Synthesis, biological testing, and binding mode prediction of 6,9-diarylpurin-8-ones as p38 MAP kinase inhibitors

Article abstract of DOI:10.1021/jm061061w

Based on the purine scaffold of ATP, derivatives of 6,9-diarylpurine-8-one were prepared and tested for their ability to inhibit p38 MAP kinase, a key enzyme in the cellular regulation of proinflammatory cytokines. The inhibitor design combines the purine

Full text of DOI:10.1021/jm061061w

2060
J. Med. Chem. 2007, 50, 2060-2066
Synthesis, Biological Testing, and Binding Mode Prediction of 6,9-Diarylpurin-8-ones as
p38 MAP Kinase Inhibitors
Dominik R. J. Hauser,† Thomas Scior,§ David M. Domeyer,† Bernd Kammerer,‡ and Stefan A. Laufer*,†
Department of Pharmaceutical and Medicinal Chemistry, Eberhard-Karls-UniVersity Tu¨bingen, Auf der Morgenstelle 8,
72076 Tu¨bingen, Germany, Department of Clinical Pharmacology, UniVersity Hospital, Eberhard-Karls-UniVersity Tu¨bingen,
Otfried-Mu¨ller-Strasse 45, 72076 Tu¨bingen, Germany, and Department of Pharmacy, Beneme´rita UniVersidad Auto´noma de Puebla,
14 Sur con AVienda San Claudio, C.P. 72570 Puebla, Me´xico
ReceiVed September 7, 2006
Based on the purine scaffold of ATP, derivatives of 6,9-diarylpurine-8-one were prepared and tested for
their ability to inhibit p38 MAP kinase, a key enzyme in the cellular regulation of proinflammatory cytokines.
The inhibitor design combines the purine system of the authentic cosubstrate ATP with various phenyl
moieties to explore the selectivity for the two hydrophobic regions of the kinase’s ATP-binding cleft. The
present study indicates a new binding mode of our scaffold to p38 MAP kinase, which comprises the desired
structural features of ATP and the N-phenyl-N-purin-6-yl ureas previously published by Wan et al.
Combinations of Autodock and FlexX docking with different scoring functions were used to assess the
postulated binding mode. The predictive power of different docking-scoring combinations was determined.
The presented results may form a solid basis for further optimization cycles since our theoretical findings
are consistent with our experimental binding data and supported by the literature.
Introduction
The p38 MAPa kinase is closely associated with the cytokine-
driven progression of rheumatoid arthritis and related inflam-
matory diseases. This enzyme tightly controls the expression
of proinflammatory cytokines like IL-1â and TNFR at the levels
of transcription and translation.1 In human immune cells the
p38 MAP kinase signaling pathway is activated in a self-
inducing manner by cytokines.2 Cellular activation of the
pathway occurs in Vitro in response to environmental stresses
such as ultraviolet light, heat, osmotic shock, and inflammatory
cytokines.1 Due to its prominent role in regulating the production
of inflammatory cytokines, the inhibition of p38 MAP kinase
is an attractive approach for the treatment of cytokine-mediated
diseases.3
Most of the p38 inhibitors published to date rely structurally
on a vicinal diaryl system which provides selectivity.4,5 As
reported for many kinase inhibitors,6 such molecules bind to
p38 MAP kinase competitively with ATP. Molecular modeling
allows the reproduction of the crystallographic binding modes
of the natural ligand7 and of the lead compounds SB 2035808
or ML 3163,5 which are typical inhibitors of the vicinal diaryl
class (Figures 1 and 2). The first aryl, a 4-pyridyl moiety, mimics
a conserved hydrogen bond, originally formed between N-1 of
ATP and the backbone amide N-H of Met 109 (Figure 1).5
The second aryl, a 4-fluorophenyl moiety occupies the kinase’s
hydrophobic region I which is pocket-shaped (Figure 2).
Intriguingly, ATP does not occupy this hydrophobic region I
(selectivity pocket) when bound to the kinase (Figure 1), thus
providing opportunities for the design of selective inhibitors with
an appropriate substitution pattern targeting region I and, in order
Figure 1. Mapping important regions within the ATP-binding-site of
kinases based on both X-ray structures and computational methods
(modified from Traxler9). The bidentate hydrogen-bond donor/acceptor
system with the backbone of the hinge region anchors the ATP-adenine
deep inside the binding cleft.
to increase affinity, region II.9 In a recent publication we
presented compounds combining the purine system from the
original cosubstrate ATP and phenyl moieties in order to explore
possible interactions with the different regions of the ATP
binding site in several disease-related protein kinases.10 In this
paper we present the synthesis, biological testing, and compound
design. In addition, we postulate and assess computationally
the binding mode for our inhibitors of p38 MAP kinase based
on a 6,9-diarylpurine-8-one scaffold.
Scaffold Design for Targeting p38 MAP Kinase. Recently
Wan et al.11 published a series of N-phenyl-N-purin-6-yl ureas
suggesting the design of directly derived adenine mimics. The
backbone amide N-H of Met109 forms a conserved hydrogen
bond even with structurally diverse ligands, e.g., (i) N-1 of ATP-
adenine (Figure 1), (ii) the pyridine moiety of either SB 203580
* Author to whom correspondence should be addressed. Telephone:
++49 7071 2972459, fax: ++49 7071 295037, e-mail: stefan.laufer@
uni-tuebingen.de.
† Eberhard-Karls-University Tu¨bingen.
‡ University Hospital Eberhard-Karls-University Tu¨bingen.
