Studies on the metabolism of troglitazone to reactive intermediates in vitro and in vivo. Evidence for novel biotransformation pathways involving quinone methide formation and thiazolidinedione ring scission

Article abstract of DOI:10.1021/tx000180q

Therapy with the oral antidiabetic agent troglitazone (Rezulin) has been associated with cases of severe hepatotoxicity and drug-induced liver failure, which led to the recent withdrawal of the product from the U.S. market. While the mechanism of this toxicity remains unknown, it is possible that chemically reactive metabolites of the drug play a causative role. In an effort to address this possibility, this study was undertaken to determine whether troglitazone undergoes metabolism in human liver microsomal preparations to electrophilic intermediates. Following incubation of troglitazone with human liver microsomes and with cDNA-expressed cytochrome P450 isoforms in the presence of glutathione (GSH), a total of five GSH conjugates (M1-M5) were detected and identified tentatively by LC-MS/MS analysis. In two cases (M1 and M5), the structures of the adducts were confirmed by NMR spectroscopy and/or by comparison with an authentic standard prepared by synthesis. The formation of GSH conjugates M1-M5 revealed the operation of two distinct metabolic activation pathways for troglitazone, one of which involves oxidation of the substituted chromane ring system to a reactive o-quinone methide derivative, while the second involves a novel oxidative cleavage of the thiazolidinedione (TZD) ring, potentially generating highly electrophilic α-ketoisocyanate and sulfenic acid intermediates. When troglitazone was administered orally to a rat, samples of bile were found to contain GSH conjugates which reflected the operation of these same metabolic pathways in vivo. The finding that metabolism of the TZD ring of troglitazone was catalyzed selectively by P450 3A enzymes is significant in light of the recent report that troglitazone is an inducer of this isoform in human hepatocytes. The implications of these results are discussed in the context of the potential for troglitazone to covalently modify hepatic proteins and to cause oxidative stress through redox cycling processes, either of which may play a role in drug-induced liver injury.

Full text of DOI:10.1021/tx000180q

62
Chem. Res. Toxicol. 2001, 14, 62-70
Stu d ies on th e Meta bolism of Tr oglita zon e to Rea ctive
In ter m ed ia tes in Vitr o a n d in Vivo. Evid en ce for Novel
Biotr a n sfor m a tion P a th w a ys In volvin g Qu in on e Meth id e
F or m a tion a n d Th ia zolid in ed ion e Rin g Scission †
Kelem Kassahun,*,‡ Paul G. Pearson,‡ Wei Tang,§ Ian McIntosh,‡ Kwan Leung,§
Charles Elmore,§ Dennis Dean,§ Regina Wang,§ George Doss,§ and
Thomas A. Baillie‡
Department of Drug Metabolism, Merck Research Laboratories, West Point, Pennsylvania 19486,
and Rahway, New J ersey 07065
Received August 17, 2000
Therapy with the oral antidiabetic agent troglitazone (Rezulin) has been associated with
cases of severe hepatotoxicity and drug-induced liver failure, which led to the recent withdrawal
of the product from the U.S. market. While the mechanism of this toxicity remains unknown,
it is possible that chemically reactive metabolites of the drug play a causative role. In an effort
to address this possibility, this study was undertaken to determine whether troglitazone
undergoes metabolism in human liver microsomal preparations to electrophilic intermediates.
Following incubation of troglitazone with human liver microsomes and with cDNA-expressed
cytochrome P450 isoforms in the presence of glutathione (GSH), a total of five GSH conjugates
(M1-M5) were detected and identified tentatively by LC-MS/MS analysis. In two cases (M1
and M5), the structures of the adducts were confirmed by NMR spectroscopy and/or by
comparison with an authentic standard prepared by synthesis. The formation of GSH conjugates
M1-M5 revealed the operation of two distinct metabolic activation pathways for troglitazone,
one of which involves oxidation of the substituted chromane ring system to a reactive o-quinone
methide derivative, while the second involves a novel oxidative cleavage of the thiazolidinedione
(TZD) ring, potentially generating highly electrophilic R-ketoisocyanate and sulfenic acid
intermediates. When troglitazone was administered orally to a rat, samples of bile were found
to contain GSH conjugates which reflected the operation of these same metabolic pathways in
vivo. The finding that metabolism of the TZD ring of troglitazone was catalyzed selectively by
P450 3A enzymes is significant in light of the recent report that troglitazone is an inducer of
this isoform in human hepatocytes. The implications of these results are discussed in the context
of the potential for troglitazone to covalently modify hepatic proteins and to cause oxidative
stress through redox cycling processes, either of which may play a role in drug-induced liver
injury.
In tr od u ction
Troglitazone (Rezulin; Figure 1), an oral antidiabetic
drug belonging to the 2,4-thiazolidinedione (TZD)1 struc-
tural class, acts by decreasing tissue resistance to insulin
via agonism of the PPAR receptor (1). Thus, troglitazone
enhances insulin-dependent glucose metabolism in muscle
and adipose tissue and inhibits hepatic gluconeogenesis.
However, despite its effectiveness as an antidiabetic
agent, use of troglitazone has been associated with severe
hepatotoxicity which, in some cases, has led to hepatic
† A preliminary account of this work was presented at the Sixth
International Symposium on Biological Reactive Intermediates, J uly
2000, Paris, France.
F igu r e 1. Structures of troglitazone and GSH conjugates
M1-M5.
* To whom correspondence should be addressed: Department of
Drug Metabolism, WP75A-203, Merck Research Laboratories, West
Point, PA 19486. E-mail: kelem_kassahun@merck.com.
‡ Department of Drug Metabolism, Merck Research Laboratories,
West Point, PA.
failure and death (2-4). As a consequence of safety
concerns over the use of this agent, troglitazone was
withdrawn from the U.S. market in March 2000.
In the course of our studies on the interaction of several
TZD derivatives with cytochrome P450 enzymes in vitro,
we found that troglitazone caused preincubation- and
§ Department of Drug Metabolism, Merck Research Laboratories,
Rahway, NJ .
