Docking-based development of purine-like inhibitors of cyclin-dependent kinase-2

Article abstract of DOI:10.1021/jm990506w

The cell division cycle is controlled by cyclin-dependent kinases (cdk), which consist of a catalytic subunit (cdk1-cdk8) and a regulatory subunit (cyclin A-H). Purine-like inhibitors of cyclin-dependent kinases have recently been found to be of potential

Full text of DOI:10.1021/jm990506w

2
506
J . Med. Chem. 2000, 43, 2506-2513
Dock in g-Ba sed Develop m en t of P u r in e-lik e In h ibitor s of Cyclin -Dep en d en t
Kin a se-2
†
‡
§
|
‡
,⊥
Michal Otyepka, Vladim ´ı r Kry sˇ tof, Libor Havl ´ı cˇ ek, V eˇ ra Siglerov a´ , Miroslav Strnad, and J aroslav Ko cˇ a*
Department of Inorganic and Physical Chemistry, Faculty of Science, Palack y´ University, t rˇ . Svobody 8,
7
71 46 Olomouc, Czech Republic; Laboratory of Growth Regulators, Faculty of Science, Palack y´ University & Institute of
Experimental Botany, Sˇ lechtitel u˚ 11, 783 71 Olomouc, Czech Republic; Central Radioisotope Laboratory, Medical Faculty 1,
Charles University, Na boji sˇ ti 3, 121 08 Praha 2, Czech Republic; Isotope Laboratory, Institute of Experimental Botany,
V ı´ de nˇ sk a´ 1083, 14220 Praha 4, Czech Republic; and Laboratory of Biomolecular Structure and Dynamics, Faculty of Science,
Masaryk University, Kotl a´ rˇ sk a´ 2, 611 37 Brno, Czech Republic
Received October 6, 1999
The cell division cycle is controlled by cyclin-dependent kinases (cdk), which consist of a catalytic
subunit (cdk1-cdk8) and a regulatory subunit (cyclin A-H). Purine-like inhibitors of cyclin-
dependent kinases have recently been found to be of potential use as anticancer drugs. Rigid
and flexible docking techniques were used for analysis of binding mode and design of new
inhibitors. X-ray structures of three (ATP, olomoucine, roscovitine) cdk2 complexes were
available at the beginning of the study and were used to optimize the docking parameters.
The new potential inhibitors were then docked into the cdk2 enzyme, and the enzyme/inhibitor
interaction energies were calculated and tested against the assayed activities of cdk1 (37
compounds) and cdk2 (9 compounds). A significant rank correlation between the activity and
the rigid docking interaction energy has been found. This implies that (i) the rigid docking can
be used as a tool for qualitative prediction of activity and (ii) values obtained by the rigid
docking technique into the cdk2 active site can also be used for the prediction of cdk1 activity.
While the resulting geometries obtained by the rigid docking are in good agreement with the
X-ray data, the flexible docking did not always produce the same inhibitor conformation as
that found in the crystal.
In tr od u ction
The cell division cycle is controlled by cyclin-depend-
ent kinases (cdk), which consist of a catalytic subunit
(cdk1-cdk8) and a regulatory subunit (cyclin A-cyclin
H). These proteins are regulated in several ways:
subunit production, complex formation, (de)phosphoryl-
ation, cellular localization, and interaction with various
natural protein inhibitors. Recently, a deregulation of
cdks has been proved in human primary tumors and in
1
tumor cell lines. The discovery evoked a strong interest
in inhibitors of cdk that could play a role in the therapy
of cancers. Several types of cdk inhibitors have so far
2
been described: staurosporine, flavopiridole (L86-
3
4
5
6
8
275), butyrolactone-1, purine derivatives, indirubin,
7
8,9
paullones, and others.
Molecular modeling techniques that can help us to
predict the leading compound are more and more
frequently used in the development of medicinal drugs.
Many drugs produce their effect by interacting with a
biological target such as active sites of enzymes or DNA,
and the interaction is in many cases caused by nonbond
forces. Therefore, molecular docking techniques based
on a nonbond force field can be used to predict orienta-
F igu r e 1. A Kabsch-Sander model of the structure of ATP-
cdk2 complex. N and C termini of the protein are denoted by
N and C, respectively; ATP molecule is shown with coordinated
Mg2 ion. The C-terminal domain consists primarily of R-he-
lices (cylinders) and N-terminal domain consists predomi-
nantly of â-sheets (ribbons).
+
1
0
tion and complementarity of the ligand to its target.
Molecular docking starts with the knowledge of the
active site or even with information about the structure
of the ligand-receptor complex. In the past few years,
the structures of some cdk2 complexes have been
*
Corresponding author. Tel: 420 5 41129310. Fax: 420 5 41211214.
11,12
published.
The cdk2 that regulates the G1/S cell
E-mail: jkoca@chemi.muni.cz.
†
cycle phase transition and DNA replication consists of
two lobes, a smaller N-terminal domain and a larger
C-terminal domain, and ATP and all inhibitors bind in
the deep cleft between the lobes (Figures 1 and 2).
Palack y´ University.
Palack y´ University & Institute of Experimental Botany.
Charles University.
Institute of Experimental Botany.
‡
§
|
⊥
Masaryk University.
1
0.1021/jm990506w CCC: $19.00 © 2000 American Chemical Society
Published on Web 06/03/2000

Purine-like Inhibitors of cdk2
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13 2507
F igu r e 2. A stereoview of cdk2 (backbone) complexed with roscovitine (represented by thick line).
