3D pharmacophore models for 1,2,3,4-tetrahydroisoquinoline derivatives acting as anticonvulsant agents

Article abstract of DOI:10.1002/ardp.200600022

A 3D pharmacophore model predicting anticonvulsant activity was obtained for a series of 6,7-dimethoxy-1,2,3,4-tetrahydroisoquinoline derivatives recently disclosed as a new class of non-competitive AMPA receptor antagonists. The training set included 17

Full text of DOI:10.1002/ardp.200600022

388
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
Full Paper
3D Pharmacophore Models for 1,2,3,4-Tetrahydroisoquinoline
Derivatives Acting as Anticonvulsant Agents
Laura De Luca1, Rosaria Gitto1, Maria Letizia Barreca1, Roberta Caruso1, Silvana Quartarone1,
Rita Citraro2, Giovambattista De Sarro2, Alba Chimirri1
1 Dipartimento Farmaco-Chimico, Universitꢀ di Messina, Messina, Italy
2 Dipartimento di Medicina Sperimentale e Clinica, Universitꢀ Magna Græcia, Catanzaro, Italy
A 3D pharmacophore model predicting anticonvulsant activity was obtained for a series of 6,7-
dimethoxy-1,2,3,4-tetrahydroisoquinoline derivatives recently disclosed as a new class of non-
competitive AMPA receptor antagonists. The training set included 17 compounds with varying
potency against audiogenic seizures in DBA/2 mice. The best statistical hypothesis, generated
with the HypoGen module of Catalyst 4.9, consisted of five features: two hydrogen bond accep-
tors, two hydrophobic features, and one hydrophobic aromatic region, providing a model with a
correlation coefficient of 0.919. The obtained model was an efficient tool in the design of some
new anticonvulsant agents containing the tetrahydroisoquinoline scaffold. Moreover, in order
to explain the different degree of efficacy of the newly designed N-substituted derivatives, exclu-
ded volumes were also considered.
Keywords: AMPA receptor / Anticonvulsant agents / HypoGen / Isoquinolines / Pharmacophore model /
Received: February 6, 2006; accepted: March 16, 2006
DOI 10.1002/ardp.200600022
chronic disorders [2, 3]. In particular, some 2,3-benzodia-
Introduction
zepines (1–3, Fig. 1) [4, 5] were identified as marked anti-
epileptic agents in various seizure models interacting
with AMPA receptor complex in a selective and noncom-
petitive fashion. Among them, talampanel 1 has shown
positive results in patients with severe epilepsy not
responsive to other drugs and phase III trials are under
way to confirm these results [6].
AMPARs are constituted as a heteromeric assembly of
the subunits GluR1-4 which, in different combinations,
produce receptors with distinct properties. The AMPAR
competitive agonist/antagonist binding site is located
within the extracellular region of membrane protein. In
particular, the application of X-ray diffraction has eluci-
dated the structure of GluR2 bound with a series of com-
petitive agonists/antagonists, providing some details
about the ligand recognition and the activation-deactiva-
tion mechanism [7, 8]. On the contrary, no X-ray struc-
ture of any complex between noncompetitive antago-
nists and their binding site has been reported as yet and
little information is available about the mechanism and
impact of positive and negative modulator interaction
[9]. Due to this lack of structural information and
Glutamate (Glu), the major excitatory neurotransmitter
in the vertebrate brain, plays an important role in neuro-
nal activity operating through ionotropic (iGluRs) and
metabotropic (mGluRs) receptors. The iGluRs are
involved in many different physiological processes such
as neuronal development, learning, and memory. How-
ever, their excessive activation is held responsible for the
destruction of neuronal cells during several neurological
diseases [1]. iGluRs are divided into N-methyl-D-aspartic
acid (NMDA), 2-amino-3-(3-hydroxy-5-methylisoxazol-4-
yl)propionic acid (AMPA), and kainate (KA) receptors,
based on their pharmacology and structural properties.
It is well known that competitive and noncompetitive
antagonists of AMPA receptor (AMPAR) show to be pro-
mising in terms of their therapeutic potential for the pre-
vention and treatment of a broad range of acute and
Correspondence: Laura De Luca and Rosaria Gitto, University of Messi-
na, Medicinal Chemistry, Viale Annunziata, Messina 98168, Italy.
E-mail: ldeluca@pharma.unime.it, gittoros@pharma.unime.it
Fax: +39 90 355-613
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
389
tures of a series of 1,2,3,4-tetrahydroisoquinoline deriv-
atives, previously designed and synthesized by us [12, 14,
15], to their biological activity. The input for HypoGen
run was a training set (TS) of 17 compounds 4–20
(Table 1).
Table 1 reports the molecules of our TS with the corre-
sponding anticonvulsant activity value against audio-
genic seizures in DBA/2 mice, which has been considered
an excellent animal model for generalized epilepsy and
for screening new anticonvulsant agents.
For our SAR study, we chose to use only the ED50 values
in the clonic phase of seizures, since the activities versus
the clonic and tonic phases were highly correlated
(r2 = 0.901).
All molecular structures were sketched within Catalyst
and minimized to their closest local energy minimum
using a molecular mechanics approach. Poled conforma-
tions were generated for each molecule using the “best”
conformer generation option and an energy cutoff of
10 kcal/mol.
Figure 1. Negative allosteric modulator of the AMPA receptor.
ongoing with our research in this field [10], we have
recently reported a hypothetical 3D ligand-based phar-
macophore model by using the hypothesis generation
approach HipHop/Catalyst [11].
Upon inspection of the structures and on the basis of
our previous studies, it was determined that four generic
types of chemical features described most of the func-
tionalities of TS molecules. Therefore, HypoGen was
instructed to select pharmacophore models using a
hydrogen bond acceptor (HBA), hydrogen bond donor
(HBD), hydrophobic (HY), hydrophobic aromatic (HYAr),
and positive ionizable (P) groups.
During the hypothesis generation process, Catalyst
selects the best hypotheses of many possibilities by apply-
ing a cost analysis.
A set of ten hypotheses was thus generated using the
data from TS compounds having correlation coefficients
ranging from 0.772 to 0.919.
In order to evaluate if the obtained correlations were
true and not due to chance, we have investigated the
three cost values calculated by HypoGen and crucial in
the validation of a reported hypothesis: the fixed cost,
the null cost, and the total cost (section 4, Experimental).
In this study, the fixed cost of 45.6 bits was well sepa-
rated from the null cost of 120.3 bits, and the difference
between these two values of 74.7 bits underscored a high
probability of generating useful models. Moreover, Cata-
lyst documentation suggested that the goal of hypothesis
generation is to obtain a set of hypotheses with total
costs as close as possible to the fixed cost; our ten phar-
macophore models fully satisfied this requirement as the
total costs of the ten top-scored hypotheses range from
66.101 to 86.270.
The aim of this study was to identify the 3D structural
requirements that are relevant in a molecule in order to
noncompetitively interact with AMPAR and to produce
anticonvulsant activity. The obtained common-features
hypothesis, consisting of two hydrophobic regions, one
hydrogen bond acceptor (HBA) and one aromatic region,
was successfully used as a framework for the design of a
new class of noncompetitive AMPAR antagonists and led
to the discovery of 6,7-dimethoxy-1,2,3,4-tetrahydroiso-
quinoline derivatives as new potent anticonvulsant
agents [11]. In particular, the 2-acetyl-1-(4-chlorophenyl)-
6,7-dimethoxy-1,2,3,4-tetrahydroisoquinoline (4, Fig. 1)
was characterized by improved pharmacological effects
when compared in in vivo and in vitro tests with other cur-
rent AMPAR antagonists 1–3 [12].
