Computer-aided structure-based design of multitarget leads for Alzheimer's disease

Article abstract of DOI:10.1021/ci500555g

Alzheimer's disease is a neurodegenerative pathology with unmet clinical needs. A highly desirable approach to this syndrome would be to find a single lead that could bind to some or all of the selected biomolecules that participate in the amyloid cascade, the most accepted route for Alzheimer disease genesis. In order to circumvent the challenge posed by the sizable differences in the binding sites of the molecular targets, we propose a computer-assisted protocol based on a pharmacophore and a set of required interactions with the targets that allows for the automated screening of candidates. We used a combination of docking and molecular dynamics protocols in order to discard nonbinders, optimize the best candidates, and provide a rationale for their potential as inhibitors. To provide a proof of concept, we proceeded to screen the literature and databases, a task that allowed us to identify a set of carbazole-containing compounds that initially showed affinity only for the cholinergic targets in our experimental assays. Two cycles of design based on our protocol led to a new set of analogues that were synthesized and assayed. The assay results revealed that the designed inhibitors had improved affinities for BACE-1 by more than 3 orders of magnitude and also displayed amyloid aggregation inhibition and affinity for AChE and BuChE, a result that led us to a group of multitarget amyloid cascade inhibitors that also could have a positive effect at the cholinergic level.

Full text of DOI:10.1021/ci500555g

Subscriber access provided by UNIV NEW ORLEANS
Article
Computer Aided Structure Based Design of
Multitarget Leads for Alzheimer’s Disease
Jose Luis Dominguez, Fernando Fernández-Nieto, Marian Castro, Marco Catto, M. Rita Paleo, Silvia
Porto, F. Javier Sardina, Jose Brea, Angelo Carotti, M. Carmen Villaverde, and Fredy S. Sussman
J. Chem. Inf. Model., Just Accepted Manuscript • Publication Date (Web): 07 Dec 2014
Downloaded from http://pubs.acs.org on December 9, 2014
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Chemical Information and Modeling is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Computer Aided Structure Based Design of Multitarget
Leads for Alzheimer’s Disease
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
José L. Domínguez,1 Fernando Fernández-Nieto,2 Marian Castro,3 Marco Catto,4 M.
Rita Paleo,1,2 Silvia Porto,2 F. Javier Sardina,2 José M. Brea,3 Angelo Carotti,4 M.
Carmen Villaverde,1* and Fredy Sussman1*
1 Departamento de Química Orgánica, Facultad de Química, Universidad de Santiago
de Compostela, 15782-Santiago de Compostela, Spain.
2 Centro Singular de Investigación en Química Biológica y Materiales Moleculares
(CIQUS), Universidad de Santiago de Compostela, 15782-Santiago de Compostela,
Spain.
3
Departamento de Farmacología, Instituto de Farmacia Industrial, Centro de
Investigación en Medicina Molecular y Enfermedades Crónicas (CIMUS), Universidad
de Santiago de Compostela, 15782-Santiago de Compostela, Spain.
4
Dipartimento di Farmacia-Scienze del Farmaco, Università degli Studi di Bari
“Aldo Moro”, 70125-Bari, Italy.
1
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 2 of 39
1
2
3
ABSTRACT
4
5
6
7
8
9
Alzheimer’s disease is a neurodegenerative pathology with unmet clinical needs. A
highly desirable approach to this syndrome would be to find a single lead that could
bind to some or all of the selected biomolecules that participate in the amyloid cascade,
the most accepted route for Alzheimer disease genesis. In order to circumvent the
challenge posed by the sizable differences in the binding sites of the molecular targets
we propose a computer assisted protocol based on a pharmacophore and a set of
required interactions with the targets, which allows for the automated screening of
candidates. We used a combination of docking and molecular dynamics protocols in
order to discard non-binders, optimize the best candidates and provide a rationale for
their potential as inhibitors. To provide a proof of concept we proceeded to screen the
literature and databases, a task that allowed us to identify a set of carbazole containing
compounds that initially showed affinity only for the cholinergic targets in our
experimental assays. Two cycles of design based on our protocol led to a new set of
analogues that were synthesized and assayed. The assay results revealed that the
designed inhibitors improve their affinity for BACE-1 by more than three orders of
magnitude, as well as displaying amyloid aggregation inhibition and affinity for AChE
and BuChE, a result that led us to a group of multitarget amyloid cascade inhibitors that
also could have a positive effect at the cholinergic level.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2
ACS Paragon Plus Environment

