LC–MS/MS and chiroptical spectroscopic analyses of multidimensional metabolic systems of chiral thalidomide and its derivatives

Article abstract of DOI:10.1002/chir.22683

Enantiomeric thalidomide undergoes various kinds of biotransformations including chiral inversion, hydrolysis, and enzymatic oxidation, which results in several metabolites, thereby adding to the complexity in the understanding of the nature of thalidomide. To decipher this complexity, we analyzed the multidimensional metabolic reaction networks of thalidomide and related molecules in vitro. Characteristic patterns in the amount of various metabolites of thalidomide and related molecules generated during a combination of chiral inversion, hydrolysis, and hydroxylation were observed using liquid chromatography–tandem mass spectrometry and chiroptical spectroscopy. We found that monosubstituted thalidomide derivatives exhibited different time-dependent metabolic patterns compared with thalidomide. We also revealed that monohydrolyzed and monohydroxylated metabolites of thalidomide were likely to generate mainly by a C-5 oxidation of thalidomide and subsequent ring opening of the hydroxylated metabolite. Since chirality was conserved in most of these metabolites during metabolism, they had the same chirality as that of nonmetabolized thalidomide. Our findings will contribute toward understanding the significant pharmacological effects of the multiple metabolites of thalidomide and its derivatives.

Full text of DOI:10.1002/chir.22683

Received: 23 September 2016
DOI: 10.1002/chir.22683
Revised: 12 December 2016
Accepted: 4 January 2017
R E G U L A R A R T I C L E
LC–MS/MS and chiroptical spectroscopic analyses of
multidimensional metabolic systems of chiral thalidomide and its
derivatives
Yoshiyuki Ogino1
| Masahito Tanaka2 | Togo Shimozawa1
| Toru Asahi1
1 Waseda University, Department of Life
Science and Medical Bioscience, Tokyo,
Japan
2 National Institute of Advanced Industrial
Science and Technology (AIST), Research
Institute of Instrumentation Frontier,
Tsukuba, Japan
Abstract
Enantiomeric thalidomide undergoes various kinds of biotransformations including
chiral inversion, hydrolysis, and enzymatic oxidation, which results in several
metabolites, thereby adding to the complexity in the understanding of the nature
of thalidomide. To decipher this complexity, we analyzed the multidimensional
metabolic reaction networks of thalidomide and related molecules in vitro.
Characteristic patterns in the amount of various metabolites of thalidomide and
related molecules generated during a combination of chiral inversion, hydrolysis,
and hydroxylation were observed using liquid chromatography–tandem mass
spectrometry and chiroptical spectroscopy. We found that monosubstituted
thalidomide derivatives exhibited different time‐dependent metabolic patterns com-
pared with thalidomide. We also revealed that monohydrolyzed and
monohydroxylated metabolites of thalidomide were likely to generate mainly by a
C‐5 oxidation of thalidomide and subsequent ring opening of the hydroxylated
metabolite. Since chirality was conserved in most of these metabolites during
metabolism, they had the same chirality as that of nonmetabolized thalidomide.
Our findings will contribute toward understanding the significant pharmacological
effects of the multiple metabolites of thalidomide and its derivatives.
Correspondence
Toru Asahi, Waseda University, Department
of Life Science and Medical Bioscience,
Tokyo, Japan.
Email: tasahi@waseda.jp
Funding Information
Grant‐in‐Aid for JSPS Fellows, Grant/Award
Number: 14J07106
KEYWORDS
chiral inversion, chiroptical spectroscopy, enzymatic hydroxylation, hydrolysis, LC–MS/MS, thalidomide
1 | INTRODUCTION
ingredient.4-6 Given the instability of drugs in vitro, not only
the static discrimination of chirality is clinically important
The intricate molecular behavior of pharmaceutical products
in vitro induced by hydrolysis, oxidation, reduction, conjuga-
tion, or isomerization leads to various pharmacological
effects that may be consequential. These transformations
could be beneficial or they could be harmful.1-3 Despite the
structural transformations caused by developing drugs via
metabolic actions including chiral inversion, which are of
course thoroughly considered during in silico drug design,
it is still difficult to predict all the effects of conformational
isomers and stereoisomers of an active pharmaceutical
but the possibility of dynamic inversion at the stereogenic
center must also be considered.7-9
Thalidomide (TD)‐induced teratogenicity, arguably the
greatest drug disaster in history, underscores the impor-
tance of fugitive stereochemistry. The resident (S)‐isomer
of orally prescribed TD was identified as the principle
cause of the adverse reaction, whereas the (R)‐isomer
exhibited
a
prescriptive sedative hypnotic effect.10
However, because of the high acidity of the stereocenter
hydrogen atom of the glutarimide ring, TD undergoes
Chirality. 2017;1–12.
wileyonlinelibrary.com/journal/chir
© 2017 Wiley Periodicals, Inc.
1

