Morphinans and isoquinolines: Acetylcholinesterase inhibition, pharmacophore modeling, and interaction with opioid receptors

Article abstract of DOI:10.1016/j.bmc.2010.05.071

Following indications from pharmacophore-based virtual screening of natural product databases, morphinan and isoquinoline compounds were tested in vitro for acetylcholinesterase (AChE) inhibition. After the first screen, active and inactive compounds were used to build a ligand-based pharmacophore model in order to prioritize compounds for biological testing. Among the virtual hits tested, the enrichment of actives was significantly higher than in a random selection of test compounds. The most active compounds were biochemically tested for their activity on μ, δ, and κ opioid receptors.

Full text of DOI:10.1016/j.bmc.2010.05.071

Bioorganic & Medicinal Chemistry 18 (2010) 5071–5080
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Morphinans and isoquinolines: Acetylcholinesterase inhibition,
pharmacophore modeling, and interaction with opioid receptors
Daniela Schuster a, Mariana Spetea a,b, Melisa Music c, Silvia Rief a, Monika Fink a, Johannes Kirchmair a,
Johannes Schütz a,b, Gerhard Wolber a,d, Thierry Langer a,d, Hermann Stuppner c, Helmut Schmidhammer a,
Judith M. Rollinger c,
*
a Department of Pharmaceutical Chemistry, Institute of Pharmacy, University of Innsbruck and Center for Molecular Biosciences Innsbruck—CMBI, Innrain 52a/c,
A-6020 Innsbruck, Austria
b AlcaSynn Pharmaceuticals GmbH, Life Science Center, Mitterweg 24, A-6020 Innsbruck, Austria
c Department of Pharmacognosy, Institute of Pharmacy, University of Innsbruck and Center for Molecular Biosciences Innsbruck—CMBI, Innrain 52c, A-6020 Innsbruck, Austria
d Inte:Ligand Softwareentwicklungs-und Consulting GmbH, Clemens Maria Hofbauer-Gasse 6, A-2344 Maria Enzersdorf, Austria
a r t i c l e i n f o
a b s t r a c t
Article history:
Following indications from pharmacophore-based virtual screening of natural product databases, mor-
phinan and isoquinoline compounds were tested in vitro for acetylcholinesterase (AChE) inhibition. After
the first screen, active and inactive compounds were used to build a ligand-based pharmacophore model
in order to prioritize compounds for biological testing. Among the virtual hits tested, the enrichment of
actives was significantly higher than in a random selection of test compounds. The most active com-
Received 1 March 2010
Accepted 26 May 2010
Available online 2 June 2010
Keywords:
Morphinans
Acetylcholinesterase inhibition
Opioid receptor
pounds were biochemically tested for their activity on
l
, d, and
j
opioid receptors.
Ó 2010 Elsevier Ltd. All rights reserved.
Pharmacophore modeling
1. Introduction
in the brain cells of AD patients.3 A therapy with AChE inhibitors
leads to a symptomatic amelioration of memory, cognition, mood,
and daily living skills. Numerous studies confirm the vital role of
this medication corresponding to the ‘cholinergic hypothesis of
learning’.4 AChE inhibition is an approved strategy to raise ACh
concentration and thereby increase the cholinergic function in
the brain. So far, four AChE inhibitors have been approved by the
European and US regulatory authorities for mild degenerative dis-
eases: tacrine (1; CognexÒ), donepezil (2; ArizeptÒ), galanthamine
(3; ReminylÒ) and rivastigmine (4; ExelonÒ) (Chart 1).
Tacrine (1) and donepezil (2) are of synthetic origin while
galanthamine (3) is a natural product from the genus Galanthus
(Amaryllidaceae). Rivastigmine (4) is structurally closely related
to physostigmine that is a prototype AChE inhibitor from the seeds
of Physostigma venenosum (Papilionaceae). Another alkaloid from
the plant kingdom—huperzine A from the clubmoss Huperzia
serrata (Lycopodiaceae)—has been approved as drug for AD treat-
ment in the People’s Republic of China.5 Apart from these natural
products, numerous naturally occurring compounds that inhibit
AChE have been identified.6
In a society with a continuous increase of the population of over
65 years of age, a dramatic increase of dementia is observed. The
age-specific prevalence of dementia is approximately 1.5% in per-
sons aged 60–69 years and almost doubles every five years until
it reaches 40% in nonagenarians.1 One consequence of the rapid in-
crease in the aging population worldwide is that the estimated 25
million people that are currently affected by dementia will double
every 20 years. The most common manifestation of dementia—Alz-
heimer’s disease (AD)—accounts for up to 75% of all dementia
cases.2
The main cause of the loss of cognitive functions in AD patients
is a continuous decline of the cholinergic neurotransmission in cor-
tical and other regions of the human brain. At the molecular level,
cholinergic neurotransmission is mediated by the neurotransmit-
ter acetylcholine (ACh) which is rapidly hydrolyzed after its pre-
synaptic release by acetylcholinesterase (AChE). Besides its well-
known role in terminating synaptic transmission, AChE has been
found to be involved in a number of other functions. For example,
AChE is a factor in neurite growth and in accelerating the assembly
of b-amyloid into amyloid fibrils which are characteristically found
As can be seen from this example, nature is a rich source of
bioactive compounds that can be used directly as drugs or as lead
compounds for drug development. Over the last 25 years, 66% of
all new chemical entities that were approved as drugs were natural
* Corresponding author. Tel.: +43 512 507 5308; fax: +43 512 507 2939.
E-mail address: Judith.Rollinger@uibk.ac.at (J.M. Rollinger).
