Biochemical and pharmacological evaluation of 4-hydroxychromen-2-ones bearing polar C-3 substituents as anticoagulants

Article abstract of DOI:10.1016/j.ejmech.2012.04.036

The objective of this study was to investigate in vitro and in vivo anticoagulant activity of sixteen 4-hydroxycoumarin derivatives bearing polar C-3 scaffolds. The activity was evaluated by measuring prothrombin time. Enhanced anticoagulant activity in vitro was observed for all tested compounds. Upon successive administration of 0.5 mg/kg of body weight to adult Wistar rats, over a period of five days, four derivatives (2b, 4c, 5c and 9c) presented anticoagulant activity in vivo. The most active compound was 2b, with PT = 30.0 s. Low or non-toxic effects in vivo were determined based on the catalytic activity of liver enzymes and the concentration of bilirubin, iron and proteins. Metabolic pathways of the most active compounds in vivo were determined after GC/MS analysis of collected rat urine samples. The excretion occurs by glucuronidation of 7-hydroxy forms of tested derivatives. In vivo results were described using PLS-based CoMFA and CoMSIA 3D-QSAR studies, which showed CoMFA-SE (q² = 0.738) and CoMSIA-SEA (q² = 0.763) to be the statistically most relevant models. Furthermore, molecular docking and DFT mechanistic studies performed on the rat VKORC1 homology model revealed interactions between the 4-OH coumarin group in the form of phenolic anion and the Cys135 catalytic site in the transition state.

Full text of DOI:10.1016/j.ejmech.2012.04.036

European Journal of Medicinal Chemistry 54 (2012) 144e158
Contents lists available at SciVerse ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Biochemical and pharmacological evaluation of 4-hydroxychromen-2-ones
bearing polar C-3 substituents as anticoagulants
,
a
ꢀ a
ꢀ b
ꢀb
ꢀ a
*
Milan Mladenovic , Mirjana Mihailovic , Desanka Bogojevic , Nenad Vukovic , Slobodan Sukdolak ,
ꢀ c
ꢀ
ꢀa
ꢀ a
ꢁ
ꢀ d
ꢀ e
Sanja Matic , Neda Niciforovic , Vladimir Mihailovic , Pavle Maskovic , Miroslav M. Vrvic ,
ꢀ a
Slavica Solujic
a Department of Chemistry, Faculty of Science, University of Kragujevac, Radoja Domanovica 12, 34000 Kragujevac, P.O. Box 60, Serbia
b Department of Molecular Biology, Institute for Biological Research, University of Belgrade, Bulevar Despota Stefana 142, 11000 Belgrade, Serbia
c Department of Biology and Ecology, Faculty of Science, University of Kragujevac, Radoja Domanovica 12, 34000 Kragujevac, P.O. Box 60, Serbia
d
ꢁ
ꢁ
ꢁ
Faculty of Agronomy, University of Kragujevac, Cara Dusana 34, 32000 Cacak, Serbia
e Faculty of Chemistry, University of Belgrade, Studentski Trg 12-16, P.O. Box 158, 11001 Belgrade, Serbia
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
< Fifteen
synthesized
coumarin
compounds were examined as
in vivo anticoagulants.
< Four compounds, 2b, 4c, 5c and 9c,
presented in vivo activity.
< Coumarin
molecular
target,
VKORC1, was homology-modeled.
< Results were discussed using 3D-
QSAR, molecular docking and DFT
mechanistic studies.
< Coumarin metabolic pathways were
obtained.
a r t i c l e i n f o
a b s t r a c t
Article history:
The objective of this study was to investigate in vitro and in vivo anticoagulant activity of sixteen
4-hydroxycoumarin derivatives bearing polar C-3 scaffolds. The activity was evaluated by measuring
prothrombin time. Enhanced anticoagulant activity in vitro was observed for all tested compounds. Upon
successive administration of 0.5 mg/kg of body weight to adult Wistar rats, over a period of five days, four
derivatives (2b, 4c, 5c and 9c) presented anticoagulant activity in vivo. The most active compound was 2b,
with PT ¼ 30.0 s. Low or non-toxic effects in vivo were determined based on the catalytic activity of liver
enzymes and the concentration of bilirubin, iron and proteins. Metabolic pathways of the most active
compounds in vivo were determined after GC/MS analysis of collected rat urine samples. The excretion
occurs byglucuronidation of 7-hydroxy forms of tested derivatives. In vivo results were described using PLS-
based CoMFA and CoMSIA 3D-QSAR studies, which showed CoMFA-SE (q2 ¼ 0.738) and CoMSIA-SEA
(q2 ¼ 0.763) to be the statistically most relevant models. Furthermore, molecular docking and DFT mech-
anistic studies performed on the rat VKORC1 homology model revealed interactions between the 4-OH
coumarin group in the form of phenolic anion and the Cys135 catalytic site in the transition state.
Ó 2012 Elsevier Masson SAS. All rights reserved.
Received 19 April 2011
Received in revised form
10 April 2012
Accepted 25 April 2012
Available online 4 May 2012
Keywords:
4-Hydroxycoumarins
In vivo anticoagulant activity
Metabolic pathways
Homology modeling
Mechanistic studies
3D-QSAR
1. Introduction
Anticoagulant activity of 4-hydroxycoumarins, confirmed by
clinical usage of the drug warfarin and expressed through long-
term treatment and elongation of International Normalized Ratio,
* Corresponding author. Tel.: þ381 34336223; fax: þ381 34335040.
ꢀ
E-mail address: mmladenovic@kg.ac.rs (M. Mladenovic).
0223-5234/$ e see front matter Ó 2012 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejmech.2012.04.036

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
145
INR [1], prevents many diseases such as venous thrombosis,
pulmonary embolism or conditions like heart attack [2]. As an
anticoagulant, warfarin acts as a competitive inhibitor of vitamin
carbonyl/carboxyl or thiazole-N-phenyl functional groups. Their
anticoagulant activity was measured by prothrombin time and the
INR. The presence of highly hydrophilic C-3 residues contributed to
high anticoagulant potential in vitro of all compounds. Four of the
compounds presented anticoagulant potential in vivo after
administration of the compounds to adult Wistar rats. The non-
toxic behavior of the administered compounds was determined
afterwards rat serum screening, using the appropriate biochemical
analyses. Monitoring of in vivo activity was accompanied by
3D-QSAR studies. The resulting CoMFA and CoMSIA models were
used to predict the anticoagulant activity of sixteen novel designed
compounds. In order to describe the in vivo results trough the
interactions of our derivatives with VKORC1, we built a homology
model of rat VKORC1 and performed molecular docking studies.
Since no results of similar studies have been presented before, we
proposed a pseudo-enzymatic mechanism of antagonist behavior
of our derivatives and studied it computationally with the aid of
DFT theory.
K-2,3-epoxide reductase subunit
1 (VKORC1), transmembrane
protein of liver endoplasmic reticulum [3]. Warfarin prevents
reformation of vitamin K-2,3-epoxide to vitamin K and, conse-
quently, vitamin K-dependent synthesis of biologically active forms
of clotting factors II, VII, IX and X [1]. The drug is transferred by
serum albumin to the active site of VKORC1 in open side chain
coumarin form [4]. Since there are still no available PDB, NMR, or
X-ray crystallography data structure of warfarin-VKORC1 complex,
the active form of warfarin, present during the anticoagulant
activity, is debatable. The hemiketal structure of the drug, which is
dominant in the crystalline form but minor in solution [5], is not
favorable for forming neither the warfarin-albumin nor the
warfarin-VKORC1 complex [6]. It has been proposed that warfarin
interacts with VKORC1 in deprotonated open side chain coumarin
form, i.e. the form of phenolic anion [6]. In this form, warfarin
shares structural similarities with vitamin K-2,3-epoxide and acts
as a competitive inhibitor. Two processes are involved in the
anticoagulant activity of warfarin. The first is binding of the 4-
phenylpentan-2-one structural part at warfarin C-3 position to
Thr138, Tyr139, and Ala140 residues of the VKORC1 active site [7].
The second is interaction between the warfarin 4-hydroxy group,
in the form of phenolic anion, and the catalytic center Cys132-
Cys135, after which a disulfide bond is formed [3]. Recently, DFT
2. Chemistry
In our previous papers, synthesis and structural characterization
of 16 substituted chromene-2H-one derivatives (1e10c) were
reported (Scheme 1) [10,11]. The compounds were characterized by
elemental analysis (C, H, N, O, and S). Structural characterization
was carried out by IR, H1 NMR and MS analyses. Over 95% of
chemical purity of the synthesized compounds was confirmed with
HPLC and MS.
computational studies [8] established
a (pseudo-enzymatic)
mechanism of vitamin K-2,3-epoxide interaction with VKORC1, on
the basis of which the essential epoxide reductase activity can be
modeled. The results suggest that the phenolic anion form of
warfarin probably acts as a competitive inhibitor of the enzyme
[5,6] and that the reaction occurs on the enzyme itself. No actual
computational studies on the warfarin inhibition mechanism have
been presented so far.
Recently, novel warfarin derivatives in which the phenylpentan-
2-one functional group at position C-3 was substituted with various
isoprenyl side chains similar to ferulenol, were evaluated as in vitro
anticoagulants [9]. Study showed that other functional groups,
structurally different in comparison with the 4-phenylpentan-2-
one, can also contribute the anticoagulant activity [9]. Therefore,
this paper presents the in vitro and in vivo anticoagulant potential of
some 4-hydroxycoumarin derivatives that bear polar C-3 substit-
uents and are structurally different from warfarin [10,11]. The C-3
position tested compounds was substituted by various prop-1-en-
3. Results and discussion
3.1. In vitro anticoagulant activity
The coagulation and prothrombin times of human plasma were
normal: CT ¼ 105 and PT ¼ 11.0e14.5 s (Table 1), respectively. In
comparison with the activity of warfarin (measured PT ¼ 18.5 s),
the introduction of prop-1-en-carbonyl/carboxyl (2e8b) and thia-
zole-N-phenyl scaffolds (2e10c) at the coumarin C-3 position
greatly increased the anticoagulant activity.
The C-3 position of the 4-xydroxycoumarin core within 2e8b
was substituted with trans-oriented carboxyl or ester groups and
various cis-oriented groups with respect to the methyl group of the
prop-1-en scaffold. Among 2e8b, the highest prolongation of
prothrombin time was noticed after testing 2b,
a cis/trans-
Scheme 1. Synthesized 4-hydroxy-chromene-2-one derivatives.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
146
Table 1
Anticoagulant activity of coumarin derivatives in vitro and in vivo (after 24 h and five-day treatment).
Comp.
