Investigation into drug release from colon-specific azoreductase-activated steroid prodrugs using in-vitro models

Article abstract of DOI:10.1111/j.2042-7158.2011.01289.x

Objectives The aim of this study was to investigate drug release from a double steroid prodrug, OPN501, which incorporates a phenylpropionate linker, and its phenylacetate analogue. The prodrugs, which were designed to deliver prednisolone to the colon fo

Full text of DOI:10.1111/j.2042-7158.2011.01289.x

Research Paper
JPP 2011, 63: 806–816
© 2011 The Authors
JPP © 2011 Royal
Pharmaceutical Society
Received August 23, 2010
Accepted March 28, 2011
DOI
Investigation into drug release from colon-specific
azoreductase-activated steroid prodrugs using
in-vitro modelsjphp_1289 806..816
10.1111/j.2042-7158.2011.01289.x
ISSN 0022-3573
Juan F. Marquez Ruiza, Kinga Kedzioraa, Henry Windlea,
Dermot P. Kelleherb and John F. Gilmera
aSchool of Pharmacy and Pharmaceutical Sciences and bDepartment of Clinical Medicine, Trinity College,
Dublin, Ireland
Abstract
Objectives The aim of this study was to investigate drug release from a double steroid
prodrug, OPN501, which incorporates a phenylpropionate linker, and its phenylacetate
analogue. The prodrugs, which were designed to deliver prednisolone to the colon for the
treatment of inflammatory bowel disease, are based on a novel design that requires sequen-
tial azoreductase activity and cyclization of an amino ester to trigger drug release. We sought
to explain the divergent effects of the two compounds in anti-inflammatory models and to
justify the selection of OPN-501 for clinical development.
Methods The compounds were incubated in mouse colonic contents (10%) fermented in
brain heart infusion under anaerobic conditions. The disappearance of the prodrugs and
release of prednisolone was monitored by HPLC. We then developed a method for assess-
ment of prodrug activation using suspensions of Clostridium perfringens, an anaerobe from
the human colon. The cyclization of the compounds was studied in various media, assess-
ing the influence of pH and bulk solvent polarity on cyclization rate using HPLC and
NMR.
Key findings The prodrugs were activated via multiple pathways releasing prednisolone
in mouse colonic ferment. The compounds released prednisolone by reduction–cyclization
in C perfringens suspension. The active OPN-501 generated a stoichiometric amount of
prednisolone following azoreductase activation, whereas its analogue did not. The pH rate
profile for the cyclization of the amino intermediates of the two compounds revealed
significant differences in rate at pH values relevant to the inflamed colon, which explain in
part the different amounts of drug produced.
Conclusions The steroid prodrug OPN-501 has optimal drug release characteristics for
colon targeting because of a kinetic advantage of a six-membered ring formation in the
aminolysis reactions of anilides. The results are relevant to the development of OPN-501 but
also to cyclization strategies in prodrug design especially for colon targeting.
Keywords azoreductase; colon; cyclization; prodrug; steroid
Introduction
Glucocorticoids are the most important therapies for controlling acute ulcerative colitis and
Crohn’s disease.[1] Their long-term use is unfortunately limited by their tendency to cause
unpleasant and serious side effects, which are well known.[2] Site-specific delivery of these
agents to the lower gut has the potential to reduce systemic exposure and side effects,
decrease dose and improve efficacy. There has accordingly been a lot of work investigating
formulation and prodrug approaches to achieving what might be considered a topical effect
from the oral route.[3–5] Despite some interesting developments there remains a strong
clinical appetite for a steroid for inflammatory bowel disease (IBD) treatment with an
improved safety profile.
The prodrug approach to colon targeting is plausible because of the success of
5-aminosalicylic acid prodrugs such as olsalazine, ipsalazide and basalazide.[6–10] These
transit to the colon and are activated by azoreductase activity associated with the luxuriant
colonic microflora. The clinical use of these agents validates azoreductase as a practical
vector in IBD patients and it underlines its ideal distribution in the human gastrointestinal
Correspondence: John F. Gilmer,
School of Pharmacy and
Pharmaceutical Sciences, Trinity
College, Dublin 2, Ireland.
E-mail: gilmerjf@tcd.ie
806

Azoreductase activated steroid prodrugs
Juan F. Marquez Ruiz et al.
807
O
O
O
O
O
HO
O
OH
HO
HO
OH
N
N
N
N
O
O
HO
1
OPN501, 2
O
O
OH
OH
Figure 1 Structural formulae for phenylacetate 1 and phenylpropionate 2 (OPN501).
tract for colon targeting. Colonic azoreductase activity has
Materials and Methods
therefore been investigated to target non-salicylate anilides to
the colon.[11] We recently reported a novel cyclization-
dependent prodrug system (Figure 1) that extends the possi-
bilities for azoreductase-mediated drug release to alcohol-
bearing drugs, including steroids.[12] The mechanism of drug
release is illustrated in Figure 2. Reduction of the azo carrier
by enzymes associated with colonic bacteria produces an
amino ester (e.g. 6), which is poised to undergo intramolecu-
lar aminolysis causing drug release. Cyclization strategies in
prodrug design are popular because of the robust chemical
predictability of the event.[13] In the prodrug design the
intramolecular arrangement promotes anilide aminolysis,
which is a very slow bimolecular reaction. There are in addi-
tion a number of competing possibilities for ester hydrolysis
on the intact prodrug or its amino ester reduction product, and
these are also shown. Overall, the design exploits the facile
cleavage of azo bonds in the colon and the predictable ami-
nolysis of the amino ester to achieve site-specific drug release.
