Synthesis, biological activity and molecular modeling studies of novel COX-1 inhibitors

Article abstract of DOI:10.1016/j.ejmech.2003.11.011

Synthesis of new potential COX-1 and/or COX-2 inhibitors, derivatives of 1,1-di-(3-carboxyphenyl)ethane, their biological activity, docking results on COX-1 enzyme and absorption, distribution, metabolism, excretion (ADME) properties are presented. In addition to known interactions between ketoprofen and ibuprofen, leading NSAID agents and COX-1 active site, the possibility of formation of additional interactions is explored. Interactions with Ala527, and with one of the water molecules situated within the active site are identified. Molecular mechanics and DFT calculations for studied compounds have revealed free rotation around two central bonds (C₁-C_3′ and C₁-C_3″), making them flexible, thus easier to enter and adjust to the active site. Further modifications of core structure have been undertaken to optimize biological activity and ADME properties. As a result, two of the compounds are indicated as novel COX-1 inhibitors.

Full text of DOI:10.1016/j.ejmech.2003.11.011

European Journal of Medicinal Chemistry 39 (2004) 141–151
www.elsevier.com/locate/ejmech
Original article
Synthesis, biological activity and molecular modeling studies
of novel COX-1 inhibitors
Miljen Martic´, Iva Tatic´ *, Stribor Markovic´, Nedjeljko Kujundžic´, Sanja Koštrun 1
PLIVA Pharmaceutical Industry, Incorporated, Prilaz Baruna Filipovic´a 25, 10000 Zagreb, Croatia
Received 1 July 2003; received in revised form 14 November 2003; accepted 19 November 2003
Abstract
Synthesis of new potential COX-1 and/or COX-2 inhibitors, derivatives of 1,1-di-(3-carboxyphenyl)ethane, their biological activity,
docking results on COX-1 enzyme and absorption, distribution, metabolism, excretion (ADME) properties are presented. In addition to known
interactions between ketoprofen and ibuprofen, leading NSAID agents and COX-1 active site, the possibility of formation of additional
interactions is explored. Interactions with Ala527, and with one of the water molecules situated within the active site are identified. Molecular
mechanics and DFT calculations for studied compounds have revealed free rotation around two central bonds (C1–C3′ and C1–C3″), making
them flexible, thus easier to enter and adjust to the active site. Further modifications of core structure have been undertaken to optimize
biological activity and ADME properties. As a result, two of the compounds are indicated as novel COX-1 inhibitors.
© 2003 Elsevier SAS. All rights reserved.
Keywords: COX-1; NSAID; Ketoprofen; Ibuprofen; Docking; ADME
1. Introduction
COX-2, although unequivocally potent for both enzymes,
can have beneficial effect on most of the chronic inflamma-
tory diseases, and could be more effective and safe in respect
to the selective COX-2 inhibitor [5,6]. Therapeutic, anti-
inflammatory effects of NSAIDs are attributable to their
ability to inhibit COX-1 and COX-2 enzyme [7]. Two of the
most studied substances, 2-(3-benzoylphenyl)propionic acid
(1), generic name ketoprofen (Structure 1), and 2-(4-isobutyl-
phenyl)propionic acid (2), generic name ibuprofen (Structure
2), are among the best NSAIDs [8–10].
From the complex crystal structure of COX-1 enzyme
with 2 (PDB ID: 1EQG), it can be seen that 2 forms three
hydrogen bonds with the catalytic amino acids; one hydrogen
bond with Tyr355 and two hydrogen bonds with Arg120.
Besides these H-bonds, large number of van der Waals inter-
2-Arylpropionic acids belong to non-steroidal anti-
inflammatory drugs (NSAIDs) with well-defined anti-
inflammatory, analgesic and anti-pyretic properties. For this
class of molecules it is found that their S-enantiomers inhibit
prostaglandin synthesis by inhibition of cyclo-oxygenase
(COX) activity of prostaglandin-endoperoxidase synthase
(PGHS) [1]. The most known COX isoforms are COX-1 and
COX-2; COX-1 is constitutively expressed in a wide variety
of tissues, while COX-2 is a highly inducible gene that is
expressed in response to a variety of proinflammatory agents,
cytokines, growth factors and tumor promoters [2]. COX-1
expression is also regulated in response to some pathological
stimuli and is involved in the processes related to carcinogen-
esis [3], atherosclerosis, thrombosis [4] and central nervous
system pathology [2]. COX-2 isoform has been implicated in
inflammatory and other chronical diseases, including cancer
[3] and rheumatic arthritis [5]. Dual inhibition of COX-1 and
Abbreviations: ADME, absorption, distribution, metabolism, excretion;
COX, cyclo-oxygenase; DFT, density functional theory; MM, molecular
mechanics; NSAID, non-steroid anti-inflammatory drug; PGHS,
prostaglandin-endoperoxidase synthase.
* Corresponding author.
E-mail address: iva.tatic@pliva.hr (I. Tatic´).
1 Previously Sanja Sekušak.
Structure 1
© 2003 Elsevier SAS. All rights reserved.
doi:10.1016/j.ejmech.2003.11.011

142
M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
inhibitors (studied compounds) to investigate their biological
activity. Studied compounds 3–9 were tested in vitro for their
inhibition properties on COX-1 and COX-2 enzymes and
compared to reference compounds 1 and 2. Further on, dock-
ing studies have been performed in order to predict and
explain their in vitro activities. Possible binding modes
within COX-1 active site have been analyzed and ligands
were ranked according to their binding affinities.
