 L aliquots of culture were added to indi-
vidual wells of a 384 well microplate, each containing 30 L of
assay media supplemented with SN34507 at twice the desired final
concentration (spanning 0–1000 M in a serial dilution). In each
lL of LB media supplemented with
l
l
l
l
3. Results
l
3.1. Drug design rationale for SN34507
l
Using the X-ray crystallographic data for AKR1C3 with flufe-
namic acid bound [34], potential binding modes for PR-104A were
predicted by molecular docking studies, and some located the
para-nitro group of PR-104A in the oxyanion hole within 3 Å of
the bound NADP cofactor, consistent with the potential for
tyrosine-55 mediated hydride transfer and observation of exclu-
sive reduction of the para-nitro by recombinant AKR1C3 [15].
The ancillary ortho nitro group of PR-104A was directed toward
the side chain hydroxyl of tyrosine-24, consistent with the absence
of nitroreduction at this site by AKR1C3 (Fig. 1B). In our preferred
binding mode, the mustard arms are predicted to pack against the
indole side chain of tryptophan-227, while the carboxamide arm
binds in part of the main active site pocket termed subpocket 1
adjacent to phenylalanine-311.
l
independent experiment cultures were assayed in duplicate. The
turbidity of the cultures (OD600) was measured, the plate returned
to 30 °C, 200 rpm for 4 h and then the OD600 was measured again.
Relative increases in OD600 for drug containing vs non-drug
containing wells were then used to calculate percentage growth
inhibition for each over-expression strain. The IC50 value (drug
concentration at which growth inhibition was 50% that of the
unchallenged control) was calculated using Graphpad Prism 6
(GraphPad Software Inc. La Jolla, CA, USA).
2.14. Enzyme kinetics
Recombinant His6-tagged NTRs were purified by nickel-affinity
chromatography (Novagen, Merck, Darmstadt, Germany). FMN
cofactors were reconstituted and proteins were desalted,
quantified, assessed for purity and stored as previously described
[21]. The extinction co-efficient of SN34507 at 400 nm
The potential for an interaction between the ancillary ortho
nitro and tyrosine-24 prompted us to hypothesize that its elimina-
tion on the PR-104A scaffold might minimize contacts within the
active site, and reduce the ability of AKR1C3 to recognize the PR-
104A dinitro motif. The novel GDEPT prodrug SN34507 was thus
created by substituting the ortho nitro group of PR-104 with hydro-
gen (Fig. 1A). Simplifying the PR-104A prodrug scaffold to a single
para-nitro group was also hypothesized to lower the one-electron
affinity (E(1)) of the remaining (para) nitro group below that
required for metabolism by human oxidoreductase enzymes such
as POR, as well as providing more reactive/cytotoxic hydroxy-
lamine and amine metabolites. We synthesized SN34507 (and its
corresponding phosphate ester pre-prodrug SN35539) as described
(Fig. 2).
(e
= 13,400 Mꢀ1 cmꢀ1) was measured as previously described for
PR-104A and steady-state enzyme kinetics for purified NTRs were
assessed as previously described [21] over a SN34507 concentra-
tion range of 5–1000
lM.
2.15. Cell cycle analysis
WT and NfsA-expressing HCT116 cells were exposed to 0.05
lM
of prodrug for four hours before washing and recovery in drug free
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

6
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
NO2
NO2
N
NO2
F
H
N
H
N
c
f
OR
OR
X
N
O
Y
O
O
Y
Br
OMs
1: X = H
2: X = NH(CH2)2OTHP
3: Y = OH, R = THP
4: Y = OMs, R = THP
5: Y = OMs, R = H
SN34507: R = H
6: R = P(O)(OtBu)2
a, b
d
e
g
h
SN35539: R = P(O)(OH)2
a) SOCl2, DMF, 80 oC; b) H2N(CH2)2OTHP, THF, -10 oC to RT; c) diethanolamine, dioxane, Et3N, 55 oC; d) MsCl,
Et3N, DCM, 0 oC; e) MsOH, MeOH; f) 1 eq LiBr, acetone; g) di-tert-butyl-N,N-diisopropylphosphoramidite, 1H-
tetrazole, DMF, 5 oC to RT then m-CPBA, 0 oC; h) TFA, DCM, 5 oC to RT
Fig. 2. The synthesis of SN34507 and SN35539.
3.2. SN34507 is not a biologically relevant substrate for human
AKR1C3
1330 lmol/kg in a growth delay assay using subcutaneous
HCT116 tumor xenografts engineered to over-express AKR1C3.
Tumors treated with SN35539 grew at a similar rate to the vehicle
treated (control) tumors, while significant tumor control was
observed following PR-104 treatment (Fig. 3d). The median time
to endpoint (four times the original treatment volume, RTV4) for
the control and SN35539 treated groups was 12 days (P = 0.70,
Log-rank test). A significant increase to 44 days was observed in
the PR-104 treated group (P = 0.001 compared to untreated
tumors, Log-rank test). The in vivo data further support the conclu-
sion that AKR1C3 does not play a biologically relevant role in sen-
sitivity to the prodrug SN34507.
