Pd-catalyzed direct C-H bond functionalization of spirocyclic σ1 ligands: Generation of a pharmacophore model and analysis of the reverse binding mode by Docking into a 3D homology model of the σ1 receptor

Article abstract of DOI:10.1021/jm300894h

To explore the hydrophobic binding region of the σ₁ receptor protein, regioisomeric spirocyclic thiophenes 9-11 were developed as versatile building blocks. Regioselective α- and β-arylation using the catalyst systems PdCl₂/bipy/Ag

Full text of DOI:10.1021/jm300894h

Article
pubs.acs.org/jmc
Pd-Catalyzed Direct C−H Bond Functionalization of Spirocyclic σ1
Ligands: Generation of a Pharmacophore Model and Analysis of the
Reverse Binding Mode by Docking into a 3D Homology Model of the
σ1 Receptor
Christina Meyer,† Dirk Schepmann,† Shuichi Yanagisawa,‡ Junichiro Yamaguchi,‡ Valentina Dal Col,§
Erik Laurini,§ Kenichiro Itami,*,‡ Sabrina Pricl,*,§,∥ and Bernhard Wunsch*,†
̈
†Institut fur Pharmazeutische und Medizinische Chemie der Westfalischen Wilhelms-Universitat Munster, Hittorfstrasse 58-62,
̈
̈
̈
̈
D-48149 Munster, Germany
̈
‡Department of Chemistry, Graduate School of Science, Nagoya University, Chikusa-ku, Nagoya 464-8602, Japan
§Molecular Simulation Engineering (MOSE) Laboratory, Department of Industrial Engineering and Information Technology (DI3),
University of Trieste, Via Valerio 10, 34127 Trieste, Italy
∥National Interuniversity Consortium for Material Science and Technology (INSTM), Research Unit MOSE-DEA, University of
Trieste, 34127 Trieste, Italy
S
* Supporting Information
ABSTRACT: To explore the hydrophobic binding region of
the σ1 receptor protein, regioisomeric spirocyclic thiophenes
9−11 were developed as versatile building blocks. Regiose-
lective α- and β-arylation using the catalyst systems PdCl2/
bipy/Ag2CO3 and PdCl2/P[OCH(CF3)2]3/Ag2CO3 allowed
the introduction of various aryl moieties at different positions
in the last step of the synthesis. The increasing σ1 affinity in the
order 4 < 5/6 < 7/8 indicates that the positions of the
additional aryl moiety and the S atom in the spirocyclic
thiophene systems control the σ1 affinity. The main features of
the pharmacophore model developed for this class of σ1
ligands are a positive ionizable group, a H-bond acceptor
group, two hydrophobic moieties, and one hydrophobic aromatic group. Docking of the ligands into a σ1 3D homology model via
molecular mechanics/Poisson−Boltzmann surface area calculations led to a very good correlation between the experimentally
determined and estimated free energy of receptor binding. These calculations support the hypothesis of a reverse binding mode
of ligands bearing the aryl moiety at the “top” (compounds 2, 3, 7, and 8) and “left” (compounds 4, 5, and 6) positions,
respectively.
Alzheimer’s and Parkinson’s diseases). Furthermore, σ1
antagonists are valuable drugs for the reduction of unpleasant
effects after withdrawal of cocaine, methamphetamine, or
alcohol from addicted animals.7−10 Due to the high
concentration of σ1 (and σ2) receptors in several tumor cells,
σ1 (and σ2) ligands may be developed for the therapy and
diagnosis of cancer.11,12
The signal transduction pathway after activation of σ1
receptors has not been completely understood so far, and
therefore, the pharmacological effects cannot be correlated with
a distinct biochemical mechanism. However, involvement of σ1
receptors in the modulation of various neurotransmitter
systems (e.g., glutamatergic, dopaminergic, and cholinergic
neurotransmission) has been shown. Additionally, some ion
1. INTRODUCTION
The class of σ receptors consists of two subtypes, which are
termed σ1 and σ2 receptors. Although the knowledge about the
structure of the σ2 receptor is rather low, it was reported very
recently that the σ2 receptor might be identical to the
progesterone receptor membrane component 1 (pgrmc1).
Cloning of pgrmc1 led to a protein of 194 amino acids with a
molecular weight (MW) of 21 670.1,2 The σ1 receptor is well
characterized on the level of DNA and amino acid sequence.
The human σ1 receptor gene encodes for a protein of 223
amino acids with a molecular weight of 25 300. The membrane-
bound σ1 receptor protein has two transmembrane domains,
and both the amino and carboxy termini are located
intracellularly.3−6
Ligands modulating the σ1 receptor activity have a potential
for the treatment of acute and chronic neurological disorders
(e.g., schizophrenia, depression, (neuropathic) pain, and
Received: June 25, 2012
Published: August 23, 2012
© 2012 American Chemical Society
8047
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
channels (i.e., K+, Na+, and Ca2+ channels) are regulated by σ1
receptors.13−19
2. CHEMISTRY
To obtain a large number of arylated spirocyclic thiophenes, we
planned to introduce various aryl moieties into spirocyclic
building blocks in the last step of the synthesis. Pd-catalyzed
cross-coupling reactions of metalated arene/heteroarene and
halogenated arene/heteroarene species as exemplified by the
Suzuki−Miyaura coupling are undoubtedly among the most
reliable methods for the synthesis of biaryls and heterobiaryls.26
For this type of cross-coupling reaction, two activated species,
i.e., a halogenated and a metalated (hetero)arene, are required.
However, the direct C−H bond arylation of thiophene
derivatives provides a step economical access to a large number
of diverse test compounds without previous activation of the
thiophene moiety.
Our interest has been focused on the development of novel
compounds with high σ1 receptor affinity and selectivity over
related receptor systems (e.g., σ2 receptor, phencyclidine
binding site, and ifenprodil binding site of the N-methyl-D-
aspartate (NMDA) receptor) in the central nervous system.
Recently, we have shown that additional lipophilic substituents
on spirocyclic σ1 ligands favor interactions with the σ1 receptor
protein (Figure 1). Replacement of the methyl moiety of 1a (Ki
The spirocyclic thiophene derivatives with the aryl moiety on
the “left” (4−6) are obtained by α-arylation of the building
blocks 9−11, whereas the corresponding β-arylation of 10 and
11 provides spirocyclic compounds 7 and 8 with the aryl
moiety at the “upper left” or “top” position (Figure 2).
Figure 1. Design of arylated spirocyclic thiophenes.
Figure 2. Plan for the synthesis of arylated spirocyclic thiophenes.
= 21 nM) with a phenyl group (1b, Ki = 1.5 nM) led to a 15-
fold increased σ1 receptor binding while maintaining high σ1/σ2
selectivity.20,21 In the series of thiophene-annulated spirocyclic
σ1 ligands 2 and 3, an additional aryl moiety in the α-position of
the S atom is well tolerated by the σ1 receptor protein.22,23
These observations are in good accordance with established
pharmacophore models indicating the acceptance of large
lipophilic substituents by the σ1 receptor protein.24,25
Herein, we report on the introduction of various aryl
moieties in the α-position (4−6) and β-position (7, 8) of
regioisomeric thiophenes by direct C−H bond arylation using
recently developed Pd catalysts. Whereas the position of the
additional aryl moiety of 7 and 8 correlates with the position in
the lead compounds 1b, 2, and 3, the aryl moiety of the
spirocyclic ligands 4−6 is shifted to the “left” position of the
thiophene ring. On the other hand, compounds 4 and 5/6
represent regioisomers with respect to the position of the S
atom in the thiophene ring. The σ1 affinities of the arylated
thiophenes 4−8 are investigated in receptor binding studies
with radioligands. The resulting data together with the affinity
data of the arylated spirocyclic σ1 ligands 2 and 3 are used for
the establishment and validation of a σ1 pharmacophore model.
Moreover, docking of the ligands into the σ1 receptor 3D
model and affinity scoring will result in the analysis of detailed
ligand receptor interactions.
The building block 9 was synthesized as described in refs 27
and 28 beginning with the formylation of 3-bromothiophene.
The synthesis of the building blocks 11 and 3a started with
(thiophen-3-yl)acetaldehyde dimethyl acetal 12 (Scheme 1),
which was prepared from 3,4-dibromothiophene.22 Halogen
metal exchange with n-BuLi and subsequent trapping of the
resulting aryllithium intermediate with 1-benzylpiperidin-4-one
led to the regioisomeric hydroxy acetals 13 and 14. After
separation of both hydroxy acetals 13 and 14, p-toluenesulfonic
acid in methanol catalyzed the intramolecular transacetalization
to form the spirocyclic thiophenes 11 and 3a.22
Formation of the hydroxy acetal 13 is explained by
wandering of lithium to the more stable α-position. This
isomerization is strongly dependent on the temperature and the
solvent used. Conducting this reaction in Et2O at −78 °C gave
exclusively the hydroxy acetal 14. However, changing the
solvent from Et2O to tetrahydrofuran (THF) or raising the
temperature to −50 or 0 °C resulted in increasing formation of
the rearranged product 13. The highest yield (41%) of the
spirocyclic thiophene derivative 11 was obtained by performing
the halogen metal exchange in THF at 0 °C and subsequent
treatment of the mixture of hydroxy acetals 13 and 14 with p-
toluenesulfonic acid in methanol.
The cyclic hemiacetal 15 was prepared by hydrolysis of the
hydroxy acetal 13 with diluted HCl (Scheme 2). Methane-
8048
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
a
Scheme 1. Preparation of Building Blocks 11 and 3a
Table 1. α-Arylation of Spirocyclic Thieno[3,2-c]pyran 9
with Various Iodoarenes
a
b
compd
aryl
C6H5
yield(GPC) (%)
yield(PTLC) (%)
4a
4b
4c
4d
57
61
66
67
46
38
53
43
p-MeOC6H4
p-CNC6H4
p-C6H5C6H4
a
b
Yield after gel permeation chromatography. Yield after preparative
thin-layer chromatography.
good yields. It should be noted that these arylation conditions
were well tolerated by the acid-labile methyl acetal and the
basic tertiary amine of 9.
The regioisomeric spirocyclic thiophenes 10 and 11 reacted
with various aryl iodides using the same catalytic system PdCl2/
2,2′-bipyridyl/Ag2CO3 to afford the α-arylated thiophenes 5
and 6 in good yields (Table 2). Again electron-rich (e.g.,
a
Reagents and reaction conditions: (a) n-BuLi, THF, −78 °C, 15 min,
then 1-benzylpiperidin-4-one, −78 °C, 3 h, 12% (13), 53% (14); (b)
n-BuLi, Et2O, 0 °C, 15 min, then 1-benzylpiperidin-4-one, 0 °C, 3 h,
33% (13), 17% (14); (c) p-TolSO3H, CH3OH, rt, 24 h, 54%; (d) p-
TolSO3H, CH3OH, rt, 24 h, 51%.
a
Scheme 2. Preparation of Building Block 10
Table 2. α-Arylation of Spirocyclic Thieno[2,3-c]pyrans 10
and 11 with Various Iodoarenes
a
Reagents and reaction conditions: (a) HCl, THF, rt, 18 h, 30%; (b)
CH3SO2Cl, NEt3, CH2Cl2, rt, 2 h, then reflux, 1 h, 67%; (c) H2, Pd/C,
CH3OH, rt, 18 h, 67%.
a
b
compd
X
aryl
C6H5
yield(GPC) (%) yield(PTLC) (%)
5a
6a
6b
6c
6d
6e
6f
H
64
38
51
69
61
38
55
42
63
97
33
40
19
23
52
32
29
49
36
48
55
14
sulfonyl chloride and triethylamine were used for the
elimination of water from lactol 15, providing the cyclic enol
ether 16, which was hydrogenated in the presence of the
catalyst Pd/C to give the thienopyran 10.
In 1990 Ohta and co-workers described for the first time the
Pd-catalyzed direct C−H bond arylation of thiophenes with
haloarenes.29 Since this pioneering work, various catalysts based
on Pd,30 Rh,31,32 Ir,33 and Cu34 have been developed for the
direct arylation of nonactivated thiophene derivatives with
haloarenes. These arylations typically occur in the α-position of
the thiophene moiety.
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
C6H5
p-MeOC6H4
p-MeC6H4
p-NO2C6H4
p-AcC6H4
p-CNC6H4
p-CF3OC6H4
1-naphthyl
p-C6H5C6H4
3-pyridyl
6g
6h
6i
6j
a
b
Yield after gel permeation chromatography. Yield after preparative
thin-layer chromatography.
In view of high α-regioselectivity, operational simplicity, and
broad substrate scope, we decided to use the catalytic system
PdCl2/2,2′-bipyridyl/Ag2CO3
35,36
for the selective α-arylation
of spirocyclic thiophenes 9−11. To learn more about the scope
of the catalytic system as well as the influence of the electron
density of the additional aryl moiety on the σ1 affinity, electron-
rich and electron-poor aryl iodides were considered for the
arylation step.
The results of the α-arylation of the spirocyclic thiophene 9
with different aryl iodides are summarized in Table 1. Both
donor (e.g., OCH3) and acceptor (e.g., CN) substituted aryl
iodides provided the corresponding α-arylated products 4b and
4c in yields similar to that of the α-arylated product provided
by the unsubstituted phenyl iodide (4a). Moreover, even the
sterically demanding biphenyl derivative 4d was obtained in
OCH3, 6b), electron-poor (e.g., NO2, 6d), and sterically
demanding (e.g., naphthyl, 6h) aryl iodides were employed in
the last arylation step. The tertiary amino moiety, the acetalic
group (10), and the ether group (11) were stable under these
reaction conditions.
In addition to α-arylation, the spirocyclic systems 10 and 11
should allow the introduction of aryl moieties in the β-position.
Only a few catalytic systems promoting selective β-arylation of
thiophenes have been reported.36−38 In this study the catalytic
system PdCl2/P[OCH(CF3)2]3/Ag2CO3, which is easy to
handle and shows high β-selectivity, was used.
8049
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
The reaction of the spirocyclic thiophenes 10 and 11 with
various aryl iodides led to a mixture of α-arylated (5, 6) and β-
arylated (7, 8) products (Table 3). Therefore, the yields of the
desired β-arylated spirocyclic thiophenes 7 and 8 were
considerably lower than the yields of the corresponding α-
substituted thiophenes 5 and 6 (compare Tables 2 and 3). In
the case of β-arylation only unsubstituted and donor-
substituted aryl iodides provided β-arylated products 7 and 8.
Ether (10), acetal (11), and tertiary amine did not inhibit the
arylation procedure and were not decomposed by the Pd
catalyst despite the high temperature of 150 °C.
