Discovery of novel GSK-3β inhibitors with potent in vitro and in Vivo activities and excellent brain permeability using combined ligand- and structure-based virtual screening

Article abstract of DOI:10.1021/jm100941j

Dysregulation of glycogen synthase kinase (GSK-3β) is implicated in the pathophysiology of many diseases, including type-2 diabetes, stroke, Alzheimers, and others. A multistage virtual screening strategy designed so as to overcome known caveats arising from the considerable flexibility of GSK-3β yielded, from among compounds in our in-house database and two commercial databases, new GSK-3β inhibitors with novel scaffold structures. The two most potent and selective validated hits, a 2-anilino-5-phenyl-1,3,4- oxadiazole (24) and a phenylmethylene hydantoin (28), both exhibited nanomolar affinity and selectivity over CDK2 and were potent enough for direct in vivo validation. Both were able to cause significant increases in liver glycogen accumulation in dose-dependent fashion. One also exhibited excellent blood-brain barrier permeability, the other adequate for a lead compound. Analogues of the oxadiazole 24 were synthesized to experimentally corroborate or rule out ligand-bound structures arising from docking studies. SAR results supported one docking study among a number of alternatives.

Full text of DOI:10.1021/jm100941j

8534 J. Med. Chem. 2010, 53, 8534–8545
DOI: 10.1021/jm100941j
Discovery of Novel GSK-3β Inhibitors with Potent in Vitro and in Vivo Activities and
Excellent Brain Permeability Using Combined Ligand- and Structure-Based Virtual Screening
Mohammad A. Khanfar, Ronald A. Hill, Amal Kaddoumi, and Khalid A. El Sayed*
Department of Basic Pharmaceutical Sciences, College of Pharmacy, University of Louisiana at Monroe,
Monroe, Louisiana 71201, United States
Received June 18, 2010
Dysregulation of glycogen synthase kinase (GSK-3β) is implicated in the pathophysiology of many
diseases, including type-2 diabetes, stroke, Alzheimer’s, and others. A multistage virtual screening
strategy designed so as to overcome known caveats arising from the considerable flexibility of GSK-3β
yielded, from among compounds in our in-house database and two commercial databases, new GSK-3β
inhibitors with novel scaffold structures. The two most potent and selective validated hits, a 2-anilino-5-
phenyl-1,3,4-oxadiazole (24) and a phenylmethylene hydantoin (28), both exhibited nanomolar affinity
and selectivity over CDK2 and were potent enough for direct in vivo validation. Both were able to cause
significant increases in liver glycogen accumulation in dose-dependent fashion. One also exhibited
excellent blood-brain barrier permeability, the other adequate for a lead compound. Analogues of the
oxadiazole 24 were synthesized to experimentally corroborate or rule out ligand-bound structures
arising from docking studies. SAR results supported one docking study among a number of alternatives.
Introduction
hyperphosphorylation of tau (not direct aggregation into
filaments under normal cellular status) is now considered to be
requisite to neurofibrillary tangle formation, and compromised
microtubule stability.19,20 Moreover, this action of GSK-3β has
been reported to be a key effector, possibly the key effector, of
the toxic consequences of excessive accumulation of β-amyloid,
a protein that aggregates as extracellular amyloid plaques in the
brains of AD patients.21 Exposure of cortical and hippocampal
primary neuronal cultures to β-amyloid-derived oligomers
induces activation of GSK-3β, tau hyperphosphorylation,
and cell death.22 Consequently, inhibition of GSK-3β renor-
malizes, in a dose-dependent manner, tau phosphorylation in
cells overexpressing tau protein.23 Suppression of GSK-3β
expression by antisense oligonucleotides, or inhibition of its
activity with lithium, averts β-amyloid-induced neurodegenera-
tion of cortical and hippocampal primary cultures.24,26
In a previous article, we reported the establishment and
validation of a pharmacophore model based on marine-derived
GSK-3β inhibitors.11 This model has the characteristic features
required for an ideal pharmacophoric query because it incorpo-
rates the important interactions known to be required for GSK-
3βinhibitors,27 works in consistent fashion with published GSK-
3βpharmacophore models28-30 and performed quite adequately
on our in-house compound database. This query enabled the
development of an efficient protocol, reported herein, for dis-
covering new GSK-3β inhibitor scaffolds based on a combina-
tion of ligand-based and structure-based (target-based) design.
The latter (SBDD) component had already been shown to be
absolutely requisite for surmounting the caveats resulting from
the known induced-fit flexibility of GSK-3β.11,27-30
Glycogen synthase kinases 3 (GSK-3R, GSK-3β, and GSK-
3β2)a play vital roles in regulating cell division, stem-cell renewal
and differentiation, apoptosis, circadian rhythm, transcription,
and insulin action.1,2 GSK-3β is a 47 kDa nonreceptor serine/
threonine kinase, which is a critical modulator of cell fate,
neuronal plasticity, and tumorigenesis.1-3 It was first discovered
by virtue of its ability to phosphorylate and inactivate glycogen
synthase, the regulatory enzyme of mammalian glycogen
synthesis.3 Its pleiotropic but unique activities have caused
GSK-3β to be prosecuted as a theoretically promising pharma-
cotherapeutic target for the treatment of several human diseases,
including type-2 diabetes,4 CNS disorders including manic
depressive disorder and Alzheimer’s disease (AD)5 as well as
other neurodegenerative diseases6 and chronic inflammatory
disorders.7 Lithium was the first GSK-3β inhibitor to be used
for therapeutic effect and has been beneficially used in tens of
thousands of patients for many years.1-3 A substantial number
of structurally diverse compounds have been reported to inhibit
GSK-3β in recent years.8-17
One well-known protein substrate of GSK-3β is the micro-
tubule-associated protein, tau. GSK-3β normally phosphor-
ylates tau at specific sites, namely serine 199 and serine 396
(human numbering sequence).18 GSK-3β seems to be a key
mediator in the pathophysiology of AD, in that abnormal
*To whom correspondence should be addressed. Phone: þ1 318 342
1725. Fax: þ1 318 342 1737. E-mail: elsayed@ulm.edu.
a Abbreviation: AD, Alzheimer’s disease; BBB, blood-brain barrier;
CDK2, cyclin-dependent kinase 2; EDCI, N-(3-dimethylaminopropyl)-
N-ethylcarbodiimide; GSK-3β, glycogen synthase kinase-3β; HBA,
hydrogen bond acceptor; HBD, hydrogen bond donor; NCI, National
Cancer Institute; SAR, structure-activity relationship; SBDD, structure-
based drug design; SD, standard deviation; SEM, standard error of
mean; VS, virtual screening;
Several structure databases (our in-house database and also
the NCI and Maybridge databases) were drawn on for iden-
tifying novel GSK-3β inhibitor scaffolds. The database com-
pounds were subjected to filtering by an extended Lipinski’s
r
pubs.acs.org/jmc
Published on Web 11/17/2010
2010 American Chemical Society

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8535
Figure 1. Pharmacophoric features of GSK-3β inhibitors generated by the DISCOtech module and used in the 3D pharmacophoric search.
AA-hydrogen bond acceptor atom, DA-hydrogen bond donor atom, HD-hydrophobic center, DS-hydrogen bond donor site, AS-hydrogen
bond acceptor site.
rule of five. Pharmacophore-based searching was subse-
quently carried out, from which initial hits were identified.
Docking studies were then undertaken, beginning from multi-
ple disparate X-ray crystallographic structures, with highly
restrictive acceptance criteria.31 Finally, many of the resulting
hits were acquired and tested in vitro as inhibitors of recom-
binant human GSK-3β. Two compounds that exhibited both
high potency, and selectivity toward GSK-3β over CDK2
were further evaluated for their in vivo (mice) abilities to
enhance hepatic glycogen reserves and to cross the blood-
brain barrier (BBB). A 2-anilino-5-phenyl-1,3,4-oxadiazole
analogue showed high in vitro and in vivo activities and
excellent BBB permeability. To aid in corroborating or else
disallowing alternative in silico structures, and for selecting
the best hypothesis-generating models for moving forward in
lead optimization, 33 new analogues were synthesized using
the recently reported, highly efficient one-pot synthesis of
substituted 2-amino-1,3,4-oxadiazoles.32 Structural modifica-
tions were mainly focused on the para position of the phenyl
rings, and substituent selections were based on varying their
lipophilic (π) and electronic (σ) characteristics in the manner
prescribed by Craig’s plot.33
refinement steps, a pharmacophore model was obtained that
includes the characteristic features known to be requisite and
ideal for ATP-competitive GSK-3β inhibition.27 Using a dis-
tance tolerance of 1 A, five pharmacophoric features were
identified, one H-bond donor (HBD), two H-bond acceptors
(HBA), and two hydrophobic sites (Figure 1), consistent with
previously reported GSK-3β pharmacophores.28-30 In applica-
tion, the model performed satisfactorily on the in-house data-
base of active GSK-3βinhibitors and decoys. This model picked
89 active GSK-3β inhibitors and three decoys from among the
200 molecules (100 decoys).
Virtual screening (VS) has for several decades been a
fruitful strategy for lead identification. Its use has increased
substantially in recent years, especially with the advent of
online databases, enabling the discovery of new hits out of
millions of virtual structures. VS offers a means to system-
atically select a small number of compounds, which can then
be tested in lower-throughput assays (in-silico (docking, e.g.)
or physically), increasing the likelihood of obtaining com-
pounds with desired activities as compared to the more
expensive and time-consuming physical high-throughput
screening. For the work herein, we employed ligand-based
VS using the validated pharmacophoric map, discussed
above, followed by structure-based VS (docking), working
from three disparate GSK-3β crystallographic structures.
As previously mentioned, this two-stage VS protocol
(Figure 2) was designed with the intent of overcoming
the limitations of each separate method and so as to
surmount the caveats known to arise because of induced-
fit flexibility with ligand binding to the ATP-binding
pocket of GSK-3β.38,39
Results and Discussion
Virtual Screening. A 3D pharmacophore model built in a
previous report was applied in the ligand-based virtual screen-
ing.11 Mapping of this model was based on three marine-derived
phenylmethylene hydantoin GSK-3β inhibitors11 using the
DISCOtech module implemented in SYBYL 8.0. Given a set
of molecules that are related by their ability to bind to the same
protein receptor, DISCOtech identifies features that could
be elements in a pharmacophore, employing clique detection
to generate pharmacophore hypotheses from up to 300 con-
formers per molecule.34-37 After several optimization and
Virtual structures in the in-house (307), National Cancer
Institute (NCI, 234,055), and Maybridge (55,541) databases
were screened to identify new GSK-3β inhibitors. Our in-house