To test whether SN34507 was cytotoxic in the presence of
human AKR1C3, the sensitivity of HCT116 WT cells (WT cells)
and HCT116 cells overexpressing AKR1C3 (AKR1C3 cells) to
SN34507 or PR-104A was examined under aerobic conditions using
a low cell density anti-proliferative IC50 assay. The WT IC50 values
for SN34507 and PR-104A were comparable, being 111
6 lM and
70 11 M respectively. AKR1C3 over-expression in this cell line
l
increased the anti-proliferative effect of PR-104A by 87.3-fold,
but had minimal effect on SN34507 (2.9-fold). To confirm elevation
in AKR1C3 catalytic activity was responsible for cellular hypersen-
sitivity to PR-104A, pharmacological inhibition of AKR1C3 was
achieved with the morpholylurea SN34037 [29]. This provided
essentially complete cytoprotection of AKR1C3 cells from
PR-104A exposure, suppressing the observed 87-fold sensitization
effect (WT:AKR1C3 IC50 ratio) to 1.7-fold (Fig. 3a). In contrast, the
effect of AKR1C3 inhibition on SN34507 treated cells was minor
(2.9-fold sensitivity modified to 1.3-fold, Fig. 3a), although this dif-
ference achieved significance (P = 0.023, Student’s t-test) suggest-
ing the possibility of trace activation. The AKR1C3 inhibitor itself
had no demonstrable anti-proliferative effect on WT cells (data
not shown). Other AKR1C family members (AKR1C1, AKR1C2,
and AKR1C4) were unable to sensitize cells to PR-104A or
SN34507 under these conditions (Fig. 3b).
A limitation of low cell density anti-proliferative assays is that a
portion of intracellular cytotoxic drug metabolites may be lost into
the extracellular media through passive diffusion and thereby con-
found the absolute IC50 value. Therefore multi-cellular layer (MCL)
culture models were employed to investigate cytotoxic potency at
tissue-like cell densities, ensuring all local metabolite redistribu-
tion contributes to the overall pharmacodynamic endpoint (loss
3.3. SN34507 is not significantly activated under anoxic conditions
The one-electron reduction potential of SN34507 was deter-
mined by pulse radiolysis. Observations at 600 nm for 3 mixtures
of compound and indicator, gave K = 0.16 0.05 which, on apply-
ing the Nernst equation and allowing for ionic strength effects,
yielded the value E(1) = ꢀ503 10 mV for SN34507. To determine
whether anoxic metabolism had been minimized, HCT116 WT cells
or POR over-expressing HCT116 cells (POR cells) were exposed to
SN34507 under anoxic or aerobic conditions. Anti-proliferative
IC50 assays were used to calculate hypoxia cytotoxicity ratios.
The HCR for PR-104A increased significantly from 18-fold in WT
cells to 126-fold in POR cells (P = 0.048, Student’s t-test), as
expected. In contrast there was minimal anoxic selectivity for
SN34507 in either WT or POR cells (1.2-fold and 2.4-fold respec-
tively, P = 0.12, Student’s t-test), indicating that SN34507 lacks sig-
nificant anoxia-selective anti-proliferative activity and is a poor
single-electron acceptor (Fig. 4). Collectively these data indicate
the low electron affinity of SN34507 eliminates the majority of
reduction by human one-electron reductases such as POR.
of clonogenic viability). After prodrug exposure (10 lM), 100%
WT MCLs demonstrated similar cell plating efficiencies (0.82 and
0.70 for PR-104A and SN34507 respectively, P = 0.12, Student’s
3.4. Use of a bacterial screen to identify nitroreductase enzymes that
activate SN34507
t-test). In 100% AKR1C3 MCLs, treatment with 10 lM PR-104A pro-
duced a >104-fold decrease (4.06 log cell kill) in plating efficiency,
in contrast to SN34507 treatment which produced a 1.1-fold
decrease in plating efficiency (Fig. 3c). Notably, the plating efficien-
cies for 100% WT and 100% AKR1C3 MCLs treated with SN34507
were not significantly different (P = 0.90, Student’s t-test). Collec-
tively, the in vitro data indicate that PR-104A is bioactivated by
AKR1C3 under aerobic conditions while the prodrug analog
SN34507 is essentially free of this activity.
To test whether AKR1C3 metabolism in vivo was biologically
relevant for SN34507 activity we determined the role of AKR1C3
in the anti-tumor efficacy of the phosphate ‘pre-prodrugs’ PR-104
or SN35539 (Fig. 1A). Each pre-prodrug was administered as a sin-
gle intraperitoneal injection at the maximum tolerated dose of
After confirming that SN34507 was not a significant substrate
for the major one- and two-electron reductases POR and AKR1C3,
we next sought to identify a suitable exogenous two-electron (type
I) nitroreductase to bioactivate SN34507. A multi-family bacterial
oxidoreductase gene library [21], including several novel candi-
dates (Table 1) was challenged with SN34507, and the level of
SOS induction for each NTR was measured (Fig. 5).
The seven E. coli NTR strains that induced the greatest SOS
response to SN34507 (i.e., those expressing Salmonella typhi NfsA,
E. coli NfsA, Citrobacter koseri NfsA, Listeria welshmerii NfsA, Enter-
obacter sakazakii NfsA, Vibrio harveyi Frp and Vibrio fischeri FRase)
were taken forward into anti-proliferative bacterial IC50 assays to
confirm that levels of SOS induction were predictive of cytotoxic
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
7
(a)
(b)
50
40
30
20
10
0
Prodrug only
100
80
60
40
20
0
Prodrug + AKR1C3 inhibitor
PR-104A
SN 34507
PR-104A
SN 34507
AKR1C1 AKR1C2 AKR1C3 AKR1C4
Control
(c)
(d)
7
HCT116 WT
HCT116 AKR1C3
101
PR-104
SN35539
6
5
4
3
2
1
0
100
10-1
10-2
10-3
10-4
10-5
0
5
10
15
20
25
30
PR-104A
SN 34507
Time (Days)
Fig. 3. The prodrug SN34507 is not a substrate for human aldo-keto reductase 1C3: (a) Relative sensitivity (interexperimental IC50 ratio at low cell density for P3
independent experiments) of HCT116 WT and its AKR1C3 expressing isogenic cell line to prodrugs PR-104A or SN34507 in the presence or absence of the selective AKR1C3
inhibitor SN34037; (b) Additional AKR1C family members AKR1C1, AKR1C2, and AKR1C4 were unable to sensitize cells to PR-104A or SN34507 under these conditions (mean
of P3 independent experiments); (c) Average plating efficiency from 3 independent experiments of HCT116 WT and its AKR1C3 expressing isogenic cell line following
exposure to prodrugs PR-104A or SN34507 (10 lM) at tissue-like cell density; (d) Relative growth inhibition of HCT116 AKR1C3-expressing tumor model in nude mice
following a single administration of the AKR1C3-activated pre-prodrug PR-104 (1330
lmol/kg, ip) or the AKR1C3-resistant pre-prodrug SN35539 (1330 lmol/kg, ip). N = 6
tumors per treatment group.
