Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298

Contents lists available at SciVerse ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

journal homepage: www.elsevier.com/locate/jpba

Structure and stability of warfarin-sodium inclusion complexes formed with

permethylated monoamino-␤-cyclodextrin
Gábor Tárkányi a,∗ , Krisztina Németh b , Réka Mizsei a , Orsolya Tőke a , Júlia Visy b , Miklós Simonyi b ,
László Jicsinszky c , Julianna Szemán c , Lajos Szente c
a
Institute of Organic Chemistry, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Pusztaszeri út 59-67, H-1025 Budapest, Hungary
b
Institute of Molecular Pharmacology, Research Centre for Natural Sciences, Hungarian Academy of Sciences, Pusztaszeri út 59-67, H-1025 Budapest, Hungary
c
CycloLab Cyclodextrin Res & Dev Lab Ltd, Illatos út 7., H-1097 Budapest, Hungary

a r t i c l e i n f o a b s t r a c t

Article history: Inclusion complexes of warfarin enantiomers with permethylated monoamino-␤-cyclodextrin

Received 16 July 2012 (PMMABCD) were characterized using CE and 1 H NMR spectroscopy in aqueous solution. These tech-
Received in revised form 31 August 2012 niques gave complementary information on the stability and the structure of the diastereomeric
Accepted 2 September 2012
host–guest inclusion complexes. The stability constants were determined from CE experiments in a wide
Available online 11 September 2012
pH range. Change in the migration order on the variation of the pH was observed. 1 H NMR assignments
have been established for the seven non-equivalent carbohydrate units of the host in the complex at pH
Keywords:
7–9. Specific H–H distance restraints were obtained from NOESY experiments and were introduced into
Cationic selector
Complex stability constant
molecular modeling to establish the geometry of the inclusion complexes. It was found that the open side
Enantiomer migration order reversal chain warfarin enters the cavity from the primary side of the CD. The orientation of the coumarin ring
Primary side inclusion within the cavity has the same preference for the two warfarin enantiomers owing to an ionic interaction
NOE distance restraints with the amino group of the CD. Accordingly, enantioselectivity at pH 8.5 arises from the difference in
the CH/␲ interactions between warfarin aromatics and the manifold of CH groups of the CD.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction warfarin enantiomers [7–9]. Our recent work details separations at

Warfarin (3-(␣-phenyl-␤-acetylethyl)-4-hydroxycoumarin) is a
Vitamin K antagonist, inhibiting Vitamin K epoxide reductase and
thereby decreasing blood coagulation by preventing the Vitamin K-
dependent synthesis of blood-clotting proteins [1]. In the form of
its sodium salt (Coumadin® , Marevan® ) it is a widely prescribed
anticoagulant for the prevention of thrombosis and embolism.
Although currently administered as the racemate, activity and
metabolism are markedly dissimilar for the two enantiomers, the
(S) enantiomer being more active than the (R) enantiomer by a
factor of 2–5 [2].
Former analytical studies have mainly focused on the success-
ful separation of warfarin enantiomers for reasons of quantitation
in biological fluids [3–6]. Among the many applications it became
evident that warfarin enantiomers can be effectively separated
on ␤-CD functionalized stationary phases under reversed phase
conditions using HPLC. In capillary electrophoresis (CE) methyl-
ated amino-␤-CD derivatives were used as successful selectors of
∗ Corresponding author. Tel.: +36 1 4381106; fax: +36 1 4381107.
E-mail address: tarkanyi.gabor@ttk.mta.hu (G. Tárkányi).
low pH (pH < 4) [10]. As a continuation of our research in design-
ing separations by exploiting stereospecific molecular interactions
we became interested in characterizing the stability and struc-
ture of the diastereomeric complexes of warfarin enantiomers in
the high pH region (7 < pH). Cyclodextrins are well known chiral
selectors among which monofunctional single isomer derivatives
are excellent candidates for molecular recognition studies [11]. In
the present work we describe results obtained for the interaction
between warfarin sodium and permethylated 6-monoamino-6-
monodeoxy-␤-cyclodextrin (PMMABCD) [12,13]. As reported by
many groups warfarin exists in various isomeric forms depend-
ing on the polarity of the solvent [14–16]. The pH is expected to
have influence on the nature of interaction between the CD and
analyte in water [17,18]. We made the assumption that the basic
amino group of PMMABCD can offer an interaction site for the open
form of warfarin bearing ionizable OH groups (Scheme 1) [19].
With this mechanistic proposal in mind we aimed at probing and
determining the nature of molecular recognition between host and
guest using the results of CE and 1 H NMR spectroscopy in aque-
ous solution. These techniques were chosen to give complementary
information on the structure and stability of the host and guest
molecules in the inclusion complex.