0968-0896/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2010.05.071

5072
D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
biological evaluation. Applications and success stories have been
reviewed recently.9–13 In our working group, we have established
an interface between the valuable resource of natural products
and the powerful VS technology by the generation of two natural
products 3D databases. The first database—the so-called DIOS data-
base—consists of 9676 structures from medicinal plants described
in the ancient ethnopharmacological source De materia medica
by Pedanios Dioscorides (1st cent AD). The second database—the
‘natural products database’ (NPD)—includes over 110,000 com-
pounds within a molecular mass between 140 and 700 Da.14 These
databases have been successfully mined for cyclooxygenase
inhibitors,14,15 acetylcholinesterase inhibitors,16–18 human rhinovi-
rus coat protein inhibitors,19 and influenza virus neuraminidase
inhibitors.20
N
NH2
N
O
O
O
1
2
tacrine
donepezil
N
N
N
O
O
HO
In our first studyon AChE inhibitors,16 a structure-based pharma-
cophore modelwas generated based on the X-ray crystal structure of
galanthamine in complex with Torpedo californica AChE (PDB entry
1qti,21 Fig. 2). This model was used to virtually screen the NPD. Bio-
logical testing of non-alkaloid virtual hits led to the successful dis-
covery of the new AChE inhibitors scopolin and scopoletin. Apart
from these coumarin derivatives, some opioids from Papaver som-
niferum, for example, ethylmorphine, were predicted as active hits.
In addition to these findings, in 1986, a study by Sim and Chua
reported opioids of the morphinan-type as AChE inhibitors (Supple-
mentary data, Chart S-1).22 However, only 15 compounds were
evaluated in this study and no further studies elaborating a struc-
ture–activity relationshipof morphinan-based AChE inhibitorswere
O
H
O
3
4
galanthamine
rivastigmine
Chart 1. Clinically used AChE inhibitors.
products, natural product derivatives, synthetic compounds mim-
icking natural products, or compounds derived from a natural prod-
ucts-based pharmacophore.7 In parallel, computer-based methods
assisting drug discovery and development have rapidly evolved
over the last 40 years.8 Virtual screening (VS) has become an estab-
lished method for selection and prioritization of compounds for
Figure 1. Study workflow for the identification of morphinan-type AChE inhibitors and elaboration of a pharmacophore model for enhancing the discovery of active
compounds.

D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
5073
reported in literature. In addition, the Ki values for the reported mor-
phinans indicated only moderate potency (12–109 M). Both the re-
sults from virtual screening and the publication by Sim and Chua
encouraged the further exploration of the effect of morphinan deriv-
atives and related scaffolds on AChE activity.
tory activity using a spectrophotometric assay with Ellman’s
l
reagent.23 At a concentration of 100
an extent of 50.4 6.2%.
lg/ml, AChE was inhibited to
2.4. AChE inhibition by in-house compounds
As a consequence of the first promising results, 416 morphinan
and non-morphinan compounds (e.g., isoquinolines) that were
readily available at the time of this study were tested at a concen-
2. Results and discussion
2.1. Study design
tration of 500
tive control, the activity of galanthamine (3; IC50 = 3.2 1.0
l
M using the spectrophotometric assay.23 As posi-
lM)
Following the strong indications from our virtual screening re-
sults of natural product databases with an established pharmaco-
phore model for AChE inhibitors (1qti-based model) and the
results from Sim and Chua,22 morphinans were chosen for biolog-
ical evaluation concerning AChE inhibition. Initially, 416 out of 812
compounds from an in-house database were evaluated for their
AChE inhibitory activity using a spectrophotometric enzyme as-
say.23 The resulting data were used to build a morphinan-focused
ligand-based pharmacophore model for AChE inhibitors. This mod-
el was used to select 14 additional compounds for biological test-
ing out of the in-house morphinan database. In parallel, further 50
compounds were chosen randomly for in vitro screening. Finally,
the results of these screening runs were compared for validating
the in silico-guided approach. An overview of the study workflow
is presented in Figure 1.
was used. Compounds that inhibited AChE more than 50% are gi-
ven in Figure 3.
2.5. Pharmacophore modeling, virtual screening, and model
validation
As a tool for the further investigation of morphinans as AChE
inhibitors, the suitability of the 1qti-model was tested. For this pur-
pose, a 3D database of 812 compounds that were already available
in-house was generated (for details see Section 4). This database
mainly consisted of morphinans (n = 542) and morphinan-based
compounds (157 indolomorphinans, 23 benzofuromorphinans, 10
dimeric morphinans, and 11 bimorphinans). In addition, it com-
prised 25 isoquinolines and 44 compounds from other chemical
scaffolds. Screening the in-house database with the 1qti-based
pharmacophore model in rigid mode resulted in a hitlist of 44 hits
including morphine. Although over 20% of all virtual hits showed an
activity in our in vitro assay (Table 2), we missed the most potent
morphinan inhibitor in our hitlist, compound 5.
2.2. Virtual hits from screening the natural products databases
The pharmacophore model derived from the PDB entry 1qti
(Fig. 2) was used for screening the DIOS and NPD.
For comparison, we generated a ligand-based pharmacophore
model based exclusively on morphinan compounds identified in
our study. The HipHop refine algorithm of Catalyst 4.11 was
employed for model generation. This algorithm does not only com-
pute the overlay of chemical features active compounds have in
common, but also considers the spatial extension of inactive com-
pounds from the training set. For example, an active compound
with a methyl substituent at a defined position and an inactive
compound with a propyl substituent at the same or a neighboring
position would provoke the positioning of exclusion volume
spheres (forbidden areas for the ligand) lining the coordinates
corresponding to the methyl group. A good composition of the
training set is crucial for representative model generation. Ideally,
Screening of the DIOS returned overall 612 hits of which 45 had
a high BestFit value of P3. Three of these highly ranked hits were
morphinan compounds: ethylmorphine, 10-oxocodeine, and 14b-
hydroxycodeine. For all these compounds, no data on AChE activity
were reported in literature. The NPD hitlist comprised 9763 hits of
which 2455 were top ranked with BestFit values of P3. The six
morphinans present in this hitlist were oripavine, codeine, dihy-
drocodeine, 14b-hydroxycodeine, neopine, and morphine.
2.3. AChE inhibition by crude opium
As a first confirmation of the results of Sim and Chua22 as well
as of our virtual hits, crude opium was tested on its AChE inhibi-
Figure 2. Pharmacophore model derived from the PDB entry 1qti (galanthamine complexed to TcAChE). Chemical features are style- and color-coded: hydrophobic feature—
yellow, hydrogen bond acceptor—red, hydrogen bond donor—green, shape restriction—wireframe. The hydrophobic feature on the 3-methoxy group was not included in the
model employed for virtual screening.