potential. Four other members of the library (E. coli NfsB, NemA,
AzoR, MdaB) were included for diversity. The calculated IC50 values
(i.e. the concentration of SN34507 that yielded 50% turbidity rela-
tive to an unchallenged control) are presented in Table 2. NfsA
family members were confirmed as the preferred bacterial NTR
[21]. Initially, anti-proliferative potency of SN34507 was compared
to PR-104A following four hour drug exposure under aerobic con-
ditions (Fig. 6a). Both PR-104A and SN34507 prodrugs were highly
active against the NfsA cell line with anti-proliferative IC50 values
of 0.025 0.004 lM and 0.065 0.015 lM respectively, confirming
enzymes for SN34507, demonstrating IC50 values of <400
(range; 282–395 M).
l
M
SN34507 as a suitable prodrug substrate for E. coli NfsA in vitro.
The four and six electron reduction products of PR-104A (PR-
104H and PR-104M, respectively) are known to alkylate DNA, caus-
ing cross-linking of DNA strands and preventing uncoiling of the
DNA double helix [10,14]. DNA replication is subsequently arrested
at the G2/M DNA damage checkpoint, with cell death occurring
when cross-links are not repaired. To test whether that the anti-
proliferative effect of SN34507 was occurring via a similar mecha-
nism of cell cycle arrest, WT and NfsA cells were exposed to
l
Based on the results of the IC50 screen, three NTRs were selected
for expression and purification as His6-tagged recombinant pro-
teins; S. typhi NfsA, E. coli NfsA and C. koseri NfsA. Apparent
SN34507 kinetic parameters were calculated for each enzyme by
measuring reaction rates at varying substrate concentrations and
a fixed concentration of NADPH (the preferred electron donor for
each enzyme) (Table 3). The kcat/Km was greatest for E. coli NfsA,
identifying this NTR as a promising NTR candidate for SN34507
activation.
equimolar concentrations (0.05 lM) of PR-104A or SN34507 before
recovery and subsequent cell cycle analysis (Fig. 6b). Following
treatment WT cell cycle redistribution was unchanged, with
G2/M population similar to the untreated control cells (21% and
25%, respectively). However with the NfsA cells, the percentage
of cells in G2/M increased from 14% (control) to 26% and 64% for
PR-104A and SN34507 treated cells, respectively. This indicates
that NfsA-mediated activation of SN34507 results in arrest of
3.5. E. coli NfsA is an efficient activator of SN34507 in vitro and in vivo
The possible utility of the E. coli NfsA/SN34507 enzyme/prodrug
combination was investigated in a human cell line model, using
previously generated HCT116 cells over-expressing E. coli NfsA
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

8
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
Anoxic
Oxic
100
80
60
40
20
0
100
80
60
40
20
0
0.01
0.1
1
10
100
1000
0.01
0.1
1
10
100
1000
PR-104A (µM)
PR-104A (µM)
100
100
80
60
40
20
0
80
60
40
20
0
0.01
0.1
1
10
100
1000
0.01
0.1
1
10
100
1000
SN34507 (µM)
SN34507 (µM)
(b)
(a)
Fig. 4. The prodrug SN34507 is a poor substrate for one-electron bioactivation under anoxic conditions: (a) Dose-response curve of HCT116 WT cells exposed to PR-104AA
(upper graph) or SN34507 (lower graph) under aerobic (s) or anoxic (d) conditions; (b) Dose-response curve of HCT116 isogenic POR-expressing cells exposed to PR-104A
(upper graph) or SN34507 (lower graph) under aerobic (s) or anoxic (d) conditions. Data are the average of three independent experiments SD.
5000
NfsA family
4500
NfsB family
4000
3500
3000
2500
2000
1500
1000
500
0
SOS-R2:NTR over-expression strain
Fig. 5. Bacterial SOS response in SOS-R2:NTR strains challenged with SN34507: SOS response levels were measured following 4 h challenge with 20 lM SN34507. SOS
response represents the difference in Miller Units (a measure of b-galactosidase activity) between challenged and unchallenged cultures of the same over-expression strain.
Data are the average of three independent assays SD.
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
9
Table 2
(T/Ac ratio of 464 for SN34507 versus 250 for PR-104A), indicating
a greater therapeutic ratio for SN34507 at physiologically relevant
cell densities. PR-104A was modestly more dose-potent than
SN34507 against NfsA-positive activators in MCL co-cultures (Acti-
Growth inhibition of NTR-overexpressing SOS-R2 after a four hour challenge with
increasing concentrations of SN34507.
Source
Family
IC50 (lM)
vator C10 values of 0.12 and 0.59 lM respectively) (Fig. 7b).
E. coli
NfsA
NfsA
NfsA
NfsA
NfsA
Frp
NfsB
FRase
MdaB
Azo
308 23
282 16
330 13
348 35
351 11
395 35
474 26
480 48
>1000
S. typhi
C. koseri
L. welshmerii
E. sakazakii
V. harveyi
E. coli
V. fischeri
E. coli
E. coli
The increase in sensitivity of the target cells in the presence of
3% activator cells was greater for PR-104A than SN34507, as
demonstrated by the increased left-shift of the target cell survival
curves in the presence of NfsA activator cells (gray filled circles,
Fig. 7). This indicates that the active metabolites of PR-104A have
superior tissue diffusion properties over those of SN34507, and
thus possess an enhanced bystander cell killing ability. The magni-
tude of this phenomenon (the bystander effect efficiency, BEE) was
calculated from the C10 values as previously described. Under these
conditions, PR-104A provided a BEE of 68%, whereas the BEE value
for SN34507 was approximately half this value, 38%, under identi-
cal conditions. Prodrug partition coefficients (LogD7.4) for PR-104A
and SN34507 were calculated to be 1.0 and 0.73 respectively, indi-
cating PR-104A and associated metabolites are more lipophilic
than SN34507. Moreover the reactivity of SN34507 reduced
metabolites will be greater due to the absence of the electron with-
drawing ortho nitro. Both of these aspects are likely to contribute
to the reduced bystander effect observed for SN34507 relative to
PR-104A.