0731-7085/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jpba.2012.09.003
 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298
Scheme 1. Dominant isomers of warfarin.
2. Materials and methods
2.1. Materials
Background electrolyte buffer components Suprapure sodium
acetate, sodium dihydrogen phosphate, sodium hydroxide, glacial
acetic acid, methanol and acetonitrile were purchased from Merck
GmbH (Darmstadt, Germany).
Racemic warfarin sodium isopropanol clathrate [20,21] was
purchased from Sigma–Aldrich. S-warfarin and R-warfarin enan-
tiomers were prepared as previously described [22]. Permethy-
lated 6A -monoamino-6A -monodeoxy-␤-cyclodextrin hydrochlo-
ride (PMMABCD·HCl) [23] and heptakis (2,3,6-tri-O-methyl)-␤-
cyclodextrin (TRIMEB) and ␣, ␤, ␥-cyclodextrins are the products
of CycloLab Ltd., Budapest, Hungary.
2.2. Capillary electrophoresis
Capillary electrophoresis was performed with an Agilent Cap-
illary Electrophoresis 3D CE system applying bare fused silica
capillary of 64.5 cm total and 56 cm effective length with bubble
cell and 50 ␮m I.D. (Agilent Technologies, Santa Clara, CA, USA).
On-line UV absorption at 209 nm and 308 nm was detected by
DAD UV–vis detector. The capillary was thermostated at 20 ◦ C.
Britton–Robinson buffer prepared from 40 mM borate, 40 mM
acetate and 40 mM phosphate in a mixture of (1:2:2, v/v/v) ratio
was applied as background electrolyte (BGE) at pH values adjusted
by NaOH. Between measurements, the capillary was rinsed by 1 M
NaOH, 0.1 M NaOH and distilled water, subsequently for 1 min and
with BGE for 8 min. Warfarin samples were dissolved in absolute
ethanol and further diluted with distilled water. Racemic warfarin
(2 × 10−6 M) was spiked with the pure (R) enantiomer (10−6 M) and
injected by 5 × 103 Pa pressure for 3 s. Runs were performed in the
positive-polarity mode with 30 kV. The resolution (Rs ) of warfarin
enantiomers is given by the following equation [9]:
1.18 · (t1 − t2 )
Rs = (1)
w(0.5)1 + w(0.5)2
where t is the migration time of the enantiomers (1, 2 in lower
index), w(0.5) is the peak width at half height.
To calculate the apparent complex stability constant (K ) of war-
farin enantiomers to the selector CDs, the mobilities of the analytes
in the absence (free ) and in the presence (eff ) of CD in five con-
centrations ([CD]) were determined from the following equation in
the range of 5–30 mM and 10–30 mM for PMMABCD and TRIMEB,
respectively (y-reciprocal calculation)
[CD] 1 1 [CD]
= · + (2)
eff − free compl − free K  compl − free
By plotting [CD]/(eff − free ) vs. [CD], the ratio of slope/intercept
value of the regression line (R2 > 0.99) equals the apparent stability
293
constant [24,25]. The mobilities of the complexes of warfarin enan-
tiomers (compl ) formed with PMMABCD were calculated from Eq.
(2). RSD values of the measured and calculated parameters were
under 10%.
2.3. NMR spectroscopy
The NMR experiments were carried out on a 600 MHz (for 1 H)
Agilent/Varian NMR SYSTEMTM spectrometer by using 5 mm direct
detection 15 N–31 P/{1 H–19 F} and 1 H{13 C/15 N} probes equipped
with Z pulse field gradient. All spectra were recorded with 80 MHz
sampling rate using the DirectDigitalTM receiver technology. The
assignments of the resonances involved homo- and heteronuclear
through-bond correlations (1 H gDQFCOSY, 1 H zqTOCSY, 1 H–13 C
gHSQC and 1 H–13 C gHMBC). 1 H NOE enhancements were nega-
tive at 278 K and intermolecular spatial contacts were effectively
probed by 1 H zqNOESY experiments (zq refers to Zero Quantum
filtered experiments providing clean J-coupling patterns of multi-
plets). Solvent presaturation was used during the recycle delay (2 s)
of the 2D experiments. 4096 complex data points were acquired
in F2 and 320 complex data points in the F1 dimension. Spectral
widths of 10 kHz were used in both dimensions. 40 and 150 ms mix-
ing times ( mix ) using spinlock field strengths (␥B1 ) of 7 kHz were
used for the 1 H zqTOCSY experiments. 4096 complex data points
were acquired in F2 and 256 complex data points in F1 dimen-
sion for the heteronuclear 2D experiments (1 H–13 C gHSQC and
1 H–13 C gHMBC). The 1 H–13 C gHSQC experiments were optimized
for 1 JCH = 140 Hz. Data were multiplied by Gaussian weighting func-
tions and zero filled to 8192 × 4096 matrixes. Digital resolution in
the F1 dimension was doubled by twofold linear prediction. An
automated polynomial baseline correction was used [26]. Spectra
were referenced to the external reference signal of 1 mM DSS in D2 O
(ı1 H,DSS = 0.00 ppm and ı13 C,DSS = 0.0 ppm). The isopropyl-alcohol
signal at 4.01 ppm was used as a secondary reference in assign-
ments of 2D NMR. NMR data were acquired with the VnmrJ 2.2C
software by using the Varian standard spectrometer pulse sequence
library. Spectra were processed by ACD/Labs 12.0 software pur-
chased from Chemicro Ltd. Hungary.
2.4. Molecular modeling
Molecular modeling was performed in HyperChem 8.0 Profes-
sional software using the CHARMM force-field [27,28]. The initial
geometries of the inclusion complexes were created by manu-
ally positioning the coumarin ring of warfarin within the cavity
to fulfill two NOEs detected for W5-A5 and W8-D5 in connection
to Fig. 3. Distance restraints of 1 kcal mol−1 Å−2 were applied for
each experimentally observed H–H contact of warfarin W5 and W8
protons as detailed in Section 3.4. This was followed by a geome-
try optimization using a 2 kcal mol−1 Å−2 force-constant for each
observed cyclodextrin H1(i) –H4(i+1) interresidue NOEs, where (i)
294 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298