PT in vitroa
INRb
PT in vivoc
INR
PT in vivod
INR
Fibrinogen. (g/L)
FT in vivoe
1
2b
14.0 ꢀ 0.2f
1.25
1.99
9.1 ꢀ 0.2
0.9
0.13
9.5 ꢀ 0.2
0.9
2.5 ꢀ 0.2
5.1 ꢀ 0.1
1.6 ꢀ 0.1
1.6 ꢀ 0.2
1.8 ꢀ 0.4
2.47 ꢀ 0.2
0.5 ꢀ 0.1
3.2 ꢀ 0.5
4.2 ꢀ 0.2
1.8 ꢀ 0.3
/
0.9 ꢀ 0.1
2.0 ꢀ 0.1
1.2 ꢀ 0.1
2.5 ꢀ 0.2
1.4 ꢀ 0.3
2.1 ꢀ 0.2
0.6 ꢀ 0.2
1.74 ꢀ 0.2
1.9 ꢀ 0.2
1.8 ꢀ 0.3
1.9 ꢀ 0.3
1.9 ꢀ 0.3
44.9 ꢀ 0.7
51.3 ꢀ 0.5
23.0 ꢀ 0.1
10.3 ꢀ 0.3
10.5 ꢀ 0.1*,y
30.0 ꢀ 0.2g,*
9.8 ꢀ 0.2
0.13
0.13
0.94
0.67
0.72
0.69
0.48
0.86
0.68
0.84
1.38
1.16
1.34
0.67
0.67
0.68
1.14
1.34
0.94
0.91
0.86
0.83
3b
4b
6b
7b
8b
2c
3c
4c
19.0 ꢀ 0.2
20.0 ꢀ 0.3
17.0 ꢀ 0.2
21.0 ꢀ 0.1
20.0 ꢀ 0.1
22.0 ꢀ 0.3
25.0 ꢀ 0.4
26.5 ꢀ 0.2
1.80
1.90
1.55
2.25
1.90
2.15
2.44
2.60
9.3 ꢀ 0.1
8.0 ꢀ 0.2
8.6 ꢀ 0.2
8.2 ꢀ 0.2
5.7 ꢀ 0.4
8.8 ꢀ 0.2
8.1 ꢀ 0.1
10.4 ꢀ 0.1
0.92
0.64
0.69
0.66
0.37
0.67
0.65
0.82
40.7 ꢀ 0.4
35.2 ꢀ 0.3
24.6 ꢀ 0.4
196.8 ꢀ 0.6
19.4 ꢀ 0.5
55.7 ꢀ 0.5
33.7 ꢀ 0.4
>500
9.0 ꢀ 0.3
9.6 ꢀ 0.2
9.2 ꢀ 0.1
8.7 ꢀ 0.1
9.1 ꢀ 0.3
9.1 ꢀ 0.4
12.4 ꢀ 0.2*,y
28.6 ꢀ 0.1g,*
9.7 ꢀ 0.2y
28.1 ꢀ 0.2g,*
10.0 ꢀ 0.3
10.0 ꢀ 0.1
8.1 ꢀ 0.2
5c
25.0 ꢀ 0.2
2.44
9.6 ꢀ 0.2y
1.16
27.6 ꢀ 0.6
6c
7c
8c
9c
22.0 ꢀ 0.3
18.5 ꢀ 0.1
18.1 ꢀ 0.2
24.4 ꢀ 0.6
2.10
1.75
1.70
2.30
8.0 ꢀ 0.2
8.0 ꢀ 0.3
8.1 ꢀ 0.4
9.6 ꢀ 0.2y
0.64
0.64
0.68
0.94
23.9 ꢀ 0.4
44.9 ꢀ 0.5
28.7 ꢀ 0.4
>500
14.4 ꢀ 0.6*,y
28.1 ꢀ 0.1g,*
9.7 ꢀ 0.1
10c
C
27.5 ꢀ 0.1
13.2 ꢀ 0.1
2.70
1.15
9.2 ꢀ 0.2
9.2 ꢀ 0.4
0.94
0.91
33.2 ꢀ 0.4
9.7 ꢀ 0.2
111.9 ꢀ 0.3
12.1 ꢀ 0.2g
12.2 ꢀ 0.4
LDi
Wh
18.5 ꢀ 0.1
1.73
10.3 ꢀ 0.2
0.83
33.4 ꢀ 0.5
y
*p <0.05 when compared with the negative control; <0.05 when compared with the positive control.
a
Prothrombin time measured on the human plasma measured in sec.
International Normalized ratio.
The prothrombin time measured after 0.1 mg/kg coumarin administration measured in sec.
The prothrombin time measured after 0.5 mg/kg coumarin administration measured in sec.
Fibrin time measured in sec.
Results are given with standard deviation, calculated from at least three measurements.
The prothrombin time measured after the second 0.5 mg/kg administration of coumarin derivatives.
Warfarin.
b
c
d
e
f
g
h
i
Lethal dose.
carboxyethyl derivative (PT ¼ 23.0 s). A minor decrease in activity
was detected for 7b, cis-cyano and trans-carboxyl derivative, with
a prothrombin time of 21.0 s. A level of activity comparable to 7b
was recorded for 8b. In the structure of 8b, besides the presence of
the trans-carboxyl group, there is a cis-carboxymethyl residue,
which additionally contributes to the activity. Further decrease in
the anticoagulant potential was observed for 4b and 3b. The lower
potential of these compounds is related with the absence of
carboxyl or cyano groups in the cis-position.
comparable with warfarin and potential in vivo anticoagulants.
With respect to the prothrombin time of the negative control (i.e.
plasma containing no anticoagulants), the activities of 2b, 4c, and
9c were 1.08-, 1.27- and 1.48-fold above the control value, respec-
tively. There was no significant statistical difference between the
prothrombin time measured in plasma containing 5c and the
negative control. The activities of 4c and 9c were 1.18-fold above,
while the activities of 2b and 5c were 86.06 and 79.51% of the
activity of warfarin as the positive control.
The substitution of prop-1-en moieties with various thiazole-N-
phenyl groups (2e10c) further enhanced the anticoagulant poten-
tial in the following order: 9c < 3c, 5c < 4c < 10c. Thus, by changing
the nitro group position from m- (9c) to p- (3c), the prothrombin
time was increased. The 5c (m-CH3) and 3c derivatives presented
equal activity (PT ¼ 25.0 s). The p-sulfo (4c) and N-naphthyl (10c,
PT ¼ 27.5 s) compounds attained the highest activity in vitro.
In view of the known behavior of warfarin to exert anticoagulant
activity after long-term administration, the test animals were
subjected to repeated five-day treatment with the selected
compounds in a concentration of 0.5 mg/kg of body weight. Such
treatment significantly increased the prothrombin time (Table 1).
Since continuous i.p. administration of warfarin induced death of
the treated animals on the final day, measurement of prothrombin
time was not possible. Statistical comparisons were therefore taken
only for the administered compounds and negative control. An
excellent in vitro-in vivo activity correlation was observed for 2b on
the rat model with prothrombin time in vivo of PT ¼ 30.0 s, i.e. 2.5-
fold above the basic value measured in the negative control. The
range of activities of 4c, 5c, and 9c was 28.1e28.6 s, 2.3- to 2.34-fold
above the negative control values.
3.2. In vivo anticoagulant activity
3.2.1. Observed activity
A
short period of intraperitoneal (i.p.) administration of
compounds 1e10c to adult Wistar rats in a concentration of 0.1 mg/
kg of body weight did not produce the noteworthy increase of
anticoagulant activity in comparison to warfarin (Table 1). There
was no significant statistical difference between the activity of
warfarin, 2b, and 4c, while the activity of 5c and 9c was 6.89% lower
comparing with the clinical drug.
Increase of the dose (0.5 mg/kg of body weight) resulted in
a better activity (Table 1) and showed four compounds, prop-1-en
diethylmalonate derivative 2b (PT ¼ 10.5 s), N-thiazole p-sulfonic
acid derivative 4c (PT ¼ 10.4 s), N-thiazole m-methyl derivative 5c
(PT ¼ 9.7 s) and N-thiazole m-nitro derivative 9c (PT ¼ 14.4 s) to be
3.2.2. Liver status after in vivo administration of coumarin
derivatives
The toxicity level of the derivatives used in this study was
evaluated by measuring the catalytic activities of ALT, AST, ALP,
and
g-GT in serum as hepatocellular markers, and by monitoring
bilirubin concentration as
a
hepatobiliary marker (Table 2)
[12e14]. In both cases, measurements were performed after i.p.
administrations (0.5 mg/kg). The ALT and AST catalytic activities

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
147
are sensitive indicators of acute liver damage [12]. Bilirubin
3.2.3. Determination of coumarin metabolites from rat urine by
GC/MS
concentration is a useful clinical clue for the diagnosis of hepa-
tocyte necrosis [15]. The amounts of Fe and serum proteins were
observed as well.
The dominant pathway in the first phase of coumarin metabo-
lism in rats and mice is hydroxylation of the coumarin core at the
C-7 position, catalyzed by the CYP2A5 enzyme, which belongs to
the CYP450 family [19]. Our GC/MS analysis was therefore directed
towards the detection of 7-hydroxy forms of the administered
compounds. The molecular ions of the tested compounds recorded
in ethanol solution were Mþ-1 m/z 345, 415, 349 and 380 for 2b, 4c,
5c, and 9c, respectively. Hence, after the first phase of metabolism,
the expected Mþ m/z values for 7-hydroxy forms were 362, 432,
366, and 397. The detection of particular ions in the mass spectra of
hydrolyzed urine samples suggested that the tested derivatives
were metabolized in the same way as the coumarin core (Scheme
2). The presence of 7-hydroxycoumarin forms of 2b, 4c, 5c,
and 9c was confirmed by correlation of the obtained spectra
with the spectrum of umbelliferone. The mass of m/z 165 for
7-hydroxycoumarin which was detected in the spectra of hydro-
lyzed urine samples containing 2b, 4c, 5c, and 9c differs from the
recorded mass of umbelliferone (Mþ m/z 162) [20] due to the
transfer of hydrogen atoms towards the coumarin core during
fragmentation of 2b, 4c, 5c, and 9c. On the other hand, the similar
intensities of the compared ions, viz., m/z 165 and m/z 162, indi-
cated that they belong to the same molecular species. Further
confirmation that m/z 165 originates from the coumarin core was
the presence of m/z 146 or 147 ions, which correspond to Mþ of the
coumarin molecule [18]. Hydroxylation may also occur in positions
C-5, -6, and -8 [5], but these transformations are not favorable for
coumarin excretion. The second phase of coumarin metabolism is
mainly catalyzed by UDP-glucuronyltransferase resulting in
formation of the coumarin-7-glucuronide form, in which the
molecule is excreted [19]. Hence, it is presumed that excretion of
the tested derivatives occurs after the formation of 2b-, 4c-, 5c-,
and 9c-glucuronides. A much smaller amount of excretion occurs
after the formation of coumarin-7-sulfate, catalyzed by sulfo-
transferase [19]. There was no evidence for the formation of other
metabolites using the specified GC/MS conditions.