§ Beneme´rita Universidad Auto´noma de Puebla.
a Abbreviations: p38 MAP kinase, p38 mitogen activated protein kinase;
ATP, adenosine triphosphate; IL, interleukine; TNF, tumor necrose factor;
FEB, free energy of binding; PI, predictive index.
10.1021/jm061061w CCC: $37.00 © 2007 American Chemical Society
Published on Web 04/06/2007

6,9-Diarylpurin-8-ones as p38 MAP Kinase Inhibitors
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9 2061
Figure 4. Chemical structures of PP1 and PP2.
Figure 2. Schematic drawing of specific interactions between a
pyridinylimidazole-based inhibitor (ML 3163) and the ATP binding
site of p38 MAP kinase.5 These are hydrogen bonds bridging the
pyridine N-atom to backbone NH atoms of Met109 and the imidazole
to the side chain of Lys53. The shaded area marks the interaction with
hydrophobic region I with the role of a selectivity pocket for one aryl
substituent. The ligand is docked in homology to SB203580 X-ray
structure 1A9U.8
Figure 5. Schematic representation of the computed binding mode
for our novel p38 MAP kinase inhibitors based on the 6,9-diarylpurin-
8-one template. Important interactions between 6,9-diarylpurin-8-one-
based inhibitors and the ATP binding site of p38 MAP kinase are
displayed. The bidentate hydrogen bond donor/acceptor system with
backbone atoms of the hinge region is achieved with these compounds
as well.
(a member of the Src family), PP2 is nearly oriented as the
adenine of bound ATP, and the chloroaromatic enters a
lipophilic pocket (region I) opening behind N7.14 PP1 and PP2
are strong inhibitors of kinases belonging to the Src family but
were found to inhibit p38 MAP kinase as well.15,16
In the present study, our goal was to verify experimentally
and computationally whether it would be possible to retain the
aforementioned key features in aryl-substituted purine deriva-
tives to create p38 MAP kinase inhibitors closely related to ATP.
The in silico design assisted in scaffold optimization, binding
mode prediction, and defining an appropriate substitution pattern
to meet the geometrical features for the most advantageous
interactions with the target protein.
We have therefore developed a series of 6,9-diarylpurin-8-
ones consistent with the aforesaid requirements (Figures 5, 6).
Compared to the compounds of Figure 3, the purine moiety is
rotated clockwise about 120°, and a carbonyl function has been
added at position 8, which brings about the hydrogen bond at
N-7 (Figure 5). Consequently, the purine moiety is oriented back
again to the adenine binding region, even though it assumes a
horizontally flipped position compared to its orientation in
kinase-bound ATP (see Figure 1). The aryl moieties are directly
attached to positions C-6 and N-9 to extend to both hydrophobic
regions I and II, respectively. The ortho mono- or disubstitution
enforces a propeller-like conformation11 of the aryl moieties for
optimized interaction with either hydrophobic region.
We postulate that the essential interactions such as binding
of the aryl moieties in hydrophobic regions I and II and the
bidentate hydrogen bond donor/acceptor system with the
backbone of the hinge region are conserved with our compound
design.
Figure 3. Schematic drawing of important interactions between urea-
based inhibitors described by Wan11 et al. and the ATP binding site of
p38 MAP kinase, derived from X-ray crystallography.
or ML 3163 (Figure 2), and (iii) the urea oxygen of Wan’s
purine compounds (Figure 3). Another hydrogen bond between
the amide carbonyl of His107 and the urea N-H on the ligand
side is not generally conserved (Figure 3). Both hydrogen bonds
mimic the bidentate hydrogen bond donor/acceptor system in
the hinge region that anchors the adenine scaffold of ATP deeply
inside the binding cleft. Like the 4-fluorophenyl moiety of ML
3163 (Figure 2), one phenyl ring in Wan’s series occupies the
hydrophobic region I (selectivity pocket) whereas another
aromatic ring resides in hydrophobic region II; neither the
pyridinyl-imidazole inhibitors nor ATP occupy this second
hydrophobic site (Figure 3). Hence, affinity to the kinase
increases. Both aryl moieties and the urea substructure are
attached to a purine scaffold which in turn aligns the substituents
to their appropriate spatial position in the binding cleft. This
purine scaffold, however, does not directly contact the adenine
binding region of the ATP-binding pocket, and the purine moiety
is not part of the essential bidentate H-bonding system.
Therefore, Wan’s compounds cannot be seen as true adenine
mimetics since such ingenious molecules should not only show
the key interactions between an appropriately substituted purine
or adenine and the backbone but also, ideally, bind within the
adenine binding region of the ATP-binding pocket. Good
examples in the literature that this goal can be achieved are the
pyrazolo[3,4-d]pyrimidines PP1 and PP2 (Figure 4).12,13 Their
structure is closely related to adenine, and their N4 position
corresponds to N6 position of ATP. When bound to Lck kinase

2062 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9
Hauser et al.
Table 1. Inhibition Data (IC50, Enzyme Assay, ATP Concentration:
100 µM).
molecule ID number
IC50 (µM)
21
22
0.9
7.8
23
2.7
24
25
26
17.5
18.1
0.5
27
1.6
28
5.2
29
7.8
30
100
0.056
SB203580
antagonistic activity. Therefore, possible side effects of our 6,9-
diarylpurin-8-ones should be minimized.
