Evaluation and Comparison of 3D-QSAR CoMSIA Models for CDK1, CDK5, and GSK-3 Inhibition by Paullones

Article abstract of DOI:10.1021/jm0308904

With a view to the rational design of selective GSK-3β inhibitors, 3D-QSAR CoMSIA models were developed for the inhibition of the three serine/threonine kinases CDK1/cyclin B, CDK5/p25, and GSK-3β by compounds from the paullone inhibitor family. The model

Full text of DOI:10.1021/jm0308904

2
2
J . Med. Chem. 2004, 47, 22-36
Articles
Eva lu a tion a n d Com p a r ison of 3D-QSAR CoMSIA Mod els for CDK1, CDK5, a n d
GSK-3 In h ibition by P a u llon es
†
†
†
†
†
‡
Conrad Kunick, Kathrin Lauenroth, Karen Wieking, Xu Xie, Christiane Schultz, Rick Gussio,
‡
§
§
|
⊥
Daniel Zaharevitz, Maryse Leost, Laurent Meijer, Alexander Weber, Flemming S. J ørgensen, and
Thomas Lemcke*
,
†
Institut f u¨ r Pharmazie, Abteilung f u¨ r Pharmazeutische Chemie, Universit a¨ t Hamburg, Bundesstrasse 45,
D-20146 Hamburg, Germany, Developmental Therapeutics Program, Division of Cancer Treatment and Diagnosis,
National Cancer Institute, Rockville, Maryland 20852, Centre National de la Recherche Scientifique, Station Biologique,
BP 74, F-29682 Roscoff, France, Department of Pharmaceutical Chemistry, University of Marburg, Marbacher Weg 6,
D-35032 Marburg, Germany, and Department of Medicinal Chemistry, Danish University of Pharmaceutical Sciences,
Universitetsparken 2, DK-2100 Copenhagen, Denmark
Received May 5, 2003
With a view to the rational design of selective GSK-3â inhibitors, 3D-QSAR CoMSIA models
were developed for the inhibition of the three serine/threonine kinases CDK1/cyclin B, CDK5/
p25, and GSK-3â by compounds from the paullone inhibitor family. The models are based on
the kinase inhibition data of 52 paullone entities, which were aligned by a docking routine
into the ATP-binding cleft of a CDK1/cyclin B homology model. Variation of grid spacing and
column filtering were used during the optimization of the models. The predictive ability of the
models was shown by a leave-one-out cross-validation and the prediction of an independent
set of test compounds, which were synthesized especially for this purpose. Besides paullones
with the basic indolo[3,2-d][1]benzazepine core, the test set comprised novel thieno[3′,2′:2,3]-
azepino[4,5-b]indoles, pyrido[2′,3′:2,3]azepino[4,5-b]indoles, and a pyrido[3′,2′:4,5]pyrrolo[3,2-
d][1]benzazepine. The best statistical values for the CoMSIA were obtained for the CDK1-
2
2
2
models (r ) 0.929 and q ) 0.699), which were clearly superior to the models for CDK5 (r )
2
2
2
0
.874 and q ) 0.652) and GSK-3 (r ) 0.871 and q ) 0.554).
In tr od u ction
as an important kinase in the intracellular signaling
pathway downstream from the insulin receptor. GSK-3
5
Glycogen synthase kinase-3 (GSK-3) is a serine/
threonine kinase, which is ubiquitously expressed in
mammalian tissues and interferes with a number of
signaling processes. Two isoforms of GSK-3 exist, des-
ignated GSK-3R and GSK-3â. While these isoforms
share a high homology at their catalytic site, the R-form
possesses an extended N-terminus with respect to the
â-form. The phosphorylation of proteins by GSK-3 is an
important link in signaling pathways that are regulat-
ing cell differentiation, cellular growth and proliferation,
metabolic processes, apoptosis control, inflammation,
and mechanisms involved in neuronal function. The
biochemical and functional properties of GSK-3 have
inactivates glycogen synthase by phosphorylation, re-
sulting in glycogen synthesis inhibition. Following the
interaction of insulin with its receptor in the plasma
membrane, a signaling cascade leads to activation of
protein kinase B (PKB, also called Akt), which in turn
deactivates GSK-3 by phosphorylation. In this situation
the unphosphorylated glycogen synthase remains active,
and glycogen synthesis is stimulated. Thus, the inhibi-
tion of GSK-3 by kinase inhibitors has been suggested
as a potential principle for the development of antidia-
3
,6
betic therapeutics.
Treatment of cultured human
skeletal muscle cells with lithium, which is thought to
7
recently been reviewed.1,2 Interference with these sig-
inhibit GSK-3 by competition for magnesium, resulted
in glycogen synthase activation and stimulation of
naling processes by the use of GSK-3-inhibiting agents
bears the potential to interfere with manifold physi-
ological or pathological situations, and consequently a
high research priority is currently dedicated toward the
development of small molecules with GSK-3-inhibitory
8
glucose incorporation into glycogen. In a number of
recent reports the discovery and development of struc-
tural diverse GSK-3 inhibitor classes has been de-
9
10
scribed, e.g. pyridyloxadiazoles, thiadiazolidinones,
1
1
properties.3 For instance, GSK-3 has been identified
,4
diamino-1,2,4-triazolecarboxylic acid derivatives, and
1
2,13
maleimides.
The maleimides SB-216763 and SB-
4
15286 have been shown to stimulate glycogen synthe-
*
To whom correspondence should be addressed. Tel: +49-40-
28383471. Fax: +49-40-428386573. E-mail: lemcke@chemie.uni-
hamburg.de.
1
2
4
sis in human liver cells. Current advances in the
search for GSK-3 inhibitors have been reviewed.4
†
Universit a¨ t Hamburg.
National Cancer Institute.
Centre National de la Recherche Scientifique.
University of Marburg.
Danish University of Pharmaceutical Sciences.
‡
§
|
⊥
Since GSK-3 is phylogenetically very closely related
3
to the cyclin-dependent kinases (CDKs), it is not
surprising that several kinase inhibitor classes have
1
0.1021/jm0308904 CCC: $27.50 © 2004 American Chemical Society
Published on Web 12/09/2003

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 23
Ta ble 1. Structures and CDK1, CDK5, and GSK-3 pIC50 Values for the Training Set
compd NSC #a
R1
R2
R3
R4
CDK1 CDK5 GSK-3
1
2
3
4
5
6
7
8
9
1
1
1
1
705701
641167 2-bromo
H
H
H
H
H
H
H
H
H
H
H
H
H
H
9-nitro
H
H
H
H
H
H
H
H
H
H
H
H
H
7.456
5.481
6.398
6.222
5.854
5.886
5.886
6.046
5.699
5.796
4.398
5.523
5.854
7.398 8.398
5.301 6.699
6.071 7.638
6.097 7.620
5.538 6.699
5.569 6.854
5.770 6.222
5.678 5.658
5.201 6.886
5.886 7.097
3.071 5.367
5.097 6.921
3.456 7.125
664704
672232
672233
672234
673433
675994
675995
675996
H
H
H
H
H
H
H
H
9-bromo
9-chloro
11-chloro
10-bromo
11-bromo
9-methoxy
9-methyl
9-fluoro
0
1
2
3
699471 4-hydroxy
699472 2,3-dihydroxy
703055
9-bromo
9-bromo
9-bromo
H
methyloxycarbonyl-
methyl
1
1
1
4
5
6
700692 2-bromo
700693 2,3-dimethoxy
H
H
9-bromo
9-bromo
9-bromo
H
H
H
6.523
6.699
5.194
5.000 6.097
6.301 7.000
5.276 6.301
701589
H
methyloxycarbonyl-
methyl
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
701590
H
H
H
H
H
9-trifluoromethyl
H
H
H
H
H
H
6.398
3.367
5.367
3.602
6.222 7.523
3.000 3.854
5.268 6.046
701591 4-methoxy
701592 2,3-dimethoxy
702373 4-methoxy
702374
702376
702377
9-bromo
3.398 4.796
b
H
H
H
tert-butyloxycarbonyl 9-bromo
methyl
benzyl
tert-butyloxycarbonyl 3.000
N.A.
3.194
9-bromo
9-bromo
9-trifluoromethyl
9-trifluoromethyl
8,10-dichloro
9-bromo
9-bromo
9-bromo
9-bromo
9-bromo
9-cyano
H
11-methyl
9-bromo
11-ethyl
H
H
H
H
H
4.699
4.456
6.553
6.620
5.602
3.886 5.678
3.569 5.000
6.367 7.125
5.523 5.699
3.456 5.301
3.097 5.638
3.854 6.523
3.398 6.398
3.620 5.398
3.585 5.824
7.357 8.000
4.996 6.208
5.046 6.301
3.097 4.620
4.638 6.155
4.824 6.456
702378 2,3-dimethoxy
703054 2-bromo
H
H
H
H
H
H
H
H
703056
703057
703059
704071
704072
704073
704626
641166
706120
706121
706122
H
H
H
H
H
H
H
H
H
H
H
tert-butyloxycarbonyl 4.155
2-hydroxyethyl
methyl
2-propenyl
ethyl
H
H
H
H
5.523
5.208
4.222
4.638
7.620
5.155
5.523
3.328
5.420
5.367
H
H
H
ethyl
H
H
H
707621 2-[2-(methyloxy-
carbonyl)ethenyl]
707623 2-iodo
H
9-trifluoromethyl
3
3
4
4
8
9
0
1
H
H
H
H
9-trifluoromethyl
H
9-cyano
H
H
H
H
6.155
5.432
7.357
5.495
5.155 5.658
5.131 6.602
7.222 7.745
5.081 5.699
707624 2-iodo
708244 2.3-dimethoxy
708245 2-[2-(1-hydroxy-
cyclohexyl)ethynyl]
708246 2-(2-cyanoethenyl)
708247 2-(3-hydroxy-1-propynyl)
709123 2-(2-cyanoethyl)
709125 2,3-dimethoxy
709126 2-(3-oxo-1-butenyl)
709129 2-bromo
684453
9-trifluoromethyl
4
4
4
4
4
4
4
4
5
5
5
2
3
4
5
6
7
8
9
0
1
2
H
H
H
H
H
H
9-trifluoromethyl
9-trifluoromethyl
9-trifluoromethyl
9-nitro
9-trifluoromethyl
9-nitro
H
H
H
H
H
H
6.569
6.523
7.328
7.620
6.495
7.276
4.481
5.638
4.367
4.292
3.886
5.022 6.398
5.699 6.699
7.481 7.481
7.678 7.886
4.469 5.854
6.921 6.699
3.155 3.456
5.097 5.699
3.347 5.921
3.066 3.745
see Figure 1
see Figure 1
see Figure 1
see Figure 1
see Figure 1
702375
704074
708241
708242
b
b
N.A.
N.A.
a
Designators from the NCI antitumor drug screen. b Defined value not available.
been characterized to act on both GSK-3 and distinct
to be a desirable feature for a potential drug, the
concomitant inhibition of especially CDK5 and GSK-3
is of high interest, because both kinases are involved
in the hyperphosphorylation of the tau protein. Hyper-
phosphorylated tau protein is the main constituent of
neurofibrillary tangles, which are found as intracellular
aggregates in brain cells of patients suffering from
1
4
15-18
CDKs, namely hymenialdisine, the indirubines,
1
9
20
the aloisines, and the paullones.
The paullones are a family of 7,12-dihydroindolo[3,2-
d][1]benzazepin-6(5H)-ones, which originally have been
2
1-24
developed as antiproliferative CDK inhibitors.
Among
the paullones, alsterpaullone (1) is the most potent
derivative toward GSK-3 (IC50 GSK-3â ) 4 nM). Al-
though the selectivity toward only one target appears
2
5
Alzheimer’s disease (AD). In AD, activation of CDK5
by the p25 protein, which is a pathological truncated