In this paper, we describe the automated generation of
reliable 3D predictive pharmacophore models for 1,2,3,4-
tetrahydroisoquinoline derivatives, in order to gain
more insight into the structure activity-relationships
(SAR) for this class of anticonvulsant drugs.
Results and discussion
Pharmacophore modeling
We used the algorithm HypoGen implemented in the
Catalyst molecular modeling software [13] with the
efforts to quantitatively correlate the chemical struc-
The above-reported theoretical parameters thus
revealed statistical significance of the optimized models
generated with the HypoGen algorithm.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

390
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
Table 1. Anticonvulsant activity of tetrahydroisoquinolines 4–20 against audiogenic seizures in DBA/2 mice.
Compd.
R1
R2
ED50lmol/kga) (l95% confidence limits)
clonic phase
tonic phase
4b)
5c)
4 – Cl
4 – Br
4.18 (2.23 – 7.84)
12.7 (6.09 – 26.3)
2.39 (1.30 – 4.40)
8.17 (4.03 – 16.6)
6d)
7d)
3 – NO2
4 – Cl
19.3 (6.10 – 61.2)
20.1 (9.65 – 41.9)
7.20 (2.45 – 21.2)
19.3 (11.8 – 31.5)
8c)
4 – Cl
24.7 (13.2 – 46.2)
29.9 (17.5 – 50.9)
32.1 (17.7 – 58.3)
36.8 (20.6 – 65.8)
11.8 (8.10 – 17.0)
17.6 (10.4 – 29.7)
21.1 (11.0 – 40.4)
18.0 (7.76 – 41.8)
9d)
3 – NH2
4 – NH2
4 – F
10b)
11b)
12d)
13c)
H
H
44.8 (23.9 – 84.0)
49.1 (36.9 – 65.5)
19.3 (9.05 – 41.3)
34.8 (19.6 – 61.6)
14b)
H
53.5 (37.6 – 76.2)
37.7 (21.2 – 67.0)
15d)
16c)
2,3 – Cl2
4 – F
63.3 (35.8 – 111)
71.7 (64.6 – 110)
32.8 (17.5 – 61.5)
46.9 (21.3 – 103)
17c)
4 – F
82.8 (54.1 – 126)
63.0 (32.8 – 101)
18d)
19d)
20d)
3 – F
4 – NO2
2,3 – F2
H
H
H
96.9 (60.5 – 155)
100 (52.1 – 194)
130 (74.0 – 233)
53.5 (38.3 – 73.8)
56.1 (26.9 – 117)
48.1 (25.4 – 91.2)
a)
All data were calculated according to the method of Litchfield and Wilcoxon [23]. At least 32 animals were used to calculate each
ED50.
b)
c)
d)
Reference [12].
Reference [15].
Reference [14].
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
391
Table 2. Experimental and estimated ED50 (lmol/kg) of TS
molecules based on the pharmacophore model Hypo1.
Compd.
Experimental
ED50 (lmol/kg)
Estimated
Error
ED50 (lmol/kg) costa)
4
5
6
7
8
4.2
13
19
20
25
30
32
37
45
49
53
63
72
83
97
100
130
4.8
23
+1.1
+1.8
+1
–1.1
–1.3
+1.2
+1.1
+1
+2.1
–1.4
–1.6
–1.3
+1
20
18
19
36
34
38
96
36
34
50
74
50
98
140
97
9
10
11
12
13
14
15
16
17
18
19
20
–1.7
+1
+1.4
–1.3
a)
The error cost values for all TS compounds, expressed as the
ratio between the estimated and experimental activities or
its negative inverse if the ratio is less than one, were found to
be a10, implying not more than one order difference
between estimated and actual activity.
Figure 2. Best HypoGen pharmacophore model Hypo1 aligned
to compound 4. The pharmacophore features are color-coded
as follows: green, hydrogen bond acceptor (HBA1 and HBA2);
blue, hydrophobic aromatic region (HYAr,); cyan, hydrophobic
group (HY1 and HY2).
cating a reliable ability to estimate the biological activity
(Table 2).
The first output hypothesis (Hypo1) showed the best
correlation coefficient (0.92), the lowest root-mean-
square divergence (1.27), and a good cost difference
(54.23) and it was chosen for further analysis.
It is worth noting that Hypo1 describes some chemical
features which were highlighted in our previous publica-
tion [11], confirming that the presence of the hydrogen-
bond acceptor (HBA2), such as an acetyl moiety, and the
Hypo1 consisted of five features: two HBAs, one HYAr, aryl substituent (HYAr) are important for the binding
and two HYs located at defined positions.
Figure 2 shows Hypo1 aligned to the most active com-
interaction with the AMPA receptor complex. In addi-
tion, this hypothesis gave new information about the
pound 4, obtained by using the BestFit compare option of influence of other substituents on the pharmacological
the View Hypothesis Workbench.
profile of tetrahydroisoquinoline derivatives.
Compound 4 mapped well onto the five chemical func-
On the basis of these results and in order to approve
tionalities of the 3D pharmacophore model: particularly, the predictive power of the quantitative model, we
the two hydrogen bond regions were occupied by the car- designed new potential AMPAR ligands (Fig. 3). Using the
bonyl oxygen of the acetyl moiety and the oxygen atom
of methoxy group at C-6, the aryl group filled the aro-
resulting five chemical feature hypothesis and the most
active compound 4 as template, we planned from time to
matic hydrophobic feature, whereas the 7-methoxy time the synthesis of different series of isoquinolines 21,
group and 49-substituent overlapped with the two hydro- 22, 23, 24, 25, 26, and 27 bearing modifications in three
phobic sites.
different positions of tetrahydroisoquinoline skeleton
(see Fig. 3). Furthermore, in order to identify the contri-
This model also accurately estimated the ED50 value of
this molecule (predicted ED50 of 4.18 lmol/kg vs. an bution of the substitution pattern on the C-1 phenyl ring,
experimental value of 4.80 lmol/kg). Moreover, compari- we also synthesized derivatives a–f for each series.
son between the estimated activities and experimentally
measured values for all compounds in the TS showed
At first, we have performed the computational design
of derivatives 21a–f (Fig. 3), in which the 1-aryl-6,7-
that the error costs (expressed as the ratio between the dimethoxyisoquinoline scaffold was maintained,
estimated and the experimental activities or its negative
inverse if the ratio is less than one) were always a10, indi-
whereas the acetyl moiety was removed and a carbonyl
group inserted on the isoquinoline skeleton. In fact, con-
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

392
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
Figure 3. Planned structural modifications.
Scheme 1. Synthesis route for 1-aryl-6,7-dimethoxy-1,2-dihydroisoquinolin-3(4H)-ones 21a–f and 22a–f.
sidering that the presence of a chemical feature able to
engage H-bond interaction (i.e. HBA2) appeared essential
for anticonvulsant activity, we thought that the acetyl
moiety could be suitably replaced by another hydrogen
bond acceptor group with the same spatial location. So,
1-aryl-6,7-dimethoxy-1,2-dihydroisoquinolin-3(4H)-ones
21a–f were synthesized, see section 2.2 (Chemistry) and
Scheme 1, and tested against audiogenic seizures in DBA/
2 mice (Table 3). In this class, compound 21c, which
mapped well onto Hypo1 (Fig. 4), was an efficacious antic-
onvulsant agent with ED50 value of 23.4 lmol/kg for the
clonic phase of the test comparable to that of talampanel
1 (ED50 = 13.4 lmol/kg) and higher than GYKI 52466 3
(ED50 = 35.8 lmol/kg). Biological data for 21c was in per-
fect agreement with the in silico activity (20.0 lmol/kg).