8536 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24
Khanfar et al.
ligands via impeded automation algorithms.42 To maximize the
enrichment and minimize the number of false-positive hits,
strict criteria were applied to select the refined hits for physical
evaluation in a GSK-3β kinase assay: (i) a ligand had to
maintain H-bonding interactions with at least three out of four
important amino acids (Val 135, Asp 133, Arg 141, and Gln
185) detected in diverse ligand-GSK-3β complexes27 in at least
one of the docking studies using the three above-mentioned
crystallographic structures; (ii) a ligand was required to exhibit
a binding orientation similar to one or more of the reported
cocrystallized inhibitors, to wit, so that the primary heterocyclic
ring (HD1 feature in the pharmacophore model, Figure 1)
bridged via at least two H-bonds to the hinge region (Asp 133,
Tyr 134, and Val 135) of GSK-3β; (iii) a ligand was required to
have a binding score equal to or higher than the cocrystallized
GSK-3β inhibitors (I-5 in 1Q4L,27 AR-A014418 in 1Q5K,40
and 6-bromoindirubin-30-oxime in 1UV541) in each docking
study. With respect to this last criterion, the choice of scoring
function is crucial. Even though scoring functions remain vastly
deficient with respect to accuracy or even rank-ordering suc-
cess, heuristic considerations and aggregate experience suggest
some that are likely to be more appropriate, versus those almost
certain to misinform, in a particular case. Here, because polar
interactions are dominant within the GSK-3β active site, scores
that consider these interactions are expected to give the most
meaningful results. Accordingly, poses generated by Surflex-
Dock were best ranked by Gold and Surflex-Dock-total
scores.39 Gold score is a GOLD-like function that focuses on
H-bonding interactions, and is known to perform well in
situations where significant polar interactions occur.44,45 Com-
plementarily, Surflex-Dock’s total score has been shown to
perform well when the ligand makes relatively many discrete
intermolecular interactions, which fits this case because the
various cocrystallized GSK-3β inhibitors exhibit at least
four.42 A total of 53 hits satisfied the aforementioned
conditions, 35 from NCI, 14 from Maybridge, and 9 from
our in-house databases. Only 25 from the two external
databases were available. Therefore, only 34 hits were physi-
cally tested for their ability to inhibit GSK-3βactivity (Figure 3,
Table 1).
GSK-3β Kinase Assay and Molecular Modeling Studies.
The 34 refined hits were tested against recombinant GSK-3β
using the Invitrogen Z0-LYTE Kinase Assay kit (Table 1).46 All
compounds were first tested at 10 μM, and hits with percentage
inhibition g50% were further tested to obtain IC50 values. A
reference compound, the marine β-carboline alkaloid manza-
mine A, gave an IC50 value of 8.4 μM, comparable to literature
reports.47 A total of 14 compounds (41%) were confirmed to be
active inhibitors, of which eight were found to have IC50 values
<1 μM. This significant success (24% of experimentally eval-
uated compounds were true actives with IC50 values <1 μM)
proved the validity and strength of our VS protocol for discover-
ing potent GSK-3β inhibitors. Strikingly, even though the initial
pharmacophore model was developed based strictly on a few
compounds having a hydantoin scaffold, diverse structural
skeletons were picked in the pharmacophoric screen, about
10% of these successfully met the strict criteria from the second
(docking) stage, and in turn nearly a fourth of these were found
to be submicromolar GSK-3β inhibitors. The three most active
hits (4, 17, and 24) are 9H-purine, pyrido[2,3-d]pyrimidine and
2-anilino-1,3,4-oxadiazole derivatives, respectively. Other di-
verse scaffolds, namely 2-thioxothiazolidin-4-one (16 and 19),
2,3-dihydro-[1,2,4]triazolo[4,3-a]pyrimidine (21), 3-hydroxypyr-
idin-2(1H)-one (23), and a 3,4-dihydro-2H-1,3-benzoxazine (30)
Figure 2. Outline of VS methodology. (A) Compounds down-
loaded from NCI, Maybridge, and in-house databases were
screened using pharmacophore-based VS, filtering also for drug-
likeness (see Experimental Section). (B) Molecular docking and
ranking by Gold and Surflex-Dock total scores, with selection
requiring docking poses that maintained multiple important
H-bonding interactions. (C) Available compounds were purchased
for testing along with in-house samples. (D) Experimental hit
validation was accomplished using a commercial GSK-3β kinase
assay kit.
database is composed of compounds of a wide variety of struc-
tural classes, mainly of natural products or their semisynthetic
or microbial transformation products thereof, which were read-
ily available for testing. Lipinski’s Rule of five31 was opted as the
filtration criterion, limiting the range for molecular weight to
e500, calculated octanol-water partition coefficient to e5, the
number of HBDs (OH’s and NH’s) to e5, and HBAs (N’s and
O’s) to e10. Additional filtering restricted the polar surface area
to <120 A2, and the number of rotatable bonds to a maximum
of 7. The pharmacophore-based VS combined with these filters
resulted in the selection of 314 hits from the NCI database,
194 hits from the Maybridge database, and 61 hits from our
in-house database (total 569). This number of compounds
remained unwieldy and undoubtedly included a large fraction
of false positives; consequently, hits obtained were further
screened via molecular docking studies, that is, structure-based
investigation, drawing on several inhibitor-bound GSK-3βcrys-
tallographic structures, was conducted. Protein induced-fit
flexibility with ligand binding and the presence of a conserved
water molecule in the active site of GSK-3β render structure-
based VS challenging.38,39 Moreover, the ATP binding site of
GSK-3β readily accommodates many structurally diverse
ligands in a variety of manners, as evidenced among the
multiplicity of cocrystal structures of GSK-3β (including
PDB codes 1Q4L, 1Q3W, 1Q5K, 1Q41, 1UV5, 1Q3D,
1PYX, 2OW3) available in the Protein Data Bank. Pre-
viously, the PDB structures of GSK-3β were classified into
three major groups based on the alternative positioning of
Gln 185.39 Therefore, we selected three different GSK-3β
X-ray structures (1Q4L,27 1Q5K,40 and 1UV541) repre-
senting each conformational group for our docking VS.
We made use of Surflex-Dock software,42 as implemented
in SYBYL 8.0,43 for the docking studies of 569 hits
generated from the 3D pharmacophore search and rule-
based filters. Surflex-Dock is a fully automatic flexible
molecular docking algorithm that combines the scoring
function from the Hammerhead docking strategy with a
search engine that relies on a surface-based molecular
similarity method as a means to rapidly generate viable
putative poses for molecular fragments carved from the

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8537
Figure 3. Structures of final hits selected from those identified by the VS scheme.