compared to the native CDs. Alternatively, the molecular recog-

nition process was also probed by the chemical shift dispersion of
the cyclodextrin signals in 1 H NMR. Addition of warfarin-sodium
to the host induced large chemical shift changes for the PMMABCD
O-methyl resonances compared to the uncomplexed form (Fig. 1).
These early qualitative findings from NMR have prompted
us to investigate the nature of enantioselective complexation
between warfarin and the monosubstituted PMMABCD. For prac-
tical reasons associated with solubility problems in NMR, capillary
electrophoresis (CE) has offered a more amenable way to quanti-
tatively determine the apparent complex stabilities (K values). In
the low pH region we encountered intolerable sensitivity problems
in NMR which prevented us from measuring accurate variations
of warfarin signals necessary for determination of K values [31].
Nonetheless, the methodologies offered by NMR spectroscopy were
exploited in the structure determination of the diastereomeric
complexes in the high pH region (7 < pH).

Fig. 1. 1 H NMR spectra of PMMABCD·HCl without (a) and with (b) 1:1 molar ratio
warfarin-sodium in D2 O (45 mM, +15 ◦ C, 399.9 MHz).
and (i + 1) represent neighboring sugar units. Conjugate gradient
(Polak–Ribiere algorithm) optimizations were performed in vacuo
until reaching 0.5 kcal/(Å mol) RMS energy gradient.
3. Results and discussion
3.1. Enantiorecognition in 1 H NMR
As reported by many authors the existence of the associa-
tion phenomena between selector and selectands can routinely
be assessed by observing the chemical shift non-equivalence of
the 1 H NMR signals for the racemic analyte in the presence of the
CD [29,30]. We monitored this effect for the methyl resonances of
racemic warfarin measured in D2 O in the presence of PMMABCD
as well as the three native CDs (see Supporting Information Figure
S1). Enantiomeric splitting was larger for the ionic PMMABCD
3.2. Complex stabilities and apparent binding constants
calculated from CE data
In our previous study good chiral resolution of warfarin has
been demonstrated by CE at low pH using PMMABCD as a selec-
tor [10]. Here, we report that the high pH range suitable for NMR
analysis shows good enantioseparations, too (Fig. 2). Table 1 sum-
marizes the apparent complex stability constants as determined
in the pH range of 4.0–9.5 together with the uncharged TRIMEB
CD analog. For PMMABCD it was found that the efficiency of res-
olution (Rs ) and the magnitude of K decreased with increasing
pH (Table 1 and Fig. 2). In our setup – independently from the
applied pH – the enantiomer forming the more stable complex was
accelerated with a higher degree. Interestingly, reversed migra-
tion order of the enantiomers (EMO) was detected at pH < 5.5
compared to pH 8.5. At pH 4.0 R-warfarin is the faster migrating
enantiomer according to its higher affinity to the PMMABCD selec-
tor. This is in line with our previous HPLC measurements at pH 4.0

Fig. 2. pH dependence of resolution and migration order of warfarin enantiomers in the presence of 15 mM PMMABCD.