Page 3 of 39
Journal of Chemical Information and Modeling
1
2
3
INTRODUCTION
4
5
6
7
8
9
Alzheimer’s disease (AD), a cerebral neurodegenerative pathology that is the main
cause of dementia in older people, is characterized by the progressive formation of
insoluble amyloid plaques and fibrillary tangles. In spite of the enormous efforts carried
out by academic institutions and pharmaceutical industry, AD is an illness with unmet
needs since the only drugs available in clinic (i.e., Acetylcholinesterase (AChE)
inhibitors and a NMDA receptor antagonist) have a palliative effect and do not modify
the course of the disease.1
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The most accepted hypothesis for the origin of AD is the one related to the amyloid
cascade, which singles out the aggregates and fibrils of the amyloid peptide (Aβ, a
peptide of 40 or 42 residues) as the cause of AD, since their presence interrupts the
synaptic connections and precludes the right inter-neuron orientation.2,3 The Abeta
peptides are produced by the hydrolysis of the amyloid precursor protein (APP) by two
aspartic proteases (γ-secretase and BACE-1). The last decade has witnessed an all out
effort to discover inhibitors of these two enzymes that could become drug leads for the
treatment of AD, but all the candidates have failed either at pre-clinical or clinical
stages.3 The inhibition of Aβ peptide aggregation has become an important target for
drug lead discovery in itself, although no Abeta aggregation inhibitor has surpassed the
clinical assays either.1 On the other hand Inestrosa et al.4 have shown that the peripheral
anionic site (PAS) in AChE could be a therapeutic target, since it is a nucleation site for
the amyloid Abeta peptide aggregation and hence its inhibition could hinder this
process. Finally, the leads that bind AChE, could also bind butirylcholinesterase
(BuChE), and hence have a bearing on the cholinergic pathway by precluding the
hydrolysis of acetylcholine and probably enhancing (albeit temporarily) cognition in
AD patients.
The multiplicity of amyloid cascade AD targets (described above) opens the door to a
new approach towards single molecule polypharmacology, which entails the search of a
molecule that could bind to all or some of the selected amyloid cascade targets. This
novel paradigm which deviates radically from the one target one molecule strategy has
recently received increasing attention.5-7 Probably the major hurdle in the search for
multitarget leads lays on the substantial structural and specificity differences amongst
the binding sites of the amyloid cascade targets, which hinders drastically this
therapeutic strategy.6 These differences are especially noticeable between BACE-1 and
the other amyloid cascade targets like AChE and Abeta peptides aggregation. It has
3
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 4 of 39
1
2
3
4
5
6
7
8
9
been shown that the compounds aimed at these latter targets share some common traits
(like the presence of aromatic moieties), an issue that explains the large body of work
on this multitarget subset.7 It would be desirable to find a systematic computer assisted
protocol leading to compounds that bind to the selected set of targets. Herein we
postulate the existence of a pharmacophore for multitarget approach to AD, which bears
some of the traits of the know leads that bind a variety of amyloid cascade targets.5-7
This pharmacophore could be used for systematic search of novel multitarget leads. As
a proof of concept, it has enabled us to identify in the literature some candidates bearing
the requirements of our proposed structural framework.8 Nevertheless, the results of our
experimental binding assays indicate that although these compounds bind AChE, they
have a modest fibril formation inhibition and display much lower affinities for BACE-1.
Our main endeavor in this work was to generate congeneric ligands with better affinities
for all the amyloid cascade targets selected here. For this sake we developed a protocol
that relies on molecular docking based screening for the enzyme targets and molecular
dynamics simulations for peptide aggregation in order to search for more potent
analogues of our starting candidates. The main novelty of the protocol elaborated by us
is that it allows for systematic search of multitarget leads and their subsequent
optimization, given the fact that our hit docking poses should comply with the set of
inhibitor-protein interactions assigned to our pharmacophore. Review of the predicted
docking poses in AChE and BACE-1 revealed possible ways of enhancing binding
affinity. Two cycles of design based on our protocol lead to a new set of analogues that
were synthesized and assayed. The assay results revealed that the designed inhibitors
improve their affinity for BACE-1 by more than three orders of magnitude, as well as
showing affinity for amyloid aggregates, AChE and BuChE, a result that led us to a
group of truly multitarget candidates that interrupt the amyloid cascade while having a
positive effect at the cholinergic level.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Furthermore, our results have allowed us to explore some basic questions that relate to
molecular recognition issues in the different amyloid cascade targets, including the
charge state of the BACE-1 main anchoring group and the mechanism by which the best
leads may interrupt the amyloid peptide aggregation.
4
ACS Paragon Plus Environment

Page 5 of 39
Journal of Chemical Information and Modeling
1
2
3
Multitarget amyloid cascade pharmacophore.
4
5
6
7
8
9
Our pharmacophore was built by identifying the specific moieties that are recognized by
the binding pockets of the different amyloid cascade targets. For instance, an essential
feature in BACE-1 inhibitors is a functional group (such as hydroxyethylene,
hydroxyethylamine, guanidium, etc.) which is able to interact through hydrogen bonds
and ion pairs with the Asp dyad, the catalytic machinery of this enzyme.9 Our choice for
this kind of functionality has been based on recent studies in our lab that have identified
the hydroxyethylamine as an Asp dyad anchor which favors good performance at
cellular level.10,11 On the other hand, an overview of the AChE inhibitors indicate that
many of them contain one or two aromatic moieties which interact through π-stacking
interactions with clusters of aromatic residues present both in the catalytic anionic site
(CAS) and in the peripheral binding site (PAS).12 Both AChE binding sites are
separated by a long gorge. Hence, our ideal AChE inhibitor should include an optimum
length spacer that will connect the aromatic moiety residing in the CAS with the one at
the PAS. If we accomplish this aim, the resulting lead should have both a palliative
effect on AD and hinder amyloid aggregation. Finally, some of the amyloid aggregation
inhibitors share with the AChE ligands a common feature, that is, the presence of
aromatic groups that target some of the residue clusters rich in aromatic residues present
in the amyloid peptide (such as the LVFFA segment). Hence, a multitarget
pharmacophore and its possible interactions in the amyloid cascade binding sites could
be described by a scheme such as the one presented in Figure 1. A number of the
features that appear in this pharmacophore have been used in the search of multitarget
cascade leads.5-7 Herein we demonstrate that the pharmacophore described could be
used for multitarget screening and subsequent lead optimization.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
BACE-1
O H
O
Asp
Asp
O
Tyr70
-
Trp84
O
H
OH
O
N
H
+
N
H
H
AChE /PAS
Ar
Ar
CAS / AChE
Phe330
Trp279
N
H
Figure 1. Schematic view of our multitarget pharmacophore (red) bound to the Asp dyad of
BACE-1 (blue) and to the CAS and PAS of TcAChE (black).
5
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 6 of 39
1
2
3
4
5
6
7
8
9
The computer assisted search for candidates based on the pharmacophore shown in
Figure 1 lead us to find some compounds (1) (see Figure 2) with neurogenerative and
neuroprotective properties in mice.8 Nevertheless, the molecular therapeutic targets for
these compounds have not been identified. The stated aim of the study that led to these
compounds in the first place was to find analogues of dimebon, a carbazole derivative
that in itself showed good promise in AD assays in animals but failed in phase three
clinical trials.13 Moreover, other carbazole derivatives have shown to be good Abeta
aggregation inhibitors.14 As seen from structure 1 in Figure 2, these compounds present
all of the structural features that make them good leads for all amyloid cascade targets.
On one hand the hydroxyethylamine moiety provides an anchor for BACE-1 binding,
while the aromatic moieties on both ends (carbazole and substituted phenyl groups)
could be a source of affinity of these compounds for AChE and Aβ peptide oligomers.
Based on this information we decided to investigate if the original group of compounds
(1) owed its beneficial properties at the CNS level to their binding to some of the
amyloid cascade targets, and whether we could produce analogues with better affinity
for a wider range of amyloid cascade targets.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
X
N
OH
H
N
N
N
N
Y
1
R
Dimebon
Figure 2. Structure of dimebon and the analogues (1) selected for this work.
COMPUTATIONAL METHODS
Pharmacophore based candidate search. The quest for compounds bearing the traits
of the pharmcophore was performed using the SMILES structure descriptors on
bibliographical databases. A variety of different structures were identified, some of
them bearing a hydroxyethylamine and others bearing unwanted fragments like amino
ketones. The unwanted candidates were pruned by using a simple AWK script.
6
ACS Paragon Plus Environment