Suzuki coupling reaction of ortho-substituted benzeneboronic
acids with 4,6-dichloro-5-nitropyrimidine yielded the 6-aryl-4-
chloro-5-nitro-pyrimidines 1-3 (see Scheme 1). Even with
equimolar amounts of the reagents, the diarylated product was
afforded as a minor side product. However, column chroma-
tography easily separated the monoarylated pyrimidines, which
eluted first (methylene chloride/hexane). Subsequent nucleo-
philic substitution of the remaining chlorine afforded the
(6-aryl-5-nitro-pyrimidin-4-yl)-aryl-amines 4-11. Reduction of
the nitro group proceeded rapidly at ambient temperature with
palladium on charcoal under a hydrogen atmosphere. However,
these reduction conditions resulted in hydrodehalogenation in
case of the chlorine-substituted aryls whereas fluorine com-
pounds remained unaffected. The chlorine remained when
platinum sulfide19 on charcoal was used as the catalyst for
hydrogenation, and the reduction of the nitro group proceeded
as quickly as with palladium on charcoal. At this step the
literature procedure18 used an iron/acetic acid system in MeOH
with refluxing. Our more versatile procedure gave nearly
quantitative yields of the diarylpyrimidinediamines 12-20. For
the following ring closure the literature procedure18 employed
phosgene in toluene under reflux conditions. The highly volatile
and toxic phosgene could be easily replaced by N,N-carbonyl-
diimidazole20 which afforded the diarylpurin-8-ones in good
yields and with short reaction times.
Figure 6. Schematic representation of the flipped binding mode for
the 6,9-diarylpurin-8-one template derived from computational docking.
The bidentate hydrogen bond donor/acceptor system with backbone
atoms of the hinge region is not achieved in the flipped orientation.
The formation of one hydrogen bond between the amide N-H of Met
109 and N1 of the purine is possible.
Scheme 1a
NMR Studies. To evaluate the rotational barrier of the ortho
monosubstituted aryl moiety at C-6 of the purin-8-one we have
performed temperature-dependent NMR measurements. As
samples we chose 24 with its freely rotateable phenyl group
and 26 with the ortho methyl-substituted phenyl. Following
Wan’s argument, the aryl at C-6 should bind deeply inside the
lipophilic selectivity pocked. Therefore a propeller-like con-
formation of the C-6 aryl would be favorable. In the NMR
studies we did not see any change in the signal line width upon
the rotational barrier (see spectra in the Supporting Information).
So we conclude a rotational barrier between 7.1 and 10.7 kcal/
mol which is in agreement with literature data.21,22 This would
mean that in solution and at a temperature relevant for biological
systems the aryl would be somewhat rotateable at least over
the N-1 of the purine scaffold, allowing it to rotate to a favored
position which would allow the entry into the binding pocket
in the sense of an induced fit. The ortho substituent should
facilitate this as already stated by Wan et al. This finding is
consistent with the difference of the inhibition values of 24 and
26.
a Reagents: (a) Appropriate boronic acid, K2CO3, toluene, Pd(PPh3)4,
90 °C; (b) appropriate aniline, triethylamine, THF, reflux; (c) PtS/C, H2, 4
bar, EtOH, rt or Pd/C, H2, 4 bar, EtOH, rt; (d) N,N-carbonyldiimidazole or
N,N-thiocarbonyldiimidazole (for 30) toluene/dioxane. b30: Purine-8-thione.
Charge distributions on oxygen and sulfur atoms at position
C-8 show significant difference. A thio analogue of 25
(Scheme 1) is synthesized (30) for probing the loss of hydrogen-
bonding in the activity tests. Isomeric structures are 21/23 and
26/27. Since the purine is substituted with similar aryl moieties
on opposite sides, the compounds show partial bilateral sym-
metry. Consequently we also expected answers about flip-flop17
behavior from our docking studies.
Chemistry. A series of 6,9-diarylpurin-8-ones was synthe-
sized according to an improved procedure described before.18
The structurally related compounds presented by Cocuzza18
et al. were reported as corticotropin-releasing hormone antago-
nists whereas their 8-oxo derivatives exhibited only weak
Biological Testing. The compounds which widely followed
the favorable substitution pattern found by Wan et al. were tested
with a recently published non-radioactive enzyme assay.23 The
test results are shown in Table 2. SB203580 is a widely utilized

6,9-Diarylpurin-8-ones as p38 MAP Kinase Inhibitors
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9 2063
Table 2. Free Energy of Binding (FEB) by Autodock versus
Experimental Affinity Data for Three Sets of Docked Conformations of
the Purin-8-one Derivatives.
FEB of best
solutions
(kcal/mol)
FEB of
orientation 1
(kcal/mol)
FEB of
orientation 2
(kcal/mol)
molecule
ID no.
IC50
(µM)
21
22
23
24
25
26
27
28
29
30
0.9
7.8
2.7
17.5
18.1
0.5
1.6
5.2
7.8
100
-8.7
-9.4
-8.8
-8.2
-9.4
-8.8
-8.9
-9.2
-8.5
-9.4
-8.7
-9.4
-8.8
-8.2
-9.4
-8.7
-8.7
-8.8
-8.5
-9.2
-8.7
-9.2
-8.7
-8.2
-8.0
-8.8
-8.9
-9.2
-6.9
-9.4
Figure 7. Schematic drawing of the 3,4-dihydro-1H-quinolin-2-one-
based inhibitor from X-ray structure 1OVE.24 This pdb entry served as
template for docking studies. (IC50 ) 0.74 nM at 2 µM ATP).
reference substance in p38 MAP kinase research and served as
a standard in our testing. The IC50 of our best compound
was 1 order of magnitude higher than that of SB203580. As
Wan et al. gave no data for a literature known reference
compound, a direct comparison of our compounds with Wan’s
is not possible. We performed molecular modeling calculations
to evaluate if it would be possible to find a proof for our
envisioned binding mode (Figure 5) which followed the binding
mode proposed by Wan (Figure 3).