Conformational analysis of studied compounds has been
done in terms of density functional theory (DFT) and mo-
lecular mechanics (MM) to explore the flexibility of the
studied compounds within the active site.
Caco-2 system is the most commonly used assay to predict
oral absorption and bioavailability of drug candidates. Since
COX inhibitors are administered by peroral route, Caco-2
assay is crucial to properly select potential candidates exerting
Structure 2
actions between various ligands and active site have been
found, including Tyr385, Ala527 and Ser530 [11].
Both 1 and 2 are relatively small molecules and occupy
only a fraction of the active site cavity. This study was,
therefore, aimed to investigate further modifications that
would result in better biological activity. Interactions be-
tween the potential inhibitor molecules 3–9 and amino acids
and/or water molecules within the active site of COX-1
enzyme have been investigated as they turned to be more
potent COX-1 inhibitors.
Besides the fact that the starting diacid 3 (Structure 3)
comprises fragments of both 1 and 2 structures, it also has
two carboxyl acid groups, which are favorable for modifica-
tions. From diacid 3 we prepared six additional potential
high oral bioavailability [12]. Due to the fact that both COX-1
and COX-2 are constitutively expressed in brain, ideal COX
inhibitor should have low blood–brain barrier (BBB) perme-
ation [2,13]. Furthermore, simultaneous optimization of bind-
ing interactions and absorption, distribution, metabolism, ex-
cretion (ADME) properties can reduce time–cost of drug
development, additional in silico ADME property predictions
(solubility, Caco-2, BBB) complemented the overall analysis
of the biological activity of the studied compounds.
2. Chemistry
Diacid 3 was used as a starting molecule for all subsequent
experiments. Scheme 1 shows reaction pathways used for the
preparation of studied compounds 3–9.
Carboxylic acid chloride of 1,1-di-(3-carboxyphenyl)
ethane (10) was used as an intermediary product and was
prepared in large quantities from diacid 3 in a fast reaction.
Structure 3
Scheme 1

M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
143
Table 2
Target molecules were selected so as to provide informa-
COX-1/COX-2 inhibition ratios for 1, 2, 5 and 9
tion on which functional group is the most promising for
further research. Methyl diester 1,1-di-(3-methoxycarbonyl-
phenyl)ethane (4), ethyl diester 1,1-di-(3-ethoxycarbonyl-
phenyl)ethane (5) and isopropyl diester 1,1-di-(3-isopro-
poxycarbonylphenyl)ethane (6) were prepared from chloride
10 in reactions with methanol, ethanol and isopropanol, re-
spectively. Diamide 1,1-di-(3-carboxyamidophenyl)ethane
(7) was prepared in the reaction of carboxyl acid chloride 10
with aqueous ammonia. As a representative of a third func-
tional group nitrile of 1,1-di-(3-carboxyphenyl)ethane (8)
was prepared by dehydration of diamide 7 by phosphorous
oxychloride in the presence of quinoline.
Compound
COX-1/COX-2 10 µM
COX-1/COX-2 1 µM
1
2
5
9
1.28
1.10
3.44
2.29
1.72
1.21
2.03
a
–
a No data for specific concentration ratio (division by zero).
Determination of IC50 values for COX-1 enzyme was
possible for compounds, which showed best inhibition prop-
erties. IC50 value for compound 9 was found to be 1.60 µmol
dm–3, while for compound 5 it is 30 µmol dm–3. Reference
compounds 1 and 2 have IC50 values of 0.30 µmol dm–3 and
0.12 µmol dm–3, respectively. For COX-2 enzyme was pos-
sible to determine IC50 value for compound 9 which was
found to be 28 µmol dm–3, while 1 has IC50 value of
0.27 µmol dm–3 and 2 has IC50 value of 0.70 µmol dm–3.
COX-1/COX-2 ratio for compounds 5 and 9, compared to
1 and 2 is presented in Table 2. Calculated values show that
two best inhibitors, 5 and 9, are more COX-1 selective.
Finally, diketone 1,1-di-(3-benzoylphenyl)ethane (9) was
prepared by Friedl-Crafts reaction from chloride 10 and
benzene in the presence of the aluminum(III)-chloride.
3. Results and discussion
3.1. Determination of COX-1 and COX-2 inhibition
All studied compounds were tested on COX-1 and COX-2
enzymes. Inhibition tests data for both COX enzymes is
presented in Table 1. Basal production of PGE-2 in untrans-
fected COS-7 cells was below the detection limit of the
ELISA kit. This basal production was checked in every
individual experiment performed.
Compound 9 gave the best inhibition response among the
studied compounds. As can be seen from Table 1, inhibition
of COX-1 enzyme for 10 µmol concentration of 9 is almost
identical as 1 and 2, but dramatically decreases for 1 µmol
concentration. The results also show substantial difference in
inhibition responses between 1, 2 and 9 for COX-1 and
COX-2.
Compounds 4, 5, 6 and 9 showed better inhibition re-
sponse on COX-1 enzyme, while compounds 3, 7 and 8 have
greater activity on COX-2 enzyme. However, inhibition of
COX-2 by compounds 3 and 7 were low, and there was no
statistically significant difference between inhibition at
10 and 1 µM. For both enzymes, compound 9 gave the best
inhibition response; 96.1% for COX-1 and 41.9% for COX-2
enzyme.