Table 1
Chiral resolution (Rs ), migration order (EMO) and apparent complex stability constants K in [M−1 ] of warfarin enantiomers in function of pH in the presence of PMMABCD
and TRIMEB.

pH Rs EMO KS KR KS KR

PMMABCD (15 mM) TRIMEB (15 mM) PMMABCD TRIMEB PMMABCD TRIMEB

4.0 4.27 0.93 R,S R,S 85 ± 1 119 ± 2 62 ± 5 92 ± 6

5.5 0 1.81 R,S 42 ± 4 48 ± 5
6.5 3.56 0 S,R 75 ± 1 61 ± 1 36 ± 1 32 ± 2
7.5 2.84 1.23 S,R S,R
8.5 2.87 1.39 S,R S,R 65 ± 5 59 ± 8 33 ± 3 29 ± 2
9.5 1.81 1.49 S,R S,R
 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298 295

Fig. 3. 1
H NMR assignments in zqTOCSY and zqNOESY experiments at 9 mM RS-warfarin-sodium/PMMABCD complex (D2 O, +15 ◦ C).

where R-warfarin – having the higher binding constant – eluted

slower than the S enantiomer on the PMMABCD functionalized chi-
ral stationary phase [10]. The compl,R was also higher compared
to that of the S-enantiomer (compl,R = 54.6 × 10−6 ± 0.7 × 10−6
vs. compl,S = 48.3 × 10−6 ± 0.8 × 10−6 in cm2 V−1 s−1 ) suggesting a
difference in the structure of the diastereomeric warfarin hemike-
tal/PMMACD complexes.
In the high pH region the migration time is less
(tmigr ) and the derived apparent complex stability con-
stant (KS ) of S-warfarin exceeds the respective values of
its enantiomer. Meanwhile, the mobilities of the warfarin
inclusion complexes were found to be identical within
experimental error (compl,R = −5.0 × 10−6 ± 2.7 × 10−6 ,
compl,S = −5.1 × 10−6 ± 1.5 × 10−6 in cm2 V−1 s−1 ) at pH 8.5. Simul-
taneous change in EMO and reversal of complex stability reinforces
the structural change in the PMMABCD/warfarin R and S complexes
associated with the shifted equilibrium between open and closed
ring forms of warfarin as a function of the pH (Scheme 1). The four Scheme 2. 1 H NMR “walkthrough” strategy in assignments of homonuclear 2D NMR
experiments.
possible hemiketal forms [14] are conformationally more rigid
than the open side chain enantiomers of warfarin and may interact
3.3. NMR resonance assignments of the host
quite differently with PMMABCD resulting in a mobility difference
compl,R =/ compl,S between the complexes at pH 4. On the con-
As introduced earlier we aimed to determine the chemical struc-
trary, the similarity of the complex mobilities (compl,R ≈ compl,S )
ture of the diastereotopic complexes of warfarin-PMMABCD in
in the high pH region suggests that both enantiomers of warfarin
the high pH region by NMR. Spatial contacts between protons of
may accommodate the cavity in a similar fashion.
the warfarin guest molecule and those of the CD located in the
296 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298