Table 3. β-Arylation of Spirocyclic Thieno[2,3-c]pyrans 10
and 11 with Various Iodoarenes
At first, all compounds were purified by gel permeation
chromatography (GPC) to isolate the arylated products. To
obtain highly pure compounds (purity >95%) for receptor
binding studies, the arylated products were further purified by
flash chromatography (FC) or preparative thin-layer chroma-
tography (PTLC) before testing.
a
b
compd
X
aryl
C6H5
yield(GPC) (%) yield(PTLC) (%)
3. RECEPTOR AFFINITY
7a
8a
8b
8c
H
21
32
30
57
6
The σ1 and σ2 receptor affinities of the arylated spirocyclic
thiophenes 4−8, the synthetic precursors 9−11, and the lead
and reference compounds were determined in competition
experiments with radioligands. The potent and selective
tritium-labeled radioligand [3H]-(+)-pentazocine was used as
a competitor for the test compounds in the σ1 assay. Membrane
preparations of guinea pig brains served as the receptor
material. In the σ2 assay, membrane preparations of rat liver
OCH3
OCH3
OCH3
C6H5
13
15
24
p-MeOC6H4
p-MeC6H4
a
b
Yield after gel permeation chromatography. Yield after preparative
thin-layer chromatography.
Table 4. σ1 and σ2 Receptor Affinities of the Spirocyclic Thiophenes and Reference Compounds
Ki SEM (nM) (n = 3)
compd
X
aryl
CH3
σ1
σ2
selectivity, σ1/σ2
a
1a
OCH3
OCH3
H
21 2.3
1.5 0.08
0.35 0.06
4.5 2.9
0.22 0.06
1.0 0.4
5.5 1.5
111 27
89 72
>1 μM
>1 μM
230
>47
>660
657
>222
3664
>1000
96
a
1b
C6H5
b
2a
H
b
2b
H
C6H5
1 μM
806
b
3a
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
H
H
b
3b
C6H5
>1 μM
530
4a
4b
4c
4d
5a
6a
6b
6c
6d
6e
6f
C6H5
p-MeOC6H4
p-CNC6H4
p-C6H4C6H5
C6H5
>1 μM
>1 μM
>1 μM
283
>9
>11
>4
220 99
2.4 0.69
16 5.8
56 23
118
>63
>18
>34
>63
>1
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
H
C6H5
>1 μM
>1 μM
>1 μM
>1 μM
>1 μM
>1 μM
>1 μM
142
p-MeOC6H4
p-MeC6H4
p-NO2C6H4
p-AcC6H4
p-CNC6H4
p-CF3C6H4
1-naphthyl
p-C6H4C6H5
3-pyridyl
C6H5
29 1.0
16 0.9
59 15
20 6.7
43 18
>50
>23
12
6g
6h
6i
12 1.8
211 74
2.5 0.46
1.6 0.7
5.4 0.97
7.1 2.4
2.7 1.6
0.32 0.10
1.0 0.3
1.9 0.44
16 2.0
3.4 1.0
3.9 1.5
>1 μM
>1 μM
445
>5
6j
>400
278
>185
>141
>370
>3000
147
>500
53
7a
8a
8b
8c
OCH3
OCH3
OCH3
OCH3
H
C6H5
>1 μM
>1 μM
>1 μM
>1 μM
147
p-MeOC6H4
p-MeC6H4
H
c
9
10
H
11
OCH3
H
>1 μM
849
15
16
OH
C4C5
H
H
236
69
haloperidol
di-o-tolylguanidine
78 2.0
42 15
20
61
8
0.7
a
b
c
Ki values are reported in ref 21. Ki values are reported in ref 22. Ki value is reported in ref 27
8050
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
Chart 1. Chemical Structures of the 27 Training Set Compounds
8051
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
were the source for σ2 receptors and [3H]di-o-tolylguanidine
was employed as a radioligand. Since di-o-tolylguanidine also
binds to σ1 receptors, an excess of nonradiolabeled, σ1-selective
(+)-pentazocine was added to mask σ1 receptors.39−42
500 nM. This result indicates a very high selectivity of these σ1
ligands over the phencyclidine and ifenprodil binding sites of
the NMDA receptor.
4. MOLECULAR MODELING OF LIGAND/PROTEIN
INTERACTIONS
Table 4 clearly shows that the introduction of an aryl moiety
in position 2 of the spirocyclic compound 9 bearing the S atom
in the top position 1 led to a significant decrease of the σ1
receptor affinity. The Ki values of the α-arylated compounds
4a−4d are 17−500-fold higher than the Ki value of the parent
thiophene 9. The very low σ1 affinity of the biphenylyl
derivative 4d indicates that the biphenylyl residue is sterically
too demanding for the binding pocket of the σ1 receptor.
α-Arylation of the regioisomeric thiophene derivatives 10
and 11 led to slightly reduced σ1 affinities compared to the σ1
affinities of their parent compounds 10 and 11. However, some
of the arylated compounds show Ki values below 20 nM (e.g.,
6a, Ki = 16 nM, 6d, Ki = 16 nM, and 6h, Ki = 12 nM),
indicating considerable σ1 affinities. High σ1 affinities are
achieved with either an unsubstituted aryl residue (e.g., 6a
(phenyl), Ki = 16 nM, and 6h (naphthyl), Ki = 12 nM) or an
electron-deficient aryl residue in the 2-position (e.g., 6d
(nitrophenyl), Ki = 16 nM, and 6j (3-pyridyl), Ki = 2.5 nM).
As seen for series 4, spirocyclic thiophenes with an electron-rich
aryl moiety in the 2-position gave reduced σ1 affinities (e.g., 6b
(methoxyphenyl), Ki = 56 nM). The most promising σ1 ligands
of this series are 5a without an acetalic methoxy group (Ki = 2.4
nM) and the 3-pyridiyl derivative 6j (Ki = 2.5 nM) showing σ1
affinities similar to those of the parent compounds 10 and 11.
The highest σ1 affinities were found for the β-arylated
spirocyclic thiophenes 7 and 8. In this compound class, even
compounds with donor substituents (e.g., 8b (methoxyphenyl),
Ki = 7.1 nM; compare the Ki values of 4b and 6b) interact in
the low nanomolar range with σ1 receptors. Unfortunately, β-
arylation of 10 and 11 with acceptor-substituted aryl iodides
failed to give the arylated products.
Generally, the σ1 affinities decreased in the order 7/8 > 5/6
> 4. This tendency depends on the orientation of the aryl
moiety. In compounds 7 and 8, the phenyl ring adopts a
position at the top of the molecule, which is favorable for
receptor interaction. Due to the increased C−S bond length
and the larger size of the S atom, the spirocyclic thiophenes 5
and 6 with the S atom in the “bottom” position direct the aryl
moiety in a position similar to its position in 7/8, whereas in
the regioisomers 4 with the S atom in the top position the
additional aryl moiety is directed in an unfavorable position.
Except for the naphthyl derivative 6h and the unsubstituted
compound 10, the σ2 affinities of the (arylated) spirocyclic
thiophenes 4−11 are very low, indicating high selectivity for the
σ1 subtype over the σ2 subtype. Despite the moderate σ2 affinity
of 10 (Ki = 147 nM), its σ1/σ2 selectivity is still high (147) due
to the very high σ1 affinity. On the contrary, the naphthyl
derivative 6h shows the lowest σ1/σ2 selectivity (12) within this
series of compounds.
To investigate the role played by the position of the aromatic
moiety on the thiophene ring in determining the affinity of our
compounds toward the σ1 receptor, we applied a sequential
molecular modeling procedure based on the following steps: (i)
generation of a 3D pharmacophore model for the present series
of compounds, (ii) 3D pharmacophore-guided docking of all
ligands into the putative binding site of our 3D σ1 receptor
model, (iii) estimation of their binding affinity via the molecular
mechanics/Poisson−Boltzmann surface area (MM/PBSA)
methodology, and (iv) comparison with the available
experimental activities.
4.1. Pharmacophore Model Generation. A pharmaco-
phore represents the three-dimensional arrangements of
structural or chemical features of a drug that may be essential
for interacting with the receptor for optimal binding. These
pharmacophore models can be used in different ways in drug
design programs, e.g., (1) as 3D query tools in virtual screening
to identify potential new compounds from 3D databases of
druglike molecules with patentable structures different from
those already discovered, (2) to predict the activities of a set of
new compounds yet to be synthesized, or, as in the present
case, (3) to understand the possible mode/mechanism of drug/
receptor interaction.
3D pharmacophore hypothesis generation by the Catalyst
software is one of the popular approaches that has been
successfully used in drug discovery and toxicology research so
far. The most critical aspect of Catalyst pharmacophore
hypothesis generation is the selection of the molecular training
set; however, some basic guidelines have been elegantly laid out
by Li et al.,46 among which (1) a minimum of 16 diverse
compounds should be selected as a training set to avoid any
chance correlation, (2) the affinity data should have a range of
4−6 orders of magnitude, (3) the most active compounds
should be included so that they provide information on the
most critical features required as a pharmacophore, (4) all
biological data must be obtained by homogeneous procedures.
On the basis of the above criteria, a set of 10 pharmacophore
hypotheses were generated with Catalyst using 27 training set
compounds listed in Chart 1. The results of the hypotheses,
which include different cost values calculated during hypothesis
generation along with the corresponding root-mean-square
deviation (rmsd), correlation coefficient (ρ), and pharmaco-
phore features, are listed in Table 5.
The total hypothesis cost of these 10 best models varies
between 122.9 bits for the best ranked model (Hypo1) and
178.9 bits for the lowest ranked one (Hypo10). Such a confined
difference (56 bits) reflects both the homogeneity of the
generated hypotheses and the adequacy of the molecular
training set. The difference between the null and the fixed costs,
which should be higher than 70 bits to guarantee a robust
correlation, is 89.6 bits in our case. This corresponds to a
chance of true correlation in the data greater than 90%.47
Furthermore, in all the generated hypotheses the total costs are
much closer to the fixed cost (115.7 bits) than to the null cost
(205.3 bits), indicating that meaningful models are obtained.
Finally, the rmsd's and the correlation coefficients (ρ) between
estimated and experimental affinities range from 0.922 to 1.874
Since the antagonistic activity of similar spirocyclic
piperidines has been shown in animal assays,27,43 we assume
that the new thiophene-annulated spirocyclic compounds
presented herein also have antagonistic activity.
To prove the receptor selectivity, the affinities of the
spirocyclic piperidines 6c, 6f, 6h, 6j, 8c, 10, and 11 toward the
phencyclidine binding site44 and the ifenprodil binding site45 of
the NMDA receptor were recorded. It was shown that the
affinity of these very promising σ1 ligands at these receptor
systems is rather low; the Ki values are generally greater than
8052
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
The affinities of the 27 compounds for the σ1 receptor
estimated using Hypo1 are reported in Table 6, along with the
experimental values and the relevant errors (expressed as the
ratio between estimated and experimental values). Table 6
clearly shows that 21 out of 27 molecules in the training set
have errors equal to or less than 2.2, while the remaining 6 have
errors less than 5.0.
Table 5. Composition (Features), Costs (Bits), and
Statistical Parameters (rmsd and ρ) Associated with the 10
Best Hypotheses (Pharmacophore Models) Generated with
the Catalyst Software Using the Training Set Molecules
(Chart 1)
total
a
b
hypothesis
cost
Δcost
rmsd
ρ
features
Pharmacophore mapping of compounds belonging to series
3 and 8all bearing an aromatic ring in the top position with
respect to the tetrahydropyran ringindicates that these
molecules can map all five pharmacophore features quite nicely
(see Figure 4). Indeed, the presence of the N-benzylic group
provides one of the aromatic functions (HY) to the molecules,
the other HY being provided by the thiophene ring. The
piperidinic nitrogen atom is aptly positioned over the PI
feature, the pyran ring oxygen fulfills the HBA requirement,
and, last, the phenyl substituent overlays the HYAr feature
independently of its α-position (3b) or β-position (8a) with
respect to the heterocyclic sulfur atom. Referring to compounds
3b and 8a of Figure 4 as an example, the in silico estimated Ki
(σ1) values (Ki = 1.2 nM for 3b and Ki = 3.4 nM for 8a) are in
excellent agreement with the corresponding experimentally
determined Ki values (1.0 nM for 3b and 5.4 nM for 8a; see
Table 6).
1
2
3
4
122.9
131.5
136.4
137.8
82.6
73.8
68.9
67.5
0.922
1.305
1.319
1.377
0.905 HBA, HYAr, HY, HY, PI
0.839 HBA, HY, HY, HY, PI
0.832 HBA, HYAr, HY, HY, PI
0.813 HBA, HYAr, HYAr, HY,
PI
5
6
139.6
142.8
148.5
153.4
173.6
178.9
65.7
62.5
56.8
51.9
31.7
26.4
1.418
1.490
1.558
1.704
1.792
1.874
0.792 HBA, HY, HY, HY, PI
0.775 HBA, HYAr, HY, PI
0.760 HBA, HY, HY, PI
7
8
0.739 HBA, HYAr, HYAr, PI
0.701 HBA, HYAr, HY, PI
0.695 HBA, HYAr, HY, PI
9
10
a
Δcost = null cost − total cost. Null cost = 205.3, fixed cost = 115.7,
b
and configurational cost = 12.5. All costs are in units of bits. HBA =
hydrogen bond acceptor, HYAr = hydrophobic aromatic, HY =
hydrophobic, and PI = positive ionizable (see the text for more
details).
On the contrary, compounds belonging to series 4 and 6
featuring the aryl substituent in the left position are all
endowed with a lower affinity toward the σ1 receptor since all
explored molecular conformations failed to access the HYAr
feature, as exemplified for compounds 4a and 6a in Figure 5.
The HYAr pharmacophore interaction point is, however, the
only feature unmapped by compounds of series 4 and 6. Thus,
given the excellent overlay between the molecular chemical
moieties and the remaining 3D pharmacophore requirements
shown in Figure 5, only a small decrease in σ1 affinity for these
two molecular series is predicted for, e.g., compounds 4a and
and from 0.905 to 0.695, respectively. As all the generated
pharmacophores map the molecules of the training set in a
similar way, the first model (Hypo1), characterized by the
highest cost difference, the lowest rmsd, and the best ρ value,
was selected for further analysis.
In analogy with other pharmacophore models formerly
developed for σ1 ligands,48−51 the main features of the present
Hypo1 are a positive ionizable group (PI), a hydrogen bond
acceptor group (HBA), two hydrophobic moieties (HY), and
one hydrophobic aromatic point (HYAr) (see Figure 3).
