Structure-based design of novel Chk1 inhibitors: Insights into hydrogen bonding and protein-ligand affinity

Article abstract of DOI:10.1021/jm049022c

We report the discovery, synthesis, and crystallographic binding mode of novel furanopyrimidine and pyrrolopyrimidine inhibitors of the Chk1 kinase, an oncology target. These inhibitors are synthetically tractable and inhibit Chk1 by competing for its ATP

Full text of DOI:10.1021/jm049022c

4332
J. Med. Chem. 2005, 48, 4332-4345
Structure-Based Design of Novel Chk1 Inhibitors: Insights into Hydrogen
Bonding and Protein-Ligand Affinity
Nicolas Foloppe,* Lisa M. Fisher, Rob Howes, Peter Kierstan, Andrew Potter, Alan G. S. Robertson, and
Allan E. Surgenor
Vernalis (R&D) Limited, Granta Park, Abington, Cambridge CB1 6GB, U.K.
Received December 1, 2004
We report the discovery, synthesis, and crystallographic binding mode of novel furanopyrimidine
and pyrrolopyrimidine inhibitors of the Chk1 kinase, an oncology target. These inhibitors are
synthetically tractable and inhibit Chk1 by competing for its ATP site. A chronological account
allows an objective comparison of modeled compound docking modes to the subsequently
obtained crystal structures. The comparison provides insights regarding the interpretation of
modeling results, in relationship to the multiple reasonable docking modes which may be
obtained in a kinase-ATP site. The crystal structures were used to guide medicinal chemistry
efforts. This led to a thorough characterization of a pair of ligand-protein complexes which
differ by a single hydrogen bond. An analysis indicates that this hydrogen bond is expected to
contribute a fraction of the 10-fold change in binding affinity, adding a valuable observation to
the debate about the energetic role of hydrogen bonding in molecular recognition.
Introduction
by the human nuclear serine/threonine kinase Chk1.17
This phosphorylation is necessary for arrest in G2 in
response to DNA damage in human cells18 and promotes
binding of the 14-3-3 protein to Cdc25C, which in turn
prevents Cdc25C from interacting functionally with
CDK1.18 These biochemical results are consistent with
genetic studies.19-22 Chk1 is activated in response to
DNA damage via the ATM/ATR pathway23-26 and is
required for checkpoint arrest.27 Inhibition of Chk1 may
also induce apoptosis.20 In sum, there is ample biological
evidence that Chk1 plays an essential role in the
regulation of the G2/M checkpoint,4,23,28-30 suggesting
that the inhibition of Chk1 should abrogate this check-
point in a manner relevant to the treatment of cancer.6
Selective Chk1 inhibitors should also prove useful to
study the regulation of the G2/M checkpoint.
Several Chk1 inhibitors are already known, including
natural products such as staurosporine (and its deriva-
tives: 7-hydroxystaurosprine (UCN-01),31,32 indolocar-
bazole analogues,33,34 isogranulatimide,35 and Go¨697636)
and debromohymenialdisine.37 This is consistent with
some of these compounds being potent abrogators of the
G2/M checkpoint in cells,8,33,37 adding to the relevance
of Chk1 as an oncology target. UCN-01 is indeed under-
going clinical trials as an anticancer agent.38,39 A limited
number of additional Chk1 inhibitors have been re-
ported;6,40,41 however, there is clearly a need to identify
more druglike Chk1 inhibitors, given the typical attri-
tion rate in drug discovery. This should be facilitated
by the structural information available for the Chk1
kinase domain.
Control of the cell cycle provides potential targets for
the development of therapeutics against cancer.1 A key
point in the cell cycle is the checkpoint between the G2
and the M phases (G2/M).2 This checkpoint is a pathway
controlling the ability of eukaryotic cells to arrest the
cell cycle in G2 in response to DNA damage, to allow
time for DNA repair.3,4 DNA damage may be due to
radiation or chemical agents, which are still a major
component of cancer therapy. Key to the relevance of
the G2/M checkpoint for cancer treatment is that, in
normal cells, DNA damage also triggers cell cycle arrest
in the G1 phase, via the tumor suppressor protein p53.5,6
This G1/S checkpoint is impaired in many tumor cells,
because of mutations in p53,5,7 and these tumor cells
can only rely on the G2/M checkpoint to repair their
DNA in response to DNA damaging agents. Therefore,
abrogation of the G2/M checkpoint in these tumor cells
would allow premature mitotic entry in the presence of
DNA damage, leading ultimately to cell death. In
contrast, normal cells would arrest in G1 and would be
minimally affected by abrogation of the G2/M check-
point. Therefore, inhibition of the G2/M checkpoint
should lead to an increased and selective sensitivity of
cancer cells to DNA damaging agents,8,9 suggesting that
small molecule inhibitors of this checkpoint may prove
useful in cancer therapy as sensitizing agents.6,10,11
Arrest in G2 is mainly regulated by the maintenance
of inhibitory phosphorylation on cyclin-dependent ki-
nase 1 (CDK1) (also called Cdc2), which prevents
activation of this mitosis promoting CDK.12-14 Dephos-
phorylation, and thereby activation, of CDK1 is carried
out in human cells by the dual specificity phosphatase
Cdc25C.15,16 Cdc25C can be phosphorylated on Ser216
Chk1 is a 54 kDa protein of 476 amino acids17
comprised of a highly conserved N-terminal kinase
domain (residues 1-265) followed by a linker and a less
conserved C-terminal region that decreases Chk1 kinase
activity.42 The crystal structure of the kinase domain
of human Chk1 and its complex with the ATP analogue
AMP-PNP have been solved at 1.7 Å resolution.42
* To whom correspondence should be addressed: N. Foloppe,
Vernalis (R&D) Ltd., Granta Park, Abington, Cambridgeshire CB1
6GB, U.K.; phone (direct): + (44) (0) 1223 895 338; fax: + (44) (0)
1223 895 556; email: n.foloppe@vernalis.com.
10.1021/jm049022c CCC: $30.25 © 2005 American Chemical Society
Published on Web 05/25/2005

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4333
Table 1. Structure and Binding Affinities of Chk1 Inhibitors
Crystal structures of Chk1 complexed with ATP-
competitive inhibitors staurosporine, 7-hydroxystauro-
sprine (UCN-01), and SB-218078 have also been pre-
sented.43 This puts structure-based inhibitor design
strategies against human Chk1 kinase on firm footing.
This work presents the discovery and elaboration of
a new type of druglike ATP-competitive Chk1 inhibitor,
in the context of the crystal structures of these mol-
ecules bound to the ATP site. These structures allowed
to rationalize the structure-activity relationship (SAR)
around the initial chemotype and to improve its affinity
by modification of their scaffold. This work also drew
on the iterative nature of molecular modeling as part
of structure-based drug design against a kinase, which
is frequently referred to but rarely illustrated. Finally,
these results enabled us to specifically probe the strength
of a hydrogen bond between compounds and the highly
conserved kinase motif with which most ATP-competi-
tive ligands interact. This interaction is found to be
moderately favorable, and this should contribute to a
better characterization of the much debated role of the
energetics of hydrogen bonding for the binding affinity
of ligands to proteins.
Results and Discussion
This report lends itself to a chronological presenta-
tion, which puts into perspective the influence of the
structural information on the medicinal chemistry
strategy.
Identification of a Novel Chk1 Inhibitor. Com-
pound 1 (Table 1) was identified as an inhibitor of
human Chk1 during a medium throughput screen of a
commercial library of compounds provided by Chemical
Diversity (ChemDiv).44 In that screen, compound 1
inhibited Chk1 with an IC50 of 15.4 µM. It was further
shown that compound 1 inhibits Chk1 in an ATP-
competitive manner by standard kinetic methods, with
a Ki of 7.2 µM.
a Measured for human PKA and CDK1, in addition to Chk1
kinase, to explore binding selectivity. b Ar is p-OMe phenyl. c Each
of these Ki was derived from three independent sets of measure-
ments; the associated standard deviations were 4.5 and 1.3 µM
for compounds 2 and 5, respectively.
Table 2. Inactivea Analogues of Compound 1
A search of the Vernalis electronic catalog of com-
45
mercially available compounds
identified four ana-
logues (8-11, Table 2) of compound 1 which were
obtained and assayed. This search aimed to identify
compounds containing the same chemical core (5,6-
diphenyl-furo[2,3]pyrimidin-4-ylamine, represented by
compound 8) as compound 1, using the substructure
search algorithm implemented in the MDL ISIS/Base
software.46 The other analogues (12-15) of compound
1 shown in Table 2 were synthesized in-house following
published procedures 47 (Scheme 1). Thus, condensation
of the commercially available aminofuran (16), with in
situ generated acetic formic anhydride gave the N-
formyl derivative (17). This was cyclized under thermal
conditions giving the oxo-compound (18) in good yield.
Chlorination was easily effected using phosphorus oxy-
chloride producing derivative 19 in 75% yield. Displace-
ment of the chlorine was achieved in refluxing ethanol
with 2 equiv of the corresponding amine, leading to the
furanopyrimidine product (20) in 70-80% yield.
a Defined as having an IC50 > 100 µM for Chk1. These analogues
differed by the nature of the listed R groups, as part of the chemical
structure shown.
The analogues in Table 2 differ from compound 1 by
the lack of methoxy substituents and the alteration of
the hydroxy-ethyl side chain, which led to a systematic
loss of affinity. A structure-based strategy was then
pursued to explain this initial SAR and suggest more
successful ways to elaborate compound 1.
Initial Structural Model. Given that in-house Chk1
X-ray crystallography was not yet available at the outset
of the project, compound 1 was docked to a publicly