>1000
>1000
>1000
E. coli
–
NemA
Empty
Table 3
Bacterial nitroreductase recombinant enzyme steady state kinetics.
kcat (sꢀ1
)
kcat/Km
a
a
b
NTR
Km
(
l
M)
E. coli NfsA
S. typhi NfsA
C. koseri NfsA
358 44
14 2.1
507 71
3.0 0.15
0.086 0.0036
1.39 0.1
8.4
6.1
2.7
a
Apparent kinetic parameters as measured at 300
(mMꢀ1 sꢀ1).
lM NADPH.
b
We have previously used mixed xenograft models expressing a
minority of NTR positive cells to compare efficacy of prodrugs for
gene therapy [22,35], as this necessitates the presence of a robust
prodrug bystander effect (cytotoxic metabolite redistribution) to
achieve an anti-tumor growth delay and mimics heterogeneous
vector distribution. Using this methodology we compared the anti-
tumor activity of PR-104 and SN35539 in vivo using xenografts
grown from 100% HCT116 WT (Fig. 8a) or from mixtures contain-
ing 30% NfsA cells (Fig. 8b). Tumor-bearing mice were treated with
a single IP dose of prodrug and the tumor size was monitored.
DNA replication and accumulation of cells at the G2/M DNA dam-
age checkpoint, as previously seen for PR-104A.
To model the ability of SN34507 to penetrate tumor tissue,
become metabolically activated by NfsA-expressing cells and
subsequently redistribute from these ‘activator’ cells to sterilize
neighboring NfsA negative ‘target’ (WT) cells, intimate mixtures
of 3% HCT116 NfsA + 97% HCT116 WT cells were grown in 3D
MCL models (Fig. 7a). Under these tissue-like densities SN34507
was more selective for activator cells than was PR-104A, as deter-
mined by the ratio of potencies against NTR negative targets (T)
grown without activators and the activators (Ac) in co-cultures
SN35539 was administered at the MTD of 1330
lmol/kg and
PR-104 was administered at 388 mol/kg, the mouse equivalent
l
dose that provides a plasma AUCinf (39 lM-h) comparable to the
human three-weekly (q3w) MTD of 1100 mg/m2 [19,20].
HCT116 WT
HCT116 E.coli NfsA
100
G1/G0
S
120
G2/M
100
10
1
80
60
40
20
0
0.1
0.01
PR-104A
SN 34507
HCT116 WT
HCT116 NfsA
(a)
(b)
Fig. 6. E. coli NfsA-dependent activation of SN34507 and PR-104A in low density cell line cultures: (a) Relative anti-proliferative activity (IC50 value (
prodrugs PR-104A and SN34507 against the HCT116 WT cell line and its E. coli NfsA-expressing isogenic cell line under aerobic conditions; (b) Cell cycle distribution of
HCT116 WT and NfsA cells following exposure to either PR-104A or SN34507 (0.05 M) under aerobic conditions.
lM), mean sem) of the
l
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

10
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
100
100
10
10
1
1
T
0.1
0.1
0.01
Tc
Ac
0.01
0.001
0.01
0.1
1
10
100
1000
0.001
0.01
0.1
1
10
100
1000
PR-104A concentration (µM)
SN 34507 concentration (µM)
(a)
(b)
Fig. 7. E. coli NfsA-dependent bystander effects of SN34507 and PR-104A in tissue-like multicellular cultures: (a) SN34507 dose-dependent clonogenic killing of NfsA-positive
‘activator’ (Ac) cells (s), WT ‘target’ (T) cells (d) and target cells (Tc) in the presence of 3% activators ( ) under hyperoxic conditions; (b) PR-104A dose-dependent clonogenic
killing of NfsA-positive ‘activator’ (Ac) cells (s), WT ‘target’ (T) cells (d) and target cells (Tc) in the presence of 3% activators ( ) under hyperoxic conditions. Data are the
average of three independent experiments SD.
As expected, tumors lacking NfsA-expressing cells (100% WT
tumors) displayed minimal sensitivity to either prodrug, with a
non-significant increase in median survival for either agent
(P = 0.90, Log-rank test). Mixed tumors composed of 30% NfsA/70%
WT cells ( 4.4% as determined by ex-vivo expression analysis)
were sensitive to both agents. Treatment with PR-104 increased
median survival by 8 days, with a tumor growth delay of 150%
(P < 0.001, Log-rank test). Treatment with SN35539 provided a
325% tumor growth delay (P < 0.001, Log-rank test), with gain in
median survival being significantly greater than PR-104 treated
tumors (8 vs 26 days, P < 0.001, Log-rank test). The body weight
loss associated with each treatment was minimal (<10% of pre-
treatment weight, data not shown).
(E(1)) of the remaining para nitro ‘switch’ to ꢀ503 10 mV for
SN34507. This value indicates net reductive metabolism by one-
electron oxidoreductases under anoxic conditions will be minimal
as single-electron transfer to SN34507 is not thermodynamically
favorable [37]. This was confirmed by the absence of hypoxia-
selective cytotoxicity for SN34507 (Fig. 4a) and the minimal influ-
ence of high level POR expression (Fig. 4b). In contrast, the more
electron affinic substrate PR-104A (E(1) = ꢀ366 mV) is a known
substrate for one-electron oxidoreductases such as cytochrome
P450 oxidoreductase (POR) [11].