cavity (H3, H5) provide the basis of such structure determination Table 2
Proton and carbon chemical shifts of PMMABCD in the presence of 45 mM RS-
[29]. In many cases this process is limited by the low dispersion
warfarin-sodium (D2 O, +10 ◦ C).
of the individual sugar signals for almost all CDs (native, sym-
metrically or random substituted). Below we demonstrate that Residue Position ıH ıC
the host/guest complex of a single isomer monosubstituted CD – 1 A1 5.280 99.33
such as PMMABCD – is suitable to establish the complete 1 H NMR 4 D1 5.258 100.71
assignment of the host before analyzing the spatial H–H contacts 7 G1 5.236 100.92
2 B1 5.202 100.20
(NOE) with the guest molecules. By assigning the seven chem-
5 E1 5.189 100.68
ically non-equivalent glucopyranosyl units (denoted with letters 3 C1 5.173 100.51
A–G) in PMMABCD one can determine the orientation of the guest 6 F1 5.152 100.48
within the CD cavity as the spatial contacts are localized to par- 1 A5 4.164 71.32
5 E6x 3.945 73.17
ticular protons of the chemically non-equivalent sugar units. First,
4 D5 3.883 73.57
we monitored the chemical shift dispersion of the H1A–G anomeric 5 E5 3.817 73.47
protons of the CD in the one dimensional 1 H NMR spectrum by titra- 4 D6x 3.814 73.17
tions in order to achieve optimal signal dispersion in 2D TOCSY NMR 2 B6x 3.774 72.80
experiments (Figure S4). Both the ratio and the total concentration 5 E4 3.761 81.33
4 D4 3.760 81.25
of the components were changed and the final conditions were
7 G5 3.752 73.57
set by varying the temperature (Figures S2–S3). Spectral overlap of 1 A3 3.751 83.08
H1A–G protons diminished in the range 10–15 ◦ C. We proceeded by 2 B4 3.743 79.18
the 1 H NMR assignments using the combination of homonuclear 2D 4 D6y 3.697 73.17
7 G4 3.661 80.43
TOCSY and NOESY experiments [32]. The “walkthrough” strategy
4 D3 3.652 83.55
of the assignment is depicted in Scheme 2 showing through-bond 2 (3-OMe)B 3.641 63.24
and through-space correlations. By varying the mixing time of 3 C4 3.636 79.06
the 1 H TOCSY experiment, the intraunit (H-1)i /(H-4)i through- 1 (3-OMe)A 3.632 63.24
bond correlations could be identified within each sugar ring. The 6 F4 3.631 80.43
5 (3-OMe)E 3.626 63.24
through-space correlations available from the 1 H NOESY spectra
5 E3 3.623 83.78
allowed us to establish the seven interunit (H-1)i /(H-4)i + 1 correla- 1 A4 3.620 83.98
tion between the sugar rings. Ring-A had unique C(1)H, C(5)H and 5 E6y 3.616 73.17
C(6)H2 chemical shifts in the 1 H–13 C HSQC experiments which clar- 4 (3-OMe)D 3.599 63.04
6 F6x 3.585 72.87
ified its identity at early stages of the assignment (Figures S5–S6).
1 A6x 3.577 42.20
The O–Me resonances were identified on the basis of their three 6 (3-OMe)F 3.576 62.86
bond 1 H–13 C J-coupling correlations from 1 H–13 C gHMBC experi- 7 G3 3.568 83.27
ments. The warfarin protons were assigned on the basis of similar 3 C3 3.561 83.72
two-dimensional 1 H–1 H and 1 H–13 C experiments. In time, deuter- 7 (3-OMe)G 3.551 62.92
2 B5 3.548 73.69
ation was found in the side-chain [33]. The full list of the 1 H, 13 C
3 (3-OMe)C 3.544 62.29
NMR assignments for the host is given in Table 2. No information 2 B3 3.525 83.72
regarding the measured spatial 1 H–1 H contacts between host and 7 G6x 3.517 73.08
guest were considered during the above assignment procedure. 4 (2-OMe)D 3.505 60.58
1 (2-OMe)A 3.503 60.91
7 G6y 3.497 73.08
3.4. Establishing the inclusion geometry of the complexes 5 (2-OMe)E 3.497 60.99
2 (2-OMe)B 3.495 61.21
Having established the NMR assignments of the host we turned 7 (2-OMe)G 3.495 60.53
our attention to the NOESY crosspeaks encoding the spatial 1 H–1 H 6 F3 3.474 83.63
6 (2-OMe)F 3.473 60.87
contacts between the host and the guest at pH 8.5. It is well doc-
3 (2-OMe)C 3.472 60.55
umented in the literature that inclusion formation is manifested 2 B6y 3.436 72.80
in the appearance of strong NOEs or ROESY cross-peaks between 3 C5 3.409 72.87
the inner H3 and H5 cyclodextrin protons and those of the ana- 3 C6x 3.393 72.80
lyte [34,35]. We looked up these correlations for the enantiomers 5 (6-OMe)E 3.384 61.13
4 (6-OMe)D 3.364 61.13
of warfarin from experiments performed on the racemate. In case 2 (6-OMe)B 3.357 61.44