2
4
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
Ta ble 2. Structures and CDK1, CDK5, and GSK-3 pIC50 Values for the Test Set
compd
NSC #a
X
R1
R2
R3
9-bromo
R4
CDK1
CDK5
GSK-3
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
7
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
707622
708243
715966
715968
716450
716451
716453
718539
718541
719342
719343
720309
720310
720311
709128
716452
716784
716977
CH
CH
CH
CH
CH
CH
N
CH
N
N
CH
CH
CH
N
2-iodo
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
6.495
5.481
6.155
4.699
6.222
5.886
4.046
6.699
4.638
5.222
5.921
6.000
7.097
3.699
5.658
5.886
3.638
6.155
4.523
3.347
5.658
4.523
6.081
5.260
3.377
6.377
4.377
4.886
6.000
6.201
7.155
3.367
5.481
5.770
3.071
3.377
5.377
5.886
6.620
4.921
7.745
4.387
2-cyano
3-methoxy
H
9-trifluoromethyl
9-amino
9-bromo
9-acetylamino
9-methoxy
9-trifluoromethyl
9-chloro
9-trifluoromethyl
9-cyano
H
3-hydroxy
H
H
b
N.A.
2-(2-oxopropyl)
7.046
3.301
5.276
6.886
7.155
7.658
H
H
3-methoxy
3-methoxy
2-methoxy
H
9-nitro
9-nitro
9-methyl
b
N.A.
see Figure 2
see Figure 2
see Figure 2
see Figure 2
5.260
6.222
3.000
7.523
a
Designators from the NCI antitumor drug screen. b Defined value not available.
version of the physiological CDK5 activator p35, leads
to tau hyperphosphorylation.26,27 On the basis of these
3
,4
and other observations, both GSK-3 and CDK5 inhibi-
2
8,29
tion
have been suggested as putative therapeutic
strategies to treat or prevent AD. It has been shown
1
4
16
that hymenialdisine, indirubin-3′-monoxime, and
2
0
alsterpaullone inhibit the phosphorylation of human
tau expressed in insect cells at phosphorylation sites
specific for Alzheimer tau.
Of the above-mentioned GSK-3/CDK-inhibitors, the
paullones have been most extensively explored by
structural modification. For a further development in
this compound family, it is not only of interest to
generate compounds with enhanced kinase inhibitory
property, but also to design compounds with indepen-
dent CDK or GSK-3 inhibitory activity. Such com-
pounds, which are not yet available, would be suitable
for evaluating the contribution of either CDK5 or GSK-3
inhibition toward tau-phosphorylation within this fam-
ily of structures. Moreover, it might make sense to
suppress within this class of compounds the inhibitory
activity on all other CDKs that are not involved in tau
phosphorylation, such as CDK1, and CDK2, to avoid
potential undesired side effects.
Here we report on the generation of a set of 3D-QSAR
models, which make it possible to guide a rational
design of further paullones with respect to GSK-3/CDK1/
CDK5 selectivity. The models were created by compara-
tive molecular similarity indices analysis (CoMSIA)
using data of 52 paullones and structurally related
F igu r e 1. Selected compounds from the training set.
the evaluation of the models reported here, we used a
test set of 23 structural diverse paullone-related com-
pounds, which covered a broad range of kinase inhibi-
tory activity (Tables 2 and 3). Of this set, 21 entities
were especially synthesized for this purpose, including
some related heterocyclic systems, namely five thieno-
3
0-32
compounds for the inhibition of GSK-3â, CDK1/cyclin
B, and CDK5/p25, respectively, which we have pub-
lished previously20 (Table 1). Recently reported 3D-
QSAR models for the interaction of CDK1 with inhibi-
tors from the roscovitin/purvalanol-family demonstrated
[3′,2′:2,3]azepino[4,5-b]indoles (71, 72, 73, 74, 75), four
the feasibility of this approach toward the molecular
pyrido[2′,3′:2,3]azepino[4,5-b]indoles (59, 61, 62, 66),
and a pyrido[3′,2′:4,5]pyrrolo[3,2-d][1]benzazepine (67).
Two additional paullones (57, 70) from a recent paper35
were also included into the test set.
design of kinase inhibitors.3
3,34
In the quoted study, the
3
D-QSAR models are evaluated using a test set of 79
compounds which were published independently. For

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 25
Sch em e 1a
a
(
i) NaH, toluene, N2, 90 °C, 39%; (ii) DMSO, H2O, 150 °C, 2 h,
1%; (iii) 1. 4-R-C6H4-NHNH2, HOAc, 70 °C; 2. HOAc, H2SO4, 70
C, 39-64%.
8
°
tocol developed by McPhillamy et al.39 for the synthesis
of benzo analogues, 76 was reacted with diethyl succi-
nate 77 in toluene in the presence of sodium hydride.
The resulting cyclic enolized â-oxoester 78 was then
dealkoxycarbonylated by heating in wet DMSO. The
subsequent indole annelation was accomplished by
applying 79 to the sulfuric acid-catalyzed Fischer reac-
tion to yield the 4-azapaullones 61, 62, and 66. When
F igu r e 2. Selected compounds from the test set.
Ta ble 3. Thieno Analogues of Paullones Included in the Test
Set
4
-methoxyphenylhydrazine was reacted with 79 in
glacial acidic acid, the methoxy-substituted 4-aza-
paullone 59 was formed immediately without previous
addition of sulfuric acid and, due to its high reactivity
in the Fischer indolization reaction, the phenylhydra-
zone precursor of 59 was not detectable in this case
(
Scheme 1).
compd
NSC #a
R1
CDK1b
CDK5b
GSK-3b
Cyano-substituted paullone congeners were prepared
7
7
7
7
7
1
2
3
4
5
703058
709124
709127
719341
720601
bromo
methyl
chloro
cyano
nitro
6.222
5.398
6.301
7.125
7.097
5.398
4.398
5.301
6.495
7.000
6.921
5.886
6.398
6.481
7.456
40,41
by means of a Rosenmund-von Braun reaction.
Upon heating appropriate bromo compounds as starting
materials with cuprous cyanide in N-methylpyrrolidone,
the desired nitriles 54, 63, and 74 were obtained, albeit
in poor to modest yields.
The 2-(2-oxopropyl)-9-trifluoromethylpaullone 60 was
prepared in a sequence starting from the previously
described iodo derivative 38, which was reacted under
a
Designators from the NCI antitumor drug screen. b pIC50
values.
Ch em istr y
23
The synthesis of all members of the training set has
been published previously.20,22,36 Furthermore, prepara-
the conditions of a palladium-catalyzed Stille coupling
with allyltributyltin to yield the 2-allyl-9-trifluoro-
methylpaullone 80. The allyl moiety of 80 was subse-
quently air-oxidized by means of a Wacker reaction in
the presence of palladium chloride and copper chloride,
furnishing the oxo compound 60 in satisfactory yield
(Scheme 2).
The preparation of new alsterpaullone derivatives is
outlined in Scheme 3. While the reduction of alster-
paullone (1) with zinc powder in ethanol in the presence
of calcium chloride led to 9-aminopaullone 56, the
usage of iron powder in refluxing glacial acetic acid
led to the acetylated analogue 58. Treatment of
alsterpaullone (1) with phosphorus pentasulfide in THF
gave thioalsterpaullone 68 which was dimethylated by
iodomethane at both the sulfur and the indole nitrogen
after deprotonation with sodium hydride to yield the
thioimidate 69.
tion procedures for compounds 57 and 70 from the test
set have recently been reported.3
5,37
Of the hitherto
unknown members of the test set, compounds 53, 55,
1, 72, and 73 were synthesized by conventional Fischer
7
indolization procedures, starting with known seven-
membered cyclic ketones and appropriate phenylhydra-
zine derivatives, which were reacted in a mixture of
glacial acetic acid and sulfuric acid following an estab-
lished procedure without isolation of the phenylhydra-
zone intermediate.36 However, namely in cases in which
electron deficient phenylhydrazine derivatives were
employed, the acid-catalyzed Fischer indolization led to
unsatisfactorily results. Hence, for the synthesis of the
paullone derivatives 64, 65, 68, and 75, the correspond-
ing phenylhydrazones were prepared first, and subse-
quently heated in refluxing diphenyl ether to yield the
target compounds by means of a thermal Fischer
indolization under neutral conditions.38
Meth od s a n d Ma ter ia l
The preparation of the cyclic ketone 79, which was
Da ta set. The training set for the CoMSIA calcula-
employed as a building block for the construction of
tions consisted of compounds that were published by
Leost et al. in 2000. From the original number of 56
compounds, four were discarded, because they were
2
0
4
2
-azapaullones, was based on commercially available
-aminonicotinic acid ethyl ester 76. Following a pro-

2
6
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
Sch em e 2a
All compounds of the training set and the test set
were manually docked into the ATP binding pocket of
the above-mentioned CDK1/cyclin B model from Gussio
2
4
et al., using the already docked kenpaullone scaffold
as a template. All inhibitor-protein complexes were
4
5,46
minimized using the MAB force field
as imple-
mented in the program MOLOC, keeping the protein
fixed and allowing the inhibitor to move.
CoMSIA. For the alignment in the CoMSIA analysis,
the relative three-dimensional orientations of the com-
pounds included in the training set and the test set
resulting from the MOLOC minimizations were used
(Figure 3a,b). This resulted in an alignment that most
likely represents the bioactive conformation of the
paullones in the CDK1 binding site. Partial atomic
charges were recalculated as indicated above, using the
AM1 as implemented in MOPAC. A 3D cubic lattice box
with a grid spacing of 2 Å or 1 Å was created around
the aligned molecules. The five CoMSIA similarity index
fields available within SYBYL (steric, electrostatic,
a
(
i) Allyltributyltin, 10 mol % PdCl2(PPh3)2, 20 mol % PPh3,
DMF, 95 °C, 2 h, N2, 70%; (ii) 10 mol % PdCl2, CuCl2, DMF, H2O,
rt, air, 16 h, 69%.
Sch em e 3a
3
0-32
hydrophobic, hydrogen bond donor and acceptor)
1
+
were calculated using a C probe of a 1 Å radius and
the default value of 0.3 as attenuation factor.
The CoMSIA fields were used as independent vari-
ables, and the pIC50 values as dependent variables in a
4
7
partial least squares (PLS) regression analysis, to
derive 3D-QSAR models. To determine the optimal
number of components the SAMPLS algorithm48 (as
implemented within SYBYL) was used. The number of
components corresponding to the lowest PRESS (predic-
tive sum of squares) value was used for deriving the
subsequent PLS models. Cross-validation was per-
formed using the leave-one-out method (LOO) with a
a
(
i) Zn, EtOH, CaCl2, H2O, reflux, 35% (ii) Fe, HOAc, reflux,
3
7%; (iii) P2S5, NaHCO3, THF, N2, reflux, 3 h, 44%; (iv) NaH, H3CI,
THF, N2, reflux, 4 h, 47%.
2
column filtering of 0.5, 1.0, and 2.0 to calculate q
2
(
predictive r ) and PRESS to check the predictiveness
of the corresponding model. Finally nonvalidated models
were produced and the corresponding conventional
correlation coefficient r , its standard error s, and the
F ratio were computed (Tables 4 and 5). The stdev*coeff
contour maps for all five CoMSIA fields were produced
and analyzed for each enzyme.
either not active (IC50 above 1000 µM) or had undefined
stereochemistry. We ended up with a list of 52 com-
pounds (Table 1). The test set comprised 21 newly
synthesized compounds described in the chemistry
section and two recently published compounds (Tables
2
2
and 3).35 The IC50 values for CDK1, CDK5, and GSK-3
For a compression of the amino acids forming the ATP
binding pocket of CDK1, CDK5, and GSK-3, respec-
tively, a multiple sequence alignment was done with the
were converted to pIC50 (-log IC50) values and used as
dependent variables in the CoMSIA calculations. The
pIC50 values of the compounds from the training set and
the test set covered an interval of more than 5 log units
for all three target enzymes.
4
9
program ClustalW at the ClustalW WWW Service at
the European Bioinformatics Institute (http://www.ebi.
5
0
ac.uk/clustalw), using the sequences of CDK1 (Q17066),
5
1
52
CDK5 (Q00535), and GSK-3 (P49841), that were
retrieved from the SwissProt database and a default
setup (default matrix type: Gonnet 250).
Molecu la r Mod elin g a n d Dock in g. The three-
dimensional structures of all compounds in the training
set and in the test set were constructed using the
SYBYL program package, version 6.7 and 6.8.4 Because
only one of the two mirror image conformations of the
kenpaullone parent ring system was suitable for docking
into the ATP binding site of a CDK1/cyclin B homology-
5
3
2
Resu lts a n d Discu ssion
Align m en t. In 3D-QSAR studies the spatial align-
ment of the compounds is usually one of the key steps
in order to obtain meaningful results. The biologically
active conformations of the structures should be aligned
in a way representing a similar binding mode. In the
case that the biologically active conformation is not
known, the lowest energy conformation of each structure
is usually aligned on the basis of a common structural
feature.
2
4
model, the corresponding azepine ring conformer was
used as a template to model all paullone derivatives
used in this study. Partial atomic charges were calcu-
lated using the semiempirical program MOPAC 6.0,
4
3
applying the AM1 Hamiltonian. Energy minimizations
were performed using the Tripos force field4 with a
distance dependent dielectric and the Powell conjugate
gradient algorithm with a convergence criterion of 0.005
kcal/(mol Å) and max. 10000 iterations, respectively,
and taking charges into account.
4
In a preliminary study with our dataset,54 an align-
ment of the compounds on the basis of the seven-
membered lactam ring system of the energy-minimized