1
Abbreviations: TZD, thiazolidinedione; PPAR, peroxisome prolif-
erator activated receptor; P450, cytochrome P450; GSH, glutathione;
DTT, dithiothreitol; CID, collision-induced dissociation.
10.1021/tx000180q CCC: $20.00 © 2001 American Chemical Society
Published on Web 12/08/2000

Metabolism of Troglitazone to Reactive Intermediates
Chem. Res. Toxicol., Vol. 14, No. 1, 2001 63
F igu r e 2. Scheme for the synthesis of GSH conjugate M1.
stirred at reflux for 1.5 h. The solution then was concentrated
to dryness at reduced pressure to afford 99 mg of 2 as a white
solid: 1H NMR (CDCl3) δ 1.42 (s, 3H), 1.88 (m, 2H), 1.99 (s,
3H), 2.03 (s, 3H), 2.09 (s, 3H), 2.14 (m, 2H), 2.34 (s, 3H), 2.63
(m, 2H), 3.15 (dd, 1H, J ) 7.3, 14.4 Hz), 3.37 (dd, 1H, J ) 4.3,
14.4 Hz), 3.87 (d, 1H, J ) 9.1 Hz), 3.98 (d, 1H, J ) 9.0 Hz), 4.47
(dd, 1H, J ) 4.5, 7.0 Hz), 5.91 (br s, 1H), 6.36 (br s, 1H), 6.87
(d, 2H, J ) 8.5 Hz), 7.17 (d, 2H, J ) 8.4 Hz); 13C NMR (CDCl3)
δ 11.8, 12.1, 12.9, 20.1, 20.5, 22.8, 28.2, 40.3, 61.0, 72.4 (br),
74.5, 114.5, 117.3, 123.1, 125.0, 127.0, 128.2, 130.6, 140.8, 148.9,
158.2, 169.7, 170.9; LC-MS (tR ) 15.53 min) m/z (relative
intensity) [M + H]+ 460.2 (100), [M + NH4]+ 477.2 (25), [M +
Na]+ 482.2 (20).
NADPH-dependent inhibition of P450 3A4 activity in
human liver microsomal preparations. This finding raised
the possibility that troglitazone undergoes P450 3A4-
mediated conversion to a chemically reactive intermedi-
ate(s), and that such a species may play a role in
troglitazone-induced hepatotoxicity. The studies reported
in this communication provide evidence that troglitazone
does indeed undergo metabolic activation, both in vitro
and in vivo, and point to both the TZD ring and the
substituted chromane moiety as sites for reactive inter-
mediate formation through cytochrome P450 catalysis.
2-Ch lor o-3-{4-[(6-h yd r oxy-2,5,7,8-tetr a m eth ylch r om a n -
2-yl)m eth oxy]p h en yl}p r op a n a m id e (3). A solution of 99 mg
(0.21 mmol) of 2 in 19 mL of MeOH was stirred at 0 °C under
N2 as 238 mg (9.9 mmol) of NaH was added in four portions
over 30 min. After the reaction mixture had been stirred at room
temperature for 2 h, the solution was concentrated to 5 mL
under a stream of N2, diluted with 10 mL of water, and extracted
four times with 10 mL of CH2Cl2. The organic layer was dried
(MgSO4), filtered, and concentrated to give 72 mg of 3 as a pale
yellow solid: 1H NMR (CDCl3) δ 1.41 (s, 3H), 1.89 (m, 2H), 2.11
(s, 3H), 2.12 (s, 3H), 2.12 (m, 2H), 2.17 (s, 3H), 2.64 (m, 2H),
3.15 (dd, 1H, J ) 7.9, 14.4 Hz), 3.37 (dd, 1H, J ) 4.5, 14.4 Hz),
3.87 (d, 1H, J ) 9.1 Hz), 3.96 (d, 1H, J ) 9.1 Hz), 4.47 (dd, 1H,
J ) 4.5, 7.9 Hz), 4.53 (bs s, 1H), 5.97 (bs s, 1H), 6.39 (br s, 1H),
6.87 (d, 2H, J ) 8.6 Hz), 7.16 (d, 2H, J ) 8.6 Hz); 13C NMR
(CDCl3) δ 11.3, 11.8, 12.2, 20.3, 22.7, 28.7, 40.3, 61.0, 72.5, 76.7,
114.5, 115.2, 117.3, 18.7, 121.4, 122.6, 128.1, 130.6, 145.0, 158.3,
171.1; LC-MS (tR ) 13.81 min) m/z (relative intensity) [M +
H]+ 418.2 (100), [M + NH4]+ 435.2 (30), [M + Na]+ 440.2 (25).
Exp er im en ta l P r oced u r es
Ch em ica ls. Troglitazone was extracted from Rezulin tablets
using acetonitrile, while the quinone metabolite of troglitazone
was synthesized at Merck Research Laboratories (5). GSH,
NADPH, and DTT were purchased from Sigma Chemical Co.
(St. Louis, MO), while ketoconazole and sulfaphenazole were
obtained from Research Biologics Inc. (Natick, MA). Quinidine
and 13-cis-retinoic acid were supplied by Aldrich Chemical Co.
(Milwaukee, WI), and fluvoxamine was obtained from the Merck
Compound Repository.
Syn th esis. The disulfide-linked GSH conjugate of troglita-
zone (M1) was synthesized as outlined in Figure 2 using the
procedure described below. 1H and 13C NMR spectra were
recorded on a Varian U-400 spectrometer and are referenced
to the residual solvent peak (7.26 and 77.0 ppm for CHCl3). LC-
MS data were recorded on an HP 1100 MSD instrument with a
gradient elution from 5 to 95% acetonitrile and 0.1% aqueous
formic acid on a Zorbax XDB C8 column.