of the monosubstituted PMMABCD we exploited the fact to see 1 A2 3.355 83.02
each inner H3A–G , H5A–G correlations of the seven individual sugar 6 F5 3.327 73.14
rings. Fig. 3 shows the appropriate spectral region relevant for the 6 F6y 3.327 72.87
7 G2 3.315 82.55
discussion.
4 D2 3.296 82.84
The TOCSY experiment (Fig. 3 top) identifies the chemical shift 7 (6-OMe)G 3.285 61.09
positions of the CD and the appropriate off-diagonal section of 6 (6-OMe)F 3.277 61.09
the NOESY experiment (Fig. 3 bottom) contains the intermolecular 3 C6y 3.276 72.80
H–H crosspeaks. We found that warfarin favors the inclusion of the 5 E2 3.273 83.14
2 B2 3.264 83.22
coumarin ring into the CD cavity in the high pH range (7 < pH < 9). To 3 C2 3.236 83.02
our surprise, exclusive H–H correlations between warfarin and the 6 F2 3.227 83.22
individual sugar protons were found suggesting a preferred orien- 1 A6y 3.223 43.20
tation of warfarin in the CD cavity. Proton W5 of warfarin is close 3 (6-OMe)C 3.148 61.09
in space to H3A , H5A , H3G , H5G of the CD. Parallel, the W8 pro-
ton gave contacts mainly with the H3D , H5D , H3E , H5E hydrogens.
Interestingly, these contacts exist for both enantiomers of warfarin
according to the 2D NMR spectra of the racemate. We established
a simplified geometrical model to depict these findings (Fig. 4).
 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298
Fig. 4. Spatial 1 H–1 H proximities and induced shielding ring currents for the inclu-
sion model of warfarin-Na/PMMA system.
Penetrating preferably from the primary side, the W4-oxygen of
warfarin is attracted by the amino group of PMMA to form a classical
ion-pair [PMMA-NH3 + ][W-O(4) − ]. According to molecular model-
ing this interaction – being available for both enantiomers – is held
responsible for the restricted motion of the deprotonated coumarin
moiety within the cavity.
3.5. Proving the mode of inclusion from chemical shift data
Inclusion of the coumarin part from the primary side of the CD
was further verified from the analysis of 1 H NMR chemical shift
dispersion data. In its uncomplexed form the methyl resonances of
PMMABCD fit in a narrow chemical shift range in the three groups
OMe protons (2, 3 or 6) in the sugar rings. In the presence of war-
farin sodium however, the (6-OMe)B–G methoxy groups located at
the primary side have shown greater chemical shift dispersion than
those at the secondary side (2-OMe)A–G suggesting guest-entry
from the primary side.
Fig. 5. Inclusion complexes of warfarin optimized in agreement with the measured 1 H–1 H spatial proximities from NOESY experiments. View from the secondary side of
the CD.
297
Fig. 6. 1 H NMR chemical shifts of H5A–G protons of the host as a function of tem-
perature.
To reinforce the orientation of the coumarin ring assessed pre-
viously on the basis of NOESY data the chemical shift dispersion
of the H5A–G cyclodextrin protons located within the cavity were
also analyzed. It was found that H5C and H5F experienced upfield
shift relative to the uncomplexed state while the H5A , H5D , H5G
protons shifted downfield in the presence of warfarin (see also
Fig. 3). This is the consequence of the well-known ring current effect
of aromatic systems which induces shielding (lower shifts) above
and below the ring plane (CH/␲ effect [36]) and causes deshield-
ing within the aromatic ring plane. To visualize this in Fig. 6 we
plotted the variation of the chemical shifts of H5 protons with the
temperature. The chemical shift difference (e.g.  = |ıH5D − ıH5F |)
of shielded and deshielded protons increases as the temperature
lowers since both enantiomers of warfarin reside for more time
in the cavity. The fitted polynomial curves in Fig. 6 help quan-
tifying the relative electron density (i.e. chemical shift) around
the H5 protons. Smaller chemical shifts represent higher electron
densities (shielded protons). Because the 1 H NMR assignments of
Section 3.3 were not based on the interpretation of chemical shifts,
this temperature dependent monitoring of the variation of H5
298 G. Tárkányi et al. / Journal of Pharmaceutical and Biomedical Analysis 72 (2013) 292–298
chemical shifts is a strong, independent proof of our inclusion
model shown in Figs. 4–5.
4. Conclusions
