Novel pyrazolopyrimidine derivatives as GSK-3 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2004.02.036

A series of [1-aryl-1H-pyrazolo[3,4-d]pyrimidin-4-yl]arylhydrazones were discovered as novel inhibitors glycogen synthase kinase-3 (GSK-3). Based on initial modeling a detailed SAR was constructed. Modification of the interior binding aryl ring (Ar₁) determined this to be a tight binding region with little room for modification. As predicted from the model, a large variety of modifications could be incorporated into the hydrazone aryl ring. This work led to GSK-3 inhibitors in the low nano-molar range.

Full text of DOI:10.1016/j.bmcl.2004.02.036

Bioorganic & Medicinal Chemistry Letters 14 (2004) 2121–2125
Novel pyrazolopyrimidine derivatives as GSK-3 inhibitors
Andrew J. Peat, Joyce A. Boucheron, Scott H. Dickerson, Dulce Garrido, Wendy Mills,
Jennifer Peckham, Frank Preugschat, Terrence Smalley, Stephanie L. Schweiker,
Jayme R. Wilson, Tony Y. Wang, Huiqiang Q. Zhou and Stephen A. Thomson*
GlaxoSmithKline Research and Development, 5 Moore Drive, Research Triangle Park, NC 27709, USA
Received 8 January 2004; revised 6 February 2004; accepted 9 February 2004
Abstract—A series of [1-aryl-1H-pyrazolo[3,4-d]pyrimidin-4-yl]arylhydrazones were discovered as novel inhibitors glycogen syn-
thase kinase-3 (GSK-3). Based on initial modelinga detailed SAR was constructed. Modification of the interior bindingaryl ring
(Ar1) determined this to be a tight binding region with little room for modification. As predicted from the model, a large variety of
modifications could be incorporated into the hydrazone aryl ring. This work led to GSK-3 inhibitors in the low nano-molar range.
Ó 2004 Elsevier Ltd. All rights reserved.
Glycogen synthase kinase-3 (GSK-3) is a serine–threo-
nine kinase ubiquitously expressed in all tissues and was
originally identified as one of five protein kinases that
phosphorylate glycogen synthase.1 The rate-limiting
step for glycogen formation is determined by the enzyme
glycogen synthase (GS) and GS activity is modulated by
GSK-3 via phosphorylation.2 Signaling via the PI-3
kinase/PKB pathway, insulin inhibits GSK-3, resulting
in GS activation and subsequent formation of glyco-
gen.3 Recent analysis has shown GSK-3 activity and
expression levels were significantly higher in the skeletal
muscle from Type 2 diabetic humans than those in
healthy subjects.4 Suggesting aberrant GSK-3 activity
may contribute to diabetes. Therefore, small molecules
that mimic the action of insulin by inhibitingGSK-3
represents a new approach for the treatment of Type 2
diabetes. GSK-3 is also known to phosphorylate the
microtubule associated protein tau,5 the hyperphosph-
orylation of which is an early event in neurodegenerative
conditions such as AlzheimerÕs disease (AD). Inhibition
of GSK-3 with potent selective small molecules has been
shown to protect primary neurons from death induced
by reduced PI-3 kinase activity.6 Therefore inhibitors of
GSK-3 may have utility in the treatment of AD.
of glycogen synthesis in vitro as well as lowering of
plasma glucose in diabetic animals by GSK-3 inhibitors
has been reported.8 We recently identified a novel class
of pyrazolopyrimidine derivatives from
a
high
throughput screen that inhibit GSK-3 (Fig. 1). The
original screening hit 1 had a pIC50 ¼ 7.0 for inhibition
of GSK-3 and was determined to bind in a competitive
manner with ATP.9 Since this template represented a
novel GSK-3 inhibitor, we proceeded to investigate this
series of compounds. Herein we report on the binding
mode and SAR of these GSK-3 inhibitors.