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 27
Ta ble 4. Summary of the CoMSIA Models
CDK1a
CDK5b
1 Å
0.5
GSK3c
step size
col. filt.
2 Å
0.5
1 Å
0.5
2 Å
0.5
2 Å
0.5
1 Å
0.5
no. used columns 1158
9275
1158
9275 1158
9275
2
q
0.699
0.663
0.929
0.321
0.691 0.652 0.649 0.546 0.548
0.671 0.846 0.849 0.861 0.859
0.929 0.874 0.871 0.874 0.874
0.323 0.509 0.515 0.454 0.453
PRESS
2
r
s
F
120.916 119.6 81.422 79.520 63.544 63.865
opt. no. compon.
stepsize
col. filt.
5
2 Å
1.0
5
1 Å
1.0
4
4
1 Å
1.0
5
2 Å
1.0
5
1 Å
1.0
2 Å
1.0
907
no. used columns 907
7139
7139 907
7139
2
q
0.697
0.689 0.651 0.649 0.548 0.550
0.673 0.847 0.849 0.859 0.857
0.928 0.872 0.870 0.873 0.874
0.325 0.512 0.517 0.455 0.454
PRESS
0.665
0.928
0.324
2
r
s
F
118.793 117.766 80.323 78.726 63.274 63.555
opt. no. compon.
stepsize
col. filt.
5
2 Å
2.0
5
1 Å
2.0
4
4
1 Å
2.0
5
2 Å
2.0
5
1 Å
2.0
2 Å
2.0
632
no. used columns 632
4971
4971 632
4971
2
q
0.692
0.686 0.652 0.649 0.552 0.554
0.676 0.846 0.850 0.855 0.853
0.926 0.871 0.866 0.870 0.871
0.329 0.514 0.525 0.460 0.458
PRESS
0.670
0.927
0.325
2
r
s
F
117.474 114.559 79.581 76.124 61.837 62.292
opt. no. compon.
5
5
4
4
5
5
a
The best CDK1 model (in italics) is called CDK1 model in the
text. b The best CDK5 model (in italics) is called CDK5 model in
the text. c The best GSK3 model (in italics) is called GSK-3 model
in the text.
Ta ble 5. Comparison of the Statistical Values and Field
Distribution within the Selected CoMSIA Models of CDK1,
CDK5, and GSK-3
CDK1 model CDK5 model GSK-3 model
step size
col. filt.
q
2 Å
0.5
2 Å
0.5
1 Å
2.0
2
0.699
0.663
0.929
0.321
120.916
5
0.652
0.846
0.874
0.509
81.422
4
0.554
0.853
0.871
0.458
62.292
5
PRESS
2
r
s
F
opt. no. components
field contribution
steric
electrostatic
hydrophobic
donor
0.087
0.253
0.181
0.188
0.290
0.092
0.236
0.227
0.169
0.277
0.084
0.178
0.232
0.213
0.293
acceptor
CDK1/cyclin B model24 as a template for our docking.
In this model the indolobenzazepinone ring system is
sandwiched between the lipophilic side chains of Ile10
and Leu134. Three hydrogen bonds are observed (azepine
NH-Leu83 backbone carbonyl, azepine carbonyl-
Leu83 backbone NH, indole NH-Asp86 side chain
carbonyl) that fix the structures in a similar spatial
position.
F igu r e 3. (a) Spacial orientation of kenpaullone 3 in the
binding site. Hydrogen bonds with Leu83 and Asp86 are
depicted by dotted lines. (b) View of the alignment of all
compounds of the training set. (c) View of the alignment of all
compounds of the test set.
After manually docking all structures of the training
set and the test set into the CDK1 binding site, we
minimized all inhibitor-protein complexes with the
MAB-force field of the MOLOC program suite which
uses geometric hydrogen bonding terms to model elec-
trostatic interactions of polar groups. By this method
all three hydrogen bonds of the semiempirical model
structures was carried out and produced only moderate
q
2
and r2 values. Therefore, we concluded, that the
ligand side chain conformation is of critical importance
to the biologically active conformations, and conse-
quently we used a docked alignment.
Since there is no X-ray crystallographic structure
information about the biologically active conformation
of paullones binding to CDK1/cyclin B, we used a model
of kenpaullone docked into the ATP-binding site of a
2
4
from Gussio et al. were nicely reproduced (Figure 3a).
The resulting relative conformations of all compounds

2
8
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
in the training set and test set were used (Figure 3b,c)
for the following CoMSIA calculations.
Ta ble 6. Residuals of the Training Set
CDK1 CDK5
fitted fitted
GSK-3
fitted
CoMSIA Mod els. The results of the initial CoMSIA
models for a 2 Å and a 1 Å grid spacing are summarized
in Table 4. For all three targets, models with a q value
compd pIC50 prediction pIC50 prediction pIC50 prediction
2
1
2
3
4
5
6
7
8
9
1
11
12
1
1
7.46
5.48
6.40
6.22
5.85
5.89
5.89
6.05
5.70
5.80
4.40
5.52
5.85
6.52
6.70
5.19
6.40
3.37
5.37
3.60
3.00
4.70
4.46
6.55
6.62
5.60
4.16
5.52
5.21
4.22
4.64
7.62
5.52
3.33
5.42
5.16
5.37
6.16
5.43
7.36
5.49
6.57
6.52
7.33
7.62
6.49
7.28
4.48
5.64
4.37
4.29
3.89
7.31
5.76
6.13
6.11
5.63
5.73
5.67
5.87
5.85
6.05
5.00
5.26
5.72
6.18
6.51
5.03
6.31
2.78
6.09
3.60
3.00
4.34
4.26
6.71
6.37
5.67
4.39
5.59
5.00
4.61
4.81
7.16
5.59
3.98
5.45
5.69
5.60
6.39
5.80
7.55
5.08
6.80
6.59
6.87
7.68
6.03
7.34
4.11
5.44
4.48
4.74
3.93
7.40
5.30
6.07
6.10
5.54
5.57
5.85
5.68
5.20
5.89
3.07
5.10
3.46
5.00
6.30
5.28
6.22
3.00
5.27
3.40
3.00
3.89
3.57
6.37
5.52
3.46
3.10
3.85
3.40
3.62
3.58
7.36
5.05
3.10
4.64
5.00
3.89
5.16
5.13
7.22
5.08
5.02
5.70
7.48
7.68
4.47
6.92
3.15
5.10
3.35
3.07
3.00
7.17
5.09
5.59
5.55
5.39
5.17
5.56
5.61
5.42
5.65
4.13
4.79
2.98
5.17
6.26
5.28
5.71
2.42
6.18
3.25
3.13
3.30
4.07
6.53
5.44
4.82
2.85
3.57
3.48
3.65
3.64
6.65
5.29
3.51
5.20
5.37
4.09
5.37
5.04
7.46
4.83
5.86
5.74
5.98
7.93
4.28
6.86
2.93
4.60
3.50
3.89
3.39
8.40
6.70
7.64
7.62
6.70
6.85
6.05
5.66
6.89
7.10
5.37
6.92
7.13
6.10
7.00
6.30
7.52
3.85
6.05
4.80
3.19
5.68
5.00
7.13
5.70
5.30
5.64
6.52
6.40
5.40
5.82
8.00
6.30
4.62
6.16
6.21
6.05
5.66
6.60
7.74
5.70
6.40
6.70
7.48
7.89
5.85
6.70
3.46
5.70
5.92
3.74
3.00
7.90
5.98
7.14
6.98
6.29
6.41
6.37
6.64
6.76
6.89
6.14
6.75
7.11
6.55
7.16
5.68
6.92
3.35
6.60
4.68
3.36
5.21
4.68
6.90
6.29
6.11
5.77
6.82
6.16
5.79
6.05
7.89
6.37
4.93
6.23
6.60
5.89
6.18
5.88
7.87
5.36
6.80
6.68
6.84
7.88
5.78
7.25
3.23
5.59
5.80
4.49
3.29
above 0.5 could be obtained, indicating that the respec-
tive models could be useful in predicting the inhibitory
activity of paullones in these targets. However, the
models for CDK1 performed much better with respect
2
2
to q and r , compared to the CDK5 or the GSK3 model.
For CDK1 and CDK5 a grid spacing of 2 Å and column
2
filtering of 0.5 resulted in the best models (CDK1: q
0
2
)
0.699, r ) 0.929, 5 components, CDK1 model;
2
2
CDK5: q ) 0.652, r ) 0.874, 4 components, CDK5
model) while for GSK-3 a grid spacing of 1 Å and column
3
4
2
2
filtering of 2.0 was best suited (GSK-3: q ) 0.554, r
0.871, 5 components, GSK-3 model). In an attempt
15
1
1
1
1
6
7
8
9
)
to further improve our models, region focusing was used
on the best initial PLS analysis, resulting in a CDK1
2
2
model with slightly better r and q values. However,
since concomitantly the number of components neces-
sary to derive the model rose from 5 to 7, this refinement
was considered to be of minor significance and therefore
the region focused CDK1 model was not pursued
further. For CDK5 and GSK-3, the application of region
focusing failed to improve the models. The further
discussion of the CoMSIA will focus on the best models
20
2
2
2
24
25
2
2
2
1
2
3
6
7
8
29
3
3
3
0
1
2
(
Table 5), which will be named CDK1 model, CDK5
model, and GSK-3 model, respectively, as mentioned
above.
33
3
3
3
4
5
6
An inspection of the fitted predictions for the best
CoMSIA models (Table 6 and Figure 4) revealed that
three compounds from the training set (44, 26, and 11)
were poorly predicted in the CDK5 and GSK-3 models.
In the case of 44 this could be caused by the relative
flexibility of the cyanoethyl side chain. Since we used a
docked alignment with the 3D structure of CDK1 as
template, the side chain conformation is optimized for
CDK1 only, but might be different in CDK5 or GSK-3.
37
3
3
4
4
8
9
0
1
42
4
4
4
46
47
4
4
5
3
4
5
1
1 is the only 4-hydroxypaullone and 26 the only 8,10-
disubstituted paullone in the dataset. This structural
uniqueness might be the reason that both structures
are poorly predicted in the CDK5 and GSK-3 models.
Even though the deletion of these compounds from the
training set would have improved the statistical results
for the CDK5 and GSK-3 models, we decided to keep
them in the dataset because they still might contribute
valuable information for the entire model.
8
9
0
51
5
2
CDK5 or GSK-3. On closer inspection some additional
details are observed.
The ultimate test for the usefulness of a 3D-QSAR
model in the drug design process is predicting the
activity of new compounds that are not included in the
dataset that was used to obtain the model. The CoMSIA
models reported here were challenged with a test set
comprising 23 newly synthesized paullone derivatives.
Among the members of this set were paullones repre-
senting structures similar to the training set compounds
and some derivatives with new structural features (4-
azapaullones 59, 61, 62, and 66, as well as the thiophene
analogues 71, 72, 73, 74, and 75). The predicted pIC50
values for the test set are summarized in Table 7 and
Figure 4. On first sight the prediction of the inhibitory
In all three models the 4-azapaullones (59, 61, 62, 66)
and the 3-methoxy-substituted paullones (63, 64) were
clearly overpredicted. This observation can be explained
with the absence of similar compounds in the training
set that could contribute required information to the
model regarding the 4- or 3-position, respectively.
Similarly, the thioimidate 69 was overpredicted, al-
though the training set comprised the thioimidate 50.
However, 50 bears an unsubstituted indolo-nitrogen at
position 12 that could be involved in hydrogen bonding,
while 69 is lacking this feature through methyl substi-
tution. Consequently, 69 has lost both hydrogen bond
donor groups, a structural modification that could be
responsible for a completely different alignment mode
for 69.
2
activities for CDK1, as was to be expected from the q
of the corresponding QSAR model, is better than for