3.2. Docking studies of receptor–ligand interactions
In vitro results have revealed that studied compounds 3–9
are better inhibitors of COX-1 than of COX-2 enzyme
(Table 1). In our modeling approach we, therefore, focused
on the COX-1 enzyme. Docking studies were performed by
FlexX [14] and GOLD [15] programs. Since FlexX, upon
multiple attempts of changing the size of the active site inside
the protein, failed to dock any of the studied compounds into
the COX-1 active site, only the results obtained by GOLD are
presented. Failure of FlexX can be attributed to two reasons:
its inability to correctly position the base fragments of stud-
ied compounds (which is considered to be the major draw-
back of the incremental docking algorithms) and large frac-
tion of hydrophobic areas in the active site [14]. Because
there is no X-ray data available for the structure of COX-1
with 1 within its active site, we used the structure of COX-1
enzyme with 2 (PDB ID: 1EQG). Active site was defined
from the coordinates of one the ibuprofen carbon atom (IBP
C8), from the complex crystal structure of COX-1 enzyme.
Radius of the active site was 15 Å. No constraints were
added. Based on the analysis of the existing complex crystal
structures of COX-1 and its known inhibitors [16], pharma-
cophore for the COX-1 active site can be deducted. Known
pharmacophore model of the active site consists of five dif-
ferent hydrogen bond interactions: two with Arg120 and one
with Tyr355, Glu524 and Ser530. Additionally, there are
three crystal water molecules in the active site, but none of
the known inhibitors has direct interaction with them [16].
The values of the scoring given by GoldScore fitness
function [15] are given in Table 3. Good qualitative agree-
ment between experimental (Table 1) and predicted (Table 3)
results is obtained for compounds with substantial inhibition
response (1, 2 and 9). In general, compounds, which showed
no activity in vitro, were given the lowest scorings, while
Table 1
Inhibition of COX-1 and COX-2 by 1–9
Compound
Concentration
COX-1 % inhibition
COX-2 % inhibition
10 µmol
96.0
96.0
0.0
1 µmol
10 µmol
74.9
87.4
12.0
13.1
21.8
22.9
6.1
1 µmol
52.2
76.3
17.5
0.0
1
2
3
4
5
6
7
8
9
90.0
92.0
2.8
27.8
75.0
27.8
0.0
16.8
24.0
16.8
0.0
11.8
0.0
18.4
0.0
11.8
96.1
0.0
17.4
41.9
74.6
0.0

144
M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
Table 3
GoldScore fitness function on COX-1 enzyme for 1–9 and list of interactions identified by docking between 1–9 and amino acids in the active site
Compound
Fitness function of the first correctly
docked ligand
55.03
The rank of the first correctly
docked ligand
Interactions with following amino acids within COX-1
active site
1
2
3
4
5
6
7
8
9
1
1
1
3
1
1
1
1
1
Arg120, Glu524, Ser530
Arg120 a, Tyr355
51.12
50.69
Arg120, Tyr355, Tyr385
Arg120, Tyr355
43.50
46.96
Tyr355
50.30
Tyr355, water
43.25
Arg120, Tyr355
51.72
Arg120
65.14
Tyr355, Ala527, water
a Compound 2 forms two hydrogen bonds with Arg120.
those with substantial activity have the highest scorings. How-
ever, some of the docked conformations were placed outside
the active site of the COX-1 enzyme. GOLD docking results
for the first correctly docked conformation, its rank and the list
of amino acids within the active site with which each com-
pound can form hydrogen bonds is presented in Table 3. Only
for methyl diester 4, two conformations with the highest
fitness function were placed outside the active site.
Position of compound 2 obtained by docking is in a very
good agreement with the one in the crystal structure of
COX-1 enzyme with 2 (PDB ID: 1EQG).
Based on the docking results, we propose two additional
interactions between the studied ligands and COX-1 active
site.
The first one invokes a hydrogen bond which links one
keto group with one of the three water molecules present
within the active site. This interaction is particularly evident
for compounds 6 and 9. In addition to this interaction, com-
pound 9 also forms a hydrogen bond with Ala527. Both of
these interactions are observed for the first time and pre-
sented in Figs. 3 and 6.
respectively, are visible. Figs. 4–6 show docked positions of
molecules 1, 2 and 9 within the active site. Particularly
effective fit of compound 9 as compared to 1 and 2, and
favoring interaction between its keto group and the water
molecule inside the active site is clearly visible.
3.3. Conformational analysis
Conformational analysis of the studied molecules has
been undertaken in order to estimate distortions caused by
the interactions within the active site.
To explore conformational space, MMFF94 [17] force
field in the combination with random search (RS) algorithm
was used. A large number of conformers for each compound
were found due to the low rotational barrier of rotable bonds
(Table 4).
Geometry of the most stable conformer was further opti-
mized at B3LYP/6-31G** [18,19] level and compared to
MMFF94 results (Table 4). Good correlation between DFT
and MM energies is observed (Fig. 7). Relatively large root
mean square distance (RMSD) values are due to the large
flexibility of the studied compounds. Although difference in
geometry exists, energy differences are very small for all:
docked, MM minimum energy and DFT conformations.