4334 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Scheme 1. Synthesis of Furanopyrimidine Derivatives
Foloppe et al.
42
available X-ray structure using molecular modeling
with the highly conserved backbone kinase motif at the
hinge between the two lobes of the kinase domain. These
hydrogen bonds are observed with almost all other
kinase inhibitors.48-50 The methoxy groups pointed
away from the active site and toward the solvent,
suggesting that it was not the lack of these groups which
imparted a loss of activity in the compounds shown in
Table 2. On the other hand, in both docking modes the
hydroxy-ethyl side chain had the correct length to
hydrogen bond to the hydroxyl oxygen of Ser147. This
was encouraging because it could possibly explain why
extending this side chain by one methylene group
abolished binding to Chk1. In particular, systematic
attempts to accommodate a hydroxy-propyl side chain
instead of its hydroxy-ethyl analogue, while keeping
docking mode A, resulted in severe steric clashes
between the side chain and the protein. Also, three
hydrogen bonds with the kinase hinge backbone were
formed in docking mode A, although only two were
formed in mode B. Some of these hydrogen bonds were
C-H‚‚‚O hydrogen bonds involving a polarized C-H
bond of the pyrimidine moiety (Figure 1), the existence
of which is well documented.51-53 Another feature in
favor of docking mode A was the presumably unfavor-
able interaction between the furan oxygen and the
carbonyl oxygen of Cys87 in docking mode B. Also, there
was skepticism about the orientation of the amino N-H
bond in mode B, where it points toward a phenyl ring
of the same compound. Although such favorable polar
interactions involving the π electrons of aromatic moi-
eties are known to exist,54-56 it was initially deemed
unlikely in the present context, assuming that a strong-
er hydrogen bond could be formed with a carbonyl
backbone in mode A. In addition, the methoxy group
closest to Ser147 accepted a hydrogen bond from Lys38
in docking mode A (not shown), with no counterpart for
this interaction in mode B. Given that the two docking
modes buried approximately the same amount of apolar
surface area, mode A scored more favorably than mode
B, because of its apparently more favorable polar
interactions. Finally, there was no room for substituents
around the hydroxy-ethyl side chain in mode A, which
was consistent with the available SAR (Table 2); that
is, all of the substituted analogues of this side chain had
lost binding to Chk1. Docking mode A was therefore
adopted as a convincing initial structural working
hypothesis for medicinal chemistry efforts.
techniques. This yielded two interesting putative bind-
ing modes (Figure 1) where compound 1 was well buried
in the ATP-binding site, while forming hydrogen bonds
Figure 1. The two initial docking modes selected for com-
pound 1. The best scoring docking mode is shown in panel A.
For both docking modes, the kinase hinge backbone (Glu85
and Cys87) is shown. Ser147 is also shown because it is
hydrogen bonded by the hydroxy-ethyl side chain. Other
residues and most hydrogens were omitted for clarity. Classical
hydrogen bonds are depicted with green dotted lines, while
CsH‚‚‚O hydrogen bonds are traced in magenta. In docking
mode B, the contact between the furan oxygen and the
carbonyl oxygen of Cys87 is shown with a blue dotted line.
Test and Revision of the Initial Structural Model.
In both docking modes A and B, the methoxy groups
are at the periphery of the binding site, pointing toward

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4335
Table 3. Crystallographic Data Collection and Refinement Statistics for Chk1 in Complex with Compounds 1-7
Chk1 + compound
structure
1
2
3
4
5
6
7
Data Collection Statistics
resolution (Å)
2.00
54186
21797
95.1/95.1
2/1
14.5/3.6
4.4/22.9
2.10
2.10
2.10
2.12
2.80
196243
7714
97.9/98.9
3.2/3.2
9.1/1.0
12.9/50.9
2.20
measured reflections
144361
20050
96.6/83.2
4/2
49021
51083
50424
48914
unique reflections
18386
19953
17903
16331
completeness: overall/in hrbc (%)
average multiplicity/in hrb
mean I/σI: overall/in hrb
Rmerge: overall/in hrb (%)
99.1/99.0
3.2/1.0
6.7/1.0
9.1/42.7
96.7/90.1
4.1/3.6
13.6/3.5
6.3/31.3
96.1/89.7
5.1/4.5
10.2/3.4
5.0/30.6
96.3/89.4
4.3/1.3
7.6/2.7
7.7/39.1
13.0/2.7
8.4/43.8
Refinement Statistics
Rfreea (%)
22.6
16.3
0.016
1.6
24.5
29.4
20.7
0.027
2.4
25.7
17.6
0.023
2.1
28.1
22.6
0.021
2.5
28.5
20.3
0.024
2.4
27.9
19.8
0.015
1.6
Rcryst (%)
18.3
0.017
1.6
rms deviations:b
bonds (Å)
angles (deg)
B factor (Å2)
PDB code
2.3
2.8
2.3
2.6
2.8
3.1
2.3
2BR1
2BRB
2BRG
2BRH
2BRM
2BRN
2BRO
a Rfree is the R factor calculated using 5% of the reflection data chosen randomly and omitted from the refinement process, whereas
Rcryst is calculated with the remaining data used in the refinement. b The rms bond lengths and angles are the deviations from ideal
values; the rms deviation in B factors is calculated between covalently bonded atoms. c Highest resolution bin.
the solvent. Although one of these methoxy groups was
hydrogen bonded to Lys38 in mode A, this interaction
was not expected to contribute significantly to the
affinity of the compound for the site, given the intrinsic
flexibility of the lysine side chains and the competition
by water for this solvent-exposed site. It was therefore
predicted that if either mode A or B were correct, an
analogue (compound 2, Table 1) lacking the methoxy
groups (while keeping the original hydroxy-ethyl side
chain) would have an affinity similar to that of com-
pound 1. A more stringent test of docking mode A,
however, was to methylate the amino nitrogen (com-
pound 3) presumed to hydrogen bond the carbonyl
oxygen of Glu85.
Compounds 2 and 3 were synthesized with high
priority, as described above and in Scheme 1. As
predicted, compound 2 was virtually as active as com-
pound 1 with respect to Chk1 (Table 1). Compound 3,
however, was slightly more active than compounds 1
and 2. This implied that docking mode A was not
correct, but uncertainty remained about mode B, given
some of its peculiarities. In-house X-ray crystallography,
however, confirmed the overall relevance of this binding
mode.
Experimental Binding Mode. The crystal struc-
tures of compounds 1, 2, and 3 bound to Chk1 were
solved, with resolutions and refinement statistics given
in Table 3. These compounds were found to bind in the
ATP-binding site, consistent with the observation that
they are competitive inhibitors. These structures re-
vealed the same overall binding mode (Figure 2) for
compounds 1, 2, and 3. In view of this experimental
binding mode, docking mode B correctly predicted the
main contacts between the compound and the kinase
backbone and the intramolecular orientation of the
amino NsH bond toward the phenyl ring. The distance
(3.7 Å) between this amino nitrogen and the centroid of
the phenyl ring and the NsH‚‚‚centroid angle (149.0°)
are compatible with an interaction with some hydrogen
bond character where the phenyl ring acts as an
acceptor.54-56 Another likely acceptor is the water
molecule positioned 3.1 Å from the amino nitrogen (Ns
H‚‚‚O angle of 126.8°). Polar interactions involving
phenyl rings as acceptors are typically not represented
in current docking scoring functions,57 and yet the
present results indicate that they are relevant to drug
design. It is interesting to note that the polar character
of this interaction can be changed to a completely apolar
interaction with only minor structural adjustments, as
seen in the structure of compound 3, where the methyl
group branching off the amino nitrogen packs against
the face of the phenyl (not shown). The possibility for
such dual character may help interpret apparently
surprising SAR data in other contexts. Other interesting
interactions present in the X-ray binding mode include
the contacts between the two phenyl carbon atoms and
the carbonyl oxygen of Cys87 (Figure 2). The orientation
of these phenyl CsH bonds toward the carbonyl oxygen
suggests that they are favorable interactions.52,53,58 This
is supported by the observation of similar interactions
involving Phe237 and the carbonyl oxygen of Val123 in
protein kinase A (PKA) (Figure 5).
The main difference between docking mode B and its
experimental counterpart is in the positioning of the
compound hydroxy-ethyl side chain. In the crystal
structures, the distance between the hydroxyl oxygen
of the compound and Ser147 is ∼4.2 Å, indicating that
there is no hydrogen bond between these groups, in
contrast to model B. Instead, the compound hydroxyl
forms an intramolecular hydrogen bond with one of the
pyrimidine nitrogens (Figure 2) with an oxygen to
nitrogen distance of 3.0 Å. The compound side chain
adopts a more staggered conformation in the crystal
structure than in docking mode B, which contributes
to explain why the intramolecular hydrogen bond is
preferred.
The crystallographic conformation of the hydroxy-
ethyl side chain explains, by and large, the lack of
binding of the compounds listed in Table 2. Clearly, the
modifications to the hydroxy-ethyl side chain in com-
pounds 9, 10, 11, and 14 would disrupt the associated
intramolecular hydrogen bond, with no alternative
favorable interactions between the side chain and the
protein. The lack of binding of compounds 12 and 15
can be explained by steric clashes between the added
methyl (15) or phenyl (12) groups with the side chains
of Val23 or Leu84. The side chain of compound 14 would
also be poorly accommodated in the binding site.

4336 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Foloppe et al.
optimized toward computational speed, to allow the
virtual screening of large collections of compounds, and
are also commonly used to propose structural hypo-
theses for ligand design. Because of speed requirements,
these scoring functions involve a number of well-known
approximations57 in their treatment of molecular rec-
ognition, in particular with respect to electrostatics. A
more physically sound (but more computationally de-
manding) approach termed MM/PBSA (molecular me-
chanics/Poisson-Boltzmann surface area) has recently
emerged as a promising technique for improved predic-
tions of protein-ligand interactions.59-61 A rigorous
treatment of electrostatics, which accounts for desolva-
tion upon complex formation, is a key feature of MM/
PBSA calculations.
It is important to test if these calculations can help
discriminate between different docking modes of the
same ligand, but only limited work has been presented
in that direction.61 It is therefore interesting to inves-
tigate whether an MM/PBSA protocol could rank the
X-ray binding mode and docking mode B more favorably
than docking mode A, especially because mode A was
initially preferred to mode B on the basis of a count of
polar (electrostatic) interactions. The calculations were
performed with both compounds 1 and 2, to assess the
robustness of the results. Indeed, in docking mode A, a
methoxy group of compound 1 is hydrogen bonded to
the amino group of Lys38, but this interaction is not
present with compound 2. In the MM/PBSA framework,
the total free energy of binding ∆Gbinding can be decom-
posed into electrostatic (∆Gelec), van der Waals (∆Gvdw),
and nonpolar (∆Gnp) components. The electrostatic part
includes both ligand-protein interactions and solvation
effects, while the nonpolar contribution is usually
interpreted as representating the hydrophobic effect.59-61
Because we aimed to compare different binding modes,
we needed to take into account the internal conforma-
tional energy of the ligand (∆Ginternal_rel). Only relative
values of ∆Gbinding are required, and it is therefore
reasonable to neglect the entropic contributions (see
Methods Section). It follows that
Figure 2. Crystallographic binding mode of compound 2 in
the ATP-binding site of Chk1. Panel A shows the compound
2Fo-Fc electron density contoured at 2.0 σ (calculated for the
final refined structure) and the interactions between the
compound and the kinase hinge backbone. Classical hydrogen
bonds are depicted with green dotted lines, while CsH‚‚‚O
hydrogen bonds are traced in magenta. The contact between
the furan oxygen and the carbonyl oxygen of Cys87 is shown
with a blue dotted line. Water molecules are shown as red
spheres, including the water located 3.1 Å from the amino
nitrogen and the three waters in the targeted pocket (see main
text). Panel B shows the same binding mode viewed in the
plane of the compound aromatic scaffold with selected residues
labeled.
∆Gbinding ) ∆Gelec + ∆Gvdw + ∆Gnp + ∆Ginternal_rel (a)
The various components to ∆Gbinding for compounds 1
and 2 in their X-ray binding mode and docking modes
A and B are given in Table 4. It should be noted that
∆Gbinding is not appropriate to compare binding affinities
of compounds 1 and 2 because the internal energetics
of the ligands would cancel out in such a comparison.
The contribution of ∆Gelec was calculated for two values
of the solute dielectric constant ꢀ, to test the effect of
this parameter on the results. The choice of ꢀ affects
the absolute values of ∆Gelec, but has much less impact
on the corresponding differences between binding modes,
and does not alter the relevant rankings. In all cases,
∆Gelec acts against binding, reflecting the importance
of desolvation. It is interesting to note that for compound
2 ∆Gelec favors the X-ray binding mode over docking
mode A, despite a presumed electrostatic clash between
the furan and the carbonyl oxygens in the X-ray
structure. ∆Gnp also favors the X-ray binding mode over
docking mode A, although by a small margin. The
largest contribution opposing docking mode A is from
∆Ginternal_rel. This is consistent with the presence of an
A more general lesson from this binding mode is that
suboptimal protein-ligand interactions ought to be tol-
erated to some degree in virtual screening scoring func-
tions used at the hit discovery stage, which is typically
concerned with the discovery of ligands with initial
affinities in the micromolar range. This is illustrated
by the contact between the furan oxygen of compounds
1-3 and the carbonyl oxygen of Cys87, which are about
3.2 Å apart in the crystal structures. The targeted
manipulation of this interaction led to improvement of
the binding affinity of the series for Chk1.
Alternate Binding Modes as a Test for More
Sophisticated Calculations. The initial docking mode
A of compound 1 (Figure 1) was obtained with a docking
empirical scoring function. Such functions are typically