Following the generation of an analog of PR-104A resistant to
reduction by key human one- and two-electron reductases, an
optimal bacterial NTR needed to be identified that could activate
SN34507. We utilized
a library of bacterial nitroreductases
expressed in a modified E. coli reporter strain to probe for active
enzymes that would yield DNA damaging species following expo-
sure to SN34507 [21]. Twenty-five enzymes (44%) from the 58
member NTR library catalyzed a DNA damage-mediated ‘SOS
response’ after exposure to SN34507; all were members of the
NfsA or NfsB families. The proportion of active NTRs was lower
than seen historically for PR-104A (72%; 34 of 47 active NTRs),
which was also a prodrug substrate for AzoR and MdaB family
members [21]. Following measurement of the steady state kinetic
parameters, the major NTR from E. coli NfsA was selected as the
preferred bacterial NTR for SN34507 bioactivation. E. coli NfsA
has previously been identified as a prodrug activating enzyme for
the prototypical NTR prodrug CB1954 [33,38,39], and PR-104A
[21]. Expression of E. coli NfsA in HCT116 colon carcinoma cells
confirmed the activation of prodrug SN34507 leading to loss of
proliferative potential mediated through a G2/M cell cycle block,
as reported for PR-104A (Fig. 6). These data are consistent with
the conclusion that the mechanism of action of SN34507 is analo-
gous to PR-104A, with DNA inter-strand crosslink formation giving
rise to replication arrest as a consequence of stalled replication
forks [14].
4. Discussion
An ideal GDEPT prodrug is substantially less cytotoxic than its
corresponding active metabolite(s) and is a unique substrate for
the chosen prodrug activating enzyme under physiological condi-
tions. PR-104 is a clinical stage nitroaromatic prodrug that is
demonstrably more active than the prototypical NTR prodrug
CB1954 in gene therapy models [24]. Nevertheless, several mecha-
nisms of PR-104A activation including hypoxia-selective metabo-
lism [10], aerobic activation by AKR1C3 [15] and oxygen-
insensitive non-CYP450 hepatic enzymes [36], indicate this pro-
drug lacks selectivity for NTRs in the systemic setting. In order to
generate an NTR-specific prodrug for gene therapy applications,
the chemical structure of PR-104A was modified to minimize acti-
vation by endogenous human enzymes. Rational design led to the
synthesis of the prodrug SN34507, an analog of PR-104A lacking
the ancillary ortho nitro moiety predicted to reduce recognition
by the AKR1C3 active site. Additionally, the potential for ortho nitro
reduction leading to intramolecular cyclization of the activate
metabolite to form non-cytotoxic by-products (as seen for
PR-104A [10]) is negated. Loss of AKR1C3-dependent cytotoxicity
was confirmed in 2D anti-proliferative IC50 assays, 3D multi-
cellular layer assays and tumor bearing murine models (Fig. 3).
Suggestion of residual activity in low cell density assays did not
result in biologically meaningful activity in high cell density MCL
and tumor models, with no evidence of AKR1C3 driven antitumor
activity.
In high cell-density (3D) culture models, SN34507 generated a
respectable bystander effect (38% efficiency) and exhibited a
464-fold differential in clonogenic cell kill (C10) between activator
and target cell populations (T/Ac; Fig. 7a). While PR-104A demon-
strated a bystander effect approaching twice that of SN34507
(BEE of 68%) this was associated with poorer selectivity for
NfsA-positive cells (T/Ac C10 ratio of 250). Of note, PR-104A is more
lipophilic than SN34507, a physiochemical property associated
Removal of the electron withdrawing ortho nitro group on the
PR-104A scaffold lowered the one-electron reduction potential
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
11
1.0
1.0
0.8
0.6
0.4
0.2
0.0
0.8
0.6
0.4
0.2
Control
PR-104
SN 35539
0.0
0
10
20
30
40
0
10
20
30
40
Time (Days)
Time (Days)
(a)
(b)
Fig. 8. Relative anti-tumor efficacy of PR-104 and SN35539: (a) 100% HCT116 WT tumor xenografts were treated with a single intraperitoneal dose of vehicle, PR-104 or
SN35539 and tumor volume was monitored trice weekly until volume exceeded 4ꢁ volume on day of prodrug administration. N = 4 tumors per treatment group; (b) HCT116
xenografts harboring 30% NfsA-positive cells were treated with a single intraperitoneal dose of vehicle, PR-104 or SN35539 and tumor volume was monitored thrice weekly
until volume exceeded 4ꢁ volume on day of prodrug administration. N = 8 tumors per treatment group. Log-rank test was employed to compare the survival distributions of
any two groups.
with improved bystander efficiency [32,35]. However, lipophilicity
(LogD7.4) is a fundamental characteristic that influences multiple
properties of pharmaceutical agents including absorption, distribu-
tion, metabolism, excretion and toxicology (ADMET), with less
lipophilic compounds typically possessing more favorable ADMET
properties [40–42]. In addition to the consideration of their relative
partition-coefficients, the major reduced metabolites of PR-104A
are predicted to be less reactive than those of SN34507 (due to
the presence of the additional electron withdrawing nitro group)
and thus are anticipated to diffuse further.
trol tumors (P < 0.001, Log-rank test), but the possibility of activity
attributable to one-electron reductases or induction of AKR1C3
activity cannot be excluded.