2
OGINO ET AL.
facile chiral inversion under physiological pH and tempera-
ture conditions.11,12 This isomerization is accompanied by
nonenzymatic hydrolysis and enzymatic hydroxylation
in vivo, prompting our interest in the toxicity of its
metabolites. Some of the ring‐opened hydrolytic products
were reported to be a more potent inhibitor of tumor
necrosis factor‐α (TNF‐α) production than unmetabolized
TD.13,14 Another report theoretically proposed the mechanism
that TD or its ring‐opened products selectively bind to
specific consensus sequences, GC boxes, and suppressed the
expression of some proteins relating to limb growth.15
Previous animal examinations revealed that various mono‐
and dihydroxylated TD metabolites are formed in the
presence of CYP450, mainly 2C19, involved in the liver
microsome, and exhibit an enhanced efficacy for the
antiinflammatory effect compared with unmetabolized
TD.16-19 In particular, oxidation of TD occurring at the
C‐5′ position in the glutarimide ring induces a new
stereoisomerism. Oxidation of TD occurring in the
phthalimide ring leads to further hepatic metabolic reactions
such as glutathione conjugation and glucuronic acid
conjugation.20-22 Thus, TD has multiple reaction sites for
hydrolysis and hydroxylation and several metabolic products
of TD and their enantiomers have been identified with
experimental analyses (Figure 1A, Figures S1, S2).
Multiple functions and behaviors relevant to TD have
already been identified in previous studies. It has been sug-
gested that the intercalation of TD into stacked intracellular
DNA basepairs via the plane structure of its phthalimide
ring induces DNA oxidation, which may trigger an antican-
cer effect.15 We previously demonstrated that TD and a
fluorinated derivative, 3′‐fluorothalidomide (3′‐FTD), show
self‐disproportionation of enantiomers, which may lead to
an enantiospecific function of TD against racemization.23,24
Recently, cereblon (CRBN) has been identified as a TD
target protein for suppressing the production of growth
factor FGF8 during the organogenesis stage and inducing
malformation or total absence of a fetal limb.25-27 Struc-
tural studies of CRBN and the binding molecules have
been reported recently, and revealed the structural charac-
teristics of the potential moiety of TD and its derivatives
that bind to the target proteins and evoke translational sig-
naling cascades.28,29 The medicinal properties of TD such
as its hypnotic, sedative, and teratogenic effects seem to
result from these multiple effects, which are still indetermi-
nate. Furthermore, several derivatives and analogs of TD
have been developed to reduce its side effects and other
properties such as its slow chiral inversion.22,30-37 However,
it is uncertain whether the existence of various structures
including TD, its derivatives, and their metabolites is the
FIGURE 1 (A) Biotransformation pathways of enantiomeric TD involving nonenzymatic chiral inversion, hydrolysis, and enzymatic hydroxylation.
Different combinations of substituents Ri (i = 1, 2, 3, 4, 5) on TD determines different derivatives and structural isomers of mono‐hydroxylated
metabolites. (B) Schematic of three‐directional and hierarchical metabolic reaction network of TD. TD undergoes hydrolysis on the illustrated location
and suffers from chiral inversion as indicated by the mirrored plane. Oxidation of TD is demonstrated by moving downward in the layers. (C) The
definitions of (i,j)‐metabolites in the present study

OGINO ET AL.
3
reason why TD is effective for various intractable diseases.
We consider that this struggle stems from the complexity
of the metabolic reaction pathways of TD, which includes
spontaneous hydrolysis and chiral inversion, as well as
enzymatic hydroxylation (Figure 1B). In our computational
approaches toward understanding of the complexity of TD
metabolic reaction pathways, we quantitatively simulated
the temporal changes in the concentration and the enantio
excess of hydrolytic metabolites of TD during spontaneous
chiral inversion and hydrolysis by using experimental and
theoretical kinetic parameters for TD metabolism.38 This
study enabled us to expansively comprehend the molecular
biological assays analyzing the antiinflammatory effects of
TD hydrolytic metabolites from the viewpoint of the
dynamic metabolism of TD. Since the metabolic reaction
network of TD has not been achieved, the preferred meta-
bolic pathway is not understood.
In this study, we experimentally analyze the progress of
the metabolic reaction network of chiral TD by employing
liquid chromatography–tandem mass spectrometry (LC–
MS/MS) and chiroptical spectroscopy. The metabolism of
TD derivatives such as 3′‐FTD and TD amino derivatives
of TD, pomalidomide (PD), are also examined to reveal
the structural effects of substituents on the TD molecule
on the progress of the metabolic reaction network.
Furthermore, we monitor the changes in the chirality of
hydrolyzed and hydroxylated metabolites by measuring
the temporal changes in the circular dichroism (CD) spectra
during the metabolic reactions. To estimate the preferred
metabolic pathway of the multiple possible pathways, we
calculate the energy barriers for the first hydrolysis step
of TD and hydroxylated TD into their glutarimide ring‐
broken products by using density functional theory
calculations.
2.2 | Hydrolysis of TD and its derivatives
In this study, hydrolysis of the samples was performed before
hydroxylation of the samples. Enantiomeric and racemic TD,
3′‐FTD, and PD were dissolved in dimethyl sulfoxide
(DMSO) (10 mM). The highly concentrated solutions were
then mixed with 100 times volume of 0.1 M potassium phos-
phate buffer (pH 7.4). The mixture (total volume of 500 μL)
was incubated at 37 °C for 1, 2, and 3 h. To avoid the influ-
ence on the function of oxidation enzymes by using organic
solvents to stop the reaction, hydrolysis was instead termi-
nated by rapid freezing of the reaction solution. These
samples were stored at −80 °C.
2.3 | Hydroxylation of TD and its derivatives
Hydroxylation reactions were performed with a typical incu-
bation mixture (total volume of 200 μL) containing 0.1 mM
EDTA (Dojindo, Japan), an NADPH‐generating system
(1.3 mM NADP+ (Oriental Yeast, Japan), 3.3 mM glucose
6‐phosphate (Wako), 0.4 unit mL−1 glucose 6‐phosphate
dehydrogenase (Oriental Yeast) and 3.3 mM MgCl2 (Wako)),
CD‐1 mouse liver microsome extract (protein content
≥20 mg mL−1) (Sigma‐Aldrich, Japan), and the hydrolyzed
samples (100 μM). To stop unnecessary degradation of TD,
3′‐FTD, and PD via hydrolysis, all the above‐mentioned pro-
cedures were carried out quickly and on ice. Each resulting
mixture was then incubated at 37 °C for 10, 30, and
60 min. Incubation was terminated by adding 200 μL of
ice‐cold CH3CN. The internal standard (ethyl‐p‐
hydroxybenzoate) was added into the sample solutions at a
final concentration of 100 μM. The solutions were centri-
fuged at 1700 g at 4 °C for 15 min, and then collected sepa-
rately into an organic supernatant. These samples were
filtrated with a spinfilter (Merckmillipore, Japan) and the
StrataX solid‐phase extract cartridge with polymer stationary
phase (Shimadzu, Japan). These samples were then applied
on LC–MS/MS systems and a circular dichroism spectropo-
larimeter to obtain absorbance and CD spectra.
2 | MATERIALS AND METHODS
2.1 | Chemicals
(R)‐(+)‐, (S)‐(−)‐, and (RS)‐TD were synthesized as previ-
ously described. The hydrolytic products of TD, (S)‐ and
(RS)‐phthaloylisoglutamine (PiG), (S)‐ and (RS)‐phthaloy-
lglutamine (PG), and (S)‐ and (RS)‐2‐(α‐carboxybenzamide)
glutarimide (CBG), were purchased from Tokyo Chemical
Industry Co., Ltd. (Japan). Racemic 3′‐FTD was synthesized
as previously reported. Racemic PD was purchased from
Sigma‐Aldrich (Japan). All other chemicals used in the
present study to dissolve chemical compounds were all of
spectroscopic analytical grade and purchased from Wako
Pure Chemical Industries, Ltd. (Japan). Ethyl‐p‐hydroxybenzoate,
which was used as an internal standard for LC–MS/MS, was also
purchased from Wako Pure Chemical Industries, Ltd. (Japan).
2.4 | LC–MS/MS assays
LC–MS/MS analyses of the complex metabolic reaction
network of TD and its derivatives were performed on a
Shimadzu Prominence HPLC system connected to an AB
SCIEX TripleTOF system mass spectrum (AB SCIEX,
Japan). TD, 3′‐FTD, PD, as well as their early‐stage metab-
olites, were analyzed in the mass spectrum (MS) positive
ion mode. Liquid chromatography conditions were as
follows: buffer A contained 0.1% HCOOH (v/v) in 5%
CH3CN and 95% H2O (v/v), and buffer B contained 0.1%
HCOOH (v/v) in 95% CH3CN and 5% H2O (v/v). The gra-
dient program used in these analyses with a flow rate of