Compound 2 is in nonclinical development. It is not absorbed
from the gastrointestinal tract in animal models, consistent
with its transport characteristics in epithelial models.[12] It
exhibits similar anti-inflammatory efficacy to prednisolone
but causes significantly lower suppression of the
hypothalamic–pituitary–adrenal (HPA) axis.[12] Compound 1
is an analogue of 2 that incorporates a shorter phenylpropi-
onate linker. It was inactive in an in-vivo anti-inflammatory
model indicating that it did not release prednisolone.[12]
The objective of this study was to find out why 2 is phar-
macologically active but 1 is not. Therefore we studied the
rates of drug release and disappearance kinetics of 1 and 2
under conditions designed to simulate colonic conditions. We
developed and validated a generally useful assay that employs
a suspension of Clostridium perfringens from the human
colon in order to study the activation and drug release from 1
and 2. We had to do this because there were no methods
available that would have allowed us to measure the intact
drugs and their metabolites. The relative rates of aminolysis of
the intermediates 3 and 6, the anilide products of azoreductase
activity, were then studied. We propose that the relative effects
of 1 and 2 arise because of differences in rates of cyclization
under conditions that justify the selection of 2 for clinical
development.
All chemicals were obtained from the Sigma-Aldrich Chemi-
cal Company (Dublin, Ireland). Brain heart infusion (BHI)
and BHI-agar was obtained from Sigma Aldrich (Dublin,
Ireland). C. perfringens and Bacteroides fragilis were
obtained from the Health Protection Agency Culture Collec-
tions (Centre for Emergency Preparedness and Response,
Porton Down, Salisbury, UK) and Escherichia coli was from
New England Biolabs (Ipswich, USA). Samples of fresh
mouse colonic contents were obtained from OPSONA, Thera-
peutics Ltd (St James’s Hospital, Dublin, Ireland).
General methods
Uncorrected melting points were obtained using a Stuart
melting point SMP11 melting point apparatus. Spectra were
obtained using a Perkin Elmer 205 FT Infrared Paragon 1000
spectrometer. Band positions are given in cm-1. Solid samples
were obtained by KBr disk; oils were analysed as neat films
on NaCl plates. 1H and 13C spectra were recorded at 27°C on
a Bruker Advance II 600 mHz spectrometer (600.13 mHz 1H,
150.91 mHz 13C) and Bruker DPX 400 mHz FT NMR spec-
1
trometer (400.13 mHz H, 100.16 mHz 13C), in either CDCl3
or CD3OD (tetramethylsilane as internal standard). For
CDCl3, 1H NMR spectra were assigned relative to the tetram-
ethylsilane (TMS) peak at 0.00 d and 13C NMR spectra were
assigned relative to the middle CDCl3 triplet at 77.00 ppm.
For CD3OD, 1H and 13C NMR spectra were assigned relative
to the centre peaks of the CD3OD multiplets at 3.30 d and
49.00 ppm, respectively. Coupling constants were reported in
1
hertz (Hz). For H NMR assignments, chemical shifts are
reported: shift values (number of protons, description of
absorption (s = singlet, d = doublet, t = triplet, q = quartet,
m = multiplet), coupling constant(s) where applicable, proton
assignment). High-resolution mass spectrometry (HRMS)
was performed on a Micromass mass spectrophotometer (EI
mode) at the Department of Chemistry, Trinity College. Thin-
layer chromatography (TLC), was performed on silica gel
Merck F-254 plates. Compounds were visually detected by
absorbance at 254 nm and/or vanillin staining. The HPLC
system used in this study consisted of a 600 controller pump
(Waters), 2996 photodiode array detector (Waters), 717 plus
autosampler (Waters) and an on-line degasser. The Empower

808
Journal of Pharmacy and Pharmacology 2011; 63: 806–816
O
O
OH
OH
OH
HO
(CH2)n
HO
OH
N
O
N
8 n=1
9 n=2
4
O
Hydrolysis
azocarrier
prednisolone
O
O
OH
O
(CH2)n
HO
OH
N
O
N
OH
1 n=1
2 n=2
O
Azoreductase
activity
O
NH2
ODRUG
NH2
n(CH2)
(CH2)n
O
5-ASA
3 n=1
6 n=2
HO
N
H
+
OH
O
5 n=1
7 n=2
Intramolecular
lactamization
Hydrolysis
O
Intramolecular
lactamization
OH
n(CH2)
(CH2)n
NH2
+
PRED
+
O
N
H
PRED
5 n=1
7 n=2
Figure 2 Competition between intramolecular lactamization and hydrolysis in 3 and 6.
software was used for system control, data acquisition and formate buffer (0.5% NH4OH, pH 10) over 20 min was used.
processing. The HPLC separation was achieved using X The assay was performed at room temperature and flow rate of
Bridge C18 column (6 ¥ 250 mm, 5 mm particle size). Mobile 1.5 ml/min. Detector l = 245 nm. The method was validated
phase with gradient elution 10–90% acetonitrile in ammonia according to the principles of ICHQ2R.

Azoreductase activated steroid prodrugs
Juan F. Marquez Ruiz et al.
809
80 ml of trifluoroacetic acid (TFA) into 30 ml of the stock
solution; the reaction was left for 6 min before evaporation
of the solvents with nitrogen. After that MeOH (50 ml) was
added and 950 ml of the buffer solution at different pH values at
37°C and monitored by HPLC. The buffers were prepared
using a universal borate/citrate/phosphate buffer system.
To study the effect of the ionic strength on the lactamiza-
tion process of 3 and 6, solutions were prepared with ionic
strength values in the range 0.06–0.6, at constant pH (7.4).
Borate buffer pH 7.4 was prepared by mixing 2.5 ml solu-
tion A (containing boric acid, citric acid, sodium chloride,
I = 0.06 m) and 4.5 ml solution B (containing potassium phos-
phate I = 0.06 m) and ionic strength of 0.06 m. Borate buffers
at ionic strength 0.12, 0.24 and 0.6 m were prepared by addi-
tion of NaCl.
(2-tert-Butoxycarbonylamino-phenyl)-acetic acid
2-(11,17-dihydroxy-10,13-dimethyl-3-oxo-6,7,8,9,
10,11,12,13,14,15,16,17-dodecahydro-3H-
cyclopenta[a]phenanthren-17-yl)-2-oxo-ethyl
ester (BOC-3)
To a solution of 3-(2-t-butoxycarbonylmethyl-phenyl)-acetic
acid (13) (0.02 g, 0.79 mmol) in DCM (5 ml), DMAP (1
equivalent, 0.009 g, 0.79 mmol) and DCC (1 equivalent,
0.016 g, 0.79 mmol) were added followed by prednisolone
(1.1 equivalent, 0.03 g, 0.86 mmol). After 3 h the reaction was
complete. DCU was removed by filtration. The solvent was
removed under reduced pressure to afford a yellowish oil.