8538 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24
Khanfar et al.
Table 1. Refined Hits Captured by VS and Their Corresponding in
Vitro Kinase Activities
to possible side effects, CDK2 was selected to represent other
related kinases, which should show low binding affinity toward
new GSK-3β hits to avoid possible toxicities. CDK2 is essential
for G1 progression and S-phase entry of the cell cycle.48
Inhibition of CDK2 will arrest the cell cycle and cellular
proliferation,48 which will complicate assessment of therapeutic
and toxicological effects of GSK-3β inhibition in GSK-3β-
related pathologies, including diabetes, inflammation, neuro-
logical disorders, and various neoplastic diseases. Therefore,
the affinities of all confirmed hits toward CDK2 were assessed.
% of GSK-3β GSK-3β % of CDK2
inhibition
at 10 μM
IC50
(nM)a
inhibition
at 10 μM
compd
source
1
NCI 146875
NCI 330187
NCI 382703
NCI 71283
NCI 619690
NCI 609243
NCI 382699
NCI 667083
NCI 609239
NCI 430239
NCI 1820652
NCI 357032
NCI 122482
NCI 90869
NCI 44685
NCI 690619
NCI 671801
NCI 667084
NCI 409027
NCI 1048917
NCI 382704
NCI 684340
M S06669SC
M SEW00923SC
M SEW00923SC
IH
71.5
33.9
48.5
100
1378 ( 281
ND
ND
ND
ND
ND
98
2
3
4
10.8 ( 1.5
ND
5
49.4
71.3
34.8
75.0
35.2
36.9
47.6
43.2
81.9
45.4
43.0
90.2
100
ND
ND
ND
ND
ND
ND
ND
ND
84
6
2400 ( 301
ND
Excluding hits 24 and 28, all other potent hits (GSK-3β IC50
<
7
8
4700 ( 199
ND
ND
1.0 μM) demonstrated strong inhibition of CDK2 at 10 μM.
The 2-anilino-1,3,4-oxadiazole 24 showed a high affinity to
GSK-3β(IC50 = 17.1 nM) but only inhibited CDK2 by 22% at
10 μM. To structurally understand the high affinity for GSK-
3β but not CDK2, compound 24 was docked inside the ATP-
binding site of GSK-3β using Surflex-Dock implemented in
SYBYL 8.0 (Figure 4). Only one pose, obtained from PDB
structure 1UV5, could well explain the SAR obtained from the
experimental IC50 values of its analogues (vide supra). In this
pose, the docking results suggest that oxadiazole 24 binds along
the hinge region of GSK-3β, making three hydrogen-bonding
interactions to main-chain (backbone) atoms of the protein.
Furthermore, in this identified pose, the nitro group of the
docked inhibitor occupies the inner part of the ATP pocket; the
closest distance between a nitro oxygen and the selectivity
residue Leu 132 is only 3.12 A. Importantly, the model indicates
that this nitro group occupies the position of a bound water
molecule from the starting PDB structure and thereby interacts
with lysine of the Asp 200-Lys 85 salt bridge via an apparent
H-bond. The docking study positions the electron-deficient
benzene ring so as to interact with the Cys 199 thiol via a
charge-transfer interaction. Finally, this pose positions the
other end group of 24, the aniline moiety, with its ring rotated
out of the plane from the oxadiazole ring, and so as to occupy a
hydrophobic pocket established by Ile 62, Leu 188, and the
methyl of Thr 138 (Figure 4A,B).
Similarly, a pose obtained starting from PDB structure
1UV5 well accounts for the experimental SAR among the
nine tested phenylmethylene hydantoin analogues. In this
pose, structure 28 binds through two hydrogen bonds to the
hinge region. The dioxolane ring of 28 is predicted to rest in
proximity to the selectivity residue Leu 132, with the meta-
oxygen bridged via two hydrogen bonds, one to Lys 85 and
the other to the backbone N-H of Asp 200. The model
furthermore predicts the aromatic ring to stack over the thiol
group of Cys 199, at 3.93 A, positioned for an effective
charge transfer interaction (Figure 4C,D).
The GSK-3β active site has been extensively characterized,
and two “selectivity residues” versus CDK2 have been iden-
tified: Leu 132 and Arg 141.27 Arg 141 forms a salt bridge
with Glu 137, therewith defining the boundary and entrance
region to the ATP pocket. The equivalent residues involved
in the salt bridge are Lys 89 and Glu 85 in CDK2. The Lys 89
side chain points inward to the ATP pocket, interacting with
a main-chain atom from a Gly-rich loop. This arrangement
would disallow occupancy by the aniline moiety of 24. The
other selectivity residue that differs between both kinases is
Leu 132 in GSK-3β vs Phe 80 in CDK2. Likely, the latter
residue and its surrounding structure in CDK2 cannot
accommodate, sterically or electronically or both, the nitro-
benzene moiety of 24, whereas the docking study suggests
that this group fits exquisitely in GSK-3β (see earlier dis-
cussion). A structural basis for the experimentally observed
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
ND
ND
214 ( 17
ND
ND
ND
ND
27
263 ( 77
113 ( 9
ND
91
35.5
90.4
56.2
71.8
37.9
96.9
100
ND
ND
ND
ND
ND
94
537 ( 110
ND
2456 ( 491
ND
325 ( 13
17 ( 2
ND
22
33.3
37.6
75.5
94.4
45.5
91.4
33.1
42.2
89.9
38.6
76.8
ND
ND
ND
27
ND
IH
IH
1975 ( 243
138 ( 26
ND
IH
IH
ND
ND
ND
ND
ND
ND
ND
1270 ( 99
ND
ND
IH
IH
IH
2810 ( 113
ND
IH
Manzamineb
8400 ( 739
a IC50 values for GSK-3β shown are the mean ( SD of duplicate or
triplicate measurements. b Used as positive control. NCI: National
Cancer Institute, M: Maybridge, IH: in-house database, ND: not
determined.
were identified. All compounds were stereochemically pure
single isomers except compound 2. This hit was virtually
screened as downloaded from the NCI database, with stereo-
chemistry as specified in Figure 3. Compound 2 was found to be
inactive (Table 1) and was not investigated further.
The chances of the validated hits to act as noncompetitive
or allosteric inhibitors, or alternatively, competitively bind
within the substrate-binding pocket, would be very remote
for the following reasons: (1) the algorithm by which the hits
were discovered, in particular considering the highly restric-
tive nature of the used in silico docking procedures, and (2)
the fact that the observed SAR (from experimental testing of
hits in an in vitro kinase assay, and subsequently a functional
assay (in vivo liver glycogen content) for the most potent
compounds) can be well rationalized by the most robust of
the docking models, which were constructed based on a
diverse set of X-ray crystallographic structures of ATP-
competitive inhibitors bound within the ATP binding pocket
of GSK-3β.
CDK2 is the most closely related kinase from a homology
perspective (overall 33% amino acid identity),25 and most
reported GSK-3β inhibitors also potently inhibit CDK2, no-
tably excepting a number of anilinomaleimides.7,12,26 Although
dual inhibition of GSK-3β and CDK2 is not necessarily linked

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8539
Figure 5. Effect of various doses of 24 (A) and 28 (B) on liver
glycogen storage in treated C57BL/6N mice measured 3 h after ip
dosing. Error bars indicate SEM, n = 3/dose.
high GSK-3β vs CDK2 selectivity of 28 cannot yet be
discerned from available structural information.
In Vivo Activities: Hepatic Glycogen Effects and BBB
Permeability. GSK-3β plays an important role in controlling
hepatic and skeletal muscular glycogen storage in that GSK-3β
phosphorylates and inactivates glycogen synthase.4 Suppression
of GSK-3β activity by insulin or small molecule inhibitors
increases glycogen storage in muscle and liver.4,11 A pre-
vious study showed that inhibition of GSK-3β selectively
reduces the expression of genes for glucose-6-phosphatase
and phosphoenolpyruvate carboxykinase.49 Overall, then,
glucose output is reduced and synthesis of glycogen from
glucose is increased.49 These findings indicate that GSK-3β
inhibitors may have valuable therapeutic potential for lowering
blood glucose levels.
The validated hits 24 and 28, which exhibited high potency
and selectivity for GSK-3β versus CDK2, were tested in mice
for their ability to increase liver glycogen reserves. The oxadia-
zole 24 significantly raised liver glycogen content in a dose-
dependent manner ((1, 5, and 15 mg/kg); Figure 5), increasing
glycogen content by 3.5- and 4.3-fold at 5 and 15 mg/kg,
respectively. Similarly, the phenylmethylene hydantoin 28 in-
creased glycogen reserves at 15 and 25 mg/kg dose levels by 5.8-
and 8.3-fold, respectively. Compounds 24 and 28 were well
tolerated by the mice; no fatalities were observed.
GSK-3β has also been linked to abnormalities associated
with AD, via hyperphosphorylation of the microtubule-
associated tau protein. For a GSK-3β inhibitor to counter
this abnormality, penetration to CNS neurons is requisite.
Figure 4. Detailed view of 24 (A/B) and 28 (C/D) docked within the
ATP binding site of GSK-3β projected over the Connolly electro-
static potential surface (blue, negative potential; red/brown, posi-
tive potential).

8540 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24
Khanfar et al.
Table 2. Plasma and Blood Concentrations of Compounds 24 and 28 at Different Dosesa
24
28
1 mg/kg
5 mg/kg
15 mg/kg
5 mg/kg
15 mg/kg
25 mg/kg
plasma (ng/mL)
brain (ng/g brain tissue)
14 ( 5
38 ( 6
31 ( 12
113 ( 54
61 ( 11
286 ( 58
<10
<50
81 ( 8
<50
3734 ( 109
12113 ( 2182
a Data are expressed as mean ( SEM of n = 3 using C57BL/6N mice following 3 h of the ip administration.
Scheme 1. Synthesis of 2-Anilino-5-phenyl-1,3,4-oxadiazolesa
a Reagents and conditions: (a) (1) N2H4, MeOH, rt, 1 h; (2) 65-80 °C, 20 min; (b) EDCI (3 equiv), CH2Cl2, rt, 12 h.
Table 3. Structures and GSK-3β Inhibitory Activities of 2-Anilino-5-
phenyl-1,3,4-oxadiazoles
Consequently, we assessed the ability of compounds 24 and
28 to cross the BBB. Using the same mice groups as in the
previous experiment, blood samples and brains were col-
lected and drug concentrations measured. Compound 24 was
detected in the brain at concentrations higher than in plasma
for all three tested doses (1, 5, 15 mg/kg; see Table 2). On the
other hand, compound 28 was not detected in plasma or brain
samples at 5 mg/kg and was only barely detectable at 15 mg/kg.
With the 25 mg/kg dose, however, the plasma levels were
substantial, with high penetration into brain. The abrupt,
disproportionate increase in plasma concentrations can be
rationalized by saturable efflux transport or biotransformation
processes or both. Any possible BBB efflux was not yet char-
acterized, as 28 is only a validated hit (albeit a potent one).
Chemistry and SAR of New 2-Anilino-5-phenyl-1,3,4-oxa-
diazole Derivatives. The validated hit 24 is a 2-anilino-5-
phenyl-1,3,4-oxadiazole derivative. Because this compound
shows high in vitro and in vivo potency against GSK-3β,
outstanding selectivity versus CDK2, and good brain perme-
ability, preliminary SAR studies were merited (i.e., initial hit-to-
lead chemistry). Moreover, a main intent was to help assess
whether or not any of the well-scoring poses obtained from the
docking studies could well explain experimental SAR results and,
thus, more profitably offer up further hypotheses for structure-
based design efforts in the absence of an X-ray structure.
A variety of synthetic methods for the preparation of
2-amino-1,3,4-oxadiazoles have been reported, almost all
involving multistep synthesis. Recently, an efficient one-pot
synthesis for preparing substituted 2-amino-1,3,4-oxadiazoles
was developed.32 A carboxylic acid and 4-phenyl-3-thiosemicar-
bazide are mixed in dichloromethane (DCM) at room tempera-
ture with three equivalents of N-(3-dimethylaminopropyl)-N1-
ethylcarbodiimide (EDCI) as coupling reagent, producing the
corresponding 2-anilino-5-phenyl-1,3,4-oxadiazoles (Scheme 1).
In our hands, as was reported by Chekler et al.,31 chromato-
graphic purification was not required in most cases. Two of the
needed thiosemicarbazides, 4-(4-methoxyphenyl)-3-thiosemi-
carbazide and 4-(4-nitrophenyl)-3-thiosemicarbazide, were not
available commercially but were readily prepared by reacting the
corresponding isothiocyanate with hydrazine under reflux in
MeOH.
compd
R1
R2
IC50 (nM) a
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
61
62
63
64
65
66
67
H
H
Cl
980 ( 63
>1000
H
H
CF3
CH3
632 ( 29
>1000
>1000
H
H
SCH3
SO2CH3
N(CH3)2
CN
H
439 ( 107
>1000
H
H
247 ( 55
209 ( 28
>1000
H
NO2
OCH3
OH
H
H
Cl
>1000
>1000
H
Cl
Cl
Cl
>1000
>1000
>1000
CF3
CH3
Cl
Cl
SCH3
SO2CH3
N(CH3)2
CN
855 ( 142
671 ( 89
>1000
Cl
Cl
Cl
Cl
719 ( 17
811 ( 58
>1000
NO2
OCH3
OH
Cl
Cl
>1000
>1000
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
OCH3
NO2
H
Cl
534 ( 29
53 ( 8
805 ( 192
>1000
CF3
CH3
SCH3
SO2CH3
N(CH3)2
CN
314 ( 33
>1000
18 ( 4
7 ( 1
662 ( 72
710 ( 39
NO2
OCH3
NO2
a IC50 values for GSK-3β shown are the mean ( SD of duplicate or
triplicate measurements.
electron-donating and more lipophilic (-σ/þπ) ones. A total
of 33 2-anilino-5-phenyl-1,3,4-oxadiazoles were synthesized
and evaluated for their in vitro GSK-3β inhibitory activity
(Table 3). For clarity of the following discussion, we denote the
aromatic ring derived from the benzoic acids as “ring A” and
the one derived from 4-phenyl-3-thiosemicarbazide as “ring B”.
Substitution of ring A with electron-withdrawing (þσ) groups
increased the activity compared to the unsubstituted ring.
Aromatic substituents were selected so as to systemati-
cally vary both lipophilic (π) and electronic (σ) character, as
prescribed by Craig’s plot.33 In this study, Cl, CF3, and NO2
served as electron-withdrawing (þσ) and lipophilic (þπ)
substituents, CN and SO2CH3 as electron-withdrawing and less
lipophilic (þσ/-π) ones, OH and OCH3 as electron-donating
and less lipophilic (-σ/-π) substituents, and CH3 and SCH3 as