Page 7 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Docking protocols. We carried out the docking simulations with the suite of modules
resident in the program GOLD.15 For each docking run we employed a minimum of
100,000 and a maximum of 1,250,000 genetic generated poses. We scored the poses
with three of the scoring functions resident in GOLD (i.e., GoldScore,16,17 ChemScore18-
20 and ChemPLP21 ).
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Docking to BACE-1. One of the most outstanding structural features of the active site
of BACE-1 is a residue segment that forms a flap (residues 69 to 75) whose
conformational variability allows for a great variety of binding poses for those
inhibitors differing in size and shape.9 This loop, which forms part of the S1 pocket,
closes in onto the active site during substrate catalysis. There are inhibitors that bind
explicitly to the flap avoiding its closure and hence hampering catalysis. Perusal of the
many BACE-1-inhibitor complex structures indicates that the flap presents a wide
variety of openings depending on the inhibitor’s chemical nature. In order to take into
account the flap variability in our docking calculations, we have carried out our docking
simulations with three protein templates that differ in the opening of this loop. The first
of these protein structures comes from the complex between BACE-1 and the peptide-
mimetic inhibitor OM99-2 (PDB entry 1FKN) and has a closed flap.22 The second
structure has a non-peptidic inhibitor bound to the enzyme with a middle range opening
of the flap (PDB entry 3KMY),23 while the last template was obtained from the
unbound structure of BACE-1, and has a fully opened flap (PDB entry 1W50).24 Figure
3 exhibits the superposition of these three structures with the flap structures displayed in
different colors.
Our previous studies, which combined SPR binding experiments with molecular
mechanics based calculations, predicted that the Asp dyad in BACE-1 has a
monoprotonated state (at pH values ranging from 4.5 to 7.4) when bound to compounds
with a hydroxyethylamine moiety.10 Thus, we assigned this Asp dyad state to BACE-1
in our docking screening calculations. On the other hand, the aniline present in our
compounds could be protonated or neutral at the acidic pH (4.5-5.0) at which the
experimental assays are carried out. For this reason docking simulations for these
compounds were carried out both with neutral and protonated anilines.
As a first step we cleaned up the target structures mentioned above, eliminating the
crystallographic water molecules, discarding alternative conformations and adding
hydrogen atoms, by using the Discovery Studio (DS) modules.25 We defined the
7
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 8 of 39
1
2
3
4
5
targeted binding site for GOLD docking to these proteins as all the atom residues that
were within 6Å of the inhibitor in the 3KMY complex.
6
7
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Superimposition of 1FKN, 3KMY and 1W50 protein ribbon structures. The flap loops
are colored in green, blue and red respectively.
For each of the structures differing in the flap opening, we then performed docking
calculations with one of the three scoring functions (ChemScore, GoldSCore and
ChemPLP). In each of the runs we searched for the presence of a single hit pose
amongst the top ten docking resulting poses. We define a hit pose as the one that will
fulfill the hydrogen bond pattern between the hydroxyethylamine fragment and the Asp
dyad shown for our pharmacophore in Figure 1. Global success rates for every
compound candidate were calculated by adding up the number of single hits of the
docking simulations carried out with the three scoring functions for our three protein
templates differing in the flap opening described above.
Docking to AChE. For the docking predictions to this target we used the X-ray
structure of Torpedo californica (TcAChE) complexed with a bis-tacrine analogue
(PDB entry 1ODC).26 A close up of the binding pose is shown in Figure 4.
8
ACS Paragon Plus Environment

Page 9 of 39
Journal of Chemical Information and Modeling
1
2
3
4
N
5
6
7
HN
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
NH
N
Figure 4. Close up to the binding of the bis-tacrine analogue indicated to both binding sites
(CAS and PAS) in TcAChE (from PDB entry 1ODC). The inhibitor is shown in blue and the
aromatic residues in green. Notice the π-π interactions between the aromatic moieties and the
side chains of residues Trp 84 and Phe 330 at the CAS and with residues Tyr 70 and Trp 279 at
the PAS.
In the same way as for BACE-1, we cleaned up the target structure (PDB id 1ODC),
eliminating the crystallographic water molecules, discarding alternative conformations
and adding hydrogen atoms, by using the Discovery Studio (DS) modules.25 We defined
the binding site as all the AChE atom residues that lay at 6 Å from the ligand of this
complex (i.e., the bis-tacrine analogue). Again, the docking conformations generated by
the genetic algorithm were evaluated by the same scoring functions used for BACE-1.
In each run the top ten poses were screened for compliance with the face-to-face π-
stacking interactions that will optimize the inhibitor affinity both at the CAS and PAS in
AChE, as shown in Figure 1 and exemplified by the bis-tacrine analogue in Figure 4. In
the same way as in BACE-1 the hit success rate was determined by searching for a hit
amongst the top ten poses for the docking simulations evaluated with our three scoring
functions. Global success rates were calculated by adding up the number of single hits.
Aβ aggregation inhibition protocol. As we explained in the introduction, one of the
therapeutic targets is the inhibition of the Aβ peptide oligomerization that leads to the
neuronal toxic species. These peptides do not have a unique binding site for the ligands
as in the case of other AD amyloid cascade enzymatic targets. Depending on the nature
of the ligand, there are multiple binding sites that share similar sequence motifs. For
instance, some aromatic moiety inhibitors tend to bind to aromatic residue clusters
9
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 10 of 39
1
2
3
4
5
6
7
8
9
present in the LVFFA segment of the amyloid peptide.27 Hence, many in silico
aggregation inhibition studies merge docking protocols with molecular dynamics (MD)
simulations.28 We have explored a different approach recently developed by the group
of Caflisch.27a It consists in analyzing lengthy MD simulations of a segment of the Aβ
peptide (Aβ12-28) in the presence of a given candidate. This peptide fragment has been
chosen for several reasons. Firstly, the first 11 residues were omitted since they lack any
definite secondary structure in some NMR amyloid aggregate structures.29 Moreover,
the selected segment contains one of the regions (LVFFA) around which a beta hairpin,
the template for Abeta aggregation forms, and is also one of the ligand’s binding spots
for a number of ligands.27 Finally, this segment has been used in NMR based
experiments to study ligand binding locations.27b
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Our extensive MD calculations were carried out with the CHARMM PARAM-19 force
field,30 which employs an extended atom approximation for all carbon atoms.
Protonation states of all titratable residues were considered at neutral pH. In particular
the side chains of the His residues were assigned a neutral charge, whereas the basic
residues (Arg/Lys) and the acidic residues (Asp/Glu) were assigned either a positive or
a negative charge respectively. We used an implicit solvation protocol called FACTS
(Fast Analytical Continuum Treatment of Solvent), an efficient generalized Born (GB)
implicit solvation model developed in Caflisch’s group,31 which includes a solvent
accessible surface of the solute for the non-polar contribution.
MD Simulations were carried out with periodic boundary conditions at a fixed peptide
concentration of 2.5 mM (87 Å cubic simulation box) using the Langevin integrator at
low friction (coefficient of 0.15 ps-1) and at a temperature of 300 K. Using a time step of
2 fs, for each system, we performed five independent runs which added up to a 5 ꢀs
trajectory. Each of the starting structures contained the peptide in an extended
conformation together with the inhibitor candidate in a different position. We also used
as a reference the simulations of a peptide with the known aggregation inhibitor 9,10-
anthraquinone and the ligand-free peptide. The initial structures were subjected to two
energy minimization runs, which began by a 500 steps steepest descent run followed by
a 50,000 steps conjugate gradient calculation. In each case the gradient tolerance was of
0.001 kcal/(mol·Å2). Then, heating stage and a thermal MD equilibration stages
spanning 0.5 ns each were carried out, followed by the production stage described
above.
10
ACS Paragon Plus Environment

