close

Вход

Забыли?

вход по аккаунту

?

Lead Identification for Modulators of Multidrug Resistance based on in silico Screening with a Pharmacophoric Feature Model.

код для вставкиСкачать
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 317
Thierry Langera,
Monika Edera,
Remy D. Hoffmannb,
Peter Chibac, *,
Gerhard F. Eckerd, *
Lead Identification for Modulators of Multidrug
Resistance based on in silico Screening with a
Pharmacophoric Feature Model
a
Considerable effort has been devoted to the characterization of P-glycoprotein
⫺ drug interaction in the past. Systematic quantitative structure-activity relationship (QSAR) studies identified both predictive physicochemical parameters and
pharmacophoric substructures within homologous series of compounds. Comparative molecular field analysis (CoMFA) led to distinct 3D-QSAR models for
propafenone and phenothiazine analogs. Recently, several pharmacophore
models have been generated for diverse sets of ligands. Starting from a training
set of 15 propafenone-type MDR-modulators, we established a chemical function-based pharmacophore model. The pharmacophoric features identified by
this model were (i) one hydrogen bond acceptor, (ii) one hydrophobic area, (iii)
two aromatic hydrophobic areas, and (iv) one positive ionizable group. In silico
screening of the Derwent World Drug Index using the model led to identification
of 28 compounds. Substances retrieved by database screening are diverse in
structure and include dihydropyridines, chloroquine analogs, phenothiazines,
and terfenadine. On the basis of its general applicability, the presented 3DQSAR model allows in silico screening of virtual compound libraries to identify
new potential lead compounds.
Institute of Pharmacy,
University of Innsbruck,
Innsbruck, Austria
b
Accelrys, Parc Club Orsay,
Université, Orsay Cedex,
France
c
Department of Medical
Chemistry, Medical University
of Vienna, Wien, Austria
d
Department of
Pharmaceutical Chemistry,
University of Vienna, Wien,
Austria
Keywords: P-glycoprotein; Pharmacophore model; in silico Screening; Propafenone; Multidrug resistance
Received: May 15, 2003; Accepted: January 23, 2004 [FP817]
DOI 10.1002/ardp.200300817
Introduction
One of the reasons for the development of multiple
drug resistance is the overexpression of plasma membrane-associated transport proteins that efflux therapeutically administered xenotoxins and thereby prevent substances from reaching their intracellular targets [1]. In cancer cells, broad spectrum resistance to
chemotherapeutic agents is mediated by ATP-driven
drug transporters such as P-glycoprotein (P-gp) [2]. Inhibition of P-gp leads to re-sensitization of multidrug
resistant tumor cells in vitro and was thus considered
a promising approach for treatment of multidrug resistant tumors [3]. Currently, several compounds are in
clinical phase III studies [4].
Use of homologous series of compounds identified
both predictive physicochemical parameters and pharmacophoric substructures [5⫺9]. 3D-QSAR studies
based on comparative molecular field analysis
(CoMFA) and comparative molecular similarity index
Correspondence: Gerhard F. Ecker, Department of Pharmaceutical Chemistry, University of Vienna, Althanstrasse 14,
A-1090 Wien, Austria. Phone: +43 1 4277-55063, Fax:
+43 1 4277-9551, e-mail: gerhard.f.ecker@univie.ac.at
analysis (CoMSIA) led to distinct 3D-QSAR models for
propafenone and phenothiazine analogs [10, 11]. Very
recently, three-dimensional pharmacophore models
have been proposed based on in vitro data for digoxin
transport in Caco-2 cells, vinblastine binding in CEM/
VLB100 cells and vinblastine and calcein accumulation
in LLC-PK1 cells [12]. Additionally, Pajeva and Wiese
proposed a hypothesis for the broad structural variety
of P-gp substrates and inhibitors interacting on the
verapamil binding site [13]. Electron cryo-microscopy
of two-dimensional crystals has led to structural resolution of hamster P-gp at approximately 10Å [14]. In
this case, ligand-based approaches rather than targetbased design represent the method of choice for drug
development. Pharmacophoric feature modeling has
gained importance, since 3D-models for structurally diverse compounds are generated [15]. In a chemical
function-based pharmacophore model, structural motifs are aligned according to their chemical behavior.
The resulting pharmacophore models consist of socalled features that are located in coordinate space as
points surrounded by a tolerance sphere. Each sphere
is meant to be occupied by a certain atom or functional
* Both authors contributed equally to this work.
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
318 Ecker et al.
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Figure 1. Chemical structure of compounds used for the training set.
group in order to match the chemical interactions
specified by the selected features. In contrast to other
3D QSAR methods such as CoMFA and CoMSIA,
models generated by means of pharmacophoric features allow in silico screening of structurally diverse
compound databases [16].
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 319
Table 1. MDR-modulating activity of compounds used
in the training set.
#
EC50 (µM)
Propafenone
GPV 002
GPV 027
GPV 195
GPV 240
GPV 092
GPV 095
GPV 512
GPV 515
GPV 522
KAR 11
B 56
GB 005
LAN 2
LAN 7
0.33
0.91
0.027
0.53
6.47
11.52
0.18
11.52
0.21
0.14
1.06
0.28
0.90
11.66
82.69
In this paper, we propose a general pharmacophoric
feature model for inhibitors of P-gp. The model is
based on a training set of 15 propafenone-type modulators and was validated using 105 compounds of our
in house library. In silico screening of the Derwent
World Drug Index gave 232 hits, which, after refining
the model, were reduced to 28. A literature survey
showed that nine out of those 28 compounds have
previously been described as P-gp-inhibitors, which
supports the validity of the model.
Results and discussion
15 compounds (Figure 1) were selected as training
set. The compounds covered a broad activity range
(EC50 = 0.027⫺82.7 µM) (Table 1) and were selected
on the basis of different core structures to introduce
structural diversity. Feature-based alignments were
obtained in a three-step procedure: First, for each molecule a conformational model was generated using a
sequential poling algorithm, which yields a balanced
coverage of the conformational space. Second, each
Table 2A. Distance matrix of pharmacophoric features in the Catalyst model shown in Figure 2; values are given
in Å.
AR-HYD 1
AR-HYD 2
HBA
pos I
HYD
AR-HYD 1†
AR-HYD 2‡
HBA§
pos I#
HYD‡‡
⫺
7.89
4.38
6.52
9.41
⫺
5.93
6.76
5.94
⫺
7.12
9.72
⫺
4.77
⫺
†
Aromatic-hydrophobic; represents the central aromatic ring.
represents the phenyl ring of the phenylpropionyl side chain.
§
H-bond acceptor.
#
positive ionizable.
‡‡
hydrophobic.
‡
Table 2B. Distance matrix (in Å) of the 3D pharmacophore model obtained by Ekins et al. [12].
AR
HYD 1
HYD 2
HBA
AR
HYD 1
HYD 2
HBA
⫺
12.1
3.5
4.8
⫺
n.d.
10.8
⫺
4.5
⫺
AR ⫺ aromatic, HYD ⫺ hydrophobic, HBA ⫺ H-bond acceptor, n.d. ⫺ not determined.
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
320 Ecker et al.
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Table 2C. Distance matrix (in Å) of the 3D pharmacophore model obtained by Pajeva and Wiese [13].
AR-HYD 1
AR-HYD 2
HBA 1
HBA 2
HBA 3
HBD
AR-HYD 1
AR-HYD 2
HBA 1
HBA 2
HBA 3
HBD
⫺
6.5
4.5
7.4
7.4
8.8
⫺
5.0
2.7
5.0
6.9
⫺
6.6
5.6
8.1
⫺
n.d.
9.1
⫺
n.d.
⫺
AR-HYD ⫺ aromatic-hydrophobic, HBA ⫺ H-bond acceptor, HBD ⫺ H-bond donor, n.d. ⫺ not determined.
Pharmacophoric features identified by the highest
scoring model were: (i) one hydrogen bond acceptor,
(ii) one hydrophobic area, (iii) two aromatic hydrophobic areas, and (iv) one positive ionizable group. A
distance matrix is given in Table 2A.
Recently, Ekins et al. [12] and Pajeva and Wiese [13]
published 3D pharmacophore models of the verapamil
binding site of P-gp. The corresponding distance matrices of the pharmacophoric features obtained are
shown in Tables 2B and 2C. Although both groups
identified several hydrophobic and/or aromatic features, as well as H-bond acceptor features, the models
are not directly comparable with each other. This might
be due to the fact that P-gp shows multiple, partly
overlapping binding sites [25]. This is also supported
by the recently published X-ray structure of AcrB,
which shows a binding cavity of 5000 Å3 being able to
bind up to three ligands simultaneously [26].
Figure 2A. GPV 027 mapped into the pharmacophoric
feature model.
Figure 2B. GPV 522 mapped into the pharmacophoric
feature model.
conformer was examined for the presence of chemical
features. Third, chemical features common to the input
molecules were aligned considering differences in biological activity. The ensuing pharmacophore models,
called hypotheses, were ranked according to rmsd
(root means square deviation).
Figure 2A and B show compounds GPV 027 and GPV
522 fitted to the hypothesis. GPV 027 represents the
most active compound in the training set. The hydrogen bond acceptor is represented by the carbonyl
group, the positive ionizable feature corresponds to
the aliphatic tertiary nitrogen atom and the two hydrophobic aromatic pharmacophores are represented by
the two phenyl rings of the phenylpropiophenone core
structure. The hydrophobic feature is located in close
vicinity of the nitrogen atom and is represented by the
aromatic methyl group.
Previously, we reported that the basic nitrogen atom
in propafenone-type modulators of P-gp represents an
important pharmacophoric group and that it interacts
on basis of H-bond acceptor features rather than
charge [27]. Most of the compounds under consideration contain a basic nitrogen atom. The software
package Catalyst includes basic primary, secondary
and tertiary amines as positive ionizable groups; in an
aqueous environment these are protonated at physio-
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 321
logical pH. An optional definition of nitrogen atoms as
H-bond acceptors is not provided by the software
package. Therefore, the basic nitrogen atoms in our
training set had to be represented as positive ionizable, otherwise an important descriptor for activity
would have been lost in the model. When we excluded
the positive ionizable function from model generation,
we obtained models containing an additional H-bond
acceptor function in proximity to the previously found
positively charged center.
Validation of the model
When searching a database with a hypothesis, by default only, compounds which exhibit mapping of all fea-
Table 3A. Experimental and predicted activity of the compounds of the test set: Core structures and aminesubstituents.
Core structures
Amines
Amines
A
A1
A8
B
A2
A9
C
A3
A10
D
A4
A11
E
A5
A12
A6
A13
A7
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
322 Ecker et al.
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Table 3B. Experimental and predicted activity of the compounds of the test set: Chemical structure and MDRmodulating activity of the compounds used in the test set. EC50 values are given in µM; inact: EC50 =
3,100,000 µM.
Code
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
005
009
012
017
019
021
023
025
029
031
045
046
048
049
050
051
057
062
073
088
090
091
093
094
128
129
134
135
149
155
156
157
159
163
164
180
181
182
184
186
189
201
206
211
216
220
226
227
231
232
233
238
Core
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
B
B
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
R1⫺R4
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R3
R1
R1
R1
R1
R1
R1
R1
R3
R2
R3
R1
R1
R2
R3
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
R1
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
COCH2CH2Ph
COCH2CH2Ph
COCH2CH3
COCH3
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH3
COCH2CH3
COCH2CH2Ph
COCH3
COCH2CH2Ph
COPh
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
CH(OH)CH2CH2Ph
CH(OCH3)CH2CH2Ph
CH2CH2Ph
CH2CH2Ph
CH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph, R3 = OH
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
CH(OH)CH2CH2Ph
CH(OCH3)CH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
CH(OH)CH2CH2Ph
CH(OCH3)CH2CH2Ph
COCH2CH2-(1-naphthyl)
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2-(1-naphthyl)
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
COCH2CH2Ph
CH(OCH3)CH2CH2Ph
CH(OH)CH2CH2Ph
CH(OCH3)CH2CH2Ph
COCH2CH2Ph, R3 = OBz
COCH2CH2Ph, R3 = OBz
COCH2CH2Ph, R3 = OH
COCH2CH2Ph
L
A
EC50
obs.
EC50
pred.
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺(CH2)2⫺
(S) ⫺CH(OH)CH2⫺
(R) ⫺CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺(CH2)3⫺
⫺(CH2)2⫺
⫺(CH2)5⫺
⫺(CH2)6⫺
⫺(CH2)7⫺
⫺(CH2)8⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
A1
⫺N(iPr)2
A1
A1
A3
A4
A5
A6
A10
A8
A8
A2
⫺N(CH3)2
A2
⫺N(CH3)nPr
⫺N(C2H5)2
A2
A12
A1
A1
A1
A1
⫺NH(nPr)
⫺NH(nPr)
A11
⫺NH(nPr)
A8
A1
A2
A8
A8
A8
⫺N(iPr)2
⫺N(iPr)2
⫺N(iPr)2
A1
⫺NH(CH2CH(Ph)2)
⫺NH(CH(Ph)2)2
⫺NH(CH2CH2CH(Ph)2)
A1
A1
A1
A1
A1
A1
⫺NH(CH2CH2CH(Ph)2)
A2
A2
⫺NH(nPr)
A1
A1
⫺NH(CH2CH2CH(Ph)2)
0.60
0.38
14.3
32.0
0.61
0.23
0.68
0.26
0.66
0.07
2.09
207
2.48
67.3
0.42
2.32
3.65
0.06
0.97
0.90
0.17
0.98
1.06
1.10
0.26
3.02
2.54
0.42
6.88
0.67
0.23
1.12
0.92
1.74
0.66
0.17
0.80
10.4
0.72
0.65
1.45
0.24
0.20
0.18
0.14
0.75
9.54
1.81
0.11
0.17
1.73
0.72
0.36
0.39
inact
inact
0.005
0.02
0.009
0.004
0.002
0.02
inact
inact
inact
inact
0.004
1.20
inact
0.01
43000
0.24
66.0
inact
inact
inact
0.20
0.002
1.60
16.0
inact
0.01
0.71
0.07
200
0.31
2.20
0.28
0.002
0.002
0.002
0.30
0.47
0.39
0.22
34.0
500
0.01
inact
inact
0.008
0.30
0.59
0.002
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 323
Table 3B. (continued)
Code
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
GPV
242
245
253
264
317
319
321
323
334
335
336
338
339
354
356
357
339
358
359
360
361
363
366
374
376
381
382
384
385
386
388
389
390
391
470
472
476
479
485
491
523
543
556
557
558
559
574
576
577
579
596
598
600
608
Core
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
E
A
E
D
C
C
C
C
A
A
A
A
A
A
A
A
R1⫺R4
L
A
EC50
obs.
EC50
pred.
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph, R3 = OH
R1 = COCH2CH2Ph, R3 = OBz
R1 = CH(OiPr)CH2CH2Ph
R1 = COPh
R1 = COPh
R1 = COCH3
R1 = COCH3
R1 = COCH2CH2(4-Cl-Ph)
R1 = COCH2CH2(4-CH3-Ph)
R1 = COCH2CH2(4-OCH3-Ph)
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph, R3 = Cl
R1 = COCH2CH2Ph, R3 = OCH3
R1 = COCH2CH2Ph, R3 = CH3
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2-(1-naphthyl)
R1 = COCH2CH2-(1-naphthyl)
R1 = COCH2CH3
R1 = COCH2CH2-(1-naphthyl)
R2 = COCH3
R2 = COCH3
R2 = COCH3
R1 = COCH2CH2Ph
R3 = COCH3
R3 = COCH3
R3 = COCH3
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph
R1 = COCH2CH2Ph, R3,R4 = Cl
R1 = COCH3, R3 = CH3
R1 = H, R2 = O, R3 = H,H
R1 = COCH2CH2Ph
R1 = CH2Ph, R2 = O, R3 = H,H
⫺
cis; R1 = CH2CH2Ph
trans; R1 = CH2CH2Ph
cis; R1 = CH2CH2Ph
trans; R1 = CH2CH2Ph
R3 = COCH3
R1 = COCH2CH2Ph
R3 = COCH3
R2 = COCH3
R2 = COCH3
R1 = COCH3
R3 = COCH2CH2Ph
R1 = COCH2CH2Ph, R4 = OCH3
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OAc)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺(CH2)4⫺
⫺(CH2)4⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2⫺
⫺CH(OH)CH2⫺
⫺CH(OH)CH2⫺
⫺CH(OH)CH2⫺
⫺CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺CH2CH(OH)CH2⫺
⫺NH(cyclohexyl)
⫺N(iPr)2
A10
A1
A12
A8
A11
A12
A7
A7
A7
A1
⫺NH(Ph-4-COOCH3)
A7
A7
A7
⫺NH(Ph-4-COOCH3)
⫺NH(Ph-4-CF3)
⫺NH(Ph-4-NO2)
⫺N(COCH2CH3)Bz
A2
A8
⫺N(nPr)COPh
A2
⫺N(iPr)2
A12
A12
A2
A1
A8
A13
A1
A8
A2
⫺NHCH(iPr)Ph
⫺NHCH(cyclohexyl)Ph
A7
A1
A1
⫺NHCH(tBu)Ph
A9
A1
A1
A1
A1
A1
A9
A9
A7
A9
A7
A9
A9
A7
0.34
1.04
0.12
0.46
0.31
0.15
0.36
3.20
0.02
0.02
0.01
0.34
1.54
0.06
0.18
0.03
1.54
8.49
2.57
1.65
1.01
0.19
1.69
0.73
0.47
0.30
0.08
128
9.07
10.1
0.13
49.0
11.9
302
0.21
0.14
0.02
5.47
9.68
0.29
0.17
176.3
0.65
1.14
2.10
1.73
1.35
0.006
1.55
0.38
0.54
0.20
0.23
0.30
0.01
0.32
0.002
0.58
0.12
0.19
2400
inact
0.004
0.004
0.004
0.71
inact
0.002
0.01
0.003
inact
inact
inact
inact
200
0.01
inact
17.0
0.53
inact
0.002
inact
inact
inact
0.09
inact
inact
inact
0.002
0.005
0.02
inact
inact
0.007
0.01
inact
inact
inact
inact
inact
12.0
0.004
3600
0.82
0.90
2000
7.3
0.06
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
324 Ecker et al.
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Table 4. Shape parameter setting applied for database search.
Min. Percent Extent
x: 0.9 Å
y: 0.9 Å
z: 0.9 Å
Max. Percent Extent x: 1.1 Å
y: 1.1 Å
z: 1.1 Å
Percent Box
Volume Match
min. 90 % max. 110 %
Similarity Tolerances min. 0.5
Grid Resolution
1
Bit Volume Padding
2
max. 1
tures are retrieved, while those with partial mapping
are omitted. Consequently, prediction of pharmacological activity by Catalyst is a rough estimate as com-
pared to QSAR models which can potentially yield
quite precise estimates. Compounds missing several
features are often classified as completely inactive
(EC50 = 3,100,000 µM) without further differentiation.
Thus, we used a classification approach rather than
regression analysis for validation of our model. For this
purpose, a test set of 105 propafenone type inhibitors
with activities ranging from 0.006 to 302 µM was used
(Table 3) [28]. These compounds were ranked according to their EC50 values and the top 30 % and bottom
30 % were identified. EC50 values of the compounds
were predicted, using the pharmacophore model and
compared to measured activities. Within the top 30 %
(n = 35; activity range 0.006⫺0.30 µM), only one compound out of 35 (2.8 %) was predicted as completely
inactive (GPV 381) and two compounds were predicted with an EC50 value more than three orders of
magnitude off the experimentally determined value
Figure 3. Chemical structure of known MDR-modulators identified in the in silico screen.
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 325
Figure 4. Chemical structure of compounds identified in the in silico screen as potential MDR-modulators.
(GPV 216: 0.14 vs. 500 µM; GPV 598: 0.20 vs. 2000
µM). Of the bottom 30 % of compounds (n = 35; activity
range 1.45⫺302.1 µM), 28 substances (80.0 %) were
predicted as completely inactive, five compounds were
predicted with EC50 values matching the observed activity (residual within one order of magnitude), and two
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
326 Ecker et al.
compounds were classified as false positives (GPV
129 and GPV 182) with predicted EC50 values of
0.0024 µM (actual 3.02 µM) and 0.0017 µM (actual
10.4 µM), respectively. Thus, more than 80 % of the
compounds were correctly classified on basis of a high
activity/low activity classification system. We considered this score high enough to use this model for
in silico screening of compound libraries.
Database screen
The pharmacophore hypothesis was used as a 3D
query in a screen of the world drug index (WDI) to test
the usefulness of the model in identifying structurally
novel P-gp inhibitors. In order to reduce the number of
hits, a shape constraint was introduced using the volume of compound GPV 027, the most highly active
compound in the training set, as an additional filter. In
a first run 232 entries were identified. A further reduction in the number of retrieved compounds was accomplished by stepwise variation of shape query tolerances. Applying the final setting reported in Table 4,
we were able to reduce the number of hits to a total of
28. Of those, nine compounds have previously been
reported being inhibitors of P-gp (Figure 3). These
were AHR-16303B [29], asocainol [25], benidipine
[30], CGP-54103 [25], CP-117227 [31], CP-215548
[32], homofenazine [33], mepacrine [34], and terfenadine [35]. These compounds belong to several different
structural classes of MDR-modulators and comprise a
chloroquine analog (mepacrine), a dihydropyridine
(benidipine), a phenothiazine (homofenazine), and
aryloxypropanolamines (CP-117227 and CP-215548).
Within the remaining 19 compounds (Figure 4), structural analogs of phenothiazines, flavonoids, quinine,
terfenadine, and opioid receptor ligands, all of which
belong to known classes of P-gp-inhibitors, are represented. This indicates that these classes of P-gpinhibitors share a common pharmacophore assembly
and might thus share the same binding domain. The
validity of the model enables screening of structurally
diverse virtual libraries and might thus be useful in the
identification of new classes of P-gp-inhibitors.
Acknowledgments
This work was supported by a grant from the Austrian
Science Fund (FWF) (Grant P17014-B11).
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
pafenone, GPV 002, GPV 027, GPV 195, GPV 240), two
benzofuranes (GPV 092, GPV 095), three indanones (GPV
512, GPV 515, GPV 522), and five pyrazolones (KAR11, B
56, GB 005, LAN 2, LAN 7). Compounds were synthesized
as described previously [17⫺22].
Pharmacological activity
Cell lines
The human T-lymphoblast cell line CCRF-CEM and the multidrug resistant lines CCRF VCR1000 and CCRF adr5000
were provided by V. Gekeler (Byk Gulden, Konstanz, Germany). The resistant lines were obtained by stepwise selection in vincristine- or daunorubicin-containing medium [23].
Cells were kept under standard culture conditions (RPMI1640
medium, supplemented with 10 % fetal calf serum). The Pgp-expressing resistant cell lines were cultured in presence
of 1000 ng/mL vincristine or 5000 ng/mL daunorubicin,
respectively. One week prior to the experiments, cells were
transferred into medium without selective agents or antibiotics.
Daunorubicin efflux studies
Daunorubicin efflux studies were performed as described
[24]. Generally, eight serial dilutions were tested for each
modulator. Parental CCRF-CEM cells were used to correct
for simple membrane diffusion, which was less than 3 % of
the efflux rate observed in resistant cells. Dose response
curves were fitted to the data points using non-linear least
squares, and EC50 values were calculated as described [23].
EC50 values of individual compounds are given in Tables 1
and 2 and represent the average and standard deviation of
at least triplicate determinations.
Pharmacophoric feature modeling
3D structures of the compounds were built interactively using
Catalyst version 4.0. The number of conformers generated
using the “best” feature of the program for each substrate
was limited within the program to a maximum of 255 with an
energy range of 15.00 kcal/mol beyond the calculated potential energy minimum. Ten hypotheses were generated using
these conformer structures for the molecules in the training
set and the EC50 values after selection of the following features: hydrogen bond acceptor (HBA), hydrophobic (H), aromatic hydrophobic (HA), and positive ionizable (PI). A couple
of constraints have been imposed on the hypothesis generator: (i) only hypotheses with five features have been kept,
because of the molecular flexibility and functional complexity
of the training set, (ii) as the literature reports a nitrogen atom
(either represented as a positive ionizable group or as HBA)
as critical substructure for P-gp inhibition, the program was
forced to include such a feature in the composition of hypotheses.
References
[1] R. L. Juliano, V. Ling, Biochim. Biophys. Acta 1976,
455, 152⫺162.
Experimental
[2] M. Horio, M. M. Gottesman, I. Pastan, Proc. Natl. Acad.
Sci. USA 1988, 85, 3580⫺3584.
Compounds
[3] M. Raderer,
3553⫺3563.
Compounds depicted in Figure 1 were selected as a training
set. These included five phenylpropiophenones (pro-
[4] S. F. Bates, C. Chen, R. Robey, M. Kang, W. D. Figg, T.
Fojo, Novartis Found. Symp. 2002, 243, 83⫺96.
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
W. Scheithauer, Cancer
1993, 72,
Arch. Pharm. Pharm. Med. Chem. 2004, 337, 317−327
Lead Identification for Modulators of Multidrug Resistence 327
[5] G. Ecker, P. Chiba, Recent Res. Devel. Medicinal Chem.
2001, 1, 121⫺137.
[6] G. Klopman, M. S. Leming, A. Ramu, Mol. Pharmacol.
1997, 52, 323⫺334.
[7] T. Osterberg, U. Norinder, Eur. J. Pharm. Sci. 2000,
10, 295⫺303.
[8] H. L. Pearce, A. R. Safa, N. J. Bach, M. A. Winter, M.
C. Cirtain, W. T. Beck, Proc. Natl. Acad. Sci. USA 1989,
86, 5128⫺5132.
[9] A. Seelig, Int. J. Clin. Pharmacol. Ther. 1998, 36,
50⫺54.
[10] I. K. Pajeva, M. Wiese, Quant. Struct.-Act. Relat. 1998,
17, 301⫺312.
[11] I. Pajeva, M. Wiese, J. Med. Chem. 1998, 41,
1815⫺1826.
[12] S. Ekins, R. B. Kim, B. F. Leake, A. H. Dantzig, E. G.
Schuetz, L. B. Lan, K. Yasuda, R. L. Shepard, M. A.
Winter, J. D. Schuetz, J. H. Wikel, S. A. Wrighton, Mol.
Pharmacol. 2002, 61, 964⫺973.
[13] I. K. Pajeva, M. Wiese, J. Med. Chem, 2002, 45,
5671⫺5686.
[14] M. F. Rosenberg, G. Velarde, R. C. Ford, C. Martin, G.
Berridge, I. D. Kerr, R. Callaghan, A. Schmidlin, C.
Wooding, K. J. Linton, C. F. Higgins, EMBO J. 2001,
20, 5615⫺5625.
[15] S. Ekins, G. Bravi, S. Binkley, J. S. Gillespie, B. J. Ring,
J. H. Wikel, S. A. Wrighton, Pharmacogenetics 1999,
9, 477⫺489.
[16] P. W. Sprague, Perspect. Drug Discovery Des, 1995,
3, 1⫺20.
[17] P. Chiba, S. Burghofer, E. Richter, B. Tell, A. Moser, G.
Ecker, J. Med. Chem. 1995, 38, 2789⫺2793.
[18] P. Chiba, M. Hitzler, E. Richter, M. Huber, C. Tmej, G.
Ecker, Quant. Struct.-Act. Rel. 1997, 16, 361⫺366.
[19] P. Chiba, D. Annibali, M. Hitzler, E. Richter, G. Ecker,
Il Farmaco 1998, 53, 357⫺364.
[20] P. Chiba, W. Holzer, M. Landau, G. Bechmann, K. Lorenz, B. Plagens, M. Hitzler, E. Richter, G. Ecker, J. Med.
Chem. 1998, 41, 4001⫺4011.
[21] G. Ecker, P. Chiba, M. Hitzler, D. Schmid, K. Visser, H.
P. Cordes, J. Csöllei, J. K. Seydel, K.-J. Schaper, J. Med.
Chem. 1996, 39, 4767⫺4774.
[22] M. Salem, E. Richter, M. Hitzler, P. Chiba, G. Ecker, Sci.
Pharm. 1998, 66, 147⫺158.
[23] R. Boer, V. Gekeler, W. R. Ulrich, P. Zimmermann, W.
Ise, A. Schodl, S. Haas, Eur. J. Cancer 1996, 32A,
857⫺861.
[24] P. Chiba, G. Ecker, D. Schmid, J. Drach, B. Tell, S.
Goldenberg, V. Gekeler, Mol. Pharmacol. 1996, 49,
1122⫺1130.
[25] S. V. Ambudkar, C. Kimchi-Sarfaty, Z. E. Sauna, M. M.
Gottesman, Oncogene 2003, 22, 7468⫺7485.
[26] E. W. Yu, G. McDermott, H. I. Zgurskaya, H. Nikaido, D.
E. Koshland jr., Science 2003, 300, 976⫺980.
[27] G. Ecker, M. Huber, D. Schmid, P. Chiba, Mol. Pharmacol. 1999, 56, 791⫺796.
[28] C. Klein, D. Kaiser, S. Kopp, P. Chiba, G. F. Ecker, J.
Computer-Aided Mol. Des. 2003, 16, 785⫺793.
[29] G. A. Bakken, P. C. Jurs, J. Med. Chem. 2000, 43,
4534⫺4541.
[30] M. Katoh, M. Nakajima, H. Yamazaki, T. Yokoi, Pharm.
Res, 2000, 17, 1189⫺1197.
[31] T. Suzuki, N. Fukazawa, K. San-nohe, W. Sat, O. Yano,
T. Tsuruo, J. Med. Chem. 1997, 40, 2047⫺2052.
[32] D. F. Tang-Wai, S. Kajiji, F. di Capua, D. de Graaf, I. B.
Roninson, P. Gros, Biochemistry 1995, 34, 32⫺39.
[33] A. Ramu, N. Ramu, Cancer Chemother. Pharmacol.
1992, 30, 165⫺173.
[34] J. M. Zamora, H. L. Pearce, W. T. Beck, Mol. Pharmacol.
1988, 33, 454⫺462.
[35] W. N. Hait, J. F. Gesmonde, J. R. Murren, J. M. Yang,
H. X. Chen, M. Reiss, Biochem. Pharmacol. 1993, 45,
401⫺406.
 2004 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Документ
Категория
Без категории
Просмотров
1
Размер файла
177 Кб
Теги
base, features, mode, resistance, screening, silicon, modulators, identification, leads, multidrug, pharmacophores
1/--страниц
Пожаловаться на содержимое документа