CoMFA and CoMSIA investigations of dopamine D3 receptor ligands leading to the prediction, synthesis, and evaluation of rigidized FAUC 365 analogues

Article abstract of DOI:10.1016/j.bmc.2006.05.025

Taking advantage of our in-house experimental data on dopamine D₃ receptor modulators, we have successfully established highly significant CoMFA and CoMSIA models (q_cv² = 0.82 / 0.76). These models were carefully investiga

Full text of DOI:10.1016/j.bmc.2006.05.025

Bioorganic & Medicinal Chemistry 14 (2006) 5898–5912
CoMFA and CoMSIA investigations of dopamine D3 receptor
ligands leading to the prediction, synthesis, and evaluation
of rigidized FAUC 365 analogues
Ismail Salama, Karin Schlotter, Wolfgang Utz, Harald Hubner,
¨
Peter Gmeiner and Frank Boeckler*
Department of Medicinal Chemistry, Emil Fischer Center, Friedrich Alexander University,
Schuhstraße 19, D-91052 Erlangen, Germany
Received 23 March 2006; revised 5 May 2006; accepted 15 May 2006
Available online 5 June 2006
Abstract—Taking advantage of our in-house experimental data on dopamine D3 receptor modulators, we have successfully estab-
lished highly significant CoMFA and CoMSIA models (q2cv ¼ 0:82=0:76). These models were carefully investigated to assure their
stability and predictivity (r2pred ¼ 0:65=0:61) and subsequently applied to guide experimental investigations on the synthesis and
receptor binding of three conformationally restricted D3 ligands. Besides the high D3 affinity, the test compound 45, incorporating
a trans-1,4-cyclohexylene partial structure, exhibited improved (ꢀ3200-fold) selectivity over the D4 subtype.
Ó 2006 Elsevier Ltd. All rights reserved.
1. Introduction
related agonists have evolved into structurally diverse
agents with high affinity, selectivity over the closely
related biogenic amine receptors, and a broad range of
intrinsic activities. In recent years, 4-phenylpiperazine
has obtained a predominant position as a building block
in D3-selective ligands. Although this ‘privileged struc-
ture’ has already been a frequently used scaffold for
other biogenic amine receptors including the dopamine
D4 receptor subtype,9–12 tuning of the substitution pat-
tern of the phenyl moiety and, in particular, tuning of
the spacer length and type between a (hetero)aromatic
moiety and the piperazine has yielded series of ligands
with highly interesting D3 receptor binding and activa-
tion profiles.8,13 Among these, for example, BP 897
(27) has been intensely investigated for its putative use
in the treatment of cocaine addiction6,14 and L-DOPA
induced dyskinesia in patients with Parkinson’s disease.5
We have recently demonstrated that the pyrazolo[1,5-
a]pyridine analogue FAUC 329 (3) exerts neuroprotec-
tive effects in the MPTP (1-methyl-4-phenyl-1,2,3,6-tet-
rahydropyridine) mouse model of Parkinson’s
disease.15 Its closely related bioisostere FAUC 365 has
been described as the most selective D3 antagonist, yet.16
In more than 15 years of research since the discovery of
the D3 receptor by Sokoloff and coworkers,1 enormous
progress has been made toward improving our under-
standing of its physiological function and pharmacolog-
ical impact. As the D3 receptor is preferentially located
in brain regions which have an impact on emotional
and cognitive functions, it is able to affect behavioral
properties, such as locomotor activity, reinforcement,
and reward. Thus, various pharmacological studies have
investigated it as an interesting therapeutic target for the
treatment of schizophrenia,2,3 Parkinson’s disease,4
drug-induced dyskinesia,5 and drug abuse (in particular
cocaine addiction).3,6 Moreover, D3 might be involved
in the cortical development during gestation when it
obviously regulates neuronal migration and differentia-
tion.7 Parallel to gaining a more detailed insight into
D3 receptor pharmacology and into the evaluation of
respective treatment opportunities, the available D3
ligands have undergone a ‘structural evolution’.8 Mostly
driven by rational drug discovery, small, dopamine-
Taking advantage of our in-house data on D3 ligands,16–18
we herein present a ligand-based 3D-QSAR approach
(Chart 1) in order to improve and refine our understand-
ing of the molecular requirements for optimized D3
Keywords: 3D-QSAR; Dopamine D3 Receptor; Phenylpiperazine;
CoMFA/CoMSIA model.
*
Corresponding author. Tel.: +49 9131 8524115; fax: +49 9131
8522585; e-mail: boeckler@pharmazie.uni-erlangen.de
0968-0896/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2006.05.025

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5899
41 (hetero)arene Carboxamides (1-25, 29-44)
external reference
ligands:
N
O
H
N
N
π
4
7
+ haloperidol (26)
+ BP 897 (27)
+ aripiprazole (28)
3
N
N
OCH3
5
6
( )n
( )n
N
R1
R2
2
H
N
O
N
1-9 and 33-35
O
N
R3
CI
CoMFA
CoMSIA
models
N
R2
N
H
X
prediction
10-25 and 36-44
OH
R1
N
S
H
N
O
N
N
Cl
Cl
Cl
Cl
trans: 45
cis:
O
46
26 (Haloperidol)
FAUC 365 (11)
F
H
pKi = 8.08
N
H
N
S
H
N
N
O
47
OCH3
N
N
H
O
Chart 1.
27 (BP897)
pKi = 8.85
N
N
CI
receptor binding, as well as to establish a working model
that allows to facilitate further drug discovery efforts. In
extension to previous 3D-QSAR analyses on D4 antag-
onists19 and D3 agonists,20 we apply a well-established
protocol on our training and test set comprising of 32
and 12 compounds, respectively. In addition to several
techniques for monitoring the statistical quality and pre-
dictivity, we demonstrate the applicability of the result-
ing model in guiding synthetic efforts toward D3 ligands
with less flexible alkyl spacers. Thus, the study involved
the prediction, synthesis, and biological testing of novel
rigidized FAUC 365 analogues (45–47). Even though
these molecules contain structural elements not directly
represented in the training set, the good correspondence
between the obtained experimental pKi values and our
previous predictions is encouraging.
CI
CI
O
N
O
H
28 (Aripiprazole)
pKi = 8.35
N
O
N
R
N
H
R =
Cl
29 pKi = 9.00
H3CO
H3CO
30 pKi = 8.77
As an extension to recent 3D-QSAR studies on D3
ligands,21,22 a vital part of our modeling strategy was
to enrich our dataset by a number of external reference
ligands, such as haloperidol (26), BP 897 (27), and ari-
piprazole (28) thus enhancing the ligands’ structural
diversity. To ensure optimal comparability of the affini-
ties for all ligands, we have obtained all receptor binding
data used in this study within our laboratory.
31 pKi = 8.51
32 pKi = 9.55
Scheme 1.
oxamide moiety, we have selected compounds 1, 2, 11,
and 4, which are representatives for the ligands bearing
an alkyl spacer of 2, 3, 4, and 5 atoms’ length,
respectively.
2. Methods
2.1. Generation of structures and exploration of the
conformational space
First, we performed a grid search, using the Tripos force
field24 with Gasteiger–Marsili charges,25,26 on the phe-
nylpiperazine fragment, iterating the bond between
piperazine and phenyl in steps of 30°. Afterwards, we
accomplished full conformational analyses of the tem-
plates, while only the lowest energy conformer of phe-
nylpiperazine was retained from the previous
Structure building and refinement for the entire set of
phenylpiperazine analogues (1–44, Scheme 1, Tables 1
and 2) was accomplished using SYBYL 6.9 molecular
modeling software.23 As templates for generating the
whole series of phenylpiperazines with (hetero)arylcarb-

5900
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
Table 1. 2-Methoxyphenylpiperazines containing a pyrazolopyridine
moiety in the training set (1–9) or in the test set (33–35) and their
measured D3 affinities16
the (hetero)aromatic moiety and the phenylpiperazine
substituents. The reference structures haloperidol (26)
and aripiprazole (28) were constructed and their confor-
mational space evaluated as described for the template
structures. Finally, all ligands were optimized with
MOPAC using a semiempirical AM1 Hamiltonian27,28
to improve the molecular geometries and ensure better
comparability of the ligand structures by providing
structures based on identical levels of calculation. The
MMOK keyword was used in these calculations to en-
sure planar amide bonds.
Compound
Position
n
pKi
Training set
1
2
3
4
5
6
7
8
9
2
2
2
2
3
3
4
5
7
1
2
3
4
1
4
3
3
3
6.60
6.19
8.37
7.29
6.44
7.19
7.70
8.55
7.54
2.2. Alignment
Test set
33
34
In general, the alignment is one of the most challenging
aspects of 3D-QSAR. Thus, various approaches involv-
ing different degrees of complexity exist in order to ad-
dress this problem. The spectrum of available
techniques ranges from simple atom-based fitting proce-
dures to sophisticated binding-site guided protocols.
However, none of these has evolved to be superior over
the others. In two recent 3D-QSAR investigations,19,20
we have found the ligand-based alignment technique
ASP29 to be very useful. Thus, we used the module
ASP as implemented in the QSAR package TSAR,30
which allows us to perform an alignment by comparison
of steric overlap and molecular electrostatic potentials.
For deriving reasonable electrostatic potentials, first,
VESPA charges31 were calculated using the semiempiri-
cal program package VAMP.32 These atomcentered par-
tial charges are obtained by a fit of the electronic wave
function to the atomic positions. Compared to other
charge schemes such as Coulson or Mulliken, they have
the advantage that the anisotropy of the electron distri-
bution around the molecule, especially for aromatic sys-
tems, is described in more detail.
3
3
6
3
2
3
7.74
6.82
8.37
35
Table 2. Substituted phenylpiperazines containing a heteroaromatic
moiety in the training set (10–25) or in the test set (36–44) and their
measured D3 receptor affinities
Compound
X
R1
R2
R3
pKi
Training set
10a,b
11a,c
12
S
H
OCH3
Cl
H
Cl
H
H
H
H
Cl
H
Cl
H
Cl
H
H
H
Cl
Cl
9.63
9.30
9.41
9.34
9.42
8.95
8.92
9.16
8.46
8.37
9.18
9.24
9.61
9.60
9.13
9.45
S
H
S
5-CCH
5-CN
6-CCH
H
OCH3
OCH3
OCH3
OCH3
Cl
13d
S
14
15a
S
O
O
O
O
O
N
N
N
N
N
N
16a
H
17d
5-Br
5-Br
5-CN
H
OCH3
Cl
18d
19d
OCH3
Cl
20d
21d
5-CN
5-Br
6-CN
5-Br
6-CN
OCH3
OCH3
OCH3
Cl
22d
23d
As the basic template onto which the other ligands are
to be superimposed, we have chosen FAUC 365, which
shows considerably high dopamine D3 receptor affinity.
To quantify the relative orientation of two molecules,
the combined similarity index based on the Carbo index
for electrostatics and the shape similarity index to ac-
count for steric differences was evaluated (with both
indices weighted equally) using three Gaussian functions
for integration. This parameter was then optimized by
overlaying the centroids of the molecules and perform-
ing a full translational and orientational search of
each rigid comparison molecule relative to the lead
compound FAUC 365 (11) by systematically rotating
around the Cartesian x-, y-, and z-axes in 10° steps.
For each new orientation, a Simplex algorithm in com-
bination with Simulated Annealing directs the six de-
grees of freedom to an alignment with optimal
similarity.29 Finally, the orientation and placement of
each ligand on the template 11 featuring the highest
score related to this search algorithm was chosen to
yield the TSAR-based alignment.
24d
25d
Cl
Test set
36d
37d
38
S
S
S
S
S
S
S
S
S
5-Br
5-Br
OCH3
Cl
H
9.59
8.49
8.51
7.85
9.27
9.60
9.24
9.48
9.59
Cl
H
5-CF3
5-CF3
5-CCH
5-CN
6-CCH
6-CN
6-CN
OCH3
Cl
Cl
39
Cl
Cl
Cl
Cl
H
40
41d
42
Cl
Cl
43d
44d
OCH3
Cl
Cl
a Ref. 16.
b FAUC 346.
c FAUC 365.
d Ref. 18.
calculation and kept fixed as an aggregate this time. The
conformations for each template obtained from these
analyses were separated into 8 to 17 conformational
families employing a hierarchical clustering algorithm.
The most reasonable low energy conformer was chosen
from each of the obtained clusters. Using these four
template structures, all other ligands of the
corresponding series were derived by modification of
2.3. CoMFA and CoMSIA
To ensure best comparability between molecules, all
pharmacological data of the ligands used in this study