This was flash columned using hexane : ethyl acetate (2 : 1)
to yield the product as off-white crystals (0.03 g, 64%), m.p.
142–144°C. IRVmax(KBr): 3327.27, 1722.56, 1626.17 and
1
1573.84 cm-1. H NMR d (CDCl3): 8.56 (1H, s, H’6), 7.32
Solution A: 0.005 m of citric acid monohydrate, 0.02 m of
boric acid and 0.059 m of NaCl in distilled water.
Solution B: 0.01 m of tri-potassium ortho-phosphate in dis-
tilled water. The solutions were mixed in the necessary pro-
portion to achieve the required pH.
(2H, m, H’4+H2), 7.23 (1H, m, H5’), 7.11 (1H, m, H3’), 6.17
(1H, d, J 10.04 Hz, H1), 5.91 (1H, s, H4), 5.60 (2H, d, J
7.56 Hz, C21CH2), 4.26 (1H, s), 3.80 (2H, s), 1.45 (9H, s,
BOC), 2.30–0.78 (prednisolone envelope, 19H). 13C NMR d
(CDCl3): 205.25, 185.16, 170.54, 170.40, 156.72 156.60,
153.51, 136.84, 130.78, 128.64, 127.34, 127.02, 124.71,
121.55, 88.63, 78.83, 68.26, 67.91, 55.34, 51.03, 47.49,
47.09, 43.75, 33.35, 32.77, 32.06, 31.42, 28.10, 25.53, 24.46,
23.57, 16.49. HRMS: Found: (M-Na)+ = 616.2894. Required:
(M-Na)+ = 616.2886.
It is possible to make a lineal correlation to make the different
solutions at different pH values using the formula below,
substitution of the pH value in the equation gave the volume
of the solution required to obtain the desired pH value.
Volume of solution A (ml) = 6.68 − 0.56 × pH
Volume of solution B (ml) = −2.00 + 0.8670 × pH
3-(2-tert-Butoxycarbonylamino-phenyl)-propionic
acid 2-(11,17-dihydroxy-10,13-dimethyl-3oxo-
6,7,8,9,10,11,12,13,14,15,16,17-dodecahydro-3H-
cyclopenta[a]phenanthren-17-yl)-2-oxo-ethyl
ester (BOC-6)
To find the buffer salt independent relationship between pH
and cyclization rate we needed to generate a range of buffers
at constant ionic strength.
In this case the buffer solutions were prepared as indicated
below:
To
a solution of 3-(2-t-butoxycarbonylmethyl-phenyl)-
propionic acid (14) (0.15 g, 0.55 mmol) in DCM (6 ml),
DMAP (1 equivalent, 0.067 g, 0.55 mmol) and DCC (1
equivalent, 0.113 g, 0.55 mmol) were added followed by
prednisolone (1 equivalent, 0.2 g, 0.55 mmol). The reaction
was left stirring overnight and TLC (EtOAc) showed comple-
tion. DCU was removed by filtration and solvent was evapo-
rated under reduced pressure to afford an oil. The crude was
flash columned using DCM : ethyl acetate (7 : 3) to yield the
product as white crystals (0.31 g, 90%), m.p. 152–154°C.
Solution 1: 0.05 m citric acid monohydrate, 0.2 m boric acid
and 0.59 m NaCl in distilled water.
Solution 2: 0.1 m tri-potassium ortho-phosphate in distilled
water.
Solution 1a: 0.025 m citric acid monohydrate, 0.1 m boric acid
and 0.59 m NaCl in distilled water.
Solution 2a: 0.05 m tri-potassium ortho-phosphate and 0.3 m
NaCl in distilled water.
Solution 1b: 0.02 m citric acid monohydrate, 0.04 m boric acid
and 0.59 m NaCl in distilled water.
Solution 2b: 0.02 m tri-potassium ortho-phosphate and 0.48 m
NaCl in distilled water.
Solution 1c: 0.005 m citric acid monohydrate, 0.02 m boric
acid and 0.59 m NaCl in distilled water.
Solution 2c: 0.01 m tri-potassium ortho-phosphate and 0.54 m
NaCl in distilled water.
IRvmax (KBr): 3443.05, 1722.69, 1657.72 cm-1. H NMR d
1
(CDCl3): 7.6 (1H, m, H-6’), 7.33 (1H, d, J 10.04 Hz, H2), 7.22
(1H, t, J 7.27 Hz, H-4), 7.16 (1H, dd, J 9.04 & 1.51 Hz, H’5),
7.09 (1H, t, J 7.53 & 1.26 Hz, H’3), 6.3 (1H, dd, J 10.29 &
1.8 Hz, H4), 6.02 (1H,s), 4.93 (2H, q, J 17.13 Hz, 21-CH2),
4.49 (1H, s), 4.15 (1H, q, J 7.27 Hz), 1.55 (9H, s, BOC), 1.30
(1H, t, J 7.28), 1.0 (3H, s), 2.93–0.98 (prednisolone envelope
21H). 13C NMR d (CDCl3): 204.85, 186.78, 171.26, 170.54,
156.74, 153.92, 135.92, 131.46, 129.59, 127.73, 127.16,
124.57, 123.60, 122.31, 89.89, 80.68, 69.99, 68.67, 55.37,
51.39, 48.05, 39.41, 34.67, 34.54, 34.12, 32.06, 31.33,
28.37, 25.33, 23.86, 21.09, 16.72. HRMS: Found:
(M-Na)+ = 630.3062. Required: (M-Na)+ = 630.3043.
The solutions were mixed in the necessary proportion to
achieve the required pH using the above formula. In this way
we prepared four solutions at constant ionic strength at each
pH. The disappearance kinetics were monitored at each dilu-
tion, the rate estimated from the exponential decay and then
the resulting kobs values plotted against buffer concentration.