2-Ch lor o-3-{4-[(6-a cetoxy-2,5,7,8-tetr a m eth ylch r om a n -
2-yl)m eth oxy]p h en yl}p r op a n a m id e (2). A solution of 104 mg
(0.213 mmol) of ethyl 3-{4-[(6-acetoxy-2,5,7,8-tetramethylchro-
man-2-yl)methoxy]phenyl}-2-chloropropionate (1), available in
five steps from 6-hydroxy-2,5,7,8-tetramethylchroman-2-car-
boxylic acid (6), in 10 mL of MeOH and 20 mL of NH3 was
2-Am in o-4-(N-{1-[N-(2,2-d ih yd r oxyeth yl)ca r ba m oyl]-2-
[(1-car bam oyl-2-{4-[(6-h ydr oxy-2,5,7,8-tetr am eth ylch r om an -
2-yl)m eth oxy]p h en yl}eth yl)d isu lfa n yl]eth yl}ca r ba m oyl)-
bu ta n oic Acid (M1). A modification of the procedure reported
by Hu and Fox was followed (7). A solution of 5 mg (12 µmol) of
chloride 3 in 1 mL of THF was stirred and cooled at -5 °C as

64 Chem. Res. Toxicol., Vol. 14, No. 1, 2001
Kassahun et al.
Ta ble 1. 1H NMR Da ta for Syn th etic M1 a n d for M5
Isola ted fr om Liver Micr osom a l In cu ba tion s of
Tr oglita zon e
20 µL (95 µmol) of hexamethyldisilathiane and 80 µL (80 µmol)
of 1 M tetrabutylammonium fluoride in THF were added in two
aliquots over the course of 1 h. A third aliquot of 20 µL (95 mol)
of hexamethyldisilathiane and 80 µL (80 µmol) of 1 M tetrabu-
tylammonium fluoride in THF were added, and the reaction
mixture was heated to 40 °C for 1 h. HPLC analysis showed
that the starting material is consumed. The solution was cooled
to room temperature and was diluted with 5 mL of saturated
aqueous NaCl solution. The resulting solution was extracted
with 10 mL of Et2O, and the organic phase was washed twice
with 3 mL of water, concentrated to approximately 0.2 mL,
diluted with 10 mL of THF, and concentrated to 1 mL. A solution
of 10 mg of NaHCO3 in 1 mL of water was added to the THF
solution followed by 10 mg (22 µmol) of glutathione thiosulfonate
4 (8). After 12 h at room temperature, the reaction was
incomplete as judged by HPLC; 20 mg of 4 was added, and the
reaction mixture was heated to 40 °C for 4 h. After 4 h, an
additional 30 mg of 4 was added and the heating continued for
4 h. HPLC analysis showed a complex reaction mixture that
contained two components with an MH+ ion at m/z 721. The
solution was concentrated to near dryness and taken up in 1
mL of THF and 0.5 mL of 0.1% trifluoroacetic acid. Purification
was effected by semipreparative HPLC (YMC ODS-AM column,
flow rate of 5 mL/min, isocratic at 30% acetonitrile and 0.1%
trifluoroacetic acid for 25 min and then 35% acetonitrile and
0.1% trifluoroacetic acid, UV at 230 nm). The column fractions
were analyzed by HPLC and LC-MS; the fraction containing
the desired product (tR ) 56.3-58.1 min) was concentrated to
remove the acetonitrile under reduced pressure and was lyoph-
ilized to give 0.5 mg of 1 that was 80% pure (determined by
HPLC, via the YMC ODS-AM column, 35% acetonitrile and 0.1%
trifluoroacetic acid, 1 mL/min, tR ) 16.0 and 16.4 min): MS m/z
721 (MH+). For the 1H NMR data, see Table 1.
chemical shift
(ppm)
chemical shift
(ppm)
assignmenta
assignmentb
a
1.25 (s)
1.76 (m)
1.78 (m)
1.89 (m)
1.92 (s)
1.94 (m)
1.98 (s)
2.00 (s)
a
1.33 (s)
1.85 (m)
1.98 (s)
2.03 (m)
2.07 (s)
2.46 (m)
2.70 (m)
b
b+d
c
Gluâ
e
Gluγ
g
Cysâ
Gluâ
Gluâ
c
d
e
f
2.91 (dd, J ) 5.4,
4.2 Hz)
Gluγ
2.26 (m)
h
3.08 (dd, J ) 9.1,
14.1 Hz)
Gluγ
g
2.31 (m)
2.52 (t,
i
k
3.32 (m)
3.95 (AB)
J ) 6.6 Hz)
2.79 (m)
h
CysR
4.48 (dd, J ) 5.4,
8.6 Hz)
Cysâ
2.83 (m)
j
4.58 (dd, J ) 4.3,
8.6 Hz)
i
2.98 (m)
3.19 (m)
3.22 (m)
m
l
6.86 (d, J ) 8.6 Hz)
7.12 (d, J ) 8.6 Hz)
Cysâ
GluR
j
3.51 (m)
GlyR
3.61 (m)
3.87 (AB)
4.47 (m)
M5 was isolated by HPLC (using a Zorbax XDB-C18 4.6 mm
× 250 mm column and a gradient containing 0.1% aqueous
formic acid and acetonitrile) following incubation of troglitazone
(100 µM) with rat liver microsomes (2 mg/mL) in the presence
of GSH (5 mM) in a total incubation volume of 50 mL: MS m/z
747 (MH+). For the 1H NMR data, see Table 1.
k
CysR
l
6.82 (d,
J ) 8.4 Hz)
7.05 (d,
m
J ) 8.4 Hz)
7.16, 6.61 (br s)
8.46 (br d)
8.66 (br)
NH2
Cys NH
Gly NH
Liver Micr osom a l P r ep a r a tion s. A human liver microso-
mal preparation (from a pool of 15 livers) was purchased from
Xenotech LLC (Kansas City, KS). The recombinant P450
enzymes and FMO3 that were used were from baculovirus-
infected insect cells and were either prepared in-house or
purchased from Gentest Corp. (Woburn, MA).
a
b
Obtained at 600 MHz in DMSO-d6. Obtained at 500 MHz in
CD3CN/17% D2O.
removed and evaporated to dryness. The residue was reconsti-
tuted in 30% acetonitrile in water (200 µL), vortex-mixed, and
centrifuged. Aliquots (10 µL) of the final supernatant were
analyzed by LC-MS and LC-MS/MS.