Page 11 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Since every MD simulation takes a month wall clock computer time we had to select the
kind and number of compounds used in our calculations. For instance, for the indole
derivatives we performed simulations only on one aniline and two benzylamine
containing compounds (see below). Our simulations were then used to calculate the
average residency time of the ligand around the peptide and around every residue of the
peptide, as well as the effect of the ligand on the secondary structure of the peptide. The
statistical significance of these results is provided by the standard deviations based on
the individual trajectories.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
RESULTS
1. First lead optimization: Carbazole derivatives.
1.1. BACE-1 ligand screening
The first set of compounds studied were the carbazole derivatives bearing an aniline
moiety with a substituent either in ortho or meta positions, that originally showed
neuroprotective and neurogenerative activity in mice.8 From this study we chose a
subset whose chemical structures and docking results data are displayed in Table S1 in
the supporting info, while the global docking success rates (see methods section for
details) are shown in Figure 5. As seen from Figure 5 and Table S1, the compounds that
have a methoxy substituent at the aniline either in the ortho (compound 1d) and at the
meta position (compound 1g) are the ones that show the widest consensus as a possible
BACE-1 inhibitor. The better fit displayed by these compounds could be the result of
reduced steric clashes, given that they are the smallest candidates in this list. For the
same reason, our predictions indicate that the smaller ligands bind to BACE-1 with any
of the three flap openings (see Table S1). Nevertheless, given their size, it is doubtful
that these ligands could span both the CAS and PAS in AChE, and hence become
multitarget leads
Perusal of the results in Figure 5 and Table S1 shows that the available hits are obtained
almost exclusively when the docking calculations are carried out with a charged ligand.
Actually, only two of the seven compounds (1e, 1g) display some hits when neutral.
This result is in line with our previous studies,32 which indicate that there is an
enrichment in the number of predicted poses close to those observed experimentally
when the ligand is charged.
11
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 12 of 39
1
2
3
4
5
X
1a: X = H, R1 = H, R2 = OMe
1b: X = H, R1 = H, R2 = OBn
1c: X = H, R1 = H, R2 = OCH2CH2Ph
1d: X = Br, R1 = H, R2 = OMe
1e: X = Br, R1 = H, R2 = OBn
1f: X = Br, R1 = H, R2 = OCH2CH2Ph
1g: X = Br, R1 = OMe, R2 = H
6
7
8
9
OH
R2
H
N
N
R1
X
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
9
8
7
6
5
4
3
2
1
0
Neut
Prot
1a
1b
1c
1d
Candidate ID
1e
1f
1g
Figure 5. Global docking success rates for BACE-1 as provided by the number of single hits
across flap openings and docking scoring functions. Neut and Prot indicate weather the ligand is
neutral or protonated.
The pKa of the protonated amino group belonging to the aniline moiety, for these
compounds in solution, is ca. 4.7, a value that is close to the pH at which the binding
assays are performed. Replacing the aniline by a benzylamine moiety would likely
increase the pKa value of the amino group and hence the likelihood of being charged, an
outcome that probably will boost the number of predicted BACE-1 binders. Figure 6
displays the global hit success for benzylamine containing compounds substituted in
ortho (upper panel) and meta (lower panel). The detailed data for these docking
calculations are shown in Tables S2 and S3. Comparison of the number of single hits
amongst the top ten exit poses between analogous aniline (Table S1 and Figure 5) and
benzylamine compounds (Tables S2/S3 and Figure 6), indicates a substantial hit rate
increase for benzylamines above those calculated for the aniline containing compounds,
an outcome that validates our design premises.
12
ACS Paragon Plus Environment

