A nitrophenyl-based prodrug type for colorectal targeting of prednisolone, budesonide and celecoxib

Article abstract of DOI:10.1016/j.bmcl.2013.01.060

Celecoxib is a COX-2 inhibitor drug that can be used to reduce the risk of colorectal adenocarcinoma. Glucocorticoids are used in the treatment of inflammatory bowel disease. A limitation to the use of both drug types is that they undergo absorption from the intestinal tract with serious side effects. The prodrug systems introduced here involve forming a nitro-substituted acylsulfonamide group in the case of celecoxib and a nitro-substituted 21-ester for the glucocorticoids. Drug release is triggered by the nitro reductase action of the colonic microflora, liberating a cyclization competent species. The release of the active parent drugs was evaluated in vitro using Clostridium perfringens and epithelial transport through Caco-2 monolayer evaluation was carried out to estimate the absorption properties of the prodrugs compared to the parental drugs.

Full text of DOI:10.1016/j.bmcl.2013.01.060

Bioorganic & Medicinal Chemistry Letters 23 (2013) 1693–1698
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
A nitrophenyl-based prodrug type for colorectal targeting
of prednisolone, budesonide and celecoxib
Juan F. Marquez Ruiz a, Kinga Kedziora a, Maria Pigott a, Brian Keogh c, Henry Windle b, Jason Gavin a,
Dermot P. Kelleher b, John F. Gilmer a,
⇑
a School of Pharmacy and Pharmaceutical Sciences, Trinity College, Dublin 2, Ireland
b School of Medicine, Trinity College, Dublin 2, Ireland
c Opsona Therapeutics Ltd, Trinity Centre for Health Sciences, Institute of Molecular Medicine, St. James’s Hospital, Dublin 8, Ireland
a r t i c l e i n f o
a b s t r a c t
Article history:
Celecoxib is a COX-2 inhibitor drug that can be used to reduce the risk of colorectal adenocarcinoma. Glu-
cocorticoids are used in the treatment of inflammatory bowel disease. A limitation to the use of both drug
types is that they undergo absorption from the intestinal tract with serious side effects. The prodrug sys-
tems introduced here involve forming a nitro-substituted acylsulfonamide group in the case of celecoxib
and a nitro-substituted 21-ester for the glucocorticoids. Drug release is triggered by the nitro reductase
action of the colonic microflora, liberating a cyclization competent species. The release of the active par-
ent drugs was evaluated in vitro using Clostridium perfringens and epithelial transport through Caco-2
monolayer evaluation was carried out to estimate the absorption properties of the prodrugs compared
to the parental drugs.
Received 5 December 2012
Revised 11 January 2013
Accepted 16 January 2013
Available online 8 February 2013
Keywords:
Prodrug
Colon
Targeting
Nitroreductase
Cancer
Ó 2013 Elsevier Ltd. All rights reserved.
