The discovery of inhibitors of Fas-mediated cell death pathway using the combined computational method

Article abstract of DOI:10.1016/j.bmcl.2013.07.018

In this study, pharmacophore and 3D-QSAR models were developed for analogues of 3-substituted-benzofuran-2-carboxylate as inhibitors of Fas-mediated cell death pathways. Our pharmacophore model has good correspondence with experimental results and can exp

Full text of DOI:10.1016/j.bmcl.2013.07.018

Bioorganic & Medicinal Chemistry Letters 23 (2013) 5155–5164
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
The discovery of inhibitors of Fas-mediated cell death pathway using
the combined computational method q
Seo Hee Jung a, Ji Hee Suh b, Eun Hee Kim c, Jun Tae Kim a, Seung-Eun Yoo a, Nam Sook Kang a,
⇑
a Graduate School of New Drug Discovery and Development, Chungnam National University, Daehakno 99, Yuseong-gu, Daejeon 305-764, Republic of Korea
b Bio-organic Science Division, Korea Research Institute of Chemical Technology, Yuseong-gu, Daejeon 305-600, Republic of Korea
c College of Biological Science and Biotechnology, Chungnam National University, Daehakno 99, Yuseong-gu, Daejeon 305-764, Republic of Korea
a r t i c l e i n f o
a b s t r a c t
Article history:
In this study, pharmacophore and 3D-QSAR models were developed for analogues of 3-substituted-ben-
zofuran-2-carboxylate as inhibitors of Fas-mediated cell death pathways. Our pharmacophore model has
good correspondence with experimental results and can explain the variance in biological activities
coherently with respect to the structure of the data set compounds. The predictive power for our synthe-
sized compounds were 0.96 for the pharmacophore model, 0.58 for the comparative molecular field anal-
ysis (CoMFA) model, and 0.57 for the comparative molecular similarity analysis (CoMSIA) model.
Ó 2013 The Authors. Published by Elsevier Ltd. All rights reserved.
Received 5 March 2013
Revised 25 June 2013
Accepted 12 July 2013
Available online 19 July 2013
Keywords:
Fas-mediated cell death pathway
Pharmacophore
3D-QSAR
The contraction or occlusion of blood vessels results in ischemia,
which is frequently associated with cardiovascular diseases, angina
pectoris, coronary artery diseases, and other heart conditions.1,2 Re-
cent studies have shown that caspase-independent cell death
(CICD) mediators including RIP1K, JNK, and PARP are significant
in ischemic cell death.3 Alternatively, caspase-dependent cell death
(CDCD) is also an important alternative pathway for ischemic cell
death.4,5 In particular, Fas-mediated cell death has an important
function in CDCD as a component of the Fas-death inducing signal-
ing complex (DISC) and it also sensitizes cells to chemotherapeutics
in chemotherapeutics-induced cell death.4 Following the Fas-DISC
formation, caspase-8 is activated, which then triggers programmed
cell death, that is, apoptosis.5 Blocking the Fas-mediated cell death
pathway has been reported to be effective in the treatment of ische-
mic strokes, myocardial hypertrophy, and traumatic brain injuries.
In this study, we optimized the inhibitors of the Fas-mediated
cell death pathway screened using pharmacophore and 3D-QSAR
models. We built a combined qualitative pharmacophore and 3D-
QSAR, comparative molecular field analysis (CoMFA)/comparative
molecular similarity analysis (CoMSIA) models of heterocyclic car-
boxylate derivatives (3-substituted-benzofuran, furo[2,3-b]pyri-
dine, thiophene and pyrazole derivatives).5–9 The pharmacophore
model and 3D-QSAR, CoMFA/CoMSIA calculative analyses were
performed using Catalyst and Accerlys Discovery Studio 3.1 and
Tripos Sybyl 2.0, respectively. All chemical inhibitors and their
biological data (e.g., cell death values) were obtained from
high-throughput screening experiments including the work in our
previous paper.5 The cell death (%) values ranged from 4.00% to
30.86% for the 32 compounds selected for the theoretical model
in this research. The biological activities (EC50) were obtained from
the known cell death (%) and EC50 data using a simple regression
analysis, and then the EC50 was converted to pEC50 (=ꢀlogEC50) in
order to obtain useful values. The biological activity values
(pEC50) of the training set and test set compounds cover a range
of more than 3log units (pEC50 = 5–7), as shown in Table 1.
The training set of the pharmacophore model included 7 com-
pounds with high-ranked activity and the remaining 25 compounds
were used as a test data set. The common pharmacophore features
were applied using the HipHop algorithm in order to generate a
qualitative common features model, which was implemented in
the Catalyst molecular modeling software.10,11 Because this algo-
rithm is a useful computational tool for building a 3D pharmaco-
phore, we performed the HipHop pharmacophore modeling using
a highly active compound. In the HipHop run, the most active com-
pound (compound 27) was considered as the reference compound,
which specifies a principal value of 2 and a maximum omitted feat
(MOF) value of 0. The initial features were specified based on an
overview of the training set including the hydrogen-bond donor
(HBD), hydrogen-bond acceptor (HBA), ring aromatic feature (RA),
and hydrophobic feature (HY).
q
This is an open-access article distributed under the terms of the Creative
Commons Attribution-NonCommercial-No Derivative Works License, which
permits non-commercial use, distribution, and reproduction in any medium,
provided the original author and source are credited.
⇑
Corresponding author.
E-mail address: nskang@cnu.ac.kr (N.S. Kang).
0960-894X/$ - see front matter Ó 2013 The Authors. Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmcl.2013.07.018