N1...HN(Leu83). The roscovitine binding mode (Figure
3
C) is characterized also by two H-bonds, (roscovitine)-
6
N7...HN(Leu83) and (roscovitine)N H...OC(Leu83). Pur-
valanol B, an efficient inhibitor, employs the roscovitine
binding mode and, moreover, it creates another H-bond
between the acidic C8 carbon and kinase backbone,
(purvalanol B)C8H...OC(Glu81). Finally, the IPA bind-
ing mode (Figure 3A) is defined by three H-bonds,
(
(
IPA)N3...HN(Leu83), (IPA)N9H...OC(Glu81), and
IPA)N7...HN(terminal group of Lys33). Recently, an
X-ray paper dealing with the determinants of the
1
4
binding modes has been published.
Here we report on the testing of molecular docking
techniques on cdk2. Using rigid docking, we predict
activities and ways to enhance them for the set of new
inhibitors based on 2,6,9-purine derivatives.
F igu r e 3. Comparison of (A) 6-(3,3-dimethylallylamino)purine
IPA), (B) ATP, and (C) olomoucine orientation (thick lines)
(
with roscovitine (thin lines) orientation in the active site.
Structures of (D) 6-(3,3-dimethylallylamino)purine, (E) olo-
moucine, and (F) roscovitine.
Molecu la r Mod elin g Meth od s
Recep tor P r ep a r a tion . X-ray structures of the apo-
cdk2 (Brookhaven Protein Data Bank access code 1hcl),
the ATP-cdk2 complex (1hck), the ATP-activated cdk2/
cyklinA (1jst), and the purvalanol B-cdk2 (1ckp) com-
plex were obtained from the Protein Data Bank. X-ray
structures of the cdk2-roscovitine (complex I),
-olomoucine (II), and -6-(3,3-dimethylallylamino)-
purine (IPA) (III) complexes were kindly provided by
Prof. S.-H. Kim. Receptors were prepared for docking
in such a way that all heteroatoms (i.e., nonreceptor
atoms such as water, ions, etc.) were removed. This was
done with one exception, complex I, where the 534th
water molecule was kept because it creates a strong
hydrogen bond with the enzyme and is positioned deep
There are no significant differences in domain orien-
tations between the ligand-enzyme complexes and the
apoenzyme, but there are considerable differences in the
orientation of different types of ligand.13 The differences
in the binding of ATP and adenine-based inhibitors are
shown in Figure 3. The molecules of 2,6,9-substituted
purines, 6-[(3-chloro)anilino]-2(R)-[[1-(hydroxymethyl)-
2
-(methyl)propyl]amino]-9-isopropylpurine (purvalanol
B), 6-(benzylamino)-2-[(hydroxyethyl)amino]-9-methyl-
purine (olomoucine, Figure 3E), and 6-(benzylamino)-
2
(R)-[[1-(hydroxymethyl)propyl]amino]-9-isopropylpu-
rine (roscovitine, Figure 3F) are very similarly posi-
tioned in the active site. On the other hand, the
positions of mono- (isopentenyladenine (IPA), Figure 3D)
and di- (ATP) substitued purines in the cavity are very
different from each other. This supports the conclusion
that a similar substitution of 2,6,9-substituted purines
should lead to a similar positioning in the active site,
and, consequently, to the same binding mode. Each
binding mode of purine derivatives is characterized by
several H-bonds to the backbone of the kinase active
site. The ATP binding mode (Figure 3B) is characterized
1
5
in the cleft. Insight II software (MSI) was used to add
hydrogen atoms and to generate Sybyl (Tripos) mol2
1
6
files. The receptor active site for DOCK was repre-
sented by spheres that were constructed on the Connolly
surface of the receptor with 1 Å probe size. The spheres
positioned outside the active site were deleted manually.
Table 1 shows the most important rigid docking param-
eters. A typical rigid docking experiment consumed
2
about 400 s of computer time (SGI Indigo Extreme,
R4400).
6
by two H-bonds, (ATP)N H...OC(Glu81) and (ATP)-

2
508 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13
Otyepka et al.
Ta ble 1. Parameters for Rigid Docking (DOCK 4.0)
parameter
value
parameter
value
parameter
value
bump•maximum
energy•cutoff•distance
dielectric•factor
4
15
4
yes
yes
yes
attractive•exponent
repulsive•exponent
energy•minimize
6
12
yes
10
0.1
100
total•orientations
nodes•minimum
nodes•maximum
distance•tolerance
distance•minimum
grid•points
5000
4
10
distance•dependent
random•search
maximum•cycles
0.25
2
energy•convergence
maximum•iterations
6
uniform•sampling
10
Ta ble 2. Parameters for Rigid and Flexible Docking (AutoDock
.4)
2
parameter
value
parameter
cycles
steps accepted
steps rejected
value
initial RT
RT reduction factor
number of runs
616
0.95
15
55
35000
35000
Liga n d P r ep a r a tion . The ligands were minimized
1
7
by the semiempirical PM3 method (HyperChem ). The
crystal structure conformations of roscovitine, olomou-
cine, and purvalanol B were used as initial geometries
for the optimization. A 0.05 kcal/molÅ gradient was used
as the terminating condition; there were no significant
differences in interaction energies on using a lower
gradient. Empirical partial atomic charges were taken
from the CVFF force field with the assistance of Insight
II software. The docking experiments, where the set of
Mulliken partial atomic charges calculated for the PM3
optimized structure gave results with less negative
binding energy in comparison with the docking experi-
ments where empirical partial charges, were employed.
Bin d in g Mod e An a lysis. The flexible docking tech-
nique was used for the binding mode analysis. As
flexible docking implemented in DOCK consumed con-
siderable computer resources, we have used the flexible
docking based on Monte Carlo simulated annealing as
implemented in AutoDock.18 The flexible docking ex-
periment with 12 active torsions (roscovitine) consumed
F igu r e 4. Interaction energies (kcal mol-1), as calculated for
the 5000 docked roscovitine-cdk2 complexes using DOCK, are
plotted against the rmsd (Å), from the roscovitine orientation
in the crystal complex. The docking experiments converge to
four binding modes (with the rmsd of 0.41, 3.5, 6.5, and 8
respectively), the best one being identical with the X-ray
structure.
for the docking of 2,6,9-purine derivatives into the cdk2
active site. The optimized parameters for the docking
are collected in Table 1.