4
OGINO ET AL.
0.2 mL min−1 was as follows: 0 to 1 min, hold at 99% A,
1 to 3 min, linear gradient from 99% A to 1% A (v/v), 3 to
5 min, hold at 1% A, 5 to 6 min, linear gradient from 1%
A to 99% A (v/v), 6 to 10 min, hold at 99% A. The
column temperature was set at 25 °C with a column oven.
The prepared samples (5 μL) were infused with an
autosampler. MS analyses were carried out in the ESI‐
TOF positive ion mode. Total ion chromatograms of the
metabolites of TD and its derivatives were acquired over
the range m/z 100–1000 with a Shiseido CAPCELL CORE
ADME adamantyl semimicro column (2.1 × 100 mm).
The m/z of the fragment ions of metabolites of TD and
its derivatives were also obtained over the range m/z
100–300. ESI conditions were as follows: capillary voltage,
2.5 V; sampling cone, 30; extraction cone, 4.1; source
temperature, 120 °C; desolvation temperature, 325 °C;
and Trap CE parameter, 6. Hydrolyzed and hydroxylated
metabolites of TD, 3′‐FTD, and PD were quantified by
the peak area of extracted ion chromatogram (XIC) at the
monoisometric exact mass displayed in Table S1 (mass tol-
erance: Æ0.1 Da). For quantification of each metabolite, the
proper m/z transitions during MS/MS fragmentation were
employed because of the lack of an authentic standard,
such as the m/z 293 → 225 transition of (1,1)‐metabolites
(Figure S3). The XICs of the internal standard were
utilized to verify adequate ionization of the sample. The
XICs of these m/z were acquired using the AB SCIEX
Analyst TF 1.7 software.
2.7 | Geometrical optimization
All quantum chemical calculations in the present study,
including structure optimization and electron transition
calculations, were performed using Gaussian09 (Revision
D.01) and GaussView5.39 The initial coordinates of
enantiomeric and racemic TD and racemic 3′‐FTD were
obtained from their crystal structure, which we previously
reported.23,40 The initial structure of racemic PD was
acquired by the modification of the crystal structure of race-
mic TD. The initial structure of one of the first‐generation
hydrolytic products, (S)‐CBG, was acquired from the crystal
structure obtained by our X‐ray crystallographic analysis
(unpublished). The B3LYP/6–311+g(d,p) basis set was used
for geometrical optimization of the substrate and the first‐
generation hydrolytic products of TD, 3′‐FTD, and PD.
2.8 | Computation of absorbance and CD
intensity
Theoretical CD and absorbance intensities were obtained by
calculating the transition intensities and rotatory strengths
of each singlet excitation at the B3LYP/6–311+g(d,p) level
of theory for 5‐hydroxylated TD (5‐OH‐TD), cis/trans‐5′‐
hydroxylated TD (cis/trans‐5′‐OH‐TD), 5‐hydroxylated
CBG (5‐OH‐CBG), and cis/trans‐5′‐hydroxylated CBG
(cis/trans‐5′‐OH‐CBG).
2.9 | Calculation of activation energy of
metabolic reactions
To estimate the activation energy for hydrolysis of TD
into CBG, the structure of the transition state (TS) was
determined. The activation energy was calculated as the
difference between the electron energy of the TS structure and
that of the substrate structures (TD + H2O). The substrate
structures were optimized using the B3LYP/6–311+g(d,p)
basis set. To reveal the effect of oxidation on TD, we also
calculated the activation energy for the reactions of 5OH‐TD,
cis‐5′‐OH‐TD, and trans‐5′‐OH‐TD into 5OH‐CBG, cis‐5′‐
OH‐CBG, and trans‐5′‐OH‐CBG. These substrate structures
were also optimized using the same level of computations.
2.5 | Circular dichroism spectroscopy
Hydrolyzed and hydroxylated samples (~200 μL) were
diluted with the same volume of diluted water. CD and absor-
bance spectra were measured using a J‐820 CD spectropolar-
imeter (Jasco, Japan) with a quartz cell under the following
conditions: scan speed, 100 nm min−1; temperature of the
cell holder, 27 °C (using a Peltier device cell holder);
sensitivity, standard (100 mdeg). Blank tests for CD and
absorbance spectra were performed with a 1:1 mixture of
diluted water and CH3CN.
3 | RESULTS AND DISCUSSION
2.6 | Data analysis
3.1 | Multiple steps of hydrolysis and
hydroxylation
The generated metabolites of TD, 3′‐FTD and PD were
quantified from the peak areas of XIC for the specified m/z
of each metabolite listed in Table S1. Absolute values of
the area of peak at retention time (tR) were calculated using
the Origin 2016 software (LightStone, Japan). All graphs
displayed in the present study were also drawn using Origin
2016 software.
We first defined the group of metabolites that suffered from
i‐times hydrolysis and j‐times hydroxylation as (i,j)
(unmetabolized molecule can be written as (0,0)).
Hydrolysis, hydroxylation, and chiral inversion are
then expressed as (i,j) → (i + 1,j), (i,j) → (i,j + 1), and