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5901
have been measured within our laboratory. We per-
formed Comparative Molecular Field Analysis
ponents. After obtaining the models, the CoMFA and
CoMSIA results were graphically interpreted by field
contribution maps.
a
(CoMFA) evaluating the typically used steric and elec-
trostatic fields implemented in SYBYL. All CoMFA
calculations were accomplished using an sp3 carbon
atom with a charge of +1, a cutoff value of 30 kcal/
mol for the Lennard-Jones and Coulomb-type potential,
and a constant dielectric function. The dimension of the
2.5. Calculation of the predictive correlation coefficient
(r2pred) and prediction of novel compounds
The predictive ability of the 3D-QSAR model was deter-
mined from a set of 12 compounds (33–44) that were not
included in the training set (Table 3). The compounds
were manually assigned to the training or test set ensur-
ing a reasonable range of pKi units and structural diver-
sity in both sets. These molecules were aligned using the
same method as described before, and their activities
were predicted using the model generated by the training
set. Accompanying a synthetic strategy to design some
rigidized analogues of the highly flexible ligands used
in the training and test set, we also predicted three li-
gands (45–47) with cycloalkyl spacers, which were sub-
sequently synthesized and tested. In addition to the
classical test set, we also included these rigidized com-
pounds in the calculation of r2pred. Based on the test set
˚
surrounding lattice (1.0 A grid spacing) was selected
with a sufficiently large margin to enclose all aligned
˚
molecules by at least 4 A.
Putative problems of the analysis can arise from the
absolute orientation of the molecules within the grid
space.33 A useful protocol to address this problem is
the AOS/APS script of Wang et al.,34 which automati-
cally rotates/translates the entire dataset within the
lattice without changing the relative orientation or
alignment of the molecules. We applied the APS proto-
˚
col varying the ligands by steps of 0.1 A in all three
˚
dimensions within a 1.0 A grid. After each of 1000 trans-
lational steps, the PLS analysis was repeated. This pro-
cedure gives detailed information about the
translational dependence of the CoMFA, helps to en-
sure that no artificial effects are included, and provides
evidence of the robustness of the model.
molecules, this predictive correlation coefficient (rp2red
is defined as
Þ
r2pred ¼ ðSD ꢁ PRESSÞ=SD;
where SD is the sum of the squared deviations of each
biological property value from their mean and PRESS
is the sum of squared differences between the predicted
and actual affinity values for every molecule in test set.36
In the CoMSIA, all five physicochemical descriptors
(electrostatic, steric, hydrophobic, and hydrogen-bond
donor and acceptor) were evaluated using a common
probe atom placed within a 3D grid. The atom was set
up with a radius of 1.0 A and charge, hydrophobic inter-
action, and hydrogen-bond donor and acceptor proper-
ties all equal to +1. Like in the CoMFA, the grid was
˚
extended beyond the molecular dimensions by 4.0 A in
the x, y, and z directions and the spacing between probe
˚
points within the grid was set at 1.0 A. For the attenua-
˚
3. Results and discussion
3.1. CoMFA
tion factor a controlling the steepness of the Gaussian
function the standard value of 0.3 was accepted.
Using our properly selected training set of 32 D3
ligands (Scheme 1 and Table 2), involving literature
compounds with extended structural diversity such as
haloperidol (26) and aripiprazole (28), we obtained sta-
tistically significant QSAR models (Table 4). The initial
PLS analysis of our aligned training set applying a
default rmin data filter of 2 kcal/mol yielded a cross-
validated q2 of 0.816 with sc/v = 0.432 using two com-
ponents (B). Increasing the minimum level of field
variation rmin to purpose a more efficient reduction
of noise, only a negligible further improvement of the
statistics of the CoMFA model is found for rmin = 3.0
and 4.0 kcal/mol (C, D). Likewise, reduction of rmin to
1.0 kcal/mol shows also only negligible effects on the
q2cv (A). In contrast, increasing rmin further to
5.0 kcal/mol results even in a reduction of the cross-
validated q2 to 0.794 (E), which presumably reflects
that some statistically relevant descriptors have been
filtered out at this rmin value. With increasing rmin
the number of remaining descriptors is noticeably de-
creased from 3603 to 1266, while at the same time
the electrostatic contribution is raised from 27.7% to
36.1%. Thus, these results demonstrate that the ob-
tained CoMFA is stable at a highly significant level
(>0.8) and not prone to deviate noticeably with varying
numbers of descriptors.
2.4. Partial least squares (PLS) analysis
The PLS method35,36 was used to linearly correlate the
CoMFA and CoMSIA fields to biological activity val-
ues. The cross-validated analysis was performed using
the leave-one-out (LOO) method in which one com-
pound is removed and its activity is predicted using
the model derived from the rest of the dataset. Comple-
menting the results obtained form the leave-one-out
cross-validation, the more robust leave-many-out proce-
dure was performed to ensure the reproducibility of q2.37
The cross-validated q2 that resulted in minimal number
of components and lowest standard error of prediction
was accepted.38,39 To speed up the analysis and reduce
noise, column filtering values (rmin) were iterated be-
tween 1.0 and 5.0 kcal/mol. Thus, only columns with a
standard deviation of more than rmin were used for
the cross-validation, resulting in approximately 5.5–
15.6% of the original data to be used in the CoMFA
and 4.1–12.7% used in the CoMSIA.
A final non-cross-validated analysis was performed
using the optimal number of previously identified com-