2
about 25 h (SGI Indigo Extreme, R4000) of computer
time. Table 2 shows the parameters used for docking.
R ela t ion sh ip s b et w een t h e cd k 2 Act ivit y a n d
th e In ter a ction En er gy. Several compounds with
various 2,6,9-substituents were tested for their inhibi-
tion activity. Interaction energies for these compounds
were taken from the rigid docking, and the relationship
between the interaction energy (E) and activity (IC50)
Resu lts a n d Discu ssion
Recep tor Hu n tin g a n d P a r a m eter Op tim iza tion .
First of all, a search for the best receptor structure for
molecular docking was performed. The structures of
apo-cdk2, ATP-cdk2, the phosphorylated complex of
ATP-cdk2/cyclin A, and roscovitine-cdk2 complexes
were used as possible candidates. They were tested by
a rigid docking of the ligand into the receptor. The key
criterion describing the quality of the dock was the
convergence in graphs in which the interaction energy
is plotted against rmsd (root-mean-square deviation).
An example of such a graph is shown in Figure 4. The
best results for olomoucine and roscovitine were ob-
tained using the receptor from the complex I. The
results for 6-(3,3-dimethylallylamino)purine were also
acceptable. The parameters for docking, especially
dielectric properties and Lennard-J ones potential, were
optimized by running several docks on the best receptor
structure. The following parameters and their combina-
tions were varied: Lennard-J ones potential 10-12,
(Table 3) was analyzed. Since there is no direct signifi-
cant relationship between the two variables, we have
calculated Spearman’s rank order correlation coefficient
F using eq 1.
n
2
6
[r(x ) - r(y )]
i i
∑i
F ) 1 -
(1)
3
n - n
where n is the number of pairs (n ) 8 in our case), r(xi)
and r(yi) are the rank of the activity and the interaction
energy of the ith sample in the testing set.
The value obtained (F ) 0.643) is the same as the
1
9
critical value at the 0.05 level of significance. It implies
that the rigid docking experiments can be used for
qualitative activity prediction. However, a larger set of
samples will be necessary to prove this statement.
Rela tion sh ip s betw een th e cd k 1 Activity a n d
th e In ter a ction En er gy. The structure of cdk1 has not
6
-12; dielectric constant scale factor D (1, 2, 3, 4, 6);
and distance-dependent dielectric scale factor D (1, 2,
, 4, 6). It was found out that the 6-12 Lennard-J ones
3
potential and the distance-dependent dielectric constant
with the dielectric factor of 4 are the best parameters

Purine-like Inhibitors of cdk2
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13 2509
Ta ble 3. Comparison of IC50 Values for Various Purine Derivatives Tested on cdk2 and cdk1 with Interaction Energies from the
Docking Experiment
IC50, µM
compound
cdk1
cdk2
Ea
1
2
3
4
5
6
7
8
9
0
6-(benzylamino)-2-[(3-hydroxypropyl)amino]-9-isopropylpurine
6-(benzylamino)-2(R)-[(2-hydroxypropyl)amino]-9-isopropylpurine
6-(3-hydroxybenzylamino)-2-[[(1-hydroxymethyl)-2-(methyl)propyl]amino]-9-isopropylpurine
6-(benzylamino)-2-[(2-hydroxyethyl)amino]-9-isopropylpurine
6-(benzylamino)-2-[(2-hydroxyethyl)amino]-9-methylpurine
6-(cyclopentylamino)-2-[(3-hydroxypropyl)amino]-9-isopropylpurine
6-amino-2-[(3-hydroxypropyl)amino]-9-isopropylpurine
2.5
1.1
0.1
9.4
4.6
2.4
10
0.8
0.2
0.03
8.0
1.2
0.9
2.4
-39.9
-34.0
-38.5
-32.5
-30.8
-30.3
-27.7
-32.7
-28.1
-34.0
6-(adamantylamino)-2-[(3-hydroxypropyl)amino]-9-isopropylpurineb
6-(benzylamino)-9-isopropylpurine
>100
>100
3.5
0.01
4.5
-
b
1
6-(2-hydroxybenzylamino)-2-[[1-(hydroxymethyl)-2-(methyl)propyl]amino]-9-ispropylpurine
a
E denotes the interaction energy of compound’s complex with cdk2 (in kcal/mol). b Not used for cdk2 energy rank order correlation.
Ta ble 4. Influence of the C2 Substituent on the Interaction
Energy of 6-(Benzylamino)-9-isopropylpurine
C2 substituent
E
∆Ea
Cl
HS
H3CS
H2NNH
HOCH2(Ph)CHNH
HOCH2[S](OH)CHNH
HO(CH2)2NH
HO(CH2)2S
-28.8
-30.1
-30.6
-30.9
-31.4
-32.5
-32.5
-32.6
-32.7
-33.0
-33.1
-33.5
-34.0
-34.4
-35.3
-7.6
-6.3
-5.8
-5.5
-5.0
-3.9
-3.9
-3.8
-3.7
-3.4
-3.3
-2.9
-2.4
-2.0
-1.1
H2N(CH2)2NH
(CH3)2N
HO(CH2)3N(CH3)
H2N(CH2)3NH
[S]HO(CH3)CHCH2NH
PhCH2NH
HOCH2[R](OH)CHNH
a
E ) interaction energy of roscovitine (-36.4 kcal mol-1) -
∆
interaction energy of the derivative.
Ta ble 5. Influence of the C8 Substituent on the Interaction
-
1
Energy (E/kcal mol ) of 6-(Benzylamino)-2(R)-
[
(
F igu r e 5. A plot of the inhibitory activity IC50 for cdk1 (µM)
[1-(hydroxymethyl)propyl]amino]-9-isopropylpurine
roscovitine)
vs the interaction energy computed for the active site of cdk2
-
1
(
kcal mol ).