Figure 3. Geometrical relationships (A, B) among the features of the top-scoring pharmacophore Hypo1 (C). The hypothesis features are portrayed
as meshed spheres, color-coded as follows: red, PI; light blue, HYAr; pink, HY; light green, HBA. HBA is actually represented as a pair of spheres
(the smaller sphere represents the location of the HBA atom on the ligand and the larger one the location of an HB donor on the receptor). Selected
distances (Å) and angles (deg) are labeled.
8053
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
moieties are endowed with comparable (if not higher) σ1
affinities with respect to the unsubstituted counterparts.
In this respect, the analysis of the top panels of Figure 6
indeed reveals how, for compounds 3e and 8b bearing the p-
methoxyphenyl substituent in the top position, all five 3D
pharmacophore features are mapped very well (as in the case of
their unsubstituted analogues; see Figure 4), resulting in high
predicted/experimental σ1 affinity values (Ki(estimated) = 3.1
nM/Ki(exptl) = 2.2 nM for 3e and Ki(estimated) = 2.2 nM/
Ki(exptl) = 7.1 nM for 8b; see Table 6).
On the other hand, not only do all derivatives characterized
by the presence of the aromatic ring in the left position miss the
HYAr feature as discussed above, but the presence of the p-
OCH3 substituent on the phenyl ring induces a further
conformational change so that all four remaining pharmaco-
phoric features are mapped less precisely with respect to the
unsubstituted compounds (compare the lower panels in Figures
5 and 6). Accordingly, for this set of compounds the model
predicts lower σ1 affinities for 4b (Ki = 50 nM) and 6b (Ki = 67
nM) in line with the experimental σ1 affinities (Ki = 111 nM for
4b and Ki = 56 nM for 6b; see Table 6).
4.2. Pharmacophore Assessment: Cross-Validation
Study. The quality of the developed pharmacophore was
assessed using the CatScramble technique available in Catalyst.
The purpose of this technique is to randomize the affinity data
among the training set compounds and to generate
pharmacophore hypotheses using the same features and
parameters used to develop the original pharmacophore
hypothesis. If the randomized sets generate pharmacophores
with similar or better cost values, rmsd's, and correlations, then
the original pharmacophore can be considered as generated by
chance. The results of the CatScramble runs are listed in Table
7. Since a 98% confidence level was selected for this test, 49
random hypothesis runs were performed. The results clearly
indicate that randomization produced hypotheses with no
predictive values similar or close to those of the corresponding
Hypo1. Indeed, none of the outcome hypotheses had a lower
cost score, better correlation, or smaller root-mean-square
deviation than the initial hypothesis. Table 7 lists the first 10
lowest total score values of the resulting 49 hypotheses for our
training set molecules. In conclusion, there is a 98% chance for
the best hypothesis to represent a true correlation in the
training set affinity data for the present classes of compounds.
Finally, a further statistical test, the leave-one-out method,
which consists of recomputing the hypothesis by excluding
from the training set one molecule at a time, was carried out.
Basically, this test is performed to verify whether the correlation
is strongly dependent on one particular compound in the
Table 6. Experimentally Determined and Estimated Ki
Values of the Training Set Compounds Calculated on the
Basis of Pharmacophore Hypothesis 1 (Hypo1)
σ1 affinity, Ki
(nM)
compd of
compd
a
this paper of ref 22
aryl
C6H5
X
estimated exptl error
2a
2b
2c
2d
3b
3e
H
H
H
H
1.8
2.7
2.8
6.9
1.2
3.1
0.35
4.5
3.6
4.0
1.0
2.2
5.0
−1.6
−1.3
1.7
p-MeC6H5
1-napthyl
C6H5
−OCH3
−OCH3
1.2
p-
1.4
MeOC6H5
3g
p-AcC6H5
p-CNC6H5
3-pyridyl
C6H5
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
3.5
0.51
1.9
13
1.6
2.2
2.0
3h
0.25
2.2
3i
−1.1
2.2
4a
4b
5.5
p-
50
111
−2.2
MeOC6H5
4c
4d
11
6a
6b
p-CNC6H5
p-biphenyl
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
60
51
5.7
23
67
89
−1.5
−4.3
3.0
220
1.9
16
C6H5
1.4
p-
56
1.2
MeOC6H5
6c
6d
6e
6f
p-MeC6H5
−OCH3
27
51
34
37
21
22
45
1.8
3.4
2.2
29
−1.1
3.2
p-NO2C6H5 −OCH3
p-AcC6H5
16
−OCH3
−OCH3
59
−1.7
1.8
p-CNC6H5
20
6g
6h
6i
p-CF3C6H5 −OCH3
43
−2.0
1.8
1-naphthyl
p-biphenyl
C6H5
−OCH3
−OCH3
H
12
211
1.6
5.4
7.1
−4.7
1.1
7a
8a
8b
C6H5
−OCH3
−OCH3
−1.6
−3.3
p-
MeOC6H5
8c
p-MeC6H5
−OCH3
3.5
2.7
1.3
a
Values in the error column represent the ratio of the estimated
affinity to experimental affinity, or its negative inverse if the ratio is less
than 1.
6a (Ki = 13 nM for 4a and Ki = 23 nM for 6a), in agreement
with the experimental binding constants (Ki = 5.5 nM for 4a
and Ki = 16 nM for 6a; see Table 6).
Importantly, however, the affinity of compounds 4 and 6 is
strongly affected by the presence of a substituent at the para
position of the thiophene-linked phenyl ring. On the contrary,
all derivatives 3 and 8 bearing the same substituted aromatic
Figure 4. Mapping of compounds 3b and 8a onto the developed σ1 3D pharmacophore model (Figure 3). Compounds are portrayed as atom-
colored sticks (red, O; yellow, S; gray, C; blue, N; white, H).
8054
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
Figure 5. Mapping of compounds 4a and 6a onto the developed σ1 3D pharmacophore model (Figure 3). Compounds are portrayed as atom-
colored sticks (red, O; yellow, S; gray, C; blue, N; white, H).
Figure 6. Mapping of compounds 3e, 8b, 4b, and 6b onto the developed σ1 3D pharmacophore model (Figure 3). Compounds are portrayed as
atom-colored sticks (red, O; yellow, S; gray, C; blue, N; white, H).
assessment of the statistical test. For each of the 27 new
hypotheses generated according to this method, we did not
obtain meaningful differences between Hypo1 and each
hypothesis resulting from the exclusion of one compound at
a time.
4.3. Pharmacophore Validation Using Test Set
Compounds. A critical step in automated pharmacophore
generation is model validation, especially in those cases where
the model has been generated for the purpose of either
predicting the affinity of external sets of compounds or
estimating the affinity of newly conceived molecular entities
prior to their synthesis. Thus, to check the robustness of our
correlation, we used the pharmacophore model (Hypo1) to
predict the affinity of a further set of molecules, also called the
test set, composed of 13 additional molecules from our series
(see Chart 2).
Table 7. Output Parameters of the 10 Lowest Cost
Hypotheses Resulting from the Statistical Evaluation
According to the CatScramble Validation Procedure
hypothesis
ρ
rmsd
total cost (bits)
1
0.812
0.717
0.700
0.724
0.622
0.603
0.576
0.558
0.515
0.507
0.905
1.154
1.389
1.397
1.453
1.496
1.601
1.657
1.708
1.755
1.823
0.922
142.9
151.9
154.0
155.1
157.1
160.6
163.8
165.7
167.8
168.9
122.9
2
3
4
5
6
7
8
9
10
Hypo1
Interestingly, a good correlation coefficient of 0.80 was
observed when a regression analysis was performed by mapping
the test set onto the features of the best pharmacophore
hypothesis (Hypo1). The predicted and the experimental
Ki(σ1) values for the test set along with the respective errors are
shown in Table 8. The average error in predicting the affinity of
training set. The test is positive if the affinity of each excluded
molecule is correctly predicted by the corresponding one-
missing hypothesis. The value of ρ, the feature composition of
the pharmacophore, and the quality of the predicted affinity of
the excluded molecule were used as measures for the
8055
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
Chart 2. Chemical Structures of the 13 Test Set Compounds
4.4. Docking of Ligands into the σ1 Receptor 3D
Model and Affinity Scoring via MM/PBSA Calculations.
Following our pioneer work,50 the putative binding site and
binding modes of all compounds considered in this work in the
σ1 receptor 3D homology model structure were retrieved by
exploiting the currently available preliminary information on
sequence−structure relationships and mutagenesis studies52
and the ligand-binding pharmacophore requirements presented
above.50,51 To summarize briefly, a protein isoform missing
residues 119−149 was found devoid of ligand-binding capacity,
and the conversion of residues Asp126 and Glu172 into glycine
led to a several-fold reduction in ligand binding of the σ1
receptor.52 Moreover, our hydrophobicity analysis50 identified,
aside for the TM regions, a third hydrophobic region matching
the SBDLII (steroid binding domain like II) region and
centered on Asp188, a residue specifically photolabeled by
Table 8. Experimentally Determined and Estimated Ki
Values of the Test Set Compounds Calculated on the Basis
of Pharmacophore Hypothesis 1 (Hypo1)
σ1 affinity: Ki
(nM)
compd of
compd
this paper of ref 22
aryl
X
estimated exptl
error
3a
2e
−OCH3
H
1.8
1.6
0.22
1.5
7.7
1.1
p-
MeOC6H5
2f
p-NO2C6H5
p-CNC6H5
p-MeC6H5
H
3.4
2.7
1.1
3.6
2.1
5.8
18
1.7
3.4
2.0
1.0
5.7
5.0
30
2.0
−1.3
−1.9
3.6
2g
H
3c
−OCH3
3f
p-NO2C6H5 −OCH3
p-CF3C6H5 −OCH3
1-naphthyl
3j
−2.7
1.2
3d
−OCH3
−OCH3
−OCH3
H
[
125I]IACoc (3-iodo-4-azidococaine).53,54 Having localized this
3k
p-biphenyl
−1.7
6.3
9
2.0
3.2
11
0.32
1.0
2.4
2.5
protein region as a possible zone for ligand binding, a thorough
search for a sequence satisfying the chemical features imposed
by Hypo1 was performed, and the sequence was successfully
retrieved. Thus, all compounds were docked into the putative
binding site of the receptor 3D model. In the set of docked
ligand conformations, for each compound a solution was found
that best reproduced the key 3D pharmacophore requirements.
Taking compound 3h as a prototypical example of the class
of compounds featuring the phenyl ring in the top position,
from the top panel of Figure 7, we can observe the striking
10
5a
6j
3.2
C6H5
H
4.6
3-pyridyl
−OCH3
5.4
2.2
the test set molecules is 3.0. Given the inherent simplicity of
the pharmacophoric approach, and considering the intrinsic
variability of the biological responses, we can conclude that the
ability of the present 3D pharmacophore model to predict the
affinity of these series of σ1 receptor ligands is quite satisfactory.
8056
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
Figure 7. (Top) Comparison between the best conformation assumed
by compound 3h upon mapping the 3D pharmacophore model
(purple sticks) and after docking in the putative σ1 binding site (green
sticks and balls). (Bottom) Equilibrated MD snapshots of the σ1
receptor in complex with 3h. The image is a zoomed view of the
receptor binding site. The ligand is portrayed in green sticks and balls,
while the protein residues mainly involved in the interaction with 3h
are highlighted as colored sticks and labeled. Salt bridge and H-bond
interactions are shown as black lines. Water, ions, and counterions are
not shown for clarity.
Figure 8. (Top) Comparison between the best conformation assumed
by compound 6a upon mapping the 3D pharmacophore model
(purple sticks) and after docking in the putative σ1 binding site (green
sticks and balls). (Bottom) Equilibrated MD snapshots of the σ1
receptor in complex with 6a. The image is a zoomed view of the
receptor binding site. The ligand is portrayed in green sticks and balls,
while the protein residues mainly involved in the interaction with 6a
are highlighted as colored sticks and labeled. Salt bridge and H-bond
interactions are shown as black lines. Water, ions, and counterions are
not shown for clarity.
similarity between the molecular conformation adopted by the
compound in mapping the 3D pharmacophore model and that
assumed by the same compound in the σ1 pocket upon binding.
Most importantly, however, the best binding mode detected for
3h after docking satisfies all ligand/protein intermolecular
interactions required by pharmacophore hypothesis Hypo1. In
detail, (i) the piperidinic nitrogen is engaged in a salt bridge
with Asp126 (PI feature), (ii) the pyran oxygen atom H-bonds
with the −NH group of the backbone peptidic linkage between
Thr151 and Val152 (HBA feature), (iii) the aryl group linked
to the thiophene ring is involved in stabilizing interactions with
the aromatic components of Trp121 and Tyr120 (HYAr
feature), and (iv) a network of hydrophobic interactions
stabilizes the position of the thiophene and the N-benzylic rings
among the residues Tyr173, Ile128, and Phe133 (two HY
features).
Considering compound 6a as a representative example of the
derivatives featuring the phenyl ring in the left position, from
the top panel of Figure 8 we can appreciate that, also in this
case, the best molecular conformation mapping the 3D
pharmacophore Hypo1 and the best binding mode of 6a into
the σ1 binding pocket are superimposable. Quite surprisingly,
however, the detailed inspection of the binding pose of 6a
reveals that, to comply with the pharmacophoric requirements
upon protein binding, this molecule must adopt a “reverse”
orientation with respect to that assumed by, e.g., 3h, as shown
in the bottom panel of Figure 8. Accordingly, the basic nitrogen
of 6a creates a salt bridge with Glu123 (PI feature), the oxygen
atom of the pyran ring is engaged in a H-bond with the side
chain −OH group of Thr131 (HBA feature), and the thiophene
ring is stabilized by hydrophobic interactions with the side
chains of both Tyr173 and E172 (HY feature). Notably, at
variance with 3h, the remaining hydrophobic pharmacophore
requirement is fulfilled by the thiophene-linked phenyl ring,
which aptly fits a small hydrophobic cavity lined by the side
chains of Ile128 and Phe133.
All 40 receptor/drug complexes obtained from the
pharmacophore-based docking procedure were then relaxed
by energy minimization and molecular dynamics (MD)
simulations, and finally, the relevant values of the free energy
of binding, ΔGbind, between all compounds and the σ1 receptor
were evaluated by applying the well-known MM/PBSA
computational ansatz,55,56 as listed in Table 9.