5902
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
used during the analysis.40 They stated that in the case
Table 3. Experimental and predicted biological affinities and residuals
obtained by the CoMFA and CoMSIA model for 32 compounds in the
training set and 15 compounds in the test seta
˚
of a 2.0 A lattice, important contributions to the corre-
lation analysis could be lost due to the required arbi-
trary cutoff values. Despite the fact that changing from
˚
a 2.0 to 1.0 A lattice spacing results in an increase in
Compound pKi (exp.)
pKi (pred.)
DpKi
CoMFA CoMSIA CoMFA CoMSIA
computing time by a factor of 8, we performed all anal-
yses using the smaller increment. Quantifying the trans-
lational dependence of q2cv helps to ensure that no
artificial effects are included in the final CoMFA and
gives evidence of the robustness of the model. Thus,
we applied a procedure published by Wang et al.,34
which systematically translates the aligned dataset in
space, followed by a PLS analysis after every retransla-
tion. The histogram plot of the resulting 1000 model
variants showed to be approximately corresponding to
a normal distribution of the obtained q2 values
(Fig. 1). The highest qc2v (0.839) was found for three
components. Thus, it should be noted in this context
that the increase of q2cv values by less than 5% with the
use of an additional component is regularly considered
inappropriate due to the ‘parsimony-principle’.41
However, what is even more interesting is the very
narrow range of q2cv values between 0.807 and 0.839 with
50% of the values lying between 0.817 and 0.825
indicating that our CoMFA model is considerably
stable. Modification of other CoMFA parameters, such
as changing the cutoff values for steric/electrostatic
energies, did not give any improvement of the model.
The final predicted/cross-validated versus experimental
pKi values for model B and their residuals (DpKi) are
given in Table 3.
Training set
1
2
6.60
6.19
8.37
7.29
6.44
7.19
7.70
8.55
7.54
9.63
9.30
9.41
9.34
9.42
8.95
8.92
9.16
8.46
8.37
9.18
9.24
9.61
9.60
9.13
9.45
8.08
8.85
8.35
9.00
8.77
8.51
9.55
6.450
7.208
8.689
7.621
7.120
6.846
8.110
8.586
8.364
9.017
9.108
9.336
9.250
9.431
8.757
8.594
8.728
9.136
9.069
9.098
9.430
9.243
9.371
9.418
9.309
7.690
8.918
8.433
8.713
8.357
8.839
9.573
6.252
7.181
8.690
7.152
7.321
7.615
8.330
8.304
8.384
8.974
8.935
9.291
9.119
9.430
8.809
8.629
9.186
9.178
9.222
9.118
9.369
9.246
9.324
9.709
9.046
7.590
9.123
7.498
8.948
8.108
8.791
9.467
0.150
0.348
ꢁ1.018 ꢁ0.991
ꢁ0.319 ꢁ0.320
3
4
ꢁ0.331
0.138
5
ꢁ0.680 ꢁ0.881
0.344 ꢁ0.425
ꢁ0.410 ꢁ0.630
6
7
8
ꢁ0.036
0.246
9
ꢁ0.824 ꢁ0.844
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
0.613
0.192
0.074
0.090
0.656
0.365
0.119
0.221
ꢁ0.011 ꢁ0.010
0.193
0.326
0.141
0.291
0.432 ꢁ0.026
ꢁ0.676 ꢁ0.718
ꢁ0.699 ꢁ0.852
0.082
0.062
ꢁ0.190 ꢁ0.129
0.367
0.229
0.346
0.279
ꢁ0.288 ꢁ0.579
0.141
0.390
0.404
0.490
ꢁ0.068 ꢁ0.273
ꢁ0.083
0.287
0.413
0.852
0.052
0.662
3.2. CoMSIA
ꢁ0.329 ꢁ0.281
ꢁ0.023 0.083
In analogy to the CoMFA, already the initial PLS
analysis of our aligned training set applying a default
rmin data filter of 2.0 kcal/mol yielded a highly signif-
icant cross-validated q2 of 0.741 with sc/v = 0.531 using
four components (Table 4, G). Variation of the col-
umn filtering parameter rmin shows that for increasing
rmin values from 1.0 to 5.0 kcal/mol, the number of
descriptors is strongly reduced from 7309 to 2344,
while the q2cv is steadily increased from 0.738 (F) to
0.759 (J). Although the q2 values of the CoMSIA
models are slightly reduced compared to those of
the CoMFA models, they are still indicating stable
analyses of high quality. Furthermore, the CoMSIA
models comprise valuable complementary information,
as they offer additional explanation for the ligands’
different receptor affinities by introducing three auxil-
iary field types, the hydrophobic field, the hydrogen
bond acceptor field, and the hydrogen bond donor
field. Interestingly, while the noise reduction in the
CoMFA model decreased the fraction of the steric
contribution from 72.3% down to 63.9%, it was in-
creased by the noise reduction in the CoMSIA model
from 10.9% up to 15.8%. The other CoMSIA field
types typically contribute to the full model in the
order: electrostatic (25.8–26.7%) > donor (22.2–
23.5%) > acceptor (18.0–20.6%) > hydrophobic (17.3–
19.3%). The final predicted/cross-validated versus
experimental pKi values for model G and their residu-
als (DpKi) are given in Table 3.
Test set
33
34
7.74
6.82
8.37
9.59
8.49
8.51
7.85
9.27
9.60
9.24
9.48
9.59
8.60
6.92
6.85
8.270
7.066
8.880
9.146
9.328
9.359
9.416
9.368
9.440
9.393
9.151
9.388
8.110
6.994
7.036
0.651
8.581
7.250
8.538
9.282
9.080
9.451
9.482
9.298
9.158
9.343
9.206
9.136
8.296
6.802
7.329
0.613
ꢁ0.530 ꢁ0.841
ꢁ0.246 ꢁ0.430
ꢁ0.510 ꢁ0.168
35
36
0.444
0.308
37
38
ꢁ0.838 ꢁ0.590
ꢁ0.849 ꢁ0.941
ꢁ1.566 ꢁ1.632
ꢁ0.098 ꢁ0.028
39
40
41
42
0.160
0.442
ꢁ0.153 ꢁ0.103
43
44
0.329
0.202
0.490
0.274
0.454
0.304
0.118
45
46
ꢁ0.074
47
r2pred
ꢁ0.186 ꢁ0.479
b
a The experimental binding affinities toward the dopamine D3 receptor
are expressed as pKi (ꢁlogKi) values. DpKi is the error of cross-
validated (training set) or predicted (test set) binding affinities and is
defined as (pKi,experimental ꢁ pK i,cross-validated/predicted).
b Predicted r2 calculated with a standard deviation (SD) obtained from
the test set only: 0.650 (CoMFA), 0.611 (CoMSIA).
Bo¨hm et al., have attributed the dependence of q2cv on
the grid spacing and the absolute position of the aligned
molecules within the lattice to the shape and steepness of
the hyperbolic Lennard-Jones and Coulomb potentials

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5903
Table 4. Summary of the results from several PLS runs after applying different levels of noise reduction by column filtering (rmin
)
PLS run
CoMFA
C
CoMSIA
H
A
B
D
E
F
G
I
J
rmin
1
2
3
0.818
0.430
0.937
0.252
215.6
2
4
5
1
2
3
0.744
0.528
0.956
0.219
146.4
4
4
5
q2cv
sPRESS
0.815
0.434
0.935
0.257
0.816
0.432
0.936
0.255
0.820
0.427
0.937
0.253
0.794
0.458
0.922
0.282
0.738
0.534
0.954
0.223
0.741
0.531
0.955
0.222
0.751
0.521
0.957
0.217
0.759
0.513
0.957
0.216
r2
S
F
208.0
211.9
215.1
170.2
141.1
142.3
150.0
151.4
Components
Descriptors
2
3603
2
2305
2
1433
2
1266
4
7309
4
5138
4
3044
4
2344
1730
3891
Fraction
Steric
0.723
0.277
0.704
0.296
0.671
0.329
0.645
0.355
0.639
0.361
0.109
0.109
0.118
0.259
0.189
0.232
0.201
0.138
0.158
Electrostatic
Hydrophob.
Donor
0.258
0.193
0.235
0.206
0.258
0.193
0.235
0.206
0.267
0.179
0.222
0.194
0.267
0.173
0.222
0.180
Acceptor
0.799 0.028 and 0.726 0.038 for CoMFA and
CoMSIA, respectively, show only a small decrease of
the mean values, although disregarding every fifth
compound. Likewise, there is only a marginal increase
in the standard deviation. Therefore, we conclude that
leave-10%-out and leave-20%-out cross-validation
clearly corroborates the good stability and robustness
of the models.
The predictive power of the CoMFA (PLS mode B) and
CoMSIA (PLS mode G) analysis was further examined
using a test set of 12 compounds (33–44, Scheme 1) that
had been omitted from the training set. In addition to
these test set ligands, we also included three rigidized
compounds (45–47) in the calculation of the r2pred value,
which were synthesized accompanied by these QSAR
predictions as described subsequently. The calculation
was performed according to the formula of Cramer
et al.,36 and gave better results for the CoMFA with
r2pred = 0.651 than for the CoMSIA model with
r2pred = 0.613. The corresponding plots of the cross-vali-
dated or predicted versus the experimental binding affin-
ities are shown for the training set (unfilled circles), the
test set (filled circles), and the three newly synthesized
compounds (filled triangles) in Figure 2.
Figure 1. Histogram showing the distribution of q2 values calculated
by SAMPLS33 leave-one-out cross-validation after systematic transla-
tion of aligned molecules within the lattice by an all placement search
(APS).
3.3. Advanced cross-validation and assessment of model
predictivity
Despite the good results obtained for CoMFA and
CoMSIA cross-validation, of course, we are aware that
there is always the danger to overemphasize the useful-
ness of cross-validation and q2 as measures of the pre-
dictive performance. High values of q2LOO can be
regarded as a necessary, but not a sufficient, condition
for a model to possess significant predictive power.42
Thus, we also performed the more critical and widely
accepted leave-10%-out and leave-20%-out cross-valida-
tions. With q2L10%O-values (mean SD) of 0.812 0.012
and 0.741 0.018 for the CoMFA and CoMSIA,
respectively, we have shown that even after discarding
every tenth compound from the training set the predic-
tivity of the models is hardly impaired. Repeating this
calculation n = 20 times for each model assesses the lia-
bility of the respective model to random effects caused
by different assignment of the ligands to the 10 groups.
The negligible standard deviation (SD) is a further indi-
cator of the model quality. Also the q2L20%O-values of
3.4. Graphical interpretation of the fields
The three-dimensional representations of the CoMFA
and CoMSIA field contributions as ‘stdev*coeff’ con-
tour plots reveal where variability in the fields of the
molecules is able to explain experimental binding differ-
ences. Thus, they can be useful to identify important fea-
tures, which contribute to interactions between ligand
and receptor in the active site.
3.4.1. CoMFA. Favored and disfavored cutoff energies
were set at the 90th and 10th percentiles for the steric
contributions and at the 85th and 15th percentiles for
the electrostatic contributions, respectively. Because it
is a representative example of the most active ligands,
compound 22 is shown embedded as a guide into the ste-
ric (A) and electrostatic fields (C) (Fig. 3) resulting from
a non-cross-validated CoMFA run (rmin = 2.0 kcal/