5074
D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
5
6
7
8
9
17
N
R4 = OCH3, R17 = allyl
IC50 = 1.0 µM
IC50 = 37.1 µM
IC50 = 46.2 µM
IC50 = 81.9 µM
IC50 = 105.1 µM
78 % inhibition
H
6
R4 = OH, R14 = OEt, R17 = CH3
R4 = OCH3, R14 = OCH3, R17 = CH3
R4 = OCH3, R14 = OCH3, R17 = allyl
R1 = OH, R4 = OH, R17 = CHO
1
3
14
2
10 R4 = OCH3, R14 = OH, R17 = allyl
11 R17 = CH3
12 R4 = OCH3, R14 = OH, R17 = CH3
13
14 R4 = OH, R14 = OCH3, R17 = CH3
15 R4 = OH, R17 = allyl
16 R4 = OCH3, R17 = CH2-cBu
4
5
inhibition
74 %
O
73 % inhibition
70 % inhibition
68 % inhibition
66 % inhibition
64 % inhibition
R4 = OCH3, R17 = CH2-cPr
O
N
R
N
N
A =
N
17
R1 = A, R17 = CH3
inhibition
59 % inhibition
64 %
18 R1 = OH
inhibition
inhibition
57 %
19 R4 = OBu, R14 = OCH3, R17 = CH2-cPr
20 R4 = OCH3, R6 = (R)-OH, R17 = CH2-cBu
59 %
21
R4 = OCH3, R14 = OCH3
inhibition
55 %
22 R1 = OCH3, R17 = CH3
23 R4 = OCH3
24 R2 = OH, R17 = CH3
51 % inhibition
50 % inhibition
50 % inhibition
25
R3 = OCH3, R14 = OH, R17 = CH3
26 R3 = OH, R14 = OCH3, R17 = CH3
IC50 = 76.2 µM
84 % inhibition
83 % inhibition
75 % inhibition
68 % inhibition
17
N
H
27 R3 = OH, R5 = CH3, R14 = OCH2CH3, R17 = CH3
28 R3 = OH, R14 = OCH3, R17 = (CH3)2
29
30 R14 = phenylpropyl
31 R3 = A, R14 = OH, R17 = CH3
32 R14 = OCH3
1
2
14
R3 = OCH3, R6 = =NNHCONH2,R14 = OCH3, R17 = CH3
6
inhibition
inhibition
inhibition
61 %
56 %
55 %
3
4
5
O
O
4
33
Δ1,2, R6 = OCH3, R3' = OH
IC50 = 43.1 µM
IC50 = 138.5 µM
55 % inhibition
50 % inhibition
6
7
4a
8a
,
34 Δ4a,8a, Δ6,7, R6 = OCH3 R2' = OH
35 Δ4a,8a, Δ6,7, R6 = OCH3, R3' = OH
36 Δ4a,8a, R2' = O-benzyl
2
NH
1
2'
3'
O
N
N
H
O
N
H
O
O
O
O
38
37
61 % inhibition
IC50 = 12.7 µM
Figure 3. Compounds with AChE inhibition higher than 50% from the initial in vitro screen (c = 500
l
M; n = 416). For compounds with AChE inhibitory activity higher than
85% the IC50 value is given. Tables with full chemical structures are available as Supplementary data (Tables S-1 and S-2).
Table 1
compounds of different activity from the same chemical scaffold
Training set for the generation of the HipHop refine model
are included in the training set. The exact activity of the training
compounds is not necessary for hypothesis generation. Instead,
an activity class is assessed to each compound. Compounds from
class 2 are highly active, class 1 indicates active compounds, and
class 0 includes inactives used for model generation. The training
set for our HipHop refine model is given in Table 1.
The hypothesis generation process returned ten models which
were evaluated for their ability to find actives and exclude inac-
tives from our 3D database of in-house compounds. The best model
(Fig. 4) retrieved 32 hits (using rigid search) from the in-house
database including compound 5 and morphine. Compared to the
1qti-model, the enrichment of actives raised considerably. Mining
of the in-house database using flexible search led to a hitlist with
71 hits. Although the enrichment of highly and medium actives
was less pronounced than in the hitlist from the rigid search, it still
performs better than the 1qti-model (Table 2).
Compound
Group
Class
5
6
7
8
9
26
47
49
S-1
S-2
S-3
S-4
S-5
S-6
S-7
S-8
S-9
S-10
S-11
S-12
S-13
S-14
S-15
S-16
Highly active
Highly active
Active
Active
Active
2
2
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Active
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
Inactive
As a further validation of the model, 14 virtual hits from the
database which have not been tested on their AChE-inhibiting ef-
fects were submitted to the spectrophotometric assay. For compar-
ison, a random set of 50 compounds, which was not retrieved as
virtual hits, was selected from the in-house database and also
screened with our assay.
Although both approaches—the virtual screening and the
random method—led to the identification of new AChE inhibitors
2D structures of the inactive compounds are available as Sup-
plementary data, Table S-1.

D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
5075
Table 2
Retrieval of hits from the in-house database (n = 812) using the AChE pharmacophore models
Activity class
1qti-model
rigid search
1qti-model
flexible search
Ligand-based model
rigid search
Ligand-based model
flexible search
Number of hits
Highly activea
Medium activeb
Weakly activec
Inactived
44
49
32
71
2 (4.5%)
2 (4.5%)
6 (13.6%)
14 (31.8%)
20 (45.5%)
2 (4.1%)
3 (6.1%)
6 (12.2%)
16 (32.7%)
22 (44.9%)
4 (12.5%)
7 (21.9%)
2 (6.3%)
8 (25%)
11 (34.4%)
7 (9.9%)
12 (16.9%)
6 (8.5%)
22 (31.0%)
24 (33.8%)
Unknown
a
b
c
>80% inhibition at a concentration of 500 lM.
50–80% inhibition at a concentration of 500
l
M.
M.
30–50% inhibition at a concentration of 500
l
d
Inhibition not detectable or <30% inhibition at a concentration of 500 lM.