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 29
Ta ble 7. Predictions of the Test Set
CDK1
CDK5
GSK-3
compd pIC50 prediction pIC50 prediction pIC50 prediction
5
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
7
7
7
7
7
7
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6.490
5.480
6.160
4.700
6.220
5.890
4.050
6.700
4.640
5.220
5.920
6.000
7.100
3.700
5.660
5.890
3.640
6.160
6.220
5.400
6.300
7.120
7.100
6.220
5.666
6.413
5.390
5.436
5.890
5.334
6.292
5.568
5.779
7.267
7.399
7.602
5.300
6.025
6.636
5.313
6.188
6.284
6.027
6.246
7.337
7.453
4.520
3.350
5.660
4.520
6.080
5.260
3.380
6.380
4.380
4.890
6.000
6.200
7.160
3.370
5.480
5.770
3.070
6.380
5.400
4.400
5.300
6.490
7.000
5.121
5.024
6.034
5.262
4.915
5.690
5.141
5.521
4.956
5.257
6.969
7.455
7.673
4.920
5.383
6.188
5.367
5.538
5.720
5.677
5.777
6.910
7.394
5.380
5.890
6.620
4.920
7.740
4.390
N.A.a
7.050
3.300
5.280
6.890
7.160
7.660
N.A.a
5.260
6.220
3.000
7.520
6.920
5.890
6.400
6.480
7.460
6.445
6.239
6.851
6.373
6.773
6.622
-
6.354
6.885
6.776
7.823
7.832
7.994
-
7.077
6.337
5.263
6.840
7.631
7.213
7.446
8.348
8.369
a
Defined values not available.
the absence of a comparable compound in the training
set is not critical for these substances. In contrast, the
GSK-3 model over predicted these compounds. The
latter observation might be due to structural differences
between the CDKs and the GSK-3 that are not ex-
plained with the alignment that was used for the
training set and test set of this study.
With exception of the above-mentioned compounds,
the inhibitory activities of the structures in the test set
are predicted reasonably well. Consequently, we believe
all three QSAR models can be used as tools in the
rational design of further GSK-3 or CDK inhibitors of
the paullone family.
An a lysis of th e 3D Con tou r Ma p s of th e CoMSIA
F ield s. For each enzyme the contour maps for all five
CoMSIA fields were produced and analyzed.
CDK1. Figure 5a,b shows the steric and electrostatic
contour maps for the CDK1 model with alsterpaullone
(1) as an example for a good inhibitor and compound
2
7 as an example for a poor inhibitor. Sterically favored
regions (colored green) appear around the nitro group
at position 9 and near position 2 and 3 of the paullone
scaffold, while disfavored areas (colored yellow) are
located near the indole-nitrogen at position 12 and
beyond the green (favorable) isopleth at position 2 and
3
of alsterpaullone. Negatively charged groups are
favored near the oxygen atoms of the 9-nitro group, near
the oxygen of the lactam-group at position 6, and at
position 2 of the paullone ring system. A blue isopleth
above the indole ring system is representing an area
where positive charge is favored. The distribution of
these electrostatic contours suggests that a negatively
charged and electron-withdrawing substituent in the
9
-position of the paullone scaffold is depleting the indole
F igu r e 4. Plot of actual pIC50 vs predicted pIC50 for the
training set (O) and the test set (b) of CDK1 model (a), CDK5
model (b), and GSK-3 model (c).
ring system of electrons, making the electrostatic po-
tential more positive and thereby leading to inhibitors
with enhanced biological activity. These findings are in
good agreement with quantum mechanics calculations
The CDK1 and the CDK5 model predicted the activity
of the thiophene analogues quite well, indicating that
2
4
published recently.

3
0
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
F igu r e 5. CoMSIA contour maps for the steric and electrostatic fields. Green contours (contour levels: CDK1: 0.006; CDK5:
.007; GSK-3: 0.0007) indicate areas where sterical bulk is favored while yellow contours (contour levels: CDK1: -0.010; CDK5:
0
-
0.007; GSK-3: -0.0015) indicate areas where it is not favored. Blue isopleths depict areas where positively charged groups
increase the activity (contour levels: CDK1: 0.021; CDK5: 0.020; GSK-3: 0.004) and red isopleths indicate where negatively
charged groups increase the activity (contour levels: CDK1: -0.023; CDK5: -0.020; GSK-3: -0.004). Contours of CDK1 model
(a), CDK5 model (c), and GSK-3 model (d) with alsterpaullone 1. (b) Contour of CDK1 model with compound 27.
Substitution of the indole nitrogen, like that in
alsterpaullone. A recently reported CoMSIA model for
CDK1-inhibition by paullones produced contour plots
with considerable different features, namely for the
electrostatic field distribution. These differences can
easily be rationalized, taking into account that the
model from the literature was constructed on the basis
of a smaller number of compounds with a higher
similarity, which were fitted by a rigid alignment
procedure.
compound 27, with a boc group causes an anticlockwise
shift of the whole paullone scaffold, when compound 27
is docked into the active site cavity. As a result of this
shift, the bromine atom at position 9 is no longer
rendered within the favored sterical area. In contrast,
the bromo substituent is forced away from the area
where negative charge is favored. Furthermore, the
butyloxy group of the 12-substituent is occupying the
sterically disfavored area, and the oxygen of the ester
group is placed inside the area where positive charge
is favored. This is reflected by a substantial loss of
inhibitory activity compared to alsterpaullone.
5
5
The contour plots of the CDK1 model nicely reflect
the situation within the ATP binding site. For example,
areas were sterical bulk is favored are located near the
entrance to the binding pocket and at the sugar pocket
that would accommodate the ribose moiety of ATP
The contour maps of the hydrogen bond donor and
acceptor fields describe the spatial arrangement of
favorable and disfavorable hydrogen bond interactions
to acceptor or donor groups of the target enzyme (Figure
(Figure 5a). The four hydrogen bonds that are assumed
to render the paullones in the binding pocket are also
accurately visualized by the contour plots. Thus, while
the cyan contours depicted in Figure 6a represent the
backbone carbonyl of Leu83 and the side chain carboxy-
late of Asp86, the orange contours match the backbone
NH of Leu83 as well as a water molecule reported to be
located near the aromatic ring of Phe80.
6
a,b). The cyan contours that are observed opposite the
lactam hydrogen at position 5 and the indole hydrogen
at position 12 describe a favorable hydrogen bond to an
acceptor group in the protein, while the orange contours
near the lactam oxygen at position 6 and the nitro group
at position 9 indicate favorable hydrogen bonds to a
hydrogen bond donor structure in the active site. Al-
sterpaullone fulfills these structural requirements per-
fectly, being able to form all four possible hydrogen
bonds to the target molecule (Figure 6a). Compound 20
should in principle also be able to form these hydrogen
bonds. However, the 4-methoxy group causes the ring
system to rotate anticlockwise, resulting in an align-
ment, that places the lactam and the indole hydrogen
near the purple contours which indicate disfavorable
interactions between a ligand hydrogen bond donor
group and the protein (Figure 6b). Consequently, com-
pound 20 is 4 orders of magnitude less active than
CDK5. For CDK5, the spatial distribution of favorable
and unfavorable contours is nearly identical, while the
relative size varies a little compared to CDK1 (Figures
5
c and 6c). This similarity is corresponding to the high
degree of conservation of the amino acid sequence within
the ATP binding sites of CDK1 and CDK5. A multiple
sequence alignment, done with ClustalW, revealed that
only three active site residues vary between the se-
quence of our CDK1 model and the sequence of CDK5
(Leu83-Cys83; Thr84-Asp84; Met85-Glu85). All three
residues present their backbone to the binding site
cavity, while their side chains point away from the

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 31
F igu r e 6. CoMSIA contour maps for the hydrogen bond donor and acceptor fields. Cyan contours (contour levels: CDK1: 0.035;
CDK5: 0.04; GSK-3: 0.009) indicate areas where an acceptor group in the protein forms favorable hydrogen bonds to a donor
group of the ligand, while orange contours (contour levels: CDK1: 0.032; CDK5: 0.04; GSK-3: 0.008) indicate areas where a
donor group in the protein forms favorable hydrogen bonds to an acceptor group of the ligand. The purple contours (contour
levels: CDK1: -0.0041; CDK5: -0.025; GSK-3: -0.0085) and the red contours (contour levels: CDK1: -0.016; CDK5: -0.021;
GSK-3: -0.0035) indicate areas where an acceptor or a donor group, respectively, in the protein would form unfavorable hydrogen
bonds to the ligand. Contours of CDK1 model (a), CDK5 model (c), and GSK-3 model (d) with alsterpaullone 1. (b) Contour of
CDK1 model with compound 20.
cavity. Hence, the architecture of the ATP binding site
of CDK1 and CDK5 can be considered as rather similar.
GSK-3. The contour maps of GSK-3 are different to
the ones of CDK1 and CDK5. The steric field shows two
additional favorable contours below the lactam oxygen
at position 6 and around the benzo anneland of the
indole substructure, respectively (Figure 5d). With
respect to the electrostatic field, a much larger blue
isopleth (positive charge favored) near the 1 position is
observed in GSK-3, while the large red contour indicat-
ing a favorable interaction for negative charges near the
position 2, as found in the CDK1 and CDK5, is missing.
Moreover, in GSK-3 the blue isopleth above the indole
ring system observed in CDK1 and CDK5 is divided into
two areas that are located above the 8- and the 11-
position of the paullone ring system. An additional area
of favorable negative charge (colored red) is found
slightly below the indole ring system.
coefficients of 0.699, 0.652, and 0.554, respectively.
Differences in the 3D contour plots for inhibition of the
three enzymes show that a rational design of paullones
directed toward selective GSK-3 or CDK inhibition
should be possible by the application of the three
models. Namely the differences in the electronic fields
between the models should be taken into account for
the further development of GSK-3-selective paullones.
In the course of our future research, new compounds
with modified electron density throughout the hetero-
cyclic ring system will be designed and be evaluated in
the CoMSIA models in silico before synthetic approaches
are undertaken.
Exp er im en ta l Section
Syn th etic Ch em istr y. Melting points (mp) were deter-
mined on an electric variable heater (Electrothermal 9100) and
were not corrected. Elemental analyses were performed in the
analytical departments of the Fachbereich Chemie, Universit a¨ t
Hamburg. Results obtained were within (0.4% unless indi-
cated otherwise. Nuclear magnetic resonance spectra were
recorded on a Bruker AMX 400 instrument, using tetrameth-
Only slight differences were found when the hydrogen
bonding contour maps of the GSK-3 model were com-
pared to the CDK1 and CDK5 models, respectively
6
ylsilane as internal standard and dimethyl sulfoxide-d as
(Figure 6d).
solvent unless stated otherwise. NMR signals are reported in
ppm on a δ scale. High-resolution FAB mass spectra (FAB-
HMRS) were determined on a Finnigan MAT 311A instrument
using 3-nitrobenzylic alcohol as matrix. EI/HRMS spectra were
recorded on a VG 70-250S double focusing mass spectrometer.
Column chromatography was performed using chromatogra-
phy grade silica gel 60 (Merck). Thin-layer chromatography
Con clu sion
Three CoMSIA models have been developed, which
describe the three-dimensional structure activity rela-
tionships for the kinase-inhibitory activity of paullones
with respect to CDK1, CDK5, and GSK-3â. The models
(TLC) was performed using fluorescent Polygram Sil G/UV254
2
are characterized with conventional r of 0.929, 0.874,
silica gel plates. Spots were visualized under 254 nm UV
illumination.
2
and 0.871, respectively, and cross-validated with q