Celecoxib
Budesonide
Prednisolone
Oral drug delivery to the colon is pursued in order to improve
therapy for colorectal diseases or as a means to increase bioavail-
ability of drugs that would be degraded in the small intestine.1,2
The conventional pharmaceutical strategy for achieving colon tar-
geting involves the use of slowly biodegradable polymers that de-
lay drug release until the oral caecal transit time. The variability in
this parameter has led to the development of more sophisticated
formulation approaches in which the matrix exhibits more or less
specific sensitivity to a chemical or biological property of the colon,
for example a property associated with its dense microflora.3 The
candidate drug may also be linked covalently to the biodegradable
polymer through a biodegradable bond.4
The more explicit prodrug approach to colorectal delivery in-
volves attaching a group that suppresses intestinal absorption
using a form of covalent linkage that is susceptible to cleavage in
the colon.5 The most well known examples of this are the azo-
linked derivatives of 5-amino salicylic acid such as olsalazine and
balsalazide, which were inspired by the older sulfasalazine. The
latter remains an important compound not just for ulcerative coli-
tis but also for arthritis. There have been several investigations into
prodrugs that target sugar metabolism of bacteria, for example
steroidal glucuronides and glycosides, cyclodextrin and even dex-
tran conjugates.6,7 Amino acid carriers have been investigated
more recently for colorectal targeting of ursodeoxycholic acid
(UDCA), celecoxib and naproxen.8,9 Sulfate conjugates of UDCA
have been explored for similar reasons.10 As of yet none of these
prodrug approaches can rival the clinical success of the azo pro-
drugs of 5-ASA. It may be that the reductive capacity of the colon
more clearly distinguishes it from the small intestine than other
potential vectors. The azoprodrug approach is directly applicable
only to primary aromatic amino drugs such as 5-ASA. However,
we showed, that by incorporating an azoreductase sensitive linker,
it is possible to adapt the approach to drugs bearing a primary
alcohol group as shown in Figure 1. These conjugates are less prone
to intestinal absorption than the parent compounds but they can
be specifically activated by murine colonic microflora and by
Clostridium perfringens, an obligate anaerobe found widely in the
human colon. A prednisolone analogue was as efficacious as its
parent in the murine dextran sodium sulfate model of ulcerative
colitis but caused less thymic atrophy, indicating reduced systemic
steroid exposure.11 We subsequently showed that the approach
could be applied to the more sterically hindered steroid budeso-
nide and to celecoxib, a sulfonamide COX-2 inhibitor, through
the use of an acylsulfonamide link.12
⇑
Corresponding author. Tel.: +353 1 896 2795; fax: +353 1 896 2793.
E-mail address: gilmerjf@tcd.ie (J.F. Gilmer).
0960-894X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmcl.2013.01.060

1694
J. F. Marquez Ruiz et al. / Bioorg. Med. Chem. Lett. 23 (2013) 1693–1698
O
O
OH
NH2
OH
ODRUG
NH2
(CH2)n
DrugO
(CH2)n
O
N
N
HO
+
OH
5-ASA
O
Azoreductase
activity
Intramolecular
lactamization
(CH2)n
O
DRUG-OH
N
H
Figure 1. Azoreductase-cyclization activation mechanism.
O2N
O
O
O
O
O
(CH2)n
HO
O
O
HO
(CH2)n
NO2
OH
O
O
O
3 n=1
4 n=2
1 n=1
2 n=2
a
O
b
n(CH2)
OH
NO2
9 n=1
10 n=2
c
O2N
O
S
O
N
(CH2)n
O H
N
N
CF3
5 n=1
6 n=1
Scheme 1. Reagents: (a) Prednisolone, DMAP, DCC in ACN; (b) budesonide, DMAP, DCC in ACN; (c) celecoxib, EDC, DMAP in tBuOH/ClCH2CH2Cl.
In this study we have further extended the cyclization approach
propionic acid (10), which was prepared as described.13 The
starting material for this was 2-nitrobenzoyl bromide which was
treated with dimethylmalonate under basic conditions to obtain
dimethyl-2-(2-nitrobenzyl)-malonate, which was heated under
acidic conditions to yield 10 (Scheme 1). The esterification reac-
tions between the carboxylic acid group of 9 and 10 and the
21-OH groups of the prednisolone and budesonide to obtain the
prodrugs 1, 2, 3 and 4 was carried out using DCC and DMAP in
ACN.14 The synthesis of the celecoxib prodrugs 515 and 6 was
achieved using as coupling agent (1-(3-dimethylaminopropyl)-3-
ethylcarbodiimide (EDC). In order to evaluate the intrinsic stability
of the ester/acylsulfonamide groups towards the C. perfringens and
the role of the nitro group in ester scission and prodrug activation
we prepared two analogs without a nitro group 7 and 8 (Scheme 2).
using a nitroreductase activation design. We have focused on nitro
substituted phenyl acetic and propionic acid esters of three drugs
that would benefit from colorectal targeting-budesonide, prednis-
olone and celecoxib (Scheme 1). The passive permeability of the
compounds was assessed using the predictive Caco-2 monolayer
model while their capacity to trigger drug release was investigated
using C. perfringens as a model intestinal anaerobe. The stability of
the prodrugs towards hydrolysis under simulated gastric and duo-
denal conditions was tested in order to predict off-target drug
release.