Therefore, additional single-point energies were calculated
at B3LYP/6-311G** [18,19] level, and rotation around C1–
C3′ bond was studied in 30° increment for all 12 conformers.
However, energy difference was found to be less than a 0.3 kJ
mol–1 even at B3LYP/6-311G** level.
Although Ala527 is not considered as a part of the active
site of COX-1 enzyme [16], it plays an important role in the
binding of compound 9. This interaction should be kept in
mind if designing new class of NSAID-type inhibitors of
COX-1.
Fitness function of compound 9 is significantly higher
from those of known drugs 1 and 2, which can be attributed to
its specific interactions withAla527 and water molecule. Due
to the long hydrophobic channel-like active site, unfavorable
interactions with Arg120 and Glu524 could be avoided. The
shift in the position inside the active site enabled the forma-
tion of H-bonds described in Table 3. Furthermore, this
positioning placed its two outer phenyl rings such as to favor
hydrophobic interactions within the active site (Fig. 6). One
of the outer rings is parallel to the plane formed by guani-
dinium group of Arg120, while connected inner ring is per-
pendicular to Tyr355 phenyl ring. Second outer ring is ori-
ented toward hydrophobic pocked formed by Leu352,
Tyr348, Tyr385, Trp387 and away from hydroxide group of
Ser530.
3.4. ADME profiling
In order to predict ADME properties of the studied com-
pounds Volsurf [20] program was used. The water probe was
used to simulate solvation–desolvation processes, while
DRY and O probe were used to simulate interactions between
the polar groups and hydrophobic core of biological mem-
brane [20]. Volsurf procedure automatically converts 3D
molecular fields into a number of molecular descriptors,
which are chemically easy to understand and interpret. Vol-
surf descriptors describe the size and shape, hydrogen bonds,
hydrophobicity, polarity and the balance between them. On
Fig. 8, hydrophilic and hydrophobic regions for the initial
Figs. 1–3 represent 2D visualization of the active site
where different binding modes for compounds 1, 2 and 9,

M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
145
Fig. 1. Interactions between 1 and COX-1 active site.
compound 3 and the most active compound 9 are given. It can
be seen that compound 9 has larger hydrophobic area as
compared to compound 3. Accordingly, lower solubility and
larger log P values are predicted.
Predictions of ADME properties for studied compounds
are given in Table 5. Compounds 3, 4 and 8 have acceptable
lipophilicity values (log P < 5), while most active compounds
5 and 9 have relatively high log P values. Their solubility
follows the trend of the log P predictions. Compounds 5 and
9 have good permeability, while the permeability of com-
pound 3 is low. Most of the compounds are predicted to have
medium BBB permeation except compounds 3 and 7 for
which low BBB permeation is estimated.
In conclusion, analysis of the predicted ADME properties
for newly prepared compounds 3–9 opens the possibility for
further optimization of studied compounds. Compound 9 is
found to be too lipophilic due to the larger number of phenyl
rings; therefore, it could serve as lead molecule for further
optimization of both specific biological activity and ADME
properties.
(k = 254 nm). IR spectra were recorded on NICOLET
MAGMA-IR 760 spectrophotometer from KBr pelleted
sample or as film. 1H and 13C NMR spectra (300 and
500 MHz) were recorded using BRUKER-AVANCE spec-
trometers with TMS as internal standard.All obtained signals
were marked as: s, singlet; d, doublet; t, triplet; q, quartet and
m, multiplet. For HPLC, THERMOQUEST FINNIGAN
LCQ DECA, was used, with built in mass detector. Mass
spectra were scanned on Micromass, Platform LCZ instru-
ment. CHN analyses were performed on Perkin-Elmer, Se-
ries II, CHNS Analyzer 2400.
For spectra interpretation, numbers were added to follow-
ing carbon atoms.
4. Experimental protocols
4.1.1. 1,1-Di-(3-carboxyphenyl)ethane (3)
A mixture of raw 3-(1-cyanoethyl)benzoic acid (CEBA)
(105.0 g, 0.60 mol), NaOH (75.0 g, 1.87 mol) and distilled
water (655.0 ml) was refluxed for 4 h.A reaction mixture was
cooled and added H2SO4 until pH 2. Formed crystals were
4.1. Chemistry
TLC was performed using Merck Silica gel 60 F254 silica
plates and components were visualized using UV light

146
M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
Fig. 2. Interactions between 2 and COX-1 active site.
isolated and stirred in boiling distilled water (1400 ml).
Undissolved substance was filtered and dried until constant
mass at 105 °C. The sample (6.10 g, m.p. 226 °C) was not
purified for further experiments. For the analysis, obtained
diacid 3 was recrystallized from EtOH and showed: IR (KBr)
mmax (cm–1): 2963, 2875, 2676, 2565, 1686, 1585 and 1420;
1H NMR (DMSO) d (ppm): 1.62 (d, 3H, C(2)H3), 4.37 (q, 1H,
C(1)H), 7.45 (t, 2H, C(5′)H), 7.55 (d, 2H, C(4′)H), 7.79 (d, 2H,
C(6′)H) and 7.83 (s, 2H, C(2′)H); 13C NMR (DMSO) d (ppm):
21.13 (C(2)), 43.30 (C(1)), 127.13 (C(6′)), 128.04 (C(2′)),
128.70 (C(5′)), 131.83 (C(4′)) and 167.20 (C(7′)).