We have also synthesized the 3-acetyl derivatives 22a–f
which proved to be less active than the corresponding N-
unsubstituted analogues 21a–f (Table 3), thus suggesting
that the contemporaneus introduction of the acetyl moi-
ety and the carbonyl function in these compounds is det-
rimental for the anticonvulsant potency.
Figure 4. Compound 21c mapped into HypoGen pharmaco-
phore model Hypo1. The pharmacophore features are labeled
as follow: hydrogen bond acceptor (HBA1 and HBA2); hydro-
phobic aromatic region (HYAr,); hydrophobic group (HY1 and
HY2).
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
393
Table 3. Anticonvulsant activity against audiogenic seizures in
DBA/2 mice.
Compd.
R1
ED50 (lmol/kg)a)
(l95% confidence limits)
clonic phase
tonic phase
21a
21b
21c
21d
21e
21f
22a
22b
22c
22d
22e
22f
H
4–F
A100
A100
70.1 (60.3–91.0)
81.7 (52.5–127)
23.4 (14.4–36.9) 17.6 (12.8–25.6)
4–Cl
4–Br
4–NO2
3–NO2
H
A100
A 100
96.9 (60.5–155)
A100
58.3 (41.3–82.3)
74.1 (50.6–108)
A 100
A100
4–F
A100
92.5 (55.1–155)
63.8 (45.1–90.2)
A 100
4–Cl
4–Br
4–NO2
3–NO2
A100
A100
84.6 (64.5–111)
A100
44.9 (26.6–75.8)
A 100
Figure 5. Structure of compound 24c superimposed on model
Hypo1. The pharmacophore features are labeled as follow:
hydrogen bond acceptor (HBA1 and HBA2); hydrophobic aro-
matic region (HYAr); hydrophobic group (HY1 and HY2).
a)
All data were calculated according to the method of Litch-
field and Wilcoxon [23]. At least 32 animals were used to cal-
culate each ED50.
Secondly, we extended our investigation and decided should show lower efficacy than the corresponding 6,7-
to explore if the two methoxy groups were important for disubstituted derivative 4 (the estimated ED50 value for
anticonvulsant activity as suggested by our predictive compound 24c is 77 lmol/kg), since one chemical feature
pharmocophore model. We have thus studied C-6-mono- of model Hypo1 is not mapped (Fig. 5).
methoxytetrahydroisoquinolines 23a–f and 24a–f
As expected, the anticonvulsant pharmacological
(Fig. 3), where the original feature interacting with the screening confirms our in silico analysis, as 24c (Table 4)
hydrophobic region HY1 was lacking, and synthesized was really less active than analogue 4. The same behavior
them via the route outlined in Scheme 2.
was observed for compounds 23a–f and 24a–f (Table 4),
Compound 24c, which is the analogue of the most which showed generally lower anticonvulsant properties
active TS compound 4 (experimental ED50 = 4.2 lmol/kg), with respect to 6,7-dimethoxy analogues [12, 14].
Scheme 2. Synthesis of C-6-monomethoxytetrahydroisoquinolines 23a–f and 24a–f.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

394
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
Table 4. Anticonvulsant activity against audiogenic seizures in
DBA/2 mice.
our pharmacophore model. In particular Hypo1 sug-
gested that some compounds, such as 25c, 26c, and 27,
could be potential anticonvulsant agents with an esti-
mated ED50 values of 24 lmol/kg, 4.3 lmol/kg and
12 lmol/kg, respectively; the superimposition of their
structures against the 3D hypothesis revealed that the
five features of the pharmacophore were well matched
by the chemical groups of the molecules (Fig. 6). So we
prepared compounds 25a–f, 26a–f, and 27 following the
synthetic pathway outlined in Scheme 3.
Compd.
R1
ED50 (lmol/kg)a)
The results of experimental ED50 values against audio-
genic seizures in DBA/2 mice (Table 5) for 25c
(82.79 lmol/kg), 26c (68.7 lmol/kg), and 27 (57.36 lmol/
kg) were not in agreement with the estimated activities.
Moreover, this new series of N-substituted tetrahydroiso-
quinolines generally demonstrates poor efficacy against
audiogenic seizures.
The discrepancy between predict and actual ED50
values could be explained considering that a long chain
around the carbonyl group might occupy a region where
the increased steric bulk alters the ability of these com-
pounds to bind to the receptor.
Since our Hypo1 model did not take into account steric
factors that might be important for the binding affinity,
we subsequently decided to optimize our pharmaco-
phore-based 3D model by including compounds 25c, 26c,
and 27 in the TS and by using the algorithm Catalyst/
HypoRefine [16], which performs automated generation
of excluded volumes in activity-based hypothesis genera-
tion. In fact, for systems where steric effect plays a role in
modulating biological activity, HypoRefine is able to add
one or more excluded volumes to HypoGen models in
order to improve both the regression coefficients and the
(l95% confidence limits)
clonic phase
84.4 (59.4–121)
tonic phase
23a
23b
23c
23d
23e
23f
24a
24b
24c
24d
24e
24f
H
4–F
35.8 (24.2–53.1)
48.4 (29.7–79.0) 16.0 (9.26–27.5)
4–Cl
4–Br
4–NO2
3–NO2
H
A100
A100
A100
56.5 (38.0–84.0)
3.30 (2.11–5.16)
32.0 (18.7–56.6)
3.60 (2.02–6.41)
A100
32.0 (21.1–48.6)
82.8 (51.2–134)
41.5 (28.2–61.1)
A100
4–F
4–Cl
4–Br
4–NO2
3–NO2
82.5 (63.6–107)
A100
53.4 (36.5–78.2)
A100
64.5 (42.7–97.3) 21.2 (15.1–29.7)
65.9 (39.3–110)
23.8 (15.0–37.9)
a)
All data were calculated according to the method of Litch-
field and Wilcoxon [23]. At least 32 animals were used to cal-
culate each ED50.
Moreover, the HypoGen pharmacophore Hypo1 was
also useful to explore the influence of N-substitution on
biological activity. Thus, we studied the N-substituted
derivatives 25a–f, 26a–f, and 27 (see Fig. 3) having func-
tionalities able to engage hydrogen bond as requested by
Figure 6. Mapping of compounds 25c (a), 26c (b), and 27 (c) onto the best five-feature hypothesis Hypo1. The pharmacophore
features are labeled as follow: hydrogen bond acceptor (HBA1 and HBA2); hydrophobic aromatic region (HYAr,); hydrophobic group
(HY1 and HY2).
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
395
Scheme 3. Synthesis of compounds 25a–f, 26a–f, and 27.
Table 5. Anticonvulsant activity against audiogenic seizures in
DBA/2 mice.
Compd.
R1
ED50 (lmol/kg)a)
(l95% confi-
dence limits)
clonic phase
tonic phase
25a
25b
25c
25d
25e
25f
26a
26b
26c
26d
26e
26f
27
H
4–F
A100
38.9 (24.2–62.4)
43.1 (23.9–77.7)
45.5 (27.8–74.3)
A100
A100
4–Cl
4–Br
4–NO2
3–NO2
H
82.8 (51.2–134)
A100
61.7 (40.8–93.5) 16.4 (10.7–25.3)
53.7 (35.9–80.2) 13.0 (8.26–20.5)
A100
36.1 (18.7–69.5) 10.8 (5.43–21.6)
68.7 (40.5–116) 21.0 (10.3–43.1)
47.2 (32.6–68.2) 13.3 (7.81–22.8)
Figure 7. The top-scoring pharmacophore HypoRef1 is mapped
to the most active compound in the training set (TS). The phar-
macophore features are color-coded as follow: green, hydrogen
bond acceptor (HBA1 and HBA2); blue, hydrophobic aromatic
region (HYAr,); cyan, hydrophobic group (HY1 and HY2); black,
excluded volumes (E1, E2, E3).