Figure 4. HipHop refine model for AChE inhibitors from the morphinan class (left). The model consists of one hydrogen bond acceptor (green), one hydrophobic feature
(cyan), one hydrophobic aromatic feature (blue), and 54 exclusion volume spheres (black). Compound 5 fitted into the model (right).
(Tables 3 and 4), the enrichment of actives was superior for the vir-
tual screening runs (Table 5).
2.6. Chemistry
Overall, 14 out of a total of 481 tested compounds (3.1%)
showed AChE inhibition of at least 80%. This proportion is accu-
rately reflected by the fraction of highly active hits identified in
the initial in vitro screening (11 out of 416 compounds, 2.6%)
and the random in vitro screening (1 out of 50 compounds, 2.0%)
that was carried out in parallel to the pharmacophore-based selec-
tion. In order to quantitatively assess the quality of the pharmaco-
phore models, enrichment factors (EFs) were calculated (Table 6).
The syntheses of 47 and 49, which have been used in the train-
ing set for the pharmacophore model, have not been previously
reported. Compound 47 has been prepared from 5b-methyldihy-
drothebainone24,25 by alkylation with 5-chloro-1-phenyl-1H-tetra-
zole in DMF (Scheme 1). Catalytic hydrogenation of 3-benzyloxy-
14-ethoxy-4-methoxy-N-methylmorphinan-6-one26 in MeOH
afforded 49 (Scheme 2). Details of these synthetical procedures
are described in the Section 4.
Table 3
AChE inhibition by virtual hits (n = 14) from the in-house database derived from a flexible search using the ligand-based model
N
OH
N
O
O
N
H
N
O
O
O
O
O
O
O
O
40
41
39
86% Inhibition
73% Inhibition
86% Inhibitiona
IC50 = 42.9 8.3
IC50 = 65.3 11.3
lM
IC50 = 140.1 18.3 lM
l
M
42
72% Inhibition
IC50 = 193.4 25.5
lM
Only hits with an AChE inhibition of at least 50% are shown.
a
% inhibition measured at a concentration of 500 lM.

5076
D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
Table 4
AChE inhibition by random compounds not included in our virtual hitlist (n = 50)
O
N
OH
N
H
H
N
O
O
O
N
O
N
N
O
O
O
O
N
43
94% Inhibition
IC50 = 60.3 16.6 lM
44
45
56% Inhibition
51% Inhibition
IC50 = 404.6 83.1 lM
Only hits with an AChE inhibition of at least 50% are shown.
a% inhibition measured at a concentration of 500
lM.
CH3
CH3
N
OC2H5
Table 5
N
Novel AChE inhibitors selected from the pharmacophore-based search vs. random
selection
OC2H5
(i)
Activity class
Pharmacophore-based
selection
Random
selection
PhH2C
O
OCH3
O
HO
OCH3
49
O
Number of tested compounds
Highly activea
14
50
2 (14.3%)
2 (14.3%)
4 (28.6%)
6 (42.9%)
1 (2.0%)
2 (4.0%)
4 (8.0%)
43 (86.0%)
48
Medium activeb
Weakly activec
Scheme 2. Reagents and conditions: (i) Catalytic hydrogenation, 10% Pd/C, MeOH,
Inactived
30 psi, rt, 3 h.
a
b
c
>80% inhibition at a concentration of 500 lM.
2.7. Molecular docking study
50–80% inhibition at a concentration of 500
l
M.
M.
30–50% inhibition at a concentration of 500
l
d
Inhibition not detectable or <10% inhibition at a concentration of 500 lM.
The binding mode of the most potent AChE inhibitor identified
in our study—compound 5—was predicted by docking experiments
using GOLD 3.1.27 For validation of the docking settings (see Section
4), galanthamine was docked into the binding site of 1qti which
yielded satisfying results (Supplementary data, Fig. S-2). Subse-
quent docking of compound 5 revealed a similar binding site of 5
and galanthamine (3); however, different interaction patterns
were predicted for the two ligands (Fig. 5). Common protein–
ligand interactions for these compounds essentially included the
basic nitrogen positioned in an aromatic cage. While galanthamine
directly forms hydrogen bonds with the catalytically active amino
acid residues Glu199 and Ser200, no such interactions were pre-
dicted for compound 5. In contrast, a hydrogen bond was observed
to Tyr130.
The predicted binding interactions of 5 with AChE were com-
pared to the HipHop refine model. The docking solution of 5 was
imported into the PDB structure 1qti and submitted to pharmaco-
phore model generation using LigandScout.28 Of course, the docked
compound established more than three chemical interactions with
the protein; however, a reduction of chemical features led to a
model closely resembling the ligand-based HipHop refine model.
It also consisted of hydrophobic features representing the aromatic
ring and the bridging ring. A hydrogen bond was established be-
tween the 6-keto group and Tyr130 (Fig. 5).
Table 6
Enrichment factors for the 1qti-model and the ligand-based pharmacophore model
Model
1qti-
model model
rigid flexible
search search
1qti-
Ligand-based Ligand-based
model
model
flexible
search
rigid search
TP (highly actives in the
hitlist)
2
2
5
9
n (hitlist size)
A (number of actives in the 14
44
49
14
32
14
71
14
database)
N (database size)
EF
481
1.56
481
1.40
481
5.37
481
4.36
The calculations are based on the retrieval of highly active compounds (AChE
inhibition >80%) from the dataset.
CH3
CH3
N
N
(i)
2.8. Opioid receptor binding
H3CO
OH
O
H3CO
OT
O
CH3
CH3
Ph
46
5-methyldihydrothebainone
Among the investigated morphinan and isoquinoline com-
pounds, some of the most active AChE inhibitors (inhibition
P80%) identified in this study were further evaluated in in vitro
47
N
T =
N
N
receptor binding assays for the interaction with l, d, and j opioid
N
receptors.29 The binding affinities at
l and d opioid receptors were
determined by inhibition of binding of [3H][D-Ala2,Me-Phe4,Gly-
ol5]enkephalin ([3H]DAMGO) and [3H][Ile5,6]deltorphin II to rat
Scheme 1. Reagents and conditions: (i) 5-Chloro-1-phenyl-1H-tetrazole, K2CO3,
DMF, rt, 72 h.