Our initial SAR around this template is shown in Table
1. Replacement of H at R1 with chloro or phenyl (2, 3)
results in a complete loss of activity. Likewise replace-
ment of H at R2 with methyl results in a complete loss of
activity. Replacement of the phenyl at R3 with hydrogen
R3
R1
N
N
N
N
N
N
N
N
N
N
H
R2
N
N
Recently, a number of publications have emerged
7
describingmolecules that inhibit GSK-3. Stimulation
R4
OMe
OH
Keywords: Glycogen synthase kinase-3.
* Correspondingauthor. Tel.: +1-919-483-6287; fax: +1-919-483-6053;
e-mail: stephen.a.thomson@gsk.com
1
Template
Figure 1. Pyrazolopyrimidine template.
0960-894X/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2004.02.036

2122
A. J. Peat et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2121–2125
Table 1. Inhibition of GSK-3beta
a
Compound
R1
R2
R3
R4
pIC50
1
2
3
4
5
6
7
8
H
Cl
Ph
H
H
H
H
H
H
H
H
Me
H
H
H
H
Ph
Ph
Ph
Ph
H
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–Ph–(3-OMe, 4-OH)
–4-pyridyl
7.0
<5
<5
<5
5.5
5.6
<5
i-Pr
Bn
Ph
7.0
a Logof concentration, which inhibits 50% of the activity of GSK-3.
or isopropyl (5, 6) results in weakly active compounds
where a benzyl (7) at this position results in a complete
loss of activity. Lastly replacement of the 3-methoxy-4-
hydroxyphenyl group at R4 with a 4-pyridyl (8) group
gives a compound of equal activity.
Based on the result with compounds 2, 3, and 4, we
reasoned that the hydrazone hydrogen and the proximal
pyrimidine nitrogen were hydrogen bonding to the hinge
region of the ATP binding pocket. In silico docking10
gave the structure shown in Figure 2, in which the
hydrazone hydrogen interacts with the carbonyl of
valine 135 and the pyrimidine nitrogen interacts with the
NH of valine 135. Although not a hydrogen bond in the
formal sense, a potential third interaction could be
occurringbetween the C2 hydrogen of the pyrimidine
ringand the backbone carbonyl of the aspartate 133
residue. This model projects the R4 group toward the
solvent front and the R3 group deeper into the binding
pocket. The loss of activity with compounds 5, 6, and 7
suggest that the binding of the R3 phenyl ring is tight.
With this bindingmode of the template in mind, our
approach was to set R1 and R2 as hydrogen and explore
the SAR of the R3 (Ar1) and R4 (Ar2) positions.
Figure 2. Bindingof pyrazolopyrimidines to GSK-3.
Ar1
N
Ar1
N
N
H2N
NC
N
i
ii, iii
N
Ar1 NHNH2
N
III
I
II
Cl
Scheme 1 outlines the general synthesis of the template.
The Ar1 group was set at the start of the synthesis, with
the startingaryl-hydrazines ( I), which upon treatment
with ethoxymethylenemalonitrile in refluxingethanol
gave the desired 1-aryl-4-cyano-5-aminopyrazoles (II) in
good yields. Condensation with aqueous formic acid
and subsequent treatment with POCl3 gave the chloro-
pyrimidine core (III). Displacement of the chloride with
hydrazine and final condensation with Ar2 aldehydes
gave the desired products (IV).11 NMR and X-ray
analysis revealed that only the E hydrazone isomer was
formed.
Ar1
N
N
N
N
iv, v
N
H
IV
N
Ar2
Scheme 1. Reagents and conditions: (i) ethoxymethylenemalonitrile,
EtOH reflux, 4 h; (ii) 88% formic acid, reflux, 24 h; (iii) POCl3, DMF
(cat.) reflux 4 h; (iv) NH2NH2–H20, EtOH, 45 °C 24 h; (v) Ar2CHO,
EtOH, pyrrolidine (cat.) reflux 5 h.
