Rational design of potent GSK3β inhibitors with selectivity for Cdk1 and Cdk2

Article abstract of DOI:10.1016/j.bmcl.2010.01.114

From an HTS hit, a series of potent and selective inhibitors of GSK3β have been designed based on a Cdk2-homology model and with the help of several crystal structures of the compounds within Cdk2.

Full text of DOI:10.1016/j.bmcl.2010.01.114

Bioorganic & Medicinal Chemistry Letters 20 (2010) 1985–1989
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Rational design of potent GSK3b inhibitors with selectivity for Cdk1 and Cdk2
Dominique Lesuisse a,e, , Gilles Dutruc-Rosset a,e, Gilles Tiraboschi c,e, Matthias K. Dreyer f,
*
Sébastien Maignan d,e, Alain Chevalier a,e, Frank Halley a,e, Philippe Bertrand b,e, Marie-Claude Burgevin b,e
Dominique Quarteronet b,e, Thomas Rooney b,e
,
a Medicinal Chemistry, 13 Quai Jules Guesde, 94300 Vitry-sur-Seine, France
b CNS Department, 13 Quai Jules Guesde, 94300 Vitry-sur-Seine, France
c CAS Molecular Modelling, 13 Quai Jules Guesde, 94300 Vitry-sur-Seine, France
d CAS Structural Biology, 13 Quai Jules Guesde, 94300 Vitry-sur-Seine, France
e Sanofi-aventis, 13 Quai Jules Guesde, 94300 Vitry-sur-Seine, France
f Sanofi-aventis, Industriepark Hoechst, 65926 Frankfurt am Main, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
From an HTS hit, a series of potent and selective inhibitors of GSK3b have been designed based on a Cdk2-
homology model and with the help of several crystal structures of the compounds within Cdk2.
Ó 2010 Elsevier Ltd. All rights reserved.
Received 27 November 2009
Revised 18 January 2010
Accepted 20 January 2010
Available online 25 January 2010
Keywords:
Glycogen synthase kinase
GSK3b
Aminoindazole
Cdk’s
Glycogen synthase 3b (GSK3b) is one of the two major tau phos-
phorylating enzymes.1 As tau hyperphosphorylation has been
linked to both cytoskeletal dysfunction and the formation of neu-
rofibrillary tangles, one of the major hallmarks of Alzheimer’s dis-
ease,2 inhibiting these events could potentially be beneficial in the
treatment of this chronic neurodegenerative disease. In the course
of a programme aimed at identifying inhibitors of this enzyme, we
initiated a high throughput screening approach. This led to the hit
monomeric Cdk2 (see Supplementary Fig. S1) and the results were
transferred into the GSK3b homology model, clearly suggesting
that the aminoindazole moiety interacted with the hinge area of
the kinase, while the carboxylic acid chain was pointing towards
the solvent area (Fig. 1). The model also showed that two important
areas of the protein were poorly or not occupied, namely the ribose
and the hydrophobic pockets, suggesting that occupying these
pockets would have the potential to greatly enhance the affinity
of this series for the kinase.
1 with an IC50 in the micromolar range (1.18 lM). The object of the
Our medicinal chemistry programme can be summarized as fol-
lows: introduce substitutions in the 5- and 6-position of the amin-
oindazole moiety to improve affinity and replace the carboxylic
acid by a chain more amenable to brain penetration (Fig. 2).
As the carboxylic acid chain sits in the solvent area, it was
hypothesized that it contributed little to the activity. Indeed re-
moval of the carboxylic acid moiety, as in 2, resulted in better inhib-
present paper is to describe how we have optimized this hit into
potent and selective inhibitors.
O
H
N
OH
N
O
N
Cl
H
itor activity (0.35
lM) than the initial hit 1 (1.18 lM). This gain of
1
IC50 = 1.18 µM
O
At the time this work was initiated, no structural data on GSK3b
was available. Based on the 28% identity (45% homology) of GSK3b
with Cdk2, we built a homology model from the reported structure
of Cdk2.3 Compound 1 was therefore crystallized with human
H
N
N
N
Cl
H
2
* Corresponding author. Tel.: +33 1 58933771; fax: +33 1 58933450.