In cu b a t ion s w it h Micr osom es Con t a in in g cDNA-
Exp r essed En zym es. Incubations with recombinant enzymes
were conducted as described above for native liver microsomes.
The expressed enzymes that were used, and the quantities
added to incubation mixtures (final volume of 1 mL), were as
follows: 300 pmol of P450 1A2, 300 pmol of P450 2A6, 500 pmol
of P450 2C8, 400 pmol of P450 2C9, 200 pmol of P450 2C19,
300 pmol of P450 2D6, 500 pmol of P450 3A4, and 200 µL of
FMO3 from a preparation having a protein concentration of 5
mg/mL.
Effect of Tr oglita zon e on Hu m a n Liver Micr osom a l
Testoster on e 6â-Hyd r oxyla se Activity. For non-preincuba-
tion-dependent enzyme inhibition, a pooled human liver mi-
crosomal preparation was incubated with testosterone (250 µM)
and troglitazone at concentrations ranging from 0.1 to 50 µM
at 37 °C for 10 min in the presence of an NADPH-generating
system. For preincubation-dependent inhibition, microsomes
were incubated with troglitazone at concentrations varying from
0.1 to 50 µM in the presence of an NADPH-generating system
at 37 °C for 30 min, after which incubation with testosterone
(250 µM) was carried out for 10 min. To determine if the
preincubation-dependent inhibition also was NADPH-depend-
ent, a third incubation in which the NADPH-generating system
was absent during the preincubation period was carried out.
In h ibition Stu d ies. The effect of P450 form-selective chemi-
cal inhibitors on the formation of troglitazone metabolites in
human liver microsomes was investigated using the following
inhibitors at the indicated concentration: fluvoxamine (10 µM;
P450 1A2), ketoconazole (2 µM; P450 3A), sulfaphenazole (5 µM;
P450 2C9), quinidine (10 µM; P450 2D6), and 13-cis-retinoic acid
(150 µM; P450 2C8). The inhibitors, which were dissolved in
50% acetonitrile in water, were incubated individually with
human liver microsomes at troglitazone concentrations of 5 and
50 µM.
In cu ba tion s w ith Hu m a n Liver Micr osom es. Troglita-
zone and its quinone metabolite were dissolved in methanol for
addition in incubation media such that the volume of methanol
in the final incubation mixture was e1%. Incubations were
carried out at 37 °C for 60 min in a shaking water bath. The
incubation volume was 1 mL and consisted of the following: 0.1
M potassium phosphate buffer (pH 7.4), magnesium chloride
(1 mM), microsomal protein (2 mg), substrate (50 µM), and
NADPH (1 mM). GSH (5 mM) was added 3 min after initiation
of the reaction with NADPH. Incubations that lacked either
NADPH or GSH served as negative controls, and reactions were
terminated by the addition of ice-cold acetonitrile (1 mL). After
centrifugation, the supernatant from each incubation was
Red u ction of Disu lfid e-Lin k ed GSH Con ju ga tes. GSH
adducts generated by human liver microsomes or recombinant
P450 3A4 were treated with DTT to explore the nature of the

Metabolism of Troglitazone to Reactive Intermediates
Chem. Res. Toxicol., Vol. 14, No. 1, 2001 65
F igu r e 3. Preincubation-dependent inhibition of P450 3A
activity (testosterone 6â-hydroxylase) in human liver microso-
mal preparations incubated with troglitazone.
linkage between the drug residue and GSH. Thus, following
incubation with microsomal preparations, termination of the
reaction, and evaporation of the supernatant as outlined above,
the residue was dissolved in 30% acetonitrile in water and
treated with a solution of DTT (100 µL of a 20 mM solution
in phosphate buffer). The resulting mixture was held at 37 °C
for 60 min, and the products were analyzed directly by
LC-MS/MS.
F igu r e 4. LC-MS/MS chromatogram (A) and product ion
spectrum (B) obtained by CID of the MH+ ion (m/z 721) of M1
detected (retention time of 21.1 min) in human liver microsomal
extracts. The origins of characteristic ions are as indicated.
and NADPH-dependent inhibitor of P450 3A activity in
human liver microsomes.
In Vivo Stu d ies. A fasted male Sprague-Dawley rat, which
had been cannulated at the common bile duct, was dosed
intraperitoneally with a solution of troglitazone in DMSO (100
mg/mL) such that the administered dose was 100 mg/kg. Bile
was collected over ice during the following time intervals: 0-1,
1-2, 2-4, 4-6, 6-8, and 8-24 h after administration. A sample
of pooled bile (0-8 h) was analyzed by LC-MS/MS for the
presence of GSH conjugates.
Id en tifica tion of Tr oglita zon e Meta bolites. LC-
MS/MS analysis of extracts of human liver microsomal
incubations containing troglitazone, NADPH, and GSH
revealed the presence of five GSH adducts (designated
M1-M5), none of which were evident in control incuba-
tions that lacked one or both cofactors. M1 exhibited an
MH+ ion at m/z 721 and yielded the product ion spectrum
shown in Figure 4B. A notable feature of this spectrum
was the presence of ions resulting from neutral losses
characteristic of GSH adducts (9), namely, m/z 646 [(MH
- 75)+] and 592 [(MH - 129)+]. When the conjugate was
treated with excess DTT, reaction occurred to yield a
product (MH+ at m/z 416) whose molecular mass was
consistent with an R-thiolamide derivative resulting from
TZD ring cleavage (data not shown). This observation
suggested that M1 was a disulfide-linked GSH conjugate,
with the structure depicted in Figure 4B. Subsequently,
an authentic sample of this conjugate was prepared by
chemical synthesis and, upon LC-MS/MS analysis,
yielded an HPLC retention time and product ion mass
spectrum identical to those obtained from the isolated
metabolite. Although several diastereomers of M1 are
possible, both the isolated metabolite and the authentic
standard eluted as a single chromatographic peak (Figure
4A) under the LC-MS conditions used in these studies.