In general, after the first animal treatment with coumarin
derivatives in concentration of 0.5 mg/kg of body weight, mostof the
compounds were weak inflammation agents, as judged from low
increase in the catalytic activity of serum ASTand the concentration
of ferrous ions [16,17]. In serum containing 2b, 4c, 5c, and 9c, cata-
lytic activity of AST was significantly increased: 1.41-, 1.59-, 1.68-,
and 1.74-fold above the activity in negative control, respectively.
Still, the administration of 2b, 4c, and 5c induced lower catalytic
activity of the enzyme than in the serum of animals treated with
warfarin (63.28e97.26% of the basal level), and therefore lower
inflammation than induced by the clinical drug. The compound 9c
was the strongest inflammation agents, with toxicity expressed by
AST catalytic activity value 1.06-fold above the positive control.
During continued 0.5 mg/kg administration, no inflammation
was perceived after treatment of rats with the compounds. The
catalytic activities of AST measured in serum containing 2b, 4c, 5c,
and 9c were respectively 6.25, 7.33, 21.62, and 5.90% lower than in
the negative control.
The low catalytic activities of ALT and ALP recorded after the
first and second 0.5 mg/kg administration indicated that there was
no chemically induced liver damage [18], which was also
confirmed by no change in the concentration of serum proteins
[18]. Further, the decrease of bilirubin concentration and normal
catalytic activity of
the bile [18].
g-GT showed that our derivatives do not affect
The in vivo anticoagulant potential of tested 4-hydroxycoumarins
was attained after five-day long administration of a 0.5 mg/kg
concentration. The results are in correlation with the fact that
coumarin presents anticoagulant activity after several days of
continuous treatment. Furthermore, analyses of the rat liver by
biochemical serum screening showed that coumarin-conditioned
inflammation decreased after the administration interval, and no
inflammation was observed on the final day of application.
Table 2
Hepatotoxic activity of coumarin derivatives and liver protein status after 24 h and five-day treatment in concentration of 0.5 mg/kg of body weight.
Comp.
Bilirubin(
m
mol/L)
AST (U/L)
ALT (U/L)
ALP (U/L)
g-GT (U/L)
Fe (
m
mol/L)
Proteins (g/L)
Albumin (g/L)
1
2b
1.7 ꢀ 0.1
2.0 ꢀ 0.2
160.5 ꢀ 0.4
238.3 ꢀ 0.2*,y
209.9 ꢀ 0.2a,*
236.9 ꢀ 0.3
154.3 ꢀ 0.2
201.4 ꢀ 0.4
/
218.4 ꢀ 0.3
232.5 ꢀ 0.2
264.4 ꢀ 0.2
269.7 ꢀ 0.2*,y
207.5 ꢀ 0.2a,*
284.6 ꢀ 0.2*,y
175.5 ꢀ 0.2a,*
231.5 ꢀ 0.2
254.8 ꢀ 0.1
263.4 ꢀ 0.2
294.6 ꢀ 0.3*,y
210.7 ꢀ 0.3a,*
345.3 ꢀ 0.2
169.0 ꢀ 0.2
223.9 ꢀ 0.2a
277.3 ꢀ 0.1
34.8 ꢀ 0.2
71.8 ꢀ 0.3
55.5 ꢀ 0.3b
65.5 ꢀ 0.2
57.2 ꢀ 0.2
56.4 ꢀ 0.4
/
51.8 ꢀ 0.6
44.3 ꢀ 0.3
68.7 ꢀ 0.4
58.4 ꢀ 0.3
63.0 ꢀ 0.3b
67.3 ꢀ 0.3
62.4 ꢀ 0.3b
52.6 ꢀ 0.2
64.7 ꢀ 0.1
59.5 ꢀ 0.3
52.4 ꢀ 0.1
65.1 ꢀ 0.1b
85.0 ꢀ 0.2
76.8 ꢀ 0.4
40.7 ꢀ 0.4b
51.5 ꢀ 0.4
200 ꢀ 0.4
513 ꢀ 0.7
3.8 ꢀ 0.1
3.1 ꢀ 0.2
2.3 ꢀ 0.2c
2.8 ꢀ 0.1
3.0 ꢀ 0.3
4.4 ꢀ 0.1
/
4.4 ꢀ 0.1
3.6 ꢀ 0.2
3.8 ꢀ 0.3
2.4 ꢀ 0.2
2.2 ꢀ 0.2c
2.2 ꢀ 0.2
3.0 ꢀ 0.2c
3.5 ꢀ 0.2
4.4 ꢀ 0.2
3.6 ꢀ 0.1
2.8 ꢀ 0.2
2.6 ꢀ 0.2c
4.3 ꢀ 0.1
4.0 ꢀ 0.2
4.5 ꢀ 0.2c
3.4 ꢀ 0.1
26.1 ꢀ 0.1
28.4 ꢀ 0.3
17.7 ꢀ 0.3d
23.9 ꢀ 0.4
54.2 ꢀ 0.5
32.4 ꢀ 0.3
/
19.4 ꢀ 0.2
22.6 ꢀ 0.4
14.8 ꢀ 0.3
19.3 ꢀ 0.2
19 ꢀ 0.2d
28.4 ꢀ 0.4
17.3 ꢀ 0.4d
27.4 ꢀ 0.6
25.0 ꢀ 0.2
29.2 ꢀ 0.4
29.7 ꢀ 0.1
17.3 ꢀ 0.1d
24.6 ꢀ 0.2
44.2 ꢀ 0.2
21.1 ꢀ 0.2d
22.3 ꢀ 0.4
63.8 ꢀ 0.6
67.2 ꢀ 0.5
31.9 ꢀ 0.6
35.2 ꢀ 0.5
3b
4b
6b
7b
8b
2c
3c
4c
1.7 ꢀ 0.3
1.4 ꢀ 0.2
1.5 ꢀ 0.4
/
1.8 ꢀ 0.2
1.9 ꢀ 0.2
1.7 ꢀ 0.5
1.6 ꢀ 0.1
285 ꢀ 0.3
333 ꢀ 0.9
223 ꢀ 0.5
/
215 ꢀ 0.6
213 ꢀ 0.4
204 ꢀ 0.8
164 ꢀ 0.6
66.4 ꢀ 0.3
65.2 ꢀ 0.4
66.5 ꢀ 0.2
68.9 ꢀ 0.6
62.8 ꢀ 0.3
65.8 ꢀ 0.4
63.0 ꢀ 0.4
64.9 ꢀ 0.2
38.9 ꢀ 0.4
39.0 ꢀ 0.1
40.6 ꢀ 0.6
38.6 ꢀ 0.3
35.8 ꢀ 0.5
37.4 ꢀ 0.3
37.2 ꢀ 0.4
40.8 ꢀ 0.4
5c
1.8 ꢀ 0.2
172 ꢀ 0.2
66.2 ꢀ 0.7
41.9 ꢀ 0.1
6c
7c
8c
9c
1.2 ꢀ 0.1
2.0 ꢀ 0.3
1.8 ꢀ 0.1
2.2 ꢀ 0.2
246 ꢀ 0.4
192 ꢀ 0.7
318 ꢀ 0.4
225 ꢀ 0.4
71.5 ꢀ 0.1
63.2 ꢀ 0.6
69.2 ꢀ 0.3
64.2 ꢀ 0.4
35.4 ꢀ 0.2
36.7 ꢀ 0.5
36.3 ꢀ 0.1
39.7 ꢀ 0.2
10c
C
2.0 ꢀ 0.2
2.4 ꢀ 0.5
210 ꢀ 0.3
516 ꢀ 0.2
72.8 ꢀ 0.2
71.8 ꢀ 0.5
41.3 ꢀ 0.2
39.3 ꢀ 0.4
W
1.3 ꢀ 0.2
170 ꢀ 0.6
59.4 ꢀ 0.4
34.5 ꢀ 0.2
*p <0.05 when compared with the negative control; y <0.05 when compared with the positive control.
a
The level of AST after the second 0.5 mg/kg coumarin administration.
The level of ALT after the second coumarin administration.
b
c
The level of
g-GT ather the second coumarin administration.
d
The level of Fe after the second coumarin administration.

ꢀ
148
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
Scheme 2. Proposed metabolic pathways of 2b and 4c.
3.2.4. CoMFA and CoMSIA results
respectively) were noticed. Correlations between the calculated
and experimental pIC50 values (from training and LOO cross-
validation) are illustrated in Fig. 2(a). The CoMFA-SE model was
also used to predict anticoagulant activity of the test-set
compounds. This model was able to describe the test-set variance
with r2 ¼ 0.724 and a standard deviation of test-set predictions of
(stest) ¼ 0.901. The predicted values of test-set are listed in Table 3.
Our analysis revealed that the proposed model was able to
successfully predict the activity of compounds that were not used
in the training process.
As a result of the 3D-QSAR studies, CoMFA and CoMSIA contour
maps were generated to rationalize the pharmacologically active
regions in 3D space around the molecules. The positions of steric,
electrostatic, and hydrogen bond fields were used to define parts of
molecules that enhance or lessen activity. Contour maps were
generated to reveal the nature of potential interactions between
the 4-OH group of coumarin anticoagulants and the catalytic site, as
well as between the C-3 polar scaffolds and amino acids of the
active site of the VKORC1 molecule.
Fig. 1. depicts aligned molecules within the grid box (grid
spacing of 2.0 Å) used to generate the CoMFA and CoMSIA columns.
The best obtained CoMFA model described the anticoagulant
activity of the evaluated compounds using both steric and elec-
trostatic fields (CoMFA-SE), had a cross-validated coefficient value
of q2 ¼ 0.738 with an optimized component number of N ¼ 6,
excellent non-validated value of r2 ¼ 0.931, and standard estimate
error (SEE) of 0.224. The model possessed a low standard deviation
(s ¼ 0.216) and a high Fisher ratio (F ¼ 108.96). Statistical param-
eters of the model suggested that it should be a useful tool for the
prediction of IC50 values, i.e. the concentrations that increase the
prothrombin time for 50%. Within the CoMFA-SE model, similar
contributions of both steric and electrostatic fields (49.1 and 51.9%,
The positions of contour plots of CoMFA steric and electrostatic
fields surrounding the 2b and 4c, as the most promising antico-
agulants in vivo, are presented in Fig. 3.