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4337
Table 4. Relative MM/PBSA Energeticsa of Compounds 1 and 2 in the X-ray Binding Modeb and in Docking Modes A and B
∆Gelec
∆Gvdw
∆Gnp
∆Ginternal_rel
∆Gbinding
∆∆Gbinding
compound 1
-22.4
X-rayb
11.7 (ꢀ ) 2)
6.0 (ꢀ ) 4)
7.8 (ꢀ ) 2)
3.5 (ꢀ ) 4)
9.2 (ꢀ ) 2)
4.6 (ꢀ ) 4)
-39.7
-40.3
-37.5
0.0c
14.3
2.0
-50.4 (ꢀ ) 2)
-56.1 (ꢀ ) 4)
-39.7 (ꢀ ) 2)
-44.0 (ꢀ ) 4)
-49.0 (ꢀ ) 2)
-53.6 (ꢀ ) 4)
docking mode A
docking mode B
-21.5
-22.7
10.7 (ꢀ ) 2)
12.1 (ꢀ ) 4)
1.4 (ꢀ ) 2)
2.5 (ꢀ ) 4)
compound 2
-20.2
X-rayb
7.2 (ꢀ ) 2)
3.5 (ꢀ ) 4)
8.1 (ꢀ ) 2)
4.3 (ꢀ ) 4)
8.4 (ꢀ ) 2)
3.8 (ꢀ ) 4)
-36.2
-38.6
-38.7
0.0c
21.3
3.7
-49.2 (ꢀ ) 2)
-52.9 (ꢀ ) 4)
-28.8 (ꢀ ) 2)
-32.6 (ꢀ ) 4)
-46.3 (ꢀ ) 2)
-50.9 (ꢀ ) 4)
docking mode A
docking mode B
-19.6
-19.7
20.4 (ꢀ ) 2)
20.3 (ꢀ ) 4)
2.9 (ꢀ ) 2)
2.0 (ꢀ ) 4)
a All energies are in kcal/mol. ∆Gelec is reported for a solute dielectric constant ꢀ of 2 or 4. ∆∆Gbinding ) ∆Gbinding (X-ray) - ∆Gbinding
(docking mode). b Refers to the X-ray conformation of the ligand, not the X-ray coordinates of the protein-ligand complex. Calculations
reported in this table were obtained with the protein X-ray structure 1IA8 from the PDB, used to generate docking modes A and B.
c Internal energies are differences relative to the most stable internal energy, set to 0.0.
Table 5. Binding Free Energiesa of Compounds 2 and 5 Calculated with MM/PBSA
protein structureb
∆Gelec
∆Gvdw
-37.3
∆Gnp
∆Gbinding
compound 2
compound 5
protein X-ray structure solved with cpd 2
“native X-ray structure”
protein X-ray structure solved with cpd 5
“exchanged X-ray structure”
protein X-ray structure solved with cpd 5
“native X-ray structure”
protein X-ray structure solved with cpd 2
“exchanged X-ray structure”
4.4 (ꢀ ) 2)
2.4 (ꢀ ) 4)
5.5 (ꢀ ) 2)
2.3 (ꢀ ) 4)
4.1 (ꢀ ) 2)
2.2 (ꢀ ) 4)
4.2 (ꢀ ) 2)
1.6 (ꢀ ) 4)
-19.2
-52.1 (ꢀ ) 2)
-54.1 (ꢀ ) 4)
-52.2 (ꢀ ) 2)
-55.4 (ꢀ ) 4)
-53.3 (ꢀ ) 2)
-55.2 (ꢀ ) 4)
-51.6 (ꢀ ) 2)
-54.2 (ꢀ ) 4)
-37.9
-38.0
-36.7
-19.8
-19.4
-9.1
a All energies are in kcal/mol. ∆Gelec is reported for a solute dielectric constant ꢀ of 2 or 4. b Protein X-ray structure used in the calculations.
Scheme 2. Synthesis of Pyrrolopyrimidine Derivatives
intramolecular hydrogen bond in the experimental
binding mode, without a counterpart in docking mode
A. This, however, is not enough to explain differences
of > 10.0 kcal/mol in ∆Ginternal_rel between docking mode
A and the other two modes. Additional analysis revealed
that these large differences also reflect unfavorable van
der Waals contacts between the hydroxy-ethyl side
chain and the adjacent phenyl ring in docking mode A
tions and confirm that the MM/PBSA strategy holds
promise to discern between multiple docking modes of
a same compound.
Structure-Based Design. To improve the affinity of
the furanopyrimidine series (compounds 1-4) for Chk1,
the first suggestion was to replace the furan by a pyrrole
in its scaffold. Indeed, simple visualization of the crystal
structures suggested that replacing the furan oxygen
by a pyrrole NsH would create an additional hydrogen
bond between the ligand scaffold and the protein. This
led us to synthesize compound 5, with a pyrrolopyrimi-
dine scaffold (Table 1).
The pyrrolopyrimidines were prepared from the com-
mercially available analogue 21 (Scheme 2). This was
debenzylated with aluminum chloride in refluxing
toluene to give the derivative 22 in 54% yield, based on
the recovered starting material. The anilino derivatives
were prepared by microwave irradiation of precursor 22
with 6 equiv of the corresponding amine in ethanol at
100 °C. Purification of the crude material using column
chromatography gave the pure pyrrolopyrimidines in
10-20% yield.
(Figure 1). Recalculation of the ∆Ginternal_rel contributions
62
with the Merck molecular force field
confirmed the
trends in ∆Ginternal_rel obtained with the CHARMm force-
field between binding mode A and the other modes (not
shown). One, however, could not rule out docking mode
A solely on the basis of its high internal strain energy.63
The net result is values of ∆Gbinding which clearly favor
the X-ray binding mode over docking mode A, by a
difference of ∆∆Gbinding which is > 10.0 kcal/mol (Table
4). These calculations also clearly discriminate between
docking modes A and B, in favor of the latter. ∆Gbinding
also favors slightly the X-ray binding mode over docking
mode B but by differences which may be of the same
order as the error margin of the calculations. Overall,
these results stress the importance of a detailed treat-
ment of the intramolecular energetics in scoring func-
Compound 5 led to a 10-fold increase in binding
affinity to Chk1 over compound 2 (Table 1). The crystal

4338 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Foloppe et al.
intermediary compounds, the crystal structures of which
would act as a guide to reach the next stage.
Addition of a methyl group to compound 5 gave
compound 6 and led to an apparent slight decrease in
affinity when tested as a racemate (Table 1). Given the
introduction of a chiral center in compound 6, and that
only one of these stereoisomers may bind to Chk1, the
actual affinity of the active stereoisomer may be the
same as that of compound 5. Importantly, however, the
crystal structure of compound 6 bound to Chk1 con-
firmed that the methyl group was added toward the
targeted pocket (Figure 3). The associated crystal-
lographic electron density could be interpreted such that
only the R stereoisomer was bound to Chk1. Compound
7 was made by building on compound 6, with the
intention that the added hydroxyl group may hydrogen
bond to Ser147. This was, however, not observed in the
crystal structure of compound 7, where the added
hydroxyl group points away from both Ser147 and from
the nearby pocket. Therefore, additional efforts will be
needed to test if this pocket can be filled, with a
concomitant favorable impact on the binding affinities.
It remains, however, that this work presents significant
progress toward that goal.
Figure 3. Crystal structure of compound 6 bound to Chk1.
The protein is represented as a gray molecular surface. The
black arrow points to the methyl group added to the hydroxy-
ethyl side chain. This methyl group projects toward a buried
pocket which contains three water molecules (red spheres).
Experimental Probing of the Strength of a
Protein-Ligand Hydrogen Bond. Like in the field
of protein folding and stability,65,66 the net energetic
contribution of ligand-protein hydrogen bonds to the
binding affinity of these complexes has raised immense
interest67-73 and remains controversial.74-76 One aspect
of the controversy is that the chemical functions which
hydrogen bond in the complex are also expected to form
hydrogen bonds with water when ligand and protein are
dissociated. Therefore, it is sometimes argued that
hydrogen bonds between solutes contribute very little,
if at all, to their free energy of association in water.77,78
Experimental studies of these questions have been
hampered by the difficulty of systematically adding or
removing exactly one hydrogen bond in the complexes,
and therefore report of such systems is of general
interest. These questions are particularly intriguing
with kinases because they have become widely targeted
in a number of therapeutic areas and most of their
ligands form several hydrogen bonds with their back-
bone in the hinge region.49 The thoroughly measured
binding affinities (Figure 4) combined with the virtually
identical binding modes of compounds 2 and 5, ascer-
tained by crystallography, provide a rare opportunity
to relate our results to these issues. The change from a
furan in compound 2 to a pyrrole in compound 5 is
arguably as minor as experimentally possible. Because
this change does not add or remove any flexible torsion,
no compound internal strain or conformational entropies
are involved, not even an entropy loss associated with
restraining a torsional motion of the hydrogen mediat-
ing the hydrogen bond.79 Also, the protein is expected
to undergo the same desolvation upon binding of each
ligand. Changing the furan to a pyrrole adds a standard
and buried hydrogen bond in the complex with com-
pound 5, while deleting an oxygen‚‚‚oxygen contact. In
addition, a simple inventory of the hydrogen bonds
present in the relevant equilibria69 indicates that one
hydrogen bond to the furan oxygen is lost when this
oxygen is transferred from water to the protein binding
structure of compound 5 bound to Chk1 confirmed that
the pyrrole N-H hydrogen bonds to the backbone
carbonyl of Cys87, with standard geometries (N‚‚‚O
distance ) 3.1 Å, N‚‚‚O-C angle ) 126.8°). The implica-
tions of this result for the contribution of hydrogen
bonds to the association free energy of ligand-protein
complexes are discussed further below. Another strategy
to improve the affinity of the compounds to Chk1 would
be to target a pocket located in the vicinity of the ligand
hydroxy-ethyl side chain (Figures 2 and 3). This pocket
is lined by Asn59, Leu84, and Phe149, and the electron
density defining the crystal structure shows that three
water molecules occupy the pocket in the present
ligand-Chk1 complexes. This region is a potentially
attractive target for gaining potency against Chk1
because this pocket is well buried at the bottom of the
ATP-binding site (Figure 3) and should allow the
formation of additional tight interactions between ligand
and protein. It might also be entropically favorable to
displace the water molecules present in the pocket into
the bulk solvent.64
To reach into this pocket of interest, simple proximity
considerations dictate that derivatization of the com-
pound’s hydroxy-ethyl side chain is the only option
available (Figures 2 and 3). In this side chain, the C-H
vectors emanating from the carbon bound to the amino
group do not point in directions which could be used to
reach the pocket. One of the C-H vectors from the other
side chain carbon, however, is projected in a direction
which could be amenable to derivatization toward the
pocket. The second C-H vector from the same carbon
could allow targeting of Leu137 and Ser147, which
would provide another region of interest to make
additional interactions with the protein. Given the strict
geometric constraints imposed by the protein environ-
ment, it was decided to adopt a stepwise approach,
whereby the hydroxy-ethyl side chain would be deriva-
tized incrementally rather than add complex substitu-
ents from the start. This should allow us to obtain useful