5904
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
mol). Likewise, compound 6 is shown enclosed in the
steric (B) and electrostatic fields (D) as a representative
of lower affinity ligands. In panel (A), the green-colored
regions specify areas where steric bulk enhances D3
affinity, while the yellow contours mark regions where
steric bulk leads to impaired biological activity. The pre-
dominant field color for the steric interactions in panel
(A) is obviously yellow. This indicates that in several
directions the tolerance for the displacement of structur-
ally divergent ligands is rather limited, in order to avoid
the occupation of any of those sterically forbidden yel-
low regions. The large yellow isopleths 1 and 2 below
the phenylpiperazine reflect a strict decrease in affinity
for all of these ‘anchor moieties’ being dislocated into
this area. The other two larger yellow isopleth areas (3
and 4), which flank the sides of the indolecarboxamide,
reveal that narrow (hetero)arylcarboxamides with a
straight-lined orientation are preferentially recognized
to yield high pKi values, while any bent or perpendicular
orientation, as well as sterically demanding systems, is
detrimental for affinity (exemplified by 6 in (B)). In con-
trast, the green areas in position 5 and 6 of the indole of
22 indicate that substituents in these positions favor
binding affinity to D3 receptors. This, for instance, ap-
plies to the indole analogues 21–25, which all bear a cya-
no or bromo substituent in position 5 or 6 and bind with
subnanomolar affinity to the D3 receptor.
In panel (B), the blue-colored regions show areas where
electropositive charged groups enhance D3 affinity,
while in red regions groups with electronegative charges
improve the receptor binding. The figure shows a blue
area (1) in the vicinity of the protonated piperazine
nitrogen. In consistency with the steric field, this blue
isopleth suggests that the phenylpiperazine ‘anchor moi-
ety’ should be fixed. Even small displacements, for
example, in ligands with shorter spacers, are not well tol-
erated and lead to a substantial decrease of the pKi. The
Figure 2. Fitted predictions versus experimental binding affinities for
the 47 compounds in the training set (1–32: open circles) and the test
set (33–44: filled circles/45–47: filled triangles). In addition to the line of
ideal correlation, dotted lines are given, which indicate deviations from
the actual pKi by 1 logarithmic unit. Outliers are labeled by their
compound numbers.
Figure 3. Stdev*coeff contour plots illustrating steric (A and B) and electrostatic features (C and D) as obtained by the final CoMFA. In (A) and (B),
regions where steric bulk will enhance affinity are shown enclosed by green contours (contribution level51: 90%), whereas regions which should be
kept unoccupied to prevent decrease of affinity are contoured in yellow (10%). This is exemplified by the high affinity ligand 22 (pKi = 9.61) in (A) and
the lower affinity ligand 6 (pKi = 7.19) in (B)s. In (C) and (D), red contours (contribution level51: 15%) encompass regions where electron-rich
fragments with negative partial charges will improve affinity. Blue contours (85%) indicate regions where reduced electron density (positive partial
charges) is predicted to increase affinity. Again, compounds 22 (C) and 6 (D) are used to exemplify the plots for a high and lower affinity ligand,
respectively.

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5905
blue contour 2 and the red contour 3 correspond exactly
to the carboxamide function. Thus, different orientation
or omission of this functional moiety exerts a negative
impact on the binding properties as shown for 6 in panel
(D). The electrostatic contour map shows a region of red
polyhedrals (4) at position 6 and in particular at position
5 of the indole ring, indicating that electron-rich substit-
uents are beneficial for the binding affinity. This field
can be exemplified by the fact that the presence of the
5-cyano group in compounds 19 and 21, the 6-cyano
group in compounds 23 and 25, and the 5-bromo group
in compounds 17, 18, 22, and 24, gives rise to higher
binding affinity than compounds (1–9).
by ligands of low and high affinity. It becomes obvious
from a direct comparison of Figures 3 and 4 that steric
and electrostatic properties in CoMFA and CoMSIA
show a high degree of similarity, however, a certain de-
gree of complementary information can be found.
3.4.2.1. Steric contributions. In panel (A), green and
yellow isopleths are drawn at a contribution level of
90% and 10%, and enclose regions favorable or unfavor-
able for bulky groups, respectively.
As in the CoMFA, the only green isocontour (1) found
in Figure 4A and B is located most distant to the phenyl-
piperazine on the far left side. It can be occupied, for
example, by the terminal ring of the biphenyl in 32 (B)
or by the six-membered ring of the heteroaromatic
3.4.2. CoMSIA. In Figure 4, as well as in the following
discussion, the CoMSIA contour plots are exemplified
Figure 4. Stdev*coeff contour plots illustrating steric (A + B), electrostatic (C + D), hydrophobic (E + F), and hydrogen bond donor (G + H) and
acceptor (I + J) properties revealed by the final CoMSIA. For all features, one ligand with low (A: 6, C: 5, E: 1, G: 5, I: 7) and another with high
affinity (B: 32, D: 22, F: 17, H: 22, J: 13) are shown in comparison. The mesh fields represent the stdev*coeff plots, whereas the transparent surfaces
indicate the fields of the particular ligand, thus, facilitating the recognition of matching or mismatching features.

5906
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
system of those active ligands, which contain a butylene
spacer and the carboxamide function in position 2 of the
benzothiophene, benzofuran or indole nucleus. In con-
trast, for ligands containing a spacer of different length
or the carboxamide function in a different position of
the heterocycle, such as 6, the green isopleth is not even
touched (A). However, 6 as an example of a bend ligand
penetrates the forbidden regions 2 and 3 much more
than straight ligands, such as 32, do. As in the CoMFA,
the isopleths 4 and 5 below the phenylpiperazine partial
structure penalize any displacement of this ‘anchor re-
gion’. Isopleth 5 can also be explained by a number of
low affinity ligands (pyrazolopyridine derivatives) bear-
ing an ortho-methoxy substituent in this region.
(transparent orange) induced by the methoxy group of
several potent ligands (10, 12, 13, 14, 15, 17, 19, 21,
22, and 23) overlaps the orange isopleth 6.
3.4.2.4. Hydrogen bond donor contributions. Cyan
contour maps (contribution level 75%) are representing
the position of H-bond donor groups which favor bio-
logical activity, while purple areas (10%) are outlining
the location of biologically unfavored donors.
The cyan isocontours 1, 2, and 3 result from the indole-
NH (ligands 20–25), from the carboxamide moiety, and
from the protonated piperazine nitrogen N1. These
maps can be occupied only by ligands showing good
D3 affinity (butylene-spacer required), such as 22 (panel
(H)), while less active compounds (in particular with
ethyl or propyl spacer) point toward the unfavored pur-
ple areas 4, 5, and 6, as exemplified by 5 in panel (G). It
should be noted that map 3 might be equivalent to
Asp3.32, which is responsible for forming a ‘reinforced
ionic bond’ (consisting of a H-bond-mediated and an
ionic electrostatic interaction) to the protonated pipera-
zine within in the D3-pocket.
3.4.2.2. Electrostatic contributions. Red and blue isop-
leths (contribution level 90%/20%) enclose regions
favorable for negative and positive charge, respectively.
A large, red isopleth (1) is located at the p-system of the
left-sided heterocyclic or isocyclic ringsystems of ligands
showing high D3 affinity (Fig. 4D). While, for example,
the indole nucleus of 22 (panel (D)) is able to place its
negative electrostatic fields (transparent red) perfectly
within this mesh, the pyrazolo[1,5-a]pyridine nucleus
of 5 (panel (C)) fails to do so. Above and below the
extended red mesh, the blue areas 2 and 3 indicate where
electron-deficient substructures should be placed. These
two areas reflect a common placement of the carboxam-
ide hydrogens (2) of the highly potent ligands and the
indole hydrogen (3), respectively. The blue isopleth 4
exhibits the influence of the protonated piperazine nitro-
gen (N1) for the active ligands, while the red one (5) aris-
es from the electron rich, tertiary piperazine nitrogen
(N2) of those compounds. Consequently, these two isop-
leths determine the preferred position of the piperazine
ring, which is of course affected by the number of car-
bons in the alkyl chain. Thus, Figure 4 shows that the
piperazine moiety of 5 is misplaced (panel (C)), because
of the short two-carbon atom spacer (the blue isopleth 4
is completely buried in the transparent red electrostatic
potential at the N2 of the piperazine).
3.4.2.5. Hydrogen bond acceptor contributions.
Magenta isopleths (90%) encompass regions where a
hydrogen bond acceptor will lead to improved biological
activity, while an acceptor located near the red regions
(10%) will result in impaired biological activity.
The 5- or 6-cyano group (H-bond acceptor) within the
active ligands 13 (panel (J)), 23, and 25 exhibits a perfect
fit to the magenta isopleth 1. Compound 19 also bears a
5-cyano substituent. However, as its heteroaromatic
moiety is flipped horizontally within the alignment, the
cyano function points toward the red map 2, which
offers an explanation for the worse biological data of
19. The red mesh 3 reveals the low activity of the 2-
substituted pyrazolopyridines, because their acceptor
nitrogen is typically located in close proximity. The re-
duced affinity of 7 (panel (I)) can be explained in part
by its molecular acceptor field (transparent magenta),
which corresponds to the pyrazolopyridine nitrogen,
being completely embedded in the unfavorable red mesh
4. The carboxamide oxygens of various, highly potent li-
gands align to a common position close to the magenta
isopleth 5, while, for instance, the low affinity compound
2 is not able to place its carbonyl oxygen in this region.
3.4.2.3. Hydrophobic contributions. Yellow and
orange isopleths (contribution level 85%/15%) enclose
regions favorable for hydrophobic and hydrophilic
groups, respectively. The yellow mesh 1 encloses the 4,
5, and 6 positions of the heterocyclic rings, which can
be ascribed, for example, to the 5-bromo-substitution
of several potent compounds (17, 18, 22, and 24). As
exemplified in panel (F) for 17, a bromo substituent
enhances the size of the molecular hydrophobic field
(transparent yellow) significantly compared to an
unsubstituted heteroaromatic moiety as shown in 1
(panel (E)). The heteroatoms within the bicyclic moieties
of the active ligands and the carboxamidic hydrogens
are responsible for the orange isopleths 2 and 3, respec-
tively. The N1 piperazine nitrogen of the low affinity
compounds (such as 1) is lying adjacent to the yellow
isosurface 4, while that of the active structures corre-
sponds to the orange one (5). In addition, exactly this
area 5 is not enclosed in the large, hydrophilic molecular
field (transparent orange) of ligand 1, as a representative
of weak D3 receptor binding. The hydrophilic field
3.5. Prediction of novel, rigidized compounds
After establishing and evaluating our 3D-QSAR mod-
els, we were intrigued by the question, whether we can
employ them also for the prediction of novel compounds
containing structural elements hitherto unrepresented
by any other ligand in the training or test set.
Rigidization is a classic and popular strategic concept in
modern medicinal chemistry. Its objective is to synthet-
ically fix a flexible ligand scaffold in a biologically rele-
vant conformation with both implications on the
enthalpy and entropy of binding to a biological target.
Thus, we envisioned to exchange the flexible butylene