A plot of the ΔGbind values calculated from simulations vs
those obtained experimentally using the corresponding Ki(σ1)
(ΔGbind(exptl), last column in Table 9) shows a very good
correlation coefficient (R2 = 0.81; see Figure 9). Although this
might be a somewhat expected result, given the structural
similarity of the compounds considered, it is interesting to
observe how MM/PBSA is able not only to rank all molecules
in the correct order with respect to their affinity toward the σ1
receptor but also to discriminate between high-affinity (e.g.,
ΔGbind values ranging from approximately −12 to >−13 kcal/
mol), intermediate-affinity (−10.5 kcal/mol < ΔGbind < −12
kcal/mol), and low-affinity (ΔGbind < −10.5 kcal/mol)
compounds. The most notable result, however, is the fact
that these calculations do support the reverse binding mode of
the molecules with the phenyl substituent in the left position
with respect to those featuring the same group in the top
position. In fact, if we consider the MD snapshot of the ligand
3h/σ1 receptor complex shown in Figure 10A, we can observe
how the main interactions detected after docking are totally
preserved along the MD trajectory: a permanent salt bridge
(average dynamic distance (ADL) = 4.03
0.12 Å) is
8057
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
a
Table 9. Experimentally Determined (ΔGbind(exptl)) and Calculated (ΔGbind) Free Energy of Ligand/σ1 Receptor Complexes
compd of this paper compd of ref 22
aryl
X
ΔH (kcal/mol)
−TΔS (kcal/mol) ΔGbind (kcal/mol) ΔGbind(exptl) (kcal/mol)
2a
H
−36.66 (0.18)
−36.90 (0.19)
−37.40 (0.18)
−38.36 (0.21)
−38.50 (0.19)
−38.70 (0.20)
−38.63 (0.20)
−39.06 (0.17)
−38.52 (0.22)
−38.71 (0.18)
−38.51 (0.20)
−38.58 (0.18)
−38.43 (0.17)
−39.11 (0.16)
−38.13 (0.19)
−38.38 (0.20)
−38.67 (0.21)
−38.08 (0.20)
−36.82 (0.20)
−36.31 (0.18)
−34.83 (0.19)
−34.78 (0.18)
−34.71 (0.22)
−36.61 (0.21)
−36.78 (0.18)
−36.07 (0.19)
−35.80 (0.19)
−35.62 (0.20)
−35.78 (0.17)
−35.61 (0.17)
−35.50 (0.17)
−35.67 (0.21)
−35.49 (0.19)
−35.72 (0.21)
−35.13 (0.20)
−36.08 (0.22)
−38.46 (0.18)
−38.49 (0.18)
−38.51 (0.21)
−38.69 (0.20)
23.56 (0.37)
23.67 (0.39)
25.76 (0.42)
25.83 (0.39)
25.69 (0.38)
25.91 (0.38)
26.02 (0.40)
26.18 (0.41)
25.81 (0.37)
25.77 (0.39)
25.79 (0.39)
25.99 (0.40)
25.56 (0.40)
25.82 (0.40)
25.64 (0.41)
26.03 (0.37)
25.98 (0.36)
26.30 (0.38)
23.61 (0.37)
24.48 (0.39)
24.82 (0.41)
24.79 (0.42)
24.95 (0.36)
23.60 (0.36)
23.71 (0.41)
24.53 (0.39)
24.74 (0.39)
24.68 (0.40)
24.61 (0.38)
24.59 (0.37)
24.67 (0.41)
24.71 (0.38)
24.70 (0.39)
24.81 (0.40)
24.98 (0.40)
24.82 (0.36)
25.73 (0.38)
25.81 (0.38)
25.91 (0.38)
25.88 (0.41)
−13.10 (0.41)
−13.23 (0.43)
−11.64 (0.45)
−12.53 (0.44)
−12.81 (0.42)
−12.79 (0.43)
−12.61 (0.45)
−12.88 (0.44)
−12.71 (0.43)
−12.94 (0.43)
−12.72 (0.44)
−12.59 (0.45)
−12.87 (0.43)
−13.29 (0.42)
−12.49 (0.43)
−12.35 (0.42)
−12.69 (0.41)
−11.78 (0.43)
−13.21 (0.42)
−11.83 (0.43)
−10.01 (0.43)
−9.99 (0.44)
−9.76 (0.42)
−13.01 (0.41)
−13.07 (0.44)
−11.54 (0.43)
−11.06 (0.43)
−10.94 (0.45)
−11.17 (0.41)
−11.02 (0.40)
−10.83 (0.42)
−10.96 (0.43)
−10.79 (0.43)
−10.91 (0.45)
−10.15 (0.45)
−11.26 (0.42)
−12.73 (0.42)
−12.68 (0.42)
−12.60 (0.43)
−12.81 (0.46)
−12.91
−13.18
−11.39
−12.04
−11.52
−11.97
−11.56
−11.46
−12.28
−11.82
−11.87
−12.28
−12.00
−13.10
−11.25
−11.33
−11.82
−10.27
−12.96
−11.27
−9.49
3a
−OCH3
H
2b
C6H5
2e
p-MeOC6H5
p-MeC6H5
p-NO2C6H5
p-CNC6H5
1-napthyl
H
2c
H
2f
H
2g
H
2d
H
3b
C6H5
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
H
3e
p-MeOC6H5
p-MeC6H5
p-NO2C6H5
p-AcC6H5
p-CNC6H5
p-CF3C6H5
1-naphthyl
3-pyridyl
3c
3f
3g
3h
3j
3d
3i
3k
p-biphenyl
9
4a
4b
4c
4d
10
11
5a
6a
6b
6c
6d
6e
6f
C6H5
p-MeOC6H5
p-CNC6H5
p-biphenyl
−9.62
−9.09
−12.28
−11.90
−11.76
−10.64
−9.90
−10.29
−10.64
−9.87
−10.51
−10.05
−10.81
−9.11
−11.74
−12.00
−11.28
−11.12
−11.69
−OCH3
H
C6H5
C6H5
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
−OCH3
H
p-MeOC6H5
p-MeC6H5
p-NO2C6H5
p-AcC6H5
p-CNC6H5
p-CF3C6H5
1-naphthyl
p-biphenyl
3-pyridyl
C6H5
6g
6h
6i
6j
7a
8a
8b
8c
C6H5
−OCH3
−OCH3
−OCH3
p-MeOC6H5
p-MeC6H5
a
ΔGbind = ΔH − TΔS. ΔH = enthalpic component of ΔGbind. −TΔS = entropic component of ΔGbind. ΔGbind(exptl) was estimated from the
corresponding available Ki(σ1) values using the following relationship: ΔGbind(exptl) = −RT ln(1/Ki(σ1)).
by an ADL of 1.93 0.18 Å, is detected between the backbone
NH group of Thr151 and Val152 and the O atom of the pyran
ring of 3h, a number of stabilizing π−π interactions take place,
involving the side chains of Trp121 and Tyr173 and the ligand
benzonitrile and thiophene rings, respectively, and last, the N-
benzyl group of 3h remains nicely encased within the
hydrophobic pocket generated by residues Ile128 and
Phe133. These plethora of favorable, stabilizing interactions
result in a calculated value of ΔGbind of −13.29 kcal/mol, in
excellent agreement with the corresponding experimental value
(−13.10 kcal/mol; see Table 9), thereby confirming the
Figure 9. Correlation displaying the experimental vs estimated free
ranking of 3h as one of the most potent σ1 ligands of the
energy of binding for all 40 ligand/σ1 receptor complexes considered
in this work.
entire series.
In the case of compound 6a, a template molecule
representative of all derivatives bearing the phenyl ring in the
left position, the protein/ligand interactions depicted at the end
of the pharmacophore-assisted docking phase are again
established between the pyperidine NH+ group and the
−COO− moiety of Asp126, a persistent H-bond, characterized
8058
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
correlation between the experimentally determined free energy
of the ligand binding (calculated from the Ki value) and the
estimated free energy from the model. The most striking result
of these calculations is the reverse binding mode of ligands with
the phenyl moiety in the top position compared with ligands
bearing the aryl moiety in the left position. Moreover, the
contribution of single amino acid residues to the overall binding
of both types of ligands was analyzed using the docking
experiments.
6. EXPERIMENTAL SECTION
Chemistry. General Procedures. Silica gel 60 F254 plates (Merck)
were used for thin-layer chromatography (TLC). 1H NMR (400, 300,
and 600 MHz) and 13C NMR (100 MHz) spectra were recorded on a
Unity Mercury Plus 400 (400 MHz) NMR spectrometer (Varian), a
JNM-ECA-400 (400 MHz) spectrometer (JEOL), a Bruker AV 300
(300 MHz) spectrometer, and a Varian Unity Plus 600 (600 MHz)
spectrometer operating at 23 °C; chemical shifts δ are reported in
parts per million (ppm) against the reference compound tetrame-
thylsilane and calculated using the chemical shift of the signal of the
residual nondeuterated solvent. Finnigan MAT 4200s, Bruker
Daltonics Micro Tof, and Waters Micromass Quatro LCZ
spectrometers were used for HRMS (ESI); peaks are given in m/z
(percentage of the base peak). The purity of all test compounds was
greater than 95%, which was determined by HPLC method A or B
(see the Supporting Information).
General Procedure A for the α-Arylation of Spirocyclic
Thiophenes with Iodoarenes. A 20 mL glass vessel was equipped
with a magnetic stirring bar and closed by a J. Young O-ring tap. The
flask was flame-dried under vacuum and filled with Ar after being
cooled to rt. Under a permanent stream of Ar, the catalyst PdCl2/bipy
(10 mol %) and Ag2CO3 (1 equiv) were filled into the vessel and
suspended in dry m-xylene (0.4 mL). This mixture was stirred at 60 °C
for 30 min. Finally, a solution of the iodoarene (1.1 equiv) and a
solution of the spirocyclic starting material (1 equiv) in dry m-xylene
(0.6 mL in total) were added dropwise. The vessel was sealed with the
O-ring tap and heated at 150 °C for 12 h in an eight-well reaction
block. After the vessel was cooled to rt, the mixture was filtered
through a short silica pad (EtOAc). The filtrate was concentrated in
vacuo, and the crude product was purified by gel permeation
chromatography (CHCl3) followed by preparative thin-layer chroma-
tography or flash chromatography (short column) to yield the
corresponding arylthiophene in high purity.
General Procedure B for the β-Arylation of Spirocyclic
Thiophenes with Iodoarenes. A 20 mL glass vessel was equipped
with a magnetic stirring bar and closed by a J. Young O-ring tap. The
flask was flame-dried under vacuum and filled with Ar after being
cooled to rt. Under a permanent stream of Ar, PdCl2 (10 mol %) and
Ag2CO3 (1 equiv) were filled into the vessel. The phosphite ligand
P[OCH(CF3)2]3 (20 mol %) and subsequently dry m-xylene (0.5 mL)
were added, and this mixture was stirred at 60 °C for 30 min to form
the active catalyst. Finally, a solution of the iodoarene (1.1 equiv) and
a solution of the spirocyclic starting material (1 equiv) in dry m-xylene
(0.5 mL in total) were added. The vessel was sealed with the O-ring
tap and heated at 150 °C for 12 h in an eight-well reaction block. After
the vessel was cooled to rt, the mixture was filtered through a short
silica pad (EtOAc). The filtrate was concentrated in vacuo, and the
crude product was purified by gel permeation chromatography
(CHCl3) followed by preparative thin-layer chromatography or flash
chromatography (short column) to yield the corresponding
arylthiophene in high purity.
Figure 10. Equilibrated MD snapshots of 3h (A) and 6a (B) in
complex with the σ1 receptor. The images are zoomed views of the
receptor binding site. The protein structure is depicted as a transparent
gray ribbon, while both ligands are shown in atom-colored sticks and
balls (C, gray; O, red; S, yellow; N, blue; H atoms are omitted). The
protein residues mainly involved in the interaction with the ligands are
highlighted as colored sticks and labeled. Salt bridge and H-bond
interactions are shown as black lines. Some water molecules, ions, and
counterions are displayed as atom-colored spheres (O, red; H, white;
Na+, purple; Cl−, green).
detected and preserved during the entire MD simulation. In
detail, a stable H-bond (ADL = 1.88
0.15 Å) is found
between the side chain −OH group of Thr181 and the pyran
ring oxygen of 6a, while a salt bridge (ADL = 3.98 0.22 Å)
between the CO2− group of Glu123 and the piperidine NH+ of
6a is also evidence for the whole simulation period. Favorable
hydrophobic interactions are generated between the phenyl
substituent and the side chains of Ile128 and Phe133, and last,
the thiophene moiety of 6a is engaged in a positive π−π
interaction with the aromatic side chain of Tyr173 and is
further stabilized by the influence of the vicinal alkyl chain of
Glu172.
5. CONCLUSION
The introduction of aryl moieties in the α- and β-positions of
complex spirocyclic σ1 ligands was performed by regioselective
direct C−H bond arylation using the catalyst systems PdCl2/
2,2′-bipyridyl/Ag2CO3 and PdCl2/P[OCH(CF3)2]3/Ag2CO3,
respectively. This late-stage functionalization allows the syn-
thesis of a large library of diversely substituted spirocyclic σ1
ligands, which represent the basis for the development of a
pharmacophore model. The main features of the pharmaco-
phore are a positive ionizable group, a hydrogen bond acceptor
group, two hydrophobic moieties, and one hydrophobic
aromatic point. Validation using test set compounds confirms
the high quality of the pharmacophore model. The interactions
of the spirocyclic σ1 ligands with the receptor protein were
analyzed by docking into a 3D homology model of the σ1
receptor. These docking experiments led to a very good
1-Benzyl-6′-methoxy-2′-phenyl-6′,7′-dihydrospiro[piperidine-
4,4′-thieno[3,2-c]pyran] (4a). According to general procedure A, the
spirocyclic thiophene 9 (33.2 mg, 0.101 mmol) was reacted with
iodobenzene (12.4 μL, 0.11 mmol), Ag2CO3 (27.5 mg, 0.01 mmol),
and PdCl2/bipy (3.2 mg, 0.01 mmol) in m-xylene (1.2 mL). The crude
product was purified by CHCl3 GPC and FC (1.5 cm, h = 5 cm,
hexane:EtOAc = 4:1, 3 mL, Rf = 0.10): colorless solid; mp 137 °C;
1
yield 23.4 mg (57%) after GPC; yield 18.9 mg (46%) after FC; H
8059
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
NMR (CDCl3) δ (ppm) = 1.84−2.01 (m, 3H, N(CH2CH2)2), 2.15
(td, J = 13.2/4.4 Hz, 1H, N(CH2CH2)2), 2.45 (td, J = 11.6/3.3 Hz,
1H, N(CH2CH2)2), 2.54 (td, J = 12.4/2.5 Hz, 1H, N(CH2CH2)2),
2.74−2.81 (m, 2H, N(CH2CH2)2), 2.86 (dd, J = 15.8/7.2 Hz, 1H,
thioph CH2CH), 2.99 (dd, J = 15.8/3.3 Hz, 1H, thioph CH2CH), 3.57
(s, 3H, OCH3), 3.57 (d, J = 13.0 Hz, 1H, NCH2Ph), 3.61 (d, J = 13.1
Hz, 1H, NCH2Ph), 4.91 (dd, J = 7.2/3.3 Hz, 1H, thioph CH2CH),
7.02 (s, 1H, 3′-H-thioph), 7.27−7.41 (m, 8H, Ph H), 7.49 − 7.54 (m,
2H, Ph H).