Table 1
Structure and biological activity of the training and test set
Compound
Structure
pEC50
Compound
Structure
pEC50
O
H
N
-O
Br
H
N
O
N+
S
1
6.24
17
6.02
O
O
O
S
N
O
O
O
O
H
N
H
N
Br
Br
2
5.98
18
5.70
O
Cl
O
S
O
O
O
HO
O
O
Cl
Br
H
N
H
N
N
Br
3
6.19
19
5.80
O
Br
S
S
O
O
O
O
O
O
Cl
Cl
H
N
H
N
S
N
4
6.20
20
5.99
O
S
O
N
O
O
O
O
O
O
H
N
H
N
O
Br
S
5
6.18
21
5.88
N
O
O
S
O
O
O
O
O

Table 1 (continued)
Compound
Structure
pEC50
Compound
Structure
pEC50
Cl
O
O
H
N
H
N
Br
6
5.94
22
5.97
O
O
S
S
N
O
O
O
O
O
O
O
O
N
H
N
H
N
7
6.31
23
5.68
S
S
O
S
N
O
N
O
O
HO
O
O
H
N
S
H
N
S
O
O
8
6.24
24
6.16
O
S
S
O
O
O
O
H
N
O
H
N
S
Br
S
O
9
5.95
25
6.06
O
S
S
O
O
HO
O
-O
H
N
H
N
Br
N
N+
S
O
O
N
S
O
O
O
O
10
6.34
26
6.05
O
O
(continued on next page)

Table 1 (continued)
Compound
Structure
pEC50
Compound
Structure
pEC50
H
N
H
N
N
Br
Br
N
S
S
S
O
O
O
O
11
6.29
27
6.37
O
N
O
NH2
H
N
N
H
N
N
S
N
S
O
O
O
O
O
12
6.29
28
6.36
O
O
H
N
H
N
S
S
S
S
O
O
13
14
6.30
6.25
29
30
5.89
5.65
O
O
O
O
Br
Br
H
N
H
N
S
S
O
O
O
HO
O
O
Cl
H
N
N
H
N
S
S
O
15
6.16
31
5.76
N
O
O
O
N
O
Cl
NH2
H
N
Br
H
N
H
N
S
O
O
16
6.02
32
5.92
O
N
O
O
N
O