C8 substituent
E
∆Ea
yet been resolved. The high degree of primary sequence
homology between cdk2 and cdk1 (65.5%) suggests that
the active sites of these enzymes could be very similar.
Therefore, the relationship between the inhibitor/cdk2
interaction energy and the cdk1 assayed activity was
tested. Thirty-seven 2,6,9-derivatives of purine were
docked into the cdk2 active site, and the interaction
energies were computed and compared with activities
using Spearman rank correlation. The calculated value
F ) 0.685 again implies that the rank of the activities
is closely bound to the rank of interaction energies (the
critical value for the 0.01 level of probability is <0.4).
Figure 5 shows the plot of measured activities (IC50) vs
calculated interaction energies. However, we did not
find any other simple direct relationship between the
activity and the interaction energy.
NH2
OH
CN
-33.7
-34.2
-34.4
-2.7
-2.2
-2.0
a
∆E ) interaction energy of roscovitine (-36.4 kcal mol-1) -
interaction energy of the derivative.
Ta ble 6. Influence of the N9 Substituent on the Interaction
Energy of 6-(Benzylamino)-2(R)-
[[1-(hydroxymethyl)propyl]aminopurinea
∆Ea
N9 substituent
E
F3C
-30.1
-32.7
-33.9
-36.3
-39.3
-39.5
-40.7
-6.3
-3.7
-2.5
-0.1
2.9
H3C
H5C2
Cl3C
NH2)2CH
Cl2CH
OOC
(
3.1
4.3
-
a
E ) interaction energy of roscovitine (-36.4 kcal mol-1) -
interaction energy of the derivative.
∆
Sea r ch in g for New In h ibitor s. We have analyzed
the influence of purine ring substitution in positions 1,
2
8
, 3, 8, and 9 on the inhibitor activity; the results for 2,
, and 9 positions are collected in Tables 4-6. The
energies than the roscovitine (Table 5). This should
imply a lower activity, although the active site flexibility
could also influence the binding of compounds substi-
tuted by bulky groups.
The docking and experimental data show that the N9-
substitution is important for the positive binding, and
isopropyl appears to have an appropriate shape (cf. ref
20). The interaction energies of methyl, ethyl, etc.
docking experiments show that the C8 substituent must
be small, because a bulky substituent causes a consider-
able difference of the ligand orientation in the cavity.
For example, methyl and chloro derivatives were posi-
tioned quite differently from roscovitine. All the small
substituents tested exhibited more negative interaction

2
510 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13
Otyepka et al.
Ta ble 7. Comparison of Observed H-Bonds and Those
Predicted by Rigid Docking for Roscovitine in Å
6
NH(Leu83)‚‚‚N7
O(Leu83)‚‚‚HN
(roscovitine)
(
roscovitine)
observed
predicted
3.38
3.28
2.82
3.21
observation results from the comparison of ligand
orientations in the active site for known ligand-cdk2
complexes (Figure 3). The results of rigid docking show
two different binding modes; the first is similar to the
crystal orientation of roscovitine and the second to ATP
orientation. The mean interaction energy of crystal
-
1
roscovitine-like orientation is -77.8 kcal mol (1.2 Å
rmsd) and the interaction energy of ATP crystal-like
-
1
orientation is -55.4 kcal mol (2.8 Å rmsd).
These two binding modes were also found by flexible
docking. The first one was very similar to roscovitine
crystal orientation and its binding energy was equal to
-
1
-
74.6 kcal mol (2.1 Å rmsd). The second one cor-
F igu r e 6. A plot of van der Waals (squares) and electrostatic
triangles) contributions to the interaction energy E (all in kcal
responded to the ATP crystal orientation (with binding
energy -73.4 kcal mol and 2.6 Å rmsd). In the ATP-
like binding mode the roscovitine was positioned a little
away from the cavity because a large substituent (N -
benzyl) was present at the N -position of adenine. It was
(
-
1
-
1
mol ).
6
derivatives were more negative than the roscovitine
interaction energy. Nevertheless, we have found a few
substituents for which the interaction energies are even
higher than those calculated for the roscovitine (see
Table 6).
6
observed that the docked conformations did not agree
with the conformation of roscovitine in the crystal.
There were considerable differences in N9-isopropyl
orientation and in C2 side chain shape (Figure 7). The
results obtained after the C2-isopropyl fixation showed
also two ATP- and roscovitine-like (Figure 8) binding
modes. The roscovitine-like interaction energy was equal
The cdk1 and cdk2 assays showed that C2-substitu-
tion may increase the inhibitory activity. The docking
experiments confirm this observation (Table 4). There-
fore, finding an appropriate C2 substituent seems to be
a suitable approach to designing new potent and selec-
tive inhibitors. The first step in any such a design is to
make sure that the space occupied by the roscovitine
C2 chain is the same as the space occupied by the ATP
N9-substituent and by the IPA 6-substituent (Figure 3).
It is known that addition of a methyl substituent at
positions 1, 3, and 7 drastically reduces the cdk1
activity. Our docking experiments are not able to
explain this experimental observation, because the
interaction energies calculated for the 1-methyl and
-1
to -77.2 kcal mol (1.5 Å rmsd), but the structure with
-
1
the lowest interaction energy (-78.2 kcal mol ) was
positioned differently from the roscovitine binding mode.
Con clu sion s
We have used rigid and flexible docking techniques
for prediction of cdk2 and cdk1 inhibitory activities. The
parameters for the rigid docking were optimized to allow
us routine work. After that, 47 docking experiments
were performed with the cdk2 active site. We have
calculated the rank correlation between the interaction
energy and the respective cdk2/cdk1 assayed activity.