4.1.2. Carboxylic acid chloride of
1,1-di-(3-carboxyphenyl)ethane (10)
A mixture of diacid 3 (2.0 g, 0.0074 mol) and SOCl2
(20 ml, 0.274 mol) was refluxed for 4 h. Obtained solution
was evaporated resulting in chloride 10 (2.3 g, 100%,
m.p. = 98 °C). Product purified by sublimation (t = 220 °C,
p = 0.67 Pa) showed: IR (film) mmax (cm–1): 3448, 2971,
1751, 1594, 1483.
4.1.3. 1,1-Di-(3-methoxycarbonylphenyl)ethane (4)
A mixture of carboxyl acid chloride 10 (0.8 g, 0.00129
mol) and MeOH (20.0 ml, 0.49 mol) was refluxed for 6 h and
evaporated resulting in raw oily product 4 (0.38 g, 98.4%).
Anal. Calc. for C16H14O4 (Mr = 270.29): C, 71.10; H,
5.22. Found: C, 71.08; H, 5.15%.
Table 4
RS results. DFT and RS calculated energies for 3–9 and RMSD between obtained geometries
Compound
Number of
conformers
DHf of the most stable
D(DHf) between 1st and 10th most E(DFT)/Hartree
RMSD
(kJ mol–1
237.424
277.761
256.067
276.841
242.399
187.557
437.437
)
stable conformer (kJ mol–1
)
3
4
5
6
7
8
9
35
0.209
0.167
1.630
2.675
0.543
0.878
3.637
–919.10
–997.71
–1076.35
–1154.98
–897.36
–726.42
–1230.73
2.31
2.37
3.18
3.37
2.22
1.76
3.64
49
194
178
117
17
175

M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
147
H-bond
dH(1)
dH(2)
dH(3)
Atoms
Distance (Å)
2.88
Angle (°)
108.31
152.21
86.44
O–H–O
O–H–O
O–H–N
3.48
2.49
Fig. 3 Interactions between 9 and COX-1 active site.
For the analysis, obtained oily methyl diester 4 was puri-
fied by redistillation (b.p. = 230 °C, p = 0.67 Pa) and showed:
and 7.85 (s, 2H, C(2′)H); 13C NMR (DMSO) d (ppm): 21.64
(C(2)), 44.50 (C(1)), 52.13 (OCH3), 127.05 (C(6′)), 127.82
(C(2′)), 129.05 (C(5′)), 132.30 (C(4′)); MS m/z: 298.88.
Anal. Calc. for C18H18O4 (Mr = 298.34): C, 72.47; H,
6.08. Found: C, 71.38; H, 6.19%.
1
IR (film) mmax (cm–1): 2952, 1720, 1587, 1433; H NMR
(DMSO) d (ppm): 1.68 (d, 3H, C(2)H3), 4.26 (q, 1H, C(1)H),
7.46 (t, 2H, C(5′)H), 7.57 (d, 2H, C(4′)H), 7.80 (d, 2H, C(6′)H)
Table 5
Selected ADME properties of compounds 3–9
Compound
c logP
4.1
a logP98
3.3
Caco-2 permeability
BBB permeation
–0.45
Solubility
–3.86
–3.80
–4.33
–5.21
–3.96
–6.32
–6.55
3
4
5
6
7
8
9
0.25
1.09
1.25
1.25
0.44
1.97
1.26
4.5
3.8
0.38
5.6
4.5
0.50
6.2
5.2
0.24
1.6
2.0
–0.38
3.5
3.8
0.23
6.7
7.1
0.23
Solubility: very low, –8.0 > log Sw; low, –8.0 < log Sw < –6.0; good, –6.0 < log Sw < –4.0; optimal, –4.0 < log Sw < –2.0; very good, –2.0 < log Sw < 0.0. Caco-2
permeability: –1, low; 0, medium; 1, high. BBB permeation: negative values, low; positive values, high; zero, medium.

148
M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
Fig. 4. The active site of COX-1 enzyme (green) with 1 (by atom type).
Fig. 6. The active site of COX-1 enzyme (green) with 9 (by atom type).
4.1.4. 1,1-Di-(3-ethoxycarbonylphenyl)ethane (5)
NMR (DMSO) d (ppm): 1.32 (t, 3H, C(9′)), 1.63 (d, 3H,
C(2)H3), 4.31 (q, 2H, C(8′)), 4.41 (q, 1H, C(1)H), 7.47 (t, 2H,
C(5′)H), 7.58 (d, 2H, C(4′)H), 7.82 (d, 2H, C(6′)H), 7.89 (s, 2H,
C(2′)H); 13C NMR (DMSO) d (ppm): 14.00 (C(9′)), 21.09
(C(2)), 43.26 (C(1)), 60.59 (C(8′)), 126.95 (C(6′)), 127.85
(C(2′)), 128.82 (C(5′)), 132.11 (C(4′)); MS m/z: 326.93.