3
2
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
Gen er a l P r oced u r e A for P r ep a r a tion of P a u llon e
(dd, 1H, 7.9/1.3 Hz), 8.45 (dd, 1H, 4.6/1.0 Hz), 10.38 (s, 1H),
Der iva tives 53, 55, 61, 62, 66, 71, 72, 73, by Acid -Ca ta lyzed
F isch er In d ole Rea ction . To a slurry of an appropriate cyclic
ketone (1 mmol) in glacial acetic acid (2 mL) is added dropwise
a suspension of the appropriate corresponding phenylhydra-
zine (1.5 mmol) [or the appropriate substituted phenylhydra-
zine hydrochloride (1.5 mmol) and sodium acetate (123 mg,
11.91 (s, 1H); Anal. (C15
63.08.
3
H10ClN O); H, N; C: calcd 63.50; found
9-Tr iflu or om eth yl-7,12-dih ydr o-pyr ido[2′,3′:2,3]azepin o-
[4,5-b]in d ol-6(5H)-on e (62). Preparation following General
Procedure A from 6,7,8,9-tetrahydro-5H-pyrido[2,3-b]azepine-
5,8-dione (79) and 4-trifluormethylphenylhydrazine yielded 3%
of a beige powder after crystallization from ethanol; mp > 330
1
.5 mmol)] in glacial acetic acid (5 mL) with stirring. After
being stirred at 70 °C for 1 h, the mixture is cooled to room
temperature. Concentrated sulfuric acid (0.1 mL) is added, and
stirring is continued at 70 °C. The reaction is monitored by
TLC (eluent: ethyl acetate or toluene/acetone 1:1, respec-
tively). Portions of concentrated sulfuric acid (0.1 mL) are
added each 1.5 h until completion of the reaction is detected.
After being cooled to room temperature, the mixture is poured
into a 5% aqueous sodium acetate solution (15 mL). A
precipitate is formed, which is filtered off with suction and
purified by crystallization from the given solvent.
°C (dec starting at 304 °C); H NMR 3.73 (s, 2H), 7.37 (dd,
1
1H, 7.6/4.6 Hz), 7.49 (d, 1H, 8.1 Hz), 7.63 (d, 1H, 8.2 Hz), 8.19-
8.21 (m, 2H), 8.47 (dd, 1H, 6.8/1.8 Hz), 10.42 (s, 1H), 12.19 (s,
1H); Anal. (C H F N O) (317.27); C, H, N. Preparation of the
1
6
10
3
3
compound according to General Procedure B yielded 39% of a
sample with identical spectral properties.
9
-Meth yl-7,12-d ih yd r o-p yr id o[2′,3′:2,3]a zep in o[4,5-b]-
in d ol-6(5H)-on e (66). Preparation following General Proce-
dure A from 6,7,8,9-tetrahydro-5H-pyrido[2,3-b]azepine-5,8-
dione (79) and 4-methylphenylhydrazine hydrochloride yielded
64% light yellow needles from ethanol; mp > 330 °C (dec
9
-Br om o-2-iod o-7,12-d ih yd r o-in d olo[3,2-d ][1]ben za ze-
1
p in -6(5H)-on e (53). Preparation following General Procedure
starting at 322 °C); H NMR 2.41 (s, 3H), 3.58 (s, 2H), 7.03
2
3
A from 7-iodo-1H-[1]benzazepine-2,5(3H,4H)-dione and 4-bro-
(dd, 1H, 8.1/1.0 Hz), 7.31-7.35 (m, 2H), 7.47 (s, 1H), 8.14 (dd,
1H, 7.6/1.5 Hz), 8.41 (dd, 1H, 4.6/1.5 Hz), 10.30 (s, 1H), 11.55
(s, 1H); Anal. (C H N O); C, H, N.
mophenylhydrazine hydrochloride yielded 56% beige crystals
1
from ethanol/toluene, mp > 330 °C; H NMR: 3.54 (s, 2H),
1
6
13
3
7
1
2
.05 (d, 8.6 Hz, 1H), 7.29 (dd, 2.0/8.6 Hz, 1H), 7.40 (d, 8.6 Hz,
H) 7.71 (dd, 2.0/8.6 Hz, 1H), 7.93 (d, 2.0 Hz, 1H), 8.07 (d,
.0 Hz, 1H) 10.19 (s, 1H), 11.88 (s, 1H); Anal. (C16H10BrIN O);
2
9
-Meth oxy-7,12-d ih yd r o-p yr id o[2′,3′:2,3]a zep in o[4,5-b]-
in d ol-6(5H)-on e (59). A mixture of 6,7,8,9-tetrahydro-5H-
pyrido[2,3-b]azepine-5,8-dione (79) (140 mg; 0.8 mmol), 4-meth-
oxyphenylhydrazine hydrochloride (210 mg, 1.2 mmol), sodium
acetate (98 mg; 1.2 mmol), and glacial acetic acid (2 mL) was
stirred 2 h at 70 °C. The mixture was poured into 5% aqueous
sodium acetate solution (20 mL). The precipitate was filtered
off, washed with water, and crystallized from ethanol/toluene
to yield 175 mg (78%) orange crystals, mp > 330 °C, (dec
C, H, N, Br, I.
-Meth oxy-9-tr iflu or m eth yl-7,12-d ih yd r o-in d olo[3,2-d ]-
1]ben za zep in -6(5H)-on e (55). Prepared following General
Procedure A from 8-methoxy-1H-[1]benzazepine-2,5(3H,4H)-
3
[
3
5
dione and 4-trifluorophenylhydrazine. Crystallization from
1
ethanol yielded 25% of a yellow powder, mp > 330 °C; H NMR
1
3
8
8
.59 (s, 2H), 3.82 (s, 3H), 6.85 (d, 1H, 2.6 Hz), 6.95 (dd, 1H,
.9/2.3 Hz), 7.43 (m, 1H), 7.59 (d, 1H, 8.6 Hz), 7.70 (d, 1H,
.7 Hz), 8.09 (s, 1H), 10.01 (s, 1H), 11.98 (s, 1H); Anal.
starting at 295 °C); H NMR 3.62 (s, 2H), 3.81 (s, 3H), 6.83
(dd, 1H, 8.9/2.3 Hz), 7.21 (d, 1H, 2.6 Hz), 7.31-7.35 (m, 2H),
8.14 (dd, 1H, 7.6/1.5 Hz), 8.41 (dd, 1H, 4.6/1.5 Hz), 10.33 (s,
(
C
18
H
13
F
3
N
2
O
2
); C, H, N.
-Hyd r oxy-8-oxo-8,9-d ih yd r o-7H-p yr id o[2,3-b]a zep in e-
-ca r boxylic Acid Eth yl Ester (78). Sodium hydride (640
13 3 2
1H), 11.55 (s, 1H); Anal. (C16H N O ) C, H, N.
8-Br om o-6,11-d ih yd r o-th ien o[3′,2′:2,3]a zep in o[4,5-b]in -
d ol-5(4H)-on e (71). Preparation according to General Proce-
5
6
5
6
mg of a 60% suspension in white oil; 16 mmol) was suspended
in toluene (5 mL) and heated to 90 °C under nitrogen. A
solution of 2-aminopyridine-3-carboxylic acid ethyl ester (76)
dure A from 4H-thieno[3,2-b]azepine-5,8(6H,7H)-dione and
4-bromophenylhydrazine hydrochloride yielded 26% of a gray
powder after crystallization from ethanol/toluene, mp > 330
1
(664 mg, 4 mmol) and succinic acid diethyl ester (77) (1045
°C (dec starting at 293 °C); H NMR 3.54 (s, 2H), 6.94 (d, 1H,
mg, 6 mmol) in toluene (20 mL) was added dropwise, and
stirring was continued at 90 °C. The reaction was monitored
by TLC (eluent: ethyl acetate). After all the sodium hydride
had reacted and completion of the reaction was detected
5.6 Hz), 7.24 (dd, 1H, 8.6/2.0 Hz), 7.34 (d, 1H, 8.6 Hz), 7.62 (d,
1H, 5.1 Hz), 7.86 (d, 1H, 1.5 Hz), 10.37 (s, 1H), 11.73 (s, 1H);
Anal. (C14
9 2
H BrN OS); C, H, N, Br, S.
8-Meth yl-6,11-d ih yd r o-th ien o[3′,2′:2,3]a zep in o[4,5-b]in -
d ol-5(4H)-on e (72). Preparation according to General Proce-
(10 h), the mixture was cooled to room temperature. A 15%
5
6
solution of acetic acid in water (20 mL) was added dropwise.
A precipitate was formed, which was filtered off and washed
successively with water and hexanes. Crystallization from
ethanol yielded 387 mg (39%) colorless needles, mp 211.5 °C;
dure A from 4H-thieno[3,2-b]azepine-5,8(6H,7H)-dione and
4-methyl-phenylhydrazine hydrochloride yielded 31% brown
crystals from ethanol, mp > 330 °C (dec starting at 260 °C);
1
H NMR 2.40 (s, 3H, CH ), 3.48 (s, 2H), 6.92 (d, 1H, 5.6 Hz),
3
1
H NMR 1.31 (t, 3H, 7.1 Hz), 3.03 (s, 2H), 4.31 (q, 2H, 7.1 Hz),
6.96 (dd, 1H, 8.4/1.3 Hz), 7.26 (d, 1H, 8.1 Hz), 7.39 (s, 1H),
7.57 (d, 1H, 5.6 Hz), 10.31 (s, 1H), 11.37 (s, 1H); Anal.
(C H N OS); C, H, N, S.
7
.34 (dd, 1H, 8.1/4.6 Hz), 8.23 (dd, 1H, 7.9/1.8 Hz), 8.58
(
(
dd, 1H, 4.6/2.0 Hz), 10.72 (bs, 1H), 12.47 (bs, 1H); Anal.
1
5
12
2
C
12
H
12
N
2
O
4
) C, H, N.
8
-Ch lor o-6,11-d ih yd r o-th ien o[3′,2′:2,3]a zep in o[4,5-b]in -
6
,7,8,9-Te t r a h yd r o-5H -p yr id o[2,3-b]a ze p in e -5,8-d i-
d ol-5(4H)-on e (73). Preparation according to General Proce-
5
6
on e (79). A solution of 5-hydroxy-8-oxo-8,9-dihydro-7H-pyrido-
2,3-b]azepine-6-carboxylic acid ethyl ester (78) (372 mg, 1.5
mmol) and water (0.5 mL) in DMSO (15 mL) was stirred at
50 °C under nitrogen. Every 2 h, portions of water (0.5 mL
each) were added. After 8 h, the mixture was poured into water
40 mL) and stored at 4 °C overnight. A precipitate was formed,
dure A from 4H-thieno[3,2-b]azepine-5,8(6H,7H)-dione and
4-chlorophenylhydrazine hydrochloride yielded 32% of a gray
powder from ethanol, mp > 330 °C (dec starting at 295 °C);
[
1
1
H NMR 3.71 (s, 2H), 6.94 (d, 1H, 5.1 Hz), 7.13 (dd, 1H, 8.6/
2.0 Hz), 7.38 (d, 1H, 8.6 Hz), 7.62 (d, 1H, 5.1 Hz), 7.73 (d, 1H,
2.0 Hz), 10.39 (s, 1H), 11.73 (s, 1H); Anal. (C H ClN OS); C,
(
1
4
9
2
which was filtered off, washed with water, and crystallized
H, N.
from ethanol to yield beige crystals (124 mg, 81%), mp 221
Gen er a l P r oced u r e B for P r ep a r a tion of P a u llon es 64,
5, 67, 75 by Th er m a l-In d u ced F isch er In d ole Rea ction
1
°
C; H NMR 2.74-2.77 and 2.93-2.96 (m, AA′XX‘, 4H), 7.26
6
(
dd, 1H, 7.6/4.6 Hz), 8.26 (dd, 1H, 7.9/1.8 Hz), 8.56 (dd, 1H,
.6/1.5 Hz), 10.39 (bs, 1H); Anal. (C ) C, H, N.
-Ch lor o-7,12-d ih yd r o-p yr id o[2′,3′:2,3]a zep in o[4,5-b]-
fr om Ar ylh yd r a zon es. A mixture of an appropriate cyclic
ketone (2 mmol), an appropriate arylhydrazine (3 mmol) [or
the appropriate substituted arylhydrazine hydrochloride (3
mmol) and sodium acetate (246 mg, 3 mmol)], and glacial acetic
acid (20 mL) is stirred for 1 h at 70 °C. After being cooled to
room temperature, the mixture is poured into a 5% aqueous
sodium acetate solution (40 mL). An arylhydrazone precipi-
tates, which is filtered off and purified by crystallization from
ethanol. The arylhydrazone (1 mmol) is then refluxed in the
4
9 8 2 2
H N O
9
in d ol-6(5H)-on e (61). Prepared following General Procedure
A from 6,7,8,9-tetrahydro-5H-pyrido[2,3-b]azepine-5,8-dione
(79) and 4-chlorophenylhydrazine hydrochloride. Crystalliza-
tion from ethanol yielded 51% of orange crystals, mp > 330
1
°
C; H NMR 3.64 (s, 2H), 7.19 (dd, 1H, 8.6/1.5 Hz), 7.35 (dd,
H, 7.9/4.8 Hz), 7.46 (d, 1H, 8.6 Hz), 7.82 (d, 1H, 1.5 Hz), 8.16
1