E-mail address: dominique.lesuisse@sanofi-aventis.com (D. Lesuisse).
IC50 = 0.35 µM
0960-894X/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2010.01.114

1986
D. Lesuisse et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1985–1989
obtained by soaking 4 with the Aurora2 kinase showed that this
interaction was not observed (Fig. 3). Nevertheless, since the resi-
dues around the DFG motif are displaced by packing interactions,
we cannot exclude the possibility that this results from the crystal
arrangement. On the other hand, 4 did not show any significant
inhibition of Aurora2 (IC50 >30 lM) so the better activity on GSK3b
could be partly explained by this extra H-bond and stabilization of
the ring orientation.
Since, as we had predicted, potentially good inhibitory activity
could be gained from the introduction of groups towards the ribose
pocket, we decided to keep the chlorine atom of 2 in position 6,
while investigating position 5. This turned out to be quite profit-
able, as some of the disubstituted analogs displayed up to 100-fold
improvement in activity, compared to the 6-chloro derivative 2. A
non-substituted phenyl was the best 5-substituent; substitution
on the phenyl ring did not improve activity (Table 2). Compound
37 was chosen for further optimization.
A classical approach in medicinal chemistry is to try and com-
bine the best substitutions in the same compound. However, com-
bining the 4-phenol in position 6 with the best substituents of
position 5 was detrimental to the activity (Table 3). When aryl
groups were introduced at both positions 5 and 6 the level of
Figure 1. Docking model of 1 into GSK3b.
R3
H
N
R5
N
R6
N
H
R7
Figure 2.
activity is probably due to less desolvation energy upon binding. We
therefore selected 2 as a starting point for further optimization.
We first investigated the hydrophobic pocket. To this end, the
chlorine atom in the 6-position was replaced by several groups.
While the majority of substitutions were detrimental to the activ-
ity, introduction of a phenolic or anilinic group in the 6-position
(compounds 3–7) increased potency by up to 10-fold (Table 1).
We hypothesized that this improvement was linked to a favor-
able interaction with the catalytic lysine in the bottom of the
hydrophobic pocket. Many attempts to obtain a co-crystal struc-
ture of 4 with Cdk2 were not successful. However, a co-structure
Figure 3. Crystals of Aurora2 kinase domain (125–399CterHistag) with 4 (2.2 Å).
Red dashes represent hydrogen bonds to the hinge region of the kinase. The
distance between the phenolic OH and the catalytic lysine nitrogen is 4.8 Å.5
Table 1
GSK3b activity (IC50
4
)
of compounds modified in 6-position (Fig. 2)a
Compd#
R6
GSK3b (
l
M)
Compd#
R6
GSK3b (lM)
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Cl
0.354
0.027
0.044
0.055
0.076
0.094
0.195
0.198
0.260
0.661
0.935
1.088
1.184
1.281
1.353
2.459
2.894
3.808
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
C6H5CH2–
4-F–C6H4–
4.140
5.008
5.413
5.498
7.165
8.525
10.546
10.892
12.247
16.784
17.103
20.532
25.038
>100
3,4-(OH)2–C6H3–
4-OH–C6H4–
4-OH-2-Cl–C6H3–
4-NH2–C6H4–
3-OH–C6H4–
CF3–
3,5-Cl2–C6H3–
4-F3CO–C6H4–
4-MeS–C6H4–
C6H5(CH2)2–
4-CF3–C6H4–
3-Py
4-Me2N–C6H4–
2-OH-5-Pyridinyl
4-Me–C6H4–
4-CN–C6H4–
3,4-Dioxolano-C6H4–
4-NO2–C6H4–
4-MeO–C6H4–
4-Et–C6H4–
Br–
3-Furyl
4-Pyridinyl
3-Thienyl
2-NH2-5-pyridinyl
4-OBn–C6H4–
C6H5–
2-Cl–C6H4–
(E)-CH3CH@CH–
3,5-F2–C6H3–
4-tBu–C6H4–
>100
>100
>100
4-Cl–C6H4–
a
R3 = C(O)n-Pr, R5 = H.