The steric fields are characterized by green and yellow isopleths,
which represent steric regions that favor and disfavor activity,
respectively. Green isopleths (Fig. 3(a) and (b)) are placed above
both carbonyl groups of the C-3 diethylmalonate scaffold of 2b as
well as the thiazole-N-Ph-SO3H system of 4c. The m-CH3 and
m-NO2 groups of 5c and 9c, respectively, are also overlaid with
green contours. These contours indicate residues that are manda-
tory for a high anticoagulant activity and expected to bind with the
VKORC1 active site. Yellow isopleths cover the space around the
methyl group of the prop-1-en moiety, ethyl group of the trans-
carboxyethyl residue of 2b, and the sulfur atom of the thiazole ring
of 4c. This suggests that regions covered with the yellow contours
are too large, i.e. not favorable for activity. Similar orientation of
unfavorable fields was observed above the structures of 5c and 9c.
The electrostatic fields are depicted in Fig. 3(c) and (d). The blue
contours illustrate electrostatic regions where an increase of posi-
tive charge and electron-donor ability of functional groups enhance
activity, while red isopleths indicate regions where negative charge
and electron-withdrawing groups are favorable for activity. The
blue electrostatic maps around the hydrogen atom of the 4-OH
group of both 2b and 4c, as well as of 5c and 9c, show that posi-
tively charged hydrogen atom is very important for the overall
activity and that it participates in the anticoagulant mechanism.
Furthermore, the blue regions that cover cis-carboxyethyl group of
2b and phenyl group of 4c suggest that the presence of the
electron-donating groups increase the activity. Location of the red
contour surrounding the cis-carboxyethyl group of 2b shows the
significance of groups containing electron-withdrawing oxygen
atoms for the activity. Also, apart from the unfavorable steric effects
of thiazole sulfur, the low positive charge of the sulfur atom is
important for the activity of 4c. Electrostatic interactions estab-
lished over the sulfur atom are important for the activity of 5c and
9c, as well.
Fig. 1. Atom-by-atom superposition used for the 3D-QSAR analysis.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
149
Fig. 2. a) Plot of predicted vs. experimental pIC50 values of the training set (squares) and the test-set (circles) by CoMFA; b) Plot of predicted versus experimental pIC50 values of the
training set (squares) and the test-set (triangles) by CoMSIA.
In comparison with CoMFA methodology, CoMSIA has the
advantage of exploring more fields. A statistically more significant
model was obtained using the CoMSIA study. The best CoMSIA
model (CoMSIA-SEA) included steric, electrostatic, and H-bond
acceptor (HA) fields and had a q2 value of 0.763 with optimized
components of N ¼ 3, a non-cross-validated correlation coefficient
value r2 of 0.916, and an SEE value of 0.224. The model possessed
a low standard deviation (s ¼ 0.791) and a high Fischer ratio
(F ¼ 78.96). Statistical parameters of the model suggested that it
should be a useful tool for the prediction of IC50 values. Contribu-
tions of the steric, electrostatic, and HB acceptor fields within the
CoMSIA-SEA model were 11.2, 42.6, and 46.2%, respectively. The
predictions of pIC50 values of the training set are shown in Table 3.
Correlations between the calculated and experimental pIC50 values
(from training and LOO cross-validation) are shown in Fig. 3(b). The
CoMSIA-SEA model was also used to predict inhibitory activities of
the test-set compounds. It was found that this model was able to
describe the test-set variance with r2 ¼ 0.817 and stest ¼ 0.766.
The contour plots of the CoMSIA steric, electrostatic, and HB
acceptor fields are presented in Fig. 4. To simplify, only interactions
between the most active compounds and the contour maps are
shown.
for 5c and 9c. The CoMSIA steric field contour maps were similar to
the corresponding CoMFA contour maps.
Electrostatic fields of 2b and 4c are depicted by Fig. 4(b) and
(e), respectively. Blue contours depict regions where positive
electrostatic interactions increase the activity, while red contours
indicate regions where negative electrostatic interactions also
increase the activity. Positive electrostatic regions are located
above carbonyl group of cis-carboxyethyl scaffold of 2b and the
sulfur atom of p-SO3H group of the 4c. The orientation of a blue
contour around the sulfo group was the main difference in
comparison with the CoMFA field, where only the phenyl group
was surrounded by the contour. Negative electrostatic fields are
positioned around the hydrophobic parts of the molecules. As for
5c and 9c, blue contours cover the sulfur atom of the thiazole
ring, while a strong red contour was noticed around the p-nitro
group of 9c, highlighting the electron-withdrawing effect as
important for activity.
The contour maps representing hydrogen bond acceptor fields
of 2b and 4c are depicted in Fig. 4(c) and (f), respectively. The
magenta isopleths represent areas where HB acceptor groups
increase activity. Large magenta contours cover both carbonyl
residues of the 2b C-3 scaffold and N-thiazole system of 4c. The red
regions decrease the ability of the compounds to accept a proton
and form a hydrogen bond. They are located around the ethyl group
of the cis-carboxyethyl scaffold of 2b and the amino group that
links the thiazole and phenyl group of 4c. The position of the red
fields indicates that the presence of ethyl group as well as
secondary amine group as a strong H-bond donor are detrimental
to biological activity. Indeed, the potential of carboxyl groups as HB
acceptors in the structure of 2b was stronger than the HB acceptor
Green isopleths around 2b and 4c, depicted by Fig. 4(a) and (d)
respectively, illustrate regions where bulky groups favor the
activity, while yellow isopleths present regions that decrease it.
Thus, carbonyl groups in the C-3 residue of 2b are surrounded by
the advantageous side of the steric field, indicating their impor-
tance in interactions with the molecular target. In the case of 4c,
green contours encircle nitrogen atoms of the thiazole and amino
group, respectively. A similar position of the fields was perceived
Table 3
Regression summaries of CoMFA and CoMSIA results and the prediction of the pIC50 of newly designed compounds.
Comp.
pIC50
Exp.
CoMFA
Pred.
CoMSIA
Pred.
Designed
Comp.
CoMFA
Pred.
CoMSIA
Pred.
Res.
Res.
Res
Res
1
2.77
3.85
2.94
2.56
2.86
2.63
2.50
2.63
2.63
3.85
4.16
3.03
3.03
2.27
3.13
3.23
2.83
3.94
2.96
2.60
2.84
2.65
2.56
2.79
3.04
3.90
4.33
2.95
2.98
2.25
3.16
3.17
0.06
0.30
0.28
0.14
0.41
0.02
0.06
0.46
0.40
0.14
0.83
0.08
0.04
0.98
0.04
0.05
2.95
3.38
3.10
2.58
2.97
2.70
2.43
2.63
2.62
3.84
4.38
2.95
2.93
2.26
3.12
3.36
0.05
0.37
0.26
0.22
0.40
0.03
0.06
0.27
0.38
0.12
0.63
0.07
0.08
0.42
0.05
0.18
1d
2d
3d
4d
5d
6d
7d
8d
9d
10d
11d
12d
13d
14d
15d
16d
3.97
3.95
4.04
3.98
4.14
4.10
4.21
4.14
4.22
4.18
4.13
3.87
4.07
4.09
4.15
4.12
0.05
0.09
0.35
0.12
0.41
0.08
0.24
0.56
0.58
0.13
0.25
0.34
0.27
0.41
0.03
0.15
3.92
3.87
3.99
3.81
4.06
4.03
4.12
4.08
4.07
4.11
3.98
3.44
4.06
4.11
4.10
4.07
0.07
0.44
0.15
0.09
0.25
0.35
0.14
0.04
0.02
0.19
0.41
0.52
0.38
0.72
0.65
0.28
2b
3b
4b
6b
7b
8b
2c
3c
4c
5c
6c
7c
8c
9c
10c

ꢀ
150
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
Fig. 3. CoMFA contour a) Steric and c) Electrostatic maps of 2b; b) Steric and d) Electrostatic maps of 4c.
ability of the N-thiazol groups in the structures of 4c, 5c, and 9c,
thereby ensuring better in vivo activity of 2b.
Regression summaries of CoMFA and CoMSIA results and the
prediction of the pIC50 of newly designed compounds.
is formed between the carbonyl group of the cis-carboxyethyl
scaffold and the hydroxyl group of Thr138 (d ¼ 2.626 Ǻ). Finally,
the ethoxy group oxygen of the trans-carboxyethyl scaffold is
positioned within hydrogen-bonding distance (d ¼ 2.115 Ǻ) from
the NH group of the peptide bond between the Thr138 and
Tyr139. The exact hydrogen-bonding orientation correlates
perfectly with the position of CoMFA and CoMSIA steric, CoMSIA
electrostatic, and hydrogen bond acceptor contours, which
showed carbonyl groups of the 2b C-3 scaffold to be particularly
important for activity. Such compact interaction of 2b with the
3.2.5. Homology modeling and study of the binding mode onto rat
VKORC1
The complete three-dimensional structure of rat VKORC1 is
not available yet, for either experimental or theoretical reasons.
For the purpose of molecular docking studies, we created
a homology model of rat VKORC1. The crystallographic structure
of Synechococcus sp. bacterial vitamin K epoxide reductase was
chosen as the template structure for computer simulations. The
goal of homology modeling was to achieve four transmembrane
enzyme influenced low Ki value of 68.9
mM. Further, advantageous
alignment of the hydrogen atom within 4-OH group of 2b towards
Cys135 (d ¼ 1.720 Ǻ) is assured trough electrostatic interactions
between the hydroxyl group of Tyr139 and the 4-OH group. This
interaction is predicted by the blue CoMFA electrostatic contour.
The obtained active site enzyme-substrate conformation points to
Cys135 as an amino acid participating in the coagulation
mechanism.
a
-helix structures (amino acid positions 11e31; 81e97; 101e123;
127e149) [3] (Fig. 5(b)) and the best possible loop alignment
(Fig. 5(a)). After successful modeling, the structure of rat VKORC1
was relaxed by molecular dynamics simulation; the calculated C
a
The spatial arrangement of 4c (Fig. 6 (b)) led to extension of
the binding space in VKORC1 to one more amino acid, Ile141. Still,
orientation of the coumarin ring itself was quite similar
compared to 2b. Inspection of the 4c binding mode reveals
several positive interactions. Thus, the p-sulfo group is narrowed
to Ile141, presenting one S]O group as an acceptor for one
strong hydrogen bond with the NH group of the peptide bond
between Ala140 and Ile141 (d ¼ 1.324 Ǻ). The interaction is
predicted by the CoMFA steric and CoMSIA electrostatic isopleths.
The coumarin ring is stabilized by two electrostatic interactions.