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4339
Figure 4. Double reciprocal (Lineweaver-Burke) plots using the Chk1 kinase assay, in the presence of compounds 5 (panel A)
and 2 (panel B), showing the inverse of the reaction rate (1/CPM where CPM is the number of radioactive counts per minute)
versus the inverse of the substrate concentration (1/[ATP]), with increasing inhibitor concentrations. The plots for each compound
intersect on the 1/CPM axis, indicative of a simple competitive inhibitor. Ki values were determined by plotting the slopes of each
plot against inhibitor concentration and the -Ki value equaling the negative intercept on the y-axis (data not shown).
site, although there is no such loss with the pyrrole.
Therefore, the higher affinity of compound 5 than that
of compound 2 for the protein reflects the removal of
unfavorable contributions to complex formation with
compound 2, as well as an additional hydrogen bond in
the complex of compound 5.
favor the protein-ligand association only weakly. This
is consistent with the components to the free energy of
binding calculated for compounds 2 and 5.
Computational Analysis of the Influence of the
Added Hydrogen Bond on Binding. Hydrogen bonds
are essentially electrostatic in nature. Therefore the
MM/PBSA formalism, introduced above, should in prin-
ciple allow us to dissect the atomistic contribution of a
hydrogen bond to the free energy of association of a
ligand and a protein. This may be approached by
comparing the components to the free energy of binding
for compounds 2 and 5 (Table 5). When the free energies
of the binding of two distinct compounds are compared,
the internal energy between bound and free ligand
cancel out for each ligand, and ∆Gbinding is reduced to
The difference in free energy of binding between these
two complexes (∆∆G), derived from their Ki values with
the law of mass action at the temperature of the
experiments (303 K), yields ∆∆G ) 1.4 kcal/mol. A
detailed physicochemical analysis of this result, includ-
ing a dissection of enthalpic and entropic contributions,
is beyond the scope of this work. Nevertheless, it is
interesting to reflect on the practical implications of this
result for ligand design. Given the above-mentioned
differences regarding the relevant equilibria, it follows
that the value of ∆∆G is an upper limit for the net
contribution of the added hydrogen bond in the complex
with compound 5. This value is significantly less than
the value of 4.0 kcal/mol reported in a similar compari-
son,67,68 possibly because the relatively weak hydrogen
bond acceptor character of a furan oxygen80 would
facilitate its desolvation upon complex formation. Our
observations are compatible with the 0.5-1.8 kcal/mol
range derived for a neutral hydrogen bond for complex
formation in an aqueous solvent.69 They are also con-
sistent with an analysis which concluded that the
binding of balanol to PKA is mostly driven by the burial
of nonpolar groups.60 In sum, the present results add
to the view that the net hydrogen-bonding energetics
∆Gbinding ) ∆Gelec + ∆Gvdw + ∆Gnp
(b)
Some authors60,81 have argued that the ∆Gvdw contribu-
tion is implicitly contained in ∆Gnp, but this discussion
has little bearing on the question at hand, because the
influence of the hydrogen bond should be primarily
reflected in ∆Gelec
.
For both compounds 2 and 5, ∆Gbinding was calculated
with the structure of the protein in which the compound
was crystallized (“native X-ray structure”, Table 5),
which yields slightly different components of ∆Gbinding
for compound 2 compared to those presented in Table
4. Values of ∆Gbinding obtained with these native X-ray
structures are more favorable for compound 5 than for

4340 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Foloppe et al.
compound 2, by about 1 kcal/mol, in agreement with the
experimental results (see above). To isolate the contri-
bution of the hydrogen bond of interest, however, it is
desirable to compare compounds 2 and 5 with the same
protein structure, which required additional calculations
(“Exchanged X-ray structure”, Table 5). Overall, the
∆Gelec contribution is only marginally more favorable
with compound 5 than with compound 2, which supports
the view that the added hydrogen bond in the complex
with compound 5 contributes only weakly to the free
energy of association with the protein.
Although the protein structure is very similar whether
crystallized with compound 2 or 5, small structural
differences are enough to impart variations to ∆Gbinding
of a given compound, which are on the same order as
the differences in ∆Gbinding between compounds. Because
of these structure-dependent variations, it may be more
accurate to extract the MM/PBSA energies from a
thermally equilibrated ensemble of structures, such as
those generated by molecular dynamics, but this is
beyond the scope of this work. A detailed experimental
characterization of the thermodynamics of binding of
compounds 2 and 5 would also be highly valuable.
Selectivity. Different kinases are involved in a
variety of cell signaling pathways,82 but X-ray structures
combined with sequence alignments have shown that
many features of the ATP-binding pocket tend to be
conserved across kinases.48-50 Therefore selectivity of
inhibition for a particular kinase is a well-known issue,
and we initiated work to investigate this question for
Chk1 and its inhibitors presented here (Table 1).
Human PKA was used as a first test for selectivity
because it is easily available and still closely related to
Chk1 phylogenetically,82 with a 29% sequence identity
between the two sequences (shortest sequence used for
total length). The IC50 of Chk1 inhibitors 1-7 for PKA
was > 200 mM, and this lack of inhibition for PKA can
be understood in structural terms. An alignment of the
crystal structures of PKA and the Chk1-compound 2
complex unambiguously reveals steric clashes between
some PKA side chains and this compound (Figure 5).
These clashes would also be present with the other
compounds in Table 1 and therefore explain the selec-
tivity between Chk1 and PKA.
A more functionally relevant test regarding selectivity
is the comparison of Chk1 inhibition with that of human
CDK1. CDK1 activity is needed for cells to go through
the G2/M checkpoint and is therefore essential for the
therapeutic strategy presented in the Introduction. It
follows that ideal Chk1 inhibitors should not target
CDK1. The data presented in Table 1 show that
compounds in the pyrrolopyrimidine series are as potent
against CDK1 as Chk1; however several furanopyrimi-
dine compounds showed a promising degree of selectiv-
ity in the desired direction. This suggests that differ-
ential selectivity between Chk1 and CDK1 is achievable
and could probably be improved with further medicinal
chemistry efforts.
Figure 5. Superposition of the crystal structure of PKA (green
carbon atoms, PDB entry 1APM) with Chk1 (gray carbon
atoms) in complex with compound 2. The conserved hinge
backbones of both proteins are very closely aligned. The PKA
side chains (Met120, Thr183, and Phe327) which would
sterically clash with compound 2 are shown and labeled, as
well as their counterpart in Chk1. The phenyl ring of Phe327
ovelaps with a phenyl ring of compound 2. The hydroxyl
oxygens of Thr183 and of the compound side chain are only
1.9 Å apart. The carbon at the tip of the Met120 side chain
(Cꢀ) is 1.7 Å from the compound side chain oxygen.
these novel ligands have the advantage of being “drug-
like” and can be synthesized without the supply of a
natural product.
A chronological account of this work has the merit to
illustrate how structural information about the target
allowed the formulation and testing of well-defined
hypotheses to guide medicinal chemistry. In particular,
a few lessons may be gleaned from an objective com-
parison of the earlier molecular modeling suggestions
and the subsequent results from crystallography. Two
reasonable model docking modes could be generated
quickly at the outset of the project for the newly
discovered furanopyrimidine inhibitor. These models
helped to prioritize medicinal chemistry efforts, for
instance in identifying parts of the ligand which con-
sistently appeared to point toward the solvent. A simple
count of presumably favorable polar interactions with
the kinase backbone led to the selection of one of these
docking modes as the initial preferred working hypo-
thesis. This hypothesis was very convincing in view of
a conventional analysis of the ligand-protein interac-
tions and most elements of the SAR available at this
stage. Despite these arguments, it was deemed neces-
sary to specifically test this structural model (compound
3). This test proved crucial in determining that the
initial structural model was incorrect. This illustrates
the importance of using molecular modeling as part of
an iterative cycle of hypothesis generation and rigorous
testing. This approach was vindicated when it became
apparent that the second docking mode was close to the
experimental binding mode determined by crystal-
lography. This experience is of interest for ligand design
against kinases in general, given that the possibilities
of alternative binding modes are particularly pertinent
with these proteins, because of the symmetry of their
conserved hinge backbone in the ATP-binding site.83
The crystal structures of the novel furanopyrimidine
inhibitors bound to Chk1 provided a firm basis for
further elaboration of these compounds. These struc-
tures not only explained most of the previously gathered
Conclusions
We reported here the discovery, synthesis, and crys-
tallographic binding mode of novel furanopyrimidine
and pyrrolopyrimidine inhibitors of the Chk1 kinase,
which is a target of great therapeutic interest for
oncology. In contast with some other Chk1 inhibitors,

