Ligand design for alpha1 adrenoceptor subtype selective antagonists.

Article abstract of DOI:10.1016/S0968-0896(99)00263-1

Alpha1 adrenoceptors have three subtypes and drugs interacting selectively with these subtypes could be useful in the treatment of a variety of diseases. In order to gain an insight into the structural principles governing subtype selectivity, ligand base

Full text of DOI:10.1016/S0968-0896(99)00263-1

Bioorganic & Medicinal Chemistry 8 (2000) 201±214
Ligand Design for ꢀ1 Adrenoceptor Subtype Selective Antagonists
John B. Bremner, a,* Burak Coban, a,y Renate Grith, a
Karina M. Groenewoud b and Brian F. Yates b
aDepartment of Chemistry, University of Wollongong, North®elds Ave, Wollongong, NSW 2522, Australia
bDepartment of Chemistry, University of Tasmania, Hobart, TAS 7005, Australia
Received 19 March 1999; accepted 15 September 1999
AbstractÐa1 Adrenoceptors have three subtypes and drugs interacting selectively with these subtypes could be useful in the treat-
ment of a variety of diseases. In order to gain an insight into the structural principles governing subtype selectivity, ligand based
drug design (pharmacophore development) methods have been used to design a novel 1,2,3-thiadiazole ring d analogue of the
aporphine system. Synthesis and testing of this compound as a ligand on cloned and expressed human a1 adrenoceptors is descri-
bed. Low binding anity was found, possibly due to an unfavourable electrostatic potential distribution. Pharmacophore models
for antagonists at the three adrenoceptor sites (a1A, a1B, a1D) were generated from a number of di€erent training sets and their
value for the design of new selective antagonists discussed. The ®rst preliminary antagonist pharmacophore model for the a1D
adrenoceptor subtype is also reported. # 2000 Elsevier Science Ltd. All rights reserved.
Introduction
(100-fold over a1A, Table 1). For the a1D subtype the
situation is not much better, with the highest selectivity
of about 200-fold over a1A binding, being reported for a
compound from the group at Synaptic Pharmaceutical
Corporation9 (25, Table 1).
The a1 adrenoceptors (a1 ARs) mediate many e€ects of
the sympathetic nervous system. Like other adrenergic
receptors, a1 ARs are activated by the catecholamines,
adrenaline and noradrenaline.1 The a1 ARs are mem-
brane proteins and members of the G-protein coupled
receptor superfamily. Three distinct subtypes, a1A, a1B
and a1D, have been con®rmed by cloning techniques.2±5
In order to design selective antagonists, a ligand-based
drug design (pharmacophore development) method may
be used by means of Catalyst software of Molecular
Simulations, Inc. (MSI). This method is applicable when
the three-dimensional structure of the receptor is
unknown and a series of compounds has been identi®ed
that show the activity of interest. The aim of this approach
is to identify a pharmacophore, which is a template
derived from the structures of these compounds, and
representing the geometry of the receptor site as a collec-
tion of functional groups in three-dimensional space.
The a1 ARs have divergent anities for many synthetic
drugs. Drugs interacting selectively as agonists, or as
antagonists, with the subtypes have been used in the
treatment of a variety of diseases including hypertension,
asthma and prostatic hypertrophy.6,7 Ligands showing
very high binding selectivity for the a1A subtype over
both a1B and a1D (>3000-fold) are available.8 However,
these ligands (SNAP compounds, derived from niguldi-
pine) usually do not distinguish very well between the
a1B and a1D subtypes (see Table 1 and Fig. 1 for exam-
ples, compounds 9±12). For the a1B subtype, the best
reported selectivity is achieved by (+)-cyclazosin (2)
A pharmacophore for antagonists at the a1 AR has
been reported.21 Considering a wide variety of chemical
structures, features linked with a1 AR antagonism were
noted. This study was based on a process which involved
choosing an active, conformationally constrained
molecule as a template to be compared with the con-
formationally ¯exible compounds. The resulting phar-
macophore had three features including an aromatic
region, a basic nitrogen and a semi-polar region or a
bulky lipophilic area.21 However, this early study did
not consider speci®c requirements for a1 AR subtypes.
Keywords: drug design; aporphine analogues; adrenergic receptors;
selectivity.
*Corresponding author. Tel.: +61-2-4221-4256; fax: +61-2-4221-
4287; e-mail: john_bremner@uow.edu.au
yCurrent address: Karaelmas Universitesi, Fen-edebiyat Fakultesi,
Kimya Bolumu, Zonguldak, Turkey.
0968-0896/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0968-0896(99)00263-1

202
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
Table 1. Anatagonists and their activities
No.
Compound name
Activity Ki(nM)
Selectivity
D/A
Ref
a1A
0.2
12
0.04
1.9
10
3.16
6.3
0.5
0.15
0.23
0.65
1.9
0.16
1.6
a1B
0.25
0.13
0.08
0.3
a1D
0.32
3.2
0.04
2.6
3.16
4.0
2.5
1.0
B/A
1.25
0.01
2
0.16
1
0.32
0.32
2
367
56.5
498
174
15.6
4.9
63.5
500
0.03
2.52
1.43
0.58
3.64
20
0.06
22.3
0.65
10
26.7
31.6
7.50
1.29
79
32
12.5
5.20
3
80
4.58
0.06
B/D
0.78
0.04
2
0.12
0.32
0.25
0.80
1
0.55
0.24
0.51
0.83
1
2
3
4
5
6
7
8
Prazosin
Cyclazosin
Abanoquil
REC-15/2615
Alfuzosin
Doxazosin
1.6
0.27
1
1.37
0.32
1.29
0.40
2
667
287
971
211
1.56
4.9
10
11
7
12
13
13
13
14
13
13
13
13
10
15
13
13
16
13
17
18
4
10
1.0
2.0
1.0
Terazosin
Bunazosin
9
(+)-Niguldipine
SNAP-5089
SNAP-5399
SNAP-5150
WB-4101
Phentolamine
5-Methyluropidil
KMD-3213
AH-11110A
BMY-7378
SKF-104856
Discretamine
Corynanthine
Benzoxathian
Spiperone
(+)-YM-617
SNAP-8719
Indoramin
RS-17053
A-131701
NAN-190
WAY-100635
RS-100,975
REC-15/2739
SNAP-1069
SL-89.0591
JHT-601
GG-818
Uropidil
Bromotopsentin
55
100
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
13
324
331
2.5
7.9
40
20
76
630
63
360
517
4.0
0.5
96
191
40
16
66
631
400
0.25
7.9
10
2.0
2750
6.3
5.2
25
253
0.4
13
22
1.6
160
16
0.97
0.8
63
100
2.5
790
2.5
1.2
25
10
1
4
0.63
0.04
15.9
50
1.1
10
2500
250
44
616
142
0.2
7.9
4.3
294
4.0
0.6
0.22
2.0
0.03
100
12.1
1.44
2.04
10
0.04
4.36
119
0.25
1
7.16
18.8
2.95
7.9
0.03
0.12
0.04
1.78
2
1.65
5.12
0.00
40
26.7
4.41
0.40
0.44
100
2.5
19
19
15
9
13
13
10
17
17
13
13
13
13
13
13
17
20
6.95
15
186
79
32
200
13
144
1.0
1.0
12.8
16
49.4
1
3
12.5
5.76
0.25
5.20
1
0.64
0.80
2.5
0.4
0.2
1.2
16
1320
740
288
12000
1660
Ð
A number of linear QSAR models have also been
developed from a1 AR antagonists.22±24 Complete geo-
metry optimisation was performed for the protonated
forms of the compounds considered, taking the exten-
ded conformations as starting geometries.22 This study
strongly suggested that the protonated amine function is
very important for binding. In addition, two aromatic
systems on either side of the nitrogen (at 4±7 A) are also
necessary for the activity, although no clear trend was
observed for the angle between these three features. In the
other studies a protonated amine function was also found
to be a crucial element of the a1 AR antagonists.23,24
designed with expected selectivity for the a1A AR over the
a1B subtype. The short distance between the basic nitro-
gen and the aromatic ring did not allow 39 to map e€ec-
tively on the a1B pharmacophore model. Synthesis and
binding studies of 39 on cloned and expressed human a1-
adrenoceptors are discussed in the current paper.
Compound 39 is an analogue of the aporphine alkaloid
system,26 in which ring d is replaced by a 1,2,3-thiadia-
zole ring. Aporphine alkaloids such as boldine and
glaucine have shown a1A selective binding,27 and an
aporphine ring d indole analogue, IQC (40),28 has
shown high anity for a1 AR sites.
In our earlier ligand-based pharmacophore study,25 a1A
and a1B AR pharmacophore models were developed by
using Apex-3D software of MSI. These models were the
®rst subtype speci®c a1 AR pharmacophores reported,
and they also indicated the importance of a protonated
amine. The additional features of the models were an
aromatic ring and a polar region for both subtypes. The
only di€erence between the pharmacophores for the two
subtypes was the distance between the aromatic ring
and the amine function. This distance was shorter for
the a1A subtype. Based on this data, a 1,2,3-thiadiazole
ring d-aporphine analogue (39) (see Scheme 1) was
Since the publication of our original study,25 several
important subtype selective a1 AR antagonists have
been developed and the selectivity of risperidone has
been queried. Even though risperidone was listed as an
a1B selective antagonist in an earlier paper,29 this
selectivity was questioned in another study,11 which
found it to be only slightly selective for the a1A subtype
over a1B. In addition, it has been found experimentally
that niguldipine analogues (which were in the training
set of the previous study25) bind at a di€erent binding
site of the receptor.29 This was supported by a later