OGINO ET AL.
5
(R)‐(i,j) ⇆ (S)‐(i,j), respectively (Figure 1C). Note that some
of the hydrolysis reactions increase the mass of metabolites
by ~1 Da, which is not caused by the addition of a hydrogen
atom but the replacement of an amino moiety with a hydroxyl
moiety (Table S1).
To consider the initial relationship between hydrolysis
and hydroxylation initially, we let TD undergo spontaneous
hydrolysis and enzyme‐driven hydroxylation with the differ-
ent reaction times for hydrolysis (1–3 h) and hydroxylation
(10–60 min). We detected various metabolites suffering from
multiple times of hydrolysis and hydroxylation. In the present
study we focused on the upstream major four metabolites
including unmetabolized TD. Figure 2A shows the XICs at
m/z 259.07, 275.06, 277.08, and 293.08 of the sample with
1 h of hydrolysis and 30 min of hydroxylation of (S)‐TD.
We plotted the time dependence of the amount of unmetabo-
lized TD (0,0), and other hydrolyzed and hydroxylated
metabolites (i,j) for i = 0, 1 and j = 0, 1 (Figure 2B). The
areas of peak at tR = 4.80 for TD, tR = 1.12 for (0,1)‐
metabolites, tR = 4.32 for (1,0)‐metabolites, and tR = 1.55
for (1,1)‐metabolites were utilized to draw the level plots.
We acquired the unique metabolic patterns for the principal
metabolites of TD. In the timescale of the metabolic reactions
employed in the present study, TD hydrolysis seems to
mainly proceed in the first step hydrolysis, TD into PiG,
PG, and CBG. In accordance with the preceding report that
FIGURE 2 (A) XICs at m/z 259.07 for
(0,0)‐metabolite (TD), m/z 275.06 for (0,1)‐
metabolites, m/z 277.08 for (1,0)‐metabolites,
and m/z 293.08 for (1,1)‐metabolites of the
sample with 1 h of hydrolysis and 30 min of
hydroxylation of (S)‐TD. (B) Level plots of
the amount of (0,0), (0,1), (1,0), and (1,1)‐
metabolites for different hydrolysis and
hydroxylation reaction times of (S)‐TD. The
plots were obtained from the peak areas of
XICs such as presented in (A) with different
metabolic reaction times. The amounts of
metabolites were normalized to the
maximum amount of each metabolite

6
OGINO ET AL.
(1,0)‐metabolites are rather stable against further degrada-
tion, few (2,0)‐ and other downstream metabolites are gener-
ated during this timescale.13 Therefore, the generation and
degradation of (1,1)‐metabolites seem to occupy an impor-
tant part of the metabolic reaction network. As hydrolysis
and hydroxylation proceeded, the amount of TD showed a
rapid decrease, resulting in the formation of downstream
metabolites including (1,0)‐ and (0,1)‐metabolites along with
both the reaction time of hydrolysis and hydroxylation.
Meanwhile, (1,1)‐metabolites were formed gradually as
hydrolysis and hydroxylation of TD proceeded. The amount
of (1,1)‐metabolites reached a maximum within 1 h of hydro-
lysis and 60 min of hydroxylation. Note that TD hydrolysis
can also occur during the hydroxylation reaction time since
TD is exposed to the aqueous buffer during the enzymatic
hydroxylation. Therefore, the hydroxylation reaction time
also encompasses hydrolysis. Following a maximum, the
(1,1)‐metabolites seemed to degrade into the downstream
metabolites owing to subsequent reactions. This suggests that
the relationship between hydrolysis and hydroxylation
produces a characteristic metabolic pattern on each metabolic
layer and affects the whole system of layered metabolic
reactions.
degrades into (1,0)‐metabolites, after which the (1,0)‐
metabolites suffer from hydroxylation mainly at C‐5 on the
phthalimide ring or C‐5′ on the glutarimide ring resulting in
(1,1)‐metabolites; 2) TD undergoes hydroxylation at C‐5
first, and then, (0,1)‐metabolites suffer from hydrolysis
resulting in (1,1)‐metabolites; and 3) TD undergoes hydrox-
ylation at C‐5′ first, and then, the (0,1)‐metabolites suffer
from hydrolysis resulting in (1,1)‐metabolites. Note that other
oxidations of TD rings such as at C‐4 or at the nitrogen atom
in the glutarimide ring can also occur. In the present study,
however, we focused on C‐5 and C‐5′ oxidation of TD
because the other hydroxylated metabolites are not dominant
during TD metabolism.17,18
To consider these three preferred pathways, we examined
whether (1,0)‐metabolites of TD are hydroxylated in the
presence of oxidation enzymes. Figure 3B illustrates the
XICs at m/z 293.08 of the hydroxylation samples of TD
and the most abundant (1,0)‐metabolites, CBG. Compared
with the XIC of the sample generated from TD, small ion
intensities were observed at the XICs of the samples gener-
ated from CBG. We then measured the amount of (1,1)‐
metabolites generated from TD and CBG using the obtained
peak area of the XICs (Figure 3C). Note that the concentra-
tions of each substrate TD and CBG were set to the same in
Figure 3B. Since part of TD was only converted to (1,0)‐
metabolites after 1 h of hydrolysis, the amount of (1,1)‐
metabolites generated from TD should be smaller than that
generated from CBG. However, the amount of (1,1)‐metabo-
lites generated from CBG was small compared with that gen-
erated from TD with experimental replications. Therefore, we
found that (1,0)‐metabolites of TD were less likely to suffer
from enzymatic hydroxylation. Other (1,0)‐metabolites were
also oxidized into (1,1)‐metabolites, which can be negligible
3.2 | Preferred pathway in temporal
formation of monohydroxylated TD metabolites
Next, we focused on the monohydrolyzed and
monohydroxylated metabolites of TD, i.e., (1,1)‐metabolites.
Our experimental results revealed that certain hydrolysis and
hydroxylation reaction times are needed to generate (1,1)‐
metabolites. During the production of (1,1)‐metabolites, the
following three pathways are possible (Figure 3A): 1) TD
FIGURE 3 (A) Possible metabolic pathways of the formation of (1,1)‐metabolites involving hydrolysis and hydroxylation. In this figure, the
pathway relavent to CBG was chosen to represent (1,0)‐metabolites because CBG is the most abundant metabolite among (1,0)‐metabolites of TD.
(B) XICs at m/z 293.08 representing the amount of (1,1)‐metabolites generated from TD and CBG. (C) The amount of (1,1)‐metabolites generated
from TD and (1,0)‐metabolites after 1 h of hydrolysis and 60 min of hydroxylation (TD) and after 60 min of hydroxylation (CBG). The values
represent mean values (n = 3) Æ standard deviation