D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
5077
Figure 5. Left: Overlay of galanthamine (green; based on the PDB entry 1qti) with the top-ranked docked binding position of compound 5. Right: Some predicted interactions
observed for 5 with the active site of AChE closely resemble the ligand-based pharmacophore model (compare to Fig. 1).
Table 8
brain membranes. The affinities of the target compounds at
oid receptors were determined by displacement of [3H]U69,593
using guinea pig brain membranes. The , d and opioid receptor
binding affinities expressed as inhibition constant (Ki) are summa-
rized in Tables 7 and 8. The selectivity for the opioid receptor
versus d and receptors was defined by the ratio of the Ki values.
j opi-
Opioid receptor binding affinities of isoquinoline compounds
l
j
Compound
Ki (nM) SEM
l
opioid receptor d opioid receptor
j opioid receptor
[3H]DAMGOa
[3H][Ile5,6]deltorphin
IIa
[3H]U69,593b
l
j
For comparison purposes, the opioid binding affinity data for mor-
phine and oxycodone (compound 25) are also included.
33
34
2318 209
>10,000
>10,000
>10,000
>10,000
>10,000
The two isoquinoline compounds identified in this study to dis-
play potent inhibitory activity of AChE (inhibition P80%) showed
a
Rat brain membranes were used.
Guinea pig brain membranes were used.
b
low or no binding to the
l, d and j opioid receptors (Table 8).
2.9. Structure–activity relationship
tives, 7 and 26, and 14-ethoxy derivatives, 6 and 27, showed high
opioid receptor affinity and selectivity, with Ki values ranging
between 0.10 and 2 nM (Table 7). The N-methyl substituted com-
pounds displayed a 6- to 65-fold enhanced affinity to the binding
site in comparison to morphine (Ki = 6.55 nM) and a 22- to 95-fold
higher receptor affinity than compound 25 (Ki = 43.6 nM).
l
Following the pattern of SAR, all morphinan compounds having
a N-methyl and N-allyl group in position 17 specifically bound to
the opioid receptors displaying selectivity for the
(Table 7). In contrast, the N-formyl substituted morphinans 9, 37,
and 43 showed no specific binding to any of the three opioid recep-
tor types (Ki >10 lM). Thus, the presence of a formyl group in posi-
l
l opioid receptor
l
Regarding the AChE inhibitory activity, no major difference in the
potency was observed between the 14-methoxy and 14-ethoxy
tion 17 significantly alters the interaction with opioid receptors.
As expected and in agreement with the previous determina-
tions of the positive influence of 14-alkoxy substituents in
N-methyl-morphinan-6-ones30,33–38 such as 14-methoxy deriva-
substituted N-methyl-14-alkoxymorphinans (IC50 of 37.1
6 and 46.2 M for 7, Fig. 3).
The N-allyl substituted morphinan-6-ones, compounds 5 and 8
lM for
l
showed both decreased
l
opioid receptor affinity and selectivity,
Table 7
Opioid receptor binding affinities of morphinan compounds
Compound
Ki (nM) SEM
Selectivity ratio
l
opioid receptor
d opioid receptor
j
opioid receptor
d/l
j/l
[3H]DAMGOa
[3H][Ile5,6]deltorphin IIa
[3H]U69,593b
Morphinec
5
6
7
8
9
6.55 0.74
20.2 1.2
1.15 0.01
2.00 0.08
67.2 1.3
>10,000
43.6 1.5
0.10 0.01
0.46 0.01
>10,000
217 19
316 26
101 14
113 9a
33
16
88
105
19
1
25
48
26
1
17
2
350
284
5
33.1 1.6
403 80
567 15
328 22
>10,000
210
7
1253
2
>10,000
1
25d
26e
27f
37
43
1087 246
4.80 0.22
12.2. 1.8f
>10,000
2658 367a
10.2 2.0a
43.2 5.7a
>10,000
61
102
94
1
>10,000
>10,000
>10,000
1
1
a
b
c
d
e
f
Rat brain membranes were used.
Guinea pig brain membranes were used.
Data from Spetea et al. 2003.29
Data from Spetea et al. 2005.30
Data from Spetea et al. 2004.31
Data from Fürst et al. 1993.32

5078
D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
being lower than that of morphine, or comparable to compound 25
(Table 7). The present receptor binding data for 8 are in agreement
with the earlier report on opioid activities in the mouse vas defer-
ens bioassay.34
When comparing N-methyl versus N-allyl 14-methoxy substi-
tuted morphinan-6-ones, 7 and 8, it was apparent that the pres-
ence of an allyl group in position 17 leads to reduced affinity and
The solvents were distilled before usage. Melting points were
determined on a Kofler melting point microscope and are uncor-
rected. IR spectra were recorded with a Mattson Galaxy Series FTIR
3000 spectrometer (in cmꢀ1). 1H NMR spectra were recorded on a
Varian Gemini 200 (200 MHz) spectrometer. Chemical shifts (d) are
reported in ppm (relative to SiMe4 as internal standard), coupling
constants (J) in Hz. Mass spectra were recorded on a Finnigan
Mat SSQ 7000 apparatus. Elemental Analyses were performed at
the Institute of Physical Chemistry at the University of Vienna,
Austria. For TLC, POLYGRAM SIL G/UV254 precoated plastic sheets
(Macherey-Nagel, Germany) were used (eluent: CH2Cl2/MeOH/
concentrated NH4OH solution, 90:9:1), and for column chromatog-
raphy, Silica Gel 60 (230–400 mesh ASTM, Fluka, Switzerland) was
used. Oxycodone (25) was purchased from Fluka, Switzerland, as
oxycodone hydrochloride solution (1 mg/mL 5% in methanol;
purity of P98%). Raw opium was used from the pharmacognostic
collection Dittrichiana, Institute of Pharmacy/Pharmacognosy,
University of Innsbruck, Austria.
selectivity for the
more favorable to an N-methyl group regarding the inhibitory
potency of AChE activity (IC50 of 46.2 M for 7 and 81.9 M for
8). On the other hand, the N-allyl 14-unsubstituted derivative com-
pound 5 (IC50 of 1 M) showed about 82-fold higher inhibitory
activity of AChE compared to its 14-methoxy analogue 8 (Fig. 3),
while both compounds display decreased opioid receptor affinity
but comparable receptor selectivity (Table 7). It appears that the
l opioid receptor. However, an N-allyl group is
l
l
l
l
l
presence of a 14-methoxy group in this class of 4-methoxy substi-
tuted N-allymorphinans causes no major alterations in binding
affinity and selectivity for
l sites.