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
203
study and it was suggested that they bind to the same
aspartate residue but the other interacting residues are
di€erent to those for the other antagonists.30 One of the
aims of the current study was to investigate the struc-
tural features of this important class of a1A selective
agents and compare them to the features encountered in
other a1A selective agents, such as KMD-3213 (16,
Table 1 and Fig. 1), with a view to contributing to the
understanding of the interaction of these compounds
with a1 ARs. In the case of the quinazoline derivatives
and close analogues 1±8, such as prazosin, we have
already stated our belief25 that these compounds also
may bind to a di€erent site on the a1 ARs. This class of
compounds, with the notable exception of (+)-cyclazo-
sin, is characterised by very high anity for all three a1
AR subtypes, with no selectivity between them. Fur-
thermore, the Apex-3D program has been superseded
by Catalyst. It was thus necessary to re-examine the
pharmacophores and this was done using the updated
software. Inclusion of further selective antagonists
described since the Apex-3D work was also undertaken,
together with an extension to a1D antagonists.
Results
Synthesis of the thiadiazole aporphine analogue 39 and
of IQC 40
The synthesis of racemic 39 was performed via the inter-
mediate tricyclic ketone 47 (Scheme 1). The carbonyl
Figure 1. Structures of a1-AR antagonists.

204
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
group in the 7-position was then used to build the d
ring. The multi-step sequence for the synthesis of 47
involved an isoquinoline precursor 41, which was pre-
pared by the method of Brossi et al.32 The N-propyl
derivative 42 was prepared in a sealed tube at 90ꢀC in
90% yield. The bromide salt ester derivative 44 could then
be made using the freshly prepared enamine derivative 43
after exposure of the iodide salt 42 to strong base. The
reduction of 44 and the preparation of the amino acid
hydrochloride salt derivative 46 was achieved in over 90%
Scheme 2. Structures of 39, 40, 49, 50 and 51 and the preparation of
racemic 40.
Scheme 1. Preparation of racemic 39.
Figure 2. The a1A pharmacophore based on Catalyst with KMD-3213 16 mapped onto it. The molecule is shown with the chemical bonds represented as
tubes and atoms colour coded according to the element type (black-C, white-H, red-O, blue-N, green-F). Mesh spheres represent the location constraints
for the feature on the ligand (or receptor) and are colour coded according to feature: red - positive ion, green±hydrogen bond acceptor (the second, larger
sphere relates to the position of the interacting group on the receptor), light brown - aromatic ring system (the second, larger sphere relates to the position
of the interacting aromatic system on the receptor; the light blue square denotes the position of the plane of the aromatic system).

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
205
Compound 39 had a negative potential section situated
on the nitrogen atom, whereas the other two aporphine
analogues, 50 and 40, had an area of positive potential
situated at the corresponding part of the lower aromatic
ring. This electron rich edge of 39 could cause an unfa-
vourable interaction with the receptor, and hence possibly
explain the poor anity of 39 for a1 ARs.
Table 2. Binding data for 39, 40 and the aporphines 50 and 51 at the
cloned and expressed human a1-ARs
No.
Compound name
Activity (nM)
a1A
a1B
a1D
39
40
50
51
Thiadiazole
IQC
Boldine
Glaucine
50 000
4
126 000
250
31 600
100
Ð
4.9a
75a
320a
1200a
Ð
Further pharmacophore development
aRef. 27.
In the light of these experimental results and the possibility
of distinct and di€erent binding sites for several important
classes of a1 AR antagonists, further pharmacophore
development was deemed necessary, before the design of
new, potentially selective ligands could be attempted.
yield. Cyclisation of 46 to 47 was completed using oleum
as the cyclising agent. Conversion of the amino ketone 47
to the semicarbazone 48 proceeded in good yield, how-
ever, reaction of 48 with thionyl chloride gave the required
39 in only 14% yield. The yield was very low in compar-
ison to that reported33 for the related derivative 49
(Scheme 2). The lower yield may possibly be due to side
reactions involving the methoxy groups.
Catalyst treats molecular structures as templates consist-
ing of strategically positioned chemical functions that will
bind to a receptor. Molecular ¯exibility is taken into
account by considering each compound as a collection of
conformers representing di€erent areas of conforma-
tional space available to the molecule within a given
energy range.35 The structures of known subtype selec-
tive AR antagonists used in this study are shown in
Figure 1 and their binding anities and selectivities
The amino ketone 47 could also be converted into the
racemic ring d indole analogue IQC 40 via the non-
isolated p-methoxyphenylhydrazone and subsequent
Fischer indolisation using glacial acetic acid and BF3-
etherate (Scheme 2). The heterocyclic skeleton had been
synthesised34 previously by this general method but with
dry hydrogen chloride as the acid catalyst. Structural
con®rmation of 39 and 40 was forthcoming from the
NMR and MS data.
Table 3. Experimental activity and estimated activities of the com-
pounds of Table 1 and Figure 1 for the a1A hypothesis
Compound name
Activity (nM)
No.
Measured
Estimated
Receptor binding results
1
2
3
4
5
6
7
8
Prazosin
Cyclazosin
Abanoquil
REC-15/2615
Alfuzosin
Doxazosin
0.2
12
0.04
1.9
10
3.16
6.3
2.8
2.8
2.8
2.8
2.9
2.8
2.8
2.8
0.16
2.8
0.49
0.086
0.3
The ring d thiadiazole analogue 39 of the aporphine
system, which was designed from the Apex-3D phar-
macophore models, was tested on the cloned and
expressed human a1A, a1B and a1D receptors (Table 2). As
can be seen, compound 39 had very low anity for the a1-
ARs. In contrast, the ring d indole analogue, 40, showed
high anity for these receptor subtypes, comparable to
those of the aporphines 50 and 51.
Terazosin
Bunazosin
0.5
9
(+)-Niguldipine
SNAP-5089
SNAP-5399
SNAP-5150
WB-4101
Phentolamine
5-Methyluropidil
KMD-3213
AH-11110A
BMY-7378
SKF-104856
Discretamine
Corynanthine
Benzoxathian
Spiperone
(+)-YM-617
SNAP-8719
Indoramin
RS-17053
A-131701
NAN-190
WAY-100635
RS-100,975
REC-15/2739
SNAP-1069
SL-89.0591
JHT-601
GG-818
Uropidil
Bromotopsentin
0.15
0.23
0.65
1.9
0.16
1.6
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
2.8
Molecular electrostatic potential (MEP) mapping
0.63
0.04
0.59
0.039
2.8
2.9
6.3
5.7
2.1
0.85
2.9
Electrostatic forces play an important role in determin-
ing the essential stereoelectronic features of biomole-
cules and drugs for QSAR studies. They also play an
important role in determining the orientations of the
molecules when approaching their receptors because of
their long-range action compared to other inter-
molecular interactions. Both these e€ects are repre-
sented in charge distributions calculated by ab initio
methods, which is the best source of information about
the electrostatic properties of a molecule.
2500
250
44
616
142
0.2
7.9
4.3
294
4.0
0.6
0.22
2.0
0.31
2.9
0.17
0.88
0.058
0.22
0.43
1.2
0.11
0.22
0.61
2.8
144
1.0
1.0
Models of the compounds 39, 40 and 50 were built, opti-
mised using ab initio calculations and the electrostatic
potentials were then mapped onto the total electron den-
sity surfaces. This showed clearly that the di€erences
between 39 and 40 or 50 lay in the electron density pat-
tern located on the lower ring system which is the thia-
diazole ring for 39, the dimethoxybenzene ring for 50
and the methoxy-indole ring for 40 (see Scheme 2).
16
2.5
0.4
0.2
0.18
1.1
0.58
288
12000