Molecular Modeling. Until now, the coordinates of Wan’s
crystal structure were not published in the PDB. Therefore we
chose PDB entry 1OVE24 as the rigid receptor model because
of its good resolution and a structural similarity of the diaryl-
substituted 3,4-dihydro-1H-quinolin-2-one with our diaryl-
substituted purin-8-one scaffold. Morever, the carbonyl group
is located in a position that corresponds closely to that found
in the purin-8-ones. The missing side chain atoms of Arg173
were added. Each ligand was built with MOE25 and minimized
using MMFF94 force field26 with standard set up. For the
ligands, Gasteiger charges27 were calculated using Autodock
Tools (MGL Tools 1.3alpha228,29). The receptor model was
prepared with Autodock Tools, deleting all water molecules,
adding polar hydrogens and loading Kollman United Atoms
charges.30 We set the following parameters and options: sample
numbers 256; survival rate 3; grid position and resolution 0.5;
bond rotation 10°; translation step size 1; cluster width 2.5, using
the GALS algorithm31 until the ligand position was uniquely
reproduced among the top hundred conformers with lowest free
energies (-12 kcal/mol). Autodock 3 allows fast docking
(100 conformations in 70 min of CPU time on a standard PC
with AMD Athlon 2.4 GHz CPU running Red Hat Linux 9).
Except for 28 with four rotatable bonds, the series were treatable
with just two rotatable bonds between the scaffold and the
aryl substitutions. FlexX docking was performed using FlexX
v.1.13.5 as implemented in Sybyl 7.0.32 FlexX was ap-
proximately 10 times faster than Autodock 3 and therefore is
better qualified for docking large compound libraries.
Figure 8. Docking results obtained using Autodock 3. All purin-8-
ones bind in a position similar to that of the cocrystallized compound
as reference (magenta) but show two orientations. Orientation 1 (red)
cf. Figure 5 with the oxygen pointing at Met109 and the 180° flipped
orientation 2 (yellow) cf. Figure 6 with the oxygen pointing at Lys53.
Figure 9. Docking of our purin-8-ones (orange) into the ATP binding
site of p38 Map kinase (PDB entry 1ove) using FlexX. Important amino
acids are colored by atom type. The cocrystallized compound as
reference is colored magenta. Figures 8 and 9 were generated with
InsightII.46
Results and Discussion
Autodock 3 and FlexX are docking programs capable of
predicting the binding geometry found in the crystallographic
complex 1OVE24 (RMSD ) 0.9 Å). As stated above 1OVE
was chosen as a reference to verify simulation because it reflects
well the structural pattern of our test series. (Figure 7).
The Autodock 3 program is better suited to explore the
unknown binding mode of our series, as it geometrically places
all ligands in the same cavities occupied by the cocrystallized
ligand of 1OVE (Figure 8). In contrast to the concise picture
Autodock 3 provides, FlexX only yields dispersed poses in the
cleft, due to its fragment-based search algorithm. As a result, it
becomes more difficult to decide which of the many conforma-
tions is of binding relevance (Figure 9).
It is possible to separate the docking results of Autodock 3
into two groups which are referred to as orientation 1 and the
flipped orientation 2: In orientation 1 (Figure 5 and red
molecules in Figure 8) each ligand bears the possibility of
forming a bidentate hydrogen bond donor/acceptor system with
the hinge region of the kinase. A third hydrogen bond may be
formed between N-1 of the purine and Lys53. Orientation 2

2064 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9
Hauser et al.
Table 3. Predictive Indices (PI) of Different Docking-Scoring Combinations. The Higher the PI Value the More Accurate Is the Prediction for
Significant Inhibitor Ranking against Mere Random Distribution (PI ) 0).
Autodock
Autodock
Autodock
scoring
best solution
orientation 1
orientation 2
Chem-score
D score
F score
G score
PMF score
Autodock
Chem score
D score
F score
G score
-0.46
-0.53
-0.54
0.36
-0.55
0.09
-0.27
-0.19
-0.32
0.55
-0.37
0.19
-0.17
-0.32
-0.38
0.73
-0.33
0.08
-0.27
-0.22
-0.07
-0.25
0.24
-0.27
-0.22
-0.07
-0.25
0.24
-0.26
-0.28
0.03
-0.25
0.51
-0.19
-0.14
-0.06
-0.14
0.36
-0.38
0.46
-0.32
-0.04
0.42
PMF score
(Figure 6 and yellow molecules in Figure 8) allows the putative
formation of a single hydrogen bond between the amide N-H
of Met109 and N-1 of the purine. In this orientation Lys53 may
form a hydrogen bond with C-8 carbonyl oxygen. Table 2
presents a comparison between the calculated energies and the
experimentally determined IC50 values. Orientations 1 and 2
show minor differences in terms of free energies of binding
(FEB) as calculated using Autodock 3 (0.04-1.63 kcal/mol).
Since the estimated FEB values vary over the narrow range from
-8.24 to -9.44 kcal/mol, no significant ranking of the different
inhibitors of our series can be performed which could reflect
the ranking of the IC50 values. Furthermore no commitment to
orientation 1 or 2 is possible. To improve the dissatisfying
scoring by Autodock 3, the scoring functions available in the
Sybyl Cscore module were applied to the three subsets of
Autodock 3 conformations shown in columns 3-5 in Table 2.