A major impact of the substitution of positions 3 and 4—highly
depending on the presence or absence of the 4,5-ether bridge—was
observed: Opening of the 4,5-ether bridge in 3-deoxygenated mor-
phinans results in compounds with enhanced AChE inhibition. In
cases where positions 3 and 4 were both substituted (e.g., with
two methoxy groups), AChE inhibitory activity was completely
lost. Interestingly, compounds with a 4,5-ether bridge and a substi-
tuent in position 3 (e.g., 31) could still inhibit AChE. Thus, it is
hypothesized that ring opening and derivatization at position 4
leads to a steric clash of the ligand with the AChE binding site
which in turn causes the inactivity of this compound.
N-Formylation in position 17 eliminates opioid activity of mor-
phinans but not AChE inhibition (e.g., 43, Fig. 3). Thus, the basic
amine nitrogen in morphinans is not essential for AChE inhibition.
The fact that the ionizable amine nitrogen was not reflected by a
chemical feature in the ligand-based pharmacophore model sup-
ports this conclusion. The substituent in position 14 seems not to
be crucial concerning AChE inhibition.
4.2.1. 5b,17-Dimethyl-3-methoxy-4-[(1-phenyl-1H-tetrazol-5-
yl)oxy]morphinan-6-one (47)
A
mixture of 5b-methyldihydrothebainone24,25 (3.8 g,
12.0 mmol), K2CO3 (3.8 g, 27.5 mmol) and 5-chloro-1-phenyl-1H-
tetrazole (2.3 g, 12.7 mmol) in anhydrous DMF (20 mL) was stirred
at room temperature for 72 h. After filtration, the filtrate was evap-
orated and the residue dissolved in AcOEt (20 mL), washed with H2O
(2 ꢁ 10 mL), dried (Na2SO4) and evaporated to give a crystalline res-
idue (4.98 g) which was recrystallized from MeOH (5 mL) to yield
2.94 g (53%) of 47. An analytical sample was obtained by recrystalli-
zation of a small amount from MeOH: mp 180–182 °C; IR (KBr):
(C@O) 1700 cmꢀ1 1H NMR (CDCl3) d 7.66 (m, 5 arom H), 6.92 (d,
;
J = 8.0, 1 arom H), 6.65 (d, J = 8.0, 1 arom H), 3.52 (s, CH3O), 2.37 (s,
CH3N), 1.11 (d, J = 7, C5–CH3). Anal. (C26H29N5O3) C, H, N.
4.2.2. 14-Ethoxy-3-hydroxy-4-methoxy-17-methylmorphinan-
6-one (49)
A
mixture of 3-benzyloxy-14-ethoxy-4-methoxy-17-meth-
3. Conclusion
ylmorphinan-6-one26 (500 mg, 1.15 mmol), MeOH (60 mL) and
10% Pd/C catalyst (60 mg) was hydrogenated at 30 psi and room
temperature for 3 h. The catalyst was filtered off and washed with
MeOH, and the filtrate was evaporated. The residue (470 mg oil)
was crystallized from MeOH (1 mL) to afford 402 mg (80%) of 49:
This is the first study that presents a large investigation of mor-
phinan and isoquinoline compounds as AChE inhibitors. Starting
from virtual screening hits, a series of AChE inhibitors were identi-
fied with activities in the low micromolar range. In parallel, a phar-
macophore model was established that may serve as a useful tool
to prioritize compounds for biological evaluation of their choliner-
gic enhancing properties and for future synthesis projects.39
mp 209–211 °C; IR (KBr): (C@O) 1705 cmꢀ1 1H NMR (CDCl3) d
;
6.78 (d, J = 8.0, 1 arom H), 6.71 (d, J = 8.0, 1 arom H), 3.79 (s,
CH3O), 2.32 (s, CH3N), 1.27 (t, J = 6, CH2CH3); CI-MS m/z 346
(M++1). Anal. (C20H27NO4ꢂ0.1MeOH) C, H, N.
Elemental analysis data for both compounds are available as
Supplementary data (Table S-4).
4. Experimental
4.1. Spectrophotometric assay for AChE inhibitory activity
4.3. Molecular modeling and virtual screening
The AChE inhibitory activity was determined using a modified
Ellman’s method23,40 with electric eel AChE (EC 3.1.1.7), acetylthio-
choliniodide, and 5,50-dithiobis-(2-nitrobenzoic acid): Sigma–Al-
drich Chemie Gmbh, Steinheim, Germany; Galanthamine. HBr
(Tocris; Cookson Ltd, Avonmouth UK Bristol)] served as the positive
Molecular modeling studies were carried out on an Intel Pen-
tium Core 2 Duo 6400 equipped with 1 GB RAM running Linux
Fedora Core 6.
control in our assay (IC50 of 3.2 1.0 lM) using a 96-well micro-
plate assay as previously described.16 The percentage of the enzyme
inhibition was calculated by determining the rate in presence of
inhibitor and the vehicle (containing 1% DMSO) compared to the
rate in the control sample (n = 4) and analyzed with Student’s t-test.
4.3.1. Data preparation
All compounds used in the in silico approach were generated
within Catalyst Version 4.1141 and submitted to conformational
analysis using the ‘best’ option with a maximum of 250 conformers
per molecule and a maximum conformer energy of 20 kcal/mol
above the calculated energy minimum. The 3D database of in-
house morphinan and isoquinoline compounds was generated
using the catDB module of Catalyst. A maximum of 255 conformers
per molecule was included into the database.
4.2. Chemistry
The required reagents as well as anhydrous DMF were pur-
chased from Fluka, Switzerland in the highest purities available.