OGINO ET AL.
7
due to the small presence of PiG and PG during metabolism
(Figure S4).13 The stability of (1,0)‐metabolites during
hydroxylation also supports our conclusion. Further, the
hydrolysis rate constant of 5‐OH‐TD has previously been
reported to be relatively smaller than that of unmetabolized
TD, whereas that of 5′‐OH‐TD is almost the same as that of
unmetabolized TD.41 Our experiments and the results
described in Yamamoto et al41 lead us to choose pathway (3)
as a major pathway. According to these analyses, ring opening
of TD and oxidation at C‐5 on the phthalimide ring possibly
inhibit subsequent hydroxylation and hydrolysis, respectively.
This implies that these properties relevant to microscopic
chemical reactions on each metabolite play an important role
in determining the hierarchy of the metabolic reactions.
In these plots, the peak areas of XICs at m/z 291.06 for (0,2)‐
metabolites of TD, m/z 309.05 for (0,2)‐metabolites of 3′‐
FTD, and m/z 306.07 for (0,2)‐metabolites of PD were nor-
malized by the maximum value of the peak areas. Di‐hydrox-
ylated metabolites were generated gradually after around
30 min of hydroxylation, but degraded with increased
hydroxylation reaction time owing to the ring breakdown of
hydroxylated TD. For 3′‐FTD and PD, the amount of (0,2)‐
metabolites reached a maximum with 30 min of hydroxyl-
ation and without any hydrolysis reaction time, whereas for
TD, (0,2)‐metabolites reached their maximum after 60 min
of hydroxylation and with a short hydrolysis reaction time.
This indicates that substitution of fluorine at the stereogenic
center causes the acceleration of P450‐driven hydroxylation.
The fluorine atom substituted at the stereogenic center acts as
an electron‐withdrawing moiety and affects the electron den-
sity of 3′‐FTD. The slight mesomeric effect of the fluorine
atom might also contribute to the rapid hydrolysis as well
as acceleration of hydroxylation. Amino moiety substitution
on the aromatic/nonaromatic ring structures seems to have
almost the same effect as the fluorine atom.
3.3 | Substituent effect on the temporal di‐
hydroxylation
Although the hydrolysis rate constants of 3′‐FTD and PD
differ from those of TD, they show almost the same behavior
as TD. To examine the effect of the substituent on the TD
molecule on the metabolic character, we next plotted the
time‐dependent formation of di‐hydroxylated metabolites of
TD, 3′‐FTD, and PD with different hydrolysis and hydroxyl-
ation reaction times (Figure 4). To understand the progress of
the metabolic reaction network in terms of the metabolic
layer, we did not discriminate between the structural isomers
of di‐hydroxylated metabolites, which we regarded as
belonging to the same metabolic layer, i.e., (0,2)‐metabolites.
Figure 4A shows that di‐hydroxylation of TD requires
some hydrolysis time, which seems to be usually unnecessary
for oxidation of substrate. Chowdhury et al. previously
reported a nonenzymatic oxidation pathway of 5‐hydroxyl-
ated TD.20 In their hypothesis, an intermediate of the addition
of a water molecule to 5‐OH‐TD can be formed during keto‐
enol tautomerism in 5‐OH‐TD evoked by electron transfer
from the hydroxyl moiety. Di‐hydroxylation of TD is
FIGURE 4 (A) Level plots of the amount
of (0,2)‐metabolites of TD, 3′‐FTD, and PD
with different hydrolysis and hydroxylation
reaction times. The amounts of metabolites
were normalized to the maximum amount of
each metabolite. (B) Structure of 3′‐FTD
and PD.