1-Benzyl-4′,5′-dihydrospiro[piperidine-4,7′-thieno[2,3-c]pyran]
(10). The enol ether 16 (106.8 mg, 0.36 mmol) was dissolved in dry
MeOH (10 mL), and 10% Pd/C (11 mg) was added. The suspension
was stirred under a H2 atmosphere (balloon) at rt for 18 h. Afterward
the catalyst was filtered off, and the remaining residue was washed first
with 2 M HCl and then with water. The filtrate was alkalized with 2 M
NaOH and extracted twice with CH2Cl2. The organic layers were
separated, combined, and dried over K2CO3. The solvent was removed
in vacuo, and the crude product was purified by FC (3.5 cm, h = 15
cm, cyclohexane:EtOAc = 9:1, 2% NEt3, 10 mL, Rf = 0.27): colorless
1-Benzyl-2′-phenyl-4′,5′-dihydrospiro[piperidine-4,7′-thieno[2,3-
c]pyran] (5a). According to general procedure A, the spirocyclic
thiophene 10 (29.4 mg, 0.098 mmol) was reacted with iodobenzene
(12.1 μL, 0.11 mmol), Ag2CO3 (29.2 mg, 0.11 mmol), and PdCl2/bipy
(3.4 mg, 0.01 mmol) in m-xylene (1.2 mL). The crude product was
purified by CHCl3 GPC and FC (1.5 cm, h = 5 cm, hexane:EtOAc =
3:2, 3 mL, Rf = 0.52): colorless resin; yield 23.6 mg (64%) after GPC;
1
solid; mp 65 °C; yield 72 mg (67%); H NMR (CDCl3) δ (ppm) =
1.90−2.05 (m, 4H, N(CH2CH2)2), 2.37−2.50 (m, 2H, N(CH2CH2)2),
2.69 (t, 5.5 Hz, 2H, thioph CH2CH2), 2.72−2.80 (m, 2H,
N(CH2CH2)2), 3.58 (s, 2H, NCH2Ph), 3.91 (t, J = 5.5 Hz, 2H,
thioph CH2CH2), 6.76 (d, 5.0 Hz, 1H, 3′-H-thioph), 7.13 (d, 5.0 Hz,
1H, 2′-H-thioph), 7.27−7.38 (m, 5H, Ph H).
1
1-Benzyl-5′-methoxy-4′,5′-dihydrospiro[piperidine-4,7′-thieno-
[2,3-c]pyran] (11). The hydroxy acetal 13 (2.1 g, 5.81 mmol) was
dissolved in dry MeOH (50 mL), and p-toluenesulfonic acid (1.25 g,
7.24 mmol) was added. The solution was stirred for 24 h at rt. Then 2
M NaOH was added until pH 8, and the solution was diluted with
water (50 mL). The aqueous layer was extracted twice with CH2Cl2.
The combined organic layers were dried (K2CO3) and filtered, and the
solvent was removed in vacuo. The remaining oil was purified by FC
(6 cm, h = 15 cm, cyclohexane:EtOAc = 4:1, 30 mL, Rf = 0.30):
yield 14.5 mg (40%) after FC; H NMR (CDCl3) δ (ppm) = 1.94−
2.09 (m, 4H, N(CH2CH2)2), 2.43 (td, J = 11.3/4.1 Hz, 2H,
N(CH2CH2)2), 2.69 (t, J = 5.5 Hz, 2H, thioph CH2CH2), 2.74 (d, J
= 11.6 Hz, 2H, N(CH2CH2)2), 3.57 (s, 2H, NCH2Ph), 3.94 (t, J = 5.5
Hz, 2H, thioph CH2CH2), 6.96 (s, 1H, 3′-H-thioph), 7.22−7.25 (m,
1H, Ph H), 7.27−7.29 (m, 1H, Ph H), 7.29−7.41 (m, 6H, Ph H),
7.51−7.56 (m, 2H, Ph H).
1-Benzyl-5′-methoxy-2′-phenyl-4′,5′-dihydrospiro[piperidine-
4,7′-thieno[2,3-c]pyran] (6a). According to general procedure A, the
spirocyclic thiophene 11 (19.5 mg, 0.060 mmol) was reacted with
iodobenzene (7.5 μL, 0.07 mmol), Ag2CO3 (17.3 mg, 0.06 mmol), and
PdCl2/bipy (2.1 mg, 0.006 mmol) in m-xylene (1.0 mL). The crude
product was purified by CHCl3 GPC and preparative TLC (h = 15 cm,
hexane:EtOAc = 3:2, Rf = 0.5): colorless solid; yield 9.2 mg (38%)
1
colorless solid; mp 77 °C; yield 1.04 g (54%); H NMR (CDCl3) δ
(ppm) = 1.87−2.02 (m, 2H, N(CH2CH2)2), 2.10 (td, J = 13.7/4.4 Hz,
2H, N(CH2CH2)2), 2.47 (td, J = 12.1/2.7 Hz, 1H, N(CH2CH2)2),
2.57 (td, J = 11.8/3.1 Hz, 1H, N(CH2CH2)2), 2.66−2.82 (m, 3H,
thioph CH2CH, N(CH2CH2)2), 2.88 (dd, J = 15.5/3.3 Hz, 1H, thioph
CH2CH), 3.56 (d, J = 13.1 Hz, 1H, NCH2Ph), 3.57 (s, 3H, OCH3),
3.61 (d, J = 13.1 Hz, 1H, NCH2Ph), 4.89 (dd, J = 7.4/3.3 Hz, 1H,
thioph CH2CH), 6.74 (d, J = 5.0 Hz, 1H, 3′-H-thioph), 7.16 (d, J = 5.0
Hz, 1H, 2′-H-thioph), 7.27−7.41 (m, 5H, Ph H).
1
after CHCl3 GPC; yield 4.5 mg (19%) after preparative TLC; H
NMR (CDCl3) δ (ppm) = 1.91−2.07 (m, 2H, N(CH2CH2)2), 2.08−
2.17 (m, 2H, N(CH2CH2)2), 2.47 (td, J = 11.5/2.8 Hz, 1H,
N(CH2CH2)2), 2.57 (td, J = 11.8/2.9 Hz, 1H, N(CH2CH2)2), 2.72
(dd, J = 15.5/7.4 Hz, 1H, thioph CH2CH), 2.77−2.84 (m, 2H,
N(CH2CH2)2), 2.87 (dd, J = 15.5/3.3 Hz, 1H, thioph CH2CH), 3.57
(s, 3H, OCH3), 3.57 (d, J = 13.2 Hz, 1H, NCH2Ph), 3.61 (d, J = 13.2
Hz, 1H, NCH2Ph), 4.92 (dd, J = 7.3/3.3 Hz, 1H, thioph CH2CH),
6.94 (s, 1H, 3′-H-thioph), 7.27−7.39 (m, 8H, Ph H), 7.49−7.54 (m,
2H, Ph H).
1-Benzyl-3′-phenyl-4′,5′-dihydrospiro[piperidine-4,7′-thieno[2,3-
c]pyran] (7a). According to general procedure B, the spirocyclic
thiophene 10 (31.3 mg, 0.105 mmol) was reacted with iodobenzene
(12.8 μL, 0.11 mmol), Ag2CO3 (28.1 mg, 0.10 mmol), PdCl2 (2.0 mg,
0.01 mmol), and P[OCH(CF3)2]3 (7.1 μL, 0.02 mmol) in m-xylene
(1.2 mL). The crude product was purified by CHCl3 GPC and
preparative TLC (h = 15 cm, hexane:EtOAc = 10:1, Rf = 0.06, four
runs): pale yellow resin; yield 8.2 mg (21%) (mixture of regioisomers)
after GPC; yield 2.5 mg (6%) after preparative TLC; 1H NMR
(CDCl3) δ (ppm) = 1.96−2.08 (m, 4H, N(CH2CH2)2), 2.44 (td, J =
11.4/4.0 Hz, 2H, N(CH2CH2)2), 2.69 (t, J = 5.4 Hz, 2H, thioph
CH2CH2), 2.75 (d, J = 11.5 Hz, 2H, N(CH2CH2)2), 3.57 (s, 2H,
NCH2Ph), 3.90 (t, J = 5.4 Hz, 2H, thioph CH2CH2), 7.13 (s, 1H, 2′-
H-thioph), 7.27−7.42 (m, 10H, Ph H).
1-Benzyl-5′-methoxy-3′-phenyl-4′,5′-dihydrospiro[piperidine-
4,7′-thieno[2,3-c]pyran] (8a). According to general procedure B, the
spirocyclic thiophene 11 (48 mg, 0.15 mmol) was reacted with
iodobenzene (17.9 μL, 0.16 mmol), Ag2CO3 (46.6 mg, 0.17 mmol),
PdCl2 (2.9 mg, 0.016 mmol), and P[OCH(CF3)2]3 (10.3 μL, 0.032
mmol) in m-xylene (1.5 mL). The crude product was purified by
CHCl3 GPC and preparative TLC (h = 15 cm, hexane:EtOAc = 14:1,
2% NEt3, Rf = 0.1, seven runs): colorless solid; yield 18.9 mg (32%)
(mixture of regioisomers) after GPC; yield 7.5 mg (13%) after
preparative TLC; 1H NMR (CDCl3) δ (ppm) = 1.97 (td, J = 12.8/3.3
Hz, 1H, N(CH2CH2)2), 2.02−2.16 (m, 3H, N(CH2CH2)2), 2.49 (td, J
= 12.1/2.4 Hz, 1H, N(CH2CH2)2), 2.59 (td, J = 11.6/2.9 Hz, 1H,
N(CH2CH2)2), 2.74−2.86 (m, 4H, N(CH2CH2)2, thioph CH2CH),
3.54 (s, 3H, OCH3), 3.58 (d, J = 13.7 Hz, 1H, NCH2Ph), 3.61 (d, J =
13.4 Hz, 1H, NCH2Ph), 4.85 (dd, J = 6.8/3.7 Hz, 1H, thioph
CH2CH), 7.16 (s, 1H, 2′-H-thioph), 7.27−7.41 (m, 10H, Ph H).
Receptor Binding Studies. The σ1 and σ2 receptor affinities were
recorded according to refs 39−42 The procedures are given in detail in
the Supporting Information.
Molecular Modeling. Pharmacophore Modeling. The model
structures of all compounds were built using the 2D−3D sketcher of
Discovery Studio Catalyst (DS, version 2.5, Accelrys, San Diego, CA).
High-quality conformational models are crucial for the development of
predictive pharmacophore models. Accordingly, in this study we
employed an ad hoc procedure to derive molecular conformations,
instead of using those automatically generated by DS Catalyst, for a
better quality in covering the low-energy conformational space.51 Each
molecular structure was subjected to energy minimization using the
generalized CHARMM force field57 until the gradient dropped below
0.05. The minimized structures were used as the starting point for
subsequent conformational searches. A 10000-step Monte Carlo
torsional sampling conformational search was conducted for each
compound. Unique low-energy conformations within 20 kcal/mol of
the corresponding global energy minimum were collected for each
molecule. A conformation was considered unique only when the
maximum displacement of at least one heavy atom was greater than 0.5
Å. A maximum of 250 unique conformations were recovered for each
compound. The classical conformational search was also carried out
using the Poling algorithm58 and the CHARMM force field as
implemented in the DS Catalyst program for comparison. The “best
quality” generation option was adopted to select representative
conformers over a 0−20 kcal/mol interval above the computed global
energy minimum in the conformational space, and again the number
of conformers generated for each compound was limited to a
maximum of 250. Comparing the results of the two conformational
searches, we verified the existence of considerable differences between
the two approaches in generating conformations for saturated six-
membered rings such as piperidine. This group is quite common in
drug molecules and constitutes a popular molecular scaffold. A survey
of the crystal structures of druglike molecules and protein/ligand
complexes available in the literature and in public databases reveals
that this saturated ring overwhelmingly adopts low-energy chair
8060
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
conformations. The conformational search conducted with the typical
DS Catalyst settings described above, however, generated predom-
inantly twisted conformations which might lead to an incorrect
mapping of this functionally important group. In comparison, the
alternative procedure of conformational search produced a consid-
erable number of chair conformations for this heterocyclic moiety. The
main drawback of this technique, however, is that it takes considerably
longer to generate the relevant conformational models. Nonetheless,
as the spirit of the work was the generation of a predictive 3D
pharmacophore model for these classes of compounds, we considered
it worthwhile to use more accurate conformational models.
On the basis of the conformations for each compound, the DS
Catalyst Hypothesis module was used to generate three-dimensional
pharmacophore models. During hypothesis generation, the software
attempts to minimize a cost function containing two main terms: the
first penalizes the deviation between the estimated affinities of the
training set molecules and their experimental values, while the second
penalizes the complexity of the hypothesis. The uncertainty factor for
each compound represents the ratio range of uncertainty in the affinity
value based on the expected statistical irregularity of biological data
collection. Uncertainty influences the first stepalso called the
constructive phaseof the hypothesis-generating process. In this
work, an uncertainty of 1.1 was preferred over the default factor of 3.0,
as the experimental affinities of our compounds barely span the
required 4 orders of magnitude.
Receptor Docking and MM/PBSA Scoring. The model structures
of the selected ligands were generated with DS. All molecules were
subjected to an initial energy minimization, with the convergence
criterion set to 10−4 kcal/(mol Å). A conformational search was
carried out using a well-validated, ad hoc developed combined
molecular mechanics/molecular dynamics simulated annealing
(MDSA) protocol60 using Amber 11.61 Accordingly, the relaxed
structures were subjected to five repeated temperature cycles (from
310 to 1000 K and back) using constant-volume/constant-temperature
(NVT) MD conditions. At the end of each annealing cycle, the
structures were again energy minimized to converge below 10−4 kcal/
(mol Å), and only the structures corresponding to the minimum
energy were used for further modeling. The atomic partial charges for
the geometrically optimized compounds were obtained using the
RESP procedure,62 and the electrostatic potentials were produced by
single-point quantum mechanical calculations at the Hartree−Fock
level with a 6-31G* basis set, using the Merz−Singh−Kollman van der
Waals parameters.63,64 Eventual ff0365 missing force field parameters
for the inhibitor molecules were generated using the general Amber
force field (GAFF)66 of Amber 11.