68.7 (40.5–116)
4–F
4–Cl
4–Br
4–NO2
3–NO2
– –
A100
A100
A100
A100
57.3 (38.4–85.6) 47.9 (35.9–64.0)
oped HypoRefine models showed correlation coefficients
ranging from 0.619 to 0.945 and a difference between the
fixed and null costs of 85.0 bits.
In particular, the most statistically significant hypoth-
esis (Hypo1Ref: correlation coefficient = 0.945; root-
a)
All data were calculated according to the method of Litch-
field and Wilcoxon [23]. At least 32 animals were used to cal-
culate each ED50.
accuracy of activity predictions, thus providing more rea- mean-square divergence = 1.00; cost difference = 66.380)
listic pharmacophore-based SAR models.
was chosen as the model for further studies.
The algorithm HypoRefine thus allowed the develop-
Hypo1Ref consisted of the same five chemical features
ment of a new set of ten hypotheses using the data from of the previous HypoGen model: two hydrogen-bond
the expanded TS and the same Catalyst parameters as in acceptors (HBA), one hydrophobic aromatic region
the previous hypothesis run. Our automatically devel- (HYAr), and two hydrophobic region (HY); moreover,
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

396
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
Table 6. Actual and estimated ED50 values for the TS molecules
based on pharmacophore model Hypo1Ref.
Compd.
Experimental
ED50 (mmol/kg) (lmol/kg)
Estimated ED50 Error costa)
4
5
6
7
8
4.2
13
19
20
25
30
32
37
45
49
54
57
63
69
72
83
83
97
100
130
4.8
14
+1.2
+1.1
+1.3
–1
25
19
19
25
41
34
93
70
38
43
67
59
68
80
59
95
120
98
–1.3
–1.2
+1.3
–1.1
+2.1
+1.4
–1.4
–1.3
+1.1
–1.2
–1.1
–1
9
10
11
12
13
14
27
15
26c
16
25c
17
18
19
20
Figure 8. Compound 25c aligned onto the lowest cost model
HypoRef1. The pharmacophore features are labeled as follow:
hydrogen bond acceptor (HBA1 and HBA2); hydrophobic aro-
matic region (HYAr,); hydrophobic group (HY1 and HY2);
excluded volumes (E1, E2, E3)
–1.4
–1
+1.2
–1.3
3(2H)-one derivatives 21a–f were further subjected to
reaction with acetic anhydride to afford N-acetyl deriv-
atives 22a–f with good yields.
a)
The error cost values for all TS compounds, expressed as the
ratio between the estimated and experimental activities or
its negative inverse if the ratio is less than one, were found to
be a10 implying a not more than one order difference
between estimated and actual activity.
In Scheme 2, the synthetic route for 1-aryl-6-methoxy-
1,2,3,4-tetrahydroisoquinolines, 23a–f and 24a–f, using
the Bischler-Napieralski method [18] is reported. The first
step was the synthesis of benzamide derivatives 32a–f
starting from 2-(39-methoxyphenyl)ethylamine 30 and
benzoylchloride derivatives 31a–f in the presence of
Na2CO3 solution. By treatment with POCl3, the benza-
mide derivatives cyclized giving isoquinolines 33a–f,
which were easily reduced into 1,2,3,4-tetrahydroisoqui-
nolines 23a–f employing NaBH4. The corresponding N-
acetyl derivatives 24a–f were prepared by reaction with
acetic anhydride.
Finally, as shown in Scheme 3, alkyl N-carboxylate-1-
aryl-6,7-dimethoxy-1,2,3,4-tetrahydroisoquinolines,
25a–f and 26a–f, were prepared starting from com-
pounds 35a–f obtained following the synthetic pathway
previously reported by us [14]. The introduction of alkyl
N-carboxy substituent was realized by reaction with right
alkylchloroformates under basic conditions [19].
Attempting to obtain the 3-chloropropionyl substituted
derivative, compound 35c reacted with 3-chloropropio-
nyl chloride affording only the dehydrohalogenated
product 27.
three excluded volumes (E1, E2, E3) were included in the
3D model. As example, the Hypo1Ref hypothesis is shown
together with the structure of the most active compound
of the TS, 4 (Fig. 7).
Using HypoRef1, the anticonvulsant activities of the
twenty TS compounds were predicted correctly within
one order of magnitude (Table 6).
In particular, the ED50 values of 25c, 26c, and 27 were
well estimated with respect to the previous hypothesis
Hypo1. In fact, as shown for compound 25c in Fig. 8, even
though these compounds had the five required chemical
groups for high potency, the inclusion of the three
excluded volume spheres prevented them from match-
ing all pharmacophore features, thus explaining their
lower activity.
Chemistry
Synthetic approaches are outlined in Schemes 1-3. Suita-
ble methods were selected to obtain the varied classes of
designed isoquinolines starting from different reagents.
The 1-aryl-6,7-dimethoxy-1,2-dihydroisoquinolin-3(4H)-
ones 21a–f were obtained, according to the literature
[17], by condensation of the (3,4-dimethoxyphenyl)aceto-
nitrile 28 with suitable aromatic aldehydes 29a–f under
acid catalysis conditions (Scheme 1). The isoquinolin-
Conclusions
This paper describes the generation of a quantitative
pharmacophore model for tetrahydroisoquinoline deriv-
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
397
structure data and the average of all activities in the TS is used as
the estimate), both expressed in bits unit. In general, the greater
the difference between these cost values, the higher the prob-
ability of generating useful models.
atives acting as anticonvulsant agents. The best HypoGen
hypothesis allowed the design of new isoquinoline deriv-
atives (e.g. 21c) able to prevent audiogenic seizures in
DBA/2 mice at dose comparable to other known AMPAR
noncompetitive antagonists. The synthetic approaches
and structural properties have also been reported for all
the designed molecules.
Even if none of the new compounds show activity
higher than the “lead compound” 4, this study suggested
interesting structure-activity relationships for this class
of molecules. In particular, the most important finding
was to demonstrate the discrepancy between predict and
experimental ED50 values for some N-substituted deriv-
atives, by using the excluded volumes approach.
The cost value of each optimized pharmacophore should lie
somewhere between these two values. Moreover, the difference
between the null cost hypothesis and the total hypothesis cost is
of particular importance. A true correlation in the data will very
likely be estimated by models that exhibit a cost difference (null
cost-total cost) of 60 bits or higher. On the other hand, Catalyst
documentation suggested that the difference between fixed and
null costs should be 70 or higher to achieve this request.
In addition to the cost analysis, two other parameters are
involved in establishing the hypothesis significance: RMS (root-
mean-square divergence) represents the deviation of the log (esti-
mated activities) from the log (measured activities) normalized
by the log (uncertainties) and indicates the quality of “predic-
tion” for the training set, while the correlation (r) value is the
linear regression derived from the geometric fit index.