D. Lesuisse et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1985–1989
1987
Table 2
GSK3b activity (IC50) of compounds modified in 5-position Fig. 2)a
Table 4
Selectivity of 37 versus a panel of kinases
Compd# R5 GSK3b (nM) Compd# R5
GSK3b (nM)
4-Me–C6H4– 14
4-NO2–C6H4– 18
4-F–C6H4– 22
4-BnO–C6H4– 23
Kinase
% Inhibition of Kinase
37 at 20
% Inhibition of
l
M
37 at 20
lM
2
37
38
39
40
41
H
C6H5–
4-NH2–C6H4–
4-OH–C6H4–
3-Furyl–
4-Et–C6H4–
354
7
8
10
11
12
42
43
44
45
46
47
EGF-tyrosine À24
kinase (h)
MAP kinase (ERK 42)
13
CAM kinase II À10
MEK1 kinase
13
14
4-Pyridinyl
Br–
50
55
Plk1
Akt
À6
À2
À2
Protein kinase C-a (h)
Protein kinase A (stimulated) 21
Aurora
Protein
kinase
p56lck
34
R3 = C(O)n-Pr, R6 = Cl.
a
Src kinase (h) À2
Casein kinase II (h)
ZAP70 kinase (h)
Jnk3
cdk4
cdk1
Cdk2
38
45
58
72
95
93
Fak
1
2
3
5
5
7
Table 3
Pak3
Abl
IGF1R
p38
GSK3b activity of compounds combining the best substitutions at 5 and 6-positions
(Fig. 2)a
Compd#
R5
R6
GSK3b (nM)
Tie2
48
49
50
51
52
53
54
55
56
2-Furyl–
3-Pyridinyl–
4-Et–C6H4–
3-Furyl–
4-OH–C6H4–
4-OBn–C6H4–
C6H5–
4-OH–C6H4–
4-OH–C6H4–
4-OH–C6H4–
4-OH–C6H4–
4-OH–C6H4–
4-OBn–C6H4–
4-OH–C6H4–
C6H5–
14
149
49
98
56
>10,000
55
1720
5
C6H5–
Br–
4-OH–C6H4–
a
R3 = C(O)n-Pr.
inhibitory activity decreased to that of one of the monosubstituted
aryl derivatives in Table 1. The only combinations that proved suc-
cessful were aryl and halogen at position 5 or 6, such as in 59 or
compounds 37–41 from Table 2.
The superposition of the crystal structures of compounds 37
and 57 in Cdk2 shows that the bulky aromatic substituent intro-
duces some small, but significant, changes (Fig. 4) that might ac-
count for the lower affinity of 57. The plane of the indazole ring,
Figure 5. Superimposition of the crystal structure of 60 (orange) in Cdk2 (cyan)
with GSK3b (grey; protein chain taken from PDB 1Q4L). Nonpolar interactions
between the 7-fluorine and Cdk2 gate keeper Phe80 are indicated by cyan dashes.
Corresponding distances to GSK3b residue Leu132 are shown in grey. Hydrogen
bonds to the backbone atoms of the hinge segment are depicted with red dashes.
including the butyramide, is inclined by about 12° with respect
to the inhibitor position of 37. In addition, the conformation of
the phenyl ring in the 5-position of 57 is rotated by 40° with re-
spect to 37 due to intramolecular repulsion between R5 and R6.
The orientation of the 6-phenol on this part is determined by the
protein environment, in particular the main chain at Asp145 and
the gate keeper residue Phe80.
In related crystal structures with smaller substituents (not
shown) both the orientation of the aromatic plane along the hinge,
as well as the orientation of an aryl moiety in the 5-position,
resemble those seen in 37 and suggest that this is a preferred con-
formation within the binding pocket.