The optimized structures of all compounds were then docked into
the σ1 putative binding pockets by applying a consolidated
procedure;50,56,60 accordingly, it will be described here only briefly.
All docking experiments were performed with Autodock 4.3/Autodock
Tools 1.4.667 on a win64 platform. DS was employed to define the size
of the binding site, using an opening site of 10 Å and a grid size of 0.7
Å. The dimensions of the Autodock grid box, based on the cavity
identified by DS, was large enough to cover all possible rotations of
each ligand. van der Waals interactions and hydrogen bonding (O−H,
N−H, and S−H) were modeled with the Amber 12-6 and 12-10
Lennard-Jones parameters, respectively, while the distance-dependent
relative permittivity of Mehler and Solmajer68 was applied in the
generation of the electrostatic grid maps. A total of 300 Monte Carlo/
simulated annealing (MC/SA) runs were performed, with 100
constant-temperature cycles for simulated annealing. The GB/SA
implicit water model69 was used in these calculations to mimic the
solvated environment. The angles of the side chains and the rotation of
the angles φ and ψ were set free during the calculations, while all
others parameters of the MC/SA algorithm were kept as default. The
structures of all compounds were subjected to cluster analysis with a 1
Å tolerance for an all-atom root-mean-square (rms) deviation from a
lower energy structure representing each cluster family. The resulting
docked conformations were clustered and visualized; then, for each
compound, only the molecular conformation satisfying the combined
criteria of (i) having the lowest (i.e., more favorable) Autodock energy,
(ii) belonging to a highly populated cluster, and (iii) satisfying the
specific 3D pharmacophore requirements was selected to carry for
further modeling.
An analysis of the functional groups characterizing our compounds
suggested that HYAr, aliphatic (HYAl) and generic HY, HBA, and PI
features could effectively map the critical chemical features and, hence,
describe the σ1 receptor affinity of our compounds. Accordingly, these
five features were selected to constitute the essential information in the
automated hypothesis generation process.
Three validation procedures were used to determine the statistical
relevance and the validity of the proposed 3D pharmacophore models:
the test set prediction method, the CatScramble method, and the
leave-one-out procedure. In this work, the first procedure consisted of
the collection of further, different compounds into a test set and
performing a regression analysis by mapping the test set molecules
onto the best pharmacophore hypothesis. The high correlation
coefficients obtained using the test set compounds revealed the
good correlation between the actual and estimated affinities and,
hence, the predictive validity of the corresponding 3D hypothesis. The
CatScramble validation procedure is based on Fisher’s randomization
test.59 The goal of this type of validation is to check whether there is a
strong correlation between the chemical structures and the binding
affinity. This is done by randomizing the affinity data associated with
the training set compounds, generating pharmacophore hypotheses
using the same features and parameters employed to develop the
original pharmacophore model. The statistical significance is calculated
according to the following formula:
Each ligand/receptor complex obtained from the docking procedure
was further refined in Amber 1161 using the quenched molecular
dynamics (QMD) method.50,56,60 According to QMD, 1 ns MD
simulations at 300 K were employed to sample the conformational
space of each ligand/receptor complex in the GB/SA continuum
solvation environment.69 The integration step was equal to 1 fs. After
each picosecond, each system was cooled to 0 K, and the structure was
extensively minimized and stored. To prevent global conformational
changes of the protein, the backbone atoms of the protein binding site
were constrained by a harmonic force constant of 100 kcal/Å, whereas
the amino acid side chains and the ligands were allowed to move
without any constraint. The best energy configuration of each complex
resulting from the previous step was subsequently solvated by a cubic
box of TIP3P70 water molecules extending at least 10 Å in each
direction from the solute. The system was then neutralized with the
addition of 21 Na+ and 15 Cl− counterions; furthermore, the solution
ionic strength was adjusted to the physiological value of 0.15 M by
adding the required amounts of Na+ and Cl− ions.
significance = 100[1 − (1 + x/y)]
where x is the total number of hypotheses having a total cost lower
than that of the original (best) hypothesis and y is the total number of
DS Catalyst Hypothesis runs (initial + random runs). Thus, 49
random spreadsheets (i.e., 49 DS Catalyst Hypothesis runs) have to be
generated to obtain a 98% confidence level. Should any randomized
data set result in the generation of a 3D pharmacophore with similar or
even better cost values, root-mean-square deviations, and correlation
coefficients, then it is likely that the original hypothesis does reflect a
chance correlation.
Finally, the leave-one-out test checks if the correlation between
experimental and computed affinities is heavily dependent on one
particular molecule of the training set by recomputing the
pharmacophore model with the exclusion of one molecule at a time.
Accordingly, 27 new training sets were derived, each composed of 30
molecules, and 27 DS Catalyst Hypothesis calculations were
performed under the same conditions. For each run, the hypothesis
characterized by the lowest total cost was employed to predict the
affinity of the excluded compound and to estimate the new correlation
coefficient.
Each solvated system was relaxed by 500 steps of steepest descent
followed by 500 other conjugate-gradient minimization steps and then
gradually heated to a temperature of 300 K in intervals of 50 ps of
NVT MD, using a Verlet integration time step of 1.0 fs. The Langevin
8061
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
thermostat was used to control temperature, with a collision frequency
of 2.0 ps−1. The SHAKE method71 was used to constrain all of the
covalently bound hydrogen atoms, while long-range nonbonded van
der Waals interactions were truncated by using dual cutoffs of 6 and 12
Å. The particle mesh Ewald (PME) method72 was applied to treat
long-range electrostatic interactions. The protein was restrained with a
force constant of 2.0 kcal/(mol Å), and all simulations were carried
out with periodic boundary conditions.
The density of each system was subsequently equilibrated via MD
runs in the isothermal−isobaric (NPT) ensemble, with isotropic
position scaling and a pressure relaxation time of 1.0 ps, for 50 ps with
a time step of 1 fs. All restraints on the protein atoms were then
removed, and each system was further equilibrated using NPT MD
runs at 300 K, with a pressure relaxation time of 2.0 ps. Three
equilibration steps were performed, each 2 ns long and with a time
step of 2.0 fs. To check the system stability, the fluctuations of the
rmsd of the simulated position of the backbone atoms of the σ1
receptor with respect to those of the initial protein were monitored.
All chemicophysical parameters and rmsd values showed very low
fluctuations at the end of the equilibration process, indicating that the
systems reached a true equilibrium condition.
yield converged mean changes in solute entropy. On the basis of the
size-reduced snapshots of the complex, we generated structures of the
uncomplexed reactants by removing the atoms of the protein and
ligand. Each of those structures was minimized, using a distance-
dependent dielectric constant ε = 4r, to account for solvent screening,
and its entropy was calculated using classical statistical formulas and
normal-mode analysis. To minimize the effects due to different
conformations adopted by individual snapshots, we averaged the
estimation of entropy over 40 snapshots.
The entire MD simulation and data analysis procedure was
optimized by integrating Amber 11 in modeFRONTIER, a multi-
disciplinary and multiobjective optimization and design environ-
ment.78
ASSOCIATED CONTENT
* Supporting Information
■
S
Physical and spectroscopic data of all new compounds, purity
data, general chemistry methods, and details of the receptor
binding studies. This material is available free of charge via the
Internet at http://pubs.acs.org.
Each equilibration phase was followed by a data production run
consisting of 4 ns of MD simulations in the canonical (constant
volume−constant temperature, NVT) ensemble. Only the last 2 ns of
each equilibrated MD trajectory weas considered for statistical data
collections. A total of 100 trajectory snapshots were analyzed for each
drug/receptor complex.
The binding free energy, ΔGbind, between each ligand and the σ1
receptor was estimated by resorting to the MM/PBSA approach.55
According to this well-validated methodology,50,56,60,73 the binding
free energy between two biological entities (e.g., a drug and its protein
target) in a solvent was obtained as the sum of the interaction energy
between the receptor and the ligand (ΔEMM), the solvation free energy
(ΔGsol), and the conformational entropy contribution (−TΔS),
averaged over a series of snapshots from the corresponding MD
trajectories:
AUTHOR INFORMATION
Corresponding Author
*Phone/fax: +81-52-788-6098 (K.I.). Phone: +39-040-5583750
(S.P.); +49-251-8333311 (B.W.). Fax: +39-040-569823 (S.P.);
+49-251-8332144 (B.W.). E-mail: itami.kenichiro@a.mbox.
nagoya-u.ac.jp (K.I.); sabrina.pricl@di3.units.it (S.P.);
wuensch@uni-muenster.de (B.W.).
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was performed within the framework of the
International Research Training Group (IRTG) “Complex
Functional Systems in Chemistry: Design, Synthesis and
Applications” in collaboration with the University of Nagoya.
Financial support of the IRTG and this project by the Deutsche
Forschungsgemeinschaft is gratefully acknowledged. E.L.,
V.D.C., and S.P. acknowledge financial support from ESTECO
through the project Drug Discovery and Optimization by
Simulation (DDOS).
ΔGbind = ΔEMM + ΔGsolv − TΔS
(1)
The ΔEMM term in eq 1 can be obtained directly from the molecular
mechanics interaction energies as
ΔEMM = ΔEint + ΔEvdW + ΔEele
(2)
where ΔEint, ΔEvdW, and ΔEele are the internal, van der Waals, and
electrostatic components of the nonbonded interaction energy,
respectively. Since in this work we adopted the “single trajectory
protocol”, ΔEint = 0 in eq 2.
The second term in eq 1, the solvation energy ΔGsol, can also be
partitioned into two different contributions:
ABBREVIATIONS USED
■
MW, molecular weight; NMDA, N-methyl-D-aspartate; MM/
PBSA, molecular mechanics/Poisson−Boltzmann surface area;
rmsd, root-mean-square deviation; PI, positive ionizable; HBA,
hydrogen bond acceptor; HY, hydrophobic moiety; HYAr,
hydrophobic aromatic point; HYAl, hydrophobic aliphatic;
MD, molecular dynamics; ADL, average dynamic length; THF,
tetrahydrofuran
ΔGsolv = ΔGPB + ΔGNP
(3)
The polar term of ΔGsolv, ΔGPB, was estimated using DelPhi,74 which
solves the Poisson−Boltzmann equations numerically and calculates
the electrostatic energy according to the electrostatic potential. In
these calculations, the interior and exterior dielectric constant values
were set equal to 1 and 80, respectively. A grid spacing of 0.5 per
angstrom, extending 20% beyond the dimensions of the solute, was
employed. The value of the nonpolar component of ΔGsolv, ΔGNP, was
calculated using the following relationship:75
REFERENCES
■
(1) Falkenstein, E.; Meyer, C.; Eisen, C.; Scriba, P. C.; Wehling, M.
Full-length cDNA sequence of a progesterone membrane-binding
protein from porcine vascular smooth muscle cells. Biochem. Biophys.
Res. Commun. 1996, 229, 86−89.
ΔGNP = γ(SA) + β
(4)
in which γ = 0.00542 kcal/(mol Å2), β = 0.92 kcal/mol, and SA is the
molecular surface area estimated by means of the MSMS software.76
The change in solute entropy upon association (−TΔS in eq 1) was
evaluated using the Nmode module of Amber 11.61 The normal-mode
analysis77 was performed for the minimized structures of the
complexes, σ1 receptor, and ligands using a distance-dependent
dielectric constant ε = 4rij. In the first step of this calculation, an 8 Å
sphere around the ligand was cut out from an MD snapshot for each
ligand/protein complex. This value was shown to be large enough to
(2) Xu, J.; Zeng, C.; Chu, W.; Pan, F.; Rothfuss, J. M.; Zhang, F.; Tu,
Z.; Zhou, D.; Zeng, D.; Vangveravong, S.; Johnston, F.; Spitzer, D.;
Chang, K. C.; Hotchkiss, R. S.; Hawkins, W. G.; Wheeler, K. T.; Mach,
R. H. Identification of the PGRMC1 protein complex as the putative
sigma-2 receptor binding site. Nat. Commun. 2011, 2, No. 380.
(3) Hanner, M.; Moebius, F. F.; Flandorfer, A.; Knaus, H.-G.;
Striessnig, J.; Kempner, E.; Glossmann, H. Purification, molecular
cloning, and expression of the mammalian sigma1-binding site. Proc.
Natl. Acad. Sci. U.S.A. 1996, 93, 8072−8077.
8062
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
(4) Seth, P.; Fei, Y.-J.; Li, H. W.; Huang, W.; Leibach, F. H.;
Ganapathy, V. Cloning and functional characterization of a σ receptor
from rat brain. J. Neurochem. 1998, 70, 922−932.
(5) Kekuda, R.; Prasad, P. D.; Fei, Y.-J.; Leibach, F. H.; Ganapathy, V.
Cloning and functional expression of the human type 1 sigma receptor
(hSigmaR1). Biochem. Biophys. Res. Commun. 1996, 229, 553−558.
(6) Aydar, E.; Palmer, C. P.; Klyachko, V. A.; Jackson, M. B. The
sigma receptor as a ligand-regulated auxiliary potassium channel
subunit. Neuron 2002, 34, 399−410.
(7) Hayashi, T.; Su, T. P. Sigma-1 receptor ligands: potential in the
treatment of neuropsychiatric disorders. CNS Drugs 2004, 18, 269−
284.
(8) Cobos, E. J.; Entrena, J. M.; Nieto, F. R.; Cendan, C. M.;
DelPezo, E. Pharmacology and therapeutic potential of sigma1
receptor ligands. Curr. Pharmacol. 2008, 6, 344−366.
(9) Maurice, T.; Su, T. P. The pharmacology of sigma-1 receptors.
Pharmacol. Ther. 2009, 124, 195−206.
(10) Ishikawa, M.; Hashimoto, K. The role of sigma-1 receptors in
the pathophysiology of neuropsychiatric diseases. J. Recept., Ligand
Channel Res. 2010, 3, 25−36.
(11) Choia, S.-R.; Yange, B.; Ploessla, K.; Chumpradita, S.; Weya, S.-
P.; Actona, P. D.; Wheelerc, K.; Mach, R. H.; Kunga, H. F.