In summary, we characterized the inclusion complexes of
warfarin-sodium enantiomers formed with the PMMABCD in the
high pH region (9 < pH < 7) with respect to their relative stabili-
ties and molecular structure. Owing to the monosubstituted, single
isomer nature of the chiral selector we were able to deduce the
structure of the complexes in solution at the atomic level of detail.
CE binding studies indicated good enantioseparation but no elec-
trophoretic mobility difference between the warfarin/PMMABCD
diastereomeric complexes at pH 8.5. Parallel, we found in NMR that
both enantiomers of ring-open form of warfarin enter the primary
side of the cyclodextrin with their coumarin moieties. The inclusion
complexes are stabilized by CH/␲ interactions as well as the ion-
pair formation between the W4 oxygen of warfarin and the single
amino group of the cyclodextrin.
We may therefore conclude that in the high pH region where
the open chain form of warfarin is more flexible the enantiomers
accommodate the cavity in a similar fashion. The diastereomeric
complexes differ by the opposite orientation of the warfarin side-
chain groups located at the primary side rim of the CD showing
difference in the network of CH/␲ interactions. This cooperativ-
ity within the network of CH/␲ hydrogen bonding interactions
makes the permethylated monoamino-␤-CD a rather efficient
enantioselective selector for warfarin both in CE and NMR. Since
significantly more information about the geometry of the inclu-
sion complexes can be drawn from the spectra of single isomer
cyclodextrins it is expected that future enantiorecognition stud-
ies supported by high-field NMR instrumentation will utilize
PMMABCD as a potent chiral selector of pharmaceutically active
compounds.
Acknowledgments
The financial support of the Chemical Research Center of HAS
and those of the Hungarian GVOP-3.2.1.-2004-04-0210/3.0 and Jed-
lik Ányos Grant NKFP 07-A3-2008-0211 NATURSEP projects are
acknowledged.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jpba.2012.09.003.
References
[1] N.A. Limdi, D.L. Veenstra, Warfarin pharmacogenetics, Pharmacotherapy 28
(2008) 1084–1097.
[2] R.H. White, H. Zhou, P. Romano, D. Mungall, Changes in plasma warfarin levels
and variations in steady-state prothrombin times, Clin. Pharmacol. Ther. 58
(1995) 588–593.
[3] D.W. Armstrong, T.J. Ward, R.D. Armstrong, T.E. Beesley, Separation of drug
stereoisomers by the formation of beta-cyclodextrin inclusion complexes, Sci-
ence 232 (1986) 1132–1135.
[4] P. Gareil, J.P. Gramond, F. Guyon, Separation and determination of warfarin
enantiomers in human plasma samples by capillary zone electrophoresis using
a methylated ␤-cyclodextrin-containing electrolyte, J. Chromatogr. 615 (1993)
317–325.
[5] X.F. Zhu, B.C. Lin, U. Epperlein, B. Koppenhoefer, Enantiomeric resolution
of some nonsteroidal antiinflammatory and anticoagulant drugs using beta-
cyclodextrins by capillary electrophoresis, Chirality 11 (1999) 56–62.
[6] B. Chankvetadze, N. Burjanadze, J. Crommen, G. Blaschke, Enantioseparation
of warfarin by capillary electrophoresis with UV and LIF detection using sin-
gle and dual cyclodextrin type chiral selectors, Chromatographia 53 (2001)
S296–S301.
[7] B. Chankvetadze, G. Blaschke, Enantioseparations in capillary electromigra-
tion techniques: recent developments and future trends, J. Chromatogr. A 906
(2001) 309–363.
[8] S. Fanali, E. Camera, Use of methylamino-beta-cyclodextrin in capillary elec-
trophoresis. Resolution of acidic and basic enantiomers, Chromatographia 43
(1996) 247–253.
[9] S. Fanali, Enantiomeric separations by electrophoretic techniques, J. Chro-
matogr. A 792 (1997) 227–267.
[10] G. Varga, G. Tárkányi, K. Németh, R. Iványi, L. Jicsinszky, O. Tőke, J. Visy, L. Szente,
J. Szemán, M. Simonyi, Chiral separation by a monofunctionalized cyclodextrin
derivative: From selector to permethyl-beta-cyclodextrin bonded stationary
phase, J. Pharm. Biomed. Anal. 51 (2010) 84–89.
[11] J. Szejtli, Introduction and general overview of cyclodextrin chemistry, Chem.
Rev. 98 (1998) 1743–1753.
[12] R. Iványi, L. Jicsinszky, Z. Juvancz, Permethyl monoamino ␤-cyclodextrin new
chiral selective agents for capillary electrophoresis, Chromatographia 53 (2001)