M2 (which appeared to be a mixture of two isomers as
shown in Figure 5A) exhibited an MH+ ion at m/z 781,
the product ion spectrum of which (Figure 5B) contained
fragment ions at m/z 706 [(MH - 75)+], 652 [(MH -
129)+], and 308 [(GSH2)+]. On the basis of the molecular
mass of 780 Da and the appearance in the mass spectrum
of the weak, but structurally informative, ion at m/z 308,
it appears that this metabolite corresponds to the car-
bamate thioester GSH adduct of the putative isocyanate
derivative of troglitazone in which the sulfur atom of the
LC-MS/MS An a lysis. Metabolites were identified by elec-
trospray LC-MS and LC-MS/MS analysis using a Finnigan
LCQ mass spectrometer (San J ose, CA). The spray voltage was
kept at 4.1 kV, and the capillary temperature was set at 200
°C. Full scan spectra (from m/z 400 to 850) were obtained in
the positive ion mode, and product ion spectra were generated
by CID of MH+ ions of interest. Analytes were introduced into
the electrospray source via an Inertsil C18 column (2 mm × 150
mm, 5 µm; MetaChem Technologies Inc., Torrance, CA) at a flow
rate of 200 µL/min. Separation of analytes was achieved using
a gradient consisting of 0.1% aqueous formic acid and acetoni-
trile. The initial solvent composition was 10% acetonitrile and
90% aqueous, and after 2 min, the proportion of acetonitrile was
increased at a rate of 2 mL/min. A modification of the above
gradient (5% acetonitrile from 0 to 5 min, 15% acetonitrile at
10 min, and 80% acetonitrile at 30 min) was used for the
analysis of bile samples from troglitazone-treated rats.
Resu lts
Effect of Tr oglita zon e on Hu m a n Liver Micr oso-
m a l Testoster on e 6â-Hyd r oxyla se Activity. Figure 3
shows the effect of varying concentrations of troglitazone
on P450 3A-mediated testosterone 6â-hydroxylation in
human liver microsomes both with and without prein-
cubation of microsomes with troglitazone. The magnitude
of enzyme inhibition was higher when the experiment
was carried out with preincubation in the presence of
NADPH, indicating that troglitazone is a preincubation-

66 Chem. Res. Toxicol., Vol. 14, No. 1, 2001
Kassahun et al.
F igu r e 5. LC-MS/MS chromatogram (A) and product ion
spectrum (B) obtained by CID of the MH+ ion (m/z 781) of M2
detected (retention time of 16.5 min) in human liver microsomal
extracts. The origins of characteristic ions are as indicated.
F igu r e 6. LC-MS/MS chromatogram (A) and product ion
spectrum (B) obtained by CID of the MH+ ion (m/z 763) of M3
detected (retention time of 19.2 min) in human liver microsomal
extracts. The origins of characteristic ions are as indicated.
original TZD ring is in the form of a sulfinic acid (Figure
5B). In support of this tentative structural assignment,
the MS/MS behavior of M2 was consistent with that
exhibited by GSH conjugates of isocyanates and isothio-
cyanates (10-12). Metabolite M3, which displayed an
MH+ ion at m/z 763 and afforded a product ion spectrum
(Figure 6B) closely similar to that of M2, appeared to
differ from M2 by the elements of H2O. While the MS/
MS characteristics of M3 are consistent with the sulfine
structure proposed in Figure 6B, the identity of this
conjugate remains speculative.
M4 exhibited an MH+ ion at m/z 737 and prominent
fragments at m/z 719 [(MH - 18)+], 662 [(MH - 75)+],
and 608 [(MH - 129)+] (Figure 7B). This conjugate was
suspected to be related to the known quinone metabolite
of troglitazone since its molecular mass was 16 Da higher
than that of the adduct (M1) obtained from the parent
compound, and since the facile loss of water in the MS/
MS spectrum suggested the presence of a free -OH
group. This conclusion was supported by the observation
that incubations of the authentic quinone derivative with
human liver microsomes fortified with GSH and NADPH
resulted in a metabolite with an HPLC retention time
and an electrospray MS/MS spectrum identical to those
of M4. In addition, treatment of M4 with DTT yielded
the corresponding thiol derivative, as evidenced by LC-
MS/MS analysis (MH+ at m/z 432).
The protonated molecular ion of M5 appeared at m/z
747, the product ion spectrum of which exhibited the
characteristic losses of 75 and 129 Da (to give ions at
m/z 672 and 618, respectively) of a GSH adduct (Figure
8B). The MS/MS spectrum of this conjugate, however,
differed from that of the adducts described above in that
the spectrum contained an intense ion at m/z 308
[(GSH2)+]. This MS/MS behavior frequently is exhibited
by GSH adducts resulting from Michael addition reac-
tions, and corresponds formally to a retro-Michael cleav-
F igu r e 7. LC-MS/MS chromatogram (A) and product ion
spectrum (B) obtained by CID of the MH+ ion (m/z 737) of M4
detected (retention time of 19.5 min) in human liver microsomal
extracts. The origins of characteristic ions are as indicated.
age in the gas phase (13). On the basis of these MS/MS
data, the structure shown in Figure 8B, which would
result from addition of GSH to an o-quinone methide
derivative of troglitazone, is proposed for M5. Following
isolation of a specimen of M5, a proton NMR spectrum
was obtained which was fully consistent with this
structural hypothesis (Table 1).