First one is the interaction between the lactone oxygen atom of
4c and the carbonyl group of the peptide bonds between Thr138
and Tyr139. The second interaction occurs between the carbonyl
oxygen atom of the lactone group and amino group of the
peptide bond of Tyr139 and Ala140. Orientation of the 4-OH
group is stabilized by electrostatic interactions with the Tyr139
hydroxyl group, as in 2b, placing the hydrogen atom in suitable
orientation towards Cys135 (d ¼ 1.735 Ǻ). The obtained CoMFA
electrostatic contours predicted this interaction. Further, positive
hydrophobic interaction between the phenyl residues of 4c and
Tyr139 strengthens binding of the compound which has a Ki
root-mean-square deviation (rmsd) value of 0.24 nm at 720 ps
confirmed that the protein reached a state of equilibration during
MD simulations. On the other hand, the relatively low sequence
identity between bacterial and rat VKORC1 in extracellular
regions made it difficult to predict exact loop conformations in
the specified regions of the homology-modeled protein. Still, high
identity (59%) in the 1e161 amino acid sequence between rat and
bacterial VKORC1 provided excellent alignment between the
modeled and template structure in the transmembrane region of
the protein. This was the basis for successful molecular docking of
the most active compounds in vivo on the selected equilibrium
structure.
Docking simulations have further revealed the considerable role
of the C-3 residues in anticoagulant activity of evaluated derivatives
(Fig. 6).
Thus, the lowest energy conformation of 2b is stabilized within
the active site of rat VKORC1 by the formation of three hydrogen
bonds (Fig. 6(a)). The carbonyl group of the lactone part of the
4-hydroxycoumarin core of 2b is oriented towards the formation
of the first, strong hydrogen bond with the NH group of the
peptide bond of Tyr139 (d ¼ 1.896 Ǻ). The second hydrogen bond
value of 44.8
mM.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
151
Fig. 4. CoMSIA a) Steric, b) Electrostatic and c) Hydrogen bond acceptor contour maps of 2b; d) Steric, e) Electrostatic and f) Hydrogen bond acceptor contour maps of 4c.
According to the results of molecular docking, higher activity of
2b in comparison with 4c can be attributed to the formation of
more hydrogen bonds between 2b and the active site of VKORC1.
epoxide form undergoes the nucleophilic attack of negatively
charged sulfur atom from the thiol group, side chain of the Cys
catalytic site of VKORC1. This results with the formation of covalent
bond between the enzyme and the substrate [21]. However, in work
resulting in discovery of the VKORC1genes, Oldenburg et al.
recently cited the Cys-SH form as the one involved in the epoxide
reduction [3]. This mechanism was considered in the DFT studies of
Deerfield et al., as well, in which they suggested the formation of
a covalent bond between the epoxide and the catalytic amino acid
[8]. The reduction process is ended by the formation of a disulfide
bond [4,21]. The SH form of catalytic amino acid was used in our
mechanistic studies, too. The anticoagulant activity of warfarin is
demonstrated by interaction of deprotonated open side chain
coumarin form of the drug with amino acids of the catalytic site,
leading to the interruption of vitamin K-2,3-epoxide reduction [3].
It was not specified whether the Cys132 or Cys135 amino acid is
involved in interaction with anticoagulant or vitamin K-2,3-
epoxide [3,21].
The inhibition constants of 5c and 9c are 74.2 and 79.6
respectively.
mM,
The binding of the lowest energy rank conformation of warfarin
to VKORC1 occurs by formation of hydrogen bonds between the
lactone carbonyl and the peptide bond of Tyr139 and Ala140, as
well as between carbonyl group of phenylpentan-2-one residue
and the peptide bond between Thr138 and Tyr139. Once again,
orientation of the 4-OH group towards Cys135 is stabilized by
Tyr139. The predicted Ki of warfarin is 5.37 mM.
3.2.6. Mechanistic studies of anticoagulant activity
The exact mechanism of warfarin anticoagulant activity is
a topic that is considered for decades. It is known that this drug acts
as a competitive inhibitor of vitamin K-2,3-epoxide during its
reduction catalyzed by VKORC1 [21]. Fasco et al. suggested that
Fig. 5. a) A cartoon presentation of the alignment of the bacterial and rat VKORC1; b)
a-Helixes and loops of homology-modeled VKORC1; c) Transmembrane rat VKORC1 system.

ꢀ
152
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
Fig. 6. Molecular docking conformations of a) 2b and b) 4c.
To become an anticoagulant agent, the coumarin molecule must
bioavailability. It has been demonstrated by NMR-spectroscopic
studies that warfarin exists in a state of dynamic equilibrium of
seven different isomeric forms, with a distribution dependent on
environmental factors, e.g. solvent polarity and pH [5]. As warfarin
is transported to VKORC1 by binding to serum albumin in the open
be transformed into a structure similar to the quinone form of
vitamin K epoxide. As an anticoagulant drug, warfarin exhibits
chameleon-like isomerism, where the environment-dependent
composition of the ensemble of structures greatly influences its
Scheme 3. The proposed mechanism of the coumarin antagonist activity.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
153
side chain form, it is believed that its deprotonated form binds to
the active site of VKORC1. Similarity between the structures of
warfarin and vitamin K quinone form can be exhibited by even
three major deprotonated open side chain forms of the drug [5]. All
of these forms are considered in our vision of the mechanism
regarding the anticoagulation activity of our compounds and
warfarin. The problem in understanding the anticoagulant activity
of warfarin starts with its inability to form the 4-hydroxycoumarin-
2,3-epoxide and therefore a covalent bond with the catalytic amino
acid. As a consequence, the warfarin antagonist form interacts with
the active site of VKORC1 in other ways that are still debatable.
There are no reported DFT studies that examine the role of warfarin
or any coumarin compound in the neither inhibition mechanism
nor coumarin interaction with the catalytic site of VKORC1.
In these DFT mechanistic studies describing anticoagulant
activity of the tested compounds, we present possible ways of
formation of the open side chain forms that may participate in
inhibition of vitamin K-2,3-epoxide reduction during the coagula-
tion cycle. In our opinion, the anticoagulant activity of evaluated
compounds begins with binding of coumarin open side chain (OSC)
form, i.e. the 4-OH coumarin form, to VKORC1 (Scheme 3) [6]. Several
mechanistic pathways of OSC interactionwith VKORC1 are proposed
in a way that corresponds to warfarin behavior in an aqueous solu-
tion. The SH group of the catalytic amino acid Cys135, identified in
the molecular docking studies, remains unchanged in the first step.
After formation of the coumarin-VKORC1 complex, the first
reaction pathway possibly involves transformation of the OSC form
to its isomeric open side chain tautomer (Scheme 3, OSC-T).
Another transformation of OSC may occur after interaction of the
coumarin inhibitor with the hydroxyl anion, a common species in
cytoplasm, and formation of the deprotonated open side chain form
(Scheme 3, DOSC). When DOSC form is formed, it is stabilized by
intarmolecular delocalization leading to formation of a phenolic
anion (Scheme 3, PA), a form that is similar to the quinone form of
vitamin K-2,3-epoxide. The transformation of PA can further
proceed to formation of the deprotonated open side chain chro-
mone form (Scheme 3, DOSCC).
Thus, all of the structures, OSC, OSC-T, DOSC, and DOSCC are
found in the warfarin aqueous solution and are structurally similar
to the vitamin K quinone form, in conformity with Oldenburg’s and
Deerfield’s consideration of the mechanism [3].
Selected coumarin derivatives are oriented towards the Cys135
in the active site of VKORC1 (Fig. 6., Scheme 3). The SH hydrogen
atom of Cys135 can therefore be attacked by the 4-hydroxy group of
OSC-T, PA, or even DOSCC (Scheme 3). During this step, the amino
acid anion is formed, while coumarin regains the neutral OSC form.
The final step of the mechanism is the spontaneous reformation of
the SH bond of Cys135 by interaction with cellular protonium ion
which is present to establish acidebase equilibrium. The process
correlates with the Oldenburg’s assertion that Cys-SH continues the
mechanistic pathway until an SeS bond is formed. As the reaction
itself occurs in the endoplasmic reticulum, the SH(Cys132)
SH(Cys135) form is reduced by GSSG, but recent studies showed
that even the negative form of SH(Cys135) can enter the reduction
process [22].
The performed DFT studies revealed the reaction pathways by
which coumarin antagonist forms arise and interact with the active
site. Geometries of the OSC forms of 2b, 4c, 5c, 9c, and warfarin,
were extracted from the complexes obtained by molecular docking
and optimized in water as solvent. The energies of the optimized
conformations were: 203.36, 207.98, 214.56, 208.89 and 197.64 kJ/
mol for 2b (Fig. 7., OSC), 4c (Fig. 8., OSC), 5c, 9c and warfarin,
respectively.
The proposed anticoagulant mechanism was computationally
investigated step by step. In the mechanistic research, we were
looking for the transition states that mediate all steps in the
mechanism. Instead, it was found that transition states mediate
only transfers of the SH hydrogen atom of Cys135 towards the OSC-
T and PA forms of 2b, 4c, 5c, 9c, and warfarin (Figs. 7 and 8).
The energy profiles of the reaction between the tautomeric
coumarin forms and Cys135 were investigated first. The energy
value of the optimized OSC-T-2b form (203.42 kJ/mol) was compa-
rable with the energy of OSC-2b. The slight difference in energy of
the two forms suggested that tautomerisation occurs spontaneously
Fig. 7. DFT mechanistic study of anticoagulant activity of 2b.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
154
(Fig. 7.). Interaction between OSC-2b and Cys135 is mediated over
the transition state TS1, formed after the amino acid SH proton has
been attacked by the quinone-shaped 4-OH group. The starting
geometries for transition state optimization were set with respect to
distances between the SH and 4-OH groups that were revealed
during the molecular docking. The activation energy for reformation
of OSC-2b formwas 4.97 kJ/mol. In the TS1 arrangement, the lengths
of Cys135$$$$H and H$$$$2b were 1.4794 and 1.4122 Å, respectively.
The proposed mechanistic pathway also predicts the formation of
unstable protonated open side chain chromone form (POSCC-2b), in
which oxygen atoms of both lactone carbonyl and 4-OH group are
substituted with a hydrogen atom. The equilibrium energy of
POSCC-2b was 8.27 kJ/mol higher than the energy of OSC-2b. The
energy difference is therefore used for the conversion of POSCC-2b
to OSC-2b. It follows that the hydrogen atom of the newly formed
lactone hydroxyl group spontaneously underlies attack of the
cellular hydroxyl group, which results in the reformation of OSC-2b.
The stabilization energy of the reaction was 196.58 kJ/mol.
the lengths of Cys135$$$$H and H$$$$2b were 1.4818 and 1.4126 Å,
respectively.
Formation of the third deprotonated form, DOSCC, was proven
to be energetically very unfavorable. The energy of the DOSCC-2b
conformation was 26.42 kJ/mol higher than the energy of OCS-2b.
As a consequence, all attempts to find the transition state between
the DOSCC and Cys135 resulted in the reverse intramolecular
stabilization of DOSCC to reformation of PA. The process is caused
by the inability of the 4-OH group oxygen to accept a hydrogen
atom due to electron deficiency of the oxygen itself.
Finally, we assumed that the negatively charged Cys135 sulfur
atom is neutralized by the GSSH/2GSH system.