8
OGINO ET AL.
completed with further electron transfer and transformation
to the keto‐form (Figure S5). The extra hydrolysis time may
contribute to this nonenzymatic oxidation, which results in
the di‐hydroxylated metabolites of TD. On the TD deriva-
tives, this nonenzymatic oxidation of monohydroxylated
molecules may be promoted by the disturbance of the
electron density induced by a substituent with a fluorine atom
or amino moiety.
The CD signal around 235 nm remained positive during
metabolism, implying that the metabolites that are stable
against chiral inversion are generated and conserved in the
solution (Figure 5A). The result of theoretical rotatory oscil-
lators and frequency intensities of the (S)‐isomer of 5‐OH‐TD
show similar signs and intensities compared with the experi-
mental spectra (Figure 5B).
We also computed theoretical absorbance and CD spectra
of the most abundant TD metabolites with the same steric
configuration, including (S)‐isomers of cis/trans‐5′‐OH‐TD,
5‐OH‐CBG, and cis/trans‐5′‐OH‐TD (Figure S6, S7). The
broad absorption band observed in the experiment may be
the result of hydroxylated CBG, especially 5‐OH‐CBG.
The summation of the absorbance spectra of the most abun-
dant metabolites and that of the CD spectra agrees with the
experimental data (Figure S8), indicating that the chirality
of the unmetabolized TD is conserved in the metabolites
during the complex metabolic reactions. This also points
toward the origin of the enantiospecific efficacy of TD; these
metabolites conserve the chirality information, whereas
original TD undergoes facile racemization and loses its
chirality information.
3.4 | Conservation of chirality in hydrolyzed
and hydroxylated metabolites
The above LC–MS/MS assays with a normal reverse phase
column did not detect the chirality of the samples. To reveal
the change in the chirality of the generated metabolites during
hydrolysis and hydroxylation, we measured absorbance and
CD spectra of the reaction solutions for different metabolic
reaction times. Figure 5 demonstrates the time dependence
of the absorbance and CD spectra of the reaction solutions.
As TD hydrolysis was under way, the CD signal around
250 nm became gradually smaller and then disappeared,
which was originated from ring opening and racemization
of TD (Figure 5A). For the reaction solution in the absence
of (S)‐TD, the obvious CD signals were not observed around
the wavelength region where the TD metabolites seem to
exhibit the CD signals (Figure 5B). This result implies that
the proteins and other compounds involved in the reaction
solutions do not interrupt the CD signals of the metabolites.
3.5 | Enantioselectivity in TD metabolic
pathway progression
Next, we examined the amounts of some metabolites gener-
ated from (R)‐ and (S)‐TD to determine whether metabolic
FIGURE 5 (A) Time‐dependent changes in absorbance and CD spectra of the sample solutions of (S)‐TD with 1–3 h of hydrolysis. The light
blue and dark blue bold lines represent the absorbance and CD spectra of unmetabolized (S)‐TD and (S)‐TD after 3 h of hydrolysis, respectively.
(B) Time‐dependent changes in absorbance and CD spectra of the sample solutions of (S)‐TD with 10–60 min of hydroxylation. The light blue and
dark blue bold lines represent the absorbance and CD spectra of (S)‐TD after 10 min and 60 min of hydroxylation, respectively. The green bars show
theoretical oscillator strength and rotatory strength of (S)‐5‐OH‐TD, one of (0,1)‐metabolites of TD

OGINO ET AL.
9
reaction enantioselectivity occurs in the complex TD
metabolic systems (Figure 6). We selected (0,0)‐ (TD),
(1,0)‐ and (1,1)‐metabolites for this measurement since these
are the dominant metabolites during the metabolic reactions.
Moreover, we treated TD for 1 h each of hydrolysis and
hydroxylation to allow sufficient time for TD to undergo
the metabolic reactions. Obvious differences in the amount
of these three metabolites generated from different enantio-
mers of TD were not observed. Therefore, we conclude that
there is no enantioselectivity in the metabolic systems of
TD. Meyring et al. reported that TD oxidation occurs
enantioselectively and that (R)‐TD preferentially undergoes
C‐5′ oxidation, whereas the opposite enantiomer of TD
mainly suffers from C‐5 oxidation.42 In this study, we did
not discriminate between the oxidation sites of TD during
LC–MS/MS analysis and regarded all (0,1)‐metabolites as
the same metabolic step. Thus, our results showed no
enantioselectivity in the metabolic reactions of TD.
Moreover, these results imply that the combination rates of
(R)‐ and (S)‐TD with P450 would be almost the same even
though the binding conformations and the oxidation sites
are different between two enantiomers.
3.6 | Activation energy of the first hydrolysis
step of TD and its hydroxylated metabolites
To consider the preferential metabolic pathway, we
calculated the energy barriers of the first hydrolysis step
because these reactions determine the general direction of
dynamical metabolic reactions of TD. Figure S9 illustrates
the estimated energy diagrams of hydrolysis of TD, 5‐OH‐
TD, and cis/trans‐5′‐OH‐TD into CBG, 5‐OH‐CBG, and
cis/trans‐5′‐OH‐CBG, respectively. We focused hydrolysis
into CBG and hydroxylated CBG since CBG, one of the three
FIGURE
6
Amounts of (0,0)‐, (1,0)‐, and (1,1)‐metabolites
generated from (R)‐TD (blue bars) and (S)‐TD (red bars) with 1 h of
hydrolysis and 60 min of hydroxylation. The values represent mean
value (n = 4) Æ standard deviation
FIGURE 7 Schematic of the preferred pathway for the generation of (1,1)‐metabolites involving hydrolysis of TD, hydroxylation of TD,
racemization of TD, hydroxylation of (1,0)‐metabolites, and hydrolysis of (0,1)‐metabolites