D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
5079
4.3.2. HipHop Refine pharmacophore model generation
The HipHop Refine algorithm of Catalyst generates a common
feature pharmacophore model from compounds labeled as highly
active from the training set. Chemical features considered in the
hypothesis generation process were hydrogen bond acceptors, pos-
itively ionizable groups, hydrophobic, and hydrophobic aromatic
features. In the first step—the constructive phase—a set of common
feature models is generated. In a second step—the optimization
phase—exclusion volume spheres are strategically placed where
steric interactions contributing to biological (in-)activity can be
approximated. This information is taken from the inactive com-
pounds included in the training set. For example, if the compound
ABC is active and ABCD is not active—even though ABCD contains
the same pharmacophore as ABC—differences in the steric bulk
are estimated to carry responsibility for the absence of ABCDs bio-
logical activity. At the 3D location of D, an exclusion volume sphere
is placed.
0.3–0.5 mg protein and different concentrations the test com-
pound as described.29 Rat brain membranes were incubated either
with [3H]DAMGO (Perkin–Elmer, Boston, MA, USA; 45 min, 35 °C)
or [3H][Ile5,6]deltorphin II (Institute of Isotopes Co. Ltd, Budapest,
Hungary; 45 min, 35 °C). Guinea pig brain membranes were incu-
bated with [3H]U69,593 (Perkin–Elmer, Boston, MA, USA; 30 min,
30 °C). Non-specific binding was determined in the presence of
10 lM unlabeled naloxone. Reactions were terminated by rapid fil-
tration through Whatman glass fiber filters GF/B pretreated with
0.1% polyethylenimine ([3H]U69,593) or GF/C ([3H]DAMGO and
[3H][Ile5,6]deltorphin II) using a Brandel M24R Cell Harvester, fol-
lowed by three washings with 5 ml of ice-cold 50 mM Tris–HCl
buffer (pH 7.4.). Inhibition constant (Ki) values were calculated
from competition binding curves using GraphPad Prism (San Diego,
CA, USA) program. The values are expressed as the mean S.E.M of
2–4 independent experiments, each performed in duplicate.
Acknowledgments
4.3.3. Virtual screening
Screening of the DIOS and NPD using the 1qti-model was per-
formed employing the fast (rigid) fitting algorithm of Catalyst
4.11.41 Fit values were computed using best fit calculation which
means that the ligand is minimized into the model before calculat-
ing the fit.
This study was granted by the Austrian Research Promotion
Agency (FFG, Bridge Project 810059). We acknowledge the techni-
cal assistance of Beatrix Jungwirth.
Supplementary data
4.3.4. Enrichment factor (EF) calculation
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2010.05.071.
The EF is a measure how well active compounds are found by
the model in comparison to inactive compounds or decoys, that
is, compounds that are supposed to be inactive. It is not only a
measure how well the model finds highly active hits from the data-
base, but also compares the fraction of actives from a hitlist with
the ratio of actives/all compounds from the entire screening data-
base. The EF is calculated using the equation42,43
References and notes
1. Qiu, C.; De Ronchi, D.; Fratiglioni, L. Curr. Opin. Psychiatry 2007, 20, 380.
2. Qiu, C.; Kivipelto, M.; von Strauss, E. Dialogues Clin. Neurosci. 2009, 11, 111.
3. Zhang, X. Curr. Drug Targets CNS Neurol. Disord. 2004, 3, 137.
4. Bartus, R. T.; Dean, R. L.; Beer, B.; Lipps, A. S. Science 1982, 217, 408.
5. Jiang, H.; Luo, X.; Bai, D. Curr. Med. Chem. 2003, 10, 2231.
6. Houghton, P. J.; Ren, Y.; Howes, M.-J. Nat. Prod. Rep. 2006, 23, 181.
7. Newman, D. J.; Cragg, G. M. J. Nat. Prod. 2007, 70, 461.
8. Ekins, S.; Mestres, J.; Testa, B. Br. J. Pharmacol. 2007, 152, 9.
9. Langer, T.; Wolber, G. Pure Appl. Chem. 2004, 76, 991.
10. Ekins, S.; Mestres, J.; Testa, B. Br. J. Pharmacol. 2007, 152, 21.
11. Kirchmair, J.; Distinto, S.; Schuster, D.; Spitzer, G.; Langer, T.; Wolber, G. Curr.
Med. Chem. 2008, 15, 2040.
12. Langer, T.; Hoffmann, R. D. Exp. Opin. Drug Discov. 2006, 1, 261.
13. Villoutreix, B. O.; Renault, N.; Lagorce, D.; Sperandio, O.; Montes, M.; Miteva, M.
A. Curr. Protein Pept. Sci. 2007, 8, 381.
TP=n
EF ¼
A=N
where TP is the number of true positives (active hits) in the hitlist, n
is the size of the hitlist, A is the number of active ligands in the
entire database, and N is the number of all compounds in the entire
database. For EF calculations, all 481 compounds for which the
AChE inhibitory potency was evaluated in vitro were considered.
14. Rollinger, J. M.; Haupt, S.; Stuppner, H.; Langer, T. J. Chem. Inf. Comput. Sci. 2004,
44, 480.
15. Rollinger, J. M.; Bodensieck, A.; Seger, C.; Ellmerer, E. P.; Bauer, R.; Langer, T.;
Stuppner, H. Planta Med. 2005, 71, 399.
16. Rollinger, J. M.; Hornick, A.; Langer, T.; Stuppner, H.; Prast, H. J. Med. Chem.
2004, 47, 6248.
17. Rollinger, J. M.; Mock, P.; Zidorn, C.; Ellmerer, E. P.; Langer, T.; Stuppner, H.
Curr. Drug Discov. Technol. 2005, 2, 185.
18. Schuster, D.; Kern, L.; Hristozov, D. P.; Terfloth, L.; Bienfait, B.; Laggner, C.;
Kirchmair, J.; Grienke, U.; Wolber, G.; Langer, T.; Stuppner, H.; Gasteiger, J.;
Rollinger, J. M. Comb. Chem. High Throughput Screening 2010, 13, 54.