The mechanism of the anticoagulant activity of 4c, 5c, 9c, and
warfarin was considered in the same manner as for 2b, according to
energy profiles of reactions occurring via the proposed mecha-
nisms. Thus, the activation barriers for the TS1 and TS2 arrange-
ments were 8.86 and 8.02, 11.26 and 10.14, 11.36 and 10.88 kJ/mol
and 17.24 and 16.35 kJ/mol, for 4c (Fig. 8.), 5c, 9c, and warfarin,
respectively.
Our attempts to optimize the interaction between the OSC form
of 2b and the hydroxide anion in the second proposed reaction
pathway resulted in the spontaneous transfer (i.e. without any
activation barrier) of a coumarin proton from the 4-OH group to the
OHꢁ anion (Fig. 7.). This rearrangement includes creation of the
DOSC-2b form and intarmolecular anion delocalization (PA-2b),
with a system stabilization value of 194.19 kJ/mol. The lower value
of equilibrium energy of PA-2b in comparison with OSC-T-2b
suggested that the whole anticoagulant reaction is probably
directed towards emergence of the PA rather than the OSC-T form.
The reaction is further guided towards the formation of a negatively
charged Cys135 sulfur atom. This process occurs after the abstrac-
tion of amino acid hydrogen atom by PA and restoration of the
native coumarin form (OSC) via transition state TS2. Here again, the
starting geometries for transition state optimization were set with
respect to distances between the SH and 4-OH groups revealed
after molecular docking. The activation energy for reformation of
the OSC-2b was 4.91 kJ/mol. The lower activation barrier energy for
the reaction between PA and Cys135, compared to OSC-T, is in
agreement with the opinion that the deprotonated form of
coumarin participates in the mechanism. In the TS2 arrangement,
The low value of the activation energy (Ea) of TS2-2b correlates
perfectly with the highest activity in vivo. The rest of the calculated
Ea values correlate nicely with the observed activity.
3.3. Design of novel coumarin anticoagulants and prediction of the
activity
The manner in which prop-1-en-carbonyl/carboxyl and
substituted N-thiazole moieties of the tested compounds contribute
to anticoagulant activity is described above. On the basis of the
material presented, we designed 16 novel improved coumarin
structures (Scheme 4). Using obtained 3D-QSAR models, we calcu-
lated their pIC50 values for potential determination of anticoagulant
activity in vivo (Table 3). The goal of the design was to successfully
implement favorable functional groups, i.e. carboxyethyl, p-SO3H,
m-CH3, and m-NO2, on various coumarin structures. Pharmaco-
phores were added to the modified warfarin core (1d, 7d, 8d, 13d,
14d, 15d, and 16d), 3-amino-4-hydroxycoumarin core (5d and 6d),
pyrano[3,2-c]chromene-2,5-dione core (9d, 10d, 11d), pyrano[3,2-c]
chromen-5(4H)-one core (12d) and 4-hydroxycoumarin core
Fig. 8. DFT mechanistic study of anticoagulant activity of 4c.

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
155
Scheme 4. Structures of newly designed coumarin anticoagulants.
(2d, 3d, and 4d). The novel structures were subjected to the same
molecular modeling and 3D-QSAR methods applied for the tested
compounds.
Use of 3D-QSAR methodology showed five of the designed
compounds, 9d, 7d, 10d, 15d and 16d, to be the most promising
candidates (Table 3) for future synthesis and in vivo administration.
The design was mainly oriented towards the introduction of groups
like p-SO3H, m-CH3 and m-NO2, which were concluded to be
generally less toxic than the carboxyethyl one.
Kragujevac, Serbia. All the coumarin samples were prepared
according to Patent Application [23] for stable warfarin sodium
liquid solutions (2 mL 96% EtOH, 1 mL 0.9% NaCl, 6 mL 20% glycerol
and 1 mL of phosphate buffer, pH ¼ 7), making 1 mg/mL solution.
5.2. Experimental conditions in vivo
5.2.1. Animals and in vivo sample preparation
Experiments were performed on 2.5 month-old adult albino
Wistar rats, weighting 220e250 g, kept at 12 h dark/light intervals,
22 ꢀ 2 ꢂC and 50% relative humidity. The animals were fed with
commercial rat food that was available ad libitum. All the animal
procedures were approved by the Committee for Ethical Animal
Care and Use of the Institute for Biological Research, Belgrade,
which acts in accordance to the Guide for the Care and Use of
Laboratory Animals, published by the US National Institute of
Health (NIH Publication No. 85/23, revised in 1986). Two different
coumarin concentrations, 0.1 and 0.5 mg/kg of body weight [24]
4. Conclusions
In evaluation of the in vivo anticoagulant activity of compounds
1e10c, four of them (2b, 4c, 5c and 9c) presented great potential. The
most active compound was 2b with a measured prothrombin time of
30 s. Evaluated compounds were confirmed as non-toxic agents by
the appropriate biochemical analyses. Compounds were metabolized
in the form of 7-hydroxycoumarin-glucuronides. The given activity
was monitored on the level of molecular target, viz., homology-
modeled rat VKORC1. 3D-QSAR and molecular docking studies
revealed 4-OH group and C-3 diethylmalonate and thiazole-N-Ph-
SO3Hscaffoldsaspharmacophoresthataremandatoryfortheactivity.
Also, DFT studies disclosed crucial interactions between the 4-Oꢁ
coumarinphenolic anionwith Cys135, the catalytic center of VKORC1,
during the prolongation of coagulation. The results of computational
studies were further used for the design of novel compounds.
were administrated to obtain
a proper doseeresponse. The
animals (three per each compound and control) were separated in
three different groups. In the first two, animals were sacrificed 24 h
after the administration of the 0.1 and 0.5 mg/kg coumarin solu-
tions, respectively. In the third group, the animals were treated for
five days, receiving one dose of 0.5 mg/kg per day. The animals were
sacrificed on the sixth day. The ethanol solutions of tested
compounds were administrated intraperitoneal (i.p.). The positive
control was warfarin solution, while negative was 96% ethanol
solution.
5. Experimental protocols
5.1. In vitro sample preparation
5.2.2. Coagulation and biochemical parameters
Plasma, for determination of prothrombin time, fibrin time and
fibrinogen, and serum, for determination of biochemical parame-
A frozen sample of human plasma, blood type Aþ, was taken
from Laboratory for Hematology, Clinical Hospital Center,
ters: bilirubin, AST, ALT,
g-GT, ALP, total amount of proteins,

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
156
albumins, globulins and Fe, were prepared by Quick method [25],
immediately immersed in liquid nitrogen and stored at ꢁ80 ꢂC.
Coagulation experiments were performed on MYTHIC 22-Orphee
hematology analyzer and THROMBOSTAT fibrintimer, while color-
Waters (Eschborn, Germany). Samples were centrifuged at
17500 rpm for 10 min. After conditioning (1 mL of methanol) and
equilibration (2 mL purified water), 1000
passed through the cartridges with gentle positive-pressure at
a flow rate of about 1.5 mL/min. After washing with 500 L of
acetonitrile/water 30/70% in 25 mM formic acid (pH ¼ 3.0), the
samples were eluted under basic conditions with 1000 L of
mL of the supernatant was
imetric measurements were recorded using
Lambda 25 UV/Vis spectrophotometer.
a
PerkineElmer
m
The prothrombin time (PT) in both in vitro and in vivo exper-
iments was measured by Quick method and the results are pre-
sented both in seconds and INR (Table 1). The IC50 values, i.e. the
concentrations that increase prothrombin time for 50%, were
calculated from linear dose-response curve [26] from the results
obtained in vivo. Average PTs were taken for the compounds that
have been administrated in 0.5 mg/kg concentration more than
once. The IC50 values were used for the purpose of the in vivo
3D-QSAR studies. The IC50 values were converted into pIC50 by
taking Log (1/IC50). The coagulation time was determined by
m
acetonitrile/water 40/60% in 25 mM ammonia (pH ¼ 10.0). The
eluates were subsequently centrifuged in vacuo for 30 min to a final
volume of 500e600 mL. The volumes of 25 mL were injected into the
GS/MS system.
5.2.4.2. GC/MS conditions. Metabolite screening was performed
using a single-quadruple mass spectrometer (Hewlett Packard
6890N, Agilent Technologies, Santa Clara, CA, USA) with HP 5MS
column, 30 m ꢃ 0.25 mm, i.d. film thickness 0.25
mm, operated by
Howell method [27]. The activities of AST, ALT and
g-GT on
HP Enhanced ChemStation software. The analyses were performed
in three consecutive steps. First step included scan mode frag-
mentation of ethanol solutions of 2b, 4c, 5c, and 9c. Spectrum of
umbelliferone, a 7-hydoxycoumarin, was also recorded for the
detection of Mþ ion. Second step included the scan and SIM
(Selected ion monitoring) mode for the recognition of each applied
compound in spiked urine samples. Final analysis included the scan
and SIM mode for the recognition of each applied compound in
hydrolyzed urine samples. The analytical conditions were set as
follows: injector (splitless mode) and transfer line temperatures of
290 ꢂC; oven temperature programmed at 200 ꢂC with an increase
of 15 ꢂC/min to 290 ꢂC (isothermal for 30 min); carrier gas helium at
340 nm, and ALP on 405 nm, were determined by UVeVis kinetic
methods according to recommendations of the Expert Panel of the
IFCC (International Federation of Clinical Chemistry) [28e30]. The
colorimetric methods were used for determination of total bili-
rubin (JendrassikeGrof 550 nm method) [31], serum iron (using
Ferene as chromogen, on 595 nm) [32] proteins and fibrinogen
concentration (biuret method on 540 nm) [33], albumin concen-
tration (bromocresol green 628 nm method) [34] and the globu-
lins as the difference in concentrations of total proteins and
albumin.
5.2.3. Statistical evaluations
1.3 ml minꢁ1; injection volume 3
mL; standard electronic impact MS
Statistical evaluation of the data was performed by 1-way
analysis (ANOVA). Variance homogeneity and data distribution
were determined with the Levene and KolmogoroveSmirnov tests,
respectively. Post-hoc comparison between control and treated
groups was performed with the T3 Dunnett’s test or with the
Bonferroni test when the variance was not homogeneous. Statis-
tical analysis was performed using the SPSS statistical software
package, version 13.0 for Windows [35]. The results were consid-
ered to be statistically significant at p < 0.05.
source temperature 230 ꢂC; MS quadruple temperature 150 ꢂC;
mass scan range 50e550 amu at 1800 eV; scan velocity 3.12 scans
s
ꢁ1. SIM mode was chosen for the determination of the 7-hydroxy
metabolites with a dwell time of 100 ms and same chromato-
graphic conditions listed above. Following ions were monitored: m/
z 91, 126, 147, 162 [20] and 362 for 2b; m/z 91, 126, 147, 162 and 432
for 4c; m/z 91, 126, 147, 162 and 366 for 5c; m/z 91, 126, 147, 162 and
397 for 9c.