10
OGINO ET AL.
(1,0)‐metabolites of TD, is preferably generated. In the chiral
inversion reaction of TD, two water molecules are involved in
deprotonation of the steric center of TD and contribute to the
energy barrier decrease for chiral inversion compared with
the reaction in an isolated system.43 Our calculation suggests
that the hydrolysis reaction involving one water molecule
causes a nucleophilic attack on the phthalimide ring and
single TS during reaction. Briefly, oxidation at C‐5 and
C‐5′ in TD results in an increase in the ring opening energy
barrier about 17–18 kcal mol−1 compared with the case of
unmetabolized TD. Although the substitution of the hydroxyl
moiety on the glutarimide ring causes only slight changes in
the molecular structure, oxidation at C‐5 on glutarimide ring
or C‐5′ on the glutarimide ring greatly affected the energy
barrier for hydrolysis.
4 | CONCLUSION
We studied the metabolic reaction network of TD and its
derivatives using LC–MS/MS, chiroptical spectroscopy, and
theoretical computations. The temporal changes in the
amount of each metabolite generated with different reaction
times revealed that the three types of biotransformation
occurring in TD and its derivatives are related. The substitu-
tion of the hydrogen atom with a fluorine atom at the stereo-
genic center and substitution of the amino moiety on the
phthalimide ring accelerated hydrolysis and may also acceler-
ate di‐hydroxylation via nonenzymatic hydroxylation. We
found the preferred pathway in the generation of (1,1)‐
metabolites, where TD was first oxidized at C‐5′ on the
glutarimide ring, followed by the degradation of (0,1)‐
metabolites into (1,1)‐metabolites. We also found no
enantioselectivity in the progress of the metabolic reactions
of TD. The chirality of the various metabolites of TD was
confirmed to be almost the same as that of unmetabolized
TD. Theoretical calculation results of energy barriers demon-
strated that oxidation at C‐5 or C‐5′ caused the increase in the
hydrolysis energy barriers. The results obtained in the present
study will enable us to comprehend the complex metabolic
network of TD and its derivatives, as well as to reveal
undiscovered drug efficacies of various metabolites of TD
and its derivatives when combined with biological studies.
3.7 | Preferred pathway among multiple chiral
inversion, hydrolysis, and hydroxylation
The theoretical results shown in Figure S9 agree with the pre-
vious experimental results,42 but also seem to be inconsistent
with our experimental results mentioned above. Our results
suggest that (1,1)‐metabolites are generated via TD oxidation
followed by ring breakdown of (0,1)‐metabolite. Conversely,
our theoretical results indicate that hydrolysis of (0,1)‐metab-
olites is energetically unfavorable. This contradictory result
can be reasonably explained with the tradeoff relationship
among hydrolysis and hydroxylation of TD and metabolites.
Hydrolysis of (0,1)‐metabolites seems to be unfavorable,
but oxidation of ring‐breakdown product (1,0)‐metabolites
are expected to be even more unfavorable since the structural
changes via ring opening substantially affect the intermolec-
ular flexibility, which results in the loss of accessibility to
P450. Thus, we can conclude again that oxidation occurring
first and degradation occurring second is a possible prefera-
ble pathway in the generation of (1,1)‐metabolites
(Figure 7). According to the results shown in Figure 5, chiral-
ity of unmetabolized TD is conserved during the metabolic
reactions due to the stability to chiral inversion of the
metabolites. Furthermore, our results presented in Figure 6
demonstrate that the series of metabolic reactions proceed
in almost the same manner between two enantiomers. There-
fore, we infer that the chirality balance in the TD metabolites
should be determined by the rate constant for racemization of
unmetabolized TD, and the manner of subsequent metabolic
reactions is likely to be almost independent of each other
between two enantiomers (Figure 7). Although the molecular
behaviors relevant to chiral inversion of TD derivatives
and their metabolites remain unclear, our scheme of the
metabolic reaction procedures also allows a simple estima-
tion of chirality balance in the metabolites of TD and its
derivatives by considering first‐order racemization of TD
and its derivatives.
ACKNOWLEDGMENTS
The authors thank Hideko Koshima and Bart Kahr for fruitful
discussions and comments on the article. The authors also
thank Hiromi Sunaga for technical support on using the
analytical apparatus. The present study was supported by
JSPS KAKENHI Grant No. 14 J07106 from MEXT, Japan.
REFERENCES
1. Denny WA. Prodrug strategies in cancer therapy. Eur J Med Chem.
2001;36:577‐595.
2. Testa B. Prodrug research: Futile or fertile? Biochem Pharmacol.
2004;68:2097‐2106.
3. Kratz F, Müller IA, Ryppa C, Warnecke A. Prodrug strategies in
anticancer chemotherapy. ChemMedChem. 2008;3:20‐53.
4. Ekins S, Mestres J, Testa B. In silico pharmacology for drug
discovery: Methods for virtual ligand screening and profiling. Br J
Pharmacol. 2007;152:9‐20.
5. Ekins S, Mestres J, Testa B. In silico pharmacology for drug
discovery: Applications to targets and beyond. Br J Pharmacol.
2007;152:21‐37.
6. Liu Z, Fang H, Reagan K, et al. In silico drug repositioning–what we
need to know. Drug Discov Today. 2013;18:110‐115.
7. Domenici E, Bertucci C, Salvadori P, Wainer IW. Use of a human
serum albumin‐based high‐performance liquid chromatography