206
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
tabulated in Table 1. The compounds are grouped
together according to structural features, so that the
quinazolines 1±8 are listed ®rst, followed by the nigul-
dipine analogues 9±12 and other antagonists.
data and that the experimental activities are nor-
mally distributed about their mean.
According to the assessment criteria, a generated
hypothesis with a score that is substantially below the
cost of the null hypothesis, is likely to be statistically
signi®cant.
In an attempt to generate general a1 subtype pharmaco-
phores, applicable to all structural classes of ligands, the
complete training set of 38 antagonists was selected. Pro-
tonated forms of the antagonists were built to prede®ne at
least this one common feature between all molecules, as
this is postulated to be an essential feature for a1 AR
ligand binding (see Introduction). Conformational
searching was performed using a 20 kcal/mol energy
threshold with respect to the global energy minimum.
The Catalyst software uses molecular mechanics (force
®eld) calculations36 and a particular way to handle the
conformational space (poling method).37,38 The com-
pounds in the training set and their associated conforma-
tional models were submitted to generate hypotheses for
general a1 models. The chemical features used in this
generation step included positive ion, hydrogen bond
acceptor and donor, aromatic ring and hydrophobic
groups. The aromatic ring function also allows con-
sideration of the directionality of p±p interactions and
constrains the plane of the aromatic system. Catalyst
provides two numbers to assess the validity of a phar-
macophore (hypothesis):
The program gives 10 of these hypotheses as a result of
an experiment by default. Sometimes however, the ®rst
of these is empty so that only nine hypotheses are
obtained, which are numbered 2±10. Activity data for
each of the three subtypes was also included in the input
of these three experiments. However, the resulting
hypotheses, which included the positive ion feature, had
several problems. The correlations between the pre-
dicted and experimental anities were fairly low for the
compounds of the training sets (0.76 for a1A, 0.84 for
a
1B, 0.82 for a1D). Of special concern in this respect was
the fact that high errors (>5-fold) were observed for
the binding anity estimation of compounds selective
for the subtype under consideration (niguldipine (9),
SNAP 5089 (10), Me-uropidil (15), KMD-3213 (16),
RS-100,975 (31) and GG-818 (36) for the a1A subtype;
AH-11110A (17), spiperone (23), and bromotopsentin
(38) for the a1B subtype; SKF-104856 (19), SNAP-8719
(25), and NAN-190 (29) for the a1D subtype). Further-
more, the ring aromatic feature was not present in the
a1A top hypothesis, but only the less speci®c hydro-
phobic feature, which, with the exception of the quina-
zolines, is not usually mapped onto an aromatic ring
system. The same is true for the aromatic ring feature
which is present in the a1B pharmacophore hypothesis.
1. The cost of an ideal hypothesis which is a lower
bound on the cost of the simplest possible
hypothesis that still ®ts the data perfectly.
2. The cost of the null hypothesis which assumes that
there is no statistically signi®cant structure in the
Figure 3. The a1B pharmacophore based on Catalyst with spiperone 23. See Figure 2 for explanation of symbols. Additionally, this pharmacophore
contains a hydrogen bond donor feature (purple spheres) and a hydrophobic group (blue sphere).

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
207
For the a1D hypothesis, the situation was slightly dif-
ferent, in that the positive charge feature and the aro-
matic ring feature were mapped together in 14 out of the
37 training set compounds, however, in nine compounds
the positive charge feature was not mapped. These
®ndings con®rm previous observations,25,30,31 suggest-
ing that not all the compounds in Table 1 share the
same binding sites at the a1 ARs. We also found that
these `inclusive' pharmacophore models were not very
successful in predicting the anity of our lead com-
training set (1.0). The error in predicting IQC 40 is 1.4-
fold. As with our previous Apex-3D pharmacophores,25
the predictions for the anities of 39 and 40 are very
similar and indicate the fundamental failing of pharma-
cophore models in not taking detrimental electrostatic
e€ects into account as discussed earlier. The hypothesis
was also used to predict the anities for all compounds
in Table 1 and the results are shown in Table 3. Overall,
the correlation (0.35) with the experimental values was
still not good but, interestingly, most of the poor pre-
dictions were for compounds which exhibit weak a1A
anity (all eight compounds with Ki>100 nM are pre-
dicted very poorly). Prazosin and abanoquil are also
very poorly predicted, whereas the other quinazolines
are predicted reasonably well. The same is true for the
niguldipines, where SNAP-5150 and SNAP-5089 are
not predicted well, whereas the other two members of
this group are predicted very well.
pound, IQC 40 (prediction errors were nine fold for a1A
18-fold for a1B, and 67-fold for a1D).
,
We thus decided to pursue this further by constructing
new models with training sets excluding both the niguldi-
pine analogues and the quinazolines and also with train-
ing sets excluding only one of those two classes of
compounds. Upon taking the quinazolines (and close
analogues) 1±8 out of the training set, the correlation
between predicted and experimental anities for the
training set molecules did indeed improve, as did the
mapping of aromatic ring systems together with the posi-
tive charge feature and the accuracy of the predictions
for IQC 40. These pharmacophore models still did not
predict highly selective compounds in the training set
very well. On the other hand, removing the niguldipines
9±12 (the inclusive term niguldipines is used for simpli-
city) from the training sets did not have a large e€ect on
the pharmacophores. For example, for the a1A model,
the aromatic ring feature was still not included in this
case. Also, taking out both the niguldipine analogues and
the quinazolines from the training sets, did not improve
the pharmacophore models signi®cantly compared to the
ones when only the quinazolines were removed. Because
we were mainly interested in the features responsible for
selectivity between a1 AR subtypes, we decided to con-
struct pharmacophore models based on training sets
containing only a1 receptor subtype selective com-
pounds, but also excluding quinazoline derivatives and
close analogues (because our results above con®rmed
our previous suspicion25 that these interact with a1 ARs
in a di€erent way), and niguldipines. These pharmaco-
phore models are discussed below. Since the training
sets used for their derivation were now small, these
hypotheses serve only as possible qualitative models.
ꢀ1B Pharmacophore model
Only three antagonists (17, 23 and 38) were used in the
training set of this experiment. These compounds dis-
play at least 10-fold selectivity over the a1A subtype, and
Table 4. Experimental activity and estimated activites of the com-
pounds of Table 1 and Figure 1 for the a1B hypothesis
Compound name
Activity (nM)
No.
Measured
Estimated
1
2
3
4
5
6
7
8
Prazosin
Cyclazosin
Abanoquil
REC-15/2615
Alfuzosin
Doxazosin
0.25
0.13
0.08
0.3
160
280
250
190
220
350
190
190
290
690
610
270
360
870
600
150
79
1800
660
5500
550
270
0.52
110
2400
100
140
150
1700
930
140
160
160
88
10
1.0
2.0
1.0
Terazosin
Bunazosin
9
(+)-Niguldipine
SNAP-5089
SNAP-5399
SNAP-5150
WB-4101
Phentolamine
5-Methyluropidil
KMD-3213
AH-11110A
BMY-7378
SKF-104856
Discretamine
Corynanthine
Benzoxathian
Spiperone
(+)-YM-617
SNAP-8719
Indoramin
RS-17053
A-131701
NAN-190
WAY-100635
RS-100,975
REC-15/2739
SNAP-1069
SL-89.0591
JHT-601
GG-818
Uropidil
Bromotopsentin
55
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
13
324
331
2.5
7.9
40
20
76
630
63
360
517
4.0
0.5
96
191
40
16
ꢀ1A Pharmacophore model
Only antagonists 15, 16, 31 and 36 which had some
selectivity for the a1A subtype over the other subtypes,
were used in the training set (required selectivities were
50-fold over a1B and 10-fold over a1D). The chemical
functions used in the hypothesis generation step were
the same as above. Nine hypotheses were obtained by
default. Even though hypothesis 3 had a slightly higher
cost than hypothesis 2, this former hypothesis was cho-
sen as the best model due to the existence of a positive
ion feature. The best hypothesis (Fig. 2) consisted of
three features including a positive charge in the middle
of the system, a hydrogen bond acceptor group and an
aromatic ring system at opposite ends of the molecules.
This hypothesis had a very good correlation between the
experimental anities and the estimated values for the
6.95
15
186
479
32
200
13
1.2
650
150
380
750
16
1320
740