Page 13 of 39
Journal of Chemical Information and Modeling
1
2
3
2a: X = Y = H, R = OMe
2b: X = Y = H, R = OBn
X
4
5
6
7
2c: X = Y = H, R = OCH2CH2Ph
2d: X = Y = Br, R = OMe
2e: X = Y = Br, R = OBn
2f: X = Y = Br, R = OCH2CH2Ph
2g: X = Br, Y = Me, R = OBn
2h: X = Br, Y = Me, R = OCH2CH2Ph
OH
H
N
N
R
Y
8
9
9
8
7
6
5
4
3
2
1
0
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
R
S
R
S
R
S
R
S
R
S
R
S
R
S
R
S
2a
2b
2c
2d
2e
2f
2g
2h
Candidate ID
3a: X = Y = H, R = OMe
3b: X = Y = H, R = OBn
3c: X = Y = H, R = OCH2CH2Ph
3d: X = Y = Br, R = OMe
3e: X = Y = Br, R = OBn
3f: X = Y = Br, R = OCH2CH2Ph
3g: X = Br, Y = Me, R = OBn
3h: X = Br, Y = Me, R = OCH2CH2Ph
X
OH
H
N
N
R
Y
9
8
7
6
5
4
3
2
1
0
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
S
R
S
R
S
3a
3b
3c
3d
3e
3f
3g
3h
Candidate ID
Figure 6. Global hit success that result from the docking to BACE-1 of protonated ortho (upper
panel) and both neutral and protonated meta benzylamines (lower panel). R and S indicate the
two possible enantiomeric configurations of these compounds.
13
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 14 of 39
1
2
3
4
5
6
7
8
9
Perusal of Figure 6 indicates that the benzylamine substitution pattern has also a bearing
in the number of possible hits. The results would seem to indicate that the ortho is
favored over the meta substitution at the benzylamine moiety. For instance amongst the
top ten poses, the number single hits with ortho substitution is twice the number of hits
with the meta substitution (see Tables S2 and S3). Nevertheless, there seems to be
exceptions (e.g., see compounds 3e when the hydroxyl group is in the R configuration),
which had a number of hits comparable to the analogous ortho derivative (i.e.,
compound 2e when the hydroxyl group is in the S configuration).
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
For the compounds bearing a benzylamine fragment we also analyzed the effectiveness
of each stereoisomer. The results shown in Figure 6 do not show a clear predilection for
a given enantiomer, as changes at the end points of the candidate compounds seem to
change the preference. For instance in the case of the compounds containing meta-
benzylamines, those compounds that bear two Br atoms on the carbazole favor a R
configuration (compounds 3e and 3f), while those that replace both Br atoms by H
atoms favor a S configuration. Nevertheless, these configuration patterns do not seem to
hold for the ortho-benzylamine bearing compounds.
1.2. AChE ligand screening
Figure 7 displays the global hit rate for protonated anilines and benzylamines
(substituted in the ortho position), while Figure 8 depicts the number of hits for the
compounds bearing benzylamines substituted in meta. As seen from these figures, there
is a wide consensus amongst the scoring function results, indicating that these
compounds are good candidates for AChE inhibitors, in most cases independently of the
stereochemistry of the hydroxyl group and the charge of the hydroxylethylamino group.
Comparison with the BACE-1 docking results described above indicates that the
number of predicted hits for AChE is substantially larger than those for BACE-1,
implying that there are fewer hurdles for finding a candidate that will fulfill the
pharmacophore requirements for the former enzyme.
14
ACS Paragon Plus Environment

Page 15 of 39
Journal of Chemical Information and Modeling
1
2
3
X
1a: X = H, n = 0, R = OMe
1b: X = H, n = 0, R = OBn
1c: X = H, n = 0, R = OCH2CH2Ph
1d: X = Br, n = 0, R = OMe
1e: X = Br, n = 0, R = OBn
2a: X = H, n = 1, R = OMe
4
5
6
7
2b: X = H, n = 1, R = OBn
2c: X = H, n = 1, R = OCH2CH2Ph
2d: X = Br, n = 1, R = OMe
2e: X = Br, n = 1, R = OBn
OH
H
N
N
n
R
1f: X = Br, n = 0, R = OCH2CH2Ph 2f: X = Br, n = 1, R = OCH2CH2Ph
X
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3
2
1
0
1a
1b
1c
1d
1e
1f
2a
2b
2c
2d
2e
2f
Candidate ID
Figure 7. Global hit rate resulting from the AChE docking screening of the protonated ortho
substituted anilines (left) and benzylamines (right). The hit search was carried amongst the top
ten poses resulting from three docking calculations each carried out with a different scoring
function (ChemScore, GoldScore and ChemPLP).
15
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 16 of 39
1
2
3
X
3b: X = Y = H, R = OBn
3c: X = Y = H, R = OCH2CH2Ph
3e: X = Y = Br, R = OBn
3f: X = Y = Br, R = OCH2CH2Ph
3g: X = Br, Y = Me, R = OBn
3h: X = Br, Y = Me, R = OCH2CH2Ph
4
5
6
7
OH
H
N
N
R
Y
8
9
3
2
1
0
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
S
3b
3c
3e
3f
3g
3h
Candidate ID
Figure 8. Global hit rate resulting from the AChE docking screening of the compounds with a
meta substituted benzylamine. For each molecule we studied the effect of the stereochemistry of
the hydroxyl group (R, S) and the protonation state of the amine group.The hit search was
carried amongst the top ten poses resulting from three docking calculations each carried out
with a different scoring function (ChemScore, GoldScore and ChemPLP).
The scoring function values for aniline and benzylamine containing compounds are
listed in Tables S4 and S5. Perusal of Table S4 and Figure 7 indicates that the shorter
compounds, like 1a, 1d, 2a and 2d (those bearing a methoxy substituent) have no hit
according to our pharmacophore definition, as shown in Figure 1. Scrutiny of the
binding poses for a methoxy as compared to a benzyloxy substituted compound (Figure
9) indicates that the former ligands derive their worse performance from their inability
to bind both to the CAS and PAS in AChE. As seen from Figure 9 the carbazole moiety
interacts in the CAS with residues Tyr 84 and Phe 330 through π stacking interactions.
The shorter ligand is not able to reach the PAS in AChE and hence to generate these
types of interactions with Tyr 70 and Trp 279
.
16
ACS Paragon Plus Environment