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5907
spacer of the highly potent and selective D3 ligand
OR
H
FAUC 365 (11) by a conformationally restricted bioiso-
stere. The first strategy was to replace the butylene by a
1,4-cyclohexylene spacer, which could provide a trans or
a cis configuration between the benzothiophenyl carbox-
amide and the 2,3-dichlorophenylpiperazine. The
second strategy involved replacement of the butylpiper-
azine by an ethyl-octahydro-2H-pyrido[1,2-a]pyrazine.
When the test compounds (45–47) were predicted
employing the CoMFA and CoMSIA models, the
trans-cyclohexyl derivative 45 was suggested to bind to
the D3 receptor with nanomolar affinity (pKi = 8.11/
8.30 predicted from CoMFA/CoMSIA). The cis-
cyclohexyl derivative 46 and the ethyl-octahydro-
2H-pyrido[1,2-a]pyrazine derivative 47, in contrast,
were supposed to give lower D3 affinities
(pKCi oMFA=CoMSIA ¼ 7:00=6:80 predicted for 46 and 7.04/
7.33 for 47, respectively).
O
a
N
N
N
N
H
Cl
Cl
50
51: R = H
52: R = OMs
b
H
N
N
H
c
d
N
N
N
N
Cl
53
Cl
Cl
Cl
Subsequently, we have proceeded with the synthesis of
all three novel ligands in order to verify or falsify our
predictions and, thus, to further evaluate, whether we
can challenge our models by predicting structurally
diverging ligands.
54
H
N
e
O
H
N
S
N
Cl
3.6. Chemistry
Cl
H
47
For the synthesis of the diastereomeric target com-
pounds 45 and 46 (Scheme 2), we started from commer-
cially available benzo[b]thiophene-2-carboxylic acid and
trans-4-aminocyclohexanol employing activation with
TBTU to yield the carboxamide 48. Subsequent oxida-
tion with the mild hypervalent iodine reagent IBX
resulted in formation of the corresponding ketone 49.
Finally, reductive amination with 1-(2,3-dichlorophe-
nyl)piperazine yielded in a mixture of regioisomers 45
and 46. Separation of the isomers was accomplished
by flash chromatography to give the trans- and cis-iso-
mers 45 and 46, respectively.
Scheme 3. Reagents and conditions: (a) 1-bromo-2,3-dichlorobenzene,
Pd2(dba)3, BINAP, NaOHt-Bu, toluene, 80 °C, 2.5 h (51%); (b) MsCl,
Et3N, THF, ꢁ30 °C to rt, 0.5 h (71%); (c) Bu4NCN, DMSO, 80 °C, 8 h
(77%); (d) LiAlH4, Et2O, 0 °C to rt, 1 h (66%); (e) benzo[b]thiophene-2-
carboxylic acid, TBTU, DIPEA, DMF, CH2Cl2, 0 °C to rt, 1 h (71%).
when we took advantage of a Buchwald–Hartwig
cross coupling44,45 as a key reaction step. Thus,
palladium promoted amination of 1-bromo-2,3-dichlo-
robenzene with the bicyclic skeleton 50 gave the
N-aryl substituted pyridopyrazine 51. Subsequent
treatment with methanesulfonyl chloride yielded 52.
Nucleophilic substitution with tetrabutylammonium
cyanide resulted in formation of the nitrile 53, which
could be reduced with LiAlH4 to furnish the primary
amine 54. Finally, coupling with benzo[b]thiophene-2-
carboxylic acid was induced by TBTU to provide
the final product 47.
Following a five-step protocol43 we prepared the bicy-
clic system 50 as a chiral building block (Scheme 3). A
synthetic route to the intermediate 51 was elaborated
H
OH
H2N
H
N
O
O
X
a
3.7. Biological testing
S
O
S
Receptor binding experiments were established to evalu-
ate the binding properties of the target compounds 45–
47 in comparison to the reference agents BP 897 and
FAUC 365 (Table 5). The obtained experimental results
showed to be in very good consistence with our previous
CoMFA/CoMSIA predictions (Table 3). All deviations
were found to be below 0.5 pKi units. In fact, the
trans-cyclohexyl derivative 45 displayed considerably
higher D3 affinity (Ki = 2.5 nM) than its cis analogue
46 or the bicyclic derivative 47. While being only moder-
ately selective for D3 over 5-HT1A or D2 receptors, selec-
tivity over the antitarget a1 was substantial. Besides a
large selectivity over D1 and 5-HT2, 45 showed a pro-
48 X = CHOH
49 X = C=O
b
H
N
N
N
45
46
c
S
Cl
Cl
O
45 (trans), 46 (cis)
Scheme 2. Reagents and conditions: (a) TBTU, DIPEA, DMF,
CH2Cl2, 0 °C to rt, 1 h (64%); (b) IBX, DMSO, rt, 1 h (56%); (c)
Na(OAc)3BH, 2,3-dichlorophenylpiperazine, 1,2-dichloroethane, rt,
3.5 h (84%); flash chromatography, CH2Cl2–MeOH, 98:2, (45: 22%;
46: 14%).

5908
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
Table 5. Receptor binding data of 45–47 compared to the reference compounds BP 897 (27) and FAUC 365 (11) utilizing human D2long, D2short, D3,
and D4.4 receptors as well as porcine D1, 5-HT1A, 5-HT2, and a1 receptorsa
H
N
N
S
H
N
H
N
N
S
H
N
Cl
Cl
Cl
Cl
trans: 45
O
47
O
cis:
46
Compound
Ki values in (nM)
[3H]SCH 23390
D1
[3H]spiperone
[3H]8-OH-DPAT
[3H]ketan serin
5-HT2
[3H]prazosin
D2long
D2short
D3
D4.4
5-HT1A
a1
45
46
4300
2200
990
76
2900
490
78
2600
1100
2600
200
2.5
8000
460
180
340
44
47
810
230
360
81
8900
18000
1200
3000
840
240
85
120
140
47
FAUC 365 (11)
BP 897 (27)
49
370
5.0
8800
760
3600
220
0.50
1.3
a Ki values in nM are based on the means of 2–5 experiments each done in triplicate.
nounced, about 3200-fold, selectivity over D4 receptors,
which was superior to both reference ligands, BP 897
5. Experimental
5.1. Methods and materials
and FAUC 365. In contrast to the interesting pharmaco-
logical receptor binding profile of 45, the cis analogue 46
and the bicyclic derivative 47 bind only with moderate
affinities to the D3 receptor (Ki = 120 and 140 nM,
respectively), as we have predicted it. Indeed, they
showed even higher affinities for the a1 receptor subtype.
Chemicals and solvents were purchased in highest purity
available. All reactions were carried out under nitrogen
atmosphere. Column chromatography was performed
using 60 lm silica gel from Merck. For TLC silica gel
60 F254 plates from Merck were used (UV, I2 or ninhy-
drin detection). Melting temperatures were determined
on a Buechi 510 apparatus and are uncorrected. NMR
data were acquired on a Bruker AM-360 or Bruker
AVANCE 360 MHz spectrometer. Chemical shifts are
noted in ppm relative to TMS. IR spectroscopy was car-
ried out on a Jasco FT/IR 410 spectrometer. EI-MS was
performed on a Finnigan MAT TSQ 70 spectrometer.
Molecular modeling investigations were performed on
a Silicon Graphics Indigo2 and Octane2 workstation.
4. Conclusion
In the current work, we have successfully established
CoMFA and CoMSIA models based on a training
set of 32 ligands. We have carefully evaluated the sta-
tistical significance of the models and found high qc2v
values of 0.82 (2 components) and 0.76 (4 compo-
nents) for CoMFA and CoMSIA, respectively, when
using the standard leave-one-out cross-validation
method. Even for the more critical leave-20%-out
method, q2cv was only slightly reduced yielding still
highly significant mean values (n = 20 runs) of 0.80
for CoMFA and 0.73 for CoMSIA. The models were
verified to be stable and robust against the variation
of the underlying parameters. Thus, an all placement
search (APS) yielded only CoMFA models in the
rather narrow range between 0.807 and 0.839. Using
a test set of 12 ligands, we were able to demonstrate
that the models are also predictive for new com-
pounds. This was extended to the successful applica-
tion of our models guiding the synthesis of novel,
rigidized derivatives (45–47). Thus, we have discov-
ered a new bioactive agent (45) with decreased flexi-
bility by introduction of a trans-cyclohexyl spacer.
In addition to good D3 affinity, the test compound
45 exhibited improved (ꢀ3200-fold) selectivity over
D4 receptors.
5.2. Benzo[b]thiophene-2-carboxylic acid 4-hydrox-
ycyclohexyl amide (48)
To a solution of benzo[b]thiophene-2-carboxylic acid
(purchased from Acros Organics, Belgium) (0.28 g,
1.20 mmol) and DIPEA (0.70 mL, 4.80 mmol) in
CH2Cl2 (10 mL) was added TBTU (0.42 g, 1.30 mmol)
in DMF (4 mL) at 0 °C. After addition of trans-4-
aminocyclohexanol (0.17 g, 2.60 mmol) (purchased from
Acros Organics, Belgium) in CH2Cl2 (5 mL), the mixture
was stirred at room temperature for 1 h. Then, CH2Cl2
and aqueous saturated NaHCO3 were added. The
combined organic layers were dried (MgSO4) and
evaporated, and the residue was purified by flash chro-
matography (CH2Cl2–MeOH 98:2) to give benzo[b]thio-
phene-2-carboxylic acid 4-hydroxycyclohexyl amide
(0.21 g, 64%) as a white solid. Mp 220–222 °C. 1H
NMR (CDCl3, 360 MHz): d 1.30–1.55 (m, 4H), 2.00–
2.09 (m, 2H), 2.10–2.20 (m, 2H), 3.64–3.71 (m, 1H),
3.97–3.99 (m, 1H), 5.85–5.87 (m, 1H), 7.38 (ddd,
J = 0.3 Hz, 5.0 Hz, 8.0 Hz, 1H), 7.40–7.43 (m,
J = 0.3 Hz, 5.0 Hz, 8.0 Hz, 1H), 7.76 (d, J = 0.1 Hz),
7.80–7.86 (m, 2H). 13C NMR (CDCl3, 90 MHz): d
The theoretical investigations presented in this study
provide a valuable tool for predicting the affinity of nov-
el compounds and, thus, for guiding and evaluating fur-
ther structural modifications.