Lactamization studies
Stock solutions of BOC-3 and BOC-6 were prepared at 10 mm Plots of kobs versus buffer concentration produced straight
in MeOH. The lactamization samples were prepared by adding lines which were used to estimate kobs at zero buffer.

810
Journal of Pharmacy and Pharmacology 2011; 63: 806–816
80% N) and the cells were kept in suspension by incubating
the gas jars in a shaking incubator (120 rpm, at 37°C). The
negative controls consisted of BHI broth containing the
appropriate test compounds at the same concentration but
without the addition of C. perfringens. The assay was also
performed in the presence of the azoreductase inhibitors
menadione (100 mm–1 mm) and iodosobenzoic acid (1 mm).
Culture of mouse colon contents
The azoreductase activity of mouse colon contents towards 1
and 2 was tested under anaerobic conditions at 37°C. Bacteria
and faeces were scraped from the regions of the colon of
BALB/C mice until 1.5 cm from the anus, weighed and inocu-
lated in BHI media to obtain a suspension 10% in weight of
colon contents. Compounds 1 and 2 stock solutions were
added (20 ml) into fermented suspension (1 980 ml) in a sterile
20 ml universal tube to reach 50 mm concentration and 1%
dimethyl sulfoxide (DMSO). The negative control consisted
of BHI media with 1 and 2 at the same concentration, also
introduced in DMSO. Immediately after the prodrugs were
added to the colon contents a 200 ml sample was taken and
added to 400 ml of acetonitrile in a 1.5 ml eppendorf; samples
were then centrifuged at 10 000 rpm for 10 min and superna-
tant transferred to a tube and stored at -20°C until analysis.
Samples were taken at time points: 0, 1.5, 3, 5 and 22 h, while
carefully maintaining the ferment under anaerobic conditions.
Samples were analysed by HPLC for remaining prodrug and
products. The cyclization reaction was also performed at 4°C
to validate the extraction and sample storage conditions. The
t1/2 for prodrug disappearance under these conditions was
>1000 min in both cases.
Data handling and statistical analyses
All data manipulations were performed using GraphPad
Prism. The experiments using mouse colon contents were
performed three times. The data points represent the mean
concentration of each species and error bars the standard
deviation of these determinations. The rate constants for
cyclization at zero buffer were estimated as described; the
error associated with these values corresponds to the SD of the
y-intercept.
Results and Discussion
Activation and hydrolysis in mouse
colon contents
To mimic colonic conditions in the murine inflammation
model, which had been previously used to assess the anti-
inflammatory activity of the two prodrugs, mouse colon
Clostridium perfringens screening assay
A 20-ml volume of test substance (2 mm in DMSO) was added contents were allowed to incubate in vitro for 24 h in BHI
to BHI broth (1 980 ml) inoculated with C. perfringens in a (37°C) under anaerobic conditions before the introduction of
sterile 20 ml universal container resulting in substrate concen- 1 or 2. Samples were taken periodically, quenched and analy-
trations of 15–50 mm (1% DMSO). Immediately after addition sed by a validated HPLC method, which permitted the sepa-
of the test compounds, a sample was taken (200 ml) and added ration and quantitation of components identified in Figure 2.
to 400 ml of acetonitrile in a 1.5 ml minifuge tube. Samples Compounds 1 and 2 underwent processing in the colonic
were then centrifuged at 10 000 rpm for 10 min. The resulting suspension with the appearance of prednisolone (4) (progress
supernatants were recovered and stored at -20°C until curves appear in Figure 3). The prodrug disappearance (and
required for analysis. Samples were taken every 2 h up to 8 h prednisolone evolution) could be attributed to a mixture of
and a final sample taken at 24 h. At all times the bacterial azo-reductase activity leading to the formation of the cycliza-
suspensions were maintained under strictly anaerobic condi- tion products (5 and 7) and ester hydrolysis producing azo
tions using a Mart Anoxomat system (Mart Microbiology acids 8 and 9 (Figure 2), which were themselves consumed
B.V., Drachten, Netherlands) (gas mix: 10% CO2, 10% H, through azo reduction. The amino products of azoreductase
(a)
(b)
110
100
90
80
70
60
50
40
30
20
10
0
110
100
90
80
70
60
50
40
30
20
10
0
0
2
4
6
8
10 12 14 16 18 20 22
Time (h)
0
2
4
6
8
10 12 14 16 18 20 22
Time (h)
Figure 3 Progress curves for the disappearance of 1 and 2 following incubation in cultured mouse colon contents (MCC) (10%) in pH 7.4 buffer
under anaerobic conditions at 37°C: A. 1 incubated in MCC ( ), 1 in BHI (᭹), prednisolone (᭝), 5 (᭞), 8 (᭛); B. 2 ( ), 2 in BHI (᭹), prednisolone
(᭝) B. 2 incubated in MCC ( ), 2 in BHI (᭹), prednisolone (4) (᭝), 7 (᭞), 9 (᭛).

Azoreductase activated steroid prodrugs
Juan F. Marquez Ruiz et al.
811
activity on 1 and 2 (3 and 6) were not detected. Ester centration; and (iv) to determine the optimum cell count as
hydrolysis was apparently more prominent in the case of 1 reflected in OD for substrate turnover.