Anal. Calc. for C20H22O4 (Mr = 326.40): C, 73.60; H,
6.79. Found: C, 73.48; H, 6.70%.
A mixture of carboxyl acid chloride 10 (0.90 g, 0.00333
mol) and 96% EtOH (50.0 ml, 0.86 mol) was refluxed for 4 h
and evaporated. Resulting oily product 5 (0.91 g, 0.00279
mol, 83.4%) was dissolved in CH2Cl2 (13 ml) and washed
with saturated solution of NaHCO3 (2 × 15.0 ml). Organic
layer was separated and evaporated, resulting in oily ethyl
diester
5 (0.61 g, 56.2%). Redistillated diester 5
(b.p. = 235 °C, p = 0.67 Pa) was used for the analysis and
showed: IR (film) mmax (cm–1): 2997, 1719, 1587, 1443; 1H
4.1.5. 1,1-Di-(3-isopropoxycarbonylphenyl)ethane (6)
A mixture of carboxyl acid chloride 10 (0.90 g, 0.00333
mol) and isopropanol (50.0 ml, 0.86 mol) was refluxed for 4 h
and evaporated. Resulting oily product 6 (0.75 g, 63.7%) was
dissolved in CH2Cl2 (30 ml) and washed with saturated
solution of NaHCO3 (2 × 15.0 ml). Organic layer was sepa-
rated and evaporated resulting in oily isopropyl diester 6
(0.63 g, 53.5%). Redistillated isopropyl diester
6
(b.p. = 238 °C, p = 0.67 Pa) was used for the analysis and
showed: IR (film) mmax (cm–1): 2979, 1715, 1587, 1453; 1H
Fig. 7. Correlation between energies obtained by DFT calculations and
MMFF94.
Fig. 5. The active site of COX-1 enzyme (green) with 2 (by atom type).

M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
149
Anal. Calc. for C16H16N2O2 (Mr = 268.32): C, 71.62; H,
6.01; N, 10.44. Found: C, 71.59; H, 5.96; N, 10.35%.
4.1.7. Nitrile of 1,1-di-(3-carboxyphenyl)ethane (8)
A mixture of diamide 7 (2.00 g, 0.00745 mol) and quino-
line (7.5 ml) was heated at 135 °C. While heated, in the
reaction mixture was slowly added POCl3 (4.57 g, 0.0298
mol) within 5 min, so the reaction temperature would not
exceed 150 °C. Obtained mixture was heated next 30 min at
150 °C, after which was cooled to the room temperature. In
cooled mixture, CH2Cl2 was added (50 ml) after which was
added distilled water (20 ml). In obtained mixture, another
50 ml of CH2Cl2 and 20 ml of water was added. Organic layer
was separated and washed with distilled water (2 × 20 ml),
after which was again separated and evaporated. Reaction
yielded in dinitrile 8 (1.11 g, 64.2%). Using column chroma-
tography and CH2Cl2 as a mobile phase, TLC pure dinitrile 8
was obtained (0.83 g, 48.40%) in a form of yellow viscose
mass. For the analysis, dinitrile 8 was additionally purified by
redistillation (b.p. = 245 °C, p = 0.67 Pa) which showed: IR
(film) mmax (cm–1): 2972, 2229, 1719, 1581; 1H NMR
(DMSO) d (ppm): 1.60 (d, 3H, C(2)H3), 4.32 (q, 1H, C(1)H),
7.47 (t, 2H, C(5′)H), 7.62 (d, 2H, C(4′)H), 7.65 (d, 2H, C(6′)H),
7.83 (s, 2H, C(2′)H); 13C NMR (DMSO) d (ppm): 21.23
(C(2)), 43.99 (C(1)), 130.49 (C(6′)), 131.03 (C(2′)), 131.71
(C(5′)), 133.27 (C(4′)); MS m/z: 233.13.
Anal. Calc. for C16H12N2 (Mr = 232.29): C, 82.73; H,
5.21; N, 12.06. Found: C, 82.60; H, 5.06; N, 12.31%.
4.1.8. 1,1-Di-(3-benzoylphenyl)ethane (9)
To a mixture of carboxyl acid chloride (10) (2.0 g, 0.00651
mol) and dry benzene (30.0 ml, 0.337 mol), AlCl3 (3.4 g,
0.0255 mol) was added. Obtained reaction mixture was re-
fluxed for 2 h, cooled to the room temperature and washed
with distilled water (2 × 20 ml). Benzene layer was separated
and evaporated. To resulting residue EtOH (40.0 ml) was
added. Obtained diphenyl 9 in a form of gummy mass was
separated and dried under vacuum (p = 10 Pa) resulting in dry
substance 9 (1.36 g, 53.0%). By column chromatography
using CH2Cl2 as a mobile phase, TLC pure diketone 9 was
obtained (0.76 g, 29.60%). Addition purification was made
for the analysis by redistillation (b.p. = 250 °C, p = 0.67 Pa).
Obtained sample showed: IR (film) mmax (cm–1): 2969, 1725,
1659, 1597; 1H NMR (DMSO) d (ppm): 1.75 (d, 3H, C(2)H3),
3.72 (q, 1H, C(1)H), 7.10–7.80 (m).
Fig. 8. GRID 3D molecular fields of (a) compound 3; and (b) compound 9
calculated with water probe. The regions shown as yellow indicate hydro-
phobic interactions at 1.5 kcal mol–1 and blue regions represent hydrophilic
interactions at –2.0 kcal mol–1
.