5.2.4. Sample preparation for the determination of coumarin
metabolites
5.3. Molecular modeling studies
Metabolic pathways of selected coumarin derivatives, 2b, 4c, 5c,
and 9c were determined by new i.p administration, collection of the
rat urine samples and following GC/MS analysis. New healthy
animals were divided into the control ant test groups for the 72 h
long application. Urine was collected on 24 h long intervals.
Samples collected from the test group were used as a control and
for assay validation. For validation, blank urine was spiked with
5.3.1. Molecular optimization of coumarin compounds
The molecular optimizations were performed to determine the
bioactive conformations of administrated coumarin compounds
1e10c. The initial structures were built in Spartan 2008 [37] and
imported in MacroModel 9.5 [38] via graphical interface Maestro
9.0. The conformations were obtained by using of the simulated
annealing molecular dynamics procedure implemented in Macro-
Model as follows [39]. Each structure was energy minimized to
a low gradient. The non-bonded cutoff distances for van der Waals
interactions were set to 8 Å while distances for electrostatic ones
were set to 20 Å. An initial random velocity to all atoms corre-
sponding to 310 K was applied. Three subsequent molecular
dynamics were then performed. The first was carried out for 10 ps
with a 1.5 fs time-step at a constant temperature of 310 K for
equilibration purposes. The next molecular dynamics was carried
out for 20 ps, during which the system is coupled to a 150 K thermal
bath with the time constant of 5 ps. The time constant represents
approximately the half-life for equilibration with the bath; conse-
quently the second molecular dynamics command caused the
molecule to slowly cool to approximately 150 K. The third and last
dynamics cooled the molecule to 50 K over 20 ps. A final energy
minimization was then carried out for 250 iterations using conju-
gate gradient. The minimizations and molecular dynamics were in
all cases performed by implicit salvation in simulated water
appropriate amounts of selected compounds (0.5
of spiked sample volume, 50 L of formic acid (1 M) was added to
mM) [36]. To 1 mL
m
give a final pH ¼ 3.0e4.0. Collected urine samples were stored in
deep freezing ꢁ80 ꢂC conditions. Animals in the test group were
treated for three days on 24 h long intervals by 1 mg/kg coumarin
concentration. The second phase of coumarin rat metabolism nor-
mally occurs by glucuronidation of 7-hydroxycoumarin, which is
the product of the first phase [19]. Therefore, after thawing,
samples were hydrolyzed in order to detect the presence of 7-
hydroxy forms of tested compounds. The 1 mL of urine samples
was adjusted to about pH ¼ 5.0 by addition of 100
(0.5 M) and incubated for 14 h at 37 ꢂC in the presence of 50
-glucuronidase (9000 U/mL). Hydrolyzed samples were adjusted
to pH ¼ 3.0e4.0 by addition of 100 L of formic acid (1 M).
mL acetate buffer
mL of
b
m
5.2.4.1. Sample purification. Sample purification [36] was achieved
using Sep Pak C18 light cartridges (130 mg, 1 mL) supplied by

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
157
aqueous solution using GB/SA (Generalized Born solvent accessible
surface area) keyword and the OPLS2005 force field.
conducted molecular dynamics simulation [46]. In the setup of the
membrane system, pre-equilibrated conformers for each DMPC
molecule were previously generated from the Monte Carlo simu-
lations. This molecule set was developed at 340 K, above the gel-
liquid-crystal phase transition temperature. After the formation of
DMPC, system was annealed to 310 K. The membrane-protein
system was built in Vega ZZ. In all, the complete membrane-
aqueous system consisted of 161 amino acid residues, 128 DMPC
molecules, 11 chlorine counter ions and 3655 water molecules.
Molecular dynamics simulations were performed with NAMD [47]
program, implemented in Vega ZZ [48] using the CHARMM22
PROT force field, after adding Gasteiger charges. Hydrogen atoms
were added using the psfgen package. The starting system was
relaxed by performing 10 000 steps of conjugate gradients energy
minimization to remove unfavorable contacts, followed by 10 000
steps of MD at 310 K (equilibration phase). NAMD was used to
perform 720 ps of MD simulation (with complete long-range elec-
trostatic interactions) and with a time-step of 2 fs. The temperature
was controlled via Langevin dynamics with a dumping factor of 5
psꢁ1. Snapshot structures were extracted for every 1 ps, resulting
720 structures from each trajectory. Analysis of the trajectories
was performed using Vega ZZ software package. All the calculations
were performed at the dielectric constant value of ε ¼ 80.4, in
water as solvent. Thermodynamic properties, such as temperature
and energy, were monitored during MD simulations to check
their convergence to stable values. The average structure after
equilibration was calculated and submitted to energy minimization
performing 100 000 steps of conjugate gradients.
5.3.2. Molecular alignment and CoMFA and CoMSIA modeling
CoMFA and CoMSIA models were obtained by using the Sybyl-X
software of Tripos [40]. All of the molecules were aligned by an
atom-by-atom least-square fit. The 2H-chromene structure was
used as a template. For the purpose of the 3D-QSAR calculations,
the compound’s set was randomly divided into the training set
(10 compounds) and the test-set (6 compounds). The subdivision
was performed in such way that both sets represent equally well
the chemical and biological properties of the whole data set. The
CoMFA descriptor fields including the steric fields and the elec-
trostatic fields were calculated at each lattice with grid spacing of
2 Å and extending to 4 Å units in all three dimensions within
defined region. The van der Waals potentials and Coulombic terms,
which represented steric and electrostatic fields, respectively, were
calculated by using the standard Tripos force field. In CoMFA
method, a sp3 hybridized carbon atom with a charge of þ1 was used
as a probe atom, the energy values of the steric and electrostatic
fields were truncated at 30 kcal/mol. The steric, electrostatic,
hydrophobic, hydrogen bond donor and hydrogen bond acceptor
CoMSIA potential fields were calculated at each lattice intersection
of a regularly spaced grid of 2 Å and extending to 4 Å using a probe
atom with radius 1.0 Å, þ1.0 charge, and hydrophobic and hydrogen
bond properties of þ1. The attenuation factor was set to the default
value of 0.3.
5.3.2.1. Partial least-squares (PLS) analysis. The partial least-
squares (PLS) analysis was used to linearly correlate the CoMFA
and CoMSIA fields to the pIC50 values. The cross-validation analysis
was performed using the leave-one-out (LOO) method in which
one molecule was removed from the data set and its activity was
predicted using the model derived from the rest of the data set. The
number of components in PLS models were optimized by using q2
value, obtained from LOO cross-validation procedure. Column
filtering was used at the default value of 2.0 kcal/mol in the cross-
validation part. The final models were developed with ONC by
using non-cross-validated analysis equal yielded the highest
correlation coefficient (r2).
5.4.2. Molecular docking and energy minimization of
ligandeenzyme complexes
The structure of newly generated rat VKORC1 protein was used
in docking experiments. The structure was further manipulated by
removing nonpolar hydrogen atoms, while Gasteiger charges and
solvent parameters were added. All of the tested compounds were
used as the ligands for docking. The rigid root and rotable bonds
were defined using AutoDockTools. The docking was performed
with AutoDock 4.2 [49]. The dimensions of the grids were thus
18 Å ꢃ 24 Å ꢃ 32 Å, with a spacing of 0.375 Å between the grid
points and center grid box coordinates x ¼ 26.863, y ¼ 68.637,
z ¼ 13.538, referring to the enzyme active site: Thr138, Tyr139 and
Ala140. The Lamarckian Genetic Algorithm was used to generate
orientations or conformations of the ligand within the binding site.
The global optimization started with a population of 200 randomly
positioned individuals, a maximum of 1.0 ꢃ 106 energy evaluations
and a maximum of 27 000 generations. A total of 100 runs were
performed with RMS Cluster Tolerance of 0.5 Å.
Finally, a molecular mechanics approach was applied to refine
the Autodock output. The computational protocol consisted the
application of 10 000 steps of the steepest descent algorithm until
the derivative convergence was 0.01 kcal/Å mol. AMBER force field
with the continuum GB/SA salvation model (water as solvent)
implemented in MacroModel was used during minimization.
Moreover, because of the large number of atoms in the model,
some additional constraints had to be imposed to correctly opti-
mize the complexes obtained by the docking. A subset, comprising
only the inhibitors and shells of residues possessing at least atom at
a distance of 5 Å from any of the inhibitor atoms, was created and
subjected to energy minimization. The inhibitors and all the amino
acids side chains of the shell were unconstrained during energy
minimization to allow the reorientation and proper hydrogen-
bonding geometries and van der Walls contacts. All the atoms
not included in the above-defined subset were fixed, but their
nonbond interactions with all of the relaxing atoms have been
calculated.
5.4. Homology modeling of rat VKORC1
The Rattus norvegicus VKORC1 sequence was retrieved from
publicly available sequence database Swiss-Prot/Uniprot (UniProt
ID: Q6TEK4) [41]. On the base of the sequence search against the
PDB, by using PSIBLAST, the crystallographic structure of Syn-
echococcus sp. bacterial vitamin K epoxide reductase, at the
resolution of 3.60 Å (PDB ID: 3KP9 chain A) [42], was identified
as only and the best structural template (24.37% sequence
identity) for the purpose of modeling the corresponding catalytic
domain (Cys132-Cys135) of rat VKORC1. The alignment of the
template and target sequences was carried out with CLUSTALW
version 1.83 [43]. Swiss-Model [44], an automated comparative
approach, was employed for the prediction of 3D rat VKORC1
structure, using alignment mode. The homology model quality
was assessed by using PROCHECK [45] G-factor, with value of
above ꢁ0.5.
5.4.1. Molecular dynamics of VKORC1
To perform a proper optimization of the newly generated
transmembrane VKORC1 homology model, a membrane-aqueous
system was built around the VKORC1 by using of phospholipide
1,2-dimiristoyleSNe3-phosphorylcholine (DMPC) model (Fig. 5(c)).
The entire protein-membrane-aqueous system was relaxed by

ꢀ
M. Mladenovic et al. / European Journal of Medicinal Chemistry 54 (2012) 144e158
158
[14] M. Shah, P. Patel, M. Phadke, S. Menon, F. Mary, T. Saner, Indian Drugs 39
(2002) 333e337.
5.5. Coumarin VKORC1 inhibition reaction DFT mechanistic studies
[15] S. Manokaran, A. Jaswanth, S. Sengottuvelu, J. Nandhakumar,
We proposed the simple pseudo-enzymatic mechanism for our
derivatives coagulation inhibition that was also applied for
warfarin (Scheme 3, Fig. 7 and 8.). The mechanism is based on the
recent views of Oldenburg et al. [3] and Deerfield et al. [8] on the
roles of vitamin K and warfarin in the coagulation cycle.