(60% at max in molar terms) than 2 (~20% at max). This
Validation of the extraction method
may reflect greater susceptibility towards hydrolysis in
the case of 1 or slower competing processes such as azo- A series of experiments were performed using different con-
reduction. Ester hydrolysis appeared to be enzymatically centrations of 1, 2 in BHI to validate an organic solvent
mediated because 1 and 2 were stable under similar condi- extraction procedure. BHI solutions (1% DMSO) were pre-
tions of pH and temperature in BHI. The esterase activity pared at three concentration levels (3.3, 16.5 and 50 mm) for
may have been associated with the bacterial component of each of the study compounds: olsalazine, prednisolone,
the preparation or with luminal or tissue-derived esterases steroid prodrugs, cyclization products 2-oxindole (5) and
from the intestinal wall artifactually introduced during DHQ (7). These solutions were each repeatedly extracted
sampling. The intestine of rodents has significantly higher (n = 3) with MeCN and analysed by a previously reported
esterase activity than the human intestine.[14,15] Notably, the high pH HPLC method, which had been fully validated for
disappearance of 2 was associated with significantly greater water/MeCN solutions according to the principles of
prednisolone evolution and the expected cyclization product ICHQ2R.[12] The recovery (accuracy) and repeatability (pre-
7 than from 1 and its cyclization product 5. The latter obser- cision) values for the extractions from BHI at the 3 mm level
vation suggested that differences in prednisolone release were, respectively, 90–110% and <10% (RSD), which was
were caused by differences in rates of reduction or lactam- acceptable for the purposes of this kind of study.
ization, although in the case of 1 there was presumably less
Azoreductase assay with C. perfringens
intact prodrug available to undergo reduction/cyclization
because of its greater susceptibility to hydrolysis. It was Routinely, 24-h-old BHI-agar grown cultures of C. perfrin-
decided to make a closer examination of the relative rates of gens were harvested using sterile plastic loops and the cells
cyclization of the intermediates 3 and 6 arising from azo suspended in BHI-broth (1.8 ml) at defined densities as
reduction in each case.
reflected in OD at 600 nm. The relationship between bacterial
cell density and the rate of substrate consumption was
measured over the OD600 range 0.1–1.1 and at substrate con-
centrations in the range 20–500 mm. These variables were
optimized using steroid prodrugs 1, 2 and Direct Blue 15 as
Development of an azoreductase assay using
bacterial suspension
To identify bacteria with appropriate levels of azoreductase standards. Optimal kinetics were observed at bacterial densi-
activity, BHI-agar plates containing either the dye Direct Blue ties in the OD600 range 0.9–1.1 and at substrate concentrations
15 (300 mm) or the prodrug olsalazine (250 mm) were inocu- of 20–50 mm (Figure 4). Following co-incubation of bacteria
lated with E. coli (K12), C. perfringens (NCTC 8237) or and substrates for different periods of time, samples were
B. fragilis (NCTC 9343). These organisms are known to removed, subjected to centrifugation in a minifuge at room
possess azoreductase activity.[16,17] The plates were incubated temperature to remove intact bacteria and the supernatants
under strictly anaerobic conditions at 37°C, together with stored at -20°C until required for analysis.
control plates without bacteria. After 24 h, the plates culti-
Compounds 1, 2 were consumed rapidly under these assay
vated with C. perfringens were completely decolourized conditions, producing prednisolone (4) and the respective
indicating substrate consumption. There was significant but cyclization products 5 (2-oxindole) and 7 (DHQ) (Figure 5).
incomplete substrate consumption in the plates inoculated Half-lives for prodrug consumption could be estimated
with B. fragilis, while E. coli plates remained unchanged.
assuming first-order kinetics as 1, 1.6 h (kobs 0.424 h-1,
Since C. perfringens demonstrated greater azoreductase r2 > 0.999) and 2, 1.2 h (kobs 0.584 h-1, r2 > 0.99). Differences
activity than B. fragilis, further assay development work was in concentration between the prednisolone and cyclization
undertaken with C. perfringens. This organism is widely dis- products (5, 7) did not achieve significance in either case
tributed in the human colon; it has high levels of an azoreduc- indicating that reduction/cyclization was the predominant
tase that exhibits wide crossed immuno reactivity with other mechanism of drug release. There was no evidence of
human flora azoreductase; it can process substrate without the hydrolysis of the ester bond between steroid and the respec-
need for additional co-factors; and its azoreductase is tive linkers. Compound 1 produced significantly less pred-
secreted, obviating the need for cell fractionation.[17–21] Colo- nisolone than 2c, consistent with the observations in the
rimetirc methods have been reported for monitoring substrate mouse colon contents and with their relative pharmacological
consumption that use C. perfringens as an azoreductase effects.[12] Olsalazine and balsalazide[9] were also reduced
source;[21] however, we required a selective assay that would under these conditions albeit more slowly. The role of
permit us to monitor steroid and cyclization adduct produc- azoreductase activity in the disappearance of 2 was con-
tion, in addition to substrate disappearance. Initial experi- firmed by co-incubating with the azoreductase inhibitor
ments with the agar plates inoculated with C. perfringens 2-iodosobenzoic acid (1 mm), which led to significant attenu-
were unsuccessful because of problems with accuracy and ation in rate (Figure 4). Co-incubation with menadione did
precision of extraction for the multiple analytes. The next not significantly depress reduction rate at <1 mm.
steps were therefore: (i) to develop bacterial suspensions in
Synthesis of BOC-3, BOC-6
BHI that could permit good precision and recovery of test
compound and products by HPLC; (ii) to validate an extrac- The intermediates BOC-3 and BOC-6 to study intramolecular
tion procedure; (iii) to determine the optimum substrate con- lactamization kinetics were synthesized as shown in Figure 6.

812
Journal of Pharmacy and Pharmacology 2011; 63: 806–816
22
20
15
10
5
(a)
(b)
OD 0.1
OD 0.2
OD 0.3
OD 0.47
OD 0.65
20
18
16
14
12
10
8
OD 0.9
OD 1.1
6
4
2
0
0
0
5
10
15
20
25
0
5
10
Time (h)
15
20
Time (h)
Figure 4 (a) Effect of C. perfringens cell count (OD) on rate of disappearance of 2 and (b) Disappearance of 2 (ꢀ) in the presence of C. perfringens
and in the presence of azoreductase inhibitor 2-iodosobenzoic acid (0.5 mm) (ᮀ).
17.5
15.0
12.5
10.0
7.5
17.5
15.0
12.5
10.0
7.5
(b)
(a)
5.0
5.0
2.5
2.5
0.0
0.0
0
5
10
Time (h)
15
20
0
5
10
Time (h)
15
20
Figure 5 Progress curves for the disappearance of (a) 1 ( ), (b) 2 and the evolution of prednisolone (ꢀ), 2-oxindole (5) (a) and (b) DHQ (7) (᭢),
1,2 in BHI (᭹) in the presence of C. perfringens (37°C, pH 7.4).