In the development of 3D hypothesis by Catalyst Hyporefine,
the differences in the steric bulk between the most active and
the inactive compounds are labeled in the ”principal” column of
the input spreadsheet; for our study; we chose 4 as the active
compound (labeled with digit 2) and the compounds with activ-
ities higher than 65 lmol/kg as the inactive ones (labeled with
digit 1).
Financial support for this research by MIUR (COFIN2004,
Roma, Italy), Fondo di Ateneo per la Ricerca (Universitꢀ di Mes-
sina) and Fondo per gli Investimenti della Ricerca di Base
(FIRB 2003) is gratefully acknowledged.
Experimental
Generation of HypoGen-based pharmacophore
models
Chemistry
Melting points were determined on a Stuart SMP10 apparatus
(Dynalab Corp. Rochester, NY, USA) and are uncorrected. Ele-
mental analyses (C, H, N) were carried out on a Carlo Erba Model
1106 Elemental Analyzer (Carlo Erba, Milan. Italy) and the
results are within l0.4% of the theoretical values. Merck silica
gel 60 F254 plates were used for analytical TLC; column chromato-
graphy was performed on Merck silica gel 60 (70–230 mesh;
Merck, Darmstadt, Germany). 1H-NMR spectra were measured in
CDCl3 or DMSO-d6 with a Varian Gemini 300 spectrometer (Var-
ian Inc., Palo Alto, CA, USA); chemical shifts are expressed in d
(ppm) relative to TMS as internal standard and coupling con-
stants (J) in Hz. All exchangeable protons were confirmed by
addition of D2O.
Pharmacophores have been generated with the HypoGen mod-
ule of Catalyst 4.9 [13] using a TS of 17 1,2,3,4-tetrahydroisoqui-
noline derivatives.
All molecular structures were sketched within Catalyst and
minimized to their closest local energy minimum using a mole-
cular mechanics approach. Poled conformations were generated
for each molecule using the “best” conformer generation option
and an energy cutoff of 10 kcal/mol.
Five chemical feature types were used in the HypoGen run:
hydrogen bond acceptor (HBA), hydrogen bond donor (HBD),
hydrophobic (HY), hydrophobic aromatic (HYAr) and positive
ionizable (P) groups.
No constraint on the minimum and maximum number of
each type of feature in the reported pharmacophore was
applied. In the model generation methodology, the interfeature
spacing penality was reduced from its default value to 50 pm. In
fact, the default setting of 297 pm is often applicable to large,
more flexible molecules, while small, rigid molecules require a
lower value. In particular, we reduced the minimum permitted
interfeature spacing in order to enable Catalyst to consider fea-
tures of the TS compounds with a minimum distance of 50 pm
as two distinct functions.
The uncertainty value of the compounds activity, which repre-
sents the ratio range of uncertainty in the activity value based
on the expected statistical variability of biological data collec-
tion, was set to 1.3.
All other parameters were kept to their default value.
During the hypothesis generation process, Catalyst selects the
best hypothesis from the various possibilities by applying a cost
analysis. In fact, HypoGen provides two important theoretical
parameters (named “cost values”), that have to be investigated in
order to valuate the simplicity of the models and their statistical
significance: the fixed cost (cost of a perfect hypothesis) and the
null cost (cost of a hypothesis for which we assume there is no
General procedure for the synthesis of 1-aryl-6,7-
dimethoxy-1,2-dihydroisoquinolin-3(4H)-ones, 21a–f
(3,4-Dimethoxyphenyl)acetonitrile 28 (1 g, 5.6 mmol) was mixed
with polyphosphoric acid (3.50 g) at room temperature and
heated at 808C. After addition of the aldehyde derivatives 29a–f
(4.2 mmol) the mixture was heated at 1008C for 2 h and then
quenched by adding water. After stirring at room temperature
for 1 h, the residue was recovered by filtration and treated with
water made basic with ammonium hydroxide. The suspension
was heated for 1 h and left overnight at room temperature. The
precipitate was filtered off and purified by crystallization with
ethyl acetate. By employing the same procedure, compounds
21a and 21c–f were previously synthesized [20].
1-(49-Fluorophenyl)-6,7-dimethoxy-1,2-dihydroiso-
quinoline-3(4H)-one, 21b
1
Mp. 185–1878C. Yield 45%. H-NMR d: 3.65 (s, 2H, CH2), 3.71 (s,
3H, MeO-6), 3.89 (s, 3H, MeO-7), 5.58 (s, 1H, H-1), 6.23 (s, 1H, NH),
6.34 (s, 1H, H-5), 6.65 (s, 1H, H-8), 7.03–7.24 (m, 4H, Ar). Anal.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

398
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
calcd. for C17H16FNO3: C, 67.76; H, 5.53; N, 4.65. Found: C, 67.96;
H, 5.43; N, 4.41.
General procedure for the synthesis of 1-aryl-6-
methoxy-1,2,3,4-tetrahydroisoquinolines, 23a–f
To a stirred mixture of 2-(39-methoxyphenyl)ethylamine 30
(1.51 g, 10 mmol) in CHCl3 and 20% aq. K2CO3 (10 mL), a solution
of the suitable benzoylchloride 31a–f (11 mmol) in CHCl3
(10 mL) was added drop-wise. The reaction mixture was stirred
at room temperature for 3 h and the organic layer was washed
consecutively with H2O, 5% HCl, 20% aq. NaHCO3, H2O, dried
over Na2SO4, and finally the solvent was removed in vacuo. The
crude product was crystallized by adding a small amount of
diethyl ether to give the N-[2-(39-methoxyphenyl)-ethyl]benza-
mide derivatives 32a–f as colorless needles. A mixture of the
obtained benzamide 32a–f (5.3 mmol), POCl3 (0.724 mL,
7.7 mmol) and dry toluene was refluxed for 60 min. The solvent
and POCl3 were removed under reduced pressure. After addition
of 10% NH4OH, the product was taken up in AcOEt and the
organic layer was washed with H2O, dried over Na2SO4, and then
concentrated in vacuo. Crystallization of the crude product from
diethyl ether gave 1-aryl-3,4-dihydro-6-methoxyisoquinoline
33a–f as colorless prisms. To a stirred solution of the obtained
dihydroisoquinoline 33a–f (5.5 mmol) in MeOH (20 mL), NaBH4
(0.224 g, 5.9 mmol) was added and the reaction mixture was stir-
red for 1 h at room temperature. The reaction mixture was
quenched by adding H2O, extracted with AcOEt, dried over
Na2SO4, and then concentrated at reduced pressure. The crude
product was crystallized by adding a small amount of diethyl
ether to give 1-aryl-6-methoxy-1,2,3,4-tetrahydroisoquinoline
23b, 23e–f as colorless needles. A similar synthetic procedure
was previously used for the synthesis of compounds 23a and
23c–d [18].
General procedure for the synthesis of 2-acetyl-1-
aryl-6,7-dimethoxy-1,2-dihydroisoquinolin-3(4H)-
ones, 22a–f
A solution of 1-aryl-6,7-dimethoxy-1,2-dihydroisoquinolin-3(4H)-
ones 21a–f (1 mmol) in acetic anhydride (7–8 mL) was refluxed
for 90 min. After cooling, the reaction mixture was quenched by
adding H2O, and the organic layer was extracted with CHCl3,
dried over Na2SO4, and then concentrated in vacuo. The crude
product was crystallized by adding a small amount of diethyl
ether. Compounds 22a was also previously obtained using a
similar synthetic procedure [20].