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8541
Figure 6. (A) SAR of 2-anilino-5-phenyl-1,3,4-oxadiazole derivatives as defined by Craig’s plot substituents. (B) Binding mode of the most
active derivative (65). (C) Docked pose of 65 in best accord with experimental SAR results.
Compounds 37, 40, 42, and 43, with CF3, SO2CH3, CN, and
NO2, respectively, on the para-position of ring A, were more
active than unsubstituted compound 35 or compounds with
electron-donating groups (38 (CH3), 39 (SCH3), 41 (N(CH3)2),
or 44 (OCH3)). Similarly, 51, 53, and 54 were more active than
46, and compounds 58, 59, 62, 64, and 65 were more active than
57. On the other hand, CN and NO2 substitution conferred
greater activity than CF3 and SO2CH3: compounds 42 and 43
were more active than 37 and 40, compounds 53 and 54 were
more active than 48, and compounds 64 and 65 were more active
than 59 and 62. Replacement of NO2 or CN with Cl dramati-
cally decreased the activity, consistent with docking results
suggesting that H-bonding interactions and suitable spatial
positioning are of paramount importance. Along with Cl,
notably neither SO2CH3 nor CF3 could substitute effectively
for CN or NO2. Methoxy-substitution of ring B was found to
generate derivatives having greater activity than those with
a halogenated or unsubstituted ring: analogue 64 exhibited
14- and 40-fold greater activity than 42 and 53, respectively,
while 65 was found to be 30- and 116-fold more potent than 43
and 54, respectively. The two most active compounds, 64 and 65,
with IC50 values of 18 and 7 nM, respectively, have p-methoxy
substitution on ring B. Docking of 65 indicated an H-bond
between an N-H of the arginine and a lone pair on the ether
oxygen of the methoxy group (Figure 6C). One cannot rule out
a secondary edge-face charge transfer interaction between the
aromatic ring of 65 and the guanidine-moiety of Arg 141 and, of
course, the protein structure was held static in the docking
studies, whereas in actuality the protein could adjust to optimize
the guanidinium:methoxybenzene interaction. The inductive
effect of the OCH3 substituent might enhance a charge-transfer
interaction with the guanidine-moiety of Arg 141; however, the
docking results suggest the alternative possibility for an HBA
interaction with Thr 138, and an X-ray structure would be
required to know which, if either, of these possibilities accounts
for the experimental observations.
pharmacophore-based searching in conjunction with suitable
druglikeness filters, followed by docking studies employing
rigorous criteria to identify novel compounds with potent
GSK-3β inhibitory activities. The validity of the pharmaco-
phore map was established by submitting the corresponding
query to a test database of active GSK-3β inhibitors and
decoys from a multiplicity of molecule classes. Subsequent
docking studies proceeding from three disparate GSK-3β
crystallographic structures were then carried out, applying
highly restrictive criteria so as to refine the initial screening set
into a sufficiently smaller subset as to readily allow experi-
mental evaluations in the kinase assay. The result was that
from the 289903 molecules in the combined databases, 34
available compounds from among 59 identified in silico hits
were selected for biological testing. With only the very modest
effort and expense involved in experimentally screening these
34 compounds, we identified eight interesting new validated
hits having nanomolar affinity. Importantly, the hits identi-
fied in our study had wide structural diversity, and little true
structural similarity to any of the other known GSK-3β
inhibitors, highlighting an important merit of the virtual
screening approach. The most potent compound, 24, was a
2-anilino-5-phenyl-1,3,4-oxadiazole, which at first take seems
to bear some structural semblance to the highly selective and
potent 2-{3-[4-(alkylsulfinyl)phenyl]-1-benzofuran-5-yl}-5-
methyl-1,3,4-oxadiazole GSK-3β inhibitor recently divulged
by the Takeda group.50,51 Our in silico results suggest, how-
ever, that the roles of oxadiazole in the bridging of 24 is very
different than for the Takeda compounds, and 24 has better
potency and selectivity. Moreover, the fact that analogues of
24 could be synthesized by a one-step reaction, in contrast to
the relatively more complex synthetic schemes required to
produce Takeda’s oxadiazoles,50,51 facilitated initial explora-
tion of the SAR. Docking studies in conjunction with the
experimental results suggest that prodigious interaction of the
optimal nitro substituent with the salt-bridged Lys 85 inside
the ATP-binding site of GSK-3β, highly likely along with the
displacement of a bound water, confers high affinity, and
along with other exquisitely suited structural attributes, ex-
cellent selectively versus CDK2. The positive outcomes of our
Conclusions
Herein we present a successful example of employing a
multistage VS scheme, beginning with robustly designed