OGINO ET AL.
11
chiral stationary phase for the investigation of protein binding:
Detection of the allosteric interaction between warfarin and
gravity‐driven achiral chromatography: Mechanistic rationale and
implications. Chem Sci. 2015;6:1043‐1048.
benzodiazepine binding sites. J Pharm Sci. 1991;80:164‐166.
24. Soloshonok VA. Remarkable amplification of the self‐dispropor-
tionation of enantiomers on achiral‐phase chromatography
columns. Angew Chem. 2006;118:780‐783.
8. Kang XF, Cheley S, Guan X, Bayley H. Stochastic detection of
enantiomers. J Am Chem Soc. 2006;128:10684‐10685.
9. Boersma A, Baylay H. Continuous stochastic detection of amino
acid enantiomers with a protein nanopore. Angew Chem Int Ed.
2012;51:9606‐9609.
25. Ito T, Ando H, Suzuki T, et al. Identification of a primary target of
thalidomide teratogenicity. Science. 2010;327:1345‐1350.
26. Lopez‐Girona A, Mendy D, Ito T, et al. Cereblon is a direct
protein target for immunomodulatory and antiproliferative
activities of lenalidomide and pomalidomide. Leukemia.
2012;26:2326‐2335.
10. Blaschke G, Kraft HP, Fickentscher K, Kohler F.
Chromatographische racemattrennung von thalidomid und
teratogene wirkung der enantiomere. Arzneim ‐forsch. 1979;29:1640
11. Nishimura K, Hashimoto Y, Iwasaki S. (S)‐form of α‐methyl‐N(α)‐
phthalimidoglutarimide, but not its (R)‐form, enhanced phorbol
ester‐induced tumor necrosis factor‐α production by human leuke-
mia cell HL‐60: Implication of optical resolution of thalidomidal
effects. Chem Pharm Bull. 1994;42:1157‐1159.
27. Lu G, Middleton RE, Sun H, et al. The myeloma drug lenalidomide
promotes the cereblon‐dependent destruction of Ikaros proteins.
Science. 2014;343:305‐309.
28. Fischer ES, Böhm K, Lydeard JR, et al. Structure of the DDB1‐
CRBN E3 ubiquitin ligase in complex with thalidomide. Nature.
2014;512:49‐53.
12. Knoche B, Blaschke G. Investigations on the in vitro racemization
of thalidomide by high‐performance liquid chromatography.
J Chromatogr A. 1994;666:235‐240.
29. Chamberlain PP, Lopez‐Gilona A, Miller K, et al. Structure of the
human cereblon–DDB1–lenalidomide complex reveals basis for
responsiveness to thalidomide analogs. Nat Struct Mol Biol.
2014;21:803‐809.
13. Schumacher H, Smith RL, Williams RT. The metabolism of
thalidomide: The spontaneous hydrolysis of thalidomide in solution.
Br J Pharmacol. 1965;25:324‐337.
30. Takeuchi Y, Shiragami T, Kimura K, Suzuki E, Shibata N. (R)‐ and
(S)‐3‐fluorothalidomides: Isosteric analogues of thalidomide. Org.
Lett. 1999;1:1571‐1573.
14. Nakamura T, Nogichi T, Miyachi H, Hashimoto Y. Hydrolyzed
metabolites of thalidomide: Synthesis and TNF‐α production‐
inhibitory activity. Chem Pharm Bull. 2007;55:651‐654.
31. Yamamoto T, Suzuki Y, Ito E, Tokunaga E, Shibata N. Asymmetric
synthesis of both mirror images of 3'‐fluorothalidomide by
enantiodivergent fluorination using a single, cinchona alkaloid.
Org Lett. 2011;13:470‐473.
15. Stephens TD, Bunde CJW, Fillmode BJ. Mechanism of action
in
thalidomide
teratogenesis.
Biochem
Pharmacol.
2000;59:1489‐1499.
16. Ando Y, Price DK, Dahut WL, Cox MC, Reed E, Figg WD.
Pharmacogenetic associations of CYP2C19 genotype with in vivo
metabolisms and pharmacological effects of thalidomide. Cancer
Biol Ther. 2002;1:669‐673.
32. Lee CJ, Shibata N, Wiley MJ, Wells PG. Fluorothalidomide: A
characterization of maternal and developmental toxicity in rabbits
and mice. Toxicol Sci. 2011;122:157‐169.
33. Kumar S, Rajkumar V. Thalidomide and lenalidomide in the
17. Ando Y, Fuse E, Figg WD. Thalidomide metabolism by the CYP2C
treatment of multiple myeloma. Eur J Cancer. 2006;42:1612‐1622.
subfamily. Clin Camcer Res. 2002;8:1964‐1973.
34. Kumar G, Lau H, Laskin O. Lenalidomide: In vitro evaluation of the
metabolism and assessment of cytochrome P450 inhibition and
induction. Cancer Chemother Pharmacol. 2009;63:1171‐1175.
18. Lu J, Helsby N, Palmer BD, et al. Metabolism of thalidomide in
liver microsomes of mice, rabbits, and humans. J Pharm Exp Ther.
2004;310:571‐577.
35. Chen N, Wen L, Lau H, Surapaneni S, Kumar G. Pharmacokinetics,
metabolism and excretion of [14C]‐lenalidomide following oral
administration in healthy male subjects. Cancer Chemother
Pharmacol. 2012;69:789‐797.
19. Nakamura T, Nogichi T, Kobayashi H, Miyachi H, Hashimoto Y.
Mono‐ and dihydroxylated metabolites of thalidomide: Synthesis
and TNF‐α production‐inhibitory activity. Chem Pharm Bull.
2006;54:1709‐1714.
36. Hoffman M, Kasserra C, Reyes J, et al. Absorption, metabolism
and excretion of [14C] pomalidomide in humans following oral
administration. Cancer Chemother Pharmacol. 2013;71:489‐501.
20. Chowdhury G, Murayama N, Okada Y, et al. Human liver
microsomal cytochrome P450 3A enzymes involved in thalidomide
5‐hydroxylation and formation of a glutathione conjugate. Chem Res
Toxicol. 2010;23:1018‐1024.
37. Li Y, Zhou S, Hoffmann M, Kumar G, Palmisano M. Modeling and
simulation to probe the pharmacokinetic disposition of
pomalidomide R‐ and S‐enantiomers. J Pharmacol Exp Ther.
2014;350:256‐272.
21. Yamazaki H, Suemizu H, Igaya S, et al. In vivo formation of a
glutathione conjugate derived from thalidomide in humanized
uPA‐NOG mice. Chem Res Toxicol. 2011;24:287‐289.
38. Ogino Y, Asahi T. Model of complex chiral drug metabolic
systems and numerical simulation of the remaining chirality toward
analysis of dynamical pharmacological activity. J Theor Biol.
2015;373:117‐131.
22. Chowdhury G, Shibata N, Yamazaki H, Guengerich FP. Human
cytochrome P450 oxidation of 5‐hydroxythalidomide and
pomalidomide, an amino analogue of thalidomide. Chem Res
Toxicol. 2014;27:147‐156.
23. Maeno M, Tokunaga E, Yamamoto T, et al. Self‐disproportionation
39. Frisch MJ, Trucks GW, Schlegel HB, et al. Gaussian 09, Revision
D.01. Wallingford CT: Gaussian, Inc.; 2009.
of enantiomers of thalidomide and its fluorinated analogue via

12
OGINO ET AL.
40. Otogawa K, Ishikawa K, Shiro M, Asahi T. Crystal structure of (S)‐
Supporting Information
4‐carbamoyl‐4‐(1,3‐dioxo‐isoindolin‐2‐yl)butanoic
Crystallogr E. 2015;71:107‐109.
acid.
Acta
Additional Supporting Information may be found online in
the supporting information tab for this article.
41. Yamamoto T, Shibata N, Sukeguchi D, et al. Synthesis, configura-
tional stability and stereochemical biological evaluations of (S)‐
and (R)‐5‐hydroxythalidomides. Bioorg Med Chem Lett.
2009;19:3973‐3976.
How to cite this article: Ogino Y, Tanaka M,
Shimozawa T, Asahi T. LC–MS/MS and chiroptical
spectroscopic analyses of multidimensional metabolic
systems of chiral thalidomide and its derivatives.
Chirality. 2017. https://doi.org/10.1002/chir.22683
42. Meyring M, Muhlbacher J, Messer K, et al. In vitro biotransfor-
mation of (R)‐and (S)‐thalidomide: Application of circular
dichroism spectroscopy to the stereochemical characterization of
the hydroxylated metabolites. Anal Chem. 2002;74:3726‐3735.
43. Tian C, Xiu P, Meng Y, Zhao W, Wang Z, Zhou R.
Enantiomerization mechanism of thalidomide and the role of water
and hydroxide ions. Chem A Eur J. 2012;18:14305‐14313.