19. Rollinger, J. M.; Steindl, T. M.; Schuster, D.; Kirchmair, J.; Anrain, K.; Ellmerer, E.
P.; Langer, T.; Stuppner, H.; Wutzler, P.; Schmidtke, M. J. Med. Chem. 2008, 51,
842.
20. Grienke, U.; Schmidtke, M.; Kirchmair, J.; Pfarr, K.; Wutzler, P.; Dürrwald, R.;
Wolber, G.; Liedl, K. R.; Stuppner, H.; Rollinger, J. M. J. Med. Chem. 2010, 53, 778.
21. Bartolucci, C.; Perola, E.; Pilger, C.; Fels, G.; Lamba, D. Protein: Struct., Funct.,
Genet. 2001, 42, 182.
4.3.5. Docking
Docking experiments were performed employing GOLD 3.1.27 The
protein structure from the PDB entry 1qti was prepared for docking
using Sybyl 8.0.44 The co-crystallized ligand galanthamine was
deleted from the file. Hydrogens were added using Sybyl’s biopoly-
mer tool. All water molecules included in the active site of the pro-
tein were set to ‘toggle’ and ‘spin’ using a perl script.45 The cavity
site was detected by the program GOLD using the coordinates of
the co-crystallized galanthamine as starting point and allowing
20 Å around this area as location for the binding. As a validation
for the docking, the bioactive conformation of galanthamine was
also submitted to docking using the same settings. The starting
conformation for compound 5 docking was a low-energy con-
former generated using Catalyst’s modified CHARMm force field-
based 3D structure minimization.
22. Sim, M. K.; Chua, M. E. Clin. Exp. Pharmacol. Physiol. 1986, 13, 159.
23. Ellman, G. L.; Courtney, K. D.; Andres, V. J.; Featherstone, R. M. Biochem.
Pharmacol. 1961, 7, 88.
24. Small, L.; Fitch, H. M.; Smith, W. E. J. Am. Chem. Soc. 1936, 58, 1457.
25. Schmidhammer, H.; Fritsch, F.; Burkard, W. P.; Eggstein-Aeppli, L.; Hefti, F.;
Holck, M. I. Helv. Chim. Acta 1988, 71, 642.
4.4. Opioid receptor binding assay
26. Schmidhammer, H.; Jennewein, H. K.; Krassnig, R.; Traynor, J. R.; Patel, D.; Bell,
K.; Froschauer, G.; Mattersberger, K.; Jachs-Ewinger, C.; Jura, P.; u.a. J. Med.
Chem. 1995, 38, 3071.
27. The Cambridge Crystallographic Data Centre—CCDC: Cambridge, 2004–2008.
28. Wolber, G.; Langer, T. J. Chem. Inf. Model. 2005, 45, 160.
29. Spetea, M.; Tóth, F.; Schütz, J.; Ötvös, F.; Tóth, G.; Benyhe, S.; Borsodi, A.;
Schmidhammer, H. Eur. J. Neurosci. 2003, 18, 290.
Membrane fractions were prepared from Sprague–Dawley rat
or guinea pig brains (Institut für Labortierkunde und Laborgenetik,
Medizinische Universität Wien, Himberg, Austria) as previously
described.29 Binding experiments were performed in 50 mM
Tris–HCl buffer (pH 7.4.) in a final volume of 1 ml containing

5080
D. Schuster et al. / Bioorg. Med. Chem. 18 (2010) 5071–5080
30. Spetea, M.; Greiner, E.; Aceto, M. D.; Harris, L. S.; Coop, A.; Schmidhammer, H. J.
Med. Chem. 2005, 48, 5052.
38. Lattanzi, R.; Spetea, M.; Schüllner, F.; Rief, S. B.; Krassnig, R.; Negri, L.;
Schmidhammer, H. J. Med. Chem. 2005, 48, 3372.
31. Spetea, M.; Friedmann, T.; Riba, P.; Schütz, J.; Wunder, G.; Langer, T.;
Schmidhammer, H.; Fürst, Z. Eur. J. Pharmacol. 2004, 483, 301.
32. Fürst, Z.; Búzás, B.; Friedmann, T.; Schmidhammer, H.; Borsodi, A. Eur. J.
Pharmacol. 1993, 236, 209.
33. Schmidhammer, H.; Aeppli, L.; Atwell, L.; Fritsch, F.; Jacobson, A. E.; Nebuchla,
M.; Sperk, G. J. Med. Chem. 1984, 27, 1575.
39. Rollinger, J. M.; Prast, H.; Schuster, D.; Langer, T.; Hornick, A.; Stuppner, H. Eur.
Pat. Appl. EP 09015938.5.
40. Ingkaninan, K.; de Best, C. M.; van der Heijden, R.; Hofte, A. J. P.;
Karabatak, B.; Irth, H.; van der Greef, J.; Veerpoorte, R. J. Chromatogr.,
2000, 872, 61.
A
41. Accelrys Software Inc.: San Diego, CA, USA, 2005–2009.
34. Schmidhammer, H.; Schratz, A.; Mittendorfer, J. Helv. Chim. Acta 1990, 73, 1784.
35. Schmidhammer, H. Curr. Top. Med. Chem. 1993, 1, 261.
36. Schütz, J.; Spetea, M.; Koch, M.; Aceto, M. D.; Harris, L. S.; Coop, A.;
Schmidhammer, H. J. Med. Chem. 2003, 46, 4182.
42. Jacobsson, M.; Liden, P.; Stjernschantz, E.; Bostroem, H.; Norinder, U. J. Med.
Chem. 2003, 46, 5781.
43. Diller, D. J.; Li, R. J. Med. Chem. 2003, 46, 4638.
44. Tripos: St. Louis, MO, USA, 2008.
37. Greiner, E.; Spetea, M.; Krassnig, R.; Schüllner, F.; Aceto, M. D.; Harris, L. S.;
Traynor, J. R.; Woods, J. H.; Coop, A.; Schmidhammer, H. J. Med. Chem. 2003, 46,
1758.
45. Schuster, D.; Maurer, E.; Laggner, C.; Nashev, L. G.; Wilckens, T.; Langer, T.;
Odermatt, A. J. Med. Chem. 2006, 49, 3454.