Page 17 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 9. Docking of compounds 2d (upper panel) and 3b (lower panel) into AChE.
17
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 18 of 39
1
2
3
4
1.3. Inhibitor Aβ12-28 peptide interaction results.
Table 1 lists the percentage of residence times for 9,10-anthraquinone (a reference
compound) and for some carbazole containing compounds, up to 7.5 Å from the
peptide. We have also calculated the residence time of these ligands around every
residue of the peptide with a cut-off distance of 4.5 Å (see Figure 10) and the residue-
residue interaction in the absence and presence of the ligands (see Figures 11 and S4).
As seen from Table 1, the candidate compounds have a much larger residence time than
the reference ligand. Moreover, the selected compounds display a bigger preference
(than the reference compound) for the binding hot spots centered on both the His cluster
at the N terminal end and on the aromatic cluster found at the LVFFA segment.
5
6
7
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1. Percentage of residence time for contacts peptide-ligand up to 7.5 Å.
Candidate
1a
Contact Time (%)
85 2.6
93 0.6
1b
1c
1d
1f
90 2.0
99 0.4
99 0.2
36 1.5
anthraquinone
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Anthraquinone
1a
1b
1c
1d
1f
V H H Q K
L V F F A E D V G S N K
Figure 10. Ligand-peptide residues fraction contact times. Cut-off distance is set at 4.5 Å.
18
ACS Paragon Plus Environment

Page 19 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
Ligand-free peptide
H H Q K L V F F A E D V G S N K
1a
A
V
V
H
H
Q
K
L
V
F
F
E D V G S N K
V
H
H
Q
K
L
V
H
H
Q
K
L
6
V
F
7
V
F
F
8
9
F
A
E
D
V
G
S
N
K
A
E
D
V
G
S
N
K
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1b
A
1c
A
V
H
H
Q
K
L
V
F
F
E
D
V
G
S
N
K
V
H
H
Q
K
L
V
F
F
E
D
V
G
S
N
K
V
H
H
Q
K
L
V
H
H
Q
K
L
V
F
V
F
F
F
A
E
D
V
G
S
N
K
A
E
D
V
G
S
N
K
1d
A
1f
A
V
H
H
Q
K
L
V
F
F
E D V G S N K
V
H
H
Q
K
L
V
F
F
E D V G S N K
V
V
H
H
Q
K
L
H
H
Q
K
L
V
F
V
F
F
F
A
E
D
V
G
S
N
K
A
E
D
V
G
S
N
K
0,00 0,05 0,10 0,15 0,20 0,25 0,30 0,35 0,40 0,45 0,50 0,55 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
Figure 11. Residue-residue contact map in the absence (top left panel) and in the presence of
inhibitors. Cut-off interaction distance is set at 4.5 Å. The color scale shown above indicates the
fraction residence times.
An important issue is the effect of the ligand on the secondary structure of the peptide.
The Aβ peptide, which originates from APP hydrolysis, originally adopts a helix
structure. As amyloid aggregates are formed, there is a change of conformation that
leads to a beta hairpin, specifically around the DVGS motif. For instance the NMR
structure of a pentamer29 shows clearly that the peptides organize themselves forming
hairpins that aggregate as sheets. As seen from Figures 11 an S4, the ligands would
seem to partially preclude the formation of this secondary structure motif (turn) in a
monomer, a process that may lead to a reduced aggregation. Hence, it may be surmised
that the residue-residue contact map in the presence of a ligand may give us a measure
of the inhibitory aggregation effect of our candidates.
19
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 20 of 39
1
2
3
4
5
6
7
8
9
1.4. Synthesis and experimental assays for carbazole containing compounds.
The synthesis of several carbazole-containing compounds was carried out following
Scheme 1, which includes the list of synthesized molecules. Carbazole-epoxide 10 was
prepared as described in the literature from carbazole (8) and epichlorohydrin 9.8,33
Regioselective ring-opening of epoxide 10 with different ortho-substituted anilines 11
in the presence of Mg(ClO4)2 in acetonitrile,34 afforded carbazole compounds 1 in
moderate to good yields. Similarly, as shown in Scheme 1, carbazole analogues 3 were
synthesized by ring opening of carbazole-epoxide 10 with meta-substituted
benzylamines 12. The choice of compounds was guided in many cases by the
availability of the reactants. Although in the case of benzylamines the compounds with
an ortho substituent seem to show a wider consensus as BACE-1 inhibitors across all
scoring functions (see Figure 6), they require a more elaborated and expensive
chemistry. For this reason, we chose the best scoring candidates with a meta substituted
benzylamine (3e and 3f).
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 1. Synthetic route and carbazole derivatives synthesized.
The results for the experimental binding assays in BACE-1, AChE and BuChE, as well
as the ThT Aβ aggregation inhibition are shown in Table 2. As seen from this table,
20
ACS Paragon Plus Environment