S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
5159
Table 2
Summary of CoMFA and CoMSIA models
CoMFA
CoMSIA
Statistical results
q2a
0.59
0.98
0.03
142.78
6
0.60
0.97
0.04
94.28
6
r2b
SEEc
F value
components
Fraction (%)
Steric
Electrostatic
Hydrophobic
Donor
70.20
29.80
9.90
16.90
40.10
15.70
17.40
Acceptor
Grid spacing: 2.0 Å
a
b
c
Cross validated correlation coefficient.
Conventional correlation coefficient.
Standard error of estimate.
Figure 1. The pharmacophore mapping of the most active compound 27. The
pharmacophoric points are color coded with green; hydrogen-bond acceptor (HBA),
magenta; hydrogen-bond donor (HBD), orange; ring aromatic (RA), Cyan; hydro-
phobic (HY).
The predictive power of the quantitative model was verified
using the 25 compounds from the test data set. The best qualitative
pharmacophore model contained six features: one hydrogen-bond
donor (HBD), two hydrogen-bond acceptors (HBA), two ring aro-
matic features (RA), and one hydrophobic feature (HY), with each
distance represented in Figure 1. This model had a maximum cor-
relation coefficient of 0.97 with a high goodness of fit. Also, an
independent test set that contained 25 external compounds was
used to validate the established model and the obtained predictive
score (0.74). The experimental and predicted activities of the train-
ing and test data set compounds are shown in Figure 2. The carbox-
ylate group nearby the 2-position of the thiophene or the
benzofuran ring indicated the importance of the hydrophobic
group in conferring the biological activity. In contrast, when the
carboxyl group was located in this region, it decreased the biolog-
ical activity (e.g., compounds 9, 18, 23, and 30). This model pro-
poses that the hydrophobic interactions between ligands and
proteins are important. In order to clarify this further, the pharma-
cophore parameters of the training data set compounds are re-
ported in Supplementary data Table 1.
6
5
4
3
2
1
Training set R2 = 0.97
Test set R2 = 0.74
0
5.5
5.7
5.9
6.1
6.3
6.5
The 3D-QSAR, CoMFA, and CoMSIA models were constructed
using derivatives based on benzofuran analogues, as shown in
Figure 3. Group 1 contains a training data set of 23 compounds that
produce the 3D-QSAR model (compounds 1–23 in Table 1), and
Group 2 is a test data set of 9 compounds (compounds 24–32 in
Table 1) in order to validate the model. This model can be used
Experimental pEC50
Figure 2. The relationship between experimental data versus pharmacophore fit
value. Training set 7 compounds and test set 25 compounds. The correlation was
calculated using spearman’s correlation efficient.
Figure 3. Stereo-view of the CoMFA/CoMSIA training and test set compounds superimposed in molecular shape analysis.

5160
S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
6.5
6.3
6.1
5.9
5.7
5.5
6.5
(A)
(B)
6.3
6.1
5.9
Training set R2 = 0.98
Test set R2 = 0.72
Training set R2 = 0.97
Test set R2 = 0.73
5.7
5.5
5.5
5.7
5.9
6.1
6.3
6.5
5.5
5.7
5.9
6.1
6.3
6.5
Experimental pEC50
Experimental pEC50
Figure 4. Plots of the experimental versus predicted pEC50 values of the training ( ) and test ( ) compounds based on the best (A) CoMFA model and (B) CoMSIA model.
Figure 5. Contour maps of (A) CoMFA and (B) CoMSIA analysis based on compound 12, high—scored conformation in training set. Green contours signify steric fields where
bulky groups increase activity, while yellow contours indicate fields where less bulky groups decrease activity. CoMFA steric fields were displayed in favored level 80% (green)
and disfavored level 20% (yellow), CoMSIA fields were in 85% and 15% contributions by using STDEV⁄COEFF field. Compound 12 is depicted in capped stick type, referentially.
Figure 6. Contour maps of (A) CoMFA and (B) CoMSIA analysis are based on compound 12. Blue contours signify electrostatic fields where positive charged groups increase
activity, while red contours indicate fields where negative charged groups increase activity. CoMFA electrostatic fields were displayed in favored level 85% (blue) and
disfavored level 15% (red), fraction of CoMSIA fields was identical.
to obtain numerical values for statistical results such as
STDEV⁄COEFF at individual lattice points. Partial atomic charges
were calculated using the Gasteiger–Hückel charge.12 The results
were statistically significant and the data from the 3D-QSAR
analyses are summarized in Table 2 and Figure 4. The best cross-
validated LOO values for the CoMFA and CoMSIA models were
computed as q2 = 0.59 and q2 = 0.60, respectively, using six compo-
nents. The non-cross-validated values for the CoMFA and CoMSIA