208
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
additionally quinazolines (cyclazosin) and niguldipines
were excluded. The same chemical features were used in
the generation step. Nine hypotheses were obtained and
hypothesis 2 (lowest cost hypothesis) was chosen as the
best hypothesis because it also contained a positive
charge and an aromatic ring feature. This hypothesis
consisted of four features including a hydrogen bond
donor group and a hydrophobic group as well as the
aromatic ring system and the positive charge (see Fig.
3). This hypothesis showed a very good correlation
(1.00) between the experimental activity and the esti-
mated values for the training set. All compounds in
Table 1 were estimated with this hypothesis and the
results are shown in Table 4. The overall correlation is
0.37. In this case, all quinazolines are predicted very
poorly with errors of 20-fold and greater. Two of the
niguldipines are predicted quite well, but they possess
relatively poor anity for the a1B site. This is a general
problem with this hypothesis. It underestimates (error
>5) the anity of 23 out of the 28 compounds with high
anity (Ki <100nM). The aporphine derived com-
pounds 39 and 40 are correctly predicted to have much
lower anity at a1B than for the a1A subsite (Ki=780nM
for 40, error is 3-fold, Ki=5300 nM for 39).
them contained an aromatic ring feature as well as a
positive ion feature. Hypothesis 5 was selected, as it was
the lowest cost hypothesis to contain a positive ion. It
also includes a hydrophobic group and a hydrogen
bond acceptor and is shown in Figure 4 with SNAP-
8719 mapped onto it. For both discretamine and SNAP-
8719 the hydrophobic feature is in fact mapped onto an
aromatic ring, but SKF-104856 has only one phenyl
ring which is much closer to the positively charged
nitrogen. The correlation between estimated and
experimental anities for the training set molecules is
excellent (1.00). Again, we have compared all molecules
in Table 1 with this hypothesis and the results are tabu-
lated in Table 5. The correlation is 0.14 (no correla-
tion). This hypothesis underestimates the anity of all
quinazolines, whereas all niguldipines are over-
estimated. The remainder of the compounds are under-
or overestimated with equal frequency. Our lead com-
pounds (39 and 40) are again estimated to have very
similar anities (58 versus 77 nM for the Ki values of
(39 and 40); error for 40 is 1.3-fold).
Discussion
We have shown that it is not possible to construct
pharmacophores for the binding of antagonists to a1
receptor subtypes that are valid for all structural classes
of antagonists in Table 1 and Figure 1. The relative
positions of the protonated nitrogen atom and the aro-
matic systems are di€erent, especially for the quinazo-
lines, and the quality of the pharmacophore improves
when these compounds are excluded from the training
sets. From these results it is quite likely that these
ligands, which generally exhibit very high anity for all
ꢀ1D Pharmacophore model
Only four compounds in Table 1 show at least 10-fold
selectivity over the a1A receptor subtype. Of these,
BMY-7378 and SNAP-8719 are structurally very similar
(see Fig. 1). Including both would have distorted the
training set, so only three compounds (19, 20 and 25)
were used. Quinazolines and niguldipines were again
excluded. Nine hypotheses were generated, but the ®rst
three did not contain a positive ion feature and none of
Figure 4. The a1D-pharmacophore with SNAP-8719 25. See Figures 2 and 3 for explanation of symbols.

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
209
Table 5. Experimental activity and estimated activites of the com-
pounds of Table 1 and Figure 1 for the a1D hypothesis
This consisted of restricting the training sets for the
pharmacophores severely by only including selective
ligands which were neither quinazolines, nor niguldi-
pines, and these are the pharmacophores presented in
Figures 2±4.
Compound name
Activity (nM)
No.
Measured
Estimated
1
2
3
4
5
6
7
8
Prazosin
Cyclazosin
Abanoquil
REC-15/2615
Alfuzosin
Doxazosin
Terazosin
Bunazosin
(+)-Niguldipine
SNAP-5089
SNAP-5399
SNAP-5150
WB-4101
Phentolamine
5-Methyluropidil
KMD-3213
AH-11110A
BMY-7378
SKF-104856
Discretamine
Corynanthine
Benzoxathian
Spiperone
(+)-YM-617
SNAP-8719
Indoramin
RS-17053
A-131701
NAN-190
WAY-100635
RS-100,975
REC-15/2739
SNAP-1069
SL-89.0591
JHT-601
0.32
3.2
0.04
2.6
3.16
4
2.5
1
100
66
631
400
0.25
7.9
10
6.7
9.1
8.1
8.6
5.8
7.5
8.5
7.3
2.6
3.2
2.7
3.9
4.8
20
All three a1 AR subtype pharmacophores so con-
structed predict IQC 40 very well and can therefore be
considered useful for the further design of more selec-
tive ligands based on that compound (see below). They
all, however, also predict the 1,2,3-thiadiazole derivative
39 to exhibit very similar anities to 40, contrary to the
experimental ®ndings (see Table 2). This can be
explained by the fact that the software is not able to
take detrimental electrostatic e€ects into account.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
These pharmacophore models are only to be considered
as qualitative models, however, as can be seen from
their generally poor predictive abilities for the com-
pounds of Table 1.
18
2
2750
2.3
5.4
2.2
4.8
24
6.3
5.2
25
253
0.4
13
22
1.6
Comparison of the pharmacophore models and
implications for ligand design
5.8
5
4.5
4.1
1.8
4.7
3.9
4.9
7.5
3.2
3.5
2.5
3.4
12
In Figure 5 the three subtype speci®c a1 pharmacophore
models derived solely from selective ligands are com-
pared. All three contain a protonated nitrogen atom,
and while this contributes to the anity of each com-
pound, in terms of selectivity it would be desirable to
design a ligand not containing such a feature. For this
ligand to exhibit enough anity it would however be
necessary to start with a pharmacophore of at least
three other features (which in our case is only possible
for the a1B subtype) and a lead compound with high
anity for this subtype. A feature unique to the phar-
macophore for ligands at the a1B subtype is a hydrogen
bond donor (HBD) group. For the a1A pharmacophore,
there is a hydrogen bond acceptor group (HBA), loca-
ted much further away from the positive ion (see Table
6 for the distances and angles between the features for
the three pharmacophores), and for the a1D pharmaco-
phore, the equivalent distance is smaller. When map-
ping IQC 39 onto the a1B pharmacophore, the HBD
feature is not mapped, but it is close to the indole
nitrogen. We have, furthermore, observed a hydrogen
bonding interaction between this indole NH of IQC and
the a1B receptor, when docking the ligand into the
computer model of the transmembrane helices of the
receptor.39 Additionally, this NH group, incorporated
into the indole ring, is not considered to be an HBA
within the Catalyst software and thus cannot be map-
ped onto the HBA feature of the a1A and a1D pharma-
cophores. A ligand selective for the a1B over the other
two a1 AR subtypes might thus be achieved by replacing
the O-methyl groups in 2 with alkyl groups, to provide
hydrophobic interactions and avoid any HBA cap-
ability. This compound would be able to interact with
all four features of the a1B pharmacophore, leading to a
predicted anity in the same order of magnitude as that
for IQC 2 (250 nM), whereas the anity for the a1A and
160
16
0.97
0.8
63
100
2.5
790
2.5
1.2
25
1660
7.5
10
11
GG-818
Uropidil
three a1 AR subtypes, but little subtype selectivity, bind
to a di€erent site, common to all a1 ARs.
Our results are, however, not as de®nitive in this respect
for the niguldipines, a class of compounds exhibiting
very high a1A anities and selectivities.
For all these `inclusive' pharmacophores, the additional
variable of the selectivity between the a1 AR subtypes
was not considered. Table 1 contains compounds with
high anity, but low selectivity (e.g. prazosin (1)) and
compounds with low anity, but high selectivity (e.g.
bromotopsentin (38)). The Catalyst program, having
only received information about the anity, dis-
criminates against compounds such as bromotopsentin
in favour of compounds such as prazosin. This was
re¯ected in the fact that all these ``inclusive'' pharma-
cophores predict the anity of selective antagonists
very poorly, and this was also true for the predictions of
our lead compounds 39 and 40. Since we were primarily
interested in the design of new subtype selective ligands
based upon our aporphine type lead compounds, we
decided to employ a di€erent approach towards phar-
macophore development.
a
1D subtypes should be reduced by two orders of magni-
tude to around 400 and 10,000 nM, respectively. As was
predicted by the original Apex 3D pharmacophores,

210
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
Figure 5. Overlap of the three pharmacophores. Letters indicate the subtype to which the feature belongs.
Table 6. Distances and angles between pharmacophore features
importance of this was shown with 39, which, despite
predictions, showed very low a1 anities. The fact that
molecular ®elds in general, and electrostatic potentials
in particular, are not considered in the pharmacophore
development is an important limitation of this approach
and has been highlighted again here. The Catalyst soft-
ware does allow the placement of excluded volume fea-
tures, the location of which can be determined using
receptor models (in the absence of any experimental 3D
structure of any adrenoceptor) and this is currently
under investigation.
Pharmacophore
model
Distances (A) and angles (ꢀ)
between featuresa
a1A
a1B
a1D
P - A 5.5; P - HBA 7.1;
A - P - HBA 100
P - A 6.2; P - H 7.8; P - HBD 4.9;
H - P - HBD 57; A - P - HBD 52
P - H 5.4; P - HBA 4.5; H - P - HBA 47
aAbbreviations used for features: P, positive charge; HBA, hydrogen
bond acceptor; A, aromatic ring centre; H, hydrophobic group; HBD,
hydrogen bond donor.
although true to a lesser extent, the distance between the
aromatic feature and the positive charge is smaller for
the a1A subtype and selectivity for a1B may be further
improved by separating the indole and the tetra-
hydroisoquinoline aromatic systems by a larger ring
and/or removing the aromatic portion of the tetra-
hydroisoquinoline system altogether. The former mod-
i®cation would also potentially strengthen a weak, long
aromatic interaction observed for the indole phenyl ring
in our docking experiments.39
Experimental
Synthesis
All melting points were determined using a Reichert hot
stage melting point apparatus and are uncorrected. The
1H nuclear magnetic resonance spectra (NMR) were
determined at 300 or 400 MHz with a Varian Unity-300
or -400 spectrometer. The 13C NMR spectra were
recorded using the same instruments at 75 or 100 MHz.
Unless otherwise stated, the spectra were obtained on
solutions in CDCl3 and referenced to TMS. The elec-
tron impact (EI) mass spectra were obtained on a Shi-
madzu QP-5000 mass spectrometer using the direct
insertion technique, with an electron beam energy of
70 eV and a source temperature of 260 ꢀC. The peak
intensities, in parentheses, are expressed as a percentage
abundance. For the chemical ionisation (CI) mass spec-
tra, isobutane was used as the ionising gas in a Shimadzu
QP-5000 mass spectrometer. The low resolution elec-
trospray mass spectra were obtained on a Vacuum
General Quattro mass spectrometer using 50% aqueous
Conclusion
Pharmacophores for selective antagonists at the a1A and
a1B adrenoceptor sites were developed together with a
preliminary antagonist pharmacophore for the a1D
adrenoceptor, which is the ®rst pharmacophore repor-
ted for this particular subtype. These pharmacophores
should be useful for designing ligands based on the
aporphine skeleton, as long as suitable care is taken not
to introduce unfavourable electrostatic e€ects. The