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4341
reaction mixture was poured onto crushed ice (100 g) and
neutralized with sodium carbonate solution. The product was
then extracted with chloroform and dried over magnesium
sulfate. The residue was purified by passing through a silica
gel column using chloroform-petroleum ether as the eluent,
to afford the product (2.00 g, 75%) as a white solid: mp 121-
SAR but led to preparation of the more potent pyrrol-
opyrimidine analogues. Changing the ligand furan
moiety to a pyrrole leads to the formation of one
additional regular hydrogen bond between the ligand
and the kinase backbone, which translates into a 10-
fold increase in affinity. A detailed qualitative analysis
of the factors involved, however, indicates that this
additional hydrogen bond is expected to contribute only
a fraction of the gain in affinity. Such observation may
be of general interest, given the ongoing debate about
the energetic role of hydrogen bonding in molecular
recognition74-76 and the rarity of structurally character-
ized pairs of ligand-protein complexes differing by only
one hydrogen bond. This adds credence to the view that
the burial of hydrophobic surface area is the main drive
for protein-ligand complex formation.
1
122 °C; MS EI m/e 307 (M+ + H); NMR (DMSO-d6) δ 7.25-
7.70 (10H, m) 8.86 (1H, s).
Method A: 2-(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-
ylamino)-ethanol (2). The chloropyrimidine 19 (0.27 g, 0.88
mmol) and ethanolamine (0.12 g, 1.9 mmol) were refluxed in
ethanol with for 8 h. After this time, the reaction was
concentrated, and the crude material was purified by passing
through a silica gel column using chloroform-petroleum ether
as the eluent to afford the product (0.25 g, 86%) as a white
solid: mp 163-165 °C; MS EI m/e 332 (M+ + H); 1NMR
(CDCl3) δ 3.54-3.64 (2H, m); 3.68-3.78 (2H, m); 3.97-4.75
(1H, m); 5.19 (1H, m); 7.21-7.64 (10H, m); 8.40 (1H, s). Anal.
(C20H17N3O2) C, H, N.
The crystal structures of the present ligand-Chk1
complexes suggest that burying added apolar groups in
a nearby pocket could be a strategy to improve the
affinity of these compounds for Chk1. This requires
derivatization of the hydroxy-ethyl side chain of these
compounds from the carbon linked to the hydroxyl
group. Indeed, preliminary results showed that the
derivatizion from this position can be achieved without
disruption of the binding mode or degradation of the
binding affinities. We suggest that longer substituents
specifically designed to reach into this nearby pocket
should lead to significant improvements in affinity and
presumably selectivity.
2-[(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-yl)-methyl-
amino]-ethanol (3). This was prepared according to method
A using N-methyl ethanolamine, affording the product as an
off-white solid: mp 79-81 °C; MS EI m/e 346 (M+ + H); 1NMR
(DMSO-d6) δ 2.58 (3H, s); 3.17-3.50 (4H, m); 4.31-4.80 (1H,
br. s); 7.25-7.62 (10H, m); 8.32 (1H, s). Anal. (C21H19N3O2) H,
N. C: calcd, 71.73; found, 73.03.
(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-ylamino)-ace-
tic Acid (4). This was prepared according to method A using
glycine, affording the product as an off-white solid: mp >240
°C; MS EI m/e 346 (M+ + H); 1NMR (DMSO-d6 δ 2.60 (1H,
br); 4.20 (2H, d, J ) 6 Hz); 5.60 (1H, br); 7.20-7.65 (10H, m);
8.40 (1H, s). Anal. (C21H19N3O2) C, H, N.
2-(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-ylamino)-2-phen-
yl-ethanol (12). This was prepared according to method A
using 2-phenyl ethanol, affording the product as an off-white
solid: mp 71-73 °C; MS EI m/e 408 (M+ + H); 1NMR (CDCl3)
δ 3.66-3.91 (2H, m); 3.76-3.84 (1H, m); 4.24-4.71 (1H, m);
5.19-5.71 (2H, m); 6.90-7.64 (15H, m); 8.39 (1H, s).
N-(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-yl)-ethane-1,2-
diamine-trifluoroacetate Salt (13). This was prepared
according to method A using (2-amino-ethyl)-carbamic acid
tert-butyl ester followed by treatment with trifluoroacetic acid.
Recrystallization of the product afforded the trifluoroacetate
salt as a white solid: mp 196-198 °C (dec); MS EI m/e 331
Experimental Section
Purchased Compounds. Every purchased compound was
subjected to an NMR and liquid chromatography-mass spec-
trometry analysis to check that they were consistent with the
structures shown in Table 2.
2-[5,6-Bis-(4-methoxy-phenyl)-furo[2,3-d]pyrimidin-4-
ylamino]-ethanol (1) was purchased from ChemDiv.
5,6-Diphenyl-furo[2,3-d]pyrimidin-4 -ylamine (8) was
purchased from Chembridge.
3-(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-ylamino)-pro-
pan-1-ol (9) was purchased from Chem Div.
Benzyl-(5,6-diphenyl-furo[2,3-d]pyrimidin-4-yl)-amine
(10) was purchased from IBS labs.
1
(M+ + H); NMR (CDCl3) δ 3.09-3.25 (2H, br. s); 3.52-3.68
(2H, br.s); 5.09 (1H, m); 7.17-7.64 (11H, m); 8.36 (1H, s); 8.69-
9.15 (2H, br. s).
(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-yl)-(2-methoxy-
ethyl)-amine (14). This was prepared according to method
A using O-methylethanolamine affording the product as a
white solid: mp 115-117 °C; MS EI m/e 346 (M+ + H); 1NMR
(DMSO-d6) δ 3.12 (3H, s); 3.24-3.41 (2H, m); 3.44-3.58 (2H,
m); 4.31-5.19 (1H, m); 7.28-7.71 (10H, m); 8.35 (1H, s).
Synthesis of Pyrrolopyrimidines. This was carried out
according to Scheme 2. 4-Chloro-5,6-diphenyl-7H-pyrrolo-
[2,3-d]pyrimidine (22). The commercially available 7-Benzyl-
4-chloro-5,6-diphenyl-7H-pyrrolo[2,3-d]pyrimidine (21, from
Bionet Research) (1.00 g, 2.50 mmol) was suspended in toluene
(20 mL). To this was added aluminum trichloride (2.40 g, 18.00
mmol) and the reaction mixture heated to 100 °C overnight.
The reaction mixture was cooled and partitioned between
water and dichloromethane. The organic layer was then dried
and evaporated. The residue was purified by passing through
a silica gel column using ethyl acetate-hexane as the eluent
to afford the product (0.25 g, 54% based on recovered starting
material) as an off-white solid: MS EI m/e 345 (M+ + H);
1NMR (DMSO-d6) 7.376-7.5 (10H, m); 9.70 (1H, s).
N-(5,6-Diphenyl-furo[2,3-d]pyrimidin-4-yl)-benzamide
(11) was purchased from IBS labs
Synthesis of Furanopyrimidines. This was carried out
according to Scheme 1. Compounds 16-19 were prepared
according to procedures found in ref 47.
N-(3-Cyano-4,5-diphenyl-furan-2-yl)-formamide (17).
Acetic formic anhydride was prepared by heating anhydrous
formic acid (60 mL) and acetic anhydride (60 mL) 60 °C for 2
h. The cyanofuran (13.00 g, 50 mmol) was then added to the
mixture and heated at 85 °C for 6 h. The reaction mixture
was evaporated to about half the original volume and neutral-
ized to pH 6 at 0 °C with 2 M NaOH solution. The precipitate
produced was filtered off and dried. This gave the crude
product as a white solid (11 g, 72%) that was used without
further purification: 1NMR (CDCl3) δ 7.27-7.59 (10, m); 8.19
(1H, s); 12.67 (1H, s).
5,6 Diphenyl-3H-furo[2,3-d]pyrimidin-4-one (18). The
formamide 17 (3.00 g, 98 mmol) was stirred for 10 min at 220
°C under nitrogen. After the crude material was cooled, it was
crystallized from methanol, giving the pure product (2.50 g,
88%) as a white solid: mp >240 °C; MS EI m/e 289 (M+ + H);
1NMR (DMSO-d6) δ 7.27-7.59 (10H, m); 8.19 (1H, s); 12.67
(1H, s).
Method B: 2-(5,6-Diphenyl-7H-pyrrolo[2,3-d]pyrimi-
din-4-ylamino)-ethanol (5). The chloropyrrolopyrimidine 22
(0.02 g, 0.08 mmol) and ethanolamine (14 µL, 0.24 mmol) were
heated to 150 °C under microwave conditions for 60 min. After
this time, the solvents were evaporated and the residue was
purified by passing through a silica gel column using ethyl
acetate as the eluent to afford the product (0.05 g, 20%) as a
4-Chloro-5,6-diphenyl-furo[2,3d-d]pyrimidine (19). The
pyrimidone 18 (2.50 g, 8.7 mmol) and phosphorus oxychloride
(25 mL) were stirred at 55 °C for 2 h. After this time, the