Table 5
Effect of
a halogen in position 7 on the selectivity of 5,6-substituted 3-
aminoindazolesa
R3
R3
H
H
N
N
R3
H
N
Ph
Cl
Ph
Cl
Ph
N
N
N
N
N
H
H
N
Cl
H
F
Cl
Compd#
37
57
58
Figure 4. (Top), Superposition of 57 (yellow) and 37 (green) in Cdk2. The hinge
segment is behind the inhibitors. For clarity, side chains are only shown for the gate
keeper residue Phe80 and for Asp145. (Bottom), Superposition of the Cdk2:57
complex (yellow) with GSK3b (grey, protein chain taken from PDB 1Q4L), showing
the potential interaction of the phenolic hydroxy group with Glu97GSK3b (blue
dashes).5,6
GSK3b (nV)
Cdk1 (nV)
Cdk2 (nV)
7
100
2500
6
2500
3160
12
—
>10,000
R3 = C(O)n-Pr.
a

1988
D. Lesuisse et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1985–1989
Pr
Pr
NH
O
O
NH
N
X,Y = Br,H
(1) (2) (3)
(4)or (6) then (5)
N
N
N
N
H
X
F
X
(Het)Ar
SEM
Y
Y
C
A
B
(Het)Ar = 4-BnOPh
(Y =H)
(7) then (4) (5)
(7)
Pr
NH
N
Pr
O
O
X,Y = Cl,H
NH
(Het)Ar
X
(4)or (6) then (5)
Br
N
N
N
X,Y = F,F
(4) (5)
X
H
Y
SEM
Y
D
Scheme 1. Reagents and conditions: (1) NH2NH2ÁH2O; EtOH; X, Y = Br, H (90%), Cl, H (93%), F, Cl (26%), F, F (42%); (2) nPrCOCl/Py (80%-Q); (3) SEMCl/NaH/DMF/RT (31–61%);
(4) (Het)ArB(OH)2/Pd[P(Ph)3]4/Na2CO3/dioxane/H2O; (5) Bu4NF/THF/reflux; (6) Pd(dba)2/PPh3/bis(pinacolato)diborane/AcOK/dioxane/H2O; (Het)ArBr/Pd[P(Ph)3]4/Na2CO3/
dioxane/H2O; (7) Br2/CHCl3/Py X,Y = Cl,H (75%), F,F (36%).
Once again, no hydrogen bond between the hydroxyl from the
phenol and the protein could be observed, but this is likely to be
different in GSK3b where the conserved Glu97GSK3b carboxylate is
near the position of Leu148CDK2 of our surrogate and may form a
hydrogen bond with the inhibitor OH group.
With this SAR, and taking into account the generally poor PK
properties of phenols resulting from phase II metabolism and elim-
ination and the propensity of anilines to give Ames positive results,
we chose to keep 37 for further optimization. An important aspect
of this programme is the selectivity of the compounds for other ki-
nases. Pan kinase inhibitors, even if tolerated in cancer treatment,
are probably not suitable for the prolonged treatment of chronic
diseases, such as Alzheimer’s disease. 37 displayed good selectivity
versus a panel of kinases7 belonging to the 4 major kinase classes,
AGC, CMGC, CLK and TK8 (Table 4).
and the crystal structure of GSK3b,16 which became available in
the later stage of our project, nicely shows the unfavourably close
contact of the 7-fluorine to the gate keeper Cdk2-Phe80, while the
corresponding Leu132 in GSK3b leaves significantly more space
(Fig. 5). Introduction of the larger chlorine was even able to totally
abolish the Cdk2 activity (58), but only marginally affected the
activity on GSK3b.
The synthesis of these compounds was straightforward and lar-
gely derivative, as illustrated in Scheme 1.17 The aminoindazole
moiety B was built from a halo substituted aryl cyanide A by reac-
tion with hydrazine, followed by propyl amide introduction and
SEM protection. The halogen atom in B was replaced by an aryl
or heteroaryl group upon either direct Suzuki coupling with an
arylboronic acid, or via an intermediate pinacolborane to afford
monosubstituted aminoindazoles C (from Table 1). A 5-bromine
could be introduced by bromination of B and C, that could be fur-
ther substituted by an aryl or heteroaryl group using the same
strategy, leading to aminoindazoles D bearing one or two aryl or
heteroaryl groups in the 5- and 6-positions (compounds from Ta-
bles 2 and 3).