8542 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24
Khanfar et al.
preliminary in vivo evaluations furthermore suggest that this
new series could be of value for further lead optimization toward
the design of therapeutically relevant GSK-3β inhibitors.
MHz) δ 7.04 (dd, 1H, J = 7.3, 1.1), 7.39 (dd, 2H, J = 7.7, 7.3),
7.62 (dd, 2H, J = 7.7, 1.1), 7.96 (d, 2H, J = 8.0), 8.10 (d, 1H, J =
8.0) 10.81 (s, 1H). 13C NMR (DMSO-d6, 100 MHz) δ 117.8
(CH), 122.7 (qC), 124.3 (CH), 126.8 (CH), 127.9 (CH), 129.4
(CH), 131.2 (qC), 138.1 (qC), 160.7 (qC), 161.5 (qC). HREIMS
m/z 306.0830 [MþH]þ (calcd for C15H11F3N3O, 306.0854).
5-(4-(Methylmercapto)phenyl)-N-phenyl-1,3,4-oxadiazol-2-amine
(39). Compound 39 was prepared according to procedure B to
Experimental Section
General Experimental Procedures. IR spectra were recorded
on a Varian 800 FT-IR spectrophotometer (Palo Alto, CA).
Optical rotation measurements were carried out on a Rudolph
Research Analytical Autopol III polarimeter (Hackettstown,
NJ). The 1H and 13C NMR spectra were recorded in DMSO-d6,
using TMS as an internal standard, on a JEOL Eclipse NMR
spectrometer (Tokyo, Japan) operating at 400 MHz for 1H and
100 MHz for 13C. The HREIMS experiments were conducted on
a 6200-TOF LCMS (Agilent, Santa Clara, CA) equipped with a
multimode source (mixed source that can ionize the samples
alternatively by ESI or APCI). Analytical HPLC analyses were
performed on a Shimadzu HPLC system (Columbia, MD) using
a 5 μm C18 column (150 mm ꢀ 4.6 mm id; Agilent, Santa Clara,
CA) and isocratic elution (CH3CN:H2O; 75:25) with UV detec-
tion set at 305 and 220 nm to verify the purity of active virtual
hits and synthetic compounds. A purity of >95% has been
established for compounds 1, 4, 6, 13, 16, 17, 19, 21, 23, 24, 27,
28, 30, 33, and 34-67. Compound 8 achieved 89% purity level.
TLC analyses were carried out on precoated Si gel 60 F254
500 μm TLC plates, using MeOH/CH2Cl2 (2:8) as a developing
solvent.
General Chemical Reaction Procedures. A. Preparation of
4-Phenyl-3-thiosemicarbazide. The appropriate phenylisothio-
cyanate (30 mmol) was added, dropwise, over a period of 1 h to a
stirred solution of hydrazine (30 mmol) in CH3OH (8 mL) at
65-80 °C. The stirring of the reaction mixture was continued for
20 min more at the same temperature. The solvent was removed
by evaporation in vacuo. The precipitate was collected by filtra-
tion and washed with petroleum ether. The compound was used
without further purification.52
1
afford 48 mg of white powder (51%). H NMR (DMSO-d6, 400
MHz) δ 2.50 (s, 1H), 7.02 (d, 1H, J = 7.3), 7.39 (dd, 2H, J = 8.4,
7.3), 7.43 (d, 2H, J = 8.4), 7.63 (d, 2H, J = 8.0), 7.81 (d, 1H, J =
8.0) 10.68 (s, 1H). 13C NMR (DMSO-d6, 100 MHz) δ 14.7 (CH3),
117.6 (CH), 120.75 (qC), 122.4 (CH), 126.5 (CH), 129.7 (CH), 139.2
(qC), 142.9 (qC), 158.1 (qC), 160.2 (qC). HREIMS m/z 284.0838
[MþH]þ (calcd for C15H14N3OS, 284.0858).
5-(4-(Methylsulfonyl)phenyl)-N-phenyl-1,3,4-oxadiazol-2-amine
(40). Compound 40 was prepared according to procedure B to
afford 32 mg of white powder (31%). 1H NMR (DMSO-d6, 400
MHz) δ 3.37 (s, 3H), 7.02 (d, 1H, J = 7.4), 7.37 (dd, 2H, J = 8.9,
7.4), 7.62 (d, 2H, J = 8.9), 8.12 (s, 4H). 13C NMR (DMSO-d6, 100
MHz) δ 40.1 (CH3), 117.0 (CH), 122.7 (CH), 129.6 (CH), 130.5
(qC), 139.1 (qC), 140.2 (qC), 159.6 (qC), 160.2 (qC). HREIMS
m/z 316.0755 [MþH]þ (calcd for C15H14N3O3S 316.0756).
N,5-Bis(4-chlorophenyl)-1,3,4-oxadiazol-2-amine (47). Com-
pound 47 was prepared according to procedure B to afford 42
mg of white powder (42%). 1H NMR (DMSO-d6, 400 MHz) δ
7.43 (d, 2H, J = 7.0), 7.64 (d, 2H, J = 7.3), 7.66 (d, 2H, J = 7.3),
7.81 (d, 1H, J = 7.0), 7.89 (d, 2H, J = 7.0), 10.92 (s, 1H). 13
C
NMR (DMSO-d6, 100 MHz) 122.6 (CH), 124.7 (qC), 127.4
(qC), 128.5 (CH), 129.6 (CH), 129.9 (CH) 134.2 (qC), 137.9
(qC), 159.6 (qC), 161.2 (qC). HREIMS m/z 306.0182 [MþH]þ
(calcd for C14H10Cl2N3O, 306.0201).
N-(4-Chlorophenyl)-5-(4-(trifluoromethyl)phenyl)-1,3,4-oxa-
diazol-2-amine (48). Compound 48 was prepared according to
procedure B to afford 32 mg of white powder (29%). 1H NMR
(DMSO-d6, 400 MHz) δ 7.45 (d, 2H, J = 8.8), 7.65 (d, 2H, J =
8.8), 7.96 (d, 2H, J = 8.4), 8.1 (d, 2H, J = 8.4), 11.02 (s, 1H).
13C NMR (DMSO-d6, 100 MHz) 120.6 (CH), 124.3 (qC), 124.7
(qC), 125.3 (CH), 127.6 (qC), 128.0 (CH) 129.0 (CH), 135.2
(qC), 143.6 (qC), 160.0 (qC), 161.4 (qC). HREIMS m/z
340.0450 [MþH]þ (calcd for C15H10ClF3N3O, 340.0464).
N-(4-Chlorophenyl)-5-p-tolyl-1,3,4-oxadiazol-2-amine (49). Com-
pound 49 was prepared according to procedure B to afford 26 mg of
B. Preparation of 2-Anilino-5-phenyl-1,3,4-oxadiazole Deriva-
tives. A carboxylic acid (0.33 mmol), a 4-phenyl-3-thiosemicar-
bazide (0.33 mmol), and EDCI (182 mg, 1.0 mmol) were mixed
in CH2Cl2 (15 mL), and the reaction mixture was stirred at room
temperature. Product precipitate formed after several hours,
and after stirring for a total of 12 h, the precipitate was collected
by filtration, washed with dichloromethane (20 mL), and dried
on a filtering funnel to afford the desired product.32
1
white powder (28%). H NMR (DMSO-d6, 400 MHz) δ 2.39 (s,
3H), 7.39 (d, 2H, J = 8.1), 7.43 (d, 2H, J = 8.1), 7.63 (d, 2H, J =
8.4), 7.79 (d, 2H, J = 8.4), 10.86 (s, 1H). 13C NMR (DMSO-d6, 100
MHz) 22.1 (CH3), 120.1 (CH), 124.2 (qC), 128.7 (CH), 129.3 (CH),
129.9 (CH), 131.0 (qC) 135.1 (qC), 135.4 (qC), 161.6 (qC), 162.0
(qC). HREIMS m/z 286.0731 [MþH]þ (calcd for C15H13ClN3O,
286.0747).
C. Preparation of Phenylmethylene Hydantoins. Hydantoin
(1.0 g, 10 mmol) was dissolved in 10 mL of H2O while heating at
70 °C in an oil bath with continuous stirring (Scheme 1).41 The
pH was adjusted to 7.0 using saturated NaHCO3. After dissolu-
tion was complete, the temperature was then raised to 90 °C.
Ethanolamine (0.9 mL, 14.9 mmol) was added, and an equimo-
lar quantity of the corresponding aldehyde as a solution in 2-5
mL of EtOH was then added dropwise with continuous stirring.
The solution was refluxed for approximately 5-8 h, during
which time a yellow or white precipitate formed. The reaction
was monitored hourly by TLC, and when complete depletion of
the starting aldehyde was apparent, the mixture was cooled, and
the precipitate was collected by filtration and washed with
EtOH-H2O (1:5) before recrystallization from EtOH.
N-(4-Chlorophenyl)-5-(4-(methylmercapto)phenyl)-1,3,4-oxa-
diazol-2-amine (50). Compound 50 was prepared according to
procedure B to afford 36 mg of white powder (34%). H NMR
1
(DMSO-d6, 400 MHz) δ 2.55 (s, 3H), 7.42 (d, 2H, J = 8.7), 7.45 (d,
2H, J = 8.8), 7.64 (d, 2H, J = 8.8), 7.81 (d, 2H, J = 8.7), 11.04 (s,
1H). 13C NMR (DMSO-d6, 100 MHz) 14.6 (CH3), 114.2 (CH),
124.5 (qC), 128.1 (CH), 129.2 (CH), 129.7 (CH), 131.2 (qC) 134.4
(qC), 143.5 (qC), 160.6 (qC), 161.8 (qC). HREIMS m/z 318.0451
[MþH]þ (calcd for C15H13ClN3OS, 318.0468).
(Z)-5-(2,3-Methylenedioxyphenyl)imidazolidine-2,4-dione (28).
Compound 28 was prepared according to procedure C to afford
1.67 g of white powder (72%). 1H NMR (DMSO-d6, 400 MHz) δ
6.12 (s, 2H), 6.31(s, 1H), 6.87 (dd, 1H, J = 1.5, 7.7), 6.90 (dd, 1H,
J = 8.0, 7.7), 8.12 (dd, 1H, J = 8.0, 1.5) 10.22 (s, 1H), 11.30 (s,
1H). 13C NMR (DMSO-d6, 100 MHz) δ 99.8 (CH2), 114.7 (CH),
116.2 (qC), 120.8 (CH), 121.9(CH), 122.4(CH), 124.7(qC), 147.2
(qC), 147.1 (qC), 158.7 (qC), 162.5 (qC). HREIMS m/z 255.0380
[MþNa]þ (calcd for C11H8N2O4Na, 255.0382).
N-(4-Chlorophenyl)-5-(4-(methylsulfonyl)phenyl)-1,3,4-oxa-
diazol-2-amine (51). Compound 51 was prepared according to
procedure B to afford 41 mg of white powder (36%). 1H NMR
(DMSO-d6, 400 MHz) δ 7.45 (d, 2H, J = 8.8), 7.65 (d, 2H, J =
8.8), 8.13 (s, 4H), 11.04 (s, 1H). 13C NMR (DMSO-d6, 100
MHz) 43.1 (CH3), 114.2 (CH), 124.5 (qC), 128.1 (CH), 129.2
(CH), 129.7 (CH), 131.2 (qC) 134.4 (qC), 143.5 (qC), 160.6
(qC), 161.8 (qC). HREIMS m/z 350.0369 [MþH]þ (calcd for
5-(4-(Trifluoromethyl)phenyl)-N-phenyl-1,3,4-oxadiazol-2-amine
(37). Compound 37 was prepared according to procedure B to
afford 8.9 mg of white powder (10%). 1H NMR (DMSO-d6, 400
C
15H13ClN3O3S, 350.0366).
N-(4-Chlorophenyl)-5-(4-(dimethylamino)phenyl)-1,3,4-oxa-
diazol-2-amine (52). Compound 52 was prepared according to