The calculated correlation coefficients imply that the
rigid docking is a suitable technique for making qualita-
tive predictions about activity. At the same time, the
results confirm the assumption that there is a consider-
able relationship between the cdk1 activity and the cdk2
interaction energy.
We have studied also the influence of the substitution
of the purine ring on ligand interaction energy and
attempted to predict new possibly active N9-derivatives.
The docking experiments did not explain the experi-
mentally observed inactivity of N1 and N3 purine
derivatives, because the interaction energies calculated
for these compounds are inadequately high.
3
-methyl roscovitine derivatives are -39.0 and -32.6
-
1
kcal mol , respectively, which is too low (even more
negative than for the roscovitine complex). These de-
rivatives keep the same binding mode as roscovitine.
Bin d in g of th e Liga n d to th e Active Site. The
interaction energy of 2,6,9-purine derivatives has two
considerable contributions, electrostatic and van der
Waals, where the electrostatic contribution includes the
H-bonds. The rigid docking experiments show that the
electrostatic contribution is small and almost substitu-
ent-independent (being in the range from -3.1 to -2.0
-
1
kcal mol ). The increase or decrease in interaction
energy is caused by the van der Waals contribution,
which is about 10-fold higher than the electrostatic
-
1
contribution (from -38.6 to -19.7 kcal mol ). Figure
shows these two contributions to the interaction
Information obtained in this study will be used for
designing new active anticancer compounds and for
additional work in the area of docking experiments.
Even now, the combination of medicinal chemistry,
X-ray data, and molecular modeling allows us to propose
new higly active purine derivatives (e.g. compound 10)
and to prove the usefulness of this approach by synthe-
sizing just a small number of designed compounds.
6
energy. The H-bonds predicted by rigid docking are in
good agreement with the experimentally observed ones
(Table 7).
Bin d in g Mod es. The active site of cdk2 offers many
different binding modes for purine derivatives that are
strongly dependent on the purine substituent. This

Purine-like Inhibitors of cdk2
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13 2511
F igu r e 7. Comparison of the conformation of the lowest energy flexible dock (thick lines) with the crystal conformation (thin
lines).
F igu r e 8. Comparison of the conformation of the second lowest energy dock after N9-isopropyl fixation (thick lines) with the
crystal conformation (thin lines).
Exp er im en ta l Section
tested compound in a final volume of 20 µL, all in reaction
buffer [50 mM Hepes (pH 7.4), 10 mM MgCl , 5 mM EGTA,
0 mM 2-glycerolphosphate, 1 mM NaF, 1 mM DTT, and
protease inhibitors]. After 10 min, the incubations were
stopped by adding SDS sample buffer, and the proteins were
separated using 12.5% SDS-PAGE. The measurement of
kinase inhibition employed the digital imaging analyzer BAS
2
cd k 2/Cyclin E Kin a se In h ibition Assa y. Nine compounds
were tested for cdk2/cyclin E inhibitory activity to determine
the basic relationships between their interaction energies in
the docked complex and the inhibitory activity. The cdk2/cyclin
E complex was produced in Sf9 insect cells coinfected with an
appropriate baculoviral construct. The cells were harvested
1
1
800. The kinase activity was expressed as a percentage of
7
0 h postinfection in lysis buffer [50 mM Tris (pH 7.4), 150
maximum activity and the IC50 value was determined by
graphic analysis. The cdk1 kinase inhibition tests were done
in a similar way.20
mM NaCl, 5 mM EDTA, 20 mM NaF, 1% Tween 20, protease
inhibitors] for 30 min on ice, and the soluble fraction was
recovered by centrifugation at 14000g for 10 min. The protein
extract was stored at -80 °C until use.
The final point test system for kinase activity measurement
was used to carry out experiments on the kinetics under linear
conditions. The assay mixture contained 1 mg/mL histone
Syn th esis. Melting points were determined on a Kofler
1
block and are uncorrected. The H NMR spectra (δ, ppm; J ,
Hz) were measured on Varian VXR-400 (400 MHz) or on
Varian Unity 200 (200 MHz) instruments. All spectra were
obtained at 25 °C using tetramethylsilane as an internal
3
2
(Sigma Type III-S), 15 µM ATP, 0.2 µCi [γ- P]ATP, and the

2
512 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13
Otyepka et al.
standard. Electron impact mass spectra m/z (rel %) were
measured on a VG 7070E spectrometer (70 eV, 200 °C, direct
inlet). Merck silica gel Kieselgel 60 (230-400 mesh) was used
for column chromatography. All compounds gave satisfactory
elemental analyses (0.4%).
(6H, m, adamantyl), 1.909 (2H, tt, J ) 7.1, J ) 6.6, CH
CH ), 2.10 (3H, bs, adamantyl), 2.19 (6H, bs, adamantyl), 3.582
(2H, t, J ) 7.1, CH CH CH ), 3.705 (2H, t, J ) 6.5, CH
), 4.608 (1H, sept, J ) 6.8, CH(CH ), 8.083 (1H, s,
2 2
CH -
2
2
2
2
2
-
CH
2
8
CH
2
3 2
)
HC ). The proton gCOSY experiment was used for the assign-
ment of signals. MS (Waters/Micromass, ZMD-detector, direct
The following compounds were prepared as described in the
literature: 6-(benzylamino)-2-[(3-hydroxypropyl)amino]-9-iso-
+
inlet, ESI, 20 eV, + ions): [M + H] ) 385.5 (100), 386.5 (22).