166–172.
[13] R. Iványi, L. Jicsinszky, Z. Juvancz, Chiral separation of pyrethroic acids with
single isomer permethyl monoamino ␤-cyclodextrin selector, Electrophoresis
22 (2001) 3232–3236.
[14] W.R. Porter, Warfarin: history, tautomerism and activity, J. Comput. Aided Mol.
Des. 24 (2010) 553–573.
[15] H. Henschel, B.C.G. Karlsson, A.M. Rosengren, I.A. Nicholls, The mechanistic
basis for warfarin’s structural diversity and implications for its bioavailability,
J. Mol. Struct. Theochem. 958 (2010) 7–9.
[16] I.A. Nicholls, B.C.G. Karlsson, A.M. Rosengren, H. Henschel, Warfarin: an
environment-dependent switchable molecular probe, J. Mol. Recognit. 23
(2010) 604–608.
[17] E.J. Valente, E.C. Lingafelter, W.R. Porter, W.F. Trager, Structure of warfarin in
solution, J. Med. Chem. 20 (1977) 1489–1493.
[18] E.J. Valente, W.R. Porter, W.F. Trager, The conformation of selected
3-substituted 4-hydroxycoumarins in solution by NMR: warfarin and phen-
procoumon, J. Med. Chem. 21 (1978) 231–234.
[19] M. He, K.R. Korzekwa, J.P. Jones, A.E. Rettie, W.F. Trager, Structural forms of
phenprocoumon and warfarin that are metabolized at the active site of CYP2C9,
Arch. Biochem. Biophys. 372 (1999) 16–28.
[20] (a) C.H. Schroeder, L.K. Paul, Warfarin sodium, US patent 3,077,481 (1963).;
(b) N. Weiner, R. Park, M. Johnson, C.H. Schroeder, Preparation of crystalline
warfarin sodium-isopropyl alcohol complex, US patent 3,246,013 (1966).
[21] A.R. Sheth, V.G. Young Jr., D.J.W. Grant, Warfarin sodium 2-propanol solvate,
Acta Crystallogr. E58 (2002) m197–m199.
[22] B.D. West, S. Preis, C.H. Schroeder, Studies on the 4-hydroxycoumarins. XVII.
The resolution and absolute configuration of warfarin, J. Am. Chem. Soc. 83
(1961) 2676–2679.
[23] L. Jicsinszky, R. Iványi, Catalytic transfer hydrogenation of sugar derivatives,
Carbohydr. Polym. 45 (2001) 139–145.
[24] Y. Tanaka, S. Terabe, Estimation of binding constants by capillary electrophore-
sis, J. Chromatogr. B 768 (2002) 81–92.
[25] G.K.E. Scriba, Fundamental aspects of chiral electromigration techniques and
application in pharmaceutical and biomedical analysis, J. Pharm. Biomed. Anal.
55 (2011) 688–701.
[26] D.E. Brown, Improved baseline recognition and modeling of FT NMR spectra, J.
Magn. Reson. Ser. A 114 (1995) 268–270.
[27] B.R. Brooks, R.E. Bruccoleri, B.D. Olafson, D.J. States, S. Swaminathan, M. Karplus,
CHARMM – a program for macromolecular energy, minimization, and dynam-
ics calculations, J. Comput. Chem. 4 (1983) 187–217.
[28] A.V. Eliseev, H.J. Schneider, Molecular recognition of nucleotides, nucleosides,
and sugars by aminocyclodextrins, J. Am. Chem. Soc. 116 (1994) 6081–6088.
[29] G. Tárkányi, Quantitative approach for the screening of cyclodextrins by
nuclear magnetic resonance spectroscopy in support of chiral separations in
liquid chromatography and capillary electrophoresis – enantioseparation of
norgestrel with alpha-, beta- and gamma-cyclodextrins, J. Chromatogr. A 961
(2002) 257–276.
[30] M.C. Ringo, C.E. Evans, Liquid chromatography as a measurement tool:
chromatographic assessment of chiral recognition, Anal. Chem. 70 (1998)
315A–321A.
[31] L. Fielding, Determination of association constants (Ka) from solution NMR
data, Tetrahedron 56 (2000) 6151–6170.
[32] E.T. Diaz, C.K. Jankowski, C. Hocquelet, P. F. del Rio, H. Barrios, The unambiguous
assignment of NMR spectra of per-O-methylated 6-mono and 6,6-diamino-␤-
cyclodextrins, Can. J. Chem. 86 (2008) 726–736.
[33] C. Tode, A. Takeuchi, S. Iwakawa, A. Tatsumi, M. Sugiura, Hydrogen–deuterium
(H–D) exchange reaction of warfarin in D2 O solution, Chem. Pharm. Bull. 57
(2009) 653–656.
[34] L.C. West, O. Wyness, B.L. May, P. Clements, S.F. Lincoln, C.J. Easton, Diazacoro-
nand linked ␤-cyclodextrin dimer complexes of brilliant yellow tetraanion and
their sodium(I) analogues, Org. Biomol. Chem. 1 (2003) 887–894.
[35] J.S. Lock, B.L. May, P. Clements, S.F. Lincoln, C.J. Easton, Cyclodextrin complex-
ation of a stilbene and the self-assembly of a simple molecular device, Org.
Biomol. Chem. 2 (2004) 337–344.
[36] M. Nishio, The CH/␲ hydrogen bond in chemistry. Conformation,
supramolecules, optical resolution and interactions involving carbohydrates,
Phys. Chem. Chem. Phys. 13 (2011) 13873–13900.