2-Acetyl-1-(49-fluorophenyl)-6,7-dimethoxy-1,2-
dihydroisoquinoline-3(4H)-one, 22b
1
Mp. 160–1648C. Yield 70%. H-NMR d: 2.65 (s, 3H, MeCO), 3.50
(AB system, 2H, J = 18.9, CH2), 3.90 (s, 3H, MeO-6), 3.91 (s, 3H,
MeO-7), 6.73 (s, 1H, H-5), 6.84 (s, 1H, H-8), 6.91 (s, 1H, H-1), 6.95–
7.01 (m, 4H, Ar). Anal. calcd. for C19H18FNO4: C, 66.46; H, 5.28; N,
4.08. Found: C, 66.44; H, 5.25; N, 4.12.
2-Acetyl-1-(49-chlorophenyl)-6,7-dimethoxy-1,2-
dihydroisoquinoline-3(4H)-one, 22c
1
Mp. 166–1708C. Yield 70%. H-NMR d: 2.65 (s, 3H, MeCO), 3.50
(AB system, 2H, J = 18.9, CH2), 3.90 (s, 3H, MeO-6), 3.91 (s, 3H,
MeO-7), 6.73 (s, 1H, H-5), 6.82 (s, 1H, H-8), 6.90 (s, 1H, H-1), 6.98 (d,
2H, J = 8.5, H29-H69), 7.23 (d, 2H, J = 8.5, H39-H59). Anal. calcd. for
C19H18ClNO4: C, 63.42; H, 5.04; N, 3.89. Found: C, 63.22; H, 5.21;
N, 3.90.
1-(49-Fluorophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 23b
Mp. 94–988C. Yield 53%. 1H-NMR d: 2.81–3.22 (m, 2H, CH2-CH2),
3.78 (s, 3H, MeO-6), 5.03 (s, 1H, H-1), 6.63 (m, 3H, H-7, H-8, H-5),
6.97–7.23 (m, 4H, Ar). Anal. calcd. for C16H16FNO: C, 74.69; H,
6.25; N, 5.44. Found: C, 74.41; H, 6.02; N, 5.79.
2-Acetyl-1-(49-bromophenyl)-6,7-dimethoxy-1,2-
dihydroisoquinoline-3(4H)-one, 22d
Mp. 175–1778C. Yield 68%.1H-NMR d: 2.65 (s, 3H, MeCO), 3.50 (AB
system, 2H, J = 18.9, CH2), 3.90 (s, 3H, MeO-6), 3.91 (s, 3H, MeO-7),
6.73 (s, 1H, H-5), 6.80 (s, 1H, H-8), 6.92 (d, 2H, J = 8.5, H29-H69), 7.39
(d, 2H, J = 8.5, H39-H59). Anal. calcd. for C19H18BrNO4: C, 56.45; H,
4.49; N, 3.46. Found: C, 56.41; H, 4.51; N, 3.44.
1-(39-Nitrophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 23e
Mp. 52–548C. Yield 61%. 1H-NMR d: 2.84–3.21 (m, 2H, CH2-CH2),
3.80 (s, 3H, MeO-6), 5.17 (s, 1H, H-1), 6.61–6.71 (m, 3H, H-8, H-5, H-
7), 7.49 (t, 1H, J = 7.7, H-5'), 7.64 (d, 1H, J = 7.7, H-69), 8.13 (s, 1H, H-
29), 8.17 (d, 1H, J = 8.0, H-49). Anal. calcd. for C16H16N2O3: C, 67.59;
H, 5.67; N, 9.85. Found: C, 67.33; H, 5.83; N, 9.62.
2-Acetyl-1-(39-nitrophenyl)-6,7-dimethoxy-1,2-
dihydroisoquinoline-3(4H)-one, 22e
1
Mp. 208–2108C. Yield 55%. H-NMR d: 2.69 (s, 3H, MeCO), 3.53
(AB system, 2H, J = 18.9, CH2), 3.92 (s, 3H, MeO-6), 3.94 (s, 3H,
MeO-7), 6.76 (s, 1H, H-5), 6.90 (s, 1H, H-8), 6.94 (s, 1H, H-1), 7.47–
8.19 (m, 4H, Ar). Anal. calcd. for C19H18N2O6: C, 61.62; H, 4.90; N,
7.56. Found: C, 61.60; H, 4.93; N, 7.52.
1-(49-Nitrophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 23f
Mp. 67–708C. Yield 60%. 1H-NMR d: 2.85–3.20 (m, 2H, CH2-CH2),
3.79 (s, 3H, MeO-6), 5.16 (s, 1H, H-1), 6.57-6.68 (m, 3H, H-8, H-5, H-
7), 7.46 (d, 2H, J = 8.5, H29-H69), 8.18 (d, 2H, J = 8.2, H39-H59). Anal.
calcd. for C16H16N2O3: C, 67.59; H, 5.67; N, 9.85. Found: C, 67.23;
H, 5.85; N, 9.57.
2-Acetyl-1-(49-nitrophenyl)-6,7-dimethoxy-1,2-
dihydroisoquinoline-3(4H)-one, 22f
1
Mp. 180–1828C. Yield 45%. H-NMR d: 2.68 (s, 3H, MeCO), 3.50
General procedure for the synthesis of 2-acetyl-1-
aryl-6-methoxy-1,2,3,4-tetrahydroisoquinolines,
24a–f
A solution of 1-aryl-6-methoxy-1,2,3,4-tetrahydroisoquinolines
23a–f (1 mmol) in acetic anhydride (7–8 mL) was refluxed for
(AB system, 2H, J = 18.9, CH2), 3.91 (s, 3H, MeO-6), 3.93 (s, 3H,
MeO-7), 6.75 (s, 1H, H-5), 6.90 (s, 1H, H-8), 6.94 (s, 1H, H-1), 7,25 (d,
2H, J = 8.8, H29-H69), 8.14 (d, 2H, J = 8.8, H39-H5k). Anal. calcd. for
C19H18N2O6: C, 61.62; H, 4.90; N, 7.56. Found: C, 61.61; H, 4.85; N,
7.60.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
3D-QSAR Pharmacophore Models for Tetrahydroisoquinolines
399
90 min and then cooled, the reaction was quenched by adding
H2O. The organic layer, extracted with CHCl3, was dried over
Na2SO4 and then concentrated in vacuo. The crude product was
crystallized by adding a small amount of diethyl ether.
General procedure for the synthesis of alkyl-1-aryl-
6,7-dimethoxy-1,2,3,4-tetrahydroisoquinoline-2-
carboxylate, 25a–f and 26a–f
To a solution of 1-aryl-6,7-dimethoxy-1,2,3,4-tetrahydroisoquino-
line derivative 35a–f (2 mmol) in CHCl3 were added triethyla-
mine (0.28 mL, 2 mmol) and the suitable alkyl chloroformate
(2 mmol). The reaction mixture was stirred at room temperature
for 90 min and then concentrated in vacuo. The crude product
was crystallized by adding a small amount of diethyl ether. Com-
pound 25a was previously obtained by other authors but using a
different synthetic approach [21]. Compounds 26a–d and 26f
were obtained according to the literature method [19].
2-Acetyl-6-methoxy-1-phenyl-1,2,3,4-
tetrahydroisoquinoline, 24a
Mp. 105–1128C. Yield 60%. 1H-NMR d: 2.17 (s, 3H, -COCH3), 2.77–
3.69 (m, 2H, CH2-CH2), 3.81 (s, 3H, MeO-6), 6.73 (m, 1H, H-5), 6.75–
6.79 (m, 1H, H-8), 6.89 (s, 1H, H-1), 7.00 (s, 1H, H-7), 7.19–7.31 (m,
5H, Ar). Anal. calcd. for C18H19NO2: C, 76.84; H, 6.81; N, 4.98.