Development of a Tc-99m labeled sigma-2 receptor-specific ligand
as a potential breast tumor imaging agent. Nucl. Med. Biol. 2001, 28,
657−666.
(12) Tu, Z.; Xu, J.; Jones, L. A.; Li, S.; Dumstorff, C.; Vangveravong,
S.; Chen, D. L.; Wheeler, K. T.; Welch, M. J.; Mach, R. H. Fluorine-18-
labeled benzamide analogues for imaging the σ2 receptor status of solid
tumors with positron emission tomography. J. Med. Chem. 2007, 50,
3194−3204.
(13) Wilke, R. A.; Lupardus, P. J.; Grandy, D. K.; Rubinstein, M.;
Low, M. J.; Jackson, M. B. K+ channel modulation in rodent
neurohypophysial nerve terminals by sigma receptors and not by
dopamine receptors. J. Physiol. 1999, 517, 391−406.
(14) Aydar, E.; Palmer, C. P.; Djamgoz, M. B. A. Sigma receptors and
cancer: possible involvement of ion channels. Cancer Res. 2004, 64,
5029−5035.
(15) Monnet, F. P. Sigma-1 receptor as regulator of neuronal
intracellular Ca2+: clinical and therapeutic relevance. Biol. Cell 2005,
97, 873−883.
(16) Zhang, H.; Cuevas, J. Sigma receptors inhibit high-voltage-
activated calcium channels in rat sympathetic and parasympathetic
neurons. J. Neurophysiol. 2002, 87, 2867−2879.
(24) Oberdorf, C.; Schmidt, T. J.; Wunsch, B. 5D-QSAR for
̈
spirocyclic σ1 receptor ligands by Quasar receptor surface modeling.
Eur. J. Med. Chem. 2010, 45, 3116−3124.
(25) Wunsch, B. Pharmacophore models and development of
̈
spirocyclic ligands for σ1 receptor. Curr. Pharm. Des. 2012, 18, 930−
937.
́
(26) Hassan, J.; Sevignon, M.; Gozzi, C.; Schulz, E.; Lemaire, M.
Aryl−aryl bond formation one century after the discovery of the
Ullmann reaction. Chem. Rev. 2002, 102, 1359−1470.
(27) Oberdorf, C.; Schepmann, D.; Vela, J. M.; Diaz, J. L.; Holenz, J.;
Wunsch, B. Thiophene bioisosteres of spirocyclic σ receptor ligands. 1.
̈
N-substituted spiro[piperidine-4,4′-thieno[3,2-c]pyrans]. J. Med.
Chem. 2008, 51, 6531−6537.
(28) Oberdorf, C.; Schepmann, D.; Vela, J. M.; Buschmann, H.;
Holenz, J.; Wunsch, B. Thiophene bioisosters of spirocyclic σ receptor
̈
ligands: relationships between substitution pattern and σ receptor
affinity. J. Med. Chem. 2012, 55, 5350−5360.
(29) Ohta, A.; Akita, Y.; Ohkuwa, T.; Chiba, M.; Fukunaga, R.;
Miyafuji, A.; Nakata, T.; Tani, N.; Aoyagi, Y. Palladium-catalyzed
arylation of furan, thiophene, benzobfuran and benzobthiophene.
Heterocycles 1990, 31, 1951.
́
(30) Liegault, B.; Lapointe, D.; Caron, L.; Vlassova, A.; Fagnou, K.
Establishment of broadly applicable reaction conditions for the
palladium-catalyzed direct arylation of heteroatom-containing aromatic
compounds. J. Org. Chem. 2009, 74, 1826−1834.
(31) Yanagisawa, S.; Sudo, T.; Noyori, R.; Itami, K. Direct C-H
arylation of (hetero)arenes with aryl iodides via rhodium catalysis. J.
Am. Chem. Soc. 2006, 128, 11748−11749.
(32) Yanagisawa, S.; Sudo, T.; Noyori, R.; Itami, K. Direct coupling
of arenes and iodoarenes catalyzed by a rhodium complex with a
strongly pi-accepting phosphite ligand. Tetrahedron 2008, 64, 6073−
6081.
(33) Benoît, J.; Yamamoto, T.; Itami, K. Iridium catalysis for C−H
bond arylation of heteroarenes with iodoarenes. Angew. Chem., Int. Ed.
2009, 48, 3644−3647.
(34) Do, H.-Q.; Khan, R. M. K.; Daugulis, O. A general method for
copper-catalyzed arylation of arene C−H bonds. J. Am. Chem. Soc.
2008, 130, 15185−15192.
(35) Yanagisawa, S.; Itami, K. Palladium/2,2′-bipyridyl/Ag2CO3
catalyst for C−H bond arylation of heteroarenes with haloarenes.
Tetrahedron 2011, 67, 4425−4430.
(36) Yanagisawa, S.; Ueda, K.; Sekizawa, H.; Itami, K. Programmed
synthesis of tetraarylthiophenes through sequential C-H arylation. J.
Am. Chem. Soc. 2009, 131, 14622−14623.
(17) Bermack, J. E.; Debonnel, G. Distinct modulatory roles of sigma
receptor subtypes on glutamatergic responses in the dorsal hippo-
campus. Synapse 2005, 55, 37−44.
(37) Ueda, K.; Yanagisawa, S.; Yamaguchi, J.; Itami, K. A general
catalyst for the β-selective C−H bond arylation of thiophenes with
iodoarenes. Angew. Chem., Int. Ed. 2010, 49, 8946−8949.
(38) Kirchberg, S.; Tani, S.; Ueda, K.; Yamaguchi, J.; Studer, A.;
Itami, K. Oxidative biaryl coupling of thiophenes and thiazoles with
arylboronic acids through palladium catalysis: otherwise difficult C4-
selective C-H arylation enabled by boronic acids. Angew. Chem., Int.
Ed. 2011, 50, 2387−2391.
(18) Gudelsky, G. A. Biphasic effect of sigma receptor ligands on the
extracellular concentration of dopamine in the striatum of the rat. J.
Neural Transm. 1999, 106, 849−856.
(19) Bowen, W. D. Sigma receptors: recent advances and new clinical
potentials. Pharm. Acta Helv. 2000, 74, 211−218.
(20) Schlager, T.; Schepmann, D.; Wurthwein, E.-U.; Wunsch, B.
̈
̈
̈
Synthesis and structure-affinity relationships of novel spirocyclic σ
receptor ligands with furopyrazole structure. Bioorg. Med. Chem. 2008,
16, 2992−3001.
(39) Maier, C. A.; Wunsch, B. Novel spiropiperidines as highly
̈
potent and subtype selective σ-receptor ligands. Part 1. J. Med. Chem.
2002, 45, 438−448.
(21) Schlager, T.; Schepmann, D.; Lehmkuhl, K.; Holenz, J.; Vela, J.
̈
(40) Maier, C. A.; Wunsch, B. Novel σ receptor ligands. Part 2. SAR
̈
M.; Buschmann, H.; Wunsch, B. Combination of two pharmacophoric
̈
of spiro[[2]benzopyran-1,4′-piperidines] and spiro[[2]benzofuran-
1,4′-piperidines] with carbon substituents in position 3. J. Med.
Chem. 2002, 45, 4923−4930.
systems: synthesis and pharmacological evaluation of spirocyclic
pyranopyrazoles with high σ1 receptor affinity. J. Med. Chem. 2011,
54, 6704−6713.
(41) Große Maestrup, E.; Wiese, C.; Schepmann, D.; Hiller, A.;
(22) Meyer, C.; Neue, B.; Schepmann, D.; Yanagisawa, S.;
Fischer, S.; Scheunemann, M.; Brust, P.; Wunsch, B. Synthesis of
̈
Yamaguchi, J.; Wurthwein, E.-U.; Itami, K.; Wunsch, B. Exploitation
̈
̈
spirocyclic σ1 receptor ligands as potential PET radiotracers, structure-
affinity relationships and in vitro metabolic stability. Bioorg. Med.
Chem. 2009, 17, 3630−3641.
of an additional hydrophobic pocket of σ1 receptors: late-stage diverse
modifications of spirocyclic thiophenes by C-H bond functionalization.
Org. Biomol. Chem. 2011, 9, 8016−8029.
(42) Holl, R.; Schepmann, D.; Frohlich, R.; Grunert, R.; Bednarski, P.
̈
̈
(23) Meyer, C.; Schepmann, D.; Yanagisawa, S.; Yamaguchi, J.; Itami,
J.; Wunsch, B. Dancing of the second aromatic residue around the 6,8-
K.; Wunsch, B. Late-stage C-H bond arylation of spirocyclic σ ligands
̈
̈
1
diazabicyclo[3.2.2]-nonane framework: influence on σ receptor affinity
and cytotoxicity. J. Med. Chem. 2009, 52, 2126−2137.
for analysis of complementary σ1 receptor surface. Eur. J. Org. Chem.
2012, in press.
8063
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
(43) Wiese, C.; Große Maestrup, E.; Schepmann, D.; Vela, J. M.;
Synergistic experimental/computational studies on arylazoenamine
derivatives that target the bovine viral diarrhea virus RNA-dependent
RNA polymerase. Bioorg. Med. Chem. 2010, 18, 6055−6068.
(d) Tonelli, M.; Boido, V.; La Colla, P.; Loddo, R.; Posocco, P.;
Paneni, M. S.; Fermeglia, M.; Pricl, S. Pharmacophore modeling,
resistant mutant isolation, docking, and MM-PBSA analysis: combined
experimental/computer-assisted approaches to identify new inhibitors
of the bovine viral diarrhea virus (BVDV). Bioorg. Med. Chem. 2010,
18, 2304−2316. (e) Carta, A.; Pricl, S.; Piras, S.; Fermeglia, M.; La
Colla, P.; Loddo, R. Activity and molecular modeling of a new small
molecule active against NNRTI-resistant HIV-1 mutants. Eur. J. Med.
Chem. 2009, 44, 5117−5122. (f) Tonelli, M.; Vazzana, I.; Tasso, B.;
Boido, V.; Sparatore, F.; Fermeglia, M.; Paneni, M. S.; Posocco, P.;
Pricl, S.; La Colla, P.; Ibba, C.; Secci, B.; Collu, G.; Loddo, R. Antiviral
and cytotoxic activities of aminoarylazo compounds and aryltriazene
derivatives. Bioorg. Med. Chem. 2009, 17, 4425−4440 and references
therein.
(57) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
Swaminathan, S.; Karplus, M. J. CHARMM: a program for
macromolecular energy, minimization, and dynamics calculations. J.
Comput. Chem. 1983, 4, 187−217.
(58) (a) Smellie, A.; Teig, S. L.; Towbin, P. Poling: promoting
conformational variation. J. Comput. Chem. 1994, 16, 171−187.
(b) Smellie, A.; Kahn, S. D.; Teig, S. L. Analysis of conformational
coverage. 1. Validation and estimation of coverage. J. Chem. Inf.
Comput. Sci. 1995, 35, 285−294. (c) Smellie, A.; Kahn, S. D.; Teig, S.
L. Analysis of conformational coverage. 2. Applications of conforma-
tional models. J. Chem. Inf. Comput. Sci. 1995, 35, 295−304.
(59) Fisher, R. The Design of Experiments; Hafner Publishing: New
York, 1966.
Holenz, J.; Buschmann, H.; Wunsch, B. Pharmacological and
̈
metabolic characterisation of the potent σ1 receptor ligand 1′-benzyl-
3-methoxy-3H-spiro[[2]benzofuran-1,4′-piperidine]. J. Pharm. Phar-
macol. 2009, 61, 631−640.
(44) Wirt, U.; Schepmann, D.; Wunsch, B. Asymmetric synthesis of
̈
1-substituted tetrahydro-3-benzazepines as NMDA receptor antago-
nists. Eur. J. Org. Chem. 2007, 462−475.
(45) Schepmann, D.; Frehland, B.; Lehmkuhl, K.; Tewes, B.;
Wunsch, B. Development of a selective competitive receptor binding
̈
assay for the determination of the affinity to NR2B containing NMDA
receptors. J. Pharm. Biomed. Anal. 2010, 53, 603−608.
(46) Li, H.; Sutter, J.; Hoffman, R. HypoGen: An automated system
for generating 3D predictive pharmacophore models. In Pharmaco-
phore Perception, Development, and Use in Drug Design; Guner, O. F.,
Ed.; International University Line: La Jolla, CA, 1999; pp 173−189.
(47) Sutter, J.; Guner, O. F.; Hoffman, R. D.; Li, H.; Wadman, M.
Effect of variable weight and tolerances on predictive model
generation. In Pharmacophore Perception, development, and Use in
Drug Design; Guner, O. F., Ed.; International University Line: La Jolla,
CA, 1999; pp 501−511.
(48) Glennon, R. A.; Ablordeppey, S. Y.; Ismaiel, A. M.; El-Ashmawy,
M. B.; Fischer, J. B.; Howie, K. B. Structural features important for σ1
receptor binding. J. Med. Chem. 1994, 37, 1214−19.
(49) Laggner, C.; Schieferer, C.; Fiechtner, B.; Poles, G.; Hoffmann,
R. D.; GLossmann, H.; Langer, T.; Moebius, F. Discovery of high-
affinity ligands of σ1 receptor, ERG2, and emopamil binding protein by
pharmacophore modeling and virtual screening. J. Med. Chem. 2005,
48, 4754−4764.
(50) Laurini, E.; Dal Col, V.; Mamolo, M. G.; Zampieri, D.; Posocco,
P.; Fermeglia, M.; Vio, L.; Pricl, S. Homology model and docking-
based virtual screening for ligands of the σ1 receptor. ACS Med. Chem.
Lett. 2011, 2, 834−839.
(51) Zampieri, D.; Mamolo, M. G.; Laurini, E.; Florio, C.; Zanette,
C.; Fermeglia, M.; Posocco, P.; Paneni, M. S.; Pricl, S.; Vio, L.
Synthesis, biological evaluation, and three-dimensional in silico
pharmacophore model for σ1 receptor ligands based on a series of
substituted benzo[d]oxazol-2(3H)-one derivatives. J. Med. Chem.
2009, 52, 5380−5393.
(52) Seth, P.; Ganapathy, M. E.; Conway, S. J.; Bridges, C. D.; Smith,
S. B.; Casellas, P.; Ganapathy, V. Expression pattern of the type 1
sigma receptor in the brain and identity of critical anionic amino acid
residues in the ligand-binding domain of the receptor. Biochim.
Biophys. Acta 2001, 1540, 59−67.