The pro-moieties/carriers selected for attachment to the ste-
roids prednisolone, budesonide and celecoxib were 2-nitrophenyl
acetic acid 9, which was commercially available, and 2-nitrophenyl

J. F. Marquez Ruiz et al. / Bioorg. Med. Chem. Lett. 23 (2013) 1693–1698
1695
O
S
O
N
O H
N
N
CF3
a
7
O
OH
O
O
O
HO
11 n=2
OH
b
O
8
Scheme 2. Synthesis of 7 and 8; Reagents: (a) celecoxib, EDC, DMAP in tBuOH/ClCH2CH2Cl; (b) prednisolone, DMAP, DCC in ACN.
The candidate compounds were tested for prodrug activation
using a suspension of C. perfringens.16 The day before the experi-
ment C. perfringens were inoculated on agar plates containing brain
heart infusion medium (BHI). On the day of experiments C. perfrin-
gens colonies were scratched from the agar plates and inoculated in
BHI media to obtain cell density between 0.9 and 1.1 at 600 nm.
the N-acyl compound (cyclic hydroxamate), whereas at smaller
chain length the OH group of the hydroxylamine is expected to
cyclise.17 In a potential ADEPT application, Liu and Hu showed that
certain nitro ester analogs of 20-deoxy-5-fluorouridine could be re-
duced by isolated Escherichia coli nitroreductase causing spontane-
ous cyclization and DFU release.18 However, whole cell reduction
by anaerobes may be more likely to produce the amino (6 electron)
products.
Test compound was introduced into suspension at 50 lM in 1%
DMSO. The negative control consisted of BHI media with 3–6 at the
same concentration.
Immediately after the compounds were added to the
C. perfringens, an aliquot was taken (200 ll) and added to ACN
In order to identify the byproduct of prodrug activation by C.
perfringens we sought to establish in vitro synthetic conditions that
could replicate it. Since 4-electron reduction seemed a likely possi-
bility and since dithionite is reported to produce hydroxylamines
from nitroimidazoles, we tried reductions several times with 1
and 2 in the presence of sodium dithionite under various states
of oxygen availability and in various solvents (ACN, IPA, H2O).19
Unfortunately, these conditions produced as byproducts, predniso-
lone and oxindole or DHQ, the expected cyclization product from
complete reduction of the nitrogroup to amine, indicating that in
the presence of sodium dithionite, the last step in the reduction
process occurred more rapidly than the cyclization of the hydrox-
ylamine intermediates.
We focussed subsequent efforts to produce a hydroxylamine or
its cyclic hydroxamate product on the methyl ester of nitrophenyl-
acetic acid as this was more synthetically accessible than the pro-
drugs. The methyl ester was treated with NaBH4 in aqueous
dioxane in the presence of palladized charcoal, conditions reported
to favour the production of hydroxylamines from aromatic nitro
compounds, rather than the aniline products. Under these condi-
tions nitrophenyl acetic acid methyl ester produced a major prod-
uct with identical relative retention time (prednisolone) to the
unknown byproduct of activation of 1. This was identified by
NMR and MS as 1,3-dihydro-1-hydroxy-2H-indole-2-one, the cyc-
lic hydroxamate resulting from cyclization of the hydroxylamine
product of incomplete reduction of the nitro ester.
and samples were centrifuged at 10,000 rpm for 10 min. Aliquots
were taken at time points 0, 2, 4, 6 and 24 h while maintaining
anaerobic conditions. Supernatant was transferred to a tube and
stored at ꢀ20 °C until analysis by a HPLC method that could
separate BHI components from the prodrug and expected byprod-
ucts of activation.
The nitro prodrugs were activated in the bacterial suspensions
(Fig. 2A–F) with substrate consumption observed within 6 h in
the prednisolone cases (1,2) and within 4 h in the celecoxib cases
(t1/2 ꢁ2 h). Processing was slower in the case of the budesonide
prodrug 3 and incomplete in the case of the propionate 4. Budeso-
nide is probably more susceptible to steric hindrance of reduction
and/or cyclization than prednisolone. All of the compounds were
stable when incubated at the same temperature in BHI (i.e., in
the absence of bacteria). Furthermore, 7 and 8, which lack the nitro
group, were not processed to a significant extent in the bacterial
model proving that nitro group metabolism triggered drug release
in the other cases (Fig. 3).