Metabolism of Troglitazone to Reactive Intermediates
Chem. Res. Toxicol., Vol. 14, No. 1, 2001 67
F igu r e 10. LC-MS/MS analysis of extracts of bile from a rat
treated with troglitazone. Product ion spectrum of m/z 722 (B)
and the corresponding product ion chromatogram (A) indicating
the detection (retention time of 25.2 min) of the proposed
carboxyl derivative of M1 (MH+ at m/z 722).
F igu r e 8. LC-MS/MS chromatogram (A) and product ion
spectrum (B) obtained by CID of the MH+ ion (m/z 747) of M5
detected (retention time of 20.6 min) in human liver microsomal
extracts. The origins of characteristic ions are as indicated.
formation of M5 was found to be catalyzed by recombi-
nant P450 2C8, 2C9, 2C19, 2D6, and 3A4.
In Vivo Stu d y. A targeted search (based on LC-MS/
MS analysis) for M1 in the bile from troglitazone-treated
rats resulted not in the detection of M1 itself but of a
metabolite whose molecular mass (721 Da) was one mass
unit higher than that of M1 (Figure 10). This biliary
metabolite most likely corresponds to the carboxylic acid
derivative of M1, formed by hydrolysis of the primary
amide group of the latter. The fact that the MS/MS
spectrum (Figure 10B) of this in vivo conjugate was
almost identical to that of M1 (with the appropriate
fragment ions shifted 1 Da to higher masses), coupled
with the observation that the metabolite eluted at a later
HPLC retention time than M1 with the acidic mobile
phase used in this work, further supported the proposal
that the biliary metabolite was the free acid derivative
of M1. Conjugates M2, M3, and M5 also were detected
as metabolites of troglitazone in rat bile.
F igu r e 9. LC-MS/MS chromatogram obtained by CID of m/z
721 to detect M1 in extracts from a human liver microsomal
incubation containing troglitazone and ketoconazole. The arrow
indicates the trace amount of M1 formed in comparison to an
incubation without ketoconazole (Figure 4A).
En zym es In volved in th e F or m a tion of GSH
Ad d u cts of Tr oglita zon e. When troglitazone (5 or 50
µM) was incubated with human liver microsomes fortified
with NADPH and GSH, ketoconazole (2 µM) strongly
inhibited the formation of M1 (Figure 9), whereas other
P450 isoform-selective inhibitors had little effect. Simi-
larly, ketoconazole inhibited the formation of conjugates
M2 and M3 to a comparable extent. The formation of both
the quinone metabolite of troglitazone and its GSH
adduct (M4) also was inhibited only by ketoconazole.
Thus, while recombinant P450 1A2, 2C9, 2C19, 2D6, and
3A4 were able to catalyze the formation of M1 from
troglitazone, it may be concluded that P450 3A4 is the
dominant enzyme in this regard. None of the isoform-
selective chemical inhibitors suppressed the formation
of M5 by >30%, suggesting that multiple P450 isoforms
contribute to the generation of this product. Indeed, the
Discu ssion
The study presented here, which focused on the
biotransformation of troglitazone to reactive metabolites,
was prompted by our initial observation that the drug
inhibits P450 3A4 activity in vitro in an NADPH- and
preincubation-dependent fashion. In previous metabolism
studies, troglitazone has been shown to undergo biotrans-
formation via conjugation of the phenolic -OH moiety
to yield the sulfate and glucuronide derivatives, as well
as oxidation of the phenol to generate a p-benzoquinone
species (14, 15). In the current study, we employed GSH
as a nucleophilic trapping agent and identified two
distinct pathways of metabolic activation of troglitazone
in liver microsomal preparations, one of which involves

68 Chem. Res. Toxicol., Vol. 14, No. 1, 2001
Kassahun et al.
F igu r e 11. Proposed scheme for the metabolic activation of
troglitazone through P450 3A4-mediated oxidation of the TZD
ring. Intermediates I-III have not been identified, but are
proposed on the basis of the structures of GSH conjugates M1-
M3.
a novel oxidative scission of the TZD ring system (Figure
11), while the second entails oxidation of the substituted
chromane moiety to an o-quinone methide (Figure 12).
Three GSH adducts (M1-M3), whose formation in hu-
man liver microsomes was shown to be mediated by P450
3A enzymes, were detected as products of the former
pathway, and their genesis is rationalized by the scheme
shown in Figure 11. Thus, it is proposed that P450 3A-
mediated oxidation of the TZD sulfur atom affords a
reactive sulfoxide intermediate (I) which undergoes
spontaneous ring opening to a highly electrophilic R-keto
iosocyanate derivative (II). Hydrolysis of the isocyanate
followed by decarboxylation to the amide, accompanied
by attack of GSH on the reactive sulfenic acid, would
afford M1. Alternatively, conjugation of the isocyanate
with GSH to yield intermediate III, followed by oxidation
to the corresponding sulfinic acid derivative, provides a
route to M2. Although speculative at present, loss of the
elements of water from the sulfinic acid moiety of M2
may lead to the formation of the sulfine-containing GSH
conjugate, M3 (Figure 11). While these mechanisms
account satisfactorily for the formation of the observed
GSH adducts M1-M3, it should be noted that a direct
attack of GSH on either the sulfur atom or the adjacent
carbonyl group of the putative TZD sulfoxide (I) also
would produce conjugates M1 and M2, respectively,
without the intermediacy of the ring-opened isocyanate
II. Regardless of the precise mechanism, it is evident that
the TZD ring of troglitazone is subject to a novel pathway
of metabolic activation, which has not been described
previously, and which leads to products of heterocyclic
ring cleavage. It is noteworthy that the three GSH
conjugates M1-M3, which were generated in vitro using
either native human liver microsomes or expressed P450
3A4, were detected in the bile of troglitazone-treated rats,
F igu r e 12. Proposed scheme for the metabolic activation of
troglitazone through cytochrome P450-mediated oxidation of the
substituted chromane ring. Intermediates shown in brackets
have not been identified. The hydroxymethyl and quinone
metabolites were not characterized in this study, but are known
to be metabolites of troglitazone.
and therefore, the TZD bioactivation pathway also oper-
ates in vivo in rats.