O
O
O
OH
(CH2)n
O
a
HO
OH
(CH2)n
NHBOC
NHBOC
O
n=1 13
n=2 14
n=1 BOC-3
n=2 BOC-6
Figure 6 a: Prednisolone (4), DCC, DMAP in DCM.
BOC-protection of the amino group was necessary to prevent
lactamization before or during the esterification with pred-
nisolone. Compound 13 was produced by BOC protection of
Nuclear magnetic resonance lactamization
in CDCl3
aminophenylacetic acid, which was in turn obtained by reduc- The lactamization of 3 and 6 was first monitored by NMR
tion (H2/Pd/C) of commercially available nitrophenyl acetic analysis in CDCl3. The BOC intermediates were dissolved in
acid. The synthesis of 14 was described previously.[12] The rate deuterated chloroform and a drop of TFA added to trigger the
of cyclization of the 2-aminophenyl propionate ester of cor- lactamization. In the case of 6, lactamization was complete by
tisone at several pH values was also reported in that study.
1H NMR after 128 min (stacked NMR spectra are shown in

Azoreductase activated steroid prodrugs
Juan F. Marquez Ruiz et al.
813
sensitive to phosphate, which can act in a concerted manner as
proton donor and acceptor. Therefore, the lactamization rate
constants were determined at four different buffer salt con-
centrations at each pH value and the rate at zero buffer esti-
mated by extrapolation. There was evidence of a hydroxide
ion-catalysed hydrolysis of the ester group in both cases
above pH 9; below this the disappearance of 3 and 6 could be
exclusively attributed to cyclization. The two outstanding fea-
tures of the pH rate profiles (Figure 8a, Table 1) are the mark-
edly higher rate constants for the cyclization of 6 (the
expected reduction product of 2) compared with 3 and sec-
ondly the general independence of pH across the pH range for
3 and at <pH 7 in the case of 6. The rate constants for the
cyclization of 6 were 5–7 fold higher than for 3 at <pH 7 and
only became similar at around pH 8 because of an apparent
HO--dependent fall in cyclization rate of 6. The maximum
rate of cyclization of 6 occurred in the pH range 3–4 followed
by a slight inflection at pH 4.3 associated with the pKa of the
anilide amino group in 6 (pKa 4.2). The cyclization of methyl
esters of 3-(2-aminophenyl) propionic and acetic acid has
been reported and there are similarities between the rate data
for the cyclization of 3 and 6 and their methyl ester ana-
logues.[22,23] The lactamization of the methyl esters is reported
to be relatively independent of pH and faster for 3-(2-
aminophenyl) propionate (leading to the 6-membered ring)
than methyl (2-aminophenyl)acetate. A mechanism was pro-
posed for intramolecular aminolysis of anilide esters by Fife
and Duddy[22] based on the general scheme for ester aminoly-
sis first proposed by Satterthwait and Jencks[24] and we assume
this is relevant to the cyclization of steroid esters 3 and 6
(Figure 8b). The anilines are shown in equilibrium with their
protonated forms. The weak nucleophilicity of the aromatic 6
means that there is a relatively small amount of the zwitteri-
onic tetrahedral intermediates 10b present. Direct breakdown
of the tetrahedral zwitterion can occur with good leaving
groups but not with the aliphatic alcohols such as predniso-
lone. In the case of the phenylpropionate 6, formation of the
neutral tetrahedral intermediate (12b) is likely to be rate deter-
mining (second-order reaction) and this may explain the
observed decrease in rates at increasing pH. For the kineti-
cally more favourable 5-membered ring closure, the concen-
tration of zwitterion 10a is expected to be higher and
breakdown of the neutral tetrahedral intermediate is likely to
be rate limiting. The slower breakdown of the intermediate in
the general 5-membered case has been attributed to the induc-
tive effect of the phenyl ring on the tetrahedral centre making
128 minutes
50 minutes
6
4
2
28 minutes
15 minutes
Compound 6 at 0 minutes
Compound BOC–6
8
6
4
2
ppm
Figure 7 Lactamization of 6 monitored by NMR in CDCl3.
Figure 7). Before deprotection, the CH2 from C-21 of the
prednisolone exhibits multiplicity forming a quadruplet at
4.93 ppm. Following removal of the BOC group the chemical
shift of the CH2 group at t0 was 5.06 ppm and a quadruplet;
this gradually transformed until it became a singlet at d
5.01 ppm. Also there was confirmation of the lactamization
when the CH2 groups from C-23 and C-24, initially triplets at
3.02 and 2.98 ppm, respectively, changed to one peak at
3.00 ppm. When the experiment was repeated with 3 lactam-
ization was not observed up to 1390 min. Under the experi-
mental conditions, the compounds were in the fully
protonated state (excess TFA) and in an anhydrous environ-
ment. The NMR experiment provided support for the identi-
fication of the lactamization products and it pointed to a
significant difference in the behaviour of the two correspond-
ing compounds that suggested an explanation for the failure of
one to exhibit biological activity. The study was repeated
more quantitatively under conditions that would more closely
mimic those in the intestine.