This study identifies the suicide gene/prodrug partnership of
E. coli NfsA/SN35539 (SN34507) as a promising combination for
development in virus-directed enzyme/prodrug therapy (VDEPT)
or bacteria-directed enzyme/prodrug therapy (BDEPT). The inclu-
sion of prodrug activating capability in tumor-tropic replicating
vectors can enhance the ability to directly kill cancer cells as
demonstrated with E. coli NfsB/SN28343 for VDEPT [22] and
E. coli NfsB/PR-104 for CDEPT [24]. Of significant interest is the
potential of armed replicating vectors as immunotherapy enhanc-
ing agents. Examples include the recent FDA approval of the HSV-1
derived product talimogene laherparepvec (T-VEC) in unresected
stage IIIB/IV melanoma. Tumor-targeted prodrug mediated cell
death should enhance vector derived danger signals, providing a
more efficient antitumor immune response [45]. There is a need
for highly selective suicide gene/prodrug combinations suitable
for application in this context. Eliminating off-target enzyme activ-
ities associated with endogenous activation of the clinical stage
prodrug analog PR-104 has provided the novel prodrug SN35539,
suitable for combination with E. coli NfsA to achieve much greater
systemic selectivity.
NfsA-dependent SN34507 potency and bystander effect arising
from the diffusion of activated metabolites can only partially
define the therapeutic utility of
a GDEPT prodrug. Overall
anti-tumor activity will also depend on the efficiency of metabolic
activation at tolerated prodrug doses, determined in part by host
toxicokinetics. The original pre-clinical testing of PR-104 was con-
ducted in murine, rat and dog models [10,20], which lack a func-
tional AKR1C3 ortholog capable of activating PR-104A [43]. This
evolutionary divergence underlies the non-predictive nature of
preclinical toxicology models with respect to PR-104 toxicology
in humans (Abbattista, Guise et al., manuscript in preparation).
Consequently, when evaluating efficacy in the ‘mixed’ NfsA/WT
xenograft model PR-104 was administered at the ‘‘human
equivalent dose” (HED) which provides a murine plasma exposure
of PR-104A identical to that achieved in human Phase I clinical tri-
als. While the achievable plasma exposure of SN34507 in humans,
following administration of SN35539, is currently unknown, it is
reasoned that the murine MTD will be representative of the human
MTD, since SN34507 is not subject to extensive activation by
endogenous reductases, the major source of toxicokinetic
discrepancy between species. Consistent with this rationale, the
(non-bioreductive) nitrogen mustard melphalan (L-PAM) achieves
equivalent plasma exposure in murine and human subjects at their
respective maximum tolerated doses [44]. Treatment with
SN35539 significantly increased median survival over both control
and PR-104 treated tumors (P < 0.001, Log-rank test), indicating
adequate prodrug penetration, metabolism and subsequent
metabolite redistribution sufficient to eliminate the majority of
NfsA naïve cells. Single-dose PR-104 displayed modest
NfsA-dependent anti-tumor activity compared to untreated con-
Author contributions
AMM, AA, JUF, CPG, DFA, JBS, and AVP conceived and designed
the experiments; AMM, AA, EMW, JNC, SS, MRB, MRA and RFA
performed the experiments; AMM, JNC, SS, JUF, CPG, JBS and AVP
analyzed the data; JNC, DFA, JUF, SS, RFA, CPG, JBS and AVP
contributed reagents/materials/analysis tools; AMM, AA, JUF,
CPG, JBS and AVP wrote the paper.
Conflicts of interest
The authors declare no conflict of interest. The funders had no
role in the study design, collection, analysis and interpretation of
data; nor in the writing of the article or the decision to submit
the work for publication.
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015

12
A.M. Mowday et al. / Biochemical Pharmacology xxx (2016) xxx–xxx
[21] G.A. Prosser, J.N. Copp, A.M. Mowday, C.P. Guise, S.P. Syddall, E.M. Williams,
Acknowledgments
et al., Creation and screening of a multi-family bacterial oxidoreductase library
to discover novel nitroreductases that efficiently activate the bioreductive
prodrugs CB1954 and PR-104A, Biochem. Pharmacol. 85 (2013) 1091–1103.
[22] D.C. Singleton, D. Li, S.Y. Bai, S.P. Syddall, J.B. Smaill, Y. Shen, et al., The
nitroreductase prodrug SN 28343 enhances the potency of systemically
administered armed oncolytic adenovirus ONYX-411(NTR), Cancer Gene
Ther. 14 (2007) 953–967.
This work was funded by the Health Research Council of New
Zealand (Grant 11/1103 and 14/289) and by scholarships from
the University of Auckland (awarded to AMM) and the Tertiary
Education Commission (a Top Achievers Doctoral award to EMW).
[23] W.R. Wilson, K.O. Hicks, S.M. Pullen, D.M. Ferry, N.A. Helsby, A.V. Patterson,
Bystander effects of bioreductive drugs: potential for exploiting pathological
tumor hypoxia with dinitrobenzamide mustards, Radiat. Res. 167 (2007) 625–
636.
References
[1] G.U. Dachs, M.A. Hunt, S. Syddall, D.C. Singleton, A.V. Patterson, Bystander or
no bystander for gene directed enzyme prodrug therapy, Molecules 14 (2009)
4517–4545.
[2] E.M. Williams, R.F. Little, A.M. Mowday, M.H. Rich, J.V. Chan-Hyams, J.N. Copp,
et al., Nitroreductase gene-directed enzyme prodrug therapy: insights and
advances toward clinical utility, Biochem. J. 471 (2015) 131–153.
[3] I.A. McNeish, N.K. Green, M.G. Gilligan, M.J. Ford, V. Mautner, L.S. Young, et al.,
Virus directed enzyme prodrug therapy for ovarian and pancreatic cancer
using retrovirally delivered E. coli nitroreductase and CB1954, Gene Ther. 5
(1998) 1061–1069.
[4] F. Friedlos, S. Court, M. Ford, W.A. Denny, C. Springer, Gene-directed enzyme
prodrug therapy: quantitative bystander cytotoxicity and DNA damage
induced by CB1954 in cells expressing bacterial nitroreductase, Gene Ther. 5
(1998) 105–112.
[5] J.I. Grove, P.F. Searle, S.J. Weedon, N.K. Green, I.A. McNeish, D.J. Kerr, Virus-
directed enzyme prodrug therapy using CB1954, Anticancer Drug Des. 14
(1999) 461–472.