The second route for the formation of reactive metabo-
lites from troglitazone is depicted in Figure 12 and
involves oxidation of the substituted chromane moiety.
Thus, cytochrome P450-mediated one-electron oxidation
of the phenolic hydroxyl group would result in the
phenoxy radical, a resonance form of which (the carbon-
centered radical shown in Figure 12) undergoes oxygen
rebound at the ipso position to form an unstable hemiket-
al. Spontaneous ring opening of this latter species would
afford the known quinone metabolite of troglitazone. A
similar mechanism has been proposed for the P450-
mediated oxidation of a number of aryl ethers (16) and
para-substituted phenols (17, 18). Once formed, the
quinone metabolite appears to undergo TZD ring oxida-
tion, in a fashion similar to that followed by the parent
compound, leading ultimately to GSH conjugate M4. The
formation of both the quinone metabolite of troglitazone
and GSH adduct M4 was found to be dependent on P450
3A isoforms, consistent with recent reports that metabo-
lism of troglitazone to the quinone is catalyzed by P450
3A (19, 20). An alternative fate for the phenoxy radical
intermediate is that it undergoes P450-mediated hydro-
gen atom abstraction from the 5-methyl group of the
substituted chromane ring, leading to the o-quinone
methide derivative IV depicted in Figure 12. Capture of
the latter intermediate by GSH provides a route to the

Metabolism of Troglitazone to Reactive Intermediates
Chem. Res. Toxicol., Vol. 14, No. 1, 2001 69
observed conjugate M5, while a reaction with water
would afford the known benzylic alcohol metabolite of
troglitazone (19), although the alcohol metabolite also
could be formed by a conventional P450 reaction involv-
ing hydrogen atom abstraction from the 5-methyl group
followed by oxygen rebound. Unlike the metabolites
derived through bioactivation of the TZD ring, the
formation of M5 was not specifically catalyzed by P450
3A enzymes, although it was found to be NADPH-
dependent.
Derek Von Langen (Merck Research Laboratories, Rah-
way, NJ ) for the synthesis of the quinone metabolite of
troglitazone.
Refer en ces
(1) Henry, R. R. (1997) Thiazolidinones. Curr. Ther. Diabetes 26,
553-573.
(2) Gitlin, N., J ulie, N. L., Spurr, C. L., Lim, K. N., and J urabe, H.
M. (1998) Two cases of severe clinical and histologic hepatotoxicity
associated with troglitazone. Ann. Intern. Med. 129, 38-41.
(3) Neuschwander-Tetri, B. A., Isley, W. L., Oki, J . C., Ramrakhiani,
S., Quiason, S. G., Phillips, N. J ., and Brunt, E. M. (1998)
Troglitazone-induced hepatic failure leading to liver transplanta-
tion. Ann. Intern. Med. 129, 36-38.
Experience to date indicates that troglitazone therapy
is associated with hepatotoxicity in diabetic patients.
During clinical trials of the drug, 1.9% of the troglitazone-
treated patients had liver enzyme level increases 3 times
the upper limit of normal compared to 0.6% of the
patients on placebo (4). Evaluations of liver biopsies from
patients that developed hepatotoxicity were consistent
with an idiosyncratic drug reaction (2, 3). Since its
introduction to the U.S. market in 1997 until its recent
withdrawal, troglitazone has been linked to 26 deaths,
and several patients required liver transplants following
troglitazone-induced hepatic injury (21, 22). While the
mechanism of this serious toxicity remains unknown, it
is possible that reactive metabolites of troglitazone play
a causative role. The results of this investigation provide
support for this hypothesis, inasmuch as indirect evi-
dence was obtained for the cytochrome P450-mediated
generation of several electrophilic intermediates of tro-
glitazone in vitro and in vivo. Although not investigated
in this preliminary study, it is reasonable to expect that
these reactive metabolites will form covalent adducts
to hepatic proteins, with potentially deleterious conse-
quences to the liver. Moreover, while the quinone me-
tabolite of troglitazone (which cannot form adducts to
proteins directly due to the substitution pattern) may
undergo redox cycling and induce oxidative stress, the
detection of GSH conjugate M4 raises the intriguing pos-
sibility of a protein-bound quinone metabolite, attached
to macromolecules through a disulfide bond to the residue
of the TZD ring. By this means, the benzoquinone moiety
would be “anchored” to hepatocellular proteins, providing
for potentially enhanced redox cycling activity. Experi-
ments for testing this hypothesis currently are in progress.
Troglitazone has been shown to act as an inducer of
P450 3A enzymes in human hepatocytes (23), and it may
be significant that the pathways for metabolic activation
of troglitazone leading to GSH adducts M1-M4 are
catalyzed by P450 3A enzymes. Hence, troglitazone acts
as an inducer of the enzymes that catalyze its biotrans-
formation to chemically reactive intermediates, and such
autoinduction of metabolism might well be a crucial
factor in the etiology of troglitazone-mediated liver injury.
Finally, it may be noted that this study is the first to
report on the metabolic activation of the TZD ring system,
a key structural element of many PPAR agonists mar-
keted, or in development, as antidiabetic agents. While
the precise details of the underlying mechanism remain
to be established, it appears that oxidation of the sulfur
atom activates the TZD ring, in a fashion analogous to
that described for thiophene derivatives (24). Whether
other sulfur-containing heterocycles undergo analogous
metabolic activation processes remains to be studied.
(4) Watkins, P. B., and Whitcomb, R. W. (1998) Hepatic dysfunction
associated with troglitazone. N. Engl. J . Med. 338, 916-917.
(5) European Patent EP 0 441 605 A2 (1991).
(6) Yoshioka, T., Fujita, T., Kanai, T., Aizawa, Y., Kurumada, T.,
Hasegawa, K., and Horikoshi, H. (1989) Studies on hindered
phenols and anologues. 1. Hypolipidemic and hypoglycemic agents
with ability to inhibit lipid peroxidation. J . Med. Chem. 32, 421-
428.