4342 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Foloppe et al.
solid: mp 245-249 °C; MS EI m/e 331 (M+ + H); 1NMR
(CDCl3) δ 3.50-3.60 (2H, q, J ) 5 Hz); 3.70-3.75 (2H, m);
5.25-5.30 (1H, m); 7.21-7.45 (10H, m); 8.20 (1H, m); 12.74
(1H, br. s). Anal. (C20H18N4O) H, N. C: calcd, 72.09; found,
72.71.
of the solutes was defined as the volume inaccessible to a
solvent probe sphere of radius 1.4 Å.98 The nonpolar contribu-
tion ∆Gnp of the free energy was calculated as in previously
described protocols,60,61 which assume that it is proportional
(γ ) 0.025 kcal/mol Å2) to the decrease in solvent accessible
surface area (SASA) upon complexation
1-(5,6-Diphenyl-7H-pyrrolo[2,3-d]pyrimidin-4-ylamino)-
propan-2-ol (6). This was prepared according to method B
using 1-amino-propan-2-ol, affording the product as solid: mp
∆Gnp
)
1
257-262 °C; MS EI m/e 345 (M+ + H); NMR (CDCl3) δ 1.20
γ(SASAcomplex - (SASAprotein alone + SASAligand alone)) (c)
(3H, d, J ) 6.3 Hz); 3.30-3.40 (1H, ddd, J ) 14.3, 7.0, 5.5
Hz); 3.50-3.60 (1H, ddd, J ) 14.0, 6.2, 2.4 Hz); 3.85-3.95 (1H,
m); 5.20-5.25 (1H, m), 7.20-7.50 (10H, m), 8.20 (1H, s); 12.95
(1H, br. s). Anal. (C21H20N4O) H, N. C: calcd, 72.55; found,
73.23.
The van der Waals interaction energy between ligand and
protein (∆Gvdw) and the ligand relative internal conformational
energy (∆Ginternal_rel) also need to be added to the final relative
free energy of binding ∆Gbinding when comparing binding modes
of a same ligand (Eqn a). ∆Ginternal_rel is the internal energy of
the bound ligand relative to that of the free ligand. The
thermodynamic cycle used assumes that the conformations of
the free and bound ligands are identical. Therefore, the bound
conformation of lowest internal energy was taken to represent
the reference internal energy of the free ligand, and the
∆Ginternal_rel components were offset relative to this reference.
∆Gbinding neglects translational, rotational, and vibrational
entropy changes upon complexation; however, it is expected
that these contributions would be very similar for different
binding modes of the same ligand and would cancel out when
comparing ∆Gbinding for these binding modes.
3-(5,6-Diphenyl-7H-pyrrolo[2,3-d]pyrimidin-4-ylamino)-
propane-1,2-diol (7). This was prepared according to method
B using 3-amino-propane-1,2-diol, affording the product as a
1
solid: 228-234 °C; MS EI m/e 361 (M+ + H); NMR (CDCl3)
δ 3.48-3.62 (2H, m); 3.68-3.76 (1H, m); 5.15-5.20 (1H, m);
7.20-7.50 (10H, m), 8.20 (1H, s); 12.95 (1H, br. s). Anal.
(C21H20N4O2) H. C: calcd, 67.94; found, 69.98. N: calcd, 15.10;
found, 15.54.
Molecular Modeling. Docking was performed with the
program rDock which is an extension of the program Ri-
boDock84 using an empirical scoring function calibrated based
on protein-ligand complexes.85,86 The Monte Carlo/simulated
annealing protocol initially used in RiboDock was replaced by
a steady-state Genetic Algorithm (GA) to improve the efficiency
of the docking search. Ligand docking poses were represented
using a conventional chromosome representation of transla-
tion, rotation, and rotatable bond dihedral angles. A single GA
population was used, of size proportional to the number of
ligand rotatable bonds, with a mutation/crossover ratio of 3:2.
The overall orientation and internal conformation of the
compounds were searched with the GA, while the receptor site
was kept fixed.
Pictures of 3D molecular structures were prepared with
PyMOL.99
Expression Vector Construction. Amino acids 1-289 of
Chk1 were amplified from IMAGE clone 5181427 with primers
MoloRT473 (TAC CGG GGA TCC ACC ATG GCA GTG CCC
TTT GTG GAA GAC T) and MoloRT474 (GTG TTC CTC GAG
TCA TTA GTG GTG GTG ATG GTG ATG GTG GTG TCC ACT
GGG AGA CTC TGA CAC ACC) by the polymerase chain
reaction (PCR) and subcloned into pCR2.1 using the TopoTA
cloning system (Invitrogen) to create the plasmid pCR2.1 Chk1
1-289 C8H. The insert was sequenced to verify that no errors
were introduced during the PCR reaction. A 900 bp EcoRI-
XhoI fragment from pCR2.1 Chk1 1-289 C8H was subcloned
into pFastBac1 to create the plasmid pFastBac1/Chk1 1-289
C8H. Recombinant bacmids were prepared using the Bac-to-
Bac system according to the manufacturer’s instructions
(Invitrogen). Bacmids were confirmed by PCR analysis (data
not shown). This bacmid encodes amino acids 1-289 of Chk1
encompassing the kinase domain with a C-terminal 8-His tag.
Chk1 Baculoviral Protein Expression and Purifica-
tion. Bacmids were transfected in Sf9 cells and used to
generate high titer stocks. Expression of Chk1 1-289 C8H was
achieved in Sf9 cells with a multiplicity of infection (MOI) of
0.5 for 3 days. Cell pellets were lysed using Popculture
(Novagen), and Chk1 was purified with the following scheme.
The initial step performed was Q-Sepharose, followed by Ni-
affinity, hydrophobic interaction chromatography (using Source-
Phenol), and finally gel filtration. This provided us with >95%
pure protein. However, we routinely obtained low yields with
this construct (< 1 mg of purified Chk1/L of Sf9 cells)
presumably because of the high endogenous activity of the
isolated kinase domain of Chk1.42
Docking used the ATP-binding site in the crystal structure
87
of apo human Chk1 (protein data bank (PDB) entry 1IA8).
Water molecules were removed from the coordinates, and polar
hydrogens were added to the protein using the CHARMm force
field.88 To specify the docking volume, the ATP-bound crystal
structure of PKA (PDB entry 1ATP) was superimposed onto
that of Chk1. There was unambiguous superimposition of the
ATP-binding sites, and the docking volume in Chk1 was
defined as the space within 8 Å of the ATP molecule. Before
docking, the three-dimensional (3D) structure of the com-
89
pounds was built with the software MOE
and energy
minimized with the Merk molecular force field.90 Each com-
pound was subjected to 200 docking runs, and the output
docking modes were analyzed by visual inspection in conjunc-
tion with the docking scores.
The MM/PBSA calculations were performed using the
program CHARMm,91,92 with the protein X-ray structures
indicated in Table 4 and Table 5. The partial atomic charges
for the compounds were derived by fitting to the molecular
electrostatic potential and the dipole moment93 calculated
quantum mechanically at the restricted Hartree-Fock/6-31G*
level with the program GAMESS.94 The CHARMm force field
for small organic molecules95 was applied to the compounds
in combination with the CHARMm all-atom protein force
field.88 Prior to MM/PBSA calculations, the compounds were
energy minimized while keeping the protein nonhydrogen
atoms fixed at their X-ray coordinates. The electrostatic
contribution ∆Gelec to the total free energy of binding ∆Gbinding
was obtained with the PBEQ module of CHARMm,96 using the
Born radii specially derived for this type of calculations.97 The
aqueous solvent was assigned a dielectric constant of 80.0. Two
values of the solute dielectric constant (2 or 4) were used, to
test the effect of this parameter on the results. The ionic
strength was set to 145 mM and the temperature to 298 K.
The electrostatic potentials were calculated using the grid
focusing technique, using two successive grids with spacings
of 1.0 and 0.3 Å. The first and second grids, extended by at
least 20 and 5 Å, respectively, around the complex. The interior
Chk1 Kinase Assay. Compounds were titrated in a twofold
dilution series in 100% DMSO in a 96-well plate (V-bottom
clear plate, VWR 007/008/257) at 50 times the final concentra-
tion. Subsequently, 1 µL of this series was added to a separate
96-well plate, to which a mixture of 10 µL of buffer (250 mM
HEPES-NaOH, pH 7.5), 0.5 µL of BSA (10 mg/mL), 2 µL of
Chktide (1 mg/mL), 0.5 µL of full-length Chk1 enzyme (10
mM), and 26 µL of double-distilled water was added per well.
To start the assay, a second mixture of 0.75 µL of 1M
magnesium acetate, 0.05 µL of ATP (100mM), 0.05 µL of 33P
γ-ATP (assuming specific activity of 10 mCi/mL), and 9.15 µL
of double-distilled water was added per well. The control wells
contained 1 µL of 100% DMSO or 1 µL of 50 µM Staurosporine
(Calbiochem #569397) to give a final concentration of 1 µM.
The assay was incubated for 30 °C for 40 min. The reaction
was stopped by adding 50 µL of 50 mM phosphoric acid, and

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4343
(3) Zhou, B.-B.; Elledge, S. J. The DNA Damage Response: Putting
Checkpoints in Perspective. Nature 2000, 408, 433-439.
(4) Melo, J.; Toczyski, D. A unified view of the DNA-damage
checkpoint. Curr. Opin. Cell Biol. 2002, 14, 237-245.
(5) Kastan, M. B.; Onyekwere, O.; Sidransky, D.; Vogelstein, B.;
Craig, R. W. Participation of p53 protein in the cellular response
to DNA damage. Cancer Res. 1991, 51, 6304-6311.
(6) Li, Q.; Zhu, G.-D. Targeting Serine/Threonine Protein Kinase
B/Akt and Cell-cycle Checkpoint Kinases for Treating Cancer.
Curr. Top. Med. Chem. 2002, 2, 939-971.
(7) Greenblatt, M. S.; Bennett, W. P.; Hollstein, M.; Harris, C. C.
Mutations in the p53 tumor suppressor gene: clues to cancer
etiology and molecular pathogenesis. Cancer Res. 1994, 54,
4855-4878.
(8) Wang, Q.; Fan, S.; Eastman, A.; Worland, P. J.; Sausville, E.
A.; et al. UCN-01: a potent abrogator of G2 checkpoint function
in cancer cells with disrupted p53. J. Natl. Cancer Inst. 1996,
88, 956-965.
(9) Bunch, R. T.; Eastman, A. Enhancement of Cisplatin-induced
Cytotoxicity by 7-Hydroxystaurosporine (UCN-01), a New G2-
Checkpoint Inhibitor. Clin. Cancer Res. 1996, 2, 791-797.
(10) Tenzer, A.; Pruschy, M. Potentiation of DNA-Damage-Induced
Cytotoxicity by G2 Checkpoint Abrogators. Curr. Med. Chem.
2003, 3, 35-46.
(11) Shao, R.-G.; Cao, C.-X.; Shimizu, T.; O’Connor, P. M.; Kohn, K.
W.; et al. Abrogation of an S-Phase Checkpoint and Potentiation
of Camptothecin Cytotoxicity by 7-Hydroxystaurosporine (UCN-
01) in Human Cancer Cell Lines, Possibly Influenced by p53
Function. Cancer Res. 1997, 57, 4029-4035.
(12) Jin, P.; Gu, Y.; Morgan, D. O. Role of inhibitory CDC2 phos-
phorylation in radiation-induced G2 arrest in human cells. J.
Cell Biol. 1996, 134, 963-970.
(13) Poon, R. Y. C.; Jiang, W.; Toyoshima, H.; Hunter, T. Cyclin-
dependent kinases are inactivated by a combination of p21 and
Thr-14/Tyr-15 phosphorylation after UV-induced DNA damage.
J. Biol. Chem. 1996, 271, 13283-13291.
90 µL of the stopped assay plate was transferred to a pre-wet
phospho-cellulose filter plate (Millipore MAPHNOB50). The
filter plate was then vacuum filtered, washed three times with
phosphoric acid and once with methanol, and dried at 65 °C.
Scintillation fluid (50 µL) (National Diagnostics LS-273) was
added per well, and the plate was counted on a Perkin-Elmer
Trilux scintillation counter. The limit of detection for this assay
is 5 nM and typical Z values are >0.7.
CDK1 and PKA Assays. Assays with human CDK1 and
PKA were performed in a manner similar to that of the Chk1
assay. For the CDK1 assay, 8 ng CDK1 enzyme were used with
30 µM Histone H1 as a substrate with a final ATP concentra-
tion of 100 µM. The CDK1 assay was incubated at 30 °C for
40 min. For the PKA assay, 10 units of PKA enzyme (Upstate)
were used with 0.75 µM Kemptide as a substrate with a final
ATP concentration of 100 µM. The PKA assay was incubated
at 30 °C for 30 min.
Crystallization and 3D Structure Determination. Chk1
protein was concentrated to 3 mg/mL using ultrafiltration into
a final buffer containing 25 mM Tris pH 7.5, 0.5 M NaCl, and
1 mM DTT. Apo-Chk1 crystals were grown using crystalliza-
tion conditions containing 10-15% PEG8K, 0.1 M Hepes pH
7.5, and 15% (v/v) 2-propanol. To obtain a crystal of Chk1 in
complex with a ligand, this ligand was soaked into an apo
crystal. Single crystals were removed from the apo-protein
crystallization drops and placed in a solution of 4 µL of
crystallization reservoir plus 0.5 µL of ligand stock solution.
The ligand stock solution was 20 mM in 100% DMSO. The
crystals were left to soak at 18°C for a period of approximately
16 h.
The complex data sets were collected at cryotemperature.
The cryoprotectant used was the same as the reservoir solution
from which the crystals grew except 20% glycerol was added.
Data were collected on beamline 19-ID at the APS (Chicago)
or beamline ID29 at the ESRF (Grenoble) and were subse-
quently processed using DENZO,100 in P21 space group with
unit-cell dimensions isomorphous to those of the previously
solved apo-Chk1 structure.42 Hence, the ligand-bound struc-
tures were solved by isomorphous replacement using the apo-
Chk1 model coordinates and the refinement program REF-
MAC5.101 Twenty cycles of rigid-body refinement followed by
twenty cycles of restrained refinement were carried out. All
model building was carried out using the molecular graphics
program O,102 and refinement calculations were performed by
REFMAC5. Difference electron density maps were calculated
for the initial model and inspected following the structure
solution. Crystallographic water molecules were added by
cycling REFMAC5 with ARP/wARP.103 After each round of
refinement, the models were adjusted and further solvent
molecules were gradually added. The progress of the refine-
ment was assessed using Rfree and the conventional R factor.
Once refinement of the structure had converged and the Rfree
(14) Rhind, N.; Furnari, B.; Russell, P. Cdc2 tyrosine phosphorylation
is required for the DNA damage checkpoint in fission yeast.
Genes Dev. 1997, 11, 504-511.
(15) Millar, J. B. A.; Blevitt, J.; Gerace, L.; Sadhu, K.; Featherstone,
C.; et al. p55CDC25 is a nuclear protein required for the initiation
of mitosis in human cells. Proc. Natl. Acad. Sci. U.S.A. 1991,
88, 10500-10504.
(16) Lee, M. S.; Ogg, S.; Xu, M.; Parker, L. L.; Donoghue, D. J.; et al.
cdc25+ encodes a protein phosphatase that dephosphorylates
p34cdc2. Mol. Biol. Cell 1992, 3, 73-84.
(17) Sanchez, Y.; Wong, C.; Thoma, R. S.; Richman, R.; Wu, Z.; et al.
Conservation of the Chk1 Checkpoint Pathway in Mammals:
Linkage of DNA Damage to Cdk Regulation Through Cdc25.
Science 1977, 277, 1497-1501.
(18) Peng, C.-Y.; P. R., G.; Thoma, R. S.; Wu, Z.; Shaw, A. S.; et al.
Mitotic and G2 Checkpoint Control: Regulation of 14-3-3
Protein Binding by Phosphorylation of Cdc25C on Serine-216.
Science 1977, 5, 1501-1505.
(19) Chen, P.; Gatei, M.; O’Connell, M. J.; Khanna, K. K.; Bugg, S.
J.; et al. Chk1 complements the G2/M checkpoint defect and
radiosensitivity of ataxia-telangiectasia cells. Oncogene 1999, 18,
249-256.
(20) Luo, Y.; Rockow-Magone, S. K.; Kroeger, P. E.; Frost, L.; Chen,
Z.; et al. Blocking CHK1 Expression Induces Apoptosis and
Abrogates the G2 Checkpoint Mechanism. Neoplasia 2001, 3,
411-419.
(21) Takai, H.; Tominaga, K.; Motoyama, N.; Minamishima, Y. A.;
Nagahama, H.; et al. Aberrant cell cycle checkpoint function and
early embryonic death in Chk1 -/- mice. Genes Dev. 2000, 14,
1439-1447.
(22) Furnari, B.; Rhind, N.; Russell, P. Cdc25 Mitotic Inducer
Targeted by Chk1 DNA Damage Checkpoint kinase. Science
1997, 277, 1495-1497.
(23) Liu, Q.; Guntuku, S.; Cui, X. S.; Matsuoka, S.; Cortez, D.; et al.
Chk1 is an essential kinase that is regulated by Atr and required
for the G2/M DNA damage checkpoint. Genes Dev. 2000, 14,
2448-2459.
(24) Zhao, H.; Piwnica-Worms, H. ATR-mediated checkpoint path-
ways regulate phosphorylation and activation of human Chk1.
Mol. Biol. Cell 2001, 21, 4129-4139.
(25) Lopez-Girona, A.; Tanaka, K.; Chen, X.-B.; Baber, B. A.; McGow-
an, C. H.; et al. Serine-345 is required for Rad3-dependent
phosphorylation and function of checkpoint kinase Chk1 in
fission yeast. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 11289-
11294.
and R factor could not be lowered further, the structures were
104
validated using PROCHECK
and various programs from
the CCP4i package.105 Full data collection and refinement
statistics are presented in Table 3.
Acknowledgment. We thank the support staff of
beamlines ID29 at the European Synchrotron Radiation
Source, Grenoble, France and 19-ID at the Advanced
Photon Source, Chicago, IL, USA for help with data
collection. We also thank A. Hold for analytical chem-
istry support, and Pr. R. Hubbard for helpful comments
on the manuscript.
Supporting Information Available: Elemental analysis
for key target compounds. This material is available free of
charge via the Internet at http://pubs.acs.org.
(26) Guo, Z. J.; Kumagai, A.; Wang, S. X.; Dunphy, W. G. Require-
ment for Atr in phosphorylation of Chk1 and cell cycle regulation
in response to DNA replication blocks and UV-damaged DNA
in Xenopus egg extracts. Genes Dev. 2000, 14, 2745-2756.
(27) Capasso, H.; Palermo, C.; Wan, S.; Rao, H.; John, U. P.; et al.
Phosphorylation activates Chk1 and is required for Checkpoint-
mediated cell cycle arrest. J. Cell Sci. 2002, 115, 4555-4564.
References
(1) Hanahan, D.; Weinberg, R. A. The Hallmarks of Cancer. Cell
2000, 100, 57-70.
(2) Nurse, P. Checkpoint Pathways Come of Age. Cell 1997, 91,
865-867.