[24] S.C. Liu, G.O. Ahn, M. Kioi, M.J. Dorie, A.V. Patterson, J.M. Brown, Optimised
clostridium-directed enzyme prodrug therapy improves the antitumor activity
of the novel DNA crosslinking agent PR-104, Cancer Res. 68 (2008) 7995–8003.
[25] G.E. Adams, E.D. Clarke, R.S. Jacobs, I.J. Stratford, R.G. Wallace, P. Wardman,
et al., Mammalian cell toxicity of nitro compounds: dependence upon
reduction potential, Biochem. Biophys. Res. Commun. 72 (1976) 824–829.
[26] S.M. Bailey, R.J. Knox, S.M. Hobbs, T.C. Jenkins, A.B. Mauger, R.G. Melton, et al.,
Investigation of alternative prodrugs for use with E. coli nitroreductase in
‘suicide gene’ approaches to cancer therapy, Gene Ther. 3 (1996) 1143–1150.
[27] S. Yang, G.J. Atwell, W.A. Denny, Synthesis of asymmetric halomesylate
mustards with aziridineethanol/alkali metal halides: application to an
improved synthesis of the hypoxia prodrug PR-104, Tetrahedron 63 (2007)
5470–5476.
[28] G.J. Atwell, W.A. Denny, Synthesis of 3H- and 2H4-labelled versions of the
hypoxia-activated
pre-prodrug
2-[(2-bromoethyl)-2,4-dinitro-6-[[[2-
(phosphonooxy)ethyl]amino]carbonyl]anilino]ethyl methanesulfonate (PR-
104), J. Labelled Comp. Radiopharm. 50 (2007) 7–12.
[6] A.H. Djeha, A. Hulme, M.T. Dexter, A. Mountain, L.S. Young, P.F. Searle, et al.,
Expression of Escherichia coli B nitroreductase in established human tumor
xenografts in mice results in potent antitumoral and bystander effects upon
systemic administration of the prodrug CB1954, Cancer Gene Ther. 7 (2000)
721–731.
[29] J.U. Flanagan, G.J. Atwell, D.M. Heinrich, D.G. Brooke, S. Silva, L.J. Rigoreau,
et al., Morpholylureas are a new class of potent and selective inhibitors of the
type 5 17-b-hydroxysteroid dehydrogenase (AKR1C3), Bioorg. Med. Chem. 22
(2014) 967–977.
[30] M.L. Verdonk, J.C. Cole, M.J. Hartshorn, C.W. Murray, R.D. Taylor, Improved
protein-ligand docking using GOLD, Proteins 52 (2003) 609–623.
[31] N.A. Helsby, G.J. Atwell, S. Yang, B.D. Palmer, R.F. Anderson, S.M. Pullen, et al.,
Aziridinyldinitrobenzamides: synthesis and structure-activity relationships
for activation by E. coli nitroreductase, J. Med. Chem. 47 (2004) 3295–3307.
[32] W.R. Wilson, S.M. Pullen, A. Hogg, N.A. Helsby, K.O. Hicks, W.A. Denny,
Quantitation of bystander effects in nitroreductase suicide gene therapy using
three-dimensional cell cultures, Cancer Res. 62 (2002) 1425–1432.
[33] G.A. Prosser, J.N. Copp, S.P. Syddall, E.M. William, J.B. Smaill, W.R. Wilson, et al.,
Discovery and evaluation of Escherichia coli nitroreductases that activate the
anticancer prodrug CB 1954, Biochem. Pharmacol. 79 (2010) 678–687.
[34] A.L. Lovering, J.P. Ride, C.M. Bunce, J.C. Desmond, S.M. Cummings, S.A. White,
Crystal structures of prostaglandin D(2) 11-ketoreductase (AKR1C3) in
complex with the nonsteroidal anti-inflammatory drugs flufenamic acid and
indomethacin, Cancer Res. 64 (2004) 1802–1810.
[35] G.J. Atwell, S. Yang, F.B. Pruijn, S.M. Pullen, A. Hogg, A.V. Patterson, et al.,
Synthesis and structure-activity relationships for 2,4-dinitrobenzamide-5
mustards as prodrugs for the Escherichia coli nfsB nitroreductase in gene
therapy, J. Med. Chem. 50 (2007) 1197–1212.
[36] Y. Gu, C.P. Guise, K. Patel, M.R. Abbattista, J. Li, X. Sun, et al., Reductive
metabolism of the dinitrobenzamide mustard anticancer prodrug PR-104 in
mice, Cancer Chemother. Pharmacol. 67 (2011) 543–555.
[37] P. Wardman, M.F. Dennis, S.A. Everett, K.B. Patel, M.R. Stratford, M. Tracy,
Radicals from one-electron reduction of nitro compounds, aromatic N-oxides
and quinones: the kinetic basis for hypoxia-selective, bioreductive drugs,
Biochem. Soc. Symp. 61 (1995) 171–194.
[38] S.O. Vass, D. Jarrom, W.R. Wilson, E.I. Hyde, P.F.E. Searle, Coli NfsA: an
alternative nitroreductase for prodrug activation gene therapy in combination
with CB1954, Br. J. Cancer 100 (2009) 1903–1911.
[39] Y. Barak, S.H. Thorne, D.F. Ackerley, S.V. Lynch, C.H. Contag, A. Matin, New
enzyme for reductive cancer chemotherapy, YieF, and its improvement by
directed evolution, Mol. Cancer Ther. 5 (2006) 97–103.
[40] P.D. Leeson, B. Springthorpe, The influence of drug-like concepts on decision-
making in medicinal chemistry, Nat. Rev. Drug Discov. 6 (2007) 881–890.
[41] J.D. Hughes, J. Blagg, D.A. Price, S. Bailey, G.A. Decrescenzo, R.V. Devraj, et al.,
Physiochemical drug properties associated with in vivo toxicological
outcomes, Bioorg. Med. Chem. Lett. 18 (2008) 4872–4875.