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8543
procedure B to afford 19 mg of white powder (19%). 1H NMR
(DMSO-d6, 400 MHz) δ 3.01 (s, 6H), 6.83 (d, 2H, J = 8.8), 7.40
(d, 2H, J = 8.8), 7.62 (d, 2H, 8.8), 7.69 (d, 2H, 8.8), 10.69 (s,
1H). 13C NMR (DMSO-d6, 100 MHz) 40.5 (CH3), 119.4 (CH),
126.3 (qC), 126.9 (CH), 128.7 (CH) 128.7 (qC), 129.6 (CH),
137.9 (qC) 143.0 (qC), 157.5 (qC), 160.7 (qC). HREIMS m/z
315.1016 [MþH]þ (calcd for C16H16ClN4O, 315.1013).
5-(4-(Dimethylamino)phenyl)-N-(4-methoxyphenyl)-1,3,4-ox-
adiazol-2-amine (63). Compound 63 was prepared according to
procedure B to afford 18 mg of white powder (16%). 1H NMR
(DMSO-d6, 400 MHz) δ 2.99 (s, 6H), 3.73 (s, 3H), 6.82 (d, 2H,
J = 7.0), 6.94 (d, 2H, J = 7.0), 7.51 (d, 2H, J = 7.0), 7.67 (d, 2H,
J = 7.0), 10.27 (s, 1H). 13C NMR (DMSO-d6, 100 MHz) 40.3
(CH3), 55.8 (CH3), 111.6 (qC), 112.5 (CH), 114.9 (CH), 118.9
(CH), 127.3 (CH), 129.4 (qC) 133.2 (qC), 155.6 (qC), 158.9 (qC),
159.2 (qC). HREIMS m/z 311.1519 [MþH]þ (calcd for C17H19-
N4O2, 311.1508).
N,5-Bis(4-methoxyphenyl)-1,3,4-oxadiazol-2-amine (66). Com-
pound 66 was prepared according to procedure B to afford 38 mg
of white powder (39%). 1H NMR (DMSO-d6, 400 MHz) δ 3.73 (s,
3H), 3.84 (s, 3H), 6.96 (d, 2H, J = 9.0), 7.13 (d, 2H, J = 8.8), 7.52
(d, 2H, J = 9.0), 7.81 (d, 2H, J = 8.8), 10.40 (s, 1H). 13C NMR
(DMSO-d6, 100 MHz) 55.8 (CH3), 65.0 (CH3), 114.9 (CH), 115.4
(CH), 117.0 (qC), 119.0 (CH), 127.8 (CH), 133.6 (qC), 154.9 (qC),
158.0 (qC), 160.3 (qC), 161.8 (qC). HREIMS m/z 298.1181
[MþH]þ (calcd for C16H16N3O3, 298.1192).
4-(5-(4-Chlorophenyl)-1,3,4-oxadiazol-2-ylamino)benzonitrile
(53). Compound 53 was prepared according to procedure B to
afford 38 mg of white powder (39%). 1H NMR (DMSO-d6, 400
MHz) δ 7.44 (d, 2H, J = 8.7), 7.65 (d, 2H, J = 8.7), 8.05 (s, 4H),
11.09 (s, 1H). 13C NMR (DMSO-d6, 100 MHz) 113.5 (qC), 119.4
(CH), 126.2 (qC), 126.7 (CH) 127.9 (qC), 128.3 (qC), 129.6 (CH)
133.9 (CH), 138.5 (qC), 158.6 (qC), 161.4 (qC). HREIMS m/z
297.0546 [MþH]þ (calcd for C15H10ClN4O, 297.0543).
N-(4-Chlorophenyl)-5-(4-nitrophenyl)-1,3,4-oxadiazol-2-amine
(54). Compound 54 was prepared according to procedure B to
afford 61 mg of orange powder (61%). 1H NMR (DMSO-d6, 400
MHz) δ 7.44 (d, 2H, J = 8.8), 7.65 (d, 2H, J = 8.8), 8.13 (d, 2H,
J = 8.4), 8.38 (d, 2H, J = 8.3), 11.08 (s, 1H). 13C NMR (DMSO-
d6, 100 MHz) 56.3 (CH3), 117.2 (CH), 122.5 (CH), 125.3 (qC)
127.9 (CH), 129.2 (CH), 129.4 (qC) 133.5 (qC), 155.5 (qC), 158.5
(qC), 161.4 (qC). HREIMS m/z 317.0430 [MþH]þ (calcd for
Molecular Modeling. Three-dimensional structure building
and all modeling were carried out using components within the
SYBYL program package, version 8.0,43 installed on Dell desktop
workstation equipped with dual 2.0 GHz Intel Xeon processor
running the Red Hat Enterprise Linux (version 5) operating
system. Energy minimizations were performed using the Tripos
force field with a distance-dependent dielectric and the Powell
conjugate gradient algorithm with a convergence criterion of 0.01
kcal/(mol-A). Partial atomic charges were computed using the
semiempirical AM1 approach within the program MOPAC 6.0.
Kinase structures were prepared using the Biopolymer-Prepare
module within SYBYL 8.0; hydrogen atoms were added, the
protonation states of charged residues within the ATP-binding
site of GSK-3β were inspected, and if necessary, altered as appro-
priate, and AMBER7 FF99 charges were loaded onto kinase
structures. In preliminary trials, docking studies gave better results
(docking scores and optimum binding orientation and inter-
actions) in the absence of all but one of the water molecules;
therefore, structure-based virtual screening was conducted without
water molecules, except the conserved water molecule (H2104).
Virtual Screening and Molecular Docking. As outlined in
Figure 2, NCI (number of molecules: 234055), Maybridge
(55541), and in-house databases (307) were searched for new
hits by employing the developed pharmacophoric queries, as
described in a previous report.11 Prior filtering was carried out
applying Lipinski’s rule of five31 and also restricting the number
of rotatable bonds to a maximum of 7 and the polar surface area
to e120 A2. A number of hits were obtained from each of the
databases, and these were further screened using molecular
docking studies with Surflex-Dock version 2.0, interfaced with
SYBYL 8.0, docking the compounds to the ATP binding site
of GSK-3β. The 3D kinase structures were taken from the
Brookhaven Protein Databank (PDB code: 1Q4L, 1Q5K,
1UV5). Surflex-Dock employs an idealized active site ligand
(protomol) as a target to generate putative poses of molecules or
molecular fragments. These putative poses were scored using the
Hammerhead and Gold score scoring functions.
C
14H10ClN4O3, 317.0441).
5-(4-Chlorophenyl)-N-(4-methoxyphenyl)-1,3,4-oxadiazol-2-
amine (58). Compound 58 was prepared according to proce-
dure B to afford 80 mg of white powder (80%). 1H NMR
(DMSO-d6, 400 MHz) δ 3.74 (s, 3H), 7.53 (d, 2H, J = 8.8), 7.64
(d, 2H, J = 9.1), 7.88 (d, 2H, J = 8.8), 7.94 (d, 2H, J = 9.1),
10.49 (s, 1H). 13C NMR (DMSO-d6, 100 MHz) 55.8 (CH3),
114.9 (CH), 119.3 (CH), 123.4 (qC), 127.8 (CH), 130.1 (CH), 132.3
(qC), 136.0 (qC), 155.1 (qC), 158.5 (qC), 160.9 (qC). HREIMS m/z
302.0694 [MþH]þ (calcd for C15H13ClN3O2, 302.0696).
N-(4-Methoxyphenyl)-5-(4-(trifluoromethyl)phenyl)-1,3,4-oxa-
diazol-2-amine (59). Compound 59 was prepared according to
procedure B to afford 69 mg of white powder (63%). H NMR
1
(DMSO-d6, 400 MHz) δ 3.74 (s, 3H), 6.96 (d, 2H, J = 9.2), 7.54 (d,
2H, J = 9.2), 7.95 (d, 2H, J = 8.4), 8.08 (d, 2H, J = 8.4), 10.6 (s,
1H). 13C NMR (DMSO-d6, 100 MHz) δ 55.8 (CH3), 114.9 (CH),
119.3 (CH), 123.4 (qC), 125.3 (qC), 126.7 (CH), 126.9 (CH), 129.4
(qC) 132.2 (qC), 155.2 (qC), 157.1 (qC), 161.2 (qC). HREIMS m/z
336.0965 [MþH]þ (calcd for C16H13F3N3O2, 336.0960).
N-(4-Methoxyphenyl)-5-p-tolyl-1,3,4-oxadiazol-2-amine (60).
Compound 60 was prepared according to procedure B to afford
80 mg of white powder (80%). 1H NMR (DMSO-d6, 400 MHz)
δ 2.39 (s, 3H), 3.74 (s, 3H), 7.33 (d, 2H, J = 8.6), 7.64 (d, 2H, J =
9.1), 7.92 (d, 2H, J = 9.1), 7.93 (d, 2H, J = 8.6), 10.39 (s, 1H).
13C NMR (DMSO-d6, 100 MHz) 55.8 (CH3), 114.9 (CH), 119.3
(CH), 123.4 (qC), 127.8 (CH), 130.1 (CH), 132.3 (qC), 136.0
(qC), 155.1 (qC), 158.5 (qC), 160.9 (qC). HREIMS m/z 282.1230
[MþH]þ (calcd for C16H16N3O2, 282.1243).
N-(4-Methoxyphenyl)-5-(4-(methylthio)phenyl)-1,3,4-oxadia-
zol-2-amine (61). Compound 61 was prepared according to
procedure B to afford 54 mg of white powder (54%). 1H
NMR (DMSO-d6, 400 MHz) δ 2.54 (s, 3H), 3.73 (s, 3H), 6.94
(d, 2H, J = 9.2), 7.43 (d, 2H, J = 8.7), 7.52 (d, 2H, J = 9.1), 7.79
(d, 2H, J = 8.7), 10.43 (s, 1H). 13C NMR (DMSO-d6, 100 MHz)
14.7 (CH3), 55.8 (CH3), 114.8 (CH), 116.5 (CH), 127.8 (CH),
129.4 (qC), 129.7 (CH), 132.5 (qC), 136.3 (qC), 155.4 (qC), 158.5
(qC), 161.2 (qC). HREIMS m/z 314.0964 [MþH]þ (calcd for
C16H16N3O2S, 314.0963).
In Vitro GSK-3β and CDK2 Inhibitory Activity Assay Using
Z0-LYTE Kinase Assay Kit.46 Recombinant GSK-3β and
CDK2 were purchased from Invitrogen (Carlsbad, CA). The
GSK-3β and CDK2 kinase assays were carried out with the
Invitrogen Z0-LYTE Kinase Assay kit. The assay was optimized
for use with GSK-3β as described in the Invitrogen protocol.
The GSK-3β concentration was optimized to obtain the desired
percent phosphorylation with an acceptable Z0-factor value,
which indicates the quality of an assay; Z0-factor values of 0.5
or greater classify an assay as excellent. A Z0-factor value of
0.74 was obtained at final kinase and ATP concentrations of
50 ng/mL and 15 μM, respectively. Tested concentrations
ranged from 1 nM to 10 μM distributed log-linearly across the
concentration range, and at least two data points from each
N-(4-Methoxyphenyl)-5-(4-(methylsulfonyl)phenyl)-1,3,4-oxa-
diazol-2-amine (62). Compound 62 was prepared according to
procedure B to afford 68 mg of white powder (60%). 1H NMR
(DMSO-d6, 400 MHz) δ 3.26 (s, 3H), 3.71 (s, 3H), 6.93 (d, 2H,
J = 8.8), 7.50 (d, 2H, J = 8.8), 8.08 (s, 4H), 10.59 (s, 1H). 13
C
NMR (DMSO-d6, 100 MHz) 43.8 (CH3), 55.8 (CH3), 114.9
(CH), 119.4 (CH), 126.7 (CH), 128.8 (CH), 128.9 (qC) 132.2
(qC), 142.70 (qC), 155.2 (qC), 157.1 (qC), 161.2 (qC). HREIMS
m/z 346.0862 [MþH]þ (calcd for C16H16N3O4S, 346.0862).