Although the nitroester compounds were presumably being re-
duced we were unable to detect amino intermediates in this pro-
cess or the corresponding lactamization byproducts when the
HPLC effluent was monitored by UV or MS. Samples of oxindole
and dihydroquinolone (DHQ) were available from our previous
work which allowed us to confirm their absence. Under reverse
phase HPLC gradient conditions, the suspensions of C. perfringens
and 1 and 2 contained unidentified components at shorter reten-
tion than oxindole (in the case of 3) and DHQ (in the case of 4).
Aromatic nitro compounds may be reduced to the correspond-
ing amines by aerobic and anaerobic bacteria. Under anaerobic
conditions the complete reduction to aromatic amine involves a
six electron transfer via nitroso (2-electron) and hydroxyl amino
(4-electron) intermediates. The latter is cyclization competent. In
the incipient five-membered ring case, the expected product is
C. perfringens reduction or biodegradation of aromatic nitro
compounds (e.g., trinitrotoluene) has been extensively studied.20
The reaction may be mediated by ferrodoxin hydrogenase systems
rather than by a specific nitroreductase enzyme. The fate of the
resultant hydroxylamine would, under colonic conditions, depend
on the relative susceptibility of the hydroxylamine to undergo
cyclization and the reductive potential for sequential production
of the corresponding amine. Clearly in the in vitro model condi-
tions used, the cyclization reaction speed exceeded the rate of
amine production. The processing of the compounds could be

1696
J. F. Marquez Ruiz et al. / Bioorg. Med. Chem. Lett. 23 (2013) 1693–1698
20
15
10
5
20
A
C
E
B
15
10
5
0
0
0
4
8
12
16
20
24
0
4
8
12
16
20
24
Time [h]
Time [h]
20
15
10
5
20
15
10
5
D
F
0
0
0
4
8
12
16
20
24
0
4
8
12
Time [h]
16
20
24
Time [h]
20
15
10
5
20
15
10
5
0
0
0
4
8
12
16
20
24
0
4
8
12
16
20
24
Time [h]
Time [h]
Figure 2. Progress curves following incubation of 1, 2, 3, 4, 5, 6 (j) in suspension of C. perfringens under anaerobic conditions at pH 7.4 and 37 °C. (A) 1 (j), prednisolone (N),
and 1 in BHI (.); (B) 2 (j), prednisolone (N), and 2 in BHI (.); (C) 3 (j), budesonide (N), and 3 in BHI (.); (D) 4 (j), budesonide (N), and 4 in BHI (.); (E) 5 (j), celecoxib (N),
and 5 in BHI (.); (F) 6 (j), celecoxib (N), and 6 in BHI (.) (n = 3).
20
A
B
12
9
15
10
5
6
3
0
0
0
4
8
12
16
20
24
0
4
8
12
16
20
24
Time [h]
Time [h]
Figure 3. Progress curves following incubation of 7 and 8 (j) in suspension of C. perfringens under anaerobic conditions at pH 7.4 and 37 °C. (A) 7 (j), celecoxib (N), and 7 in
BHI (.); (B) 8 (j) and 8 in BHI (.)-no prednisolone was observed.
studied in conditions more closely related to the colon to predict
byproduct identity in order to characterise potential off-target ef-
fects. However the model data taken together with the stability
in simulated fluid suggest that the prodrugs would be selectively
activated in the colon if able to resist absorption during transit.