(7) Hu, J ., and Fox, M. A. (1999) A convenient trimethylsilylthioxy-
dehalogenation reaction for the preparation of functionalized
thiols. J . Org. Chem. 64, 4959-4961.
(8) Hart, T. W., Vine, M. B., and Walden, N. R. (1985) Thiolsulpho-
nate derivatives of amino acids. Tetrahedron Lett. 26, 3879-3882.
(9) Baillie, T. A., and Davis, M. R. (1993) Mass spectrometry in the
analysis of glutathione conjugates. Biol. Mass Spectrom. 22, 319-
325.
(10) Davis, M. R., Kassahun, K., J ochheim, C. M., Brandt, K. M., and
Baillie, T. A. (1993) Glutathione and N-acetylcysteine conjugates
of 2-chloroethyl isocyanate. Identification as metabolites of N,N′-
Bis(2-chloroethyl)-N-nitrosourea in the rat and inhibitory proper-
ties toward glutathione reductase in vitro. Chem. Res. Toxicol. 6,
376-383.
(11) Kassahun, K., Davis, M., Hu, P., Martin, B., and Baillie, T. (1997)
Biotransformation of the naturally occurring isothiocyanate sul-
foraphane in the rat: Identification of Phase I metabolites and
glutathione conjugates. Chem. Res. Toxicol. 10, 1228-1233.
(12) Borel, A. G., and Abbott, F. S. (1993) Identification of carbomoy-
lated thiol conjugates as metabolites of the antineoplastic 1-(2-
chloroethyl)-3-cyclohexyl-1-nitrosourea, in rats and humans. Drug
Metab. Dispos. 21, 889-901.
(13) Kassahun, K., Hu, P., Grillo, M. P., Davis, M. R., J in, L., and
Baillie, T. A. (1994) Metabolic activation of unsaturated deriva-
tives of valproic acid. Identification of novel glutathione adducts
formed through coenzyme A-dependent and -independent pro-
cesses. Chem.-Biol. Interact. 90, 253-275.
(14) Kawai, K., Kawasaki-Tokui, Y., Odaka, T., Tsuruta, F., Kazui,
M., Iwabuchi, H., Nakamura, T., Kinoshita, T., Ikeda, T., Yoshio-
ka, T., Komai, T., and Nakamura, K. (1997) Disposition and
metabolism of the new oral antidiabetic drug troglitazone in rats,
mice and dogs. Arzneim.-Forsch. 47, 356-368.
(15) Loi, C. M., Alvey, C. W., Randinitis, E. J ., Abel, R., Young, M. A.,
and Koup, J . R. (1997) Meta-analysis of steady-state pharmaco-
kinetics of troglitazone and its metabolites. J . Clin. Pharmacol.
37, 1038-1047.
(16) Ohe, T., Mashino, T., and Hirobe, M. (1994) Novel metabolic
pathways of arylethers by cytochrome P450: Cleavage of the
oxygen-aromatic ring bond accompanying ipso-substitution by the
oxygen atom of the active species in cytochrome P450 models and
cytochrome P450. Arch. Biochem. Biophys. 310, 402-409.
(17) Ohe, T., Mashino, T., and Hirobe, M. (1997) Substituent elimina-
tion from p-substituted phenols by cytochrome P450. ipso-
Substitution by the oxygen atom of the active species. Drug Metab.
Dispos. 25, 116-122.
(18) Ohe, T., Hirobe, M., and Mashino, T. (2000) Novel metabolic
pathway of estrone and 17â-estradiol catalyzed by cytochrome
P-450. Drug Metab. Dispos. 28, 110-112.
(19) Yamazaki, H., Shibata, A., Suzuki, M., Nakajima, M., Shimada,
N., Guengerich, F. P., and Yokoi, T. (1999) Oxidation of trogli-
tazone to a quinone-type metabolite catalyzed by cytochrome
P-450 2C8 and P-450 3A4 in human liver microsomes. Drug
Metab. Dispos. 27, 1260-1266.
(20) He, K., Woolf, T. F., Kindt, E. K., Fielder, A., and Talaat, R. E.
(1999) Troglitazone quinone formation catalyzed by human and
rat CYP3A4: An atypical P450 oxidation reaction. Proceedings
of the 9th North American ISSX Meeting, #207, Nashville, TN.
Ack n ow led gm en t. We thank Dr. Raju Subramanian
(Merck Research Laboratories, West Point, PA) for
providing the NMR spectrum of metabolite M5 and Dr.

70 Chem. Res. Toxicol., Vol. 14, No. 1, 2001
Kassahun et al.
(21) Polsker, G. L., and Faulds, D. (1999) Troglitazone. A review of
its use in the management of type 2 diabetes mellitus. Drugs 57,
409-438.
R. (1999) Troglitazone increases cytochrome P-450 3A protein and
activity in primary cultures of human hepatocytes. Drug Metab.
Dispos. 27, 1194-1199.
(22) Kohlroser, J ., Mathai, J ., Reichheld, J ., Banner, B. F., and
Bonkovsky, H. L. (2000) Hepatotoxicity due to troglitazone:
Report of two cases and review of adverse events reported to the
United States Food and Drug Administration. Am. J . Gastroen-
terol. 95, 272-276.
(24) Mansuy, D., Valadon, P., Erdelmeier, I., Lopez-Garcia, P., Amar,
C., Girault, J . P., and Dansette, P. M. (1991) Thiophene S-oxides
as new reactive metabolites; formation by cytochrome P450
dependent oxidation and reaction with nucleophiles. J . Am. Chem.
Soc. 113, 7825-7826.
(23) Ramachandran, V., Kostrubsky, V. E., Komoroski, B. J ., Zhang,
S., Dorko, K., Esplen, J . E., Strom, S. C., and Venkataramanan,
TX000180Q