Mechanisms of trovafloxacin hepatotoxicity: Studies of a model cyclopropylamine-containing system

Article abstract of DOI:10.1016/j.bmcl.2007.10.070

The mechanism for the hepatotoxicity of trovafloxacin remains unresolved. Trovafloxacin contains a cyclopropylamine moiety which has a potential to be oxidized to reactive intermediate(s) although other putative elements may exist. In this study, a drug model of trovafloxacin containing the cyclopropylamine substructure was synthesized. Chemical oxidation of the drug model by K₃Fe(CN)₆ and NaClO revealed that both oxidants oxidize this drug model to a reactive α,β-unsaturated aldehyde, 11. The structure of 11 was fully elucidated by LC/MS/MS and NMR analysis. These results suggested that P450s with heme-iron center and myeloperoxidase generating hypochlorous acid in the presence of chloride ion are capable of bioactivating the cyclopropylamine moiety of trovafloxacin. This deleterious metabolism may lead to eventual hepatotoxicity.

Full text of DOI:10.1016/j.bmcl.2007.10.070

Available online at www.sciencedirect.com
Bioorganic & Medicinal Chemistry Letters 17 (2007) 6682–6686
Mechanisms of trovafloxacin hepatotoxicity: Studies
of a model cyclopropylamine-containing system
Qin Sun, Ran Zhu, Frank W. Foss, Jr.ꢀ and Timothy L. Macdonald*
Department of Chemistry, University of Virginia, McCormick Road, PO Box 400319, Charlottesville, VA 22904, USA
Received 18 September 2007; revised 18 October 2007; accepted 19 October 2007
Available online 24 October 2007
Abstract—The mechanism for the hepatotoxicity of trovafloxacin remains unresolved. Trovafloxacin contains a cyclopropylamine
moiety which has a potential to be oxidized to reactive intermediate(s) although other putative elements may exist. In this study, a
drug model of trovafloxacin containing the cyclopropylamine substructure was synthesized. Chemical oxidation of the drug model
by K3Fe(CN)6 and NaClO revealed that both oxidants oxidize this drug model to a reactive a,b-unsaturated aldehyde, 11. The
structure of 11 was fully elucidated by LC/MS/MS and NMR analysis. These results suggested that P450s with heme-iron center
and myeloperoxidase generating hypochlorous acid in the presence of chloride ion are capable of bioactivating the cyclopropylamine
moiety of trovafloxacin. This deleterious metabolism may lead to eventual hepatotoxicity.
ꢀ 2007 Elsevier Ltd. All rights reserved.
Trovafloxacin (TrovanTM) 1 is a naphthyridone agent of
the fluoroquinolone antibiotics. Major advantages of 1
over traditional agents such as ciprofloxacin and later
fluoroquinolone sparfloxacin lie in its broader spectrum
and superior pharmacokinetic profiles.1 Since 1’s mar-
keting in 1998, rare but severe hepatotoxicities have
been reported, which ultimately led to Trovan’s with-
drawal in some countries and a black-box warning in
United States since 1999.2 Clinical reports characterized
the hepatotoxicity of trovafloxacin as an idiosyncratic
drug reaction since it was unpredictable and possibly
immunologically based.3
The cyclopropylamine moiety is the most suspect sub-
structure to induce hepatotoxicity of trovafloxacin. Pre-
vious studies showed that P450s,5–8 monoamine
oxidases (MAOs),9–12 and horseradish peroxidase
(HRP)13 can all oxidize cyclopropylamine to a carbon-
centered radical, which can be subsequently oxidized
to a reactive a,b-unsaturated aldehyde. Although the
difluoroanilino functional group was suggested to gener-
ate reactive metabolites,14 there has been no corroborat-
ing experimental evidence. To exclude interferences
from other possible oxidizable positions, especially the
difluoroanilino moiety, a drug model of trovafloxacin
2 (Fig. 1) containing the cyclopropylamine substructure
was synthesized. Chemical oxidations, which mimic the
oxidation modes of different enzymes, were then per-
formed to help propose responsible enzymes that can
oxidize 2 and also to investigate reactive intermediate(s)
that can be generated from the respective oxidation.
The etiology of idiosyncratic hepatotoxicity involves
multiple factors. It is well recognized that metabolic
activation of some drugs may lead to the formation of
a reactive species capable of covalent binding to pro-
teins, ultimately culminating in a toxic response.4 Struc-
turally, trovafloxacin possesses two elements that have
the potential to generate reactive intermediate(s): a
cyclopropylamine moiety and a difluoroanilino system.