However, 37 displayed potent inhibition of Cdk1 and Cdk2 (0.1
and 2.5 lM, respectively). As Cdk’s and GSK3b belong to the same
family of CMGC kinases, they share a high degree of sequence and
structural homology. Indeed, several reported inhibitors of GSK3b
are non-selective vs. Cdks.9,10 Particularly noteworthy in this re-
spect are the reported series of pyrazolo[3,4-b]pyridine and pyraz-
From a homology model built from Cdk2 and several co-crystal
structures of inhibitors with Cdk2, a series of potent GSK3b inhib-
itors were designed. The model was also used to understand and
build in selectivity versus Cdk’s for these compounds. These com-
pounds displayed good cellular inhibition of Tau phosphoryla-
tion.18 However, their poor ADME properties made it difficult to
obtain reliable in vivo activities. Optimization of these parameters
will be the object of a following paper.
11
olo[3,4-b]pyridazine GSK3b inhibitors,
as these compounds are
quite similar to the ones described in this Letter in terms of struc-
tures and binding modes. These authors were able to successfully
address the Cdk2 selectivity of their series by introducing a basic
function in a position equivalent to our position 3 which resulted
in a steric/electronic clash with a salt bridge present in Cdk2 but
not in GSK3b.12 This concept was also applied successfully in the
case of a series of 4-acylamino-6-arylfuro[2,3-d]pyrimidines.13 In
our case, a close comparison of the ATP-binding site of both kinases
revealed a significant difference regarding the nature of their gate
keeper residues. This residue controls access to a hydrophobic
pocket not exploited by ATP.14 This structural difference has been
previously used to tune the selectivity for protein kinases.15 While
GSK3b has a leucine, Cdks have a phenylalanine at this position.
This amino acid is bulkier (see Fig. 5) and when looking at the model
of our aminoindazoles in the protein, it clearly appeared that the
interaction of position 7 was pointing right towards the gate keeper
residue and that introduction of a small substituent in this position,
while acceptable for GSK3b, could de detrimental to Cdks’ activity.
This turned out to be the case, as shown in Table 5. When a fluo-
rine was introduced in the 7-position (60), activity on GSK3b was
maintained but activity on Cdk1 was lowered by 20-fold. The
superposition of the crystal structure of 60 complexed to Cdk2
Acknowledgements
We are grateful to the analytical department for compound
analyses.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmcl.2010.01.114.
References and notes
1. Flaherty, D. B.; Soria, J. P.; Tomasiewicz, H. G.; Wood, J. G. J. Neurosci. Res. 2000,
62, 463; Brunden, K. R.; Trojanowski, J. Q.; Lee, V. M.-Y. Nat. Rev. Drug Disc.
2009, 8, 783.

D. Lesuisse et al. / Bioorg. Med. Chem. Lett. 20 (2010) 1985–1989
1989
2. Maccioni, R. B.; Munoz, J. P.; Barbeito, L. Arch. Med. Res. 2001, 32, 367.
3. De Bondt, H. L.; Rosenblatt, J.; Jancarik, J.; Jones, H. D.; Morgan, D. O.; Kim, S. H.
Nature 1993, 363, 595.
4. Technical note: measurement of GSK3b enzymatic activity by Scintillation
Proximity Assay (SPA). Partially purified recombinant human GSK3b enzyme is
12. Witherington, J.; Bordas, V.; Haigh, D.; Hickey, D. M. B.; Ife, R. J.; Rawlings, D. A.;
Slingsby, B. P.; Smith, D. G.; Ward, R. W. Bioorg. Med. Chem. Lett. 2003, 13, 1581.
13. Yutaka Maeda, Y.; Nakano, M.; Sato, H.; Miyazaki, Y.; Schweiker, S. L.; Smith, J.
L.; Truesdale, A. T. Bioorg. Med. Chem. Lett. 2004, 14, 3907.