Page 21 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
some of the aniline-containing compounds showed some promise as multitarget
inhibitors. For instance, compounds 1e and 1f display ꢀM affinity for AChE and Aβ
inhibition percentage of ca. 50% at 100 ꢀM peptide concentration, values that are
superior to 9,10-anthraquinone (30% at 100 ꢀM).35 On the other side compound 1a,
displays ꢀM binding for BuChE, and 58% percentage of inhibition of the fibril
formation. Nevertheless, none of the carbazole and aniline containing compounds,
which were originally shown to be neuroprotective and neurogenerative in mice,8
resulted in multitarget leads across all the chosen amyloid cascade targets.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 2. Experimental results for the multitarget assays.
BACE-1
EeAChE
hsBuChE
Aβ(1-40)
%Inh.@
100µM
%Inh.@
10µM
%Inh.@
10µM
--
%Inh.
Compound
IC50(µM)
IC50(µM)
IC50(µM)
@100µM
1.3 0.3
2.4 1.4
1.5 0.3
1.6 1.5
0.5 0.3
0.9 0.9
--
1a
1b
1c
1d
1e
1f
30
22
49
15
--
--
48
--
1
--
6.0 1.0
--
58
51
36
41
46
49
11
28
2
5
5
3
3
1
3
3
--
1
2
1
--
--
--
34
33
12
20
10
1
1
1
1
1
--
--
--
--
--
--
--
--
--
7.2 0.4
7.8 0.2
--
3e
3f
1
--
--
1.1 0.2
7.1 0.7
3.1 0.4
3.1 0.3
14
1
--
Finally, as seen from this table the benzylamine amine bearing compounds (3e and 3f)
have improved their affinity for BACE-1 by three orders of magnitude over the aniline
containing compounds. This result validates and supports the outcome of our
calculations (see Figures 5 and 6), which indicate that the addition of a CH2 fragment
provides hits across the set of scoring functions used in our calculations. As mentioned
before this effect is probably due to the raise in the pKa in the benzylamine’s amino
group, which favors the formation of an ion pair with one of the Asp residues of the
active site Asp dyad. These latter compounds (3e and 3f) have improved their affinity
for BuChE and 3f displays as well a micromolar affinity for AChE. Nevertheless, their
Abeta aggregation inhibition has dropped below the 30% inhibition displayed by the
reference compound (9,10-anthraquinone).
2. Screening a third generation candidates: Indole based multitarget candidates.
The carbazole moiety present in the candidates synthesized and assayed above is quite a
bulky group. Perusal of the resulting binding poses showed some steric clashes between
the ligand and some protein side chains. The close van der Waals contacts are allowed
by the GOLD docking protocol as a way of compensating for the lack of protein
21
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 22 of 39
1
2
3
4
5
6
7
8
9
flexibility.15 Figures S1 and S2 display the steric clashes of some ligands with BACE-1
and AChE. An option to avoid the steric clashes altogether will be to search for smaller
aromatic groups in place of the carbazole, like an indole. Based on this idea we
designed a new generation of multitarget candidates bearing this fragment in one end
and in the other end the same aromatic moieties which were already used in the
previous sections (substituted anilines and benzylamines). For these ligands we have
chosen the OBn substituent for the anilines and benzylamines, since the analysis of the
results for the carbazole-containing compounds described above has shown that this
substituent provides the leads with the right length to span the distance from the CAS to
the PAS in AChE.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2.1. BACE-1 ligand screening
Figure 12 lists the global hit success for the indole-based derivatives, while the scoring
function values for this set of compounds are listed in Tables S6 and S7. The structural
variables analyzed in these figures are the same as in the study of the carbazole
derivatives in the previous section and include the moiety to which the amino group
belongs (aniline or benzylamine), its protonation state, the substitution of the indole
group, the substituent position (ortho or meta) of the other aromatic ring and the
stereochemistry of the OH group.
Perusal of our results indicates that the indole-based ligands have a higher probability of
binding to BACE-1 with a medium open or fully open flap (see Tables S6 and S7). The
number of hits with a closed flap is very small, especially when using a ChemPLP
scoring function, indicating that these molecules fit in the active site of BACE-1 thanks
to the plasticity of its flap.
In the case of the carbazole-based compounds, the best ranking compounds where those
that contained a benzylamine moiety, one that assured the existence of a protonated
amino group. A global comparison of the compounds that contain an aniline (see Figure
12, upper panel) with those that contain a benzylamine (Figure 12, lower panel) indicate
that the latter present a larger global hit rate, a pattern similar to that observed for the
carbazole containing compounds. Nevertheless, the preferred benzylamine substitution
pattern in the indole-based compounds (predicted by our calculations) differs from the
one predicted for the carbazole containing ligands, since these latter compounds favor
an ortho rather than meta substitution pattern for benzylamines. On the other hand
perusal of Figure 12 suggests that the indole-based compounds favors a meta over the
22
ACS Paragon Plus Environment

Page 23 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
4a: R1 = H, R2 = OBn, X = H
4b: R1 = H, R2 = OBn, X =Br
4c: R1 = H, R2 = OBn, X = Me
5a: R1 = OBn, R2 = H, X = H
5b: R1 = OBn, R2 = H, X =Br
5c: R1 = OBn, R2 = H, X = Me
X
OH
R2
H
N
N
R1
Br
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
9
8
7
6
5
4
3
2
1
0
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
S
4a
4b
4c
5a
5b
5c
Candidate ID
6a: R1 = H, R2 = OBn, X = H
6b: R1 = H, R2 = OBn, X =Br
6c: R1 = H, R2 = OBn, X = Me
7a: R1 = OBn, R2 = H, X = H
7b: R1 = OBn, R2 = H, X =Br
7c: R1 = OBn, R2 = H, X = Me
X
OH
H
N
N
R1
Br
R2
9
8
7
6
5
4
3
2
1
0
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
S
6a
6b
6c
7a
7b
7c
Candidate ID
Figure 12. Global hit success that result from the docking to BACE-1 of indole derivatives,
anilines (upper panel) and benzylamines (lower panel).
23
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 24 of 39
1
2
3
4
5
6
7
ortho substitution, a substitution pattern also followed by the aniline bearing
compounds. Moreover, for these latter compounds, the largest increase is observed for
the hits containing a neutral ligand.
We have also studied the effect of the ligand protonation state on the number of hits.
For the aniline as well as the benzylamine containing compounds (Figure 12), the
number of hits increases (for the most part) when the amino group is protonated, rather
than neutral.
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
2.2 AChE ligand screening.
The global hit rate resulting from the AChE docking based screening calculations for
the indole derivatives that contain a substituted aniline or benzylamine in ortho or meta
positions are depicted in Figure 13. Comparison of the effect of the substitution pattern
in both aniline and benzylamine (shown in this figure) indicate that changing the
substitution from ortho to meta increases the number of hits. As seen from Figure 13,
the number of hits obtained and the consensus reached across the three scoring
functions indicate that our indolebenzylamine-containing compounds fit well inside
AChE spanning both the main site (CAS) and the peripheral one (PAS). Moreover,
comparison of the aniline bearing compounds with the benzylamine ones indicate that
the latter present a larger number of hits across all scoring functions, probably due to an
increase in size of the ligands that allow them to span better both binding sites.
One interesting issue is the docking exit poses. As we mentioned above in our
pharmacophore depiction for AChE, we seek to have dual inhibitors in which one of the
end aromatic fragments is to be found at the CAS and the other end aromatic group
positions itself in the PAS, both producing π stacking interactions with the aromatic
residue clusters residing in both sites (see Figure 1). There are two possible pose
orientations that fulfill this hit. In the first one the indole moiety is predicted to be found
in the CAS while the terminal benzyloxy group is to be found at the PAS. In the second
possible pose, we have the inverse option in which the indole fragment resides at the
PAS. We have listed in Figure 13 both poses and have referred to them as ‘CAS’ and
‘PAS’ respectively. As seen from these plots the number of ‘CAS’ hits (indole docked
into CAS) is larger than the one that places this fragment in the PAS (both for aniline
and benzylamine containing compounds), a result that we will try to verify through X-
ray crystallography. In Figure 14 we depict the most favored pose obtained for
compound 7a with the Chemscore fitness function. Notice that the indole fits neatly in
24
ACS Paragon Plus Environment