S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
5161
Figure 7. Contour map of CoMSIA analysis is based on compound 12. Yellow and white contours (A) signify favorable and unfavorable hydrophobic fields in same proportion.
Hydrogen bond donor contours (B) were displayed in favored level 75% (cyan) and disfavored level 25% (purple). And hydrogen bond acceptor contours (C) were indicated
favorable fraction 90% (magenta) and unfavorable fraction 10% (red).
Table 3
Biological activity and predicted values of synthesized compounds
Compound
Structure
Cell death (%)
9.48
pEC50
6.10
Fit value
2.25
CoMFA
6.29
CoMSIA
6.33
H
N
S
S
33
O
O
O
H
N
Br
Br
34
14.27
5.95
1.05
5.95
5.95
S
S
O
HO
O
H
N
S
S
35
36
10.45
7.57
6.07
6.18
1.95
3.12
6.17
6.30
6.23
6.36
O
N
O
H
N
S
N
O
O
O
S
H
N
S
37
8.38
6.14
3.07
6.25
6.30
O
O
O
(continued on next page)

5162
S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
Table 3 (continued)
Compound
Structure
Cell death (%)
9.00
pEC50
Fit value
3.07
CoMFA
6.16
CoMSIA
6.11
S
H
N
Br
S
38
39
6.12
6.23
O
O
O
S
H
N
S
O
6.45
3.89
6.21
6.48
O
O
Br
Br
Br
H
N
40
16.12
5.91
0.62
5.90
5.97
Br
S
O
HO
O
Br
H
N
Br
41
11.43
13.48
8.47
6.03
1.81
1.38
2.48
6.04
5.91
5.94
6.20
6.16
6.42
S
N
O
O
O
O
O
H
N
42
5.97
O
S
N
O
O
O
H
N
Br
43
6.14
O
S
O
HO
O
H
N
S
S
44
8.57
6.14
3.82
6.09
6.24
O
O
O
Br
H
N
N
N
O
O
45
3.99
6.37
5.19
6.37
6.29
O