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5909
30.9, 33.9, 55.3, 69.8, 122.7, 124.9, 125.0, 125.2, 126.7,
138.5, 139.1, 140.7, 161.6. IR (NaCl): 3385, 3321,
2925, 2852, 1621 cmꢁ1. EI-MS m/z: 275 M+. Anal.
(C15H17NO2S) C, H, N.
scribed before the mixture of isomers was purified
another time by flash chromatography (CH2Cl2–MeOH
97:3) yielding in the pure trans- (14.0 mg, 22% of the
mixture) and cis-isomer (9.00 mg, 14%). Mp 140–
1
143 °C. H NMR (CDCl3, 360 MHz): d 1.64–2.02 (m,
5.3. Benzo[b]thiophene-2-carboxylic acid 4-oxocyclohexyl
amide (49)
8H), 2.31–2.36 (m, 1H), 2.74–2.78 (m), 3.08–3.11 (m,
4H), 4.22–4.29 (m, 1H), 6.16–6.19 (m, 1H), 6.95–698
(m), 7.14–7.16 (m, 2H), 7.39–7.44 (m, 2H), 7.76 (s,
1H), 7.80–7.86 (m, 2H).). 13C NMR (CDCl3, 90 MHz):
d 15.2, 25.2, 28.6, 49.8, 51.5, 65.8, 118.5, 122.7, 124.6,
124.9, 125.0, 125.1, 126.3, 127.4, 127.5, 134.1, 138.8,
139.1 161.5. IR (NaCl): 3274, 3060, 2944, 2801, 2762,
1620 cmꢁ1. EI-MS m/z: 487, 489 (M+) HREIMS calcd
for C25H27Cl2N3OS: 487.1252. Found: 487.1249 (M+).
To a solution of benzo[b]thiophene-2-carboxylic acid
4-hydroxycyclohexyl amide (0.10 g, 0.36 mmol) was
added IBX ((1-hydroxy-1,2-benziodoxol-3(1H)-on-1-ox-
ide) 0.12 g, 0.43 mmol) in dry DMSO (4 mL) and stirred
at rt for 1 h. Then, Et2O and aqueous saturated NaH-
CO3 were added. The organic layer was dried (MgSO4)
and evaporated to yield in benzo[b]thiophene-2-carbox-
ylic acid 4-oxocyclohexyl amide (60.0 mg, 56%) as a
white solid. Mp 185–188 °C. 1H NMR (CDCl3,
360 MHz): d 1.75–1,78 (m, 2H), 1.83–1.86 (m, 2H),
2.33–2.41 (m, 2H), 2.44–2.56 (m, 2H), 4.43–446 (m,
1H), 6.16–6.18 (m, 1H), 7.38 (ddd, J = 0.3 Hz, 5.0 Hz,
8.0 Hz, 1H), 7.42 (ddd, J = 0.3 Hz, 5.0 Hz, 8.0 Hz,
1H), 7.78 (d, J = 0.3 Hz, 1H), 7.80–7.87 (m, 2 H) 13C
NMR (CDCl3, 90 MHz): d 32.1, 39.1, 47.2, 122.7,
125.0, 125.5, 126.5, 138.0, 139.0, 140.8, 161.9, 209.1.
IR (NaCl): 3313, 2942, 2852, 1716, 1625 cmꢁ1. EI-MS
m/z: 273 (M+). Anal. (C15H15NO2S Æ 0.16H2O) C, H, N.
5.5. [(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyri-
do[1,2-a]pyrazine-7-yl]methanol (51)
[(5S,7S)-1-(Octahydropyrido[1,2-a]pyrazine-7-yl]metha-
nol43 was synthesized according to Ref. 43. To a suspen-
sion of [(5S,7S)-1-(octahydropyrido[1,2-a]pyrazine-7-
yl]methanol (0.50 g, 2.90 mmol), 1-bromo-2,3-dichloro-
benzene (0.55 g, 2,50 mmol), Pd2(dba)3 (22.5 mg,
0.50 mol%), and BINAP (( )-2,20-bis(diphenylphosphi-
no)-1,10-binaphthyl; 45.0 mg 2 mol%) in toluene
(8 mL) was added NaOHt-Bu (1.00 g, 12.0 mmol). The
reaction mixture was heated for 2.5 h to 80 °C, cooled
to room temperature, filtered through Celite, and evap-
orated, yielding [(7S,9aS)-2-(2,3-dichlorophenyl)octahy-
dropyrido[1,2-a]pyrazine-7-yl]methanol (0.47 g, 51%) as
5.4. N-[4-[4-(2,3-Dichlorophenyl)piperazin-1-yl]cyclohex-
yl]benzo[b]thiophene-2-carboxamide (45 + 46)
1
To a solution of benzo[b]thiophene-2-carboxylic acid
4-oxocyclohexyl amide (0.60 mg, 0.21 mmol) and
Na(OAc)3BH (60.0 mg, 0.28 mmol) in dry C2H5Cl2
(5 mL) was added 1-(2,3-dichlorophenyl)piperazine
(0.10 g, 0.42 mmol) and stirred at rt for 3.5 h. After
addition of Et2O and aqueous saturated NaHCO3, the
aqueous layer was extracted with CH2Cl2. The
combined organic layers were dried (MgSO4) and evap-
orated, and the residue was purified by flash chromatog-
raphy (CH2Cl2–MeOH 98:2) to give N-[4-[4-(2,3-
dichlorophenyl)piperazine-1-yl]cyclohexyl]benzo[b]thio-
phene-2-carboxamide as a mixture of cis-, trans-isomers
(64.0 mg, 84%).
a yellow oil. H NMR (CDCl3, 360 MHz): d 1.58–1.56
(m, 1H), 1.69–1.91 (m, 4H), 2.34 (br dd, J = 10.3 Hz,
10.3 Hz, 1H), 2.48 (ddd, J = 2.5 Hz, 11.3 Hz, 11.3 Hz,
1H), 2.49–2.52 (m, 1H), 2.53 (dd, J = 10.6 Hz, 10.6 Hz,
1H), 2.88 (ddd, J = 2.5 Hz, 5.2 Hz, 11.3 Hz, 1H), 2.92
(dd, =2.7 Hz, 11.4 Hz, 1H), 3.01 (br d, J = 11.6 Hz,
1H), 3.23 (ddd, J = 2.6 Hz, 2.6 Hz, 10.6 Hz, 1H), 3.31
(br dd, J 2.7 Hz, 11.4 Hz, 1H), 3.89 (ddd, J = 1.4 Hz,
3.4 Hz, 10.4 Hz, 1H), 4.02 (dd, J = 4.5 Hz, 10.4 Hz,
1H), 6.95 (m, 1H), 7.14–7.20 (m, 2H). 13C NMR
(CDCl3, 90 MHz): d 26.6, 27.5, 34.2, 51.4, 55.2, 57.3,
58.5, 61.0, 68.6, 118.5, 124.6, 127.4, 127.5, 134.2,
151.0. IR (NaCl): 3379, 2924, 2853, 2824, 1578 cmꢁ1
.
EI-MS m/z: 314, 316 (M+).
5.4.1.
trans-N-[4-[4-(2,3-Dichlorophenyl)-piperazin-1-
yl]cyclohexyl]benzo[b]thiophene-2-carboxamide (45). The
mixture of isomers was purified another time by flash
chromatography (CH2Cl2–MeOH 97:3) yielding in the
pure trans- (14.0 mg, 22% of the mixture) and cis-isomer
(9.00 mg, 14%). Mp 266–268 °C. 1H NMR (CDCl3,
360 MHz): d 1.35 (m, 4H), 2.06–2.08 (m, 2H), 2.25–
2.28 (m, 2H), 2.40–2.43 (m, 1H), 2.79–2.80–2.83 (m,
4H), 3.10–3.13 (m, 4H), 3.96–3.99 (m, 1H), 5.92–5.94
(m, 1H), 6.99–7.05 (m, 1H), 7.13–7.16 (m, 2H), 7.37–
7.44 (m, 2H), 7.75 (s, 1H), 7.81–7.86 (m, 2H). 13C
NMR (CDCl3, 90 MHz): d 27.3, 32.3, 49.1, 49.3, 51.8,
62.7, 118.6, 122.7, 124.5, 124.9, 125.0, 125.1, 126.3,
127.4, 127.5, 134.0, 138.6, 139.1, 140.7, 151.3, 161.5. IR
(NaCl): 3274, 3060, 2944, 2801, 2762, 1620 cmꢁ1. EI-
MS m/z: 487, 489 (M+). Anal. (C25H27Cl2N3OS) C, H, N.
5.6. [(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyri-
do[1,2-a]pyrazine-7-yl] methyl-sulfonate (52)
[(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyrido[1,2-
a]pyrazine-7-yl]methanol (0.20 g, 0.65 mmol) was dis-
solved in dry THF (4 mL), cooled to ꢁ30 °C. Then were
added Et3N (0.14 mL, 0.97 mmol) and methanesulfonyl
chloride (0.08 mL, 6.00 mmol). The reaction mixture
was stirred at room temperature for 30 min, filtered
through Celite, and evaporated, yielding [(7S,9aS)-2-
(2,3-dichlorophenyl)octahydropyrido[1,2-a]pyrazine-7-
yl] methyl-sulfonate (0.18 g, 71%) as a white solid. Mp
1
110–112 °C. H NMR (CDCl3, 360 MHz): d 1.36 (ddd,
J = 3.5 Hz, 10.6 Hz, 13.6 Hz, 1H), 1.48 (dddd,
J = 2.5 Hz, 3.5 Hz, 3.5 Hz, 13.6 Hz, 1H), 1.68 (dddd,
J = 4.8 Hz, 4.8 Hz, 13.9 Hz, 13.9 Hz, 1H), 1.86 (dddd,
J = 1.6 Hz, 1.6 Hz, 2.7 Hz, 13.9 Hz, 1H), 2.17–2.30 (m,
1H), 2.27 (ddd, J = 2.7 Hz, 10.3 Hz, 10.6 Hz, H-4a),
5.4.2. cis-N-[4-[4-(2,3-Dichlorophenyl)piperazin-1-yl]cy-
clohexyl]benzo[b]thiophene-2-carboxamide (46). As de-