NMR (DMSO) d (ppm): 1.30 (d, 6H, C(9′,10′)H3), 1.62 (d, 3H,
C(2)H3), 4.40 (q, 1H, C(1)H), 5.12 (m, 1H, C(8′)H), 7.46 (t, 2H,
C(5′)H), 7.57 (d, 2H, C(4′)H), 7.80 (d, 2H, C(6′)H), 7.85 (s, 2H,
C(2′)H); 13C NMR (DMSO) d (ppm): 21.20 (C(9′,10′)), 21.51
(C(2)), 43.25 (C(1)), 68.03 (C(8′)), 126.92 (C(6′)), 127.83
(C(2′)), 128.82 (C(5′)), 132.07 (C(4′)); MS m/z: 253.15.
Anal. Calc. for C22H26O4 (Mr = 354.45): C, 74.55; H,
7.39. Found: C, 74.25; H, 7.13%.
4.1.6. 1,1-Di-(3-carboxyamidophenyl)ethane (7)
Anal. Calc. for C28H22O2 (Mr = 390.49): C, 86.13; H,
5.68. Found: C, 86.21; H, 5.56%.
A mixture of carboxyl acid chloride 10 (2.5 g, 0.00926
mol) and concentrated solution of aqueous ammonia
(130.0 ml) was refluxed 4 h and evaporated, resulting in
diamide 7 (2.11 g, 84.90%).After recrystallization from 96%
EtOH, diamide 7 was obtained in a form of white powder
(m.p. = 240 °C) which showed: IR (film) mmax (cm–1): 3401,
4.2. Determination of COX-1 and COX-2 inhibition
COX-1 gene was amplified with PCR using primer pairs
containing HindIII and BamHI restriction site: 5′
ATATAAGTCTTGCGCCATGAGCCGGAGTCTC(3), and
5′ ATATGGATCCTCAGAGCTCTGTGGATGGTCGC(3).
COX-2 primer pairs also contained HindIII and BamHI
restriction sites: 5′ ATATAAGCTTGCTGGCGATGCT-
CGCCCGC(3) and 5′ ATATGGATCCCTACAGTTCAGTC-
GAACGTTC(3).
1
3176, 2955, 1662, 1626, 1582, 1400; H NMR (DMSO) d
(ppm): 1.65 (d, 3H, C(2)H3), 3.37 (t, 4H, NH2) 4.28 (q, 1H,
C(1)H), 7.39 (t, 2H, C(5′)H), 7.45 (d, 2H, C(4′)H), 7.72 (d, 2H,
C(6′)H), 7.83 (s, 2H, C(2′)H); 13C NMR (DMSO) d (ppm):
21.15 (C(2)), 43.88 (C(1)), 125.11 (C(6′)), 126.40 (C(2′)),
128.19 (C(5′)), 130.19 (C(4′)), 167.80 (C(7′)); MS m/z: 269.08.

150
M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
Human placenta cDNA library (Clontech) was used as
ADME properties were predicted using Volsurf [20]. We
template in PCR reaction performed with pfx Platinum DNA
polymerase (Invitrogen). Products were cloned in pcDNA3.1
hygro(+) plasmid (Invitrogen) and sequenced.
COS-7 cells were transfected using Lipofectamine 2000
(Invitrogen). Cells were grown in 24-well plate in the incu-
bator (37 °C, 95% relative humidity, 5% CO2) until they
reached 95% confluency. Transfection was performed using
1 µg of plasmid DNA containing COX-1 or COX-2 gene and
2 µl of Lipofectamine 2000 reagent.
Twenty-four hours after transfection, dilutions of tested
compounds were added in duplicate. Each plate contained
positive and negative control. Cells were left for 20 min in the
incubator before addition of arachidonic acid in final concen-
tration of 20 µM. Supernatants were collected after 30 min
and stored on –70 °C.
PGE-2 was measured using PGE-2 competitive ELISA kit
(Amersham) following manufacturer’s recommendation.
Spontaneous production of PGE-2 from endogenously-
released arachidonic acid in both transfected and untransfected
cells was monitored and was below the detection limit of the
kit (20 pg ml–1). PGE-2 level was below the detection limit
also for the untransfected cells treated with exogenous arachi-
donic acid. The total amount of PGE-2 was at average of
610 pg ml–1 for COX-1 transfected cells, and 530 pg ml–1 for
COX-2 transfected cells, which is in the linearity range of the
ELISA detection. These data demonstrate that all detectable
conversion of exogenous arachidonic acid to PGE-2 was due to
the presence of plasmid driven synthesis of COX-1 and COX-2
in transfected COS-7 cells, and that the inhibition data repre-
sents the inhibition of cloned human COX-1 and COX-2 en-
zymes expressed in transiently transfected COS-7 cells.
IC50 values were determined using GraphPadPrism Soft-
ware.2
used water, DRY and O probe with GRID4 force field to
generate the 3D interaction energies. Grid space of 0.5 Å was
used. Caco-2 permeation, BBB permeation and solubility for
these molecules were calculated. Solubility is expressed in
mol l–1 at 25 °C and transformed as a negative logarithm. PLS
analysis [23] was used as a chemometric tool to correlate the
data and build the models of interest. Cerius2 QSAR module
was used for calculations of c log P and a log P98 [24].
GOLD v2.0 and Jaguar v4.2, release 77 were run on
Kayak XM 600, while Sybyl v6.8 (FlexX andVolsurf v3.0.4)
were run on SGI Octane.