8544 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24
Khanfar et al.
concentration were collected. The IC50 value for each experiment
was obtained using nonlinear regression of the log(concentration)
versus percent inhibition values (GraphPad Prism 5.0). Assays
were conducted in triplicate, average values were calculated, and
no outliers were removed.
Supporting Information Available: Analytical data of com-
pounds 1, 4, 6, 8, 13, 16, 17, 19, 21, 23, 24, 27, 28, 30, and GSK-3β
inhibitory activity graphs of 24 and 28. This material is available
free of charge via the Internet at http://pubs.acs.org.
In-Vivo Evaluation. Six-week-old male C57BL/6N mice with
weights averaging 22 g were assigned into seven groups in
randomized fashion, and fed ad libitum with standard food
and water, except when fasting was included in the course of a
study. All animals were housed under the same conditions. On
the day of the experiment, food and water were removed 4 h
before the injection. Three groups (3 mice/group) were used to
investigate each of the two test compounds (24 and 28), with an
additional group (n = 3) serving as control. Animals were
injected intraperitoneally (ip) with one of three doses (1, 5, and
15 mg/kg for compound 24; 5, 15, and 25 mg/kg for compound 28).
The control group was injected ip with the PBS vehicle (n = 3).
After 3 h, blood, liver, and brain were rapidly collected.
References
(1) Ishiguro, K.; Shiratsuchi, A.; Sato, S.; Omori, A.; Arioka, M.;
Kobayashi, S.; Uchida, T.; Imahori, K. Glycogen synthase kinase 3
beta is identical to tau protein kinase I generating several epitopes
of paired helical filaments. FEBS Lett. 1993, 325, 167–172.
(2) (a) Lovestone, S.; Reynolds, C. H.; Latimer, D.; Davis, D. R.;
Anderton, B. H.; Gallo, J.-M.; Hanger, D.; Mulot, S.; Marquardt,
B. Alzheimer’s disease-like phosphorylation of the microtubule-
associated protein tau by glycogen synthase kinase-3 in transfected
mammalian cells. Curr. Biol. 1994, 4, 1077–1086. (b) Klein, P. S.;
Melton, D. A. A molecular mechanism for the effect of lithium on
development. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 8455–8459.
(c) Pap, M.; Cooper, G. M. Role of glycogen synthase kinase-3 in the
phosphatidylinositol 3-kinase/Akt cell survival pathway. J. Biol.
Chem. 1998, 273, 19929–19932.
Determination of Liver Glycogen. Liver glycogen content was
characterized following a reported procedure.53 Briefly, livers
were removed immediately after the animals were killed and
were homogenized (IKA-T8 Ultra-Turrax, Germany) in an
appropriate volume of 5% trichloroacetic acid (TFA) for
5 min. The homogenate was centrifuged (Eppendorf centrifuge
5804 R, Germany) at 3000 rpm for 5 min. The supernatant fluid
was filtered using acid-washed paper, and the residues were
homogenized again with another volume of 5% TFA for 1-3
min to ensure relatively complete extraction of glycogen. The
glycogen contained in 1.0 mL of this filtrate was precipitated
using ethanol (95%, 5 mL), incubating in a water bath at 37-40 °C
for 3 h. Upon centrifugation at 3000 rpm for 15 min, the clear
liquid was gently decanted from the packed glycogen, and the
tubes were allowed to drain in an inverted position for 10 min.
The glycogen was redissolved in distilled water (2 mL), and
the solution mixed with 10 mL of anthrone reagent (0.05%
anthrone, 1.0% thiourea in 72% H2SO4). The mixture was
incubated in boiling water for 30 min, and subsequently, the
absorbance was measured spectrophotometrically at 620 nm
(Cintra 20, GBC). Blank and standard solutions were prepared,
respectively, by adding 10 mL of anthrone reagent to 2 mL of
H2O (blank) or to 2 mL of glucose solution (standard) contain-
ing 0.1 mg of glucose in saturated benzoic acid. The liver
glycogen content was estimated using the following formula:
(3) Welsh, G. I.; Proud, C. G. Glycogen synthase kinase-3 is rapidly
inactivated in response to insulin and phosphorylates eukaryotic
initiation factor eIF-2B. Biochem. J. 1993, 294, 625–629.
(4) Wagman, A. S.; Johnson, K. W.; Bussiere, D. E. Discovery and
development of GSK3 inhibitors for the treatment of type 2
diabetes. Curr. Pharm. Des. 2004, 10, 1105–1137.
(5) Li, X.; Lu, F.; Tian, Q.; Yang, Y.; Wang, Q.; Wang, J. Z. Activation
of glycogen synthase kinase-3 induces Alzheimer-like tau hyper-
phosphorylation in rat hippocampus slices in culture. J. Neural
Transm. 2006, 113, 93–102.
(6) Neary, J. T. Kang, Yuan. P2 purinergic receptors signal to glyco-
gen synthase kinase-3beta in astrocytes. J. Neurosci. Res. 2006, 84,
515–524.
(7) Martinez, A.; Castro, A.; Dorronsoro, I.; Alonso, M. Glycogen
synthase kinase 3 (GSK-3) inhibitors as new promising drugs for
diabetes, neurodegeneration, cancer, and inflammation. Med. Res.
Rev. 2002, 22, 373–384.
(8) Martinez, A.; Alonso, M.; Castro, A.; Perez, C.; Moreno, F. J. First
non-ATP competitive glycogen synthase kinase 3beta (GSK-3beta)
inhibitors: thiadiazolidinones (TDZD) as potential drugs for the
treatment of Alzheimer’s disease. J. Med. Chem. 2002, 45, 1292–
1299.
(9) Naerum, L.; Norskov-Lauritsen, L.; Olesen, P. H. Scaffold hop-
ping and optimization towards libraries of glycogen synthase
kinase-3 inhibitors. Bioorg. Med. Chem. Lett. 2002, 12, 1525–1528.
(10) Witherington, J., 3-Amino Pyrazoles as Potent and Selective
Glycogen Kinase Synthase 3 (GSK-3) Inhibitors. In Glycogen
Synthase Kinase 3 (GSK-3) and Its Inhibitors; Martinez, A.; Castro,
A.; Medina, M., Eds.; John Wiley & Sons, Inc.: New York, 2006; pp
281-305.
amount ðmgÞof glycogen in liver tissue ¼ ðDU=DSÞ
ꢀðvolume of extract ðmLÞ=weight of liver tissue ðgÞÞ ꢀ 0:09
(11) Khanfar, M. A.; Abuasal, B.; Mudit, M; Kaddoumi, A; El Sayed,
K. Marine natural-derived inhibitors of glycogen synthase kinase-
3β phenylmethylene hydantoins: In vitro and in vivo activities,
pharmacophore modeling, and virtual screening studies. Bioorg.
Med. Chem. 2009, 17, 6032–6039.
where DU is the absorbance of the unknown sample and DS is
the absorbance of the glucose standard.
(12) Martinez, A.; Alonso, M.; Castro, A.; Dorronsoro, I.; Gelpi, J. L.;
Luque, F. J.; Perez, C.; Moreno, F. J. SAR and 3D-QSAR studies
on thiadiazolidinone derivatives: exploration of structural require-
ments for glycogen synthase kinase 3 inhibitors. J. Med. Chem.
2005, 48, 7103–7112.
(13) Olesen, P. H.; Sorensen, A. R.; Urso, B.; Kurtzhals, P.; Bowler,
A. N.; Ehrbar, U.; Hansen, B. F. Synthesis and in vitro char-
acterization of 1-(4-aminofurazan-3-yl)-5-dialkylaminomethyl-
1H-[1,2,3]triazole-4-carboxylic acid derivatives. A new class of
selective GSK-3 inhibitors. J. Med. Chem. 2003, 46, 3333–3341.
(14) Berg, S.; Hellberg, S. Preparation of N-(4-methoxybenzyl)-N0-
(5-nitro-1,3-thiazol-2-yl)urea for treating conditions associated
with glycogen-synthase kinase-3 (GSK3). PCT Int. Appl. WO
2003004478, 2003.
(15) O’Neill, D. J.; Shen, L.; Prouty, C.; Conway, B. R.; Westover, L.;
Xu, J. Z.; Zhang, H. C.; Maryanoff, B. E.; Murray, W. V.;
Demarest, K. T.; Kuo, G. H. Design, synthesis, and biological
evaluation of novel 7-azaindolyl-heteroaryl-maleimides as potent
and selective glycogen synthase kinase-3β (GSK-3β) inhibitors.
Bioorg. Med. Chem. 2004, 12, 3167–3185.
(16) Nuss, J. M.; Harrison, S. D.; Ring, D. B.; Boyce, R. S.; Brown,
S. P.; Goff, D.; Johnson, K.; Pfister, K. B.; Ramurthy, S.;
Renhowe, P. A.; Seely, L.; Subramanian, S.; Wagman, A. S.; Zhou,
X. A. Preparation of aminopyrimidines and -pyridines as glycogen
synthase kinase 3 inhibitors. PCT Int. Appl. WO 9965897, 1999.
Brain and Plasma Concentration in Mice. Compounds 24 and
28 were administered to fasted C57BL/6N mice as described
above. About 3 h after ip injection, blood and brain samples
were collected. The blood samples were centrifuged to obtain the
plasma fraction. Brain samples were homogenized in saline to
obtain the brain homogenate. The plasma and brain homoge-
nate samples were deproteinized with an equal volume of
CH3CN. Compound concentrations in the supernatant were
measured using a Shimadzu HPLC system (Columbia, MD)
consisting of a SIL 20-AHT autosampler, SPD-20A UV/vis
detector, and LC-20AB pump connected to a Dgu-20A3 degas-
ser. Data acquisition was achieved by LC Solution software
version 1.22 SP1. Separations were achieved on an Agilent 5-μm
C18 column (150 mm ꢀ 4.6 mm id). Chromatographic condi-
tions for compound 24: mobile phase of CH3CN-H2O (75:25)
delivered at 1.0 mL/min, detection wavelength 305 nm. Chro-
matographic conditions for 28: CH3CN-H2O (40:60) delivered
at 1.0 mL/min, detection at 318 nm.
Acknowledgment. This study was supported in part by the
Louisiana Biomedical Research Network.