O
N
O
F
F
OH
H
H
N
N
N
N
H2N
H2N
H
H
F
Keywords: Trovafloxacin; Idiosyncratic drug reaction; Cyclopropyl-
amine; K3Fe(CN)6; NaClO; P450s; MPO.
2
1
*
Corresponding author. Tel.: +1 434 924 7718; fax: +1 434 982
2302; e-mail: tlm@virginia.edu
Figure 1. Structures of trovafloxacin (TrovanTM) 1 and the drug model
of trovafloxacin 2. The bold lines highlight the similarities between
trovafloxacin and the drug model.
ꢀ
Present address: Department of Chemistry, Columbia University,
New York, NY 10027, USA.
0960-894X/$ - see front matter ꢀ 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2007.10.070

Q. Sun et al. / Bioorg. Med. Chem. Lett. 17 (2007) 6682–6686
6683
O
N
O
K3Fe(CN)6 was chosen as a single electron transfer
(SET) oxidant, mimicking heme-iron centered enzymes
such as P450s. NaClO was applied to mimic myeloper-
oxidase (MPO), which can generate hypochlorous acid
in the presence of chloride ion. Our interest in the oxida-
tion of 2 by MPO arose from the fact that neutrophils,
which release MPO, have been associated with idiosyn-
cratic drug reaction of trovafloxacin in a LPS-trovaflox-
acin idiosyncrasy animal model.15
O
N
O
H
H
H
H
a
b
O2N
O2N
N
5
4
3
H
H
c
d
e
O
HN
H
H
H2N
N
O
N
6
7
O
HN
H
H
H
H
O
HN
f
g
H
Monoamine oxidases (MAOs) should also be considered
for the bioactivation of the drug model, especially con-
sidering that MAOs oxidize a variety of primary cyclo-
propylamines to reactive species.9–12 Some flavin
models, such as 3-methyllumiflavin16 and 3-methyl-5-
ethyllumiflavinium perchlorate,17 have been applied to
mimic the oxidation of MAOs. However, these model
systems failed to fully mimic MAO activity. MAOs cat-
O
O
.
H2N
TFA
N
NH
N
N
N
H
9
10
8
Scheme 1. Reagents and conditions: (a) CH2BrNO2, K2CO3, ACN;
(b) BH3 THF, THF, reflux; (c) Zn, 1 N HCl, i-PrOH; (d) Boc2O,
*
TEA, DCM; (e) H2, 20% Pd(OH)2/C, MeOH; (f) 2-fluoropyridine,
TEA, 90 ꢁC; (g) 50% TFA/DCM.
11.42
T: + c ESI Full ms2 175.00@30.00 [ 50.00-200.00]
100
157.2
100
90
90
b
a
80
80
70
60
50
40
30
147.2
70
60
50
175.1
40
30
20
10
20
10
0
10.81
10.34
11.66
14.97
15
15.33
16.91
3.93
9.31
3.02
130.0
146.6
140
147.8
172.9
176.1
180
19.52 23.47
20
0
100
120
160
200
0
5
10
m/z
Time (min)
3.66
3.68
100
90
80
70
60
50
40
30
20
10
0
T: + c ESI Full ms2 193.00@30.00 [ 50.00-250.00]
149.0
100
90
80
70
60
50
40
30
20
10
0
175.1
c
d
137.1
3.84
4.00
4.69
193.0
121.2
162.9
11.10
10.60
14.81
16.47
3.03
17.66
6.08
1.74
50
100
150
m/z
200
250
0
5
10
Time (min)
15
T: + c ESI Full ms3 193.00@30.00 175.00@40.00 [ 50.00-250.00]
147.2
100
157.2
90
80
70
60
50
40
30
20
10
0
e
145.4
140
107.0
157.8
160
100
120
180
200
220
240
m/z
Figure 2. LC chromatogram and mass spectrum of oxidized products by K3Fe(CN)6with MH+ at 175 and 193. (a) LC chromatogram of 175 from
oxidation by K3Fe(CN)6 for 5 h. (b) MS/MS spectrum of 175. (c) LC chromatogram of 193 from oxidation by K3Fe(CN)6 for 10 h. (d) MS/MS
spectrum of 193. (e) MS3 spectrum of 175 from MS2 of 193.