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 33
given amount of diphenyl ether under nitrogen. The reaction
is monitored by TLC (eluent: ethyl acetate or toluene/acetone
dihydro-indolo[3,2-d][1]benzazepine-6(5H)-thione (68) (108 mg,
0.35 mmol) and sodium hydride (15 mg of a 60% suspension
in white oil) in dry THF (10 mL) was stirred for 15 min under
nitrogen. After addition of iodomethane (60 mg, 0.42 mmol),
the mixture was refluxed under nitrogen. After 1, 1.5, and 3
h, the addition of iodomethane (60 mg, 0.42 mmol) was
repeated. Furthermore, sodium hydride (15 mg of a 60%
suspension in white oil) was added after 2 h of refluxing. The
reaction was monitored by TLC (eluent: ethyl acetate). After
completion of the reaction 4 h, the mixture was poured into
ice water (50 mL). The precipitate was filtered off and
crystallized from ethanol to yield 55 mg (47%) light brown
1
2
:1, respectively). After completion of the reaction, which takes
-13 h, the solution is cooled to room temperature. Hexane
is added until a solid precipitates. The precipitate is filtered
off and purified by crystallization from the given solvent.
8
-Nitr o-6,11-d ih yd r o-th ien o[3′,2′:2,3]a zep in o[4,5-b]in -
d ol-5(4H)-on e (75). According to General Procedure B, 4H-
5
6
thieno[3,2-b]azepine-5,8(6H,7H)-dione and 4-nitrophenylhy-
drazine hydrochloride were reacted to give an arylhydrazone
(51%; mp 277 °C (dec)), which was converted in refluxing
diphenyl ether (150 mL, 3 h) to yield 189 mg (63%) of a brown
1
1
powder after crystallization from ethanol, mp > 330 °C; H
crystals, mp 237 °C; H NMR 2.33 (s, 3H), 2.84 (bs, 1H), 3.95
NMR 3.67 (s, 2H), 6.97 (d, 1H, 5.1 Hz), 7.54 (d, 1H, 8.7 Hz),
(s, 3H), 4.31 (bs, 1H), 7.33 (“dt”, 1H, 7.6/1.3 Hz), 7.41 (d, 1H,
8.1 Hz), 7.49 (“dt”, 1H, 7.6/1.0 Hz), 7.77 (d, 1H, 9.2 Hz), 7.83
(d, 1H, 7.6 Hz), 8.13 (dd, 1H, 9.2/2.5 Hz), 8.88 (d, 1H, 2.0 Hz);
Anal. (C H N O S) C, H, N.
7
.70 (d, 1H, 5.1 Hz), 8.04 (dd, 1H, 8.9/2.3 Hz), 8.68 (d, 1H, 2.0
Hz), 10.50 (s, 1H), 12.34 (s, 1H); EI/HRMS (C14 S) calcd
99.0365, found 299.0363.
-Meth oxy-9-n itr o-7,12-dih ydr o-in dolo[3,2-d][1]ben zaze-
p in -6(5H)-on e (64). According to General Procedure B,
9 3 3
H N O
2
1
8
15
3
2
3
2
-(2-P r op en yl)-9-tr iflu or om eth yl-7,12-d ih yd r oin d olo-
3,2-d ][1]ben za zep in -6(5H)-on e (80). To a mixture of 2-iodo-
-trifluoromethyl-7,12-dihydro-indolo[3,2-d][1]benzazepin-6(5H)-
[
9
3
5
8
-methoxy-1H-(1)benzazepine-2,5(3H,4H)-dione and 4-nitro-
phenylhydrazine hydrochloride were reacted to give an aryl-
hydrazone (64%, mp 283-284 °C (dec)), which was converted
in refluxing diphenyl ether (40 mL, 2h) to yield the title
compound. Purification by column chromatography (ethyl
acetate) and subsequent crystallization from ethanol/toluene
23
one (38) (482 mg, 1.09 mmol), PdCl
2 3 2
(PPh ) (70 mg, 0.1
mmol), PPh (52 mg, 0.2 mmol) and 2,4-di-tert-butyl-4-
3
methylphenol (6 crystals) in DMF (10 mL) was added allyl-
tributyltin (410 mg, 1.24 mmol) under nitrogen. The corre-
sponding mixture was stirred at room temperature for 5 min
and then heated to 95 °C for 1.5 h. After being cooled to room
temperature, the mixture was poured into ice water. A solid
precipitated, which was filtered off with suction, washed with
cyclohexane and water, and purified by column chromatogra-
phy to give 132 mg of 2-tributylstannyl-9-trifluoromethyl-
1
yielded 13% dark brown crystals, mp > 330 °C; H NMR 3.63
(s, 2H), 3.82 (s, 3H), 6.86 (d, 1H, 2.0 Hz), 6.96 (dd, 1H, 8.9/2.3
Hz), 7.56 (d, 1H, 9.1), 7.70 (d, 1H, 8.6 Hz), 8.04 (dd, 1H, 9.2/
.0), 8.69 (d, 1H, 2.0 Hz), 10.13 (s, 1H), 12.29 (s, 1H);
(323.31); HRFAB-MS: calcd 323.0906; found
2
17 13 3 4
C H N O
3
23.0909.
-Meth oxy-9-n itr o-7,12-dih ydr o-in dolo[3,2-d][1]ben zaze-
p in -6(5H)-on e (65). According to General Procedure B,
7
,12-dihydroindolo[3,2-d][1]benzazepin-6(5H)-one (20%) as a
2
side product and 273 mg (70%) of the title compound, mp >
1
3
1
7
30 °C; H NMR 3.45 (d, 6.6 Hz, 2H), 3.59 (s, 2H), 5.11 (dd,
0.0/1.0 Hz, 1H), 5.16 (dd, 16.8/1.6 Hz, 1H), 5.99-6.09 (m, 1H),
.20-7.26 (m, 2H), 7.45 (dd, 8.6 Hz/1.0 Hz, 1H), 7.59-7.62
3
5
7
-methoxy-1H-(1)benzazepine-2,5(3H,4H)-dione and 4-nitro-
phenylhydrazine hydrochloride were reacted to give an aryl-
hydrazone (60%, mp 274 °C (dec)), which was converted in
refluxing diphenyl ether (40 mL, 2h) to yield the title com-
pound. Purification by column chromatography (ethyl acetate)
(
m, 2H), 8.13 (s, 1H), 10.10 (s, 1H), 12.06 (s, 1H); Anal.
OF ) C, H, N.
-(2-Oxop r op yl)-9-tr iflu or om eth yl-7,12-d ih yd r oin d olo-
3,2-d ][1]ben za zep in -6(5H)-on e (60). To a solution of 2-(2-
(C
20
H
15
N
2
3
2
and subsequent crystallization from ethanol/toluene yielded
[
1
2
3
7
3% dark brown crystals, mp > 330 °C; H NMR 3.61 (s, 2H),
.86 (s, 3H), 7.06 (dd, 1H, 8.9/2.8 Hz), 7.22 (d, 1H, 8.6 Hz),
.32 (d, 1H, 2.6 Hz), 7.61 (d, 1H, 8.6 Hz), 8.08 (dd, 1H, 9.2/2.0
propenyl)-9-trifluoromethyl-7,12-dihydroindolo[3,2-d][1]-
benzazepin-6(5H)-one (80) (108 mg, 0.3 mmol) in DMF (6 mL)
and H
2 2 2
O (1 mL) was added PdCl (6 mg, 0.03 mmol) and CuCl
Hz), 8.76 (d, 1H, 1.5 Hz), 10.03 (s, 1H), 12.41 (s, 1H); Anal.
); H, N; C calcd. 63.16, found 62.55.
,12-Dih yd r o-p yr id o[3′,2′:4,5]p yr r olo[3,2-d ][1]ben za ze-
p in -6(5H)-on e (67). 1H-[1]Benzazepine-2,5(3H,4H)-dione (175
(48 mg, 0.36 mmol). After the mixture was stirred overnight
17 13 3 4
(C H N O
at room temperature, it was poured into 30 mL of water and
extracted with ethyl acetate (50 mL × 3). The organic layers
were combined, washed with water (10 mL × 3), and dried
7
5
6
mg, 1 mmol) and 2-hydrazinopyridine (164 mg, 1.5 mmol)
were stirred in 10 mL glacial acetic acid at 70 °C for 7 h. Water
2 4
over Na SO . Evaporation of the solution yielded a residue,
which was purified by flash chromatography (ethyl acetate:
(50 mL) was added and the mixture was neutralized with
petroleum ether ) 70:30) to yield 78 mg (69%) of the title
concentrated ammonium hydroxide solution. An arylhydrazone
precipitated, which was filtered off and purified by crystal-
lization from ethanol to yield 51% colorless crystals, mp 233
1
compound, mp 295-297 °C (ethanol); H NMR 2.20 (s, 3H),
3
1
1
.58 (s, 2H), 3.78 (s, 2H), 7.18-7.30 (m, 2H), 7.46 (d, 8.6 Hz,
H), 7.57 (s, 1H), 7.61 (d, 8.1 Hz, 1H), 8.13 (s, 1H), 10.14 (s,
°
C. This material was converted following General Procedure
H), 12.07 (s, 1H); Anal. (C20
15 2 2 3
H N O F ) C, H, N.
B in refluxing diphenyl ether (20 mL, 2 h) to yield the title
compound. Crystallization from ethanol/toluene yielded 58%
Gen er a l P r oced u r e C for th e Syn th esis of Cya n o
1
Der iva tives 54, 63, 74, fr om Cor r esp on d in g Br om o Com -
p ou n d s. A mixture of the bromo-substituted paullone deriva-
tive (1 mmol) and copper(I) cyanide (180 mg, 2 mmol) in
N-methyl-2-pyrrolidone (10 mL) is refluxed for 2 h. After the
mixture is cooled to room temperature, water (10 mL) is added.
The precipitate is filtered off with suction and washed with
water. The precipitate is then suspended in a mixture of water
colorless crystals, mp > 330 °C; H NMR 3.52 (s, 2H), 7.13
(
7
1
dd, 8.1/4.6 Hz), 1H), 7.27 (“t”, 8.4 Hz, 2H), 7.38-7.42 (m, 1H),
.80 (d, 7.6 Hz, 1H), 8.12 (d, 7.6 Hz, 1H), 8.27 (dd, 4.6/1.0 Hz,
H), 10.16 (s, 1H), 12.20 (s, 1H); Anal. (C15 O) C, H, N.
-Nitr o-7,12-dih ydr o-in dolo[3,2-d][1]ben zazepin e-6(5H)-
11 3
H N
9
th ion e (68). To a stirred solution of 9-nitro-7,12-dihydro-
indolo[3,2-d][1]benzazepin-6(5H)-one (alsterpaullone (1); 293
mg, 1 mmol) in dry THF (30 mL) were added successively P S
2 5
(
10 mL) and 1,2-diaminoethane (2.5 mL). After stirring for 15
min, the solid is filtered off with suction, washed twice with
0% aqueous sodium cyanide solution, and purified by crystal-
(250 mg; 1.12 mmol) and NaHCO
3
(370 mg; 4.4 mmol) at 50
1
°
C under nitrogen. The mixture was refluxed for 3 h, cooled
lization from the given solvent.
to room temperature, and poured on crushed ice (50 mL). The
precipitate was filtered off, washed with water, and crystal-
lized from ethanol to yield 136 mg (44%) of a yellow powder,
3-Met h oxy-6-oxo-5,6,7,12-t et r a h yd r o-in d olo[3,2-d ][1]-
ben za zep in e-9-ca r bon itr ile (63). Preparation following Gen-
eral Procedure C from 9-bromo-3-methoxy-7,12-dihydro-indolo-
[3,2-d][1]benzazepin-6(5H)-one35 yielded 25% brown crystals
1
mp > 330 °C (dec starting at 285 °C); H NMR 4.04 (s, 2H),
7
8
1
.43-7.52 (m, 3H), 7.63 (d, 1H, 8.6 Hz), 7.83 (d, 1H, 7.1 Hz),
.10 (dd, 1H, 9.1/2.0 Hz), 8.68 (d, 1H, 2.5 Hz), 12.17 (s, 1H),
after purification by column chromatography (ethyl acetate)
1
2.51 (s, 1H); Anal. (C16
2-Met h yl-6-m et h ylt h io-9-n it r o-7,12-d ih yd r o-in d olo-
3,2-d ][1]ben za zep in e (69). A suspension of 9-nitro-7,12-
H
11
N
3
O
2
S) C, H, N.
and crystallization from ethanol/toluene; mp > 330 °C;
H
1
NMR 3.57 (s, 2H), 3.82 (s, 3H), 6.85 (d, 1H, 2.52 Hz), 6.94 (dd,
1H, 8.64/2.56 Hz), 7.48 (dd, 1H, 8.36/1.26 Hz), 7.56 (d, 1H, 8.64
[