These scoring functions are (1) The FlexX scoring function,33
(2) The PMF scoring function,34 (3) The Dock scoring
function,35 (4) the Chemscore scoring function,36 and (5) the
GOLD scoring function.37 The predictive index (PI) introduced
by Pearlman38 is used to assess the capability of each scoring
function to predict the ranking of the inhibitors. The results are
shown in the first three columns of Table 3 and discussed below.
For a set of experimental pKi (or pIC50) values, E(i), and
corresponding predicted scores P(i) the predictive index (PI) is
defined as follows:38
molecule using FlexX as the docking engine. Then each of the
scoring functions was used to select the best of the 600
conformations of every molecule. This results in five sets of
conformations of each of the ten molecules. Each of the five
scoring functions was applied to these five sets of docking
results and once more ranked by the other four scoring functions.
Again the PI is used to assess the concordance between the
ranking by each scoring function and the IC50. Table 3 (columns
4-8) shows the PI results of the different combinations of
docking methods and scoring functions. Most of the combina-
tions show results that are random. Only the combination of
the five docking methods with PMF score yields always a
positive PI which means a better prediction. Columns 1-3 show
the PI results for the combination of the three sets of Autodock
conformations with the different scoring functions including the
Autodock scores in row 1. The scoring of orientation 2 obtained
by Autodock 3 alternatively using the FlexX scoring function
attains the highest predictive index (PI ) 0.73). Applying the
FlexX scoring function to orientation 1 results in an inhibitor
ranking which is still significantly better than random distribu-
tion (PI ) 0.55). For the lowest energy Autodock 3 conforma-
tions, the highest PI is again obtained by using the FlexX scoring
function (PI ) 0.36). The other docking/scoring combinations
do not generate better results.
We examined the H-bonding properties of diarylpurin-8-ones
and diarylpurine-8-thiones. The calculated difference between
the Autodock 3 scores of 25 and its thio analogue 30 does not
correspond to the observed difference in their biological
activities. Compared to oxygen (EN ) 3.5), sulfur is less
electronegative (EN ) 2.4), and hence a relatively more positive
charge can be expected. To evaluate the electrostatic influence
on the score ranking, we tried several thio models by modifying
the partial charges of sulfur (e.g.: Gasteiger Marsili: -0.05 vs
MMFF94: -0.40) and its adjacent carbon atom (Gasteiger
Marsili: +0.17 vs MMFF94: +0.50) adjusting charges for
electroneutrality.26,27 Since no significant improvement of
sensitivity could be observed, we conducted AM1-based
semiempirical MOPAC39 calculations with full geometry opti-
mization to increase the precision level of the electronic
description of ”oxo” and “thio” groups as potential hydrogen-
bond acceptors. The rationale behind the decision to select the
AM1 Hamiltonian among current quantum-chemical methods
is supported by the literature.39-43 The computed charges
revealed that there is significantly more negative net atomic
partial charge on the oxygen atom (-0.32) of 25 than on the
equilocated sulfur atom (-0.25) of 30. This means for oxygen
and sulfur, respectively, any increase in the net atomic partial
charge decreases of the potential to build up hydrogen bonds
for homologous O or S dotted series. Thio analogues are less
potent H-bond acceptors and form a weaker H-bond than that
of the 6,9-diarylpurin-8-ones, which clearly correlates with the
experimentally observed loss in binding affinity (Table 2). This
is consistent with the data published by Masunov and Dannen-
PI )
w C /
w
(1)
(2)
∑∑ ij ∑∑
ij
ij
j>i
i
j>i i
with
and
wij ) |E(j) - E(i)|
[E(j) - E(i)]/[P(j) - P(i)] < 0
[E(j) - E(i)]/[P(j) - P(i)] > 0
if
[P(j) - P(i)] ) 0
1
if
Cij ) -1 if
(3)
{
0
P(i) is the model score assigned to molecule i, and E(i) is
the experimental pIC50. This index ranges from -1 to +1,
depending on how well the predictions track the order of
experiment: +1 arises from perfect prediction, -1 arises from
predictions that are always wrong, and 0 arises from predictions
that are completely random. This function includes a weighting
term that depends on the difference between the experimental
values which in turn reflects the fact that a good function should
be able to differentiate between changes that result in large
differences in binding.
As an additional approach to compare the performance of
the five scoring functions mentioned above, we used each of
them to determine the docked position of each molecule. For
this purpose 600 docked conformations were generated of each

6,9-Diarylpurin-8-ones as p38 MAP Kinase Inhibitors
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9 2065
berg for urea (-30 to -40 kJ/mol) and thiourea (-20 to
-30 kJ/mol) dimerization energies.44
Conclusion
The binding mode of inhibitors in the ATP binding site of
p38 MAP kinase is determined by both hydrogen bonds and
lipophilic interactions. Computational methods to predict the
ranking of biological activities of different inhibitors should
balance both types of interaction. Autodock 331 performs well
in the case of p38 MAP kinase giving similar docking results
for the structurally similar inhibitors of our purin-8-one series.
In good agreement with the reference complex, 1OVE, the
hydrophobic aryl moieties fit tightly into the hydrophobic
pockets. On one hand, Autodock predicts favorable geometries
but on the other hand the binding energies calculated by
Autodock 3 reflect only partially the ranking of all inhibitors.