Initially we set Ar2 as 4-pyridyl and examined the effect
of substitution on the phenyl ringat Ar (Table 2). The
meta-Me analog( 10), led to a 10-fold improvement in
potency (pIC50 ¼ 8.2) as compared to the unsubstituted
phenyl derivative 8 (pIC50 ¼ 7.0). Therefore we exam-
ined other substituents at the meta-position (compounds
15–23). Since the meta-OMe analog 13 had improved
potency over the unsubstituted analog 8, we first looked
at other alkoxy derivatives. Surprisingly, the addition of
the extra methylene in the meta-OEt group (15) signifi-
cantly decreased potency relative to the meta-OMe
(pIC50 ¼ 7.0 vs 8.2, respectively). Replacingthe meta-
OMe with the electron-withdrawing–OCF 3 (16) also led
to a dramatic drop in potency. Similarly, both the
1
addition of ortho-substituents, –Me or –OMe (9 and 12,
respectively), led to a decrease in activity as compared to
the unsubstituted analog( 8). ortho-Substitution in Ar1
with –Me or –OMe likely results in twistingof the aryl
ringout-of-plane with the pyrazolopyrimidine core,
suggesting that the preferred binding configuration is
co-planar. In addition, there appears to be little room at
the back of the bindingpocket since para-substitution of
Ar1 (11 and 14) dramatically weakened potency. Inter-
estingly, the meta-OMe substituent (13), but not the

A. J. Peat et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2121–2125
2123
Table 2. Inhibition of GSK-3beta
Table 3. Inhibition of GSK-3beta
Ar1
Ar1
N
N
N
N
N
N
N
N
N
H
N
H
N
N
N
N
a
a
Compound
Ar1
pIC50
Compound
Ar1
pIC50
5.9
Initial SAR
8
Substituted phenyl
H
o-Me
7.0
6.8
24
9
N
10
11
12
13
14
m-Me
p-Me
6.8
<4.5
5.6
o-OMe
m-OMe
p-OMe
N
25
26
7.5
8.5
8.2
6.0
m-Substitution Substituted phenyl
N
15
16
17
18
19
20
21
22
23
m-OEt
m-OCF3
m-NHnPr
7.0
6.5
5.4
5.7
m-NH(CH2)cyclopropyl
m-NHAc
OMe
6.8
6.2
27
28
7.4
7.4
m-NH(CO)nPr
m-F
m-Br
N
S
6.5
6.6
m-NO2
<4.5
N
a Logof concentration, which inhibits 50% of the activity of GSK-3.
a Logof concentration, which inhibits 50% of the activity of GSK-3.
electron-donatingamine analogs ( 17, 18) and the elec-
tron-withdrawingamides ( 19, 20) substantially lost
potency relative to compound 13. Substitution at the
meta-position with the relatively small but electron-
methoxyphenyl containingcompound ( 13). We rea-
soned that the lone pair of the 2-pyridyl is repulsed by
either of the neighboring nitrogen lone pairs of the
pyrazolopyrimidine core, thus disfavoringa co-planar
arrangement of these ring systems. The 3-pyridyl (25) is
modestly more potent then the phenyl, where the 4-
pyridyl (26) is substantially more potent. The reason for
the increase in potency of the 4-pyridyl is not clear, but
as stated above for the meta-methoxyphenyl group, an
interaction with the electropositive lysine 85 side chain
could play a role.
withdrawingsubstituents –F or –NO , analogs 21 and
2
23, respectively, also weakened potency as compared to
compound 13. From this data set, it appears that elec-
tron donating, sterically small substituents are favored
at the meta-position. For example, meta-OMe is optimal
for potency but not meta-OEt or meta-N(H)R, which
are perhaps too sterically large. In addition, the meta-
Me or meta-F substituents (examples 10 and 21,
respectively), are disfavored since they are weakly elec-
tron donatingcompared to methoxy. SAR data within
this series support that Ar1 is bindingin a narrow pocket
of the enzyme, in which co-planarity of the phenyl ring
with the pyrazolopyrimidine is optimal. In addition,
closer examination of the model suggests that the elec-
tronegative oxygen atom of the meta-methoxy group
may interact with the electropositive side chain of the
lysine 85 residue. This modeling does not suggest a
formal hydrogen bond, but the relative proximity would
support an electrostatic interaction.