The potential intestinal permeability of 3, 4, 5 and 6 was then as-
sessed using the Caco-2 assay (budesonide results were expected
to be relevant to the properties of prednisolone). The transport of
budesonide and celecoxib was tested for comparison. These clini-
cally used drugs were at some point selected from amongst their

J. F. Marquez Ruiz et al. / Bioorg. Med. Chem. Lett. 23 (2013) 1693–1698
1697
7.0×10-7
2.0×10-5
1.5×10-5
1.0×10-5
5.0×10-6
0
A
B
6.0×10-7
5.0×10-7
4.0×10-7
3.0×10-7
P
P
2.0×10-7
1.0×10-7
0
Budesonide
3
Celecoxib
5
6
Figure 4. (A) Apparent permeability coefficient of 5, 6 and celecoxib; (B) Apparent permeability coefficient transport of 3 and budesonide. There was no evidence of transport
of 4 under similar conditions.
50
40
30
20
10
0
50
40
30
20
10
0
A
B
0
250
500
750 1000 1250 1500
0
250
500
750 1000 1250 1500
Time [min]
Time [min]
Figure 5. (A) Stability of 1 (j), 2 (D D), 3 (ꢀ), 4 (s), 5 (h) and 6
), 3 (ꢀ), 4 (s), 5 (h) and 6 (N) in simulated human gastric juices at 37 °C.22 (B) Stability of compounds 1 (j), 2 (
(N) in simulated human duodenal fluid at 37 °C.22
development rivals in part because of their permeability character-
istics. Therefore we were confident that the relatively small struc-
tural changes involved in the prodrug design would be sufficient to
significantly attenuate passive diffusion potential. Absorptive
transport from apical to basolateral side (AP > BL) was evaluated
for appearance of the test compound and in the case of the pro-
drugs, the potential metabolic byproducts. Aliquots were collected
after 0, 30, 60, and 120 min from the acceptor reservoir and fresh
buffer was added to maintain the initial volume. The donor solu-
tion was also collected after 120 min. Caffeine was tested in two
wells in the AP > BL direction as a routine control. Trans-epithelial
byproducts of the reduction/cyclization process under simulated
colonic conditions; (ii) the potential for elaboration of the linker
unit in order to influence (i) as well as pharmaceutical properties
such as stability and permeability; the rates of reduction of the ni-
tro group and its electronic and steric control.24
References and notes
1. Krishnaiah, Y. S.; Khan, M. A. Pharm. Dev. Technol. 2012, 17, 521.
2. Xiao, B.; Merlin, D. Expert Opin. Drug Deliv. 2012, 9, 1393.
3. Patel, M.; Amin, A. Crit. Rev. Ther. Drug Carrier Syst. 2011, 28, 489.
4. Vaidya, A.; Agarwal, A.; Jain, A.; Agrawal, R. K.; Jain, S. K. Curr. Pharm. Des. 2011,
17, 1108.
electric resistance (TEER) was over 350
X
cm2 in all wells used in
the experiments prior to the addition of the test compounds.21
Compounds 3, 5, 6 possessed significantly lower permeability
than the parents celecoxib and budesonide (Fig. 4). No transport
of budesonide prodrug 4 could be detected. This data indicates that
the compounds would be able to resist absorption and transit the
intestinal tract if delivered orally. Nitroreductase-triggered cycliza-
tion and drug release may therefore be a generally effective ap-
proach to achieving colorectal drug targeting. The prodrugs were
processed in the presence of C. perfingens present in the human co-
lon, releasing celecoxib, prednisolone or budesonide. The stability
of the candidate prodrugs in this study was tested using the USP
gastric and intestinal fluid model, which mimics the enzymatic
and chemical conditions in the stomach and small intestine.22
Amounts of remaining compound was assessed using HPLC. The
compounds were found to be stable over the expected orocaecal
transit time (Fig. 5).23 Finally, they are predicted by the standard
Caco-2 permeability assay to be poorly absorbed relative to their
parents suggesting that they might arrive as intact substrates for
reductase activity associated with the microflora. Several aspects
deserve further study: (i) the identity and biological profile of
5. Dhaneshwar, S. S.; Vadnerkar, G. Curr. Top. Med. Chem. 2011, 11, 2318.
6. Varshosaz, J.; Emami, J.; Fassihi, A.; Tavakoli, N.; Minaiyan, M.; Ahmadi, F.;
Mahzouni, P.; Dorkoosh, F. Int. J. Colorectal Dis. 2010, 25, 1159.