S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
5163
Table 3 (continued)
Compound
Structure
Cell death (%)
pEC50
Fit value
CoMFA
6.27
CoMSIA
H
N
N
Br
N
O
O
O
46
2.97
6.45
6.01
6.45
O
X
X
X
Y
Br
Y
Y
R
NH
R
NH
R
R
NH
NH
R
NH2
O
N
O
O
O
O
O
OH
O
S
S
d
S
c
S
a
b
S
O
O
O
O
O
35: R = H, X = Br, Y = S
33: R = H, X = H, Y = S
36: R = H, X = H, Y = NCH3
34: R = CH3, X = Br, Y = S
Br
X
O
O
S
O
A
n
S
n
NH
O
S
n
NH
A
O
O R
NH
S
O
O
R
Z
X
O
40: A = CH, n = 1, R = H, X = Br, Z = 3,5-di-Br
41: A = N, n = 2, R = CH3, Z = 3,5-di-Br
42: A = N, n = 2, R = CH3, Z = 3,4,5-tri-OCH3
37: n = 1, X = H
38: n = 1, X = Br
39: n = 2, X = Br
43: A = O, n = 1, R = H
44: A = S, n = 2, R = CH3
O
O
O
O
O
NH2
O
NO2
NH
N
N
N
N
O
e, f
g
N
H
N
R'
R'
45
a
O
Br
O
O
NH
O
NH
Br
O
N
O
h
N
O
N
N
R'
R'
46
R' = Ph(CH2)2Br
Scheme 1. Reagents and conditions: (a) bromoacethyl bromide, NEt3, THF, rt, 83–93%; (b) (i) benzene thiol, NEt3, THF, rt or (ii) N-methyaniline, K2CO3, DMF, rt, 42–91%; (c)
2 N NaOH, THF, reflux, 90%; (d) (i) di(2-pyridyl)carbonate, DMAP, THF, rt (ii) 40 wt % dimethylamine, rt, 50%; (e) (2-bromoethyl)benzene, Cs2CO3, DMF, rt, 74%; (f) H2(g), 10%
Pd/C, MeOH, 50 PSI, rt, 94%; (g) hydrocinnamoyl chloride, NEt3, THF, rt, 87%; (h) 4-bromophenol, K2CO3, DMF, rt, 69%. And refer to Supplementary data for details.
models were calculated to be r2 = 0.98 and r2 = 0.97 with an esti-
mated standard error of 0.03, 0.04 and F values of 142.78 and
94.28, respectively. The CoMFA and CoMSIA models were validated
by substituting the test set compounds and obtaining sufficient
prediction (r2predictive = 0.72 and 0.73, respectively).
The CoMFA and CoMSIA models were developed based on a
superimposition obtained using the pharmacophoric map. The
contours provide an interpretation of the correlation obtained in
terms of the field contributions. The CoMFA model, which includes
steric and electrostatic fields, is presented as 3D contour map using
the STDEV⁄COEFF field in Figure 5a and Figure 6a. The respective
contributions of the steric and electrostatic fields for the CoMFA
model are 70% and 30%. The green contours signify the steric fields
where the bulky groups increase activity, while the yellow
contours indicate fields where the less bulky groups decrease
activity. As mentioned above, the green contours shown near the