Next we turned to examiningthe effect of substituents
on the hydrazone aryl ring(Ar ). As stated above, the
2
proposed bindingmodel orientates the Ar
group
2
toward the solvent front of the enzyme. We therefore
predicted that a large variety of groups would be
allowed at Ar2. Since the Ar2 group is installed in the
last step of the synthesis we were able to prepare a large
number of compounds. The meta-methoxyphenyl was
retained as the Ar1 substituent, and the various Ar2
groups prepared are shown in Table 4. As expected a
large variety of groups were tolerated at Ar2. Replace-
ment of the 4-pyridyl of compound 13 with an unsub-
stituted phenyl ring( 29) did not affect the activity,
suggesting that the pyridyl nitrogen does not make any
significant contacts with the enzyme. It is of interest to
examine what decreased activity in Table 4. Addition of
an ortho substituent (30, 33) resulted in a decrease in
With Ar2 still set as the 4-pyridyl, we examined the
8
effects of replacingthe phenyl group of compound
with heterocycles (Table 3). Replacement of the phenyl
with a 2-pyridyl (24) group results in a large loss of
activity, possibly due to a disfavoringof co-planarity of
the pyridyl ringwith the pyrazolopyrimidine core.
Addition of a 6-OMe substituent to the 2-pyridyl ring
restores activity (27), but is still less active than the meta-

2124
A. J. Peat et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2121–2125
Table 4. Inhibition of GSK-3beta
Table 5. Inhibition of GSK-3beta
OMe
OMe
N
N
N
N
N
N
N
N
H
R1
N
N
H
N
N
R2
Ar2
N
a
Compound
Ar2
pIC50
a
Compound
R1
R2
pIC50
Initial SAR
Substituted phenyl
H
29
30
31
32
33
34
35
36
37
38
39
40
8.0
<4.6
7.6
7.5
6.9
8.1
8.1
7.8
8.0
8.6
8.4
8.6
13
58
59
60
61
62
63
H
H
8.2
8.0
7.9
8.0
7.5
7.2
5.2
o-OMe
m-OMe
p-OMe
o-F
Me
Et
H
H
N-Pr
i-Pr
H
H
H
m-F
p-F
Me
Me
Me
3,4-diF
m-NH2
p-NH2
a Logof concentration, which inhibits 50% of the activity of GSK-3.
m-SO2Me
p-SO2Me
We next looked at the effect of substituents at the 3-
position of the pyrazole ring, and at the hydrazone
carbon, as shown in Table 5. Incorporation of small n-
alkyl group in the 3-position of the pyrazole ring (R1)
has little effect on the activity of the compounds.13 With
R1 as iso-propyl a slight decrease in activity was
observed. From the model, groups at this position
would project into an empty region of the enzyme-
bindingpocket. Due to the branchingof the iso-propyl
group it is possible that an unfavorable steric interaction
arises with the hydrazone group. Substitution at R2 with
a methyl group results in a significant, but not complete
loss of activity (62). It is interestingthat compound 63 in
which a methyl group has been incorporated at both R1
and R2 is nearly devoid of activity.