5910
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
2.36 (dd, J = 3.4 Hz, 11.8 Hz, 1H), 2.49 (ddd,
J = 2.9 Hz, 11.3 Hz, 11.3 Hz, 1H, H-6a), 2.51 (dd,
J = 10.2 Hz, 10.7 Hz, 1H), 2.74 (ddd, J = 2.4 Hz,
2.4 Hz, 11.3 Hz, 1H, H-6e), 2.86 (ddd, J = 1.8 Hz,
1.8 Hz, 12.0 Hz, 1H), 2.89 (ddd, J = 2.6 Hz, 11.4 Hz,
11.4 Hz, 1H), 3.06 (s, 3H), 3.16 (ddd, J = 2.6 Hz,
2.6 Hz, 10.7 Hz, 1H), 3.25 (ddd, J = 2.6 Hz, 2.6 Hz,
11.4 Hz, 1H), 4.40 (dd, J = 7.5 Hz, 9.5 Hz, 1H), 4.51
(dd, J = 7.5 Hz, 9.5 Hz, 1H), 6.96 (dd, J = 3.2 Hz,
6.4 Hz, 1H), 7.14–7.19 (m, 2H). 13C NMR (CDCl3,
90 MHz): d 24.3, 24.8, 33.5, 37.1, 51.3, 54.95, 55.6,
57.1, 61.1, 71.1, 118.5, 124.6, 127.4, 127.5, 134.0,
151.1. IR (NaCl): 2929, 2818, 2774, 1578, 1450, 1244,
1355, 1174 cmꢁ1. EI-MS m/z: 392, 394 (M+).
(CDCl3, 360 MHz): d 1.33–1.46 (m, 2H), 1.55–1.68 (m,
3H), 1.78–1.84 (m, 2H), 2.14–2.18 (m, 1H), 2.27 (dd,
J = 3.0 Hz, 11.3 Hz, 1H), 2.43 (ddd, J = 2.7 Hz,
11.4 Hz, 11.4 Hz), 2.50 (dd, J = 10.4 Hz, 10.4 Hz, 1H),
2.67–2.76 (m, 4H), 2.91 (ddd, J = 2.7 Hz, 11.4 Hz,
11.4 Hz), 3.14 (ddd, J = 2.5 Hz, 2.5 Hz, 10.4 Hz, 1H),
3.25 (ddd, J = 2.7 Hz, 5.0 Hz, 11.4 Hz, 1H), 6.67–6.71
(m, 1H), 6.73–6.80 (m, 1H), 6.92–6.99 (m, 1H, Phenyl).
13C NMR (CDCl3, 90 MHz): d 25.0, 27.8, 31.2, 35.4,
40.7, 51.3, 55.3, 57.1, 59.2, 61.6, 118.5, 124.4, 127.4,
127.5, 133.9, 151.3. IR (NaCl): 3361, 2928, 2854,
2817 cmꢁ1
.
(C16H23Cl2N3) C, H, N.
EI-MS m/z: 327, 329 (M+). Anal.
5.9. Benzo[b]thiophene-2-carboxylic acid [(7S,9aS)-2-
(2,3-dichlorophenyl)octahydro-pyrido[1,2-a]pyrazine-7-
yl]ethyl amide (47)
5.7. [(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyri-
do[1,2-a]pyrazine-7-yl]acetonitrile (53)
[(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyrido[1,2-
a]pyrazine-7-yl] methyl-sulfonate (0.10 g, 0.26 mmol)
was dissolved in dry DMSO (3 mL) and toluene
(5 mL). The mixture was added to tetrabutylammonium
cyanide (0.69 g, 2.55 mmol) and heated at 80 °C for 8 h.
Then, the reaction mixture was cooled to room temper-
ature and aqueous saturated NaHCO3 was added. After
extraction with ethyl acetate, the combined organic lay-
ers were dried (MgSO4) and evaporated to give
[(7S,9aS)-2-(2,3-dichlorophenyl)octahydropyrido[1,2-
a]pyrazine-7-yl]acteonitrile (64.0 mg, 77%) as a white
To a solution of benzo[b]thiophene-2-carboxylic acid
(purchased from Acros Organics, Belgium) (17.8 mg,
0.10 mmol), and DIPEA (0.07 mL, 0.42 mmol) in
CH2Cl2 (3 mL) was added TBTU (42.0 mg, 0.13 mmol)
in DMF (0.3 mL) at 0 °C. After addition of [(7S,9aS)-2-
(2,3-dichlorophenyl)octahydropyrido[1,2-a]pyrazine-7-
yl]ethylamine (40.0 mg, 0.13 mmol) in CH2Cl2 (5 mL),
the mixture was stirred at room temperature for 1 h.
Then, CH2Cl2 and aqueous saturated NaHCO3 were
added. The organic layer was dried (MgSO4) and evap-
orated, and the residue was purified by flash chromatog-
raphy (CH2Cl2–MeOH 98:2) to give benzo[b]thiophene-
2-carboxylic acid [(7S,9aS)-2-(2,3-dichlorophenyl)octa-
hydropyrido-[1,2-a]pyrazine-7-yl)]ethyl amide as a white
1
solid. Mp 146–148 °C. H NMR (CDCl3, 360 MHz): d
1.36 (dddd, J = 4.0 Hz, 11.0 Hz, 13.5 Hz, 14.0 Hz, 1H),
1.49 (dddd, J = 2.5 Hz, 2.5 Hz, 4.7 Hz, 13.5 Hz, 1H),
1.74 (dddd, J = 4.5 Hz, 4.5 Hz, 13.5 Hz, 13.5 Hz, 1H),
1.84 (dddd, J = 2.5 Hz, 2.5 Hz, 4.5 Hz, 13.5 Hz, 1H),
2.21–2.30 (m, 2H), 2.41 (dd, J = 3.1 Hz, 11.9 Hz, 1H),
2.49 (dd, J = 10.2 Hz, 10.9 Hz, 1H), 2.52 (ddd,
J = 3.0 Hz, 11.2 Hz, 11.2 Hz, 1H), 2.67 (dd, J = 7.7 Hz,
16.8 Hz, 1H), 2.74–2.81 (m, 4H), 2.91 (ddd, J = 2.7 Hz,
11.3 Hz, 11.3 Hz, 1H), 3.21 (ddd, J = 2.6 Hz, 2.6 Hz,
10.9 Hz, 1H), 3.29 (ddd, J = 2.7 Hz, 5.2 Hz, 11.3 Hz,
1H), 6.97 (dd, J = 3.0 Hz, 6.6 Hz, 1H), 7.14–7.19 (m,
2H). 13C NMR (CDCl3, 90 MHz): d 19.6, 24.1, 26.9,
31.4, 51.3, 54.8, 57.1, 57.8, 61.1, 118.5, 119.8, 124.6,
127.4, 127.5, 134.1, 151.0. IR (NaCl): 3419, 2931,
2818, 2242 cmꢁ1. EI-MS m/z: 323, 325 (M+). Anal.
(C16H19Cl2N3 Æ 0.25H2O) C, H, N.
1
solid (35.0 mg, 71%). Mp 74–76 °C. H NMR (CDCl3,
360 MHz): d 1.42–1.57 (m, 2H), 1.62–1.71 m, 2H),
1.84–1.94 (m, 2H), 2.00–2.18 (m, 1H), 2.25 (dddd,
J = 3.0 Hz, 3.0 Hz, 10.2 Hz, 10.2 Hz, 1H), 2.34 (dd,
J = 3.0 Hz, 11.7 Hz, 1H), 2.47 (ddd, J = 3.0 Hz,
11.3 Hz, 11.3 Hz, 1H), 2.55 (dd, J = 10.2 Hz, 10.8 Hz,
1H, H-1a), 2.74 (ddd, J = 2.3 Hz, 2.3 Hz, 11.3 Hz, 1H,
H-4e), 2.77 (d, J = 11.7 Hz, H-6e), 2.89 (ddd,
J = 2.2 Hz, 2.2 Hz, J = 11.5 Hz), 3.17 (ddd, J = 2.5 Hz,
2.5 Hz, 10.8 Hz, 1H, H-1e), 3.22 (dd, J = 2.2 Hz,
11.5 Hz, 1H), 3.48 (ddd, J = 6.3 Hz, 6.5 Hz, 13.0 Hz,
1H), 3.59 (ddd, J = 6.3 Hz, J = 13.0 Hz, 13.0 Hz, 1H),
6.51 (br s, 1H), 6.88 (m, 1H, Phenyl), 7.13–7.16 (m,
2H, Phenyl), 7.39 (ddd, J = 1.5 Hz, 7.4 Hz, 7.4 Hz,
1H), 7.42 (ddd, J = 1.5 Hz, 7.4 Hz, 7.4 Hz, 1H), 7.75
(s, 1H), 7.81–7.86 (m, 2H). 13C NMR (CDCl3,
90 MHz): d 24.9, 27.8, 30.9, 31.6, 38.9, 51.2, 55.2, 57.1,
58.6, 61.8, 118.5, 122.7, 124.5, 124.9, 125.0, 126.2,
127.4, 127.5, 130.6, 134.0, 138.7, 139.1, 140.7, 151.1,
5.8. [(7S,9aS)-2-(2,3-Dichlorophenyl)octahydropyri-
do[1,2-a]pyrazine-7-yl]ethylamine (54)
A solution of [(7S,9aS)-2-(2,3-Dichlorophenyl)octahy-
dropyrido[1,2-a]pyrazine-7-yl]acteonitrile
(45.0 mg,
162.3. IR (NaCl): 3358, 3064, 2926, 2853, 1627 cmꢁ1
.
0.14 mmol) in dry Et2O (5 mL) was cooled to 0 °C and
a solution of LiAlH4 (1 M in Et2O, 0.28 mL, 0.28 mmol)
was added slowly. The mixture was allowed to warm up
to room temperature and stirred for 1 h. Afterwards it
was cooled again to 0 °C and quenched by cautious
dropwise addition of water. The mixture was then fil-
tered, and the filter cake was washed with diethyl ether
(50 mL). The combined organic layers were washed with
saturated NaHCO3 and evaporated to give [(7S,9aS)-2-
(2,3-dichlorophenyl)octahydro-pyrido[1,2-a]pyrazine-7-
EI-MS m/z: 487, 489 (M+). Anal. (C25H27Cl2N3OS) C,
H, N.
5.10. Receptor binding experiments
Receptor binding studies were carried out as described
in the literature.46,47 In brief, the dopamine D1 receptor
assay was done with porcine striatal membranes at a fi-
nal protein concentration of 40 lg/assay tube and the
radioligand [3H]SCH 23390 at 0.3 nM (Kd = 0.95 nM).
1
yl]ethylamine as a yellow oil (30.0 mg, 66%). H NMR
Competition experiments with the human D2long,