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 24 8545
(17) Gentles, R. G.; Hu, S.; Dubowchik, G. M. Recent advances in the
discovery of GSK-3 inhibitors and a perspective on their utility for
the treatment of Alzheimer’s disease. Annu. Rep. Med. Chem. 2009,
44, 3–26.
(18) Godemann, R.; Biernat, J.; Mandelkow, E.; Mandelkow, E. M.
Phosphorylation of tau protein by recombinant GSK-3beta: pro-
nounced phosphorylation at select Ser/Thr-Pro motifs but no
phosphorylation at Ser262 in the repeat domain. FEBS Lett.
1999, 454, 157–164.
(19) Gong, C. X.; Lidsky, T.; Wegiel, J.; Zuck, L.; Grundke-Iqbal, I.;
Iqbal, K. Phosphorylation of microtubule-associated protein tau is
regulated by protein phosphatase 2A in mammalian brain. Im-
plications for neurofibrillary degeneration in Alzheimer’s disease.
J. Biol. Chem. 2000, 275, 5535–5544.
(20) Iqbal, K.; Grundke-Iqbal, I. Inhibition of neurofibrillary degen-
eration: A promising approach to Alzheimer’s disease and other
tauopathies. Curr. Drug Targets 2004, 5, 495–502.
Pannacci, M.; Fraschini, F.; Stankov, B. M. Conformationally
restrained melatonin analogues: synthesis, binding affinity for the
melatonin receptor, evaluation of the biological activity, and
molecular modeling study. J. Med. Chem. 1997, 40, 1990–2002.
(37) Gadakar, P. K.; Phukan, S.; Dattatreya, P.; Balaji, V. N. Pose
prediction accuracy in docking studies and enrichment of actives in
the active site of GSK-3beta. J. Chem. Inf. Model. 2007, 47, 1446–5149.
(38) Polgar, T.; Baki, A.; Szendrei, G. I.; Keseru, G. M. Comparative
virtual and experimental high-throughput screening for glycogen
synthase kinase-3beta inhibitors. J. Med. Chem. 2005, 48, 7946–
7959.
(39) Bhat, R.; Xue, Y.; Berg, S.; Hellberg, S.; Ormoe, M.; Nilsson, Y.;
Radesaeter, A.; Jerning, E.; Markgren, P. O.; Borgegard, T.;
Nyloef, M.; Gimenez-Cassina, A.; Hernandez, F.; Lucas, J. J.;
Diaz-Nido, J.; Avila, J. Structural insights and biological effects of
glycogen synthase kinase 3-specific inhibitor AR-A014418. J. Biol.
Chem. 2003, 278, 45937–45945.
(21) Phiel, C. J.; Wilson, C. A.; Lee, V.M-Y.; Klein, P. S. GSK-3alpha
regulates production of Alzheimer’s disease amyloid-beta peptides.
Nature 2003, 423, 435–439.
(40) Meijer, L.; Skaltsounis, A. L.; Magiatis, P.; Polychronopoulos, P.;
Knockaert, M.; Leost, M.; Ryan, X. P.; Vonica, C. A.; Brivanlou,
A.; Dajani, R.; Crovace, C.; Tarricone, C.; Musacchio, A.; Roe,
S. M.; Pearl, L.; Greengard, P. GSK-3-Selective inhibitors derived
from tyrian purple indirubins. Chem. Biol. 2003, 10, 1255–1266.
(41) Jain, A. N. Surflex: fully automatic flexible molecular docking
using a molecular similarity-based search engine. J. Med. Chem.
2003, 46, 499–511.
(42) Tripos Associates. SYBYL Molecular Modeling Software, version
8.0; Tripos Associates: St. Louis, MO, 2007; http://www.tripos.com.
(43) Jones, G.; Willett, P.; Glen, R. C. Molecular recognition of
receptor sites using a genetic algorithm with a description of
desolvation. J. Mol. Biol. 1995, 245, 43–53.
(22) Takashima, A.; Noguchi, K.; Michel, G.; Mercken, M.; Hoshi, M.;
Ishiguro, K.; Imahori, K. Exposure of rat hippocampal neurons
to amyloid beta peptide (25-35) induces the inactivation of phos-
phatidyl inositol-3 kinase and the activation of tau protein kinase
I/glycogen synthase kinase-3 beta. Neurosci. Lett. 1996, 203, 33–36.
(23) Culbert, A, A.; Brown, M. J.; Frame, S.; Hagen, T.; Cross, D. A. E.;
Bax, B.; Reith, A. D. GSK-3 inhibition by adenoviral FRAT1
overexpression is neuroprotective and induces tau dephosphoryla-
tion and beta-catenin stabilization without elevation of glycogen
synthase activity. FEBS Lett. 2001, 507, 288–294.
(24) Alvarez,G.;Munoz-Montano, J. R.; Satrustegui, J.; Avila, J.; Bogonez,
E.; Diaz-Nido, J. Lithium protects cultured neurons against beta-
amyloid-induced neurodegeneration. FEBS Lett. 1999, 453, 260–264.
(25) Medina, M.; Castro, A. Glycogen synthase kinase-3 (GSK-3) 1045
inhibitors reach the clinic. Curr. Opin. Drug Discovery Dev. 2008,
11, 533–543.
(26) Bertrand, J. A.; Thieffine, S.; Vulpetti, A.; Cristiani, C.; Valsasina,
B.; Knapp, S.; Kalisz, H. M.; Flocco, M. Structural characteriza-
tion of the GSK-3beta active site using selective and non-selective
ATP-mimetic inhibitors. J. Mol. Biol. 2003, 333, 393–407.
(27) Taha, M. O.; Bustanji, Y.; Al-Ghussein, M. A. S.; Mohammad, M.;
Zalloum, H.; Al-Masri, I. M.; Atallah, N. Pharmacophore model-
ing, quantitative structure-activity relationship analysis, and
in silico screening reveal potent glycogen synthase kinase-3beta
inhibitory activities for cimetidine, hydroxychloroquine, and
gemifloxacin. J. Med. Chem. 2008, 51, 2062–2077.
(44) Shultz-Gasch, T.; Stahl, M. Binding site characteristics in structure-
based virtual screening: evaluation of current docking tools.
J. Mol. Model. 2003, 9, 47–57.
(45) Rodems, S. M.; Hamman, B. D.; Lin, C.; Zhao, J.; Shah, S.;
Heidary, D.; Makings, L.; Stack, J. H.; Pollok, B. A. A FRET-
based assay platform for ultra-high density drug screening of
protein kinases and phosphatases. Assay Drug Dev. Technol.
2002, 1, 9–19.
(46) Hamann, M. T.; Alonso, D.; Martı
´
n-Aparicio, E.; Fuertes, A.;
ꢀ
Perez-Puerto, J.; Castro, A.; Morales, S.; Navarro, M. L.; del
ꢀ
Monte-Millan, M.; Medina, M.; Pennaka, H.; Balaiah, A.; Peng,
J.; Cook, J.; Wahyuono, S.; Martınez, A. Glycogen synthase
´
kinase-3 (GSK-3) inhibitory activity and structure-activity rela-
tionship (SAR) studies of the manzamine alkaloids. Potential for
Alzheimer’s disease. J. Nat. Prod. 2007, 70, 1397–1405.
(47) Lawrie, A. M.; Noble, M. E.; Tunnah, P.; Brown, N. R.; Johnson,
L. N.; Endicott, J. A. Protein kinase inhibition by staurosporine
revealed in details of the molecular interaction with CDK2. Nature
Struct. Biol. 1997, 4, 796–801.
(28) Kim, H.-J.; Choo, H.; Cho, Y. S.; No, K. T. i.; Pae, A. N. Novel
GSK-3beta inhibitors from sequential virtual screening. Bioorg.
Med. Chem. 2008, 16, 636–643.
(29) Patel, D. S.; Bharatam, P. V. New leads for selective GSK-3
inhibition: pharmacophore mapping and virtual screening studies.
J. Comput.-Aided Mol. Des. 2006, 20, 55–66.
(30) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.
Experimental and computational approaches to estimate solubility
and permeability in drug discovery and development settings. Adv.
Drug Delivery Rev. 2001, 46, 3–26.
(48) Ravitz, M. J.; Wenner, C. E. Cyclin-dependent kinase regulation
during G1 phase and cell cycle regulation by TGF-beta. Adv.
Cancer Res. 1997, 71, 165–207.
(49) Lochhead, P. A.; Coghlan, M.; Rice, S. Q. J.; Sutherland, C.
Inhibition of GSK-3 selectively reduces glucose-6-phosphatase
and phosphatase and phosphoenol-ypyruvate carboxykinase gene
expression. Diabetes 2001, 50, 937–946.
(31) Chekler, E. L. P.; Elokdah, H. M.; Butera, J. Efficient one-pot
synthesis of substituted 2-amino-1,3,4-oxadiazoles. Tetrahedron
Lett. 2008, 49, 6709–6711.
(32) van der Waterbeemd, H. Quantitative approaches to structure-
activity relationships. In The Practice of Medicinal Chemistry;
Wermuth, C. G., Ed.; Academic Press: London, 1996; p 367.
(33) Martin, Y. C.; Bures, M. G.; Danaher, E. A.; DeLazzer, J.; Lico, I.;
Pavlik, P. A. A fast new approach to pharmacophore mapping and
its application to dopaminergic and benzodiazepine agonists.
J. Comput.-Aided Mol. Des. 1993, 7, 83–102.
(50) Saitoh, M.; Kunitomo, J.; Kimura, E.; Hayase, Y.; Kobayashi, H.;
Uchiyama, N.; Kawamoto, T.; Tanaka, T.; Mol, C. D.; Dougan,
D. R.; Textor, G. P.; Snell, G. P.; Itoh, F. Design, synthesis and
structure-activity relationships of 1,3,4-oxadiazole derivatives as
novel inhibitors of glycogen synthase kinase-3β. Bioorg. Med.
Chem. 2009, 17, 2017–2029.
(51) Saitoh, M.; Kunitomo, J.; Kimura, E.; Iwashita, H.; Uno, Y.;
Onishi, T.; Uchiyama, N.; Kawamoto, T.; Tanak, T.; Mol, C. D.;
Dougan, D. R.; Textor, G. P.; Snell, G. P.; Takizawa, M.; Itoh, F.;
Kori, M. 2-{3-[4-(Alkylsulfinyl)phenyl]-1-benzofuran-5-yl}-5-
methyl-1,3,4-oxadiazole derivatives as novel inhibitors of glycogen
synthase kinase-3β with good brain permeability. J. Med. Chem.
2009, 52, 6270–6286.
(34) Clark, R. D. OptiSim: an extended dissimilarity selection method
for finding diverse representative subsets. J. Chem. Inf. Comput.
Sci. 1997, 37, 1181–1188.
(35) Marriott, D. P.; Dougall, I. G.; Meghani, P.; Liu, Y.-J.; Flower,
D. R. Lead generation using pharmacophore mapping and three-
dimensional database searching: application to muscarinic M(3)
receptor antagonists. J. Med. Chem. 1999, 42, 3210–3216.
(36) Spadoni, G.; Balsamini, C.; Diamantini, G.; Di Giacomo, B.;
Tarzia, G.; Mor, M.; Plazzi, P. V.; Rivara, S.; Lucini, V.; Nonno, R.;
(52) Mylonas, S.; Mamalis, A. Synthesis and antitumor activity of new
thiosemicarbazones of 2-acetylimidazo[4,5-b]pyridine. J. Heterocycl.
Chem. 2005, 42, 1273–1281.
(53) Carroll, N. V.; Longley, R. W.; Roe, J. H. The determination of
glycogen in liver and muscle by use of anthrone reagent. J. Biol.
Chem. 1956, 220, 583–593.