Acknowledgements
The authors are thankful to Professor Vitomir Šunjic´,
Head of Division of Organic Chemistry and Biochemistry,
RuWer Boškovic´ Institute, Zagreb and Donatella Verbanac
and Zrinka Ivezic´ from Pliva Inc. for valuable advice and
assistance throughout this project.
References
[1] S.S. Adams, P. Bresloff, C.G. Mason, J. Pharm. Pharmacol. 28 (1976)
256–257.
[2] S. Taniura, H. Kamitani, T. Watanabe, T.E. Eling, J. Biol. Chem. 277
(2002) 16823–16830.
[3] N. Rioux, A. Castonguay, Carcinogenesis 21 (2000) 1745–1751.
[4] G. De Gaetano, M.B. Donati, C. Cerletti, Trends Pharmacol. Sci. 24
(2003) 245–252.
[5] I. Foeldvari, R. Burgos-Vargas, A. Thon, D. Tuerck, J. Rheumatol. 29
(2002) 1079–1083.
[6] K.J. Sales, A.A. Katz, B. Howard, R.P. Soeters, R.P. Millar, H.N. Jab-
bour, Cancer Res. 62 (2002) 424–432.
[7] T.D. Warner, F. Giuliano, I. Vojnovic, A. Bukasa, J.A. Mitchell,
J.R. Vane, Proc. Natl. Acad. Sci. 96 (1996) 7563–7568.
[8] J.G. Lombardino, I.G. Otterness, E.H. Wiseman, Arzneim.-Forsch. 25
(1975) 1629–1635.
4.3. Computational methods
The studied molecules were first minimized using DFT
calculations. We used B3LYP [18] functional and 6-31G**
basis set [19] for the minimization, and 6-311G** basis set
for single-point energy calculations of the minimized struc-
tures. All DFT calculations have been performed using Jag-
uar program package [21]. Conformational analysis has been
performed using random search method and MMFF94 [17]
force field, as implemented in Sybyl program package.3
FlexX [14] and GOLD [15] docking programs were used
to study binding modes of studied molecules within the
COX-1 active site [16]. Structure of the COX-1 enzyme was
used as given in PDB ID: 1EQG. FlexX was used in the
combination with DrugScore scoring function [22], while
GoldScore [15] was used in GOLD. FlexX failed to dock any
of the molecules into the active site, therefore only GOLD
docking results have been presented.
[9] I.M. Hunneyball, Recent developments in disease-modifying anti-
rheumatic drugs, in: E. Jucker (Ed.), Progress in Drug Research, vol.
24, Birkhäuser, Verlag, Basel, 1980, pp. 101–216.
[10] F. Jamali, R. Mehvar, F.M. Pasutto, J. Pharm. Sci. 78 (1989) 695–715.
[11] R.G. Kurumbail, A.M. Stevens, J.K. Gierse, J.J. McDonald,
R.A. Stegeman, J.Y. Pak, D. Gildehaus, J.M. Miyashiro, T.D. Pen-
ning, K. Seibert, P.C. Isakson, W.C. Stallings, Nature 384 (1996)
644–648.
[12] V. Meunier, M. Bourrié, Y. Berger, G. Fabre, Cell Biol. Toxicol. 11
(1995) 187–194.
[13] K. Brune, A. Neubert, Clin. Exp. Rheumatol. 19 (2001) S51–S57.
[14] M. Rarey, B. Kramer, T. Lengauer, G. Klebe, J. Mol. Biol. 261 (1996)
470–489.
[15] G. Jones, P. Willett, R.C. Glen, A.R. Leach, R. Taylor, J. Mol. Biol.
267 (1997) 727–748.
[16] B.S. Selinsky, K. Gupta, C.T. Sharkey, P.J. Loll, Biochemistry 40
(2001) 5172–5180.
[17] T. Halgren, J. Am. Chem. Soc. 112 (1990) 4710–4723.
[18] V. Barone, C. Adamo, J. Chem. Phys. 105 (1996) 24–37.
2 Prism, GraphPad Software, San Diego, CA, USA.
SYBYL® 6.7.1 Tripos Inc., 1699 South Hanley Rd., St. Louis, MI
4 GRIDVersion 19, Molecular Discovery LTD, Berkeley Street, Mayfair,
London.
3
63144, USA.

M. Martic´ et al. / European Journal of Medicinal Chemistry 39 (2004) 141–151
151
[19] P.W. Atkins, R.S. Friedman, Molecular Quantum Mechanics, Oxford
University Press, Oxford, New York, Tokyo, 1997.
[20] G. Cruciani, P. Crivori, P.A. Carrupt, B. Testa, J. Mol. Struct.
Theochem. 503 (2000) 17–30.
[23] G. Klebe, Structural Alignment of Molecules, in: H. Kubinyi (Ed.),
3D-QSAR in Drug Design. Theory, Methods and Applications,
SCOM, Leiden, 1993.
[21] Jaguar 4.2 (1991–2000) Schrödinger, Inc., Portland, OR.
[22] H. Gohlke, M. Hendlich, G. Klebe, J. Mol. Biol. 295 (2000) 337–356.
[24] A.K. Ghose, V.N. Viswanadhan, J.J. Wendoloski, J. Phys. Chem. 102
(1998) 3762–3772.