6684
Q. Sun et al. / Bioorg. Med. Chem. Lett. 17 (2007) 6682–6686
alyze the oxidation of primary, secondary, and tertiary
amines. The models oxidized primary amines, but not
tertiary amine substrates.16,17 Considering the fact that
there is still some uncertainty about the mechanism of
oxidation by MAOs, the known mimics are not suitable
as oxidation models for mimicking the bioactivation of
MAOs.
N
N
+
+
;
M+H-H2O
m/z 157
M+H-CO
m/z 147
+H+
H
O
m/z 175
11
+
N
N
+
+
HO
H
;
M+H-H2O
m/z 175
;
M+H-HCHO
m/z 163
M+H-OH-CHO+2H
m/z 149
+H+
O
m/z 193
The synthesis of 2 is illustrated in Scheme 1. The exo-
nitrocyclopropane derivative 4 was synthesized by reac-
tion of N-benzylmaleimide 3 with bromonitromethane
in the presence of a base. Earlier trials of the same trans-
formation with a variety of bases gave yields lower than
15%.18 A modified method using 1,2-dimethyl-1,4,5,6-
tetrahydropyrimidine (DMTHP) improved the yield to
30–35%. However, as reported by the authors, a draw-
back of the reaction is the tar formation due to the
undesired side reaction of the base with both the N-ben-
zylmaleimide and the bromonitromethane.19 Ballini
et al.20 increased the yield of this reaction to 70% by
using potassium carbonate as base and acetonitrile as
solvent. Addition of 1.5 equivalent of bromonitrome-
thane in portions over 20 h was claimed to be essential
for a good yield. In our experiment, 1.2 equivalent of
bromonitromethane was added dropwise over 5 h, and
a 67% yield was obtained. Imide carbonyl groups of 4
were then reduced by boraneÆTHF complex to give 5.
Cyclopropyl nitro group of 5 was subsequently reduced
by Zn and 1 N HCl (aq) with i-PrOH as co-solvent to
produce the primary cyclopropylamine 6.21 The primary
amine was protected by Boc to give 7, and 7 was then
debenzylated to 8 by hydrogenation, using Pd(OH)2/C
as catalyst. Selectively protected secondary amine 8
was then coupled with 2-fluoropyridine to generate 9.
Boc deprotection of 9 provided the drug model of trova-
floxacin in TFA salt form 10. The stereochemistry of
cyclopropylamine has been reported to favor the exo-
form to a very significant extent and the exo/endo ratio
is approximately 98:2.22
12
+
+
+
;
175-H2O
m/z 157
;
175-CO-2H
m/z 145
175-CO
m/z 147
Figure 3. Proposed principal mass spectral fragments of 175 and 193.
tion was scaled up, providing ample amounts of pure 11,
which were isolated for H and 13C NMR analysis. The
1
NMR assignment of 11, as summarized in Table 1, con-
firmed the structure proposed based on MS/MS analy-
sis. These results demonstrated that K3Fe(CN)6 can
oxidize the drug model 2 to a reactive a,b-unsaturated
aldehyde 11, suggesting that enzymes with heme-iron
center, such as P450s, may bioactivate the cyclopropyl-
amine moiety of trovafloxacin and possibly be responsi-
ble for the toxicity.
Chemical oxidation of drug model by NaClO was also
performed. To date, there has been no reported oxida-
tion of cyclopropylamines by MPO, although large
numbers of normal aliphatic and aromatic amines are
good substrates of MPO. In that scenario, identifying
the possible reactive intermediate(s) from oxidation of
cyclopropylamine by MPO is crucial in understanding
the influence of neutrophils on the toxicity of trovaflox-
acin in the animal model.14 Chemical oxidation by Na-
ClO was carried out in the same way as K3Fe(CN)6
oxidation except 430 ll NaClO (aq) containing 0.5%
chlorine (1 M equiv) was added instead of K3Fe(CN)6.
Chemical oxidation of the drug model by K3Fe(CN)6
was performed as follows. Ten milligrams of compound
9 was deprotected prior to use by 50% TFA/CH2Cl2 to
give 3.64 · 10À5 mol drug model in TFA salt form.