Lactamization studies in aqueous buffer
The rate of lactamization as a function of pH for compounds C-O bond breaking more difficult than in the six-membered
3 and 6 was studied in a universal buffer system at pH range case. This difference in cyclization rate constants between 3
2.5–8.5 (m = 0.12, 37°C). To commence lactamization, depro- and 6 has a significant effect on half-lives that are relevant to
tection of BOC-3 and BOC-6 was carried out in DCM at 0°C release rates in the key pH region 4–6 (74 min vs 564 min at
with TFA for several minutes, the volatiles removed under a pH 4.1). In the case of ulcerative colitis, the pH values in
stream of nitrogen and the residue dissolved immediately in stomach, proximal small intestine and ileum are similar to
the appropriate buffer solution. HPLC was used to measure healthy subjects but the pH in the colon is lower in many
the disappearance of the parent and the appearance of pred- individuals, especially when the disease is active.[25–27] Thus,
nisolone 4, dihydroquinolone 7 or 2-oxindole 5. Data for the pH values for the right colon/caecum from 2.3 to 3.4 have
disappearance of 3 and 6 could be analysed using first-order been recorded. Similar observations were made in Crohn’s
kinetics over the pH range and rates were determined by Disease patients and the pH values in the right and left colon
non-linear regression to an exponential decay. The lactamiza- had a mean value of 5.3 in both areas. Colonic pH values in
tion rate was found to be highly buffer species dependent in IBD patients therefore fall in the region where the cyclization
some regions. In particular, the cyclization of 6 was highly rate of the amino ester products of reduction of 1 and 2 are

814
Journal of Pharmacy and Pharmacology 2011; 63: 806–816
(a)
0.0100
0.0075
0.0050
0.0025
0.0000
1
2
3
4
5
6
7
8
pH
(b)
+
OPred
O–
+
N
H2
OPred
NH2
OPred
NH3
k1
k–1
Ka
O
O
(CH2)n
(CH2)n
(CH2)n
BH+ forms of 3, 6
10a n=1
10b n=2
n=1, 3
n=2, 6
k–b
k–2
k2
kb
H
N
H
N
OPred
OH
OPred
O–
H+
–H+
(CH2)n
(CH2)n
11a n=1
11b n=2
12a n=1
12b n=2
k3
kr
Prednisolone (4) + 5 or 7
Figure 8 (a) Plot of kobs vs pH for lactamization of 3 ( ) and 6 (᭺) producing 2-oxindole (5) and dihydroquinolone (7), respectively, at 37°C and
m = 0.12 m adjusted with NaCl. Rate constants were obtained by extrapolation to zero buffer over n = 4 experiments at a range of buffer concentrations
using a universal borate/citrate/phosphate system. (b) Mechanism for intramolecular aminolysis of 3 and 6 adapted from.[22]
most different indicating that in addition to its demonstrably tinal fluid ionic strength is 0.07.[28] As evident in Figure 9a the
better activity in animal models, 2 is a better candidate for higher the ionic strength, the longer the time to complete the
human development based on its activation and drug release lactamization reaction although the effect was small (2-fold).
rate in vitro.
We also evaluated the effect of increasing the proportion of
methanol (5–90%) in PBS at pH 7.4 (m = 0.12 at 37°C)
(Figure 9b, Table 2). We were interested in the relative behav-
iour of the compounds under conditions prevailing in the
Lactamization studies: effect of the ionic
strength and organic solvent
The relationship between ionic strength and rate of lactam- colon where free water content is low (13 Ϯ 12, fed; 18 Ϯ 26
ization of 3 and 6 was studied in the range m = 0.06–0.6 while fasted).[29,30] There was a marked 10–15-fold decrease in the
maintaining constant pH (7.4). The ionic strength throughout rate of cyclization the cases of both 3 and 6 with increasing
the human gastrointestinal tract is likely to vary significantly MeOH. This can attributed mechanistically to destabilization
depending on diet and water intake. The USP simulated intes- of the initial zwitterionic intermediate in the increasingly

Azoreductase activated steroid prodrugs
Juan F. Marquez Ruiz et al.
815
Table 1 Kinetic data for the lactamization of compounds 3 and 6 at 37°C (m = 0.15) in the pH range 2.4–8.5
pH
3
6
kobs (min-1)
t1/2 (min)
sd*
kobs (min-1)
t1/2 (min)
sd
2.4
3.4
0.0014
0.0014
0.0012
0.0012
0.0012
0.0012
0.0013
0.0012
0.0010
490.55
494.75
563.99
553.63
547.94
535.24
499.74
541.94
2243.19
0.0001
0.0001
0.0001
0.0001
0.0001
0.0003
0.0004
0.0002
0.0003
0.0090
0.0091
0.0093
0.0081
0.0078
0.0074
0.0065
0.0015
0.0012
76.30
70.75
74.22
84.87
88.76
93.17
106.62
437.86
573.32
0.0005
0.0040
0.0017
0.0070
0.0080
0.0006
0.0015
0.0018
0.0008
4.10
4.50
4.95
6.49
7.40
8.08
8.5
*sd is the standard deviation of the y intercept from plots of kobs vs [buffer] estimated using GraphPad Prism5.
0.015
0.010
0.005
0.000
0.010
0.009
0.008
0.007
0.006
0.005
0.004
(b)
(a)
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0
10 20 30 40 50 60 70 80 90 100
%MeOH
Ionic strength
Figure 9 Lactamization rate (kobs) vs (a) ionic strength (m = 0.05-0.6) and, (b) % MeOH (5-95%) for 3 ( ) and 6 (᭺) at 37°C at pH 7.4 (37°C).
Table 2 Kinetic data for the lactamization of 3 and 6 at pH 7.4 (37°C) at different proportions of MeOH
% MeOH
3
6
kobs
t1/2 (min)
sd*
kobs
t1/2 (min)
sd*
5
25
50
80
90
0.0091
0.0077
0.0059
0.0043
0.0013
75.97
89.41
116.0
159.6
513.8
0.0004
0.0002
0.0001
0.0001
0.0001
0.0135
0.0090
0.0054
0.0036
0.0016
51.00
76.36
127.1
187.9
432.6
0.0004
0.0003
0.00001
0.00005
0.00001
*sd is standard deviation of the kobs values: n = 3.
apolar environment (Figure 8b). Interestingly the behaviour of (3 and 6) arising from the actions of anaerobic bacteria on 1
the cyclization intermediates was most different at low and 2. There were significant differences in the rates of
organic (5%) and low ionic strength at which the pH rate cyclization in the colonic pH range found in IBD patients and
profile was determined, suggesting that these are closest to this may explain the relative yield of prednisolone from the
colonic conditions considering the divergent pharmacological two compounds in the culture media and their pharmacologi-
characteristics of 1 and 2.
cal activity.