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
211
acetonitrile as solvent and a skimmer cone voltage of
25 V. Where electrospray MS data is reported, it refers
to the positive ion mode. The high resolution mass
spectra (HRMS) were obtained using either a Vacuum
General Autospec-OA-TOF mass spectrometer using
the same conditions as for the Quattro mass spectro-
meter and a resolution of 5000 or a Vacuum General
Micromass 7070F spectrometer employing the direct
insertion technique, with an electron beam energy of
70 eV and a source temperature of 200 ꢀC. Analytical
thin layer chromatography (TLC) was performed on
Merck Kieselgel 60PF254 silica on plastic backed sheets.
Rf values were recorded from the centre of the spots. All
chromatographic solvent proportions are volume by
volume. Column chromatography was performed using
Merck silica gel (mesh size=0.032±0.064 mm) under
medium pressure. All reagents and solvents were pur-
i®ed and dried by standard techniques.40 The drying of
chloroform, diethyl ether or dichloromethane extracts
was done with anhydrous magnesium sulfate or sodium
sulfate. Solvents were removed under reduced pressure
in a rotary evaporator. Light petroleum used had a
boiling point range of 40±60 ꢀC.
which were combined, washed with water (5 mL) and
dried. The product was collected after evaporation as a
yellow powder (10.86g, 53mmol, 76%), which yielded 41
as colourless crystals after recrystallisation from diethyl
ether; m.p. 98±100ꢀC (Found: C, 70.0; H, 7.5; N, 7.0%.
C12H15NO2 requires C, 70.2; H, 7.4; N, 6.8%. MS (EI):
m/z 205 (M+, 100%), 190 (55), 174 (15), 160 (12), 147 (10),
131 (8). HRMS: m/z 205 (M+, 100%, accurate mass
205.110. C12H15NO2 requires 205.110). 1H NMR d:
6.991 (s, 1H, ArH); 6.694 (s, 1H, ArH); 3.923 (s, 6H,
2ÂAr-OCH3); 3.662 (t, J=7.6 Hz, 2H, H3); 2.636 (t,
J=7.6 Hz, 2H, H4); 2.372 (s, 3H, C-10). 13C NMR d:
163.4 (C-1), 150.6, 147.2 (C-6, C-7), 130.9 (C-8a), 122.3
(C-4a), 110.0, 108.8 (C5, C8), 56.0, 55.8 (2ÂAr-OCH3),
46.8 (C-3), 25.6 (C-4), 23.2 (C-10).
Preparation of 6,7-dimethoxy-1-methyl-2-propyl-3,4-di-
hydroisoquinolinium iodide (42). Excess 1-iodopropane
(1.6 mL; freshly distilled over anhydrous potassium car-
bonate) and the ®nely powdered 41 (1.86 g, 9 mmol)
were placed in a sealed tube. This mixture was heated at
90 ꢀC for 24 h. The solid was removed from the tube
with a small amount of dichloromethane and diethyl
ether was added to recrystallise the crude product. This
was ®ltered and washed with diethyl ether to give 42
(2.85 g, 7.6 mmol, 84%) as a pale yellow powder; m.p.
179±180 ꢀC. Found: C, 47.7; H, 5.9; N, 3.8. C15H21NO2
requires C, 48.0; H, 5.9; N, 3.7%. MS (ES): m/z 247 (375-
HI, 45%), 232 (30), 216 (9), 205 (97), 190 (49), 174 (18),
160 (12), 142 (100). HRMS: m/z 247 (M-HI+, 79%,
accurate mass 247.159. C15H21NO2 requires 247.157). 1H
NMR d: 7.282 (s, 1H, ArH); 6.917 (s, 1H, ArH); 4.112 (t,
J=7.6Hz, 4H, 2CH2); 4.013, 3.916 (s, 6H, 2ÂAr-OCH3);
3.253 (t, J=7.7 Hz, 2H, H4); 3.032 (s, 3H, CH3); 1.987±
1.903 (m, 2H, H-20); 1.092 (t, J=7.4 Hz, 3H, CH3). 13C
NMR d: 173.7 (C-1); 156.3, 148.6 (C-6, C-7); 137.8 (C-8a);
119.7 (C-4a); 112.4, 110.6 (C-5, C-8); 59.2 (C-3); 56.9, 56.8
(2ÂAr-OCH3); 51.1 (C-10); 26.1 (C-4), 21.2 (C-20); 20.3
(CH3); 11.1 (C-30).
Micro analyses were carried out on a Carlo Erba,
CHNS-O EA1108 Elemental Analyser in the Central
Science Laboratory, University of Tasmania.
Preparation of 6,7-dimethoxy-1-methyl-3,4-dihydroiso-
quinoline (41). To a stirred mixture of 2-(3,4-dimethox-
yphenyl)ethylamine (30.09g, 145 mmol) and dry pyridine
(15 mL; distilled over potassium hydroxide) under nitro-
gen was added acetic anhydride (19 mL) dropwise to
maintain a reaction temperature of 60±70 ꢀC. The tem-
perature was then increased to 90ꢀC for 1 h. The solution
was left at room temperature overnight. Pyridine and
acetic acid were distilled and the crude 2-[(3,4-dimethoxy-
phenyl)ethyl]ethanamide recrystallised from ethyl ace-
tate to a€ord colourless crystals (23.81g, 107 mmol,
79%); m.p. 98±100 ꢀC. MS (EI): m/z 223 (M+Á, 100%).
HRMS: m/z 223 (M+, 28%, accurate mass 223.119.
C12H17NO3 requires 223.121). 1H NMR d: 6.822 (d,
J=8.3Hz, 1H, ArH); 6.753 (s, 1H, ArH); 6.732 (d,
J=7.5Hz, 1H, ArH); 3.876 (s, 6H, 2ÂAr-OCH3); 3.503
(q, J=6.6Hz, 2H, H-20); 2.773 (d, J=6.9Hz, 2H, H-10);
1.952 (s, 3H, CH3); 5.560 (s, NH). 13C NMR d: 169.9
(CˆO), 148.8 (C-3), 147.5 (C-4), 131.2 (C-1), 120.4, 111.7,
111.2 (C-2, C-5, C-6), 55.7 (2ÂAr-OCH3), 40.6 (C-20), 35.0
(C-10), 23.1 (CH3).
Preparation of 6,7-dimethoxy-1-methylidene-2-propyl-
1,2,3,4 - tetrahydroisoquinoline (43). A solution of 42
(16.75 g, 45 mmol) in water (50 mL) was basi®ed with
32% sodium hydroxide. This mixture was then extrac-
ted rapidly with portions of diethyl ether (3Â20 mL), the
extracts combined, dried and the solvent evaporated to
give crude 43 (9.79 g, 40mmol, 89%) as a yellow±orange
syrup. MS (EI): m/z 247 (M+Á, 45%), 232 (30), 219 (8),
205 (100), 190 (45), 174 (15), 160 (10), 142 (8). HRMS: m/z
247 (M+, 69%, accurate mass 247.257. C15H21NO2
1
To a mixture of dry toluene (90 mL) and 2-[(3,4-dime-
thoxyphenyl)ethyl]ethanamide (15.4 g, 69 mmol), at
60 ꢀC, freshly distilled phosphoryl chloride (18 mL) was
added dropwise over 30 min. The mixture was re¯uxed
until hydrogen chloride evolution ceased (2.5 h), and
then was cooled on ice. The toluene was decanted and
the crystals washed with light petroleum containing a
few drops of toluene. The crystals were dissolved in
water (50 mL) and the solution cooled on ice. Basi®ca-
tion to pH 11 of the mixture with 32% sodium hydrox-
ide (50 mL) was undertaken with ice added to maintain
a reaction temperature below 30 ꢀC. The product was
extracted with portions of dichloromethane (3Â20 mL)
requires 247.257). H NMR d: 7.203 (s, 1H, ArH); 6.573
(s, 1H, ArH); 4.30 (bs, integrated less than 2H, C-10); 3.90
(s, 3H, Ar-OCH3); 3.84 (s, 3H, Ar-OCH3); 3.23±3.18 (m,
4H, 2CH2); 2.80 (t, J=5.9Hz, 2H, H4); 1.73±1.60 (m, 2H,
C-20); 0.94 (t, J=7.4 Hz, 3H, C-30). 13C NMR d: 148.5 (C-
1), 147.2, 146.2 (C-6, C-7), 127.5 (C-8a), 124.8 (C-4a),
110.2, 107.9 (C-5, C-8), 75.3 (C-10), 55.7, 55.5 (2ÂAr-
OCH3), 54.0 (C-3), 47.7 (C10), 29.8 (C-4), 18.6 (C-20), 11.5
(C-30).
Preparation of 1-[2-(ethoxycarbonyl)ethyl]-6,7-dimethoxy-
2-propyl-3,4-dihydroisoquinolinium bromide (44). To the
crude enamine (43) (9.79 g, 40 mmol), at 0 ꢀC and under