The dried residual oil was dissolved in 300 ll MeOH
and 400 ll H2O. The pH was adjusted to 9 by 150 ll
of 1 M NaOH and 24 mg K3Fe(CN)6 (2 equiv) was
added. After 5 h, a new peak from LC/MS analysis23
was observed with tR = 11.42 min (Fig. 2a) and MH+
at m/z 175. MS/MS analysis showed that 175 gave rise
to typical aldehyde fragments at 157 (loss of H2O) and
147 (loss of CO) (Fig. 2b). After the oxidation was run
for 10 h, another product with tR = 3.68 min (Fig. 2c)
and MH+ at m/z 193 appeared, which was not observed
at 5 h. Fragmentation of 193 generated product peaks at
175, 163, 149, 137, and 121 (Fig. 2d). Further fragmen-
tation of product peak 175 gave peaks at 157, 147, 145,
and 107 (Fig. 2e). The proposed structures for 175 and
193 are shown in Figure 3 and the principal mass spec-
tral fragments were also assigned. Compound 12 can be
formed by adding water to the a,b-unsaturated aldehyde
11 via Michael addition. To confirm the structure of the
reactive a,b-unsaturated aldehyde 11, K3Fe(CN)6 oxida-
Table 1. 1H NMR and 13C NMR (CDCl3) assignments for reactive
a,b-unsaturated aldehyde 11
3
2
1
4
6
5
N
N
7
9
8
H10
11
O
a
Position
dC
dH (multiplicities,b J (Hz))
a
1
2
148.5
112.6
137.7
106.4
8.12 (d, 5.1)
6.56 (dd, 5.1, 7.2)
7.44 (m)
3
4
6.34 (d, 8.7)
5
6
54.8
145.1
142.8
51.0
4.54 (m)
6.96 (m)
7
8
9
10
4.34 (m)
9.77 (s)
187.7
a Chemical shifts in ppm.
b Notations: dd, doublet of doublets; m, multiplet; d, doublet; s,
singlet.

Q. Sun et al. / Bioorg. Med. Chem. Lett. 17 (2007) 6682–6686
6685
11.43
100
90
80
70
60
50
40
30
20
10
0
T: + c ESI Full ms2 175.00@30.00 [ 50.00-200.00]
157.1
100
90
175.1
80
147.1
70
60
50
40
30
20
10
0
10.98
10.60
10.07
11.63
145.2
140
4.11
158.2
133.3
9.73
14.16
15
107.2
3.91
7.56
119.2
120
15.90
148.0
175.9
180
100
160
200
0
5
10
20
m/z
Time (min)
Figure 4. Left panel: LC chromatogram with mass filter m/z 175 showing peak (tR = 11.43 min) associated with oxidized product 11 by NaClO,
which has same retention time as Figure 2a. Right panel: MS/MS spectrum of oxidized product 11 by NaClO, which yields the same fragment ions as
Figure 2b.
The LC/MS analysis23 showed that an oxidized product
with MH+ at 175 was also generated by NaClO oxida-
tion. The product showed the same retention time and
fragment ions (Fig. 4) as 11 from K3Fe(CN)6 oxidation,
which means both NaClO and K3Fe(CN)6 can oxidize
the drug model to the reactive a,b-unsaturated aldehyde
11. This suggests that MPO is also possibly responsible
for the hepatotoxicity of trovafloxacin.
in vitro enzymatic metabolism of the drug model and
also trovafloxacin will be reported separately.
Acknowledgment
The authors thank University of Virginia Department of
Chemistry for the support.
The mechanism of idiosyncratic hepatotoxicity of trova-
floxacin has intrigued both academia and industry for
more than 8 years. There are several oxidizable positions
that can be bioactivated and lead to the toxicity of
trovafloxacin. To simplify the problem, a drug model
of trovafloxacin containing the cyclopropylamine sub-
structure was synthesized and the chemical oxidations
by K3Fe(CN)6 and NaClO were performed. The results
presented here showed that both K3Fe(CN)6 and Na-
ClO can oxidize the drug model to a reactive a,b-unsat-
urated aldehyde 11, implicating that heme-iron centered
enzymes such as P450s and MPO may possibly be
responsible enzymes for the hepatotoxicity of trovaflox-
acin. Chemical oxidation provides an economic and effi-
cient way of probing the possible reactions of
cyclopropylamine and is important for helping to
choose the expensive biological enzymes for in vitro
metabolism research. Considering the limited amount
of metabolite generated from enzyme metabolism, full
structure elucidation of the reactive metabolite or its
conjugate by NMR is typically challenging and expen-
sive as a result of needing to isolate the product from
enzyme metabolism mixtures. In addition, the total
synthesis of the a,b-unsaturated aldehyde 11 is also
not readily available. Chemical oxidation proved to be
a good way to provide a standard compound for struc-
ture elucidation of in vitro enzyme metabolites. Under
the guidance of chemical oxidation results, especially
with the useful information provided by the chemical
oxidized product—a,b-unsaturated aldehyde 11, our
subsequent in vitro enzyme metabolism studies of the
drug model have revealed that one P450 enzyme and
also MPO are able to oxidize the drug model to the reac-
tive a,b-unsaturated aldehyde 11 and may be the respon-
sible enzymes for the toxicity. The detailed results for
Supplementary data
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/j.bmcl.