4344 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13
Foloppe et al.
(28) Rhind, N.; Russell, P. Chk1 and Cds1: linchpins of the DNA
damage and replication checkpoint pathways. J. Cell Sci. 2000,
113, 3889-3896.
(29) Chen, Z.; Xiao, Z.; Chen, J.; Ng, S.-C.; Sowin, T.; et al. Human
Chk1 Expression Is Dispensable for Somatic Cell Death and
Critical for Sustaining G2 DNA Damage Checkpoint. Mol. Cancer
Ther. 2003, 2, 543-548.
(30) Zachos, G.; Rainey, M. D.; Gillespie, D. A. F. Chk1-deficient
tumour cells are viable but exhibit multiple checkpoint and
survival defects. EMBO J. 2003, 22, 713-723.
(31) Graves, P. R.; Yu, L.; Schwarz, J. K.; Gales, J.; Sausville, E. A.;
et al. The Chk1 Protein Kinase and the Cdc25C Regulatory
Pathways are Targets of the Anticancer Agent UCN-01. J. Biol.
Chem. 2000, 275, 5600-5605.
(32) Busby, E. C.; Leistritz, D. F.; Abraham, R. T.; Karnitz, L. M.;
Sarkaria, J. N. The Radiosensitizing Agent 7-Hydroxystauro-
sporine (UCN-01) Inhibits the DNA Damage Checkpoint Kinase
hChk1. Cancer Res. 2000, 2000, 2108-2112.
(33) Jackson, J. R.; Gilmartin, A.; Imburgia, C.; Winkler, J. D.;
Marshall, L. A.; et al. An Indolocarbazole Inhibitor of Human
Checkpoint Kinase (Chk1) Abrogates Cell Cycle Arrest Caused
by DNA Damage. Cancer Res. 2000, 60, 566-572.
(34) Eastman, A.; Kohn, E. A.; Brown, M. K.; Rathman, J.; Living-
stone, M.; et al. A Novel Indolocarbazole, ICP-1, Abrogates DNA
Damage-induced Cell Cycle Arrest and Enhances Cytotoxicity:
Similarities and Differences to the Cell Cycle Checkpoint Ab-
rogator UCN-01. Mol. Cancer Ther. 2002, 1, 1067-1078.
(35) Roberge, M.; Berlinck, R. G. S.; Xu, L.; Anderson, H. J.; Lim, L.
Y.; et al. High-throughput assay for G2 checkpoint inhibitors
and identification of the structurally novel compound isogranu-
latimide. Cancer Res. 1998, 58, 5701-5706.
(36) Kohn, E. A.; Yoo, C. J.; Eastman, A. The Protein Kinase C
Inhibitor Go¨6976 Is a Potent Inhibitor of DNA Damage-induced
S and G2 Cell Cycle Checkpoints. Cancer Res. 2003, 63, 31-35.
(37) Curman, D.; Cinel, B.; Williams, D. E.; Rundle, N.; Block, W.
D.; et al. Inhibition of the G2 DNA Damage Checkpoint and of
Protein Kinases Chk1 and Chk2 by the Marine Sponge Alkaloid
Debromohymenialdisine. J. Biol. Chem. 2001, 21, 17914-17919.
(38) Sausville, E. A.; Arbuck, S. G.; Messmann, R.; Headlee, D.; Lush,
R. D.; et al. Phase I Trial of 72-Hour Continuous Infusion UCN-
01 in Patients with Refractory Neoplasms. J. Clin. Oncol. 2001,
19, 2319-2333.
(54) Levitt, M.; Perutz, M. F. Aromatic Rings Act as Hydrogen Bond
Acceptors. J. Mol. Biol. 1988, 201, 751-754.
(55) Meyer, E. A.; Castellano, R. K.; Diederich, F. Interactions with
Aromatic Rings in Chemical and Biological Recognition. Angew.
Chem. 2003, 42, 1210-1250.
(56) Steiner, T.; Koellner, G. Hydrogen Bonds with pi-Acceptors in
Proteins: Frequencies and Role in Stabilizing Local 3D Struc-
ture. J. Mol. Biol. 2001, 305, 535-557.
(57) Bo¨hm, H. J.; Stahl, M. The use of Scoring Functions in Drug
Discovery Applications. Reviews in Computational Chemistry;
Wiley-VCH: New York, 2002; pp 41-87.
(58) Thomas, K. A.; Smith, G. M.; Thomas, T. B.; Feldmann, R. J.
Electronic distributions within protein phenylalanine aromatic
rings are reflected by the three-dimensional oxygen atom
environments. Proc. Natl. Acad. Sci. U.S.A. 1982, 79, 4843-
4847.
(59) Kollman, P. A.; Massova, I.; Reyes, C.; Kuhn, B.; Huo, S.; et al.
Calculating Structures and Free Energies of Complex Mol-
ecules: Combining Molecular Mechanics and Continuum Mod-
els. Acc. Chem. Res. 2000, 33, 889-897.
(60) Hu¨nenberger, P. H.; Helms, V.; Narayana, N.; Taylor, S. S.;
McCammon, J. A. Determinants of Ligand Binding to cAMP-
Dependent Protein Kinase. Biochemistry 1999, 38, 2358-2366.
(61) Sirockin, F.; Sich, C.; Improta, S.; Schaefer, M.; Saudek, V.; et
al. Structure Activity Relationship by NMR and by Computer:
A Comparative Study. J. Am. Chem. Soc. 2002, 124, 11073-
11084.
(62) Halgren, T. Merck Molecular Force Field. I. Basis, Form, Scope,
Parametrization, and Performance of MMFF94. J. Comput.
Chem. 1996, 17, 490-519.
(63) Perola, E.; Charifson, P. S. Conformational Analysis of Drug-
Like Molecules Bound to Proteins: An Extensive Study of
Ligand Reorganization upon Binding. J. Med. Chem. 2004, 47,
2499-2510.
(64) Dunitz, J. D. The Entropic Cost of Bound Water in Crystals and
Biomolecules. Science 1994, 264, 670-670.
(65) Dill, K. A. Dominant forces in protein folding. Biochemistry 1990,
29, 7133-7155.
(66) Fersht, A.; Shi, J. P.; Knill-Jones, J. W.; Lowe, D. M.; Wilkinson,
A. J.; et al. Hydrgen bonding and biological specificity analysed
by protein engineering. Nature 1985, 314, 235-238.
(67) Bartlett, P. A.; Marlowe, C. K. Evaluation of Intrinsic Binding
Energy from a Hydrogen Bonding Group in an Enzyme Inhibitor.
Science 1987, 235, 569-571.
(68) Tronrud, D. E.; Holden, H. M.; Matthews, B. W. Structure of
Two Thermolysin-Inhibitor Complexes That differ by a Single
Hydrogen Bond. Science 1987, 235, 571-574.
(69) Fersht, A. The hydrogen bond in molecular recognition. Trends
Biochem. Sci. 1987, 12, 301-304.
(70) Fersht, A. Relationships between Apparent Binding Energies
Measured in Site-Directed Mutagenesis Experiments and En-
ergetics of Binding and Catalysis. Biochemistry 1988, 27, 1577-
1580.
(71) Williams, D. H.; Searle, M. S.; Mackay, J. P.; Gerhard, U.;
Maplestone, R. A. Toward an estimation of binding constants
in aqueous solution: Studies of associations of vancomycin group
antibiotics. Proc. Natl. Acad. Sci. U.S.A. 1993, 90, 1172-1178.
(72) Connelly, P. R.; Aldape, R. A.; Bruzzese, F. J.; Chambers, S. P.;
Fitzgibbon, M. J.; et al. Enthalpy of hydrogen bond formation
in a protein-ligand binding reaction. Proc. Natl. Acad. Sci.
U.S.A. 1994, 91, 1964-1968.
(73) Abraham, M. H.; Duce, P. P.; Prior, D. V.; Barrat, D. G.; Morris,
J. J.; et al. Hydrogen Bonding. Part 9. Solute Proton Donor and
Proton Acceptor Scales for Use in Drug Design. J. Chem. Soc.,
Perkin Trans. 2 1989, 1355-1375.
(74) Tame, J. R. H. Scoring functions: A view from the bench. J.
Comput.-Aided Mol. Des. 1999, 13, 99-108.
(75) Davis, A. M.; Teague, S. J. Hydrogen Bonding, Hydrophobic
Interactions, and Failure of the Rigid Receptor Hypothesis.
Angew. Chem., Int. Ed. 1999, 38, 736-749.
(76) Kubinyi. Hydrogen Bonding, the Last Mystery in Drug Design?
Pharmacokinetic Optimization in Drug Research. Biological,
Physicochemical, and Computational Strategies; Helvetica Chim-
ica Acta and Wiley-VCH: Zu¨rich, Switzerland, 2001; pp 513-
524.
(39) Senderowicz, A. M. The Cell Cycle as a Target for Cancer
Therapy: Basic and Clinical Findings with the Small Molecule
Inhibitors Flavopiridol and UCN-01. Oncologist 2002, 7, 12-19.
(40) Lyne, P. D.; Kenny, P. W.; Cosgrove, D. A.; Deng, C.; Zabludoff,
S.; et al. Identification of Compounds with Nanomolar Binding
Affinity for Checkpoint Kinase-1 Using Knowledge-Based Vir-
tual Screening. J. Med. Chem. 2004, 47, 1962-1968.
(41) Kania, R. S.; Bender, S. L.; Borchardt, A.; Braganza, J. F.;
Cripps, S. J.; et al. Indazole compounds and pharmaceutical
compositions for inhibiting protein kinases and methods for their
use. Patent WO 0102369, 2001.
(42) Chen, P.; Luo, C.; Deng, Y.; Ryan, K.; Register, J.; et al. The 1.7
Å Crystal Structure of Human Cell Cycle Checkpoint Kinase
Chk1: Implications for Chk1 Regulation. Cell 2000, 100, 681-
692.
(43) Zhao, B.; Bower, M. J.; McDevitt, P. J.; Zhao, H.; Davis, S. T.;
et al. Structural Basis for Chk1 Inhibition by UCN-01. J. Biol.
Chem. 2002, 277, 46609-44615.
(44) http://www.chemdiv.com.
(45) Baurin, N.; Baker, R.; Richardson, C.; Chen, I.; Foloppe, N.; et
al. Drug-like annotation and duplicate analysis of a 23-supplier
chemical database totalling 2.7M compounds. J. Chem. Inf.
Comput. Sci. 2004, 44, 643-651.
(46) http://mdli.com.
(47) Johannsen, F.; Jorgensen, A.; Pedersen, E. B. Reactions of
Heterocyclic o-Aminonitriles with Acetic Formic Anhydride.
Chim. Scripta 1986, 26, 347-351.
(48) Toledo, L. M.; Lydon, N. B.; Elbaum, D. The Structure-Based
Design of ATP-Site Directed Protein Kinase Inhibitors. Curr.
Med. Chem. 1999, 6, 775-805.
(49) Noble, M. E.; Endicott, J. A.; Johnson, L. N. Protein Kinase
Inhibitors: Insights into Drug Design from Structure. Science
2004, 303, 1800-1805.
(50) Williams, D. H.; Mitchell, T. Latest developments in crystal-
lography and structure-based design of protein kinase inhibitors
as drug candidates. Curr. Opin. Pharmacol. 2002, 2, 567-573.
(51) Steiner, T. Donor and acceptor strengths in C-H‚‚‚O hydrogen
bonds quantified from crystallographic data of small solvent
molecules. New J. Chem. 1998, 1099-1103.
(52) Pierce, A. C.; Sandretto, K. L.; Bemis, G. W. Kinase Inhibitors
and the Case for CH‚‚‚O Hydrogen Bonds in Protein-Ligand
Binding. Proteins 2002, 49, 567-576.
(77) Williams, D. H.; Westwell, M. S. Aspects of weak interactions.
Chem. Soc. Rev. 1998, 27, 57-63.
(78) Sippl, M. J. Helmholtz Free Energy of Peptide Hydrogen Bonds
in Proteins. J. Mol. Biol. 1996, 260, 644-648.
(79) Gidofalvi, G.; Wong, C. F.; McCammon, J. A. Entropy Loss of
Hydroxyl Groups of Balanol upon Binding to Protein Kinase A.
J. Chem. Educ. 2002, 79, 1122-1126.
(80) Bo¨hm, H. J.; Brode, S.; Hesse, U.; Klebe, G. Oxygen and Nitrogen
in Competitive Situations: Which Is The Hydrogen-Bond Ac-
ceptor? Chem. -Eur. J. 1996, 2, 1509-1513.
(53) Desiraju, G. R.; T., S. The Weak Hydrogen Bond In Structural
Chemistry and Biology; Oxford Science Publications: New York,
1999.
(81) Honig, B.; Sharp, K.; Yang, A.-S. Macroscopic Models of Aqueous
Solutions: Biological and Chemical Applications. J. Phys. Chem.
1993, 97, 1101-1109.