212
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
nitrogen gas, ethyl bromoacetate (40 mL) was added
dropwise with stirring. The solution was then heated to
60 ꢀC for 30 min. Dry toluene was added and the pro-
duct was collected as cream crystals (13.64 g, 33 mmol,
83%). Recrystallisation from dry ethanol/diethyl ether
a€orded pure 44 as a yellow powder; m.p. 176±178 ꢀC.
Found C, 54.8; H, 7.2; N, 3.63. C19H28BrNO4 requires C,
55.1; H, 6.8; N, 3.4%. MS (ES): m/z 334 (M-Br, 100%).
HRMS: m/z 333 (M-HBr+, 27%, accurate mass 333.195.
C19H27NO4 requires 333.194). 1H NMR d: 7.502 (s, ArH);
6.963 (s, ArH); 4.262±4.059 (m, 3CH2); 4.011 (s, OCH3);
4.098 (s, OCH3); 3.752 (t, J=7.7 Hz, H4); 3.235 (t,
J=7.5Hz, H-3); 2.942 (t, J=7.1 Hz, CH2); 2.097±1.924
(m, 2H, CH2); 1.223 (t, J=7.1 Hz, CH3); 1.097 (t,
J=7.4Hz, CH3). 13C NMR d: 176.0 (CˆO), 171.2 (C-1),
156.3, 148.4 (C-6, C-7), 133.7 (C-8a), 118.1 (C-4a), 112.6,
110.8 (C-5, C-8), 61.4 (C-3), 59.2 (C-400), 57.1, 56.7 (2ÂAr-
OCH3), 51.5 (C-10), 32.7 (C-200), 26.4 (C-4), 26.0 (C-100),
21.6 (C20), 14.0 (C-500), 11.1 (C-3').
J=7.4Hz, CH3). 13C NMR d: 174.4 (CˆO), 149.3, 149.1
(C-6, C-7), 122.0, 121.0 (C-8a, C-4a), 110.1, 110.8 (C-5, C-
8), 67.6 (C-3), 64.3 (C-1), 61.7 (C-10), 56.2, 55.9 (2ÂAr-
OCH3), 42.2 (C-4), 30.8 (C-200); 21.4 (C-100) 17.9 (C-20),
11.1 (C-30).
Preparation of 5,6-dimethoxy-1-propyl-2,3,7,8,9,9a-hex-
ahydro - 1H - benzo[de]quinolin - 7 - one (47). Dry, ®nely
powdered 46 (0.96 g, 2.79mmol) was sprinkled over the
surface of stirred oleum (6mL; 20% free sulfur trioxide) at
room temperature. The reaction mixture was stirred for a
further 5 min. It was then quenched with ice and the solu-
tion basi®ed to pH 11 with 32% sodium hydroxide solu-
tion. The basic solution was extracted with dichloro-
methane (3Â50mL), the extracts combined, dried and the
solvent evaporated to leave 5,6-dimethoxy-1-propyl-
2,3,7,8,9,9a-hexahydro-1H-benzo[de]-quinolin-7-one (47)
as a dark green gum (0.61 g, 2.12mmol, 74%). MS (CI):
m/z 290 (MH+, 100). HRMS (EI): m/z 288 (M-H+, 16%,
accurate mass 288.1599. C17H22NO3 requires 288.160). 1H
NMR d: 6.842 (s, 1H, ArH); 3.864, 3.852 (s, 6H, 2ÂAr-
OCH3); 3.33 (`apparent doublet', J=11.6 Hz, 1H, H-9a);
3.203±3.157 (m, 1H); 3.111±2.948 (m, 1H); 2.832±2.379
(m, 7H); 1.712±1.562 (m, 3H); 0.936 (t, J=6.8Hz, CH3).
13C NMR d: 196.9 (C-7); 151.8 (C-6); 147.5 (C-5); 133.4
(C-6a); 129.4 (C-9b); 125.6 (C-3b); 116.7 (C-4); 61.3 (C-
9a); 59.7, 55.9 (2ÂAr-OCH3); 55.1 (C-2); 49.1 (C-10); 36.4
(C-8); 28.7, 27.3 (C-3, C-9); 19.4 (C-20); 11.8 (C-30).
Preparation of ethyl 3-[6,7-dimethoxy-2-propyl-1,2,3,4-
tetrahydro-isoquinolin-1-yl]-propanoate (45). To a stir-
red solution of 44 (2.00 g, 4.8 mmol) in ethanol (15 mL)
at 0 ꢀC, was added sodium borohydride (0.42g) over 1 h.
The reaction mixture was left overnight at room tem-
perature. The ethanol was then evaporated and the
residue dissolved in water (4 mL). The solution was
extracted with portions of diethyl ether (3Â10 mL)
which were combined, dried and the solvent evaporated
leaving a clear, pale green syrup (1.54 g, 4.6 mmol,
95%). This was puri®ed by medium pressure column
chromatography to yield pure 45 as a colourless syrup.
MS (ES): m/z 336 (MH+, 100%). HRMS: m/z 334 (M-
H+, 3%, accurate mass 334.200. C19H28NO4 requires
Synthesis of 1,2-dimethoxy-6-propyl-5,6,6a,7-tetrahydro-
4H-benzo[de][1,2,3]thiadiazolo[4,5-g]quinoline
(39).
Anhydrous sodium ethanoate (180 mg, 3.0 mmol) was
added to a hot solution of semicarbazide hydrochloride
(180 mg, 1.6 mmol) in ethanol (20 mL). The sodium
chloride precipitate which formed was removed by ®l-
tration. To the ®ltrate was added the amino ketone 47
(360 mg, 1.2 mmol) and this mixture was heated under
re¯ux for 1 h. After cooling, the solid was ®ltered,
washed with cold ethanol and dried to a€ord, after
recrystallisation from methanol, the semicarbazone 48
(180 mg, 0.5 mmol, 42%) m.p. 188.7±189 ꢀC. MS (CI):
m/z 347 (MH+, 100%). HRMS (ES): m/z 347 (MH+,
100%, accurate mass 347.2081. C18H27N4O3 requires
1
334.202). H NMR d: 6.583 (s, 1H, ArH); 6.552 (s, 1H,
ArH); 4.127 (q, J=7.1 Hz, 2H, CH2); 3.856 (s, 6H,
2OCH3); 3.544 (t, J=6.5 Hz, 1H, H-1); 3.156±3.108 (m,
1H); 2.801±2.744 (m, 2H); 2.543±2.402 (m, 5H); 2.033±
1.949 (m, 2H); 1.535±1.462 (m, 2H, CH2); 1.264 (t,
J=7.1Hz, 3H, CH3); 0.891 (t, J=7.4 Hz, 3H, CH3). 13
C
NMR d: 174.3 (CˆO), 147.2 (C-6, C-7), 129.9 (C-8a),
126.7 (C-4a), 111.2, 110.5 (C-5, C-8), 63.1 (C-3), 62.5 (C-
1), 60.0 (C-400), 59.8 (C-10), 55.9, 55.8 (2ÂAr-OCH3), 55.4
(C-11), 43.7 (C-4); 31.0 (C-200), 23.7 (C-100), 21.1 (C-20),
14.2 (C-500), 11.8 (C-30).
1
347.2083). H NMR d: 9.392 (s, 1H, NH), 6.624 (s, 1H,
H-4), 3.849, 3.773 (s, 6H, 2ÂArO-CH3), 3.237±2.985 (m,
3H), 2.910±2.335 (m, 7H), 2.041 (s, 2H, NH2), 1.750±
1.280 (m, 3H), 0.995±0.890 (t, 3H, CH3). 13C NMR d:
159.0 (CˆO), 151.9 (C-5), 145.4 (C-6), 143.9 (C-7),
130.3 (C-3a), 128.46 (C-6b), 124.9 (C-6a), 112.5 (C-4),
60.7, 59,6 (2ÂAr-OCH3), 55.8 (C-9a), 55.5 (C-8), 48.7
(C-2), 28.5 (C-3), 26.0 (C-10), 25.0 (C-9), 19.2 (C-20),
11.9 (C-30).
Preparation of 1-[2-(carboxyethyl)]-6,7-dimethoxy-2-pro-
pyl- 1,2,3,4 -tetrahydroisoquinolinium chloride (46). The
compound 45 (1.04 g, 3.10 mmol) was dissolved in 2 M
hydrochloric acid (3 mL). This solution was stirred at
room temperature under nitrogen gas and shielded from
light for 4 days. The solvents were removed in vacuo
over 24 h, the crude product collected as a white foam
(0.96 g, 2.79 mmol, 90%). This was recrystallised from
dry methanol/diethyl ether to yield pure 46 as a colourless
powder; m.p. 168±170 ꢀC. Found: C, 59.3; H, 7.7; N, 4.2.
C17H26ClNO4 requires C, 59.4; H, 7.6; N, 4.1%. MS (ES):
m/z 308 (M-35Cl, 1%), 292 (2), 249 (8). HRMS: m/z 306
(M-H325Cl+, 1%, accurate mass 306.171. C17H24NO4
To thionyl chloride (2 mL) was added the semi-
carbazone 48 (40 mg, 0.12 mmol) portionwise at 0 ꢀC
with stirring. The mixture was stirred for a further 3 h at
0 ꢀC and then kept overnight at room temperature.
Dichloromethane (25 mL) was added and the mixture
was poured with stirring into a cooled solution of
sodium carbonate (6 g in 25 mL of water). The organic
phase was separated, dried and the dichloromethane
evaporated to give crude product. The crude product
was puri®ed via ¯ash column chromatography (silica
1
requires 306.170). H NMR d: 6.804 (s, 1H, ArH); 6.644
(s, 1H, ArH); 4.502 (t, 1H, H1); 3.891, 3.871 (s, 6H,
OCH3); 3.752 (m, 1H); 3.51 (m, 1H); 3.031 (m, 4H); 2.854
(m, 1H); 2.697 (m, 2H); 2.123±2.052 (m, 4H); 0.943 (t,