3
4
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
Hz), 7.69 (d, 1H, 8.64 Hz), 8.27 (s, 1H), 10.10 (s, 1H), 12.13 (s,
H); HRFAB-MS (C18 ) calcd 303.1008; found 303.1012.
-Oxo-4,5,6,11-tetr a h yd r o-th ien o[3′,2′:2,3]a zep in o[4,5-
BSA per mL 10 mM dithiothreitol, with 5 µL 40 µM GS-1
peptide as a substrate, in buffer A, in the presence of 15 µM
1
13 3 2
H N O
3
2
-1
-1
5
[γ- P]ATP (3000 Ci‚mmol ; 1 mCi‚mL ) in a final volume
of 30 µL. The GS-1 peptide (YRRAAVPPSPSLSRHSSPHQS-
pEDEEE) was synthesized by the Peptide Synthesis Unit,
Institute of Biomolecular Sciences, University of Southampton,
Southampton SO16 7PX, U.K. After 30 min incubation at 30
°C, 25 µL aliquots of supernatant were spotted onto 2.5 × 3
cm pieces of Whatman P81 phosphocellulose paper, and, 20 s
later, the filters were washed five times (for at least 5 min
each time) in a solution of 10 mL ofphosphoric acid per liter
of water. The wet filters were counted in the presence of 1 mL
of ACS (Amersham) scintillation fluid.
b]in d ole-8-ca r bon itr ile (74). Preparation following General
Procedure C from 8-bromo-6,11-dihydro-thieno[3′,2′:2,3]aze-
pino[4,5-b]indol-5(4H)-one (53) yielded 23% of a beige powder
1
after crystallization from ethanol, mp > 330 °C; H NMR 3.60
(s, 2H), 6.96 (d, 1H, 5.1 Hz), 7.48 (dd, 1H, 8.6/1.6 Hz), 7.55 (d,
1
1
2
H, 8.1 Hz), 7.67 (d, 1H, 5.1 Hz), 8.27 (s, 1H), 10.46 (s, 1H),
2.15 (s, 1H); EI/HRMS (C15
79.0458.
9 3
H N OS) calcd 279.0466; found
6
-Oxo-5,6,7,12-tetr a h yd r o-in d olo[3,2-d ][1]ben za zep in e-
2
-ca r bon itr ile (54). Preparation following General Procedure
CDK1/Cyclin B. CDK1/cyclin B was extracted in homog-
enization buffer from M phase starfish (Marthasterias glacia-
C from 2-bromo-7,12-dihydro-indolo[3,2-d][1]benzazepin-6(5H)-
2
2
one yielded 47% beige crystals after crystallization from
CKShs1
1
lis) oocytes and purified by affinity chromatography on p9
-
ethanol/toluene, mp > 330 °C; H NMR 3.62 (s, 2H), 7.09-
CKShs1
Sepharose beads, from which it was eluted by free p9
as
7
7
.12 (m, 1H) 7.19-7.24 (m, 1H), 7.39 (d, 8.1 Hz, 1H), 7.46 (d,
.6 Hz, 1H), 7.71 (d, 7.6 Hz, 1H), 7.81 (dd, 8.4/1.8 Hz, 1H), 8.2
previously described.5
8,59
The kinase activity was assayed in
buffer C, with 1 mg histone H1 (Type III-S, Sigma Chemicals)
11 3
(d, 2.0 Hz, 1H), 10.56 (s, 1H), 11.72 (s, 1H); Anal. (C17H N O)
3
2
-1
per mL, in the presence of 15 µM [γ- P]ATP (3000 Ci‚mmol
;
C, H; N: calcd 15.38, found 14.84.
-
1
1
mCi‚mL ) in a final volume of 30 µL. After a 10-min
9
-Am in o-7,12-dih ydr o-in dolo[3,2-d][1]ben zazepin -6(5H)-
incubation at 30 °C, 25 µL aliquots of supernatant were spotted
onto P81 phosphocellulose papers and treated as described
above.
CDK5/p 25. CDK5/p25 was reconstituted by mixing equal
amounts of recombinant mammalian CDK5 and p25 expressed
in E. coli as GST (glutathione S-transferase) fusion proteins
and purified by affinity chromatography on glutathione-
agarose (vectors kindly provided by J . H. Wang, Department
of Medical Chemistry, University of Calgary, Alberta, Canada)
on e (65). To a suspension of 9-nitro-7,12-dihydro-indolo[3,2-
d][1]benzazepin-6(5H)-one (alsterpaullone (1); 293 mg, 1 mmol)
in ethanol (10 mL) a suspension of CaCl (70 mg) and zinc
2
powder (2.1 g) in water (1.8 mL) was added. After being
refluxed for 1 h, the mixture was filtered. The filter cake was
extracted with ethyl acetate (10 mL). The combined organic
layers were dried by means of sodium sulfate and evaporated
to dryness. The residue was crystallized from ethanol to yield
1
3
4
1
1
1
5% brown crystals, mp 254 °C (dec); H NMR 3.33 (br. s. 2H),
.58 (s, 2H) 6.57 (dd, 1H, 8.4/1.8 Hz), 6.72 (m, 1H), 7.13 (d,
H, 8.6 Hz), 7.20-7.34 (m, 3H), 7.68 (d, 1H, 7.6 Hz), 10.01 (s,
(p25 is a truncated version of the p35, the 35-kDa CDK5
activator). Its activity was assayed in buffer C as described
for CDK1/cyclin B.
H), 11.06 (s, 1H); Anal. (C16
4.94; found 13.31.
H
13
N
3
2
O‚1H O); C, H; N: calcd
Ack n ow led gm en t. X.X. and C.K. gratefully ac-
knowledge financial funding by the National Cancer
Institute. This research was also supported by a grant
from the “Association pour la Recherche sur le Cancer”
(ARC 5343) (L.M.) and a grant (“Mol e´ cules & Cibles
Th e´ rapeutiques”) from the “Ministere de la Recherche/
INSERM/CNRS” (L.M.).
9
-Aceta m id o-7,12-d ih yd r o-in d olo[3,2-d ][1]ben za zep in -
6
(5H)-on e (58). To a refluxing solution of 9-nitro-7,12-dihydro-
indolo[3,2-d][1]benzazepin-6(5H)-one (alsterpaullone (1); 220
mg, 0.75 mmol) and water (0.35 mL) in glacial acetic acid (10
mL) was added iron powder (229 mg) portionwise over 75 min.
After 45 min, another portion of water (0.35 mL) was added.
After being refluxed for overall 6 h, the mixture was cooled to
room temperature and diluted with water (10 mL) and
extracted with ethyl acetate (3 × 10 mL). The combined
organic layers were washed with water and aqueous 10%
Refer en ces
Na
2
CO
3
solution, dried (Na
2
SO
4
), and evaporated. The residue
(1) Ali, A.; Hoeflich, K. P.; Woodgett, J . R. Glycogen synthase kinase-
3
: properties, function and regulation. Chem. Rev. 2001, 101,
was crystallized from ethanol to yield 37% of a gray powder,
mp > 330 °C; H NMR 2.05 (s, 3H), 3.41 (s, 2H), 7.24-7.38
m, 5H), 7.72 (d, 1H, 7.1 Hz), 7.94 (s, 1H), 9.81 (s, 1H), 10.10
1
2527-2540.
(
2) Doble, B. W.; Woodgett, J . R. GSK-3: tricks of the trade for a
(
(
multi-tasking kinase. J . Cell Sci. 2003, 116, 1175-1186.
(3) Frame, S.; Cohen, P. GSK3 takes center stage more than 20
s, 1H), 11.50 (s, 1H); Anal. (C18
H
15
N
3
O
2
) H, N; C: calcd 70.81;
found 70.30.
years after its discovery. Biochem. J . 2001, 359, 1-16.
(
4) 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.
Kin a se Assa ys. Rea gen ts. Homogenization buffer: 60 mM
â-glycerophosphate, 15 mM p-nitrophenyl phosphate, 25 mM
Mops (pH 7.2), 15 mM EGTA, 15 mM MgCl
threitol, 1 mM sodium vanadate, 1 mM NaF, 1 mM phenyl
phosphate, 10 µg leupeptin mL , 10 µg aprotinin mL , 10
2
, 1 mM dithio-
(5) Cohen, P. The Croonian Lecture 1998. Identification of a protein
cascade of major importance in insulin signal transduction.
Philos. Trans. R. Soc. London B 1999, 354, 485-495.
-
1
-1
-
1
µg soybean trypsin inhibitor mL , and 100 µg benzamidine.
Buffer A: 10 mM MgCl , 1 mM EGTA, 1 mM dithiothreitol,
5 mM Tris/HCl pH 7.5, 50 µg heparin mL-1. Buffer C:
(
6) Wagmann, A. S.; Nuss, J . M. Current therapies and emerging
2
targets for the treatment of diabetes. Curr. Pharm. Des. 2001,
2
7
, 417-450.
homogenization buffer but 5 mM EGTA, no NaF and no
protease inhibitors.
(7) Ryves, W. J .; Harwood: A. J . Lithium inhibits glycogen synthase
kinase-3 by competition for magnesium. Biochem. Biophys. Res.
Commun. 2001, 280, 720-725.
Kinase activities were assayed in duplicates in buffer A or
C at 30 °C, at a final ATP concentration of 15 µM. The order
of mixing the reagents was buffers, substrate, enzyme, and
inhibitor. There was no preincubation at 30 °C. Addition of
ATP was considered as time 0 of the incubation period. Assays
were run under conditions where less than 5% of the radio-
labeled phosphate was incorporated. Blank values were sub-
tracted and activities calculated as pmol of phosphate incor-
porated for a 10-min incubation. The activities are usually
expressed in percentage of the maximal activity, i.e., in the
absence of inhibitors. Controls were performed with appropri-
ate dilutions of DMSO.
(
8) Nikoulina, S. E.; Ciaraldi, T. P.; Mudaliar, S.; Carter, L.;
J ohnson, K.; Henry, R. R. Inhibition of glycogen synthase kinase
3
improves insulin action and glucose metabolism in human
skeletal muscle. Diabetes 2002, 51, 2190-2198.
(9) Naerum, L.; Norskov-Lauritsen, L.; Olesen, P. H. Scaffold
hopping and optimization towards libraries of glycogen synthase
kinase-3 inhibitors. Bioorg. Med. Chem. Lett. 2002, 12, 1525-
1
528.
(
10) Martinez, A.; Alonso, M.; Castro, A.; Perez, C.; Moreno, F. First
non-ATP competitive glycogen synthase kinase-3b (GSK-3b)
inhibitors: thiazolidinones (TDZD) as potential drugs for the
treatment of Alzheimer’s disease. J . Med. Chem. 2002, 45, 1292-
1
299.
(
11) Smith, D. G.; Ward, R. W. Preparation of diamino-1,2,4-triazole-
carboxylic acid and derivatives as GSK-3 inhibitors. PCT Int.
Appl. WO 0109106, 2001.
GSK-3â. GSK-3â was expressed in and purified from insect
5
7
sf9 cells. It was assayed, following a 1/100 dilution in 1 mg