Found: C, 76.53; H, 6.97; N, 5.21.
Methyl-1-(49-fluorophenyl)-6,7-dimethoxy-1,2,3,4-
2-Acetyl-1-(49-fluorophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 24b
tetrahydroisoquinoline-2-carboxylate, 25b
Mp. 69–738C. Yield 48%. 1H-NMR d: 2.62–3.83 (m, 4H, CH2-CH2),
3.62 (s, 3H, MeO-6), 3.65 (s, 3H, COOMe), 3.73 (s, 3H, MeO-7), 6.20
(s, 1H, H-5), 6.69 (s, 1H, H-8), 6.79 (s, 1H, H-1), 7.09–7.20 (m, 4H,
Ar). Anal. calcd. for C19H20FNO4: C, 67.08; H, 5.84; N, 4.06. Found:
C, 66.82; H, 5.51; N, 4.29.
1
Mp. 94–988C. Yield 55%. H-NMR d: 2.17 (s, 3H, -COCH3), 2.77–
3.79 (m, 2H, CH2-CH2), 3.82 (s, 3H, MeO-6), 6.72 (m, 1H, H-5), 6.75–
6.79 (m, 1H, H-8), 6.85 (s, 1H, H-1), 6.91-7.21 (m, 5H, Ar & H-7).
Anal. calcd. for C18H18FNO2: C, 72.22; H, 6.06; N, 4.68. Found: C,
71.86; H, 6.41; N, 4.51.
2-Acetyl-1-(49-chlorophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 24c
Methyl-1-(49-chlorophenyl)-6,7-dimethoxy-1,2,3,4-
tetrahydroisoquinoline-2-carboxylate, 25c
1
1
Mp. 76–798C. Yield 58%. H-NMR d: 2.17 (s, 3H, -COCH3), 2.77–
Mp. 101–1048C. Yield 65%. H-NMR d: 2.68–3.81 (m, 4H, CH2-
3.78 (m, 2H, CH2-CH2), 3.81 (s, 3H, MeO-6), 6.73 (s, 1H, H-5), 6.75–
6.79 (m, 1H, H-8), 6.84 (s, 1H, H-1), 6.97 (s, 1H, H-7), 7.15 (d, 2H, J =
8.5, H29-H69), 7.23 (d, 2H, J = 8.2, H39-H59). Anal. calcd. for
C18H18ClNO: C, 68.46; H, 5.75; N, 4.44. Found: C, 68.32; H, 5.96; N,
4.21.
CH2), 3.63 (s, 3H, MeO-6), 3.65 (s, 3H, COOMe), 3.73 (s, 3H, MeO-7),
6.19 (s, 1H, H-5), 6.72 (s, 1H, H-8), 6.80 (s, 1H, H-1), 7.16 (d, 2H, J =
8.2, H29-H69), 7.37 (d, 2H, J = 8.2, H39-H59). Anal. calcd. for
C19H20CNO4: C, 63.07; H, 5.57; N, 3.87. Found: C, 63.32; H, 5.28; N,
3.55.
2-Acetyl-1-(49-bromophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 24d
Methyl-1-(49-bromophenyl)-6,7-dimethoxy-1,2,3,4-
tetrahydroisoquinoline-2-carboxylate, 25d
1
Mp. 85–888C. Yield 62%. H-NMR d: 2.17 (s, 3H, -COCH3), 2.76–
1
Mp. 109–1118C. Yield 58%. H-NMR d: 2.67–3.82 (m, 4H, CH2-
3.72 (m, 2H, CH2-CH2), 3.81 (s, 3H, MeO-6), 6.72 (m, 1H, H-5), 6.75–
6.79 (m, 1H, H-8), 6.82 (s, 1H, H-1), 6.97 (s, 1H, H-7), 7.09 (d, 2H, J =
8.5, H29-H69), 7.38 (d, 2H, J = 8.2, H39-H59). Anal. calcd. for
C18H18BrNO2: C, 60.10; H, 5.04; N, 3.89. Found: C, 59.89; H, 5.27;
N, 3.78.
CH2), 3.63 (s, 3H, MeO-6), 3.65 (s, 3H, COOMe), 3.73 (s, 3H, MeO-7),
6.17 (s, 1H, H-5), 6.72 (s, 1H, H-8), 6.79 (s, 1H, H-1), 7.10 (d, 2H, J =
8.5, H29-H69), 7.50 (d, 2H, J = 8.5, H39-H59). Anal. calcd. for
C19H20BrNO4: C, 56.17; H, 4.96; N, 3.45. Found: C, 56.48; H, 4.41;
N, 3.72.
2-Acetyl-1-(39-nitrophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 24e
Methyl-1-(39-nitrophenyl)-6,7-dimethoxy-1,2,3,4-
tetrahydroisoquinoline-2-carboxylate, 25e
Mp. 62–638C. Yield 60%. 1H-NMR: 2.20 (s, 3H, -COCH3), 2.80–3.78
(m, 2H, CH2-CH2), 3.83 (s, 3H, MeO-6), 6.75 (m, 1H, H-5), 6.76–6.82
(m, 1H, H-8), 6.90 (s, 1H, H-1), 7.00 (s, 1H, H-7), 7.46 (t, 1H, J = 8.0,
H-59), 7.70 (s, 1H, H-69), 7.96 (s, 1H, H-29), 8.09 (d, 1H, J = 8.0, H-49).
Anal. calcd. for C18H18N2O4: C, 66.25; H, 5.56; N, 8.58. Found: C,
66.13; H, 5.82; N, 8.79.
1
Mp. 155–1578C. Yield 58%. H-NMR d: 2.66–3.84 (m, 4H, CH2-
CH2), 3.63 (s, 3H, MeO-6), 3.67 (s, 3H, COOMe), 3.75 (s, 3H, MeO-7),
6.33 (s, 1H, H-5), 6.83 (s, 2H, H-8, H-1), 7.59–7.63 (m, 2H, H-5‘, H-
69), 7.97 (s, 1H, H-29), 8.13 (d, 1H, J = 5.50, H-49). Anal. calcd. for
C19H20N2O6: C, 61.28; H, 5.41; N, 7.52. Found: C, 61.05; H, 5.12; N,
7.83.
2-Acetyl-1-(49-nitrophenyl)-6-methoxy-1,2,3,4-
tetrahydroisoquinoline, 24f
Methyl-1-(49-nitrophenyl)-6,7-dimethoxy-1,2,3,4-
tetrahydroisoquinoline-2-carboxylate, 25f
1
Mp. 65–688C. Yield 65%. H-NMR d: 2.20 (s, 3H, -COCH3), 2.80–
1
3.78 (m, 2H, CH2-CH2), 3.83 (s, 3H, MeO-6), 6.75 (m, 1H, H-5), 6.76–
6.82 (m, 1H, H-8), 6.90 (s, 1H, H-1), 7.00 (s, 1H, H-7), 7.46 (d, 2H, J =
8.5, H29-H69), 8.18 (d, 2H, J = 8.2, H39-H59). Anal. calcd. for
C18H18N2O4: C, 66.25; H, 5.56; N, 8.58. Found: C, 66.02; H, 5.34; N,
8.47.