2
1
Anal. (C H N O‚H O) C, H, N.
propylpurine (1), 6-(benzylamino)-2(R)-[(2-hydroxypropyl)-
amino]-9-isopropylpurine (2) [mp 142-143 °C, [R] ) -8,0 (c
21 32
6
2
6-Ben zyla m in o-9-isop r op ylp u r in e (9). 6-Benzylamino-
purine (Fluka) was alkylated by isopropyl bromide analogously
to 2-chloropurines mentioned above (in DMSO): Column
)
3
0.11, CHCl ); prepared in the same way as its racemate
synthesized previously by using (R)-(-)-1-amino-2-propanol
2
0
instead of racemate], 6-(benzylamino)-2-[(2-hydroxyethyl)-
3 3
chromatography 2% MeOH in CHCl ; crystallization CHCl -
2
0
+•
amino]-9-isopropylpurine (4), and 6-benzylamino-2-[(2-hy-
Et O; yield 80%; mp 117-118 °C. MS: 267 (100, M ), 266 (14),
2
droxyethyl)amino]-9-methylpurine (5, olomoucine).2
2,23
The
225 (20), 224 (61), 162 (12), 120 (17), 119 (9), 106 (59), 93 (10),
1
straightforward three- or two-step synthesis of other C2, C6,
and N9-trisubstituted purines was started from commercially
available 2,6-dichloropurine or from 2,6-dichloro-9-isopropyl-
91 (33), 65 (10). H NMR (200 MHz, DMSO): 1.45 (6H, d, J )
6.8, (CH ) CH), 4.53 (1H, sept, J ) 6.8, CH(CH ) ), 4.63 (2H,
3
2
3 2
bs, CH NH), 8.12 and 8.18 (each s 1H, H-8 and H-2). Anal.
2
2
1
purine. The latter was used as a parent compound to avoid
alkylation of phenolic OH or C6-NH in the synthesis of
(C H N ) C, H, N.
1
5
17
5
2
6
-(2-H yd r oxyb en zyla m in o)-2-[[1-(h yd r oxym et h yl)-2-
m eth yl)pr opyl]am in o]-9-ispr opylpu r in e (10): Column chro-
matography stepwise 0, 0.5, 1.0, 2.0% MeOH in CHCl ; the
compounds 3 or 7 and 10. The starting compounds were
(
reacted with appropriate alkylamine in 1-butanol in the
3
2
0,22
presence of triethylamine (115-120 °C, 3 h)
4
or NH OH (85
syrup-like product crystallized after several days, yield 35%;
°
C, 9 h, preparation comp.7). The N9-H derivatives were then
+•
mp 126-138 °C. MS: 384 (19, M ), 366 (7), 353 (25), 341 (10),
2 3
alkylated with isopropylbromide (K CO
, DMF, or DMAA).2
0,24
2
98 (13), 274 (38), 260 (25), 217 (20), 192 (15), 175 (32), 134
The resulting 9-isopropyl-6-alkylamino-2-chloropurine was
1
(
20), 107 (36), 78 (35), 69 (35), 55 (44), 43 (98), 41 (100). H
NMR (400 MHz, CDCl ): 1.060 (3H, d, J ) 6.8, (CH CHC*),
.065 (3H, d, J ) 6.8, (CH CHC*), 1.515 (3H, d, J ) 6.8,
CH CHN), 2.040 (1H,
further reacted with 3-amino-1-propanol (160 °C, 3 h) or (R/
3
3 2
)
2
0,22-25
S)-2-amino-3-methyl-1-butanol (160 °C, 12 h).
ucts were purified by column chromatography.
-(3-H yd r oxyb en zyla m in o)-2-[[1-(h yd r oxym et h yl)-2-
m eth yl)pr opyl]am in o]-9-isopr opylpu r in e (3): Column chro-
matography (second principal spot) stepwise 0.5, 1, 1.5%
MeOH in CHCl with a trace of concentrated NH OH; crystal-
lization CHCl -Et O; yield 30% (based on 2,6-dichloro-9-
isopropylpurine); mp 180-181 °C MS: 384 (11, M ), 366 (35),
54 (22), 353 (72), 341 (23), 323 (100), 298 (12), 175 (11), 134
The prod-
1
(
3 2
)
3
)
2
3 2
CHN), 1.523 (3H, d, J ) 6.8, (CH )
6
sept, J ) 6.8, (CH )CHC*), 3.743 (1H, dd, J ) 7.4, 10.6,
3
(
CHHOH), 3.824 (1H, m, CH C*H), 3.890 (1H, dd, J ) 2.5, 10.6,
2
CHHOH), 4.505 (1H, dd, J ) 6.6, 15.0, CHHNH), 4.570 (1H,
sept, J ) 6.8, NCH(CH ) ), 4.653 (1H, dd, J ) 7.0, 15.0,
3
4
3
2
3
2
CHHNH), 5.028 (1H, d, J ) 6.9, NHCH), 6.474 (1H, bs,
+•
NHCH ), 6.845 (1H, dt, J ) 1.2, 7.4, ArH), 6.918 (1H, dd, J )
2
3
1.2, 8.3, ArH), 7.174-7.222 (2H, m, ArH), 7.498 (1H, s, HC ).
8
1
(14), 122 (12), 107 (67). H NMR (399.90 MHz, CDCl
3
30 °C):
CHC*), 1.025 (6H, d, J ) 6.9,
CHC*), 1.549 (6H, d, J ) 6.8, (CH CHN), 1.557 (6H,
CHN), 1.969 (1H, sept, J ) 6.8, (CH )CHC*),
.983 (1H, sept, J ) 6.8, (CH CHC*), 3.692 (2H, dd, J ) 8.0,
0.7, CHHOH), 3.869 (2H, dd, J ) 2.8, 10.7, CHHOH), 3.917
OH), 4.633 (2H, sept, J ) 6.8, NCH(CH ),
NH), 4.933 (2H, d, J ) 7.2, NHC ), 6.234
The proton 2D-COSY, TOCSY, and HMQC experiments were
used for the assignment of signals. Anal. (C H N O ) C, H,
1
3 2
.016 (6H, d, J ) 6.9, (CH )
2
0
28
6
2
(CH
3
)
2
3
)
2
N.