(53) Pal, A.; Fontanilla, A. R.; Ramachandran, S.; Chu, U. B.;
Mavlyutov, T.; Ruoho, A. E. Identification of regions of the sigma-1
receptor ligand binding site using a novel photoprobe. Mol. Pharmacol.
2007, 72, 921−933.
(54) Pal, A.; Chu, U. B.; Ramachandran, S.; Grawoig, D.; Guo, L. W.;
Hajipour, A. R.; Ruoho, A. E. Juxtaposition of the steroid binding
domain-like I and II regions constitutes a ligand binding site in the σ-1
receptor. J. Biol. Chem. 2008, 283, 19646−19656.
(55) Srinivasan, J.; Cheatham, T. E., III; Cieplak, P.; Kollman, P. A.;
Case, D. A. Continuum solvent studies of the stability of DNA, RNA,
and phosphoramidate−DNA helices. J. Am. Chem. Soc. 1998, 120,
9401−9409.
(60) (a) Laurini, E.; Zampieri, D.; Mamolo, M. G.; Vio, L.; Zanette,
C.; Florio, C.; Posocco, P.; Fermeglia, M.; Pricl, S. A 3D-
pharmacophore model for sigma2 receptors based on a series of
substituted benzo[d]oxazol-2(3H)-one derivatives. Bioorg. Med. Chem.
Lett. 2010, 20, 2954−2957. (b) Mazzei, M.; Nieddu, E.; Miele, M.;
Balbi, A.; Ferrone, M.; Fermeglia, M.; Mazzei, M. T.; Pricl, S.; La
Colla, P.; Marongiu, F.; Ibba, C.; Loddo, R. Activity of Mannich bases
of 7-hydroxycoumarin against Flaviviridae. Bioorg. Med. Chem. 2008,
16, 2591−2605. (c) Zampieri, D.; Mamolo, M. G.; Vio, L.; Banfi, E.;
Scialino, G.; Fermeglia, M.; Ferrone, M.; Pricl, S. Synthesis, antifungal
and antimycobacterial activities of new bis-imidazole derivatives, and
prediction of their binding to P450(14DM) by molecular docking and
MM/PBSA method. Bioorg. Med. Chem. 2007, 15, 7444−7458.
(d) Carta, A.; Loriga, M.; Paglietti, G.; Ferrone, M.; Fermeglia, M.;
Pricl, S.; Sanna, T.; Ibba, C.; La Colla, P.; Loddo, R. Design, synthesis,
and preliminary in vitro and in silico antiviral activity of [4,7]-
phenantrolines and 1-oxo-1,4-dihydro-[4,7]phenantrolines against
single-stranded positive-sense RNA genome viruses. Bioorg. Med.
Chem. 2007, 15, 1914−1927. (e) Carta, A.; Loriga, G.; Piras, S.;
Paglietti, G.; Ferrone, M.; Fermeglia, M.; Pricl, S.; La Colla, P.; Secci,
B.; Collu, G.; Loddo, R. Synthesis and in vitro evaluation of the anti-
viral activity of N-[4-(1H(2H)-benzotriazol-1(2)-yl)phenyl]-
alkylcarboxamides. Med. Chem. 2006, 2, 577−589. (f) Frecer, V.;
́ ̌ ̌
Kabelac, M.; De Nardi, P.; Pricl, S.; Miertus, S. Structure-based design
of inhibitors of NS3 serine protease of hepatitis C virus. J. Mol.
Graphics Modell. 2004, 22, 209−220. (g) Felluga, F.; Pitacco, G.;
Valentin, E.; Coslanich, A.; Fermeglia, M.; Ferrone, M.; Pricl, S.
studying enzyme enantioselectivity using combined ab initio and free
energy calculations: α-chymotrypsin and methyl cis- and trans-5-oxo-
2-pentylpirrolidine-3-carboxylates. Tetrahedron: Asymmetry 2003, 14,
3385−3399.
(56) For a selected list of recent, successful applications of the MM/
PBSA methodology in related topics from our group see, for instance:
(a) Carta, A.; Briguglio, I.; Piras, S.; Boatto, G.; La Colla, P.; Loddo,
R.; Tolomeo, M.; Grimaudo, S.; Di Cristina, A.; Pipitone, R. M.;
Laurini, E.; Paneni, M. S.; Posocco, P.; Fermeglia, M.; Pricl, S. 3-Aryl-
2-[1H-benzotriazol-1-yl]acrylonitriles: a novel class of potent tubulin
inhibitors. Eur. J. Med. Chem. 2011, 46, 4151−4167. (b) Liu, X.; Chen,
H.; Laurini, E.; Wang, Y.; Dal Col, V.; Posocco, P.; Ziarelli, F.;
Fermeglia, M.; Zhang, C. C.; Pricl, S.; Peng, L. 2-Difluoromethylene-4-
methylenepentanoic acid, a paradoxical probe able to mimic the
signaling role of 2-oxoglutaric acid in cyanobacteria. Org. Lett. 2011,
13, 2924−2927. (c) Giliberti, G.; Ibba, C.; Marongiu, E.; Loddo, R.;
Tonelli, M.; Boido, V.; Laurini, E.; Posocco, P.; Fermeglia, M.; Pricl, S.
(61) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C.
L.; Wang, J.; Duke, R. E.; Luo, R.; Walker, R. C.; Zhang, W.; Merz, K.
M.; Roberts, B.; Wang, B.; Hayik, S.; Roitberg, A.; Seabra, G.;
Kolossvary, I.; Wong, K. F.; Paesani, F.; Vanicek, J.; Wu, X.; Brozell, S.
́
R.; Steinbrecher, T.; Gohlke, H.; Cai, Q.; Ye, X.; Wang, J.; Hsieh, M.-
J.; Cui, G.; Roe, D. R.; Mathews, D. H.; Seetin, M. G.; Sagui, C.; Babin,
V.; Luchko, T.; Gusarov, S.; Kovalenko, A.; Kollman, P. A. AMBER 11;
University of California: San Francisco, CA, 2010.
8064
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065

Journal of Medicinal Chemistry
Article
(62) (a) Bayly, C. I.; Cieplak, P.; Cornell, W. D.; Kollman, P. A. A
well-behaved electrostatic potential based method using charge
restraints for determining atom-centered charges: the RESP model.
J. Phys. Chem. 1993, 97, 10269−10280. (b) Cornell, W. D.; Cieplak,
P.; Bayly, C. I.; Kollman, P. A. Application of RESP charges to
calculate conformational energies, hydrogen bond energies and free
energies of solvation. J. Am. Chem. Soc. 1993, 115, 9620−9631.
(c) Cieplak, P.; Cornell, W. D.; Bayly, C. I.; Kollman, P. A. Application
of the multimolecule and multiconformational RESP methodology to
biopolymers: charge derivation for DNA, RNA and proteins. J.
Comput. Chem. 1995, 16, 1357−1377.
2011, 8, 161−170. (f) Pavan, G. M.; Posocco, P.; Tagliabue, A.; Maly,
M.; Malek, A.; Danani, A.; Ragg, E.; Catapano, C. V.; Pricl, S. PAMAM
dendrimers for siRNA delivery: computational and experimental
insights. Chem.Eur. J. 2010, 16, 7781−7795. (g) Dileo, P.; Pricl, S.;
Tamborini, E.; Negri, T.; Stacchiotti, S.; Gronchi, A.; Posocco, P.;
Laurini, E.; Coco, P.; Fumagalli, E.; Casali, P. G.; Pilotti, S. Imatinib
response in two GIST patients carrying two hitherto functionally
uncharacterized PDGFRA mutations: an imaging, biochemical and
molecular modeling study. Int. J. Cancer 2011, 128, 983−990.
(h) Jones, S. P.; Pavan, G. M.; Danani, A.; Pricl, S.; Smith, D. K.
Quantifying the effect of surface ligands on dendron-DNA
interactions: insights into multivalency through a combined
experimental and theoretical approach. Chem.Eur. J. 2010, 16,
4519−4532. (i) Pierotti, M. A.; Negri, T.; Tamborini, E.; Perrone, F.;
Pricl, S.; Pilotti, S. Targeted therapies: the rare cancer paradigm. Mol.
Oncol. 2010, 4, 19−37. (j) Conca, E.; Negri, T.; Gronchi, A.;
Fumagalli, E.; Tamborini, E.; Pavan, G. M.; Fermeglia, M.; Pierotti, M.
A.; Pricl, S.; Pilotti, S. Activate and resist: L576P-KIT in GIST. Mol.
Cancer Ther. 2009, 8, 2491−2495. (k) Woodman, S. E.; Trent, J. C.;
Stemke-Hale, K.; Lazar, A. J.; Pricl, S.; Pavan, G. M.; Fermeglia, M.;
Gopal, Y. N.; Yang, D.; Podoloff, D. A.; Ivan, D.; Kim, K. B.;
Papadopoulos, N.; Hwu, P.; Mills, G. B.; Davies, M. A. Activity of
dasatinib against L576P KIT mutant melanoma: molecular, cellular,
and clinical correlates. Mol. Cancer Ther. 2009, 8, 2079−2085.
(l) Pavan, G. M.; Danani, A.; Pricl, S.; Smith, D. K. Modeling the
multivalent recognition between dendritic molecules and DNA:
understanding how ligand “sacrifice” and screening can enhance
binding. J. Am. Chem. Soc. 2009, 131, 9686−9694 and references
therein.
(63) Singh, U. C.; Kollman, P. A. An approach to computing
electrostatic charges for molecules. J. Comput. Chem. 1984, 5, 129−
145.
(64) Besler, B. H.; Merz, K. M.; Kollman, P. A. Atomic charges
derived from semiempirical methods. J. Comput. Chem. 1990, 11, 431−
439.
(65) Duan, Y.; Wu, C.; Chowdhury, S.; Lee, M. C.; Xiong, G.; Zhang,
W.; Yang, R.; Cieplak, P.; Luo, R.; Lee, T.; Caldwell, J.; Wang, J.;
Kollman, P. A. Point-charge force field for molecular mechanics
simulations of proteins based on condensed-phase quantum
mechanical calculations. J. Comput. Chem. 2003, 24, 1999−2012.
(66) Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D.
A. Development and testing of a general amber force field. J. Comput.
Chem. 2004, 25, 1157−1174.
(67) Morris, G. M.; Huey, R.; Lindstrom, W.; Sanner, M. F.; Belew,
R. K.; Goodsell, D. S.; Olson, A. J. AutoDock4 and AutoDockTools4:
automated docking with selective receptor flexibility. J. Comput. Chem.
2009, 30, 2785−2791.
(68) Mehler, E. L.; Solmajer, T. Electrostatic effects in proteins:
comparison of dielectric and charge models. Protein Eng. 1991, 4,
903−910.
(74) Gilson, M. K.; Sharp, K. A.; Honig, B. H. Calculating the
electrostatic potential of molecules in solution: method and error
assessment. J. Comput. Chem. 1988, 9, 327−335.
(75) Sitkoff, D.; Sharp, K. A.; Honig, B. Accurate calculation of
hydration free energies using macroscopic solvent models. J. Phys.
Chem. 1994, 98, 1978−1988.
(69) (a) Onufriev, A.; Bashford, D.; Case, D. A. Modification of the
generalized Born model suitable for macromolecules. J. Phys. Chem. B
2000, 104, 3712−3720. (b) Feig, M.; Onufriev, A.; Lee, M. S.; Im, W.;
Case, D. A.; Brooks, C. L. Performance comparison of generalized
Born and Poisson methods in the calculation of electrostatic solvation
energies for protein structures. J. Comput. Chem. 2004, 25, 265−284.
(70) Jorgensen, W. L.; Chandrasekhar, J.; Madura, J. D.; Impey, R.
W.; Klein, M. L. Comparison of simple potential functions for
simulating liquid water. J. Chem. Phys. 1983, 79, 926−935.
(71) Ryckaert, J.-P.; Ciccotti, G.; Berendsen, H. J. C. Numerical
integration of the cartesian equations of motion of a system with
constraints: molecular dynamics of n-alkanes. J. Comput. Phys. 1977,
23, 327−341.
(76) Sanner, M. F.; Olson, A. J.; Spehner, J. C. Reduced surface: an
efficient way to compute molecular surfaces. Biopolymers 1996, 38,
305−320.
(77) Wilson, E. B., Decius, J. C., Cross, P. C. Molecular Vibrations;
McGraw-Hill: New York, 1995.
(78) http://www.esteco.com/home/mode_frontier/mode_frontier.
html.
(72) Toukmaji, A.; Sagui, C.; Board, J.; Darden, T. Efficient particle-
mesh Ewald based approach to fixed and induced dipolar interactions.
J. Chem. Phys. 2000, 113, 10913−10927.
(73) For a further selection of successful MM/PBSA applications in
ligand binding of our group see also: (a) Liu, X.; Liu, C.; Laurini, E.;
Posocco, P.; Pricl, S.; Qu, F.; Rocchi, P.; Peng, L. Efficient delivery of
sticky siRNA and potent gene silencing in a prostate cancer model
using a generation 5 triethanolamine-core PAMAM dendrimer. Mol.
Pharm. 2012, 9, 470−481. (b) Karatasos, K.; Posocco, P.; Laurini, E.;
Pricl, S. Poly(amidoamine)-based dendrimer/siRNA complexation
studied by computer simulations: effects of pH and generation on
dendrimer structure and siRNA binding. Macromol. Biosci. 2012, 12,
225−240. (c) Liu, X.; Wu, J.; Yammine, M.; Zhou, J.; Posocco, P.; Viel,
S.; Liu, C.; Ziarelli, F.; Fermeglia, M.; Pricl, S.; Victorero, G.; Nguyen,
C.; Erbacher, P.; Behr, J. P.; Peng, L. Structurally flexible triethanol-
amine core PAMAM dendrimers are effective nanovectors for DNA
transfection in vitro and in vivo to the mouse thymus. Bioconjugate
Chem. 2011, 22, 2461−7243. (d) Barnard, A.; Posocco, P.; Pricl, S.;
Calderon, M.; Haag, R.; Hwang, M. E.; Shum, V. W.; Pack, D. W.;
Smith, D. K. Degradable self-assembling dendrons for gene delivery:
experimental and theoretical insights into the barriers to cellular
uptake. J. Am. Chem. Soc. 2011, 133, 20288−20300. (e) Pierotti, M. A.;
Tamborini, E.; Negri, T.; Pricl, S.; Pilotti, S. Targeted therapy in GIST:
in silico modeling for prediction of resistance. Nat. Rev. Clin. Oncol.
8065
dx.doi.org/10.1021/jm300894h | J. Med. Chem. 2012, 55, 8047−8065