p-Alkoxy
Substituted phenyl
p-OH
p-O-allyl
41
42
43
44
7.9
6.8
5.6
8.2
p-O(4-F)Bn
p-O(CH2)2NMe2
p-Amines
Substituted phenyl
p-NEt2
p-CH2NMe2
45
46
47
48
49
50
5.6
8.3
8.4
7.9
7.8
8.3
p-CH2NHEt
p-NHAc
p-NHSO2Me
p-NH(CO)(CH2)2NMe2
p-Carbon
Substituted phenyl
p-Me
p-Ph
51
52
53
54
55
56
57
<4.5
<4.5
7.8
In conclusion, we have described the bindingmode and
SAR of a novel class of GSK-3 inhibitors. The binding
mode, initially proposed by in silico modeling, was
supported by the observed SAR. Within the series the
Ar1 group was shown to be highly sensitive to substitu-
tion, with the meta-methoxyphenyl and 4-pyridyl groups
optimal for potency. As predicted by the model, the Ar2
group is much more tolerant of substitution, with a large
variety of groups allowed. Incorporation of small alkyl
groups at the 3-position of the pyrazole ring was also
allowed. The ability of these compounds to effect gly-
cogen synthesis and their in vivo antidiabetic properties
will be described in subsequent publications.
p-CN
p-CO2H
7.5
7.8
p-C(O)NH(CH2)2NMe2
p-C(O)NH(CH2)2SO2Me
p-SO2NH(CH2)2NMe2
8.1
8.0
a Logof concentration, which inhibits 50% of the activity of GSK-3.
activity, where meta and para substitution is generally
allowed. Nonpolar substituents in the para-position can
lead to a large decrease in activity, such as with 4-Me, 4-
phenyl, 4-allyloxy, and 4-(4-fluorobenzyloxy) (51, 52,
42, and 43, respectively). It is likely that these larger
hydrophobic residues are projectinginto the aqueous
environment outside the bindingpocket. Although, at
best, modest improvements in potency were realized
with changes at Ar2 (38, 40), other parameters of the
molecules could be effected. Improvements in solubility
and cellular efficacy were realized with some of the
modifications. For example, in a cell based glycogen
accumulation assay, in which the effect of the compound
on 14C-glucose incorporation into glycogen was mea-
sured, compound 13 was 10-fold more potent that
compound 29, even though their activity at the enzyme
was very similar.12
References and notes
1. (a) Cohen, P. In The Enzymes; Academic: New York,
1986; Vol. XVII, pp 461–497; (b) Woodgett, J. R.; Cohen,
P. Biochim. Biophys. Acta 1984, 788, 339.
2. For recent reviews, see: (a) Eldar-Finkelman, H.; Ilouz, R.
Expert Opin. Invest. Drugs 2003, 12, 1511; (b) Eldar-
Finkelman, H. Trends Mol. Med. 2002, 8, 126.

A. J. Peat et al. / Bioorg. Med. Chem. Lett. 14 (2004) 2121–2125
2125
3. (a) Grimes, C. A.; Jope, R. S. Prog. Neurobiol. 2001, 65,
391; (b) Frame, S.; Cohen, P. Biochem. J. 2001, 359, 1; (c)
Cross, D. A.; Alessi, D. R.; Vandenheede, J. R.; McDo-
well, H. E.; Hundal, H. S.; Cohen, P. Biochem. J. 1994,
303, 21.
4. Nikoulina, S. E.; Ciaraldi, T. P.; Mudailar, S.; Mohideen,
P.; Carter, L.; Henry, R. R. Diabetes 2000, 49, 263.
5. Lovestone, S.; Reynolds, C. H.; Latimer, D.; Davis, D. R.;
Anderton, B. H.; Gallo, J. M.; Hanger, D.; Mulot, S.;
Marquardt, B. Curr. Biol. 1994, 4, 1077.
6. Cross, D. A. E.; Culbert, A. A.; Chalmers, K. A.; Facci,
L.; Skaper, S. D.; Reith, A. D. J. Neurochem. 2001, 77, 94.
7. (a) Witherington, J.; Bordas, V.; Garland, S. L.; Hickey,
D. M. B.; Ife, R. J.; Liddle, J.; Saunders, M.; Smith, D. G.;
Ward, R. W. Bioorg. Med. Chem. Lett. 2003, 13, 1577; (b)
Olesen, P. H.; Sorensen, A. R.; Urso, B.; Kurtzhals, P.;
Bowler, A. N.; Ehrbar, U.; Hansen, B. F. J. Med. Chem.