Structure-Based Design of Chk1 Inhibitors
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 13 4345
(82) Manning, G.; Whyte, D. B.; Martinez, R.; Hunter, T.; Sudar-
sanam, S. The Protein Kinase Complement of the Human
Genome. Science 2002, 298, 1912-1934.
(83) Furet, P.; Meyer, T.; Strauss, A.; Raccuglia, S.; Rondeau, J.-M.
Structure-Based Design and Protein X-ray Analysis of a Protein
Kinase Inhibitor. Bioorg. Med. Chem. Lett. 2002, 12, 221-224.
(84) Morley, S. D.; Afshar, M. Validation of an empirical RNA-ligand
scoring function for fast flexible docking using RiboDock. J.
Comput.-Aided Mol. Des. 2004, 18, 189-208.
(95) Momany, F. A.; Rone, R. Validation of the General Purpose
QUANTA 3.2/CHARMm Force Field. J. Comput. Chem. 1992,
13, 888-900.
(96) Im, W.; Beglov, D.; Roux, B. Continuum Solvation Model:
computation of electrostatic forces from numerical solutions to
the Poisson-Boltzmann equation. Comput. Phys. Commun.
1998, 111, 59-75.
(97) Nina, M.; Beglov, D.; Roux, B. Atomic radii for continuum
electrostatics based on molecular free energy simulations. J.
Phys. Chem. B 1997, 101, 5239-5248.
(85) Barril, X.; Hubbard, R. E.; Morley, D. Virtual Screening in
Structure-Based Drug Discovery. Mini-Rev. Med. Chem. 2004,
4, 779-791.
(98) Lee, B.; Richards, F. M. The interpretation of protein struc-
tures: Estimation of static accessibility. J. Mol. Biol. 1971, 55,
379-400.
(86) Morley, D. Unpublished work.
(99) Delano, W. L. The PyMOL User’s Manual; DeLano Scientific:
San Carlos, CA, 2002.
(87) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T.
N.; et al. The Protein Data Bank. Nucleic Acids Res 2000, 28,
235-242.
(100) Otwinowski, Z.; Minor, W. Processing of X-ray Diffraction Data
Collected in Oscillation Mode. Methods Enzymol. 1997, 276,
307-326.
(88) MacKerell, A. D., Jr.; Bashford, D.; Bellott, M.; Dunbrack, R.
L., Jr.; Evanseck, J. D.; et al. All-Atom Empirical Potential for
Molecular Modeling and Dynamics Studies of Proteins. J. Phys.
Chem. B 1998, 102, 3586-3616.
(101) Murshudov, G. N.; Vagin, A. A.; Dodson, E. J. Refinement of
Macromolecular Structures by the Maximum-Likelihood Method.
Acta Crystallogr. 1997, D53, 240-255.
(89) http://www.chemcomp.com.
(102) Jones, T. A.; Kjelgaard, M. O, Version 6.1.2, the manual;
Uppsala, Sweden, 1993.
(90) Halgren, T. MMFF VI. MMFF94s Option for Energy Minimiza-
tion Studies. J. Comput. Chem. 1999, 20, 720-729.
(91) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
Swaminathan, S.; et al. CHARMM: A Program for Macromo-
lecular Energy, Minimization and Dynamics calculations. J.
Comput. Chem. 1983, 4, 187-217.
(103) Perrakis, A.; Morris, R. J. H.; Lamzin, V. S. Automated protein
model building combined with iterative structure refinement.
Nature Struct. Biol. 1999, 6, 458-463.
(104) Laskowwski, R. A.; MacArthur, M. W.; Moss, D. S.; Thornton,
J. M. PROCHECK - a program to check the stereochemical qual-
ity of protein structures. J. Appl. Crystallogr. 1993, 26, 283-291.
(105) Collaborative Computational Project, N. The CCP4 Suite: Pro-
grams for Protein Crystallography. Acta Crystallogr. 1994, D50,
760-763.
(92) http://www.accelrys.com.
(93) Spackman, M. A. Potential Derived Charges Using a Geodesic
Point Selection Scheme. J. Comput. Chem. 16, 17, 1-18.
(94) Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.;
Gordon, M. S.; et al. General Atomic and Molecular Electronic
Structure System. J. Comput. Chem. 13, 14, 1347-1363.
JM049022C