3
D-QSAR CoMSIA Models
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1 35
(
12) Coghlan, M. P.; Culbert, A. A.; Cross, D. A. E.; Corcoran, S. L.;
Yates, J . D.; 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 . Selective
small molecule inhibitors of glycogen synthase kinase-3 modu-
late glycogen metabolism and gene transcription. Chem. Biol.
(30) Klebe, G.; Abraham, U.; Mietzner, T. Molecular similarity indices
in a comparativ Analysis (CoMSIA) of drug molecules to cor-
relate and predict their biological activity. J . Med. Chem. 1994,
37, 4130-4146.
(31) B o¨ hm, M.; St u¨ rzebecher, J .; Klebe, G. Three-Dimensional Quan-
titative Structure-Activity Relationship Analyses Using Com-
parative Molecular Field Analysis and Comparative Molecular
Similarity Indices Analysis To Elucidate Selectivity Differences
of Inhibitors Binding to Trypsin, Thrombin, and Factor Xa. J .
Med. Chem. 1999, 42, 458-477.
(32) Klebe, G.; Abraham, U. Comparative molecular similarity index
analysis (CoMSIA) to study hydrogen-bonding properties and
to score combinatorial libraries. J . Comput.-Aided Mol. Des.
1999, 13, 1-10.
(33) Ducrot, P.; Legraverend, M.; Grierson, D. S. 3D-QSAR CoMFA
on cyclin-dependent kinase inhibitors. J . Med. Chem. 2000, 43,
2
000, 7, 793-803.
(
13) Smith, D. G.; Buffet, M.; Fenwick, A. E.; Haigh, D.; Ife, R. J .;
Saunders: M.; Slingsby, B. P.; Stacey, R.; Ward, R. W. 3-Anilino-
4
-arymaleimides: potent and selective inhibitors of glycogen
synthase kinase-3 (GSK-3). Bioorg. Med. Chem. Lett. 2001, 11,
35-639.
6
(
14) Meijer, L.; Thunissen, A.-M. W. H.; White, A. W.; Garnier, M.;
Nikolic, M.; Tsai, L.-H.; Walter, J .; Cleverley, K. E.; Salinas, P.
C.; Wu, Y.-Z.; Biernat, J .; Mandelkow, E. M.; Kim, S.-H.; Pettit,
G. R. Inhibition of cyclin-dependent kinases, GSK-3â and CK1
by hymenialdisine, a marine sponge constituent. Chem. Biol.
4
098-4108.
(
34) Naumann, T.; Matter, H. Strucural classification of protein
kinases using 3D molecular interaction field analysis of their
ligand binding sites: target family landscapes. J . Med. Chem.
2
000, 7, 51-63.
(
15) Hoessel, R.; Leclerc, S.; Endicott, J .; Nobel, M. E. M.; Lawrie,
A.; Tunnah, P.; Leost, M.; Damiens, E.; Marie, D.; Marko, D.;
Niederberger, E.; Tang, W.; Eisenbrand, G.; Meijer, L. Indirubin,
the active constituent of a Chinese antileukaemia medicine,
inhibits cyclin-dependent kinases. Nature Cell Biol. 1999, 1, 60-
2
002, 45, 2366-2378.
(
35) Wieking, K.; Knockaert, M.; Leost, M.; Zaharevitz, D. W.; Meijer,
L.; Kunick, C. Synthesis of paullones with aminoalkyl side
chains. Arch. Pharm. Pharm. Med. Chem. 2002, 335, 311-317.
6
7.
(
36) Kunick, C. Synthese von 7,12-Dihydro-indolo[3,2-d][1]benzazepin-
(
16) Leclerc, S.; Garnier, M.; Hoessel, R.; Marko, D.; Bibb, J . A.;
Snyder, G. L.; Greengard, P.; Biernat, J .; Wu, Y.-Z.; Mandelkow,
E.-M.; Eisenbrand, G.; Meijer, L. Indirubins inhibit glycogen
kinase-3â and CDK5/p25, two protein kinases involved in
abnormal tau phosphorylation in Alzheimer’s disease. J . Biol.
Chem. 2001, 276, 251-260.
17) Damiens, E.; Baratte, B.; Marie, D.; Eisenbrand, G.; Meijer, L.
Anti-mitotic properties of indirubin-3′-monoxime, a CDK/GSK-3
inhibitor: induction of endoreplication following prophase arrest.
Oncogene 2001, 20, 3786-3797.
18) Marko, D.; Sch a¨ tzle, S.; Friedel, A.; Genzlinger, A.; Zankl, H.;
Meijer, L.; Eisenbrand, G. Inhibition of cyclin-dependent kinase
6
5
(5H)-onen und 6,11-Dihydro-thieno[3′,2′:2,3]azepino[4,5-b]indol-
(4H)-on. Arch. Pharm. (Weinheim) 1992, 325, 297-299.
(
37) Knockaert, M.; Wieking, K.; Schmitt, S.; Leost, M.; Grant, K.
M.; Mottram, J . C.; Kunick, C.; Meijer, L. Intracellular targets
of paullones: Identification following affinity purification on
immobilized inhibitor. J . Biol. Chem. 2002, 277, 25493-25501.
38) Nguyen, C. H.; Marchand, C.; Delage, S.; Sun, J .-S.; Garestier,
T.; H e´ l e` ne, C.; Bisagni, E. Synthesis of 13H-benzo[6,7]- and 13H-
benzo[4,5]indolo[3,2-c]quinolines: a new series of potent specific
ligands for triplex DNA. J . Am. Chem. Soc. 1998, 120, 2501-
(
(
(
(
(
2
507.
39) MacPhillamy, H. B.; Dziemian, R. L.; Lucas, R. A.; Kuehne, M.
E. The alkaloids of Tabernanthe iboga. Part VI. The synthesis
of the selenium dehydrogenation products from ibogamine. J .
Am. Chem. Soc. 1958, 80, 2172-2178.
40) von Braun, J .; Manz, G. Fluoranthen und seine Derivate. Liebigs
Ann. Chem. 1931, 488, 111-126.
41) Newman, M. S.; Boden, H. N-methylpyrrolidone as solvent for
reaction of aryl halides with cuprous cyanide. J . Org. Chem.
1
(CDK1) by indirubin derivatives in human tumour cells. Br.
J . Cancer 2001, 84, 283-289.
19) Mettey, Y.; Gompel, M.; Thomas, V.; Garnier, M.; Leost, M.;
Ceballos-Picot, I.; Noble, M.; Endicott, J .; Vierfond, J .-M.; Meijer,
L. Aloisines, a new family of CDK/GSK-3 inhibitors. SAR study,
crystal structure in complex with CDK2, enzyme selectivity, and
cellular effects. J . Med. Chem. 2003, 46, 222-236.
(
(
(
20) Leost, M.; Schultz, C.; Link, A.; Wu, Y.-Z.; Biernat, J .; Man-
delkow, E.-M.; Bibb, J . A.; Snyder, G. L.; Greengard, P.;
Zaharevitz, D. W.; Gussio, R.; Senderowicz, A. M.; Sausville, E.
A.; Kunick, C.; Meijer, L. Paullones are potent inhibitors of
glycogen synthase kinase-3b and cyclin-dependent kinase 5/p25.
Eur. J . Biochem. 2000, 267, 5983-5994.
21) Zaharevitz, D. W.; Gussio, R.; Leost, M.; Senderowicz, A. M.;
Lahusen, T.; Kunick, C.; Meijer, L.; Sausville, E. A. Discovery
and initial characterization of the paullones, a novel class of
small-molecule inhibitors of cyclin-dependent kinases. Cancer
Res. 1999, 59, 2566-2569.
1
961, 26, 2525.
(
(
(
(
42) SYBYL molecular modeling software; version 6.8; Tripos As-
sociated Ltd.: St. Louis, MO.
43) Stewart, J . J . MOPAC:
a semiempirical molecular orbital
program. J . Comput.-Aided Mol. Des. 1990, 4, 1-105.
44) Clark, M.; Cramer, R. D., III; Van Obdenbosch, N. The Tripos
Forcefield. J . Comput. Chem. 1989, 10, 982-1012.
45) Gerber, P. R.; M u¨ ller, K. MAB, a generally applicable molecular
force field for structure modelling in medicinal chemistry. J .
Comput.-Aided Mol. Des. 1995, 9, 251-268.
46) Gerber, P. R. Charge distribution from a simple molecular orbital
type calculation and non-bonding interaction terms in the force
field MAB. J . Comput.-Aided Mol. Des. 1998, 12, 37-51.
47) Wold, S.; Albano, C.; W. J . Dunn, I.; Ellund, U.; Esbensen, K.;
Geladi, P.; Hellberg, S.; J ohansson, E.; Lindberg, W.; Sjostrom,
M. Multivariate Data Analysis in Chemistry. Chemometrics:
Mathematics and Statistics in Chemistry; Reidel: Dordrecht,
Netherlands, 1984.
(
(
(
(
(
22) Schultz, C.; Link, A.; Leost, M.; Zaharevitz, D. W.; Gussio, R.;
Sausville, E. A.; Meijer, L.; Kunick, C. Paullones, a series of
cyclin-dependent kinase inhibitors: Synthesis, evaluation of
CDK1/Cyclin B inhibition, and in vitro antitumor activity. J .
Med. Chem. 1999, 42, 2909-2919.
(
23) Kunick, C.; Schultz, C.; Lemcke, T.; Zaharevitz, D. W.; Gussio,
R.; J alluri, R. K.; Sausville, E. A.; Leost, M.; Meijer, L.
2
-Substituted paullones: CDK1/cyclin B-inhibiting property and
in vivo antiproliferative activity. Bioorg. Med. Chem. Lett. 2000,
0, 567-569.
(
48) Bush, B. L.; Nachbar, R. B., J r. Sample-distance partial least
squares: PLS optimized for many variables, with application
to CoMFA. J . Comput.-Aided Mol. Des. 1993, 7, 587-619.
49) Thompson, J . D.; Higgins, D. G.; Gibson, T. J . CLUSTAL W:
improving the selectivity of progressive multiple sequence
alignment through sequence weighting, position-specific gap
penalties and weight matrix choise. Nucleic Acid Res. 1994, 22,
1
24) Gussio, R.; Zaharevitz, D. W.; McGrath, C. F.; Pattabiraman,
N.; Kellogg, G. E.; Schultz, C.; Link, A.; Kunick, C.; Leost, M.;
Meijer, L.; Sausville, E. A. Structure-based design modifications
of the paullone molecular scaffold for cyclin-dependent kinase
inhibition. Anti-Cancer Drug Des. 2000, 15, 53-66. Kenpaullone
was manually docked into the CDK1/cyclin B model and the
protein-ligand complex was iteratively minimized using several
cycles of minimization with the CFF91 force field and refining
with hydropathic analysis (HINT).
(
4
673-4680.
(
50) Okumura, E.; Sekiai, T.; Hisanaga, S.; Tachibana, K.; Kishimoto,
T. Initial triggering of M-phase in starfish oocytes: a possible
novel component of maturation-promoting factor besides cdc2
kinase. J . Cell. Biol. 1996, 132, 125-135.
(
25) Mandelkow, E. M.; Mandelkow, E. Tau in Alzheimer’s disease.
Trends Cell Biol. 1998, 8, 425-427.
(51) Meyerson, M.; Enders, G. H.; Wu, C. L.; Su, L. K.; Gorka, C.;
Nelson, C.; Harlow, E.; Tsai, L. H. A family of human cdc2-
related protein kinases. EMBO J . 1992, 11, 2909-2917.
(52) Stambolic, V.; Woodgett, J . R. Mitogen inactivation of glycogen
synthase kinase-3 beta in intact cells via serine 9 phosphoryla-
tion. Biochem. J . 1994, 303 (Pt 3), 701-704.
(53) Boeckmann, B.; Bairoch, A.; Apweiler, R.; Blatter, M. C.;
Estreicher, A.; Gasteiger, E.; Martin, M. J .; Michoud, K.;
O’Donovan, C.; Phan, I.; Pilbout, S.; Schneider, M. The SWISS-
PROT protein knowledgebase and its supplement TrEMBL in
2003. Nucleic Acid Res. 2003, 31, 365-370.
(54) Lemcke, T.; Kunick, C.; J oergensen, F. S. 3D-QSAR Analysis of
the CDK-inhibiting activity of paullones. Arch. Pharm. Pharm.
Med. Chem. 2001, 334, 41.
(
26) Patrick, G. N.; Zukerberg, L.; Nikolic, M.; de la Monte, S.; Dikkes,
P.; Tsai, L.-H. Conversion of p35 to p25 deregulates Cdk5 activity
and promotes neurodegeneration. Nature 1999, 402, 615-622.
27) Mandelkow, E. The tangled tale of tau. Science 1999, 402, 588-
(
(
5
89.
28) Coleman, K. G.; Lyssikatos, J . P.; Ynag, B. V. Chemical inhibitors
of cyclin-dependent kinases. Annual Reports in Medicinal Chem-
istry; Bristol, J . A., Ed.; Academic Press: San Diego, 1997; Vol.
3
2, pp 171-179.
(
29) Meijer, L. Cyclin-dependent kinase inhibitors as potential
anticancer, antineurodegenerative, antiviral and antiparasitic
agents. Drug Resist. Update 2000, 3, 83-88.

3
6
J ournal of Medicinal Chemistry, 2004, Vol. 47, No. 1
Kunick et al.
(
55) Zhu, L.; Hou, T.; Xu, X. Three-dimensional quantitative structure-
(58) Meijer, L.; Borgne, A.; Mulner, O.; Chong, J . P. J .; Blow, J . J .;
Inagaki, N.; Inagaki, M.; Delcros, J .-G.; Moulinoux, J .-P. Bio-
chemical and cellular effects of roscovitine, a potent and selective
inhibitor of the cyclin-dependent kinases cdc2, cdk2 and cdk5.
Eur. J . Biochem. 1997, 243, 527-536.
activity relationship study on paullones as cdk inhibitors using
CoMSIA and CoMFA. J . Mol. Model. 2001, 223-230.
(
56) Kunick, C. Synthese [b]-kondensierter Azepindione durch De-
alkoxycarbonylierung. Arch. Pharm. (Weinheim) 1991, 324, 579-
(
59) Borgne, A.; Ostvold, A. C.; Flament, S.; Meijer, L. Intra-M
phase promoting factor phosphorylation of cyclin B at the
prophase/metaphase transition. J . Biol. Chem. 1999, 274,
5
81.
(
57) Hughes, K.; Pulverer, B. J .; Theocharous, P.; Woodgett, J . R.
Baculovirus-mediated expression and characterisation of rat
glycogene synthase kinase-3b, the mammalian homologue of the
Drosophila melanogaster zestewhite homeotic gene product. Eur.
J . Biochem. 1992, 203, 305-311.
1
1977-11986.
J M0308904