2003, 46, 3333; (c) Kuo, G.-H.; Prouty, C.; DeAngelis, A.;
Shen, L.; OÕNeill, D. J.; Shah, C.; Connolly, P. J.; Murray,
W. V.; Conway, B. R.; Cheung, P.; Westover, L.; Xu, J.
Z.; Look, R. A.; Demarest, K. T.; Emanuel, S.; Middle-
ton, S. A.; Jolliffe, L.; Beavers, M. P.; Chen, X. J. Med.
Chem. 2003, 46, 4021; (d) Martinez, A.; Castro, A.;
Dorronsoro, I.; Alonso, M. Med. Res. Rev. 2002, 22, 373.
8. (a) Henriksen, E. J.; Kinnick, T. R.; Teachey, M. K.;
OÕKeefe, M. P.; Ring, D.; Johnson, K. W.; Harrison, S. D.
Am. J. Physiol. Endocrinol. Metab. 2003, 284, E892; (b)
Coghlan, M. P.; Culbert, A. A.; Cross, D. A. E.;
Corcoran, S. L.; Yates, J. W.; Pearce, N. J.; Rausch, O.
L.; Murphy, G. J.; Carter, P. S.; Cox, L. R.; Mills, D.;
Brown, M. J.; Haigh, D.; Ward, R. W.; Smith, D. G.;
Murray, K. J.; Reith, A. D.; Holder, J. C. Chem. Biol.
2000, 7, 793–803.
peptide substrate (Biotin––Ahx-AAAKRREILSRRP-
S(PO3)YR-amide), 2.5 lM ATP and 0.2 lCi/well c33P.-
ATP. After 40 min, the reaction was stopped by addition
of 100 mM EDTA and 1 mM solution in 100 mM HEPES
followed by a solution of diluted streptavidin coated SPA
beads (Amersham) in PBS to give a final concentration of
0.25 mgof beads per assay well. The wells were counted on
a Packard TopCount NXT microplate counter. The
standard GSK-3 inhibitor SB-415286 gave similar IC50
values in this assay as compared to those reported in the
literature, see: Smith, D. G.; Buffet, M.; Fenwick, A. E.;
Haigh, D.; Ife, R. J.; Saunders, M.; Slingsby, B. P.; Stacey,
R.; Ward, R. W. Bioorg. Med. Chem. Lett. 2001, 11, 635.
10. The molecular modelingwas performed with the Maestro
software package from Schrodinger. The ligand was
manually docked into the active site and then the complex
was energy minimized using the Merck molecular force
field. Structural co-ordinates for the GSK-3 structure was
taken from: Bax, B.; Carter, P. S.; Lewis, C.; Guy, A. R.;
Bridges, A.; Tanner, R.; Pettman, G.; Mannix, C.;
Culbert, A. A.; Brown, M. J. B.; Smith, D. G.; Reith, A.
D. Structure 2001, 9, 1143.
11. All compounds were determined to be >95% pure by both
1H NMR and LC/MS spectroscopy; the spectral data was
consistent with the reported structure. Aryl-hydrazines
and aryl-aldehydes were either commercially available or
were prepared by reported methods.
12. Compound treatment effects on glycogen accumulation in
differentiated L6 cells were determined by an adaptation of
the method of Thomas, J. A.; Schlender, K. K.; Larner
J. Anal. Biochem. 1968, 25, 486–499.
13. Compounds 58–63 were prepared in a similar manner as
shown in Scheme 1. Ethoxymethylenemalonitrile was
replaced with the required alkyl derivative, and for
compounds 62 and 63 4-acetylpyridine was used in the
final condensation step. NMR and LC/MS spectroscopy
were consistent with the reported structure.
9. GSK-3b was assayed in 96-well microtitre plates at a final
concentration of 20 nM in 100 mL HEPES at pH 7.2
containing 10 mM MgCl, 0.1 mg/mL bovine serum albu-
min, 1 mM dithiothreitol, 0.3 mg/mL heparin, 2.8 lM