Mp. 157–1608C. Yield 48%. H-NMR d: 2.65–3.84 (m, 4H, CH2-
CH2), 3.65 (s, 3H, MeO-6), 3.66 (s, 3H, COOMe), 3.74 (s, 3H, MeO-7),
6.29 (s, 1H, H-5), 6.82 (s, 2H, H-8, H-1), 7.46 (d, 2H, J = 8.5, H29-H69),
8.17 (d, 2H, J = 8.5, H39-H59). Anal. calcd. for C19H20N2O6: C, 61.28;
H, 5.41; N, 7.52. Found: C, 61.12; H, 5.22; N, 7.31.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com

400
L. De Luca et al.
Arch. Pharm. Chem. Life Sci. 2006, 339, 388–400
pound administered, and dose-response curves were fitted using
a computer program by Litchfield and Wilcoxon's method [23].
Ethyl-1-(39-nitrophenyl)-6,7-dimethoxy-1,2,3,4-
tetrahydroisoquinoline-2-carboxylate, 26e
1
Mp. 66–708C. Yield 59%. H-NMR d: 1.21 (t, 3H, J = 6.9, -CH2CH3),
2.72–3.85 (m, 4H, CH2-CH2), 3.62 (s, 3H, MeO-6), 3.75 (s, 3H, MeO-
7), 4.10 (q, 2H, J = 6.9, -CH2CH3), 6.32 (s, 1H, H-5), 6.83 (s, 2H, H-8, H-
1), 7.61–7.62 (m, 2H, H-59, H-69), 8.00 (s, 1H, H-29), 8.11–8.15 (m,
1H, H-49). Anal. calcd. for C20H22N2O6: C, 62.17; H, 5.74; N, 7.25.
Found: C, 62.25; H, 5.43; N, 7.32.
References
[1] R. Dingledine, K. Borges, D. Bowie, S. F. Traynelis, Pharma-
col. Rev. 1999, 51, 7–61.
[2] R. Gitto, M. L. Barreca, L. De Luca, A. Chimirri, Expert
Opin. Ther. Patents 2005, 14, 1199–1213.
Synthesis of 1-(1-(49-chlorophenyl)-6,7-dimethoxy-
1,2,3,4-tetrahydroisoquinolinyl)prop-2-en-1-one, 27
To a solution of 1-(49-chlorophenyl)-6,7-dimethoxy-1,2,3,4-tetra-
hydroisoquinoline 35c (0.8 g, 2.64 mmol) in CHCl3 were added
triethylamine (1.1 mL, 2.64 mmol) and 3-chloropropionyl chlor-
ide (0.25 mL, 2.64 mmol). The reaction mixture was stirred at
room temperature for 90 min and then concentrated in vacuo.
The crude product was crystallized by adding a small amount of
diethyl ether. Mp. 100–1028C. Yield 50%. 1H-NMR d: 2.38–3.50
(m, 4H, CH2-CH2), 3.77 (s, 3H, MeO-6), 3.89 (s, 3H, MeO-7), 5.75 (m,
1H, CH=CH2a), 6.45 (m, 1H, CH=CH2b), 6.51 (s, 1H, H-5), 6.61 (m, 1H,
CH=CH2), 6.66 (s, 1H, H-8), 6.89 (s, 1H, H-1), 7.19–7.24 (m, 4H, Ar).
Anal. calcd. for C20H20ClNO3: C, 67.13; H, 5.63; N, 3.91. Found: C,
67.25; H, 5.41; N, 3.80.
[3] G. De Sarro, R. Gitto, E. Russo, G. F. Ibbadu, et al., Curr.
Top. Med. Chem. 2005, 5, 31–42.
[4] A. Chimirri, G. De Sarro, A. De Sarro, R. Gitto, et al., J.
Med. Chem. 1997, 40, 1258–1269.
[5] R. Gitto, V. Orlando, S. Quartarone, G. De Sarro, et al., J.
Med. Chem. 2003, 46, 3758–3761.
[6] F. Andrꢀsi, Drugs Future 2001, 26, 754–756.
[7] N. Armstrong, E. Gouaux, Neuron 2000, 28, 165–181.
[8] Y. P. Auberson, Drugs Future 2001, 26, 463–471.
[9] L. De Luca, A. Macchiarulo, G. Costantino, M. L. Barreca,
et al., Farmaco 2003, 58, 107–113.
[10] A. Chimirri, G. De Sarro, S. Quartarone, M. L. Barreca, et
al., Pure Appl. Chem. 2004, 76, 931–939.
Pharmacology
Testing of anticonvulsant activity
[11] M. L. Barreca, R. Gitto, S. Quartarone, L. De Luca, et al., J.
Chem. Inf. Comput. Sci. 2003, 43, 651–655.
All experiments were performed with DBA/2 mice which are
genetically susceptible to sound-induced seizures. DBA/2 mice
(8–12 g; 22–25 days-old) were purchased from Harlan Italy (Cor-
ezzano, Italy). Groups of ten mice of either sex were exposed to
auditory stimulation 30 min following administration of vehi-
cle or each dose of drugs studied [22]. The compounds were given
intraperitoneally (ip.) (0.1 mL/10 g of body weight of the mouse)
as a freshly-prepared solution in 50% dimethylsulfoxide (DMSO)
and 50% sterile saline (0.9% NaCl). Individual mice were placed
under a hemispheric perspex dome (diameter 58 cm), and 60 s
were allowed for habituation and assessment of locomotor activ-
ity. Auditory stimulation (12–16 kHz, 109 dB) was applied for
60 s or until tonic extension occurred, and induced a sequential
seizure response in control DBA/2 mice, consisting of an early
wild running phase, followed by generalized myoclonus and
tonic flexion and extension sometimes followed by respiratory
arrest. The control and drug-treated mice were scored for latency
to and incidence of the different phases of the seizures.
[12] R. Gitto, M. L. Barreca, L. De Luca, G. De Sarro, et al., J.
Med. Chem. 2003, 46, 197–200.
[13] Catalyst, 4.9, Accelrys Inc., San Diego, CA, USA.
[14] R. Gitto, R. Caruso, V. Orlando, S. Quartarone, et al., Farm-
aco 2004, 59, 7–12.
[15] M. L. Barreca, R. Caruso, L. De Luca, R. Gitto, et al., 14th
Camerin–Noordwijkerouth Symposium: Ongoing pro-
gress in the receptor chemistry Proceedings 2003, 84.
[16] http://www.accelrys.com/catalyst/hyporefine.htmL.
[17] G. Deak, K. Gall–Istok, L. Sterk, Acta Chim. Acad. Sci.
Hung. 1976, 88, 87–92.
[18] S. Doi, N. Shirai, Y. Sato, J. Chem. Soc., Perkin Trans I 1997,
15, 2217–2221.
[19] A. Venkov, N. Mollov, Dokle Bulg. Akad. Nauk. 1979, 32,
The experimental protocol and all the procedures involving
animals and their care were conducted in conformity with the
institutional guidelines and the European Council Directive of
laws and policies.
895–897.
[20] V. St Georgiev, R. G. Van Inwegen, P. Carlson, Eur. J. Med.
Chem. 1990, 25, 375–378.
[21] G. K. Cheung, M. J. Earle, R. A. Fairhurst, H. Heaney, et al.,
Synlett 1991, 10, 721–723.
Statistical analysis
[22] A. G. Chapman, M. J. Croucher, B. S. Meldrum, Arzneimit-
telforschung 1984, 34, 1261–1264.
Statistical comparisons between groups of control and drug-
treated animals were made using Fisher's exact probability test
(incidence of the seizure phases). The ED50 values of each phase
of audiogenic seizures was determined for each dose of com-
[23] J. T. Litchfield, F. Wilcoxon, J. Pharmacol. Exp. Ther. 1949,
96, 99–113.
i 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.archpharm.com