J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
213
gel, methanol:dichloromethane 5:95). Recrystallisation
from methanol gave 1,2-dimethoxy-6-propyl-5,6,6a,7-
tetrahydro-4H-benzo[de][1,2,3]thiadiazolo[4,5-g]quino-
line (39) (8 mg, 0.02 mmol, 16%); m.p. 185±186 ꢀC. MS
(CI): m/z 332 (MH+, 100%). HRMS (ES): m/z 332
(MH+, 79%, accurate mass 332.1432. C17H22N3O2S
requires 332.1433). 1H NMR d: 6.711 (s, 1H, H-4);
3.987, 3.900 (s, 6H, Ar-OCH3); 3.727 (dd, J1=6.0, J2=
15.0 Hz, 1H, H-6a); 3.56±3.50 (dd, J1=3.0, J2=15.0 Hz,
1H, one of C-7); 3.24±3.16 (m, 1H, one of C-5); 3.122±
3.014 (m, 1H, one of C-5); 2.90±2.647 (m, 2H, CH2);
2.587±2.430 (m, 2H, CH2); 1.639±1.543 (m, 2H, C-20);
0.955 (t, J1=6.0 Hz, 3H, C-30). 13C NMR d: 152.6 (C-
10a), 147.7 (C-7a, Ar-OCH3), 144.2 (Ar-OCH3), 130.8
(C-10b), 125.2 (C-10c), 121.6 (C-3a), 112.8 (C-3), 61.5
(C-6a), 59.2 (C-7), 56.3, 56.0 (2ÂAr-OCH3), 49.4, 29.1
(C-10, C-3), 26.3 (C-4), 19.4 (C-20), 12.0 (C-30).
method.42 The ligand binding characteristics of
the expressed receptors were determined in a series of
radioligand binding studies using the a1-antagonist
radioligand
[
125I]HEAT as described previously.41
Nonspeci®c binding was determined in the presence of
10 4 M phentolamine. Binding data were analysed by
the iterative curve-®tting program LIGAND.
Molecular electrostatic potential mapping. This parti-
cular molecular modeling work was performed on a
Silicon Graphics Indy workstation using Spartan SGI
version 4.0.3. GL software, under the IRIX 5.2 system.
The ligands were built and optimised by using an ab
initio calculation at the Hartree±Fock level with the
minimal STO-3G basis set. The electron density of the
molecules was also calculated in three dimensions gen-
erating the isosurface (0.002 electron/au3) and the elec-
trostatic potential at each point was then determined
and visualised colour coded onto this surface from the
most negative (red) to the most positive (blue) values.
Preparation of 1,2,9-trimethoxy-6-propyl-4,5,6,6a,7,12-
hexahydroisoquinolino[8,1-ab]carbazole (40). To a solu-
tion of the ketone 47 (1.2 g, 4.2 mmol) in dry ethanol
(2.5 mL) and glacial acetic acid (three drops) was added
fresh p-methoxyphenylhydrazine (0.36 g, 7.5 mmol) in
dry methanol (2 mL). The reaction mixture was re¯uxed
for 2 h under nitrogen and then the solvents evaporated.
To the residue was added glacial acetic acid (4 mL) and
redistilled boron tri¯uoride etherate (0.5 mL). The mix-
ture under nitrogen was brought to re¯ux as rapidly as
possible and heated for 10 min. Ice was added (50 g), the
solution was basi®ed to pH 11 with 32% sodium hy-
droxide solution, and the mixture was then extracted
with chloroform (3Â20 mL). The extracts were com-
bined, washed with water and dried. Evaporation of the
solvent a€orded the crude product as a cream foam
(1.05 g, 2.6 mmol, 64%). Recrystallisation from dichlor-
omethane/light petroleum a€orded 1,2,9-trimethoxy-6-
propyl-4,5,6,6a,7,12-hexahydroisoquinolino[8,1-ab]car-
bazole 40 as tan crystals; m.p. 120±122 ꢀC (Found; C,
73.2; H, 7.2; N, 7.2. C24H28N2O3 requires C, 73.4; H7.2;
N, 7.1%). MS (CI): m/z 393 (MH+, 100%). HRMS
(EI): m/z 392 (M+, 100%, accurate mass 392.212.
C24H28N2O3 requires 392.210) 391 (71), 390 (52), 377
Catalyst experiments
All molecular modelling work was performed on a
Silicon Graphics O2 workstation using the Catalyst
package of Molecular Simulations, Inc. (version 4.0),
under an IRIX 6.3 operating system. A generalised
CHARMM-like36 force ®eld was used throughout.
Training sets were selected as described in the Results
section. Molecules were built within Catalyst and con-
formational models for each compound were generated
automatically using the poling algorithm.37,38 This
emphasises representative coverage over a 20 kcal/mol
energy range above the estimated global minimum and
the `best searching procedure' was chosen. The training
subset of molecules with their associated conforma-
tional models and activities was submitted to Catalyst
by using the `generate hypothesis' command. The che-
mical functional groups used in this generation step
included hydrogen bond donor and acceptor, hydro-
phobic, positive ion and aromatic ring groups. The statis-
tical relevance of the various hypotheses so obtained was
assessed on the basis of their cost relative to the null
hypothesis and their correlation coecients R.37,38 The
null hypothesis is an ideal hypothesis which is the simplest
possible hypothesis that ®ts the data and for which there is
no statistically signi®cant structure including that the
experimental activities are normally distributed about
their mean. The cost function consists of two terms. The
®rst penalises the deviation between the estimated activity
and the experimental activity and the second penalises the
complexity of the hypothesis. The hypotheses were then
used to estimate the activities of the test compounds.
These activities are derived from those conformers dis-
playing the smallest root-mean square (RMS) deviations
when projected onto the hypothesis.
1
(61), 375 (27), 346 (21), 303 (23). H NMR d: 9.243, (s,
1H, NH); 7.311, (d, J=8.8 Hz, 1H, ArH); 7.032, (s, 1H,
ArH); 6.858 (d, J=8.7 Hz, 1H, ArH); 6.519, (s, 1H, H-
3); 3.901, (s, 3H, Ar-OCH3); 3.877 (s, 6H, 2ÂAr-OCH3);
3.682±3.624, (m, 1H); 3.408±3.337, (m, 1H); 3.232±2.942,
(m, 3H); 2.671±2.514, (m, 4H); 1.663±1.593, (m, CH2);
0.997, (t, J=7.2Hz, CH3). 13C NMR d: 154.2 (C-1),
151.0 (C-2), 142.3 (C-9), 131.5, 131.2 (C11a, C12a), 126.5
(C-7b), 126.2 (C-12c), 121.4 (C-12b), 112.6 (ArH), 112.0
(ArH), 110.6 (ArH), 99.9 (ArH), 60.8 (C-6a), 60.1 (Ar-
OCH3), 56.0 (CH2), 55.9, 55.7 (2ÂAr-OCH3), 47.3 (C-10),
41.5 (C-5), 29.4 (C-7), 25.2 (C-4), 18.9 (C-20), 12.1 (C-30).
Pharmacological tests. This testing was undertaken at
the Victor Chang Cardiac Research Institute, Sydney,
Australia. COS-1 membranes were prepared as descri-
bed in the literature.41 The cDNAs encoding the human
a1-adrenergic receptors were sub-cloned into the mam-
malian expression vector pMT20.41 Plasmid DNA was
used to transfect cells. Transient expression in COS-1
cells was accomplished by the diethylaminoethane-dextran
Acknowledgements
The award of a scholarship from Zonguldak Karaelmas
University in Turkey to Burak Coban, and research