14. Klebl, B. M.; Müller, G. Exp. Opin. Ther. Targets 2005, 9, 975.
15. Shimamuraa, T.; Shibataa, J.; Kuriharaa, H.; Mitaa, T.; Otsukia, S.; Sagaraa, T.;
Hiraia, H.; Iwasawa, Y. Bioorg. Med. Chem. Lett. 2006, 16, 3751. and references
cited therein; Gambacorti-Passerini, C. J. Med. Chem. 2006, 49, 5759; Shokat, K.
M. Nat. Chem. Biol. 2007, 229.
incubated for 30 min at 20 °C in 96-W microplates with 1
50 nCi M of pGS-2 peptide substrate (Upstate
33P]-33P]-ATP/well) and 0.8
Biotechnology Inc.) in the presence of tested compound, and in HEPES buffer
(40 mM, pH 7.4, supplemented with 200 M EDTA, 10 mM MgCl2, 1 mM DTT
lM ATP (including
c
-
l
l
and 0.1 mg/ml BSA). Enzymatic reaction is stopped by adding SPA-
streptavidine beads (GE Healthcare) concentrated (6 mg/ml) in PBS
(containing 20 mM ATP, 25 mM EDTA and 0.2% triton X-100) under 10 min.
agitation. Radioactivity is measured by using a scintillation liquid reader (1450
Microbeta™, Perkin–Elmer). IC50 values are calculated by non-linear regression
analysis with Excel/XL fit software (IDBS).
16. Bertrand, J. A.; Thieffine, S.; Vulpetti, A.; Cristiani, C.; Valsasina, B.; Knapp, S.;
Kalisz, H. M.; Flocco, M. J. Mol. Biol. 2003, 333, 393.
17. Dutruc, Rosset, G.; Lesuisse, D.; Rooney, T.; Halley, F. Fr. Demande FR 2836915
A1, 2003; Chem. Abstr. 2003, 139, 246027; Lesuisse, D.; Dutruc-Rosset, G.;
Halley, F.; Babin, D.; Rooney, T.; Tiraboschi, G. PCT Int. Appl. WO 2004062662
A1, 2004; Chem. Abstr. 2004, 141, 140438.
5. Figures were produced using PyMOL: DeLano, W.L. The PyMOL Molecular
Graphics System (2002) on World Wide Web http://www.pymol.org, Crystal
structure of 4 with Aurora has been deposited in the PDB with accession codes
3LAU.
6. Crystal structures of 37, 57 and 60 with Cdk2 have been deposited in the PDB
with accession codes 3LFS, 3FLN, and 3LFQ.
18. Briefly, brain slices (300
rats, and slice suspension (50
pyruvate and 4.5 g/l glucose) with tested compound in a final volume of 500
l
m) are prepared from cerebral cortex of 10 week-old
l/microtube) is incubated in DMEM (containing
l,
l
l
for 120 min at 37 °C under mild agitation. Incubation is stopped by
centrifugation, followed by lysis and sonication at 4 °C. Lysates are then
centrifuged (18,000g, 15 min, 4 °C), and the supernatants are used for protein
dosage, and subsequent western blotting analysis. SDS–PAGE electrophoresis is
performed in MOPS-SDS denaturating conditions; the phosphorylation of tau
protein is detected by using AD2 mouse monoclonal antibody specific for Ser396/
404 phosphoepitope as primary antibody, and peroxydase-labelled mouse
antiserum as secondary antibody for measuring signal by chemiluminescence.
Analysis of autoradiogrammes by ‘GeneTools’ software (Syngene, GeneGnome,
Ozyme) was performed for IC50 estimation. In these conditions, 37 and 57
displayed IC50’s of inhibition of 400 and 250 nM, respectively.
7. Internal profiling.
8. Niwa, T. J. Chem. Inf. Model. 2006, 46, 2158.
9. See for instance: Mettey, Y.; Gompel, M.; Thomas, V.; Garnier, M.; Leost, M.;
Ceballos-Picot, I.; Noble, M.; Endicott, J.; Vierfond, J.-M.; Meijer, L. J. Med. Chem.
2003, 46, 222.
10. Bain, J.; McLauchlan, H.; Elliott, M.; Cohen, P. Biochem. J. 2003, 371, 199.
11. 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. 1577.