In contrast the docking program FlexX does not predict a
consistent binding mode but scatters all molecules over the
binding site due to its fragment-based cavity exploration and
its overemphasis on hydrogen bond contacts during the docking
process.45 Nonetheless, the FlexX scoring function is most
valuable to predict the ranking of biological activities among
inhibitors. The predictive index of 0.73 achieved by the
combination of Autodock docking geometries with FlexX
scoring is close to perfect prediction of the inhibitor ranking.
In analogy to Wan et al.11 we designed a diarylpurin-8-one
scaffold with a very similar binding mode which includes two
possible orientations of the central purin-8-one scaffold. The
first of the orientations allows the formation of a bidentate
H-bond network with the conserved hydrogen bonds to Met109
and adjacent His107 in complete pharmacophoric equivalency
to the known ATP binding while orientation 2 allows only a
single conserved hydrogen bond to Met109. Moreover the
binding feature of the two aryl moieties is in good agreement
with the structural requirements described in the literature which
are represented by the cocrystallized molecule of the template
1OVE. With both orientations of the purin-8-one scaffold, the
aryl moieties are able to address not only the hydrophobic region
I as already observed in complexed SB 2035808 but additionally
the hydrophobic region II. The ortho substituents of the aromatic
moieties may force them into a twisted position relative to the
plane of the purine scaffold, resulting in the earlier described
propeller-like conformations.11 This allows an optimal access
of the purine C-6 aryl group to hydrophobic region I (pocket)
and improved interaction between the second aryl (N-9) and
hydrophobic region II which is formed by the surfaces above
and below the ATP-adenine at the point where the binding cleft
opens to the solvent. The methyl group of the most active
compound 26 is optimally buried by the lipophilic side chains
of Val38, Ala51, and Lys53 (Figure 10, showing 26 in
orientation 1).
Figure 10. The most active compound 26 docked into the ATP binding
site of p38 MAP kinase. The solvent-accessible surface47 is calculated
using the Molcad module of Sybyl 7.032 and colored by lipophilicity48
(brown ) lipophilic, blue ) hydrophilic). The o-methyl group is in
optimal interaction with the hydrophobic region formed by the side
chains of Val30 and Val38. Important hydrogen bonds with the
backbone of Met109 and His107 are marked by yellow lines.
drug development cycles regarding the competitive inhibition
of ATP-consuming enzymes.
Experimental Section
Chemistry. All reagents and solvents were of commercial quality
and used without further purification. General procedures for the
preparation of compounds A-D (find procedures and spectroscopic
data for the particular compounds in Supporting Information):
General Procedure A for the Preparation of 4-Chloro-5-nitro-
6-phenyl-pyrimidines (1-3). A mixture of 4,6-dichloro-5-nitro-
pyrimidine (1 equiv), the appropriate benzeneboronic acid (1.1
equiv), K2CO3 (1.25 equiv), Pd(PPh3)4 (0.01-0.02 equiv), and a
sufficient amount of toluene was stirred under argon, protected from
light at 90 °C for 5-9 h. The mixture was cooled to ambient
temperature and filtered and the residue washed with dry ethanol.
The filtrate was evaporated in Vacuo and the viscous residue purified
by column chromatography (SiO2 60, CH2Cl2/hexane 8:2) to afford
1-3.
General Procedure B for the Preparation of (5-Nitro-6-
phenyl-pyrimidin-4-yl)-phenyl-amines (4-11). The appropriate
4-chloro-5-nitro-6-phenyl-pyrimidines A were refluxed together
with the suitable anilines (1.5-4 equiv) and triethylamine (1.5-4
equiv) in dioxane for 7-10 h. The volatiles were evaporated in
Vacuo, and the residue was crystallized from cyclohexane. The
crystals were collected and dissolved in CH2Cl2, and the solution
was filtered over SiO2 using CH2Cl2 as eluent. The filtrate was
evaporated in Vacuo to afford 4-11.
General Procedure C for the Preparation of 6-(Aryl)-N’4’-
(aryl)-pyrimidine-diamines (12-20). The appropriate (5-nitro-6-
phenyl-pyrimidin-4-yl)-phenyl-amines B were dissolved in ethanol,
palladium on charcoal (Pd-C, 10% on charcoal) or sulfidated
platinum on charcoal (PtS-C, 5% on charcoal) in case of 12 and
13 was added, and the mixture stirred at ambient temperature for
30 min at a hydrogen pressure of 4 bar. The mixture was filtered,
and the volatiles were evaporated in Vacuo to afford 12-20.
General Procedure D for the Preparation of 6,9-Diaryl-
purine-8-ones (21-29). The appropriate 6-(Aryl)-N’4’-(aryl)-
pyrimidine-diamines C and N,N-carbonyldiimidazole were dissolved
in a mixture of dioxane and toluene, and the mixture was refluxed
for 2.5-6 h. The volatiles were evaporated in Vacuo, and the residue
was purified by column chromatography (SiO2 60, toluene/acetone
8:2) to afford 21-29.
New synthetic work will be directed toward derivatization
of the scaffold and side chains. The combination of different
computational methods, e.g., the docking geometries from
Autodock and the scoring by FlexX allows to eliminate possible
poor candidates prior to synthetic attempts. Consequently, it is
possible to focus on the most promising candidates for further
improvement of the affinity of purine derivatives to p38 MAP
kinase. We were able to compute two possible binding modes
for our new purine series with ligand docking studies and verify
the model by reference to a related crystal structure and to rate
possible conformations by means of Pearlman’s predictive index.