Page 25 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
the CAS, making π stacking interactions with residues W84 and F330, while the Bn
group interacts with W279 and Y70 in the PAS.
6
7
8
9
Finally, our results (shown in Figure 13) indicate that there is scarcely a preference for
one of the enantiomers, neither there is a clear proclivity for a neutral or charged state
for the amino group.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
4a: R1 = H, R2 = OBn, X = H
4b: R1 = H, R2 = OBn, X =Br
4c: R1 = H, R2 = OBn, X = Me
5a: R1 = OBn, R2 = H, X = H
5b: R1 = OBn, R2 = H, X =Br
5c: R1 = OBn, R2 = H, X = Me
X
OH
R2
H
N
N
R1
Br
3
2
1
0
Neut
Prot
R
S
R
PAS
S
R
S
R
S
R
S
R
PAS
S
CAS
CAS
PAS
CAS
4a
4b
Candidate ID
4c
3
2
1
0
Neut
Prot
R
S
R
PAS
S
R
S
R
S
R
S
R
PAS
S
CAS
CAS
PAS
CAS
5a
5b
Candidate ID
5c
Figure 13a. Docking results of the indole bearing compounds with ortho (upper panel) and
meta (lower panel) substituted anilines on AChE.
25
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 26 of 39
1
2
3
4
5
6
7
8
9
6a: R1 = H, R2 = OBn, X = H
6b: R1 = H, R2 = OBn, X =Br
6c: R1 = H, R2 = OBn, X = Me
7a: R1 = OBn, R2 = H, X = H
7b: R1 = OBn, R2 = H, X =Br
7c: R1 = OBn, R2 = H, X = Me
X
OH
H
N
N
R1
Br
R2
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
3
2
1
0
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
PAS
S
CAS
PAS
CAS
PAS
CAS
6a
6b
6c
Candidate ID
3
2
1
0
Neut
Prot
R
S
R
S
R
S
R
S
R
S
R
PAS
S
CAS
PAS
CAS
PAS
CAS
7a
7b
Candidate ID
7c
Figure 13b. Docking results for indole bearing compounds with ortho (upper panel) and meta
(lower panel) substituted benzylamines on AChE.
26
ACS Paragon Plus Environment

Page 27 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 14. Pose for ligand 7a in which the indole fragment interacts (through π stacking) with
residues Trp 84 and Phe 330 in the CAS, while the treminal Bn group interacts with Tyr 70 and
Trp 279 at the PAS site
.
2.3. Inhibitor Aβ12-28 peptide interaction results.
We have performed the MD simulations on three of our indole candidates: one
containing an aniline fragment (4b) and the other two containing a benzylamine moiety
(7b and 7c). The results indicate that percentage residency times around the peptide
with 7.5 Å radii for these compounds are relatively high (88 2.0, 96 2.0 and 86 1.3
respectively).
Figure 15 displays the ligands residency time around every residue of Aβ12-28. As seen
from this figure, the ligands favor two binding spots (populated by aromatic residues), a
result already observed for the carbazole containing compounds. The most frequented
site is around the LVFFA segment, and the second one contains the two His residues at
the start of the sequence. Another important feature is that these ligands present much
higher residence times than the reference ligand, the 9,10-anthraquinone, a compound
that has a 30% fibril formation inhibition at 100 ꢀM concentration. Furthermore,
comparison of the intra-peptide residue interaction when these two ligands are present
(see Figures 16 and S4) with the residue contact when they are absent (see Figures 11
and S4) indicates that these compounds clearly interrupt the formation of the hairpin
structures that are the hallmark of Aβ peptide aggregates, indicating that these ligands
may inhibit amyloid aggregation. Finally, we predict that the benzylamine-containing
27
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 28 of 39
1
2
3
4
5
compounds (7b and 7c) reduce the formation of turns (around the DVSG segment) more
efficiently than the aniline based compound (4b).
6
7
8
9
1.0
0.9
0.8
0.7
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
0.6
Anthraquinone
4b
7b
7c
0.5
0.4
0.3
0.2
0.1
0.0
V
H
H Q K
L
V
F
F A E D V G S N K
Figure 15. Ligand-peptide residues fraction contact times. Cut-off distance is set at 4.5 Å.
4b
V
H H Q K L V F F A E D V G S N K
V
H
H
Q
K
L
V
F
F
A
E
D
V
G
S
N
K
7b
H H Q K L V F F A E D V G S N K
7c
A E D
V
H
H
Q
K
L
V
F
F
V G S N K
V
V
H
H
Q
K
L
V
H
H
Q
K
L
V
F
V
F
F
F
A
E
D
V
G
S
N
K
A
E
D
V
G
S
N
K
0,00 0,05 0,10 0,15 0,20 0,25 0,30 0,35 0,40 0,45 0,50 0,55 0,60 0,65 0,70 0,75 0,80 0,85 0,90 0,95 1,00
Figure 16. Residue-residue interaction contact map in the presence of candidates 4b, 7b and
7c. Cut-off interaction distance is set at 4.5 Å. The color scale shown above indicates the
residence times.
28
ACS Paragon Plus Environment

Page 29 of 39
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
2.4. Synthesis and experimental assays for indole containing compounds.
The indole derivatives whose structures contain either an aniline or a benzylamine
fragment were obtained by the same synthetic route as the carbazole containing
compounds indicated in Scheme 2, which includes the list of synthesized molecules.
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
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 2. Synthetic route and indole derivatives synthesized.
The results of the binding assays to BACE-1, AChE and BuChE for the selected
compounds, together with the fibril inhibition assay are listed in Table 3. As seen from
this table the meta substituted benzylamine compounds display altogether higher
activity than the ortho substituted aniline ones for AChE, a result that validates the
predictive power of our in silico protocol presented above. More importantly, the
replacement of a carbazole by an indole moiety with a benzylamine substituted in meta
has lead to candidates that bind to all chosen amyloid cascade targets, (BACE-1,
amyloid aggregates and the peripheral site in AChE), as well as to AChE’s main site
and BuChE, implying also a positive cholinergic effect in AD. Hence, compounds like
7c are truly multitarget compounds.
Finally, the results displayed in Tables 2 and 3 also afford us with important structural
information for the binding poses of both carbazole and indole bearing compounds to
the cholinergic targets. For instance compounds 7a, 7b and 7c bind even better to
29
ACS Paragon Plus Environment