[7] G. Chung-Faye, D. Palmer, D. Anderson, J. Clark, M. Downes, J. Baddeley, et al.,
Virus-directed, enzyme prodrug therapy with nitroimidazole reductase:
a
phase I and pharmacokinetic study of its prodrug, CB1954, Clin. Cancer Res. 7
(2001) 2662–2668.
[8] G.M. Anlezark, R.G. Melton, R.F. Sherwood, B. Coles, F. Friedlos, R.J. Knox, The
bioactivation of 5-(aziridin-1-yl)-2,4-dinitrobenzamide (CB1954)
–
I.
Purification and properties of a nitroreductase enzyme from Escherichia coli
– a potential enzyme for antibody-directed enzyme prodrug therapy (ADEPT),
Biochem. Pharmacol. 44 (1992) 2289–2295.
[9] M.H. Tang, N.A. Helsby, W.R. Wilson, M.D. Tingle, Aerobic 2- and 4-
nitroreduction of CB 1954 by human liver, Toxicology 216 (2005) 129–139.
[10] A.V. Patterson, D.M. Ferry, S.J. Edmunds, Y. Gu, R.S. Singleton, K. Patel, et al.,
Mechanism of action and preclinical antitumor activity of the novel hypoxia-
activated DNA crosslinking agent PR-104, Clin. Cancer Res. 13 (2007) 3922–
3932.
[11] C.P. Guise, A. Wang, A. Thiel, D. Bridewell, W.R. Wilson, A.V. Patterson,
Identification of human reductases that activate the dinitrobenzamide
mustard prodrug PR-104A:
a
role for NADPH:cytochrome P450
oxidoreductase under hypoxia, Biochem. Pharmacol. 74 (2007) 810–820.
[12] C.P. Guise, M.R. Abbattista, S.R. Tipparaju, N.K. Lambie, J. Su, D. Li, et al.,
Diflavin oxidoreductases activate the bioreductive prodrug PR-104A under
hypoxia, Mol. Pharmacol. 81 (2012) 31–40.
[13] Y. Gu, A.V. Patterson, G.J. Atwell, S.B. Chernikova, J.M. Brown, L.H. Thompson,
et al., Roles of DNA repair and reductase activity in the cytotoxicity of the
hypoxia-activated dinitrobenzamide mustard PR-104A, Mol. Cancer Ther. 8
(2009) 1714–1723.
[14] R.S. Singleton, C.P. Guise, D.M. Ferry, S.M. Pullen, M.J. Dorie, J.M. Brown, et al.,
DNA crosslinks in human tumor cells exposed to the prodrug PR-104A:
relationships to hypoxia, bioreductive metabolism and cytotoxicity, Cancer
Res. 69 (2009) 3884–3891.
[15] C.P. Guise, M. Abbattista, R.S. Singleton, S.D. Holford, J. Connolly, G.U. Dachs,
et al., The bioreductive prodrug PR-104A is activated under aerobic conditions
by human aldo-keto reductase 1C3, Cancer Res. 70 (2010) 1573–1584.
[16] Y. Jin, T.M. Penning, Aldo-keto reductases and bioactivation/detoxication,
Annu. Rev. Pharmacol. Toxicol. 47 (2007) 263–292.
[17] J. Birtwistle, R.E. Hayden, F.L. Khanim, R.M. Green, C. Pearce, N.J. Davies, et al.,
The aldo-keto reductase AKR1C3 contributes to 7,12-dimethylbenz(a)
anthracene-3,4-dihydrodiol mediated oxidative DNA damage in myeloid
cells: implications for leukemogenesis, Mutat. Res. 662 (2009) 67–74.
[18] M.B. Jameson, D. Rischin, M. Pegram, J. Gutheil, A.V. Patterson, W.A. Denny,
et al., A phase I trial of PR-104, a nitrogen mustard prodrug activated by both
hypoxia and aldo-keto reductase 1C3, in patients with solid tumors, Cancer
Chemother. Pharmacol. 65 (2010) 791–801.
[19] M.J. McKeage, Y. Gu, W.R. Wilson, A. Hill, K. Amies, T.J. Melink, et al., A phase I
trial of PR-104, a pre-prodrug of the bioreductive prodrug PR-104A, given
weekly to solid tumour patients, BMC Cancer 11 (2011) 432.
[20] K. Patel, S.F. Choy, K.O. Hicks, T.J. Melink, N.H.G. Holford, W.R. Wilson, A
combined pharmacokinetic model for the hypoxia-targeted prodrug PR-104A
in humans, dogs, rats and mice predicts species differences in clearance and
toxicity, Cancer Chemother. Pharmacol. 67 (2011) 1145–1155.
[42] J.A. Arnott, S.L. Planey, The influence of lipophilicity in drug discovery and
design, Expert Opin. Drug Discov. 7 (2012) 863–875.
[43] P. Velica, N.J. Davies, P.P. Rocha, H. Schrewe, J.P. Ride, C.M. Bunce, Lack of
functional and expression homology between human and mouse aldo-keto
reductase 1C enzymes: implications for modelling human cancers, Mol. Cancer
8 (2009) 121–132.
[44] J.K. Peterson, P.J. Houghton, Integrating pharmacology and in vivo cancer
models in preclinical and clinical drug development, Eur. J. Cancer 40 (2004)
837–844.
[45] H.L. Kaufman, F.J. Kohlhapp, A. Zloza, Oncolytic viruses:
a new class of
immunotherapy drugs, Nat. Rev. Drug Discov. 14 (2015) 642–662.
Please cite this article in press as: A.M. Mowday et al., Rational design of an AKR1C3-resistant analog of PR-104 for enzyme-prodrug therapy, Biochem.
Pharmacol. (2016), http://dx.doi.org/10.1016/j.bcp.2016.07.015