d, J ) 6.8, (CH
1
1
3
)
2
3
Ha r d w a r e a n d Softw a r e. The docking experiments as
3 2
)
well as receptor and ligand preparations were performed on
2
SGI Indigo Extreme (R4400), SGI Onyx 2 (R4000), SGI Power
(2H, m, C*HCH
2
3 2
)
Challenge (12 × R10000), DEC Alpha (2 × 250 MHz), and PC
2
4
.56-4.70 (4H, m, CH
2
(
CPU Intel PIII 450) machines. The Insight II (MSI Biosym)
(2H, br s, NHCH
2
), 6.746 (2H, ddd, J ) 1.1, 2.5, 8.1, ArH),
15
software was used for structure manipulation and visualiza-
tion of results. The HyperChem 5.01 (HyperCube) software
6
7
2
.848 (2H, ddd, J ) 1.0, 1.7, 7.5, ArH), 6.871 (2H, m, ArH),
17
8
.131 (2H, dd, J ) 7.5, 8.1, ArH), 7.560 (2H, s, HC ). The proton
was used for ligand construction and PM3 semiempirical
calculations. The molecular docking was performed using
D-COSY spectrum was used for the assignment of signals.
Anal. (C20
-(Cyclop en tyla m in o)-2-[(3-h yd r oxyp r op yla m in o)a m -
in o]-9-isop r op ylp u r in e (6): Column chromatography (step-
wise 0, 1, 2% MeOH in CHCl ); crystallization from ethyl
H
28
N
6
O
2
) C, H, N.
16
18
DOCK 4.0.1 and AutoDock 2.4 software packages.
6
Ack n ow led gm en t. One of the authors (M.O.) would
like to thank Assoc. Prof. J an Lasovsk y´ (Olomouc) for
creating excellent working conditions for his Ph.D. study
and also Dr. J . Damborsk y´ , J . P rˇ idal, R. Sˇ tefl, and other
members of the Laboratory of Biomolecular Structure
and Dynamics for their help. We thank the supercom-
puting centers in Brno and Olomouc for the computer
time. This work was supported by Grants 31703013
(Palacky University), VS96095 and VS96154 (Ministry
of Education), and 201/98/K041, 303/99/1541, and 204/
3
+
•
acetate yield 64%; mp 137-139 °C. MS: 318 (100, M ), 317
9), 288 (10), 287 (10), 275 (12), 274 (18), 273 (33), 250 (7), 219
17), 206 (33), 205 (21), 192 (13), 163 (14), 134 (15), 43 (29), 41
(
(
(
1
16). H NMR (200 MHz, CDCl
3
): 1.53 (6H, d, J ) 6.8, (CH
3
)
2
-
-
CH), 1.55-1.80 (8H, m, cyclopentyl), 2.09 (2H, m, CH
CH ), 3.65 (4H, m, CH N + CH
cyclopentyl), 4.59 (1H, sept, J ) 6.8, CH(CH
exch H, OH or NH), 5.21 and 5.60 (each bs 1H, NH or OH),
2
CH
2
2
2
2
O), 4.41 (1H, bm, CH in
), 4.95 (1H, bt,
3 2
)
8
7
.51 s (1H, HC ). Anal. (C16
-Am in o-2-(3-h yd r oxyp r op yla m in o)-9-isop r op ylp u -
r in e (7): Crystallization from water; recrystallization from
26 6
H N O) C, H, N.
6
9
6/K235 (Grant Agency of the Czech Republic).
+
•
MeOH-Et
2
(
(
2
O; yield 70%; mp 143-144 °C. MS: 250 (83, M ),
19 (60), 206 (73), 205 (98), 192 (25), 177 (49), 163 (100), 150
Su p p or tin g In for m a tion Ava ila ble: Contact information
1
50), 134 (65), 108 (37). NMR (400 MHz, CDCl
6H, d, J ) 6.8, (CH CH, 4.63 (1H, sept, J ) 6.8, (CH
CH CH ), 3.59-3.67 (4H, m, CH
)], COSY[7.97 (1H, bt, J ) 6.7, NH), 3.59-3.67m], 5.4
3
): COSY[1.546
regarding DOCK, Auto Dock, Insight II, and Hyperchem. This
material is available free of charge via the Internet at http://
pubs.acs.org.
3
)
2
3
)CH)],
COSY[1.73 (2H, m, CH
CH CH
bs, 2H, NH
H, N.
-(Ad a m a n t yla m in o)-2-[(3-h yd r oxyp r op yl)a m in o]-9-
isop r op ylp u r in e (8): Column chromatography stepwise 0, 1,
2
2
2
2
-
2
2
8
(
2
), 7.575 (1H, HC ). Anal. (C11
H
18
N
6
2
O‚1.5H O) C,
Refer en ces
(
1) De Azevedo, W. F.; Leclerc, S.; Meijer, L.; Havl ´ı cˇ ek, L.; Strnad,
M.; Kim, S.-H. Inhibition of Cyclin-dependent Kinases by Purine
Analogues. Eur. J . Biochem. 1997, 243, 518-526 (ref 12, Cordon-
Cardo, C. Am. J . Pathol. 1995 147, 545-560, and ref 13, Karp,
J . E.; Broder, S. Nat. Med. 1995, 1, 309-320).
6
1
2
(
% MeOH in CHCl
3
; syruplike product; yield 31% H NMR
CH), 1.67-1.76
400 MHz, D O): 1.549 (6H, d, J ) 6.8, (CH )
2
3 2

Purine-like Inhibitors of cdk2
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 13 2513
(
(
(
2) Pindur, U.; Kim, Y.-S.; Mehrabani, F. Advances in Indolo[2,3-
a]carbazole Chemistry: Design and Synthesis of Protein Kinase
C and Topoisomerase I Inhibitors. Curr. Med. Chem. 1999, 6,
(12) De Azevedo, W. F.; Leclerc, S.; Maijer L.; Havl ´ı cˇ ek, L.; Strnad,
M.; Kim, S.-H. Inhibition of Cyclin-Dependent Kinases by Purine
Analogues: Crystal Structure of Human Cdk2 Complexed With
Roscovitine. Eur. J . Biochem. 1997, 243, 518-526.