This model contributes new and promising aspects for future
Acknowledgment. We thank Klaus Albert and his co-
workers for the execution of the temperature-dependent NMR

2066 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 9
Hauser et al.
(22) Bott, G.; Field, L. D.; Sternhell, S. Steric effects. A study of a
rationally designed system. J. Am. Chem. Soc. 1980, 102, 5618-
5626.
(23) Laufer, S.; Thuma, S.; Peifer, C.; Greim, C.; Herweh, Y.; Albrecht,
A.; Dehner, F. An immunosorbent, nonradioactive p38 MAP kinase
assay comparable to standard radioactive liquid-phase assays. Anal.
Biochem. 2005, 135-137.
experiments. The authors are grateful to Sabine Luik for
assistance in biological testing and would like to thank Merckle
GmbH, Ulm, Germany, and Fonds der Chemischen Industrie,
Germany, for their generous support of this work. Also thanks
are given to Hossam Elgabarty for discussion.
(24) Fitzgerald, C. E.; Patel, S. B.; Becker, J. W.; Cameron, P. M.; Zaller,
D.; Pikounis, V. B.; O’Keefe, S. J.; Scapin, G. Structural basis for
p38a MAP kinase quinazolinone and pyridol-pyrimidine inhibitor
specificity. Nat. Struct. Biol. 2003, 10, 764-769.
(25) MOE. Chemical Computing Group Inc. Montreal, 2004.
(26) Halgren, T. A. Maximally diagonal force constants in dependent
angle-bending coordinates. II. Implications for the design of empirical
force fields. J. Am. Chem. Soc. 1990, 112, 4710-4723.
(27) Gasteiger, J.; Marsili, M. Prediction of proton magnetic resonance
shifts: the dependence on hydrogen charges obtained by iterative
partial equalization of orbital electronegativity. Org. Magn. Reson.
1981, 15, 353-360.
Supporting Information Available: Experimental procedures,
spectroscopic data, table of elemental analyses, and temperature-
dependent NMR spectra for 24 and 26. This material is available
free of charge via the Internet at http://pubs.acs.org.
References
(1) Kumar, S.; Boehm, J.; Lee, J. C. p38 MAP kinases: key signalling
molecules as therapeutic targets for inflammatory diseases. Nat. ReV.
Drug DiscoVery 2003, 2, 717-726.
(2) Johnson, G. L.; Lapadat, R. Mitogen-activated protein kinase
pathways mediated by ERK, JNK, and p38 protein kinases. Science
2002, 298, 1911-1912.
(3) Smolen, J. S.; Steiner, G. Therapeutic strategies for rheumatoid
arthritis. Nat. ReV. Drug DiscoVery 2003, 2, 473-488.
(4) Jackson, P. F.; Bullington, J. L. Pyridinylimidazole based p38 MAP
kinase inhibitors. Curr. Top. Med. Chem 2002, 2, 1011-1020.
(5) Laufer, S. A.; Wagner, G. K. From Imidazoles to Pyrimidines: New
Inhibitors of Cytokine Release. J. Med. Chem. 2002, 45, 2733-2740.
(6) Cohen, P. The development and therapeutic potential of protein kinase
inhibitors. Curr. Opin. Chem. Biol. 1999, 3, 459-465.
(7) Bellon, S.; Fitzgibbon, M. J.; Fox, T.; Hsiao, H. M.; Wilson, K. P.
The structure of phosphorylated p38gamma is monomeric and reveals
a conserved activation-loop conformation. Structure Fold. Des. 1999,
7, 1057-1065.
(8) Wang, Z.; Canagarajah, B. J.; Boehm, J. C.; Kassisa, S.; Cobb, M.
H.; Young, P. R.; Abdel-Meguid, S.; Adams, J. L.; Goldsmith, E. J.
Structural basis of inhibitor selectivity in MAP kinases. Structure
1998, 6, 1117-1128.
(9) Traxler, P. Tyrosine kinase inhibitors in cancer treatment (part II).
Expert Opin. Ther. Pat. 1998, 8, 1599-1625.
(10) Laufer, S. A.; Domeyer, D. M.; Scior, T. R.; Albrecht, W.; Hauser,
D. R. Synthesis and biological testing of purine derivatives as
potential ATP-competitive kinase inhibitors. J. Med. Chem. 2005,
48, 710-722.
(11) Wan, Z.; Boehm, J. C.; Bower, M. J.; Kassis, S.; Lee, J. C.; Zhao,
B.; Adams, J. L. N-Phenyl-N-purin-6-yl ureas: The design and
synthesis of p38a MAP kinase inhibitors. Bioorg. Med. Chem. Lett.
2003, 13, 1191-1194.
(12) Hanke, J. H.; Gardner, J. P.; Dow, R. L.; Changelian, P. S.; Brissette,
W. H.; Weringer, E. J.; Pollok, B. A.; Connelly, P. A. Discovery of
a novel, potent, and Src family-selective tyrosine kinase inhibitor.
Study of Lck- and FynT-dependent T cell activation. J. Biol. Chem.
1996, 271, 695-701.
(13) Liu, Y.; Bishop, A.; Witucki, L.; Kraybill, B.; Shimizu, E.; Tsien,
J.; Ubersax, J.; Blethrow, J.; Morgan, D. O.; Shokat, K. M. Structural
basis for selective inhibition of Src family kinases by PP1. Chem.
Biol. 1999, 6, 671-678.