R. Duraisamy, D. Karthikeyan, R. Mallegaswari, Res. J. Pharm. Tech.
1
(2008) 398e401.
[16] Y.K. Gupta, M. Sharma, G. Chaudhary, Methods Find. Exp. Clin. Pharmacol. 24
(2002) 451e497.
[17] Y.K. Gupta, M. Sharma, G. Chaudhary, C.K. Katiyar, Phytother. Res. 18 (2004)
362e364.
[18] R.K. Murray, D.K. Granner, P.A. Mayes, V.W. Rodwell, Enzymes: mechanism of
action, in: V.W. Rodwell, P.J. Kennelly (Eds.), Harper’s Illustrated Biochemistry,
twenty-sixth ed. The McGraw-Hill Companies/Lange Medical Books, New
York, 2003, pp. 49e60.
[19] O. Pelkonen, H. Raunio, A. Rautio, J. Maenpaa, M.A. Lang, J. Irch. Coll. Phys.
Surg. 22 (1993) 24e28.
Mechanistic calculations on the selected structures were per-
formed with Gaussian 03 software [50]. Starting geometries of
most active compounds in vivo, 2b, 4c, 5c, and 9c, were optimized
using DFT B3LYP functional and 6-311G(d,p) level of theory. The
geometrical parameters of all stationary points were optimized in
water (ε ¼ 78.36) using the CPCM solvation model. All calculated
structures were confirmed to be local minima (all positive eigen-
values) for ground state structures. The transition state geometries
(TS) were built by Gaussview 3.07 and calculated with DFT B3LYP
functional and 6-311Gþþ(d,p) level of theory, in water (ε ¼ 78.36)
by using the CPCM solvation model. All calculated transition state
structures were confirmed to be first-order saddle points
(one negative eigenvalue), by frequency calculations. The intrinsic
reaction coordinates (IRCs), from the transition states down to the
two lower energy structures, were traced using the IRC routine in
Gaussian in order to verify that each saddle point is linked with two
putative minima. The results of the IRC calculations for one crucial
transition state, TS, are presented in Fig. 7. Evolution of relevant
bonds along the reaction pathway was estimated using the natural
bond orbital analysis.
[20] N.S. Vul’fson, V.I. Zaretskii, V.G. Zaikin, Russ. Chem. Bull. 12 (1963)
2046e2049.
[21] M.J. Fasco, L.M. Principe, J. Biol. Chem. 10 (1982) 4894e4901.
[22] C.S. Sevier, C.A. Kaiser, Nat. Rev. Mol. Cell Biol. 3 (2002) 836e847.
[23] M.E. Anderson, D. Delmarre, D. Gao, M.A. El-khateeb, C.S. Certeno, S.H. Pathak,
PCT/US2006/023438. 2006.
ꢁ
ꢀ
ꢀ
[24] M. Kataranovski, J. Zivanovic, J. Vrankovic, I. Mirkov, D. Kataranovski, Arch.
Biol. Sci. 59 (2007) 53e54.
[25] A.J. Quick, M. Stanley-Brown, F.W. Bancroft, Am. J. Med. Sci. 190 (1935) 501.
[26] M.S. Bhatia, K.B. Ingale, P.B. Choudhari, N.M. Bhatia, R.L. Sawant, Bioorg. Med.
Chem. 17 (2009) 1654e1662.
[27] W.H. Howell, Physiol. Rev. 15 (1935) 435e470.
[28] H.U. Bergmeyer, G.N. Bowers, M. Hørder, D.W. Moss, Clin. Chim. Acta 70
(1976) 19e42.
[29] H.U. Bergmeyer, Clin. Chim. Acta 105 (1980) 147e172.
[30] M.I. Walters, H.W. Gerarde, Microchem. J. 15 (1970) 231e243.
[31] L. Jendrassik, P. Gróf, Biochem. Z. 297 (1938) 82e89.
[32] F. Ceriotti, G. Ceriotti, Clin. Chem. 26 (1980) 327e331.
[33] A.G. Gornall, C.J. Bardawill, M.M. David, J. Biol. Chem. 177 (1949)
751e766.
[34] P.H. Lolekha, W. Charoenpol, Clin. Chem. 20 (1974) 617e619.
[35] S.J. Coakes, L. Steed, P. Dzidic, SPSS Version 13.0 for Windows: Analysis
Without Anguish, John Wiley & Sons Australia, Milton, Qld., 2006.
[36] M. Ufer, B. Kammerer, J. Kirchheiner, A. Rane, J. Swensson, J. Chromat. B 809
(2004) 217e226.
Acknowledgments
This work was financed by Government of the Republic of
Serbia, Ministry of Education and Science, Grants No. III 43004, III
41010 and OI 173020. The authors would like to thank Professor Dr
[37] Software Spartan 2006 for Windows, Wavefunction, Inc., USA, 2006.
[38] F. Mohamadi, N. Richards, W. Guida, R. Liskamp, M. Lipton, C. Caulfield,
G. Chang, T. Hendrikson, W. Still, J. Comput. Chem. 11 (1990) 440e467.
[39] R. Ragno, S. Simeoni, D. Rotili, A. Caroli, G. Botta, G. Brosch, S. Massa, A. Mai,
Eur. J. Med. Chem. 43 (2008) 621e632.
ꢀ
Svetlana Markovic, Department of Chemistry, Faculty of Science,
[40] Sybyl Version X1.1, St. Tripos Associates Inc, Louis (MO), 2010.
[41] B. Boeckmann, A. Bairoch, R. Apweiler, M.C. Blatter, A. Estreicher, E. Gasteiger,
M.J. Martin, K. Michoud, C. O’Donovan, I. Phan, S. Pilbout, M. Schneider,
Nucleic Acids Res. 31 (2003) 365e370.
Kragujevac, Serbia, for useful suggestions on Gaussian computa-
tional modeling. The authors have declared no conflict of interest.
[42] W. Li, S. Schulman, R.J. Dutton, D. Boyd, J. Beckwith, T.A. Rapoport, Nature 463
(2010) 507e512.
[43] J.D. Thompson, D.G. Higgins, T.J. Gibson, Nucleic Acids Res. 22 (1994)
4673e4680.
[44] J. Kopp, T. Schwede, Nucleic Acids Res. 32 (2004) 230e234.
[45] R.A. Laskowski, M.W. MacArthur, D.S. Moss, J.M. Thornton, J. Appl. Cryst. 26
(1993) 283e291.
[46] Y. Zhang, Y.Y. Sham, R. Rajamani, J. Gao, P. Portoghese, ChemBioChem. 6
(2005) 853e859.
[47] J.C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa,
C. Chipot, R.D. Skeel, L. Kale, K. Schulten, J. Comput. Chem. 26 (2005)
1781e1802.
[48] A. Pedretti, A. Villa, G. Vistoli, J. Mol. Graph. Model. 21 (2002) 47e49.
[49] M.G. Morris, S.D. Goodsell, R. Huey, J.A. Olson, J. Comp. Aid. Mol. Des. 10
(1996) 293e304.
[50] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb,
J.R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G.A. Petersson,
H. Nakatsuji, M. Caricato, X. Li, H.P. Hratchian, A.F. Izmaylov, J. Bloino,
G. Zheng, J.L. Sonnenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda,
J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven,
J.A. Montgomery, J.E. Peralta Jr., F. Ogliaro, M. Bearpark, J.J. Heyd,
E. Brothers, K.N. Kudin, V.N. Staroverov, R. Kobayashi, J. Normand,
K. Raghavachari, A. Rendell, J.C. Burant, S.S. Iyengar, J. Tomasi, M. Cossi,
N. Rega, J.M. Millam, M. Klene, J.E. Knox, J.B. Cross, V. Bakken, C. Adamo,
J. Jaramillo, R. Gomperts, R.E. Stratmann, O. Yazyev, A.J. Austin, R. Cammi,
C. Pomelli, J.W. Ochterski, R.L. Martin, K. Morokuma, V.G. Zakrzewski,
G.A. Voth, P. Salvador, J.J. Dannenberg, S. Dapprich, A.D. Daniels, Ö. Farkas,
J.B. Foresman, J.V. Ortiz, J. Cioslowski, D.J. Fox, Gaussian 03 C02, Gaussian,
Inc., Wallingford CT, 2004.
References
[1] P.M. Ridker, S.Z. Goldhaber, E. Danielson, Y. Rosenberg, C.S. Eby, S.R. Deitcher,
M. Cushman, S. Moli, C.M. Kessler, C.G. Eliot, R. Paulson, T. Wong, K.A. Bauer,
B.A. Shwartz, J.P. Miletich, H. Baounameaux, R.J. Glyn, N. Engl. J. Med. 348 (2003)
1425e1434.
[2] J. Hirch, V. Fuster, J. Aussel, J.L. Halpelin, J. Am. Coll. Cardion. 41 (2003)
1633e1652.
[3] J. Olednburg, M. Watzka, S. Rost, C.R.Muller, J. Thromb. Haemost.5 (s1)(2007) 1e6.
[4] I. Petitpas, A.A. Bhattacharya, S. Twine, M. Twine, M. East, S. Curry, J. Biol.
Chem. 276 (2001) 22804e22809.
[5] B.C.G. Karlsson, A.M. Rosengren, P.O. Andersson, I.A. Nicholls, J. Phys. Chem.
111 (2007) 10520e10528.
[6] H. Henschel, B.C.G. Karlsson, A.M. Rosengren, I.A. Nicholls, J. Mol. Struc:-
THEOCHEM 958 (2010) 7e9.
[7] D. Lafitte, V. Lamour, P. Tsvetkov, A.A. Markov, M. Deprez, P. Klich, D. Moras,
C. Briand, R. Gilli, Biochemistry 41 (2002) 7217e7223.
[8] D.E. Deerfield II, C.H. Davis, T. Wymore, D.W. Stafford, L.G. Pedesen, Int. J.
Quantum Chem. 106 (2006) 2944e2952.
[9] M. Grauber, Bioorg. Med. Chem. 15 (2007) 2414e2420.
[10] M. Mladenovic, N. Vukovic, N. Niciforovic, S. Sukdolak, S. Solujic, Molecules 14
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
(2009) 1495e1512.
ꢀ
ꢀ
ꢀ
[11] N. Vukovic, S. Sukdolak, S. Solujic, T. Milosevic, Arch. Pharm. 341 (2008)
491e496.
[12] N.P. Singh, Mutat. Res. 455 (2000) 111e127.
[13] B.C. Tennant, Clinical biochemistry of laboratory animals, in: W.F. Loeb,
F.W. Quimby (Eds.), Assessment of Hepatic Function, Taylor and Francis,
London, 1999, pp. 501e517.