214
J. B. Bremner et al. / Bioorg. Med. Chem. 8 (2000) 201±214
support from the ARC and NHMRC are gratefully
acknowledged. We also thank Professor R. M. Graham
and colleagues at the Victor Chang Cardiac Research
Institute, Sydney, Australia for the receptor binding
assays. The input of the late Dr. E. J. Browne (University
of Tasmania) into this project was also invaluable.
18. Ko, F. N.; Guh, J. H.; Yu, S. M.; Hou, Y. S.; Wu, Y. C.;
Teng, C. M. Br. J. Pharmacol. 1994, 112, 1174.
19. Ford, A. P. D. W.; Williams, T. J.; Blue, D. R.; Clarke, D.
E. Trend. Pharmacol. Sci. 1994, 15, 167.
20. Phife, D. W.; Ramos, R. A.; Feng, M.; King, I.; Gunase-
kera, S. P.; Wright, A.; Patel, M.; Pachter, J. A.; Coval, S. J.
Bioorg. Med. Chem. Lett. 1996, 6, 2103.
21. De Marinis, R. M.; Wise, M.; Hieble, J. P; Ru€olo, Jr., R.
R. In The alpha-1 Adrenergic Receptor; Ru€olo, Jr., R. R. Ed.;
Humana Press: New Jersey, 1987; pp. 211±258.
22. DeBenedetti, P. G.; Fanelli, F.; Menziani, M. C.; Cocchi,
M. J. Mol. Struc. 1995, 305, 101.
References and Notes
1. Beck-Sickinger, A. Drug Discovery Today 1996, 1, 502.
2. Graham, R. M.; Perez, D. M.; Piasciks, M. T.; Riek, R. P.;
Hwa, J. Pharmacol. Commun. 1995, 6, 15.
3. Hieble, J. P.; Bondinell, W. E.; Ru€olo, Jr., R. R. J. Med.
Chem. 1995, 38, 1838.
23. Cocchi, M.; Fanelli, F.; Menziani, M. C.; DeBenedetti, P.
G. J. Mol. Struc. 1995, 331, 79.
24. Venturelli, M.; Menziani, M. C.; Cocchi, M.; Fanelli, F.;
DeBenedetti, P. G. (Theochem)-Journal of Molecular Struc-
ture 1992, 276, 327.
4. Ru€olo, Jr., R. R.; Bondinell, W.; Hieble, J. P. J. Med.
Chem. 1995, 38, 3681.
25. Bremner, J. B.; Coban, B.; Grith, R. J. Comp. Aid. Mol.
Des. 1996, 10, 545.
5. Alexander, S. P. H.; Peters, J. A. Eds., The 1998-receptor
and ion channel nomenclature supplement, In: Trend. Phar-
macol. Sci. 9th, Elsevier Trends Journals: Cambridge, 1998.
6. Bylund, D. B.; Eikenberg, D. C.; Hieble, J. P.; Langer, S.
Z.; Lefkowitz, R. J.; Minneman, K. P.; Molino€, P. B.; Ruf-
folo, Jr., R. R. Trendelenburg, U. Pharmacol. Rev. 1994, 46,
121.
7. Forray, C.; Wetzel, J. M.; Branchek, T. A.; Bard, J. A.;
Chiu, G.; Shapiro, E.; Tang, R.; Lepor, H.; Hartig, P. R.;
Weinshank, R. L.; Gluchowski, C. Mol. Pharmacol. 1994, 45,
703.
8. Gluchowski, C.; Wetzel, J. M.; Chiu, G.; Marzabadi, M. R.;
Wong, W. C.; Nagarathnam, D. International Patent No. WO
94/22829, 1994.
9. Konkel, M. J.; Wetzel, J. M.; Cahir, M.; Craig, D.; Noble,
S. A.; Gluchowski, C. The 216th ACS National Meeting,
Boston, 23±27 August, 1998, MEDI#129.
26. Shamma, M. The Isoquinoline Alkaloids. Chemistry and
Pharmacology Academic Press Inc: New York, 1972.
27. Madrero, Y.; Elorriaga, M.; Martinez, S.; Noguera, M. A.;
Cassels, B. K.; D'Ocon, P.; Ivorra, M. D. Brit. J. Pharmacol.
1996, 119, 1563.
28. Groenewoud, K. M. Ring-D aporphine analogues. B.Sc.
honours thesis, Department of Chemistry, University of Tas-
mania, 1991.
29. Sleight, A. J.; Koek, W.; Bigg, D. C. H. Eur. J. Pharmacol.
1993, 238, 407.
30. Wetzel, J. M.; Salon, S. A.; Tamm, J. A.; Forray, C.;
Craig, D.; Nakanishi, H.; Cui, W.; Vaysse, P. J. J.; Chiu, G.;
Weinshank, R. L.; Hartig, P. R.; Branchek, T. A.; Glu-
chowski, C. Receptors and Channels 1996, 4, 165.
31. Hamaguchi, N.; True, T. A.; Saussy, D. L.; Je€s, P. W.
Biochemistry 1996, 35, 14313.
32. Brossi, A.; Dolan, L. A.; Teitel, S. Org. Synth. 1977, 56, 3.
33. Berney, D. Helv. Chim. Acta 1982, 65, 1694.
34. Bremner, J. B.; Browne, E. J. J. Heterocyc. Chem. 1975,
12, 301.
10. Meyer, M. D.; Altenbach, R. J.; Basha, F. Z.; Carroll, W.
A.; Drizin, I.; Elmore, S. W.; Ehrlich, P. P.; Lebold, S. A.;
Tietje, K.; Sippy, K. B.; Wendt, M. D.; Plata, D. J.; Plagge, F.;
Buckner, S. A.; Brune, M. E.; Hancock, A. A.; Kerwin, J. F.,
Jr. J. Med. Chem. 1997, 40, 3141.
35. Greene, J.; Kahn, S.; Savoj, H.; Sprague, P.; Teig, S. J.
Chem. Inf. Comp. Sci. 1994, 34, 1297.
11. Giardina, D.; Crucianelli, M.; Melchiorre, C.; Taddei, C.;
Testa, R. Eur. J. Pharmacol. 1995, 287, 13.
12. Testa, R.; Guarnery, L.; Angelico, P.; Poggesi, E.; Taddei,
C.; Sironi, G.; Colombo, D.; Sulpizio, A. C.; Naselsky, D. P.;
Hieble, J. B.; Leonardi, A. J. Pharmacol. Exp. Ther. 1997, 281,
1284.
36. Brooks, B. R.; Brucolleri, R. E.; Olafson, B. D.; States, D.
J.; Swaminathan, S.; Karplus, M. J. Comput. Chem. 1983, 4,
187.
37. Smellie, A.; Kahn, S. D.; Teig, S. L. J. Chem. Inf. Comp.
Sci. 1995, 35, 285.
38. Smellie, A.; Kahn, S. D.; Teig, S. L. J. Chem. Inf. Comp.
Sci. 1995, 35, 295.
13. Kenny, B.; Ballard, S.; Blagg, J.; Fox, D. J. Med. Chem.
1997, 40, 1293.
39. Grith, R.; Bremner, J. B.; Coban, B. In Pharmacophore
Perception, Development, and Use in Drug Design; Guner, O.,
Ed.; International University Line: La Jolla, in press.
40. Perrin, D. D.; Armarego, W. L. F. Puri®cation of Labora-
tory Chemicals. 3rd Ed. Pergamon Press: Oxford.
41. Perez, D. M.; Piascik, M. T.; Graham, R. M. Mol. Phar-
macol. 1991, 40, 876.
42. Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular clon-
ing: A laboratory manual. Cold Spring Harbor Laboratory:
Cold Spring Harbour, New York, 1989.
14. Gluckowski C. The discovery of selective a1A antagonists
using recombinant human a1-adrenoceptors. IBC-GPCR
Conference, Philadelphia, PA, 11±13 December 1995.
15. Foglar, R.; Shibata, K.; Horie, K.; Hirasawa, A.; Tsuji-
muto, G. Eur. J. Pharmacol. 1995, 288, 201.
16. King, H. K.; Goetz, A. S.; Ward, S. D. C.; Saussy, Jr., D.
L. Soc. Neuro. Abs. 1994, 20, 526.
17. DeBenedetti, P. G.; Fanelli, F.; Menziani, M. C.; Cocchi,
M.; Testa, R.; Leonardi, A. Bioorg. Med. Chem. 1997, 5, 809.