(14) Zhu, X.; Kim, J. L.; Newcomb, J. R.; Rose, P. E.; Stover, D. R.;
Toledo, L. M.; Zhao, H.; Morgenstern, K. A. Structural analysis of
the lymphocyte-specific kinase Lck in complex with non-selective
and Src family selective kinase inhibitors. Structure 1999, 7, 651-
661.
(15) Bain, J.; McLauchlan, H.; Elliott, M.; Cohen, P. The specificities of
protein kinase inhibitors: an update. Biochem. J. 2003, 371, 199-
204.
(16) Cohen, P.; Goedert, M. Engineering protein kinases with distinct
nucleotide specificities and inhibitor sensitivities by mutation of a
single amino acid. Chem. Biol. 1998, 5, R161-R164.
(17) Doytchinova, I.; Petrova, S. \“N6-N7\” - a modification of the
\“N6-C8\” model for the binding site on adenosine A1 receptors
with improved steric and electrostatic fit. Med. Chem. Res. 1998, 8,
143-152.
(18) Cocuzza, A. J.; Chidester, D. R.; Culp, S.; Fitzgerald, L.; Gilligan,
P. Use of the Suzuki reaction for the synthesis of aryl-substituted
heterocycles as corticotropin-releasing hormone (CRH) antagonists.
Bioorg. Med. Chem. Lett. 1999, 9, 1063-1066.
(28) Sanner, M. F. Python: A programming language for software
integration and development. J. Mol. Graphics Modell. 1999, 17,
57-61.
(29) Michel, F. Sanner; Daniel Stoffler; Arthur, J. Olson ViPEr a Visual
Programming Environment for Python. Presented at the 10th
International Python Conference, Feb 4-7, 2002. (http://
www.scripps.edu/∼sanner/html/papers/IPC02.pdf).
(30) Weiner, S. J.; Kollman, P. A.; Nguyen, D. T.; Case, D. A. An all
atom force field for simulations of proteins and nucleic acids. J.
Comput. Chem. 1986, 7, 230-252.
(31) Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.; Hart, W.
E.; Belew, R. K.; Olson, A. J. Automated docking using a Lamarckian
genetic algorithm and an empirical binding free energy function. J.
Comput. Chem. 1998, 19, 1639-1662.
(32) Sybyl 7.0. Tripos Inc. St.Louis, MO, 2004.
(33) Rarey, M.; Kramer, B.; Lengauer, T.; Klebe, G. A fast flexible
docking method using an incremental construction algorithm. J. Mol.
Biol. 1996, 261, 470-489.
(34) Muegge, I.; Martin, Y. C. A General and Fast Scoring Function for
Protein-Ligand Interactions: A Simplified Potential Approach. J.
Med. Chem. 1999, 42, 791-804.
(35) Kuntz, I. D.; Blaney, J. M.; Oatley, S. J.; Langridge, R.; Ferrin, T.
E. A geometric approach to macromolecule-ligand interactions. J.
Mol. Biol. 1982, 161, 269-288.
(36) Eldridge, M. D.; Murray, C. W.; Auton, T. R.; Paolini, G. V.; Mee,
R. P. Empirical scoring functions: I. The development of a fast
empirical scoring function to estimate the binding affinity of ligands
in receptor complexes. J. Comput.-Aided Mol. Des. 1997, 11, 425-
445.
(37) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R.
Development and validation of a genetic algorithm for flexible
docking. J. Mol. Biol. 1997, 267, 727-748.
(38) Pearlman, D. A.; Charifson, P. S. Are free energy calculations useful
in practice? A comparison with rapid scoring functions for the p38
MAP kinase protein system. J. Med. Chem. 2001, 44, 3417-3423.
(39) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P.
Development and use of quantum mechanical molecular models. 76.
AM1: a new general purpose quantum mechanical molecular model.
J. Am. Chem. Soc. 1985, 107, 3902-3909.
(40) Stewart, J. J. P. Semiempirical molecular orbital methods. In ReViews
in Computational Chemistry; VCH Publishers: New York, 1990;
Volume 1, pp 45-81
(41) Nowak, W.; Wietzbowska, M. A theoretical study of geometry and
transition moment directions of flexible fluorescent probes - Acetoxy
derivatives of phenylanthracene. THEOCHEM 1996, 368, 223-234.
(42) Testa, A. C. Hydrogen bonding and the protonated water dimer.
Spectrosc. Lett. 1999, 32, 819-828.
(43) Bures, M.; Bezus, J. Study of hydrogen bonding in carboxylic acids
by the MNDO/M method. Collect. Czech. Chem. Commun. 1994,
59, 1251-1260.
(44) Masunov, A.; Dannenberg, J. J. Theoretical study of urea and thiourea.
2. Chains and ribbons. J. Phys. Chem. B 2000, 104, 806-810.
(45) Stahl, M.; Rarey, M. Detailed Analysis of Scoring Functions for
Virtual Screening. J. Med. Chem. 2001, 44, 1035-1042.
(46) InsightII 2000. Accelrys Inc. San Diego, CA, 2000.
(47) Connolly, M. L. Solvent-accessible surfaces of proteins and nucleic
acids. Science 1983, 221, 709-713.
(48) Ghose, A. K.; Viswanadhan, V. N.; Wendoloski, J. J. Prediction of
hydrophobic (lipophilic) properties of small organic molecules using
fragmental methods: An analysis of ALOGP and CLOGP methods.
J. Phys. Chem. A 1998, 102, 3762-3772.
(19) Davoll, J. Pyrrolo[2,3-d]pyrimidines. J. Chem. Soc., Abstr. 1960,
131-138.
(