2007.10.070.
References and notes
1. Brighty, K. E.; Gootz, T. D. J. Antimicrob. Chemother.
1997, 39, 1.
2. US Food and Drug Administration Website. Public
Health Advisory. Trovan (trovafloxacin/alatroflocacin
mesylate). June 9, 1999. Available at: http://www.fda.
gov/cder/news/trovan.
3. Ball, P.; Mandell, L.; Niki, Y.; Tillotson, G. Drug Safety
1999, 21, 407.
4. Uetrecht, J. P. Chem. Res. Toxicol. 1999, 12, 387.
5. Macdonald, T. L.; Zirvi, K.; Burka, L. T.; Peyman, P.;
Guengerich, F. P. J. Am. Chem. Soc. 1982, 104, 2050.
6. Guengerich, F. P.; Macdonald, T. L. Acc. Chem. Res.
1984, 17, 9.
7. Bondon, A.; Macdonald, T. L.; Harris, T. M.; Guenge-
rich, F. P. J. Biol. Chem. 1989, 264, 1988.
8. Cerny, M. A.; Hanzlik, R. P. J. Am. Chem. Soc. 2006, 128,
3346.
9. Silverman, R. B.; Hiebert, C. K. Biochemistry 1988, 27,
8448.
10. Silverman, R. B. Acc. Chem. Res. 1995, 28, 335.
11. Zhong, B.; Silverman, R. B. J. Am. Chem. Soc. 1997, 119,
6690.
12. Vintem, A. P.; Price, N. T.; Silverman, R. B.; Ramsay, R.
R. Bioorg. Med. Chem. 2005, 13, 3487.
13. Sayre, L. M.; Naismith, R. T.; Bada, M. A.; Li, W.; Klein,
M. E.; Tennant, M. D. Biochim. Biophys. Acta 1996, 1296,
250.

6686
Q. Sun et al. / Bioorg. Med. Chem. Lett. 17 (2007) 6682–6686
14. Ball, P.; Ed, F. Curr. Ther. Res. 2003, 64, 646.
15. Waring, J. F.; Liguori, M. J.; Luyendyk, J. P.; Maddox, J.
F.; Ganey, P. E.; Stachlewitz, R. F.; North, C.; Blomme,
E. A.; Roth, R. A. J. Pharmacol. Exp. Ther. 2006, 316,
1080.
21. Wurz, R. P.; Charette, A. B. J. Org. Chem. 2004, 69, 1262.
22. Norris, T.; Braish, T. F.; Butters, M.; DeVries, K. M.;
Hawkins, J. M.; Massett, S. S.; Rose, P. R.; Santafianos,
D.; Sklavounos, C. J. Chem. Soc., Perkin Trans. 1 2000,
1615.
16. Kim, J.-M.; Bogdan, M. A.; Mariano, P. S. J. Am. Chem.
Soc. 1993, 115, 10591.
17. Kim, J.-M.; Hoegy, S. E.; Mariano, P. S. J. Am. Chem.
Soc. 1995, 117, 100.
18. Braish, T. F. U.S. Patent 5,256,791, 1994.
19. Braish, T. F.; Castaldi, M.; Chan, S.; Fox, D. E.; Keltonic,
T.; McGarry, J.; Hawkins, J. M.; Norris, T.; Rose, P. R.;
Sieser, J. E.; Sitter, B. J.; Watson, H., Jr. Synlett 1996,
1996, 1100.
23. The LC/MS and LC/MS/MS analyses were conducted on
a Thermo-Finnigan LCQ classic ion trap mass spectrom-
eter with an electrospray ionization source and a Waters
2690 Separations Module using a Phenomenex Gemini
C18 (2.0 mm · 250 mm) column. The mobile phase con-
sisted of 0.03% aqueous NH4OH (solvent A) and 0.03%
NH4OH in acetonitrile (solvent B), and the gradient was
as follows: 0 min, 2% B; 4 min, 40% B; 4.5 min, 80% B;
10 min, 100% B; 15 min, 100% B; 16 min, 2% B; and
25 min, 2% B.
20. Ballini, R.; Fiorini, D.; Palmieri, A. Synlett 2003, 11, 1704.