Conclusions
Declarations
OPN-501 (2) produced more prednisolone than its analogue 1
in cultured mouse colon contents and in a newly developed
Conflict of interest
assay employing C. perfringens suspension. To explain this, The Author(s) declare(s) that they have no conflicts of interest
we studied the rates of cyclization of the amino intermediates to disclose.

816
Journal of Pharmacy and Pharmacology 2011; 63: 806–816
15. Berry LM et al. Esterase activities in the blood, liver and
intestine of several preclinical species and humans. Drug Metab
Lett 2009; 3: 70–77.
16. Khan AA et al. Identification of predominant human and animal
anaerobic intestinal bacterial species by terminal restriction frag-
ment patterns (TRFPs): a rapid, PCR-based method. Mol Cell
Probes 2001; 15: 349–355.
17. Rafii F, Cerniglia CE. Comparison of the azoreductase and
nitroreductase from Clostridium perfringens. Appl Environ
Microbiol 1993; 59: 1731–1734.
Funding
This work was supported by the Enterprise Ireland and
OPSONA Therapeutics LTD under the Innovations Partner-
ship Program (IP/2007/503).
References
1. Cohen RD et al. A meta-analysis and overview of the lite-
rature on treatment options for left-sided ulcerative colitis
and ulcerative proctitis. Am J Gastroenterol 2000; 95: 1263– 18. Rafii F, Cerniglia CE. Reduction of azo dyes and nitroaromatic
1276.
compounds by bacterial enzymes from the human intestinal
tract. Environ Health Perspect 1995; 103: 17–19.
19. Rafii F, Coleman T. Cloning and expression in Escherichia coli
of an azoreductase gene from Clostridium perfringens and com-
parison with azoreductase genes from other bacteria. J Basic
Microbiol 1999; 39: 29–35.
2. Campieri M. New steroids and new salicylates in inflammatory
bowel disease. Gut 2002; 50: iii43–iii46.
3. Basit AW. Oral colon-specific drug delivery using amylose-
based film coatings. Pharm Technol Eur 2000; 12: 30–36.
4. Haeberlin B et al. In vitro evaluation of dexamethasone-beta-D-
glucuronide for colon-specific drug delivery. Pharm Res 1993; 20. Rafii F et al. Azoreductase activity of anaerobic bacteria isolated
11: 1553–1562.
from human intestinal microflora. Appl Environ Microbiol 1990;
5. Hu Z et al. New preparation method of intestinal pressure-
56: 2146–2151.
controlled colon delivery capsules coating machine and evalua- 21. Semdé R et al. Study of important factors involved in azo deriva-
tion in beagle dogs. J Controlled Release 1998; 56: 293–302.
6. Jung YJ et al. Prednisolone 21-sulfate sodium: a colon-specific
tive reduction by Clostridium Perfringens. Int J Pharm 1998;
161: 45–54.
pro-drug of prednisolone. J Pharm Pharmacol 2003; 55: 1075– 22. Fife TH, Duddy NW. Intramolecular aminolysis of esters.
1082.
Cyclization of esters of (o-Aminophenyl)acetic acid. J Am Chem
Soc 1983; 105: 74–79.
7. Yano H et al. Colon specific delivery of prednisolone-appended
a-cyclodextrin conjugate: alleviation of systemic side effect 23. Kirby AJ, Mujahid TG. Anilide formation from an aliphatic
after oral administration. J Controlled Release 2002; 79: 103–
112.
8. Sinha VR, Kumria R. Colonic drug delivery: prodrug approach.
Pharm Res 2001; 18: 557–564.
ester. The mechanism of cyclisation of methyl 3-(2-
aminophenyl)propionate. J Chem Soc Perkin Trans 1979; 2:
1610–1616.
24. Satterthwait AC, Jencks WP. The mechanism of the aminolysis
of acetate esters. J Am Chem Soc 1974; 96: 7018–7031.
9. Wiggins JB, Rajapakse R. Balsalazide: a novel 5-aminosalicylate
prodrug for the treatment of active ulcerative colitis. Expert Opin 25. Evans DF et al. Measurement of gastrointestinal pH profiles in
Drug Metab Toxicol 2009; 5: 1279–1284. normal ambulant human subjects. Gut 1988; 29: 1035–1041.
10. Jain A et al. Azo chemistry and its potential for colonic delivery. 26. Fallingborg J et al. Very low intraluminal colonic pH in patients
Crit Rev Ther Drug Carrier Syst 2006; 23: 349–400. with active ulcerative colitis. Dig Dis Sci 1993; 38: 1989–1993.
11. Carceller E et al. Novel azo derivatives as prodrugs of 27. Nugent S et al. Intestinal luminal pH in inflammatory bowel
5-aminosalicylic acid and amino derivatives with potent platelet
activating factor antagonist activity. J Med Chem 2001; 44:
3001–3013.
disease: possible determinants and implications for therapy with
aminosalicylates and other drugs. Gut 2001; 48: 571–577.
28. Stippler E et al. Comparison of US pharmacopeia simulated
intestinal fluid TS (without pancreatin) and phosphate standard
buffer pH 6.8, TS of the international pharmacopoeia with
respect to their use in in vitro dissolution testing. Dissolution
Technol 2004; 11: 6–10.
12. Márquez Ruiz JF et al. Design, synthesis, and pharmacological
effects of a cyclization-activated steroid prodrug for colon tar-
geting in inflammatory bowel disease. J Med Chem 2009; 59:
3205–3211.
13. Gomes P et al. Cyclization-activated prodrugs. Molecules 2007; 29. Diakidou A et al. Characterization of the contents of ascending
12: 2484–2506.
colon to which drugs are exposed after oral administration to
14. Temple NJ, El-Khatib SM. High-fat diets and fecal level of
healthy adults. Pharm Res 2009; 26: 2141–2151.
reductase and colon mucosal level of ornithine decarboxylase, 30. Schiller C et al. Intestinal fluid volumes and transit of dosage
beta-glucuronidase, 5′-nucleotidase, ATPase, and esterase in
mice. J Natl Cancer Inst 1984; 72: 679–684.
forms as assessed by magnetic resonance imaging. Aliment
Pharmacol Ther 2005; 22: 971–979.