2
9-69.
3) Losiewicz, M. D.; Carlson, B. A.; Kaur, G.; Sausville, E. A.;
Worland, P. Potent Inhibition of Cdc2 Kinase Activity by the
Flavonoid L86-8275. Biochim. Biophys. Res. Commun. 1994,
(13) Kim, S.-H. Structure-Based Inhibitor Design for Cdk2, a Cell
Cycle Controlling Protein Kinase. Pure Appl. Chem. 1998, 70,
5
55-565.
2
01, 589-595.
(
14) Noble, M. E. M.; Endicott, J . A. Chemical Inhibitors of Cyclin-
Dependent Kinases: Insights into Design from X-ray Crystal-
lographic Studies. Pharmacol. Theor. 1999, 82, 269-278.
15) Insight II program 1995 Biosym/MSI, San Diego, USA.
16) Ewing, T. J . A.; Kuntz, I. D. Critical Evaluation of Search
Algorithm for Automated Molecular Docking and Database
Screening J . Comput. Chem. 1997, 18, 1175-1189.
4) Kitagawa, M.; Okabe, T.; Ogino, H.; Matsumoto, H.; Suzuki-
Takahashi, I.; Kokubo, T.; Higashi, H.; Saitoh, S.; Taya, Y.;
Yasuda, H.; Ohba, Y.; Nishimura, S.; Tanaka, N.; Okuyama, A.
Butyrolactone I. A Selective Inhibitor of Cdk2 and Cdc2 kinase.
Oncogene 1993, 8, 2433-2441.
(
(
(
5) Vesel y´ , J .; Havl ´ı cˇ ek, L.; Strnad, M.; Blow, J . J .; Donella-Deana,
A.; Pinna, L.; Letham, D. S.; Kato, J .-Y.; Detivaud, L.; Leclerc,
S.; Meijer, L. Inhibition of Cyclin-Dependent Kinases by Purine
Analogues. Eur. J . Biochem. 1994, 224, 771-786.
6) Hoessel, R.; Leclerc, S.; Endicott, J . A.; Noble, M.; Lawrie, A.;
Tunnah, P.; Leost, M.; Damiens, E.; Dominique, M.; 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,
(17) HyperChem 5.01 program 1997 Hypercube Inc., Ontario, Canada.
(18) Goodsell, D. S.; Olson, A. J . Automated Docking of Substrates
to Proteins by Simulated Annealing Proteins. Struct. Func.
Genet. 1990, 8, 195-202.
(19) Siegel, S. Nonparametric Statistics; McGraw-Hill: New York,
1956.
(20) Havl ´ı cˇ ek, L.; Hanu sˇ , J .; Vesel y´ , J .; Leclerc, S.; Meijer, L.; Shaw,
G.; Strnad, M. Cytokinin-Derived Cyclin-Dependent Kinase
Inhibitors: Synthesis and cdc2 Inhibitory Activity of Olomoucine
and Related Compounds. J . Med. Chem. 1997, 40, 408-412.
(
6
0-67.
(
7) Zaharevitz, D. W.; Gussio, R.; Leost, M.; Sanderowitz, 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.
(
(
(
21) Havl ´ı cˇ ek, L.; Hajd u´ ch, M.; Strnad, M. Cyclin-dependent kinase
inhibitor. Patent Applied in Australia, Canada, J apan, Korea
and USA, 1998, AU5646598.
22) Hocart, C. H.; Letham, D. S.; Parker, C. W. Substituted
Xanthines and Cytokinin Analogues as Inhibitors of Cytokinin
N-Glucosylation. Phytochemistry 1991, 30, 2477-2486.
23) Parker, C. W.; Entsch, B.; Letham, D. S. Inhibitors of Two
Enzymes which Metabolize Cytokinins. Phytochemistry 1986, 25,
(
8) Kent, L. L.; Hull-Campbell N. E.; Lau T.; Wu, J . C.; Thompson,
S. A.; Nori, M. Characterization of Novel Inhibitors of Cyclin-
Dependent Kinases. Biochem. Biophys. Res. Commun. 1999, 260,
7
68-774.
(9) Gray, N.; D e´ tivaud, L.; Doerig, Ch.; Meijer, L. ATP-site Directed
Inhibitors of Cyclin-dependent Kinases. Curr. Med. Chem. 1999,
3
03-310.
6
, 859-875.
(
10) Lybrand, T. P. Ligand-Protein Docking and Rational Drug
Design. Curr. Opin. Struct. Biol. 1995, 5, 224-228.
11) Schulze-Gahmen, U.; Brandsen, J .; J ones, H. D.; Morgan, D. O.;
Meijer, L.; Vesel y´ , J .; Kim, S.-H. Multiple Modes of Ligand
Recognition: Crystal Structure of Cyclin-Dependent Protein
Kinase 2 in Complex With ATP and Two Inhibitors, Olomoucine
and Isopenthenyladenine. Proteins: Struct. Func. Genet. 1995,
(24) Imbach, P.; Capraro, H. G.; Furet, P.; Mett, H.; Meyer, T.;
Zimmermann, J . 2,6,9-Trisubstituted Purines: Optimization
Towards Highly Potent and Selective CDK1 Inhibitors. Bioorg.
Med. Chem. Lett. 1999, 9, 91-96.
(
(25) Raiford, L. Ch.; Clark, E. P. Diacyl Derivatives of ortho-
hydroxybenzylamine J . Am. Chem. Soc. 1923, 45, 1738-1745.
2
2, 378-391.
J M990506W