5164
S. H. Jung et al. / Bioorg. Med. Chem. Lett. 23 (2013) 5155–5164
meta-position of the benzofuran signify that any bulky group at
this site could increase activity. The yellow contours near the
40-position of the benzene ring suggest that any bulky group could
decrease activity. Compounds 14 and 25 showed a negative rela-
tionship between the predicted pEC50 and the character of the con-
tours. The blue contours signify electrostatic fields where
positively charged groups increase activity, while the red contours
indicate fields where the negatively charged groups increase activ-
ity. Compounds 1 and 10 were accurately interpreted as red con-
tours; thus, nitro-benzene, which contains a negative group, had
the highest activity.
Similar to the CoMFA fields, the CoMSIA model was also devel-
oped using the STDEV⁄COEFF field, as shown in Figure 5(b), Fig-
ure 6(b), and Figure 7. The respective contributions of the steric
and electrostatic fields for the CoMSIA model were 10% and 17%.
The CoMSIA fields were aligned similarly to the CoMFA fields. In
addition, the CoMSIA model contained hydrophobic, hydrogen-
bond donor, and hydrogen-bond acceptor fields. The respective
contributions of these three fields were 40%, 16%, and 17%. In the
hydrophobic contour map, the yellow contours indicate the sites
where the hydrophobic groups are favorable and the white con-
tours show unfavorable positions. In Figure 7a, the two yellow con-
tours are located in the vicinity of the benzofuran and amine
groups, which signify that the hydrophobic groups facilitate higher
activity. Compounds 10–12 are well fitted to our model, which had
highly predictive ability. The yellow contours near the 10-position
of the pyrazole ring suggest that any hydrophobic substitution
could increase activity. In the hydrogen bond donor contour map,
the cyan contours indicate the regions where the hydrogen bond
donor groups are favorable and the purple contours show unfavor-
able positions in Figure 7(b). Figure 7(c) depicts the hydrogen bond
acceptor contour map. The magenta contours indicate the sites
where the hydrogen bond acceptor groups are favorable for activ-
ity, while the red contours present the regions where hydrogen
bond acceptors are unfavorable. Herein, all contour maps were de-
picted with compound 12 in Figures 5–7.
above employing the appropriate methyl 3-aminothiophene-2-
carboxylate, benzenethiol, and pyrazole. The modification of the
aromatic ring or benzofuran core was achieved using a similar
method as described for the synthesis of compound 33. Following
these procedures, numerous analogues (e.g., benzofuran and ben-
zene derivatives) were prepared in a short period in a parallel syn-
thesis fashion. Consequently, the 14 synthesized compounds were
well-matched with two key ring aromatic centers and they ob-
tained reasonable fitted values compared with the experimental
activity, as summarized in Table 3.
In summary, this study developed pharmacophore and 3D-
QSAR models for analogues of 3-substituted-benzofuran-2-carbox-
ylic ester as inhibitors of the Fas-mediated cell death pathway. The
proposed pharmacophore model has good agreement with the syn-
thesized compounds and is capable of explaining the variance in
biological activities coherently with respect to the structures of
the data set compounds. The predictive power for the synthesized
compounds was 0.96 for the pharmacophore model, 0.58 for the
CoMFA model, and 0.57 for the CoMSIA model. Consequentially,
the prediction of the ligand-based pharmacophore hypothesis pro-
vided significant structural insights and also illuminated the
important binding features of heterocyclic–carboxylic ester ana-
logues. In future research, we will develop strategies for the opti-
mization of the inhibitors of the Fas-mediated cell death
pathway based on these results.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2013.07.
018.
References and notes
1. Hearse, D. J. Cardiovasc. Res. 1994, 28, 1737.
2. Margaix Muñoz, M.; Jiménez Soriano, Y.; Poveda Roda, R.; Sarrión, G. Med. Oral.
Patol. Oral. Cir. Bucal. 2008, 13, E296.
3. Ryu, S. W.; Chae, S. K.; Lee, K. J.; Kim, E. H. Biochem. Biophys. Res. Commun. 1999,
262, 388.
4. Menges, C. W.; Altomare, D. A.; Testa, J. R. Cell Cycle 2009, 8, 2528.
5. Suh, J. H.; Yi, K. Y.; Lee, Y. S.; Kim, E. H.; Yum, E. K.; Yoo, S. E. Bioorg. Med. Chem.
Lett. 2010, 20, 6362.
6. Cramer, R. D.; Patterson, D. E.; Bunce, J. D. J. Med. Chem. Soc. 1988, 110, 5959.
7. Klebe, G.; Abraham, U.; Mietzner, T. J. Med. Chem. 1994, 37, 4130.
8. Kim, S. O.; Cho, I. S.; Gu, H. K.; Lee, D. H.; Lim, H.; Yoo, S. E. Eur. J. Pharmacol.
2004, 487, 81.
9. Suh, J. H.; Yoo, S. E.; Yi, K. Y.; Kim, N. J.; Kim, E. H.; Jung, Y. S.; Lee, Y. S.; Kim, H.
Y. Patent PCT WO 20,09,048,274, 2009.
10. Barnum, D.; Greene, J.; Smellie, A.; Sprague, P. J. Chem. Inf. Comput. 1996, 36,
563.
11. Hecker, E. A.; Duraiswani, C.; Andrea, T. A.; Diller, D. J. J. Chem. Inf. Comput.
2002, 42, 1204.
Based on the well-accepted pharmacophore and 3D-QSAR mod-
els, the 14 new compounds shown in Table 3 were designed and
synthesized in order to predict accurate activities, as presented
in Supplementary data Table 2. Based on our computational mod-
eling and investigation of the SAR, two aromatic features were key
points in the biological activity. According to Figure 1, the distance
between two RA central points was 9.14 Å and we synthesized 14
novel compounds with a carbon linker (n = 4, 5) while adhering to
the distance rule. For this reason, the thiophene and pyrazole
derivatives were synthesized from the corresponding 3-amino-thi-
ophene-2-carboxylic acid methyl ester and 4-nitro-pyrazole-3-car-
boxylate as readily available starter materials, as described in
Scheme 1 and the Supplementary data. The following compounds
were prepared according to the general procedures described
12. Gasteiger, J.; Marsili, M. Tetrahedron 1980, 36, 3219.