7. Varshosaz, J. Expert Opin. Drug Deliv. 2010, 9, 509.
8. Asciutti, S.; Castellani, D.; Nardi, E.; Morelli, O.; Clementi, M.; Chistolini, F.;
Gentili, G.; Setchell, K. D.; O’Connell, N.; Pellicciari, R.; Clerici, C. Anticancer Res.
2009, 29, 4971.
9. Mishra, A.; Agrawal, S.; Pathak, K. Drug Deliv. 2012, 19, 102.
10. Rodrigues, C. M.; Kren, B. T.; Steer, C. J.; Setchell, K. D. Gastroenterology 1995,
109, 1835.
11. Ruiz, J. F.; Radics, G.; Windle, H.; Serra, H. O.; Simplicio, A. L.; Kedziora, K.;
Fallon, P. G.; Kelleher, D. P.; Gilmer, J. F. J. Med. Chem. 2009, 52, 3205.
12. Ruiz, J. F.; Kedziora, K.; Keogh, B.; Maguire, J.; Reilly, M.; Windle, H.; Kelleher, D.
P.; Gilmer, J. F. Bioorg. Med. Chem. Lett. 2011, 21, 6636.
13. Hao, J.; Ge, F.; Liu, Y.; Qian, Y. Peop. Rep. China Patent CN 1760173, 2006.
14. Boden, E. P.; Keck, G. E. J. Org. Chem. 1985, 50, 2394.
15. Spectroscopic data for 5. 1H NMR (CDCl3) d: 8.70 (1H, s., 8.10 (1H, d, J 8.52 Hz),
8.05 (2H, d, J 8.56 Hz), 7.61 (1H, t, J 7.52 Hz), 7.51 (3H, m), 7.38 (2H, d, J
7.52 Hz), 7.27 (1H, s), 7.16 (2H, d, J 7.52 Hz), 7.12 (2H, d, J 8.04 Hz), 6.74 (1H, s),
3.89 (2H, s), 2.38 (3H, s). 13C NMR (CDCl3) d: 166.37, 147.89, 144.93, 143.14,
139.38, 136.94, 133.72, 133.28, 129.35, 129.09, 128.93, 128.28, 127,55, 125.15,
125.05, 124.71, 105.98, 41.14, 29.24, 20.87. Expected (MꢀH+) = 545.1151,
Found (MꢀH+) = 545.1113.
16. Ruiz, J. F.; Kedziora, K.; Windle, H.; Kelleher, D. P.; Gilmer, J. F. J. Pharm.
Pharmacol. 2011, 63, 806.

1698
J. F. Marquez Ruiz et al. / Bioorg. Med. Chem. Lett. 23 (2013) 1693–1698
17. Henning, L.; Tabakovic, I. In Advances in Heterocyclic Chemistry; Katritzky, A. R.,
Ed.; Academic Press: Orlando, 1984; Vol. 36, p 253.
18. Liu, B.; Hu, L. Bioorg. Med. Chem. 2003, 11, 3889.
Papp = J/AsC0
Where is the rate of appearance of the compound on the acceptor
compartment mol/s), C0 is the initial concentration on the apical side
J
(l
19. Kedderis, G. L.; Argenbright, L. S.; Miwa, G. T. Arch. Biochem. Biophys. 1988, 262, 40.
20. Spain, J. C.; Hughes, J. B.; Knackmuss, H. J. Biodegradation of Nitroaromatic
Compounds and Explosives; CRC Presss LLC: Florida, 2000.
(mM) as determined by the analysis of the donor solution, and As is the surface
area of the monolayer (cm2).
22. United States Pharmacopoeia 32/National, Formulary 27, 2009.
23. Clegg, M.; Shafat, A. Br. J. Nutr. 2010, 104, 554.
21. The apparent permeability coefficients (Papp
) expressed in cm/s were
determined according to the following equation:
24. Saphier, S.; Haft, A.; Margel, S. J. Med. Chem. 2012, 55, 1081.