I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
5911
Neuroscience, San Diego, CA, Nov 11–16, 1995; Society
of Neuroscience: Washington, DC, 1995.
8. Boeckler, F.; Gmeiner, P. Pharmacol. Ther. 2006, in press.
D2short, D3, and D4.4 receptors were run with prepara-
tions of membranes from CHO cells expressing the cor-
responding receptor and [3H]spiperone at a final
concentration of 0.1 nM. The assays were carried out
with a protein concentration of 3–10 lg/assay tube and
Kd values of 0.06–0.14 nM for D2long, 0.10–0.15 nM
for D2short, 0.22–0.35 nM for D3, and 0.28–0.33 nM
for D4.4.
9. Lo¨ber, S.; Hubner, H.; Utz, W.; Gmeiner, P. J. Med.
¨
Chem. 2001, 44, 2691.
10. Lo¨ber, S.; Aboul-Fadl, T.; Hubner, H.; Gmeiner, P.
¨
Bioorg. Med. Chem. Lett. 2002, 12, 633.
11. Lo¨ber, S.; Hubner, H.; Gmeiner, P. Bioorg. Med. Chem.
¨
Lett. 1999, 9, 97.
12. Ortner, B.; Waibel, R.; Gmeiner, P. Angew. Chem., Int.
Ed. 2001, 40, 1283.
13. Newman, A. H.; Grundt, P.; Nader, M. A. J. Med. Chem.
2005, 48, 3663.
14. Aston-Jones, G.; Druhan, J. Nature 1999, 400, 317.
15. Boeckler, F.; Leng, A.; Mura, A.; Bettinetti, L.; Feldon, J.;
Gmeiner, P.; Ferger, B. Biochem. Pharmacol. 2003, 66,
1025.
16. Bettinetti, L.; Schlotter, K.; Hubner, H.; Gmeiner, P.
¨
J. Med. Chem. 2002, 45, 4594.
The investigation of serotonin 5-HT1A and 5-HT2 and
adrenergic a1 binding was performed as described in
the literature.48 In brief, porcine cortical membranes
were subjected to the binding assay at a concentration
of 80 lg/assay tube for determination of 5-HT1A and
5-HT2 binding utilizing [3H]8-OH-DPAT and [3H]ke-
tanserin each at a final concentration of 0.5 nM with
KD values of 2.6 nM (for 5-HT1A) and 1.9 nM (for 5-
HT2). Cortical membranes at 55 lg/assay tube and the
radioligand [3H]prazosin at a final concentration of
0.4 nM were applied to determine adrenergic a1 binding
with KD values of 0.08–0.10 nM.
17. Bettinetti, L.; Lo¨ber, S.; Hubner, H.; Gmeiner, P.
¨
J. Comb. Chem. 2005, 7, 309.
18. Gmeiner, P.; Huebner, H.; Schlotter, K. WO2004004729,
2004.
19. Lanig, H.; Utz, W.; Gmeiner, P. J. Med. Chem. 2001, 44,
1151.
20. Boeckler, F.; Ohnmacht, U.; Lehmann, T.; Utz, W.;
Protein concentration was established by the method of
Lowry using bovine serum albumin as standard.49
Hubner, H.; Gmeiner, P. J. Med. Chem. 2005, 48, 2493.
¨
21. Cha, M. Y.; Lee, I. Y.; Cha, J. H.; Choi, K. I.; Cho, Y. S.;
Koh, H. Y.; Pae, A. N. Bioorg. Med. Chem. 2003, 11,
1293.
22. Hackling, A.; Ghosh, R.; Perachon, S.; Mann, A.; Holtje,
H. D.; Wermuth, C. G.; Schwartz, J. C.; Sippl, W.;
Sokoloff, P.; Stark, H. J. Med. Chem. 2003, 46, 3883.
23. Sybyl 6.9, Tripos Inc., 1699 South Hanley Road, St.
Louis, Missouri, 63144, U.S.A.
Data analysis of the resulting competition curves was
accomplished by non-linear regression analysis using
the algorithms in PRISM (GraphPad Software, San
Diego, CA). Ki values were derived from the corre-
sponding EC50 data utilizing the equation of Cheng
and Prusoff.50
24. Clark, M.; Cramer, R. D., III; Van Opdenbosch, N.
J. Comput. Chem. 1989, 10, 982.
Acknowledgments
25. Gasteiger, J.; Marsili, M. Tetrahedron 1980, 36, 3219.
26. Hinze, J.; Jaffe, H. H. J. Am. Chem. Soc. 1962, 84, 540.
27. Stewart, J. J. J. Comput. Aided Mol. Des. 1990, 4, 1.
28. Stewart, J. J. P.; Frank, J. MOPAC, Seiler Bokstav
Research Laboratory, United Air Force Academy, CO
80840.
29. ASP V3.22, Oxford Molecular Ltd, The Medawar Centre,
Oxford Science Park, Sandford-on-Thames, Oxford OX4
4GA, U.K.
The authors wish to thank Dr. H. H. M. Van Tol
(Clarke Institute of Psychiatry, Toronto), Dr. J.-C. Sch-
wartz and Dr. P. Sokoloff (INSERM, Paris) as well as J.
Shine (The Garvan Institute of Medical Research, Syd-
ney) for providing dopamine D4, D3, and D2 receptor
expressing cell lines, respectively, and Dr. R. Huff (Pfiz-
er, Kalamazoo, MI) for providing dopamine D4.2
receptor expressing cells. This work was supported by
the BMBF and Fonds der Chemischen Industrie.
30. TSAR V3.2, Oxford Molecular Ltd, The Medawar Centre,
Oxford Science Park, Sandford-on-Thames, Oxford OX4
4GA, U.K.
31. Beck, B.; Clark, T.; Glen, R. C. J. Mol. Model. 1995, 1,
176.
References and notes
32. Clark, T.; Alex, A.; Beck, B.; Chandrasekar, J.; Gedeck,
P.; Horn, A.; Huttern, M.; Martin, B.; Rauhut, G.; Sauer,
W.; Schindler, T.; Steinke, T. VAMP 7.0 Program
Package; Oxford Molecular Plc: Oxford, 1998.
33. Tropsha, A.; Cho, S. J.. In 3D QSAR in Drug Design;
Kubinyi, H., Ed.; ESCOM: Leiden, 1998; Vol. 3, pp 57–
69.
34. Wang, R.; Gao, Y.; Liu, L.; Lai, L. J. Mol. Model. 1998, 4,
276.
35. Wold, S. Quant. Struct.-Act. Relat. 1991, 10, 191.
36. Cramer, R. D., III; Patterson, D. E.; Bunce, J. D. J. Am.
Chem. Soc. 1988, 110, 5959.
37. Shao, J. J. Am. State Assoc. 1993, 88, 486.
38. Baroni, M.; Clementi, S.; Cruciani, G.; Costantino, G.;
Riganelli, D.; Oberrauch, E. J. Chemom. 1992, 6, 347.
39. Cruciani, G.; Baroni, M.; Clementi, S.; Costantino,
G.; Riganelli, D.; Skagerberg, B. J. Chemom. 1992, 6,
335.
1. Sokoloff, P.; Giros, B.; Martres, M. P.; Bouthenet, M. L.;
Schwartz, J. C. Nature 1990, 347, 146.
2. Richtand, N. M.; Woods, S. C.; Berger, S. P.; Strakowski,
S. M. Neurosci. Biobehav. Rev. 2001, 25, 427.
3. Joyce, J. N.; Millan, M. J. Drug Discov. Today 2005, 10,
917.
4. Joyce, J. N.; Ryoo, H. L.; Beach, T. B.; Caviness, J. N.;
Stacy, M.; Gurevich, E. V.; Reiser, M.; Adler, C. H. Brain
Res. 2002, 955, 138.
5. Bezard, E.; Ferry, S.; Mach, U.; Stark, H.; Leriche, L.;
Boraud, T.; Gross, C.; Sokoloff, P. Nat. Med. 2003, 9, 762.
6. Pilla, M.; Perachon, S.; Sautel, F.; Garrido, F.; Mann, A.;
Wermuth, C. G.; Schwartz, J. C.; Everitt, B. J.; Sokoloff,
P. Nature 1999, 400, 371.
7. Unis, A. S.; Roberson, M. D.; Hill, K.; Hamblin, M. W.;
Dorsa, D. W. In: 25th Annual Meeting of the Society of

5912
I. Salama et al. / Bioorg. Med. Chem. 14 (2006) 5898–5912
40. Bo¨hm, M.; Sturzebecher, J.; Klebe, G. J. Med. Chem.
¨
1999, 42, 458.
46. Hubner, H.; Haubmann, C.; Utz, W.; Gmeiner, P. J. Med.
¨
Chem. 2000, 43, 756.
41. Thibaut, U.; Folkers, G.; Klebe, G.; Kubinyi, H.;
Merz, A.; Rognan, D. In 3D QSAR in Drug
Design. Theory, Methods and Applications; Kubinyi,
H., Ed.; ESCOM: Leiden, The Netherlands, 1993;
pp 711–716.
42. Hattotuwagama, C. K.; Doytchinova, I. A.; Flower, D. R.
J. Chem. Inf. Mod. 2005, 45, 1415.
43. Bright, G. M.; Desai, K. A. WO9008148, 1990.
44. Hartwig, J. F. Angew. Chem., Int. Ed. 1998, 37, 2046.
45. Wolfe, J. P.; Tomori, H.; Sadighi, J. P.; Yin, J.; Buchwald,
S. L. J. Org. Chem. 2000, 65, 1158.
47. Lenz, C.; Haubmann, C.; Hubner, H.; Boeckler, F.;
¨
Gmeiner, P. Bioorg. Med. Chem. 2005, 13, 185.
48. Schlotter, K.; Boeckler, F.; Hubner, H.; Gmeiner, P. J.
Med. Chem. 2005, 48, 3696.
¨
49. Lowry, O. H.; Rosebrough, N. J.; Farr, A. L.; Randall, R.
J. J. Biol. Chem. 1951, 193, 265.
50. Cheng, Y.; Prusoff, W. H. Biochem. Pharmacol. 1973, 22,
3099.
51. Clark, R. D.; Sprous, D. G.; Leonard, J. M. In Rational
Approaches to Drug Design; Hoeltje, H.-D., Sippl, W.,
Eds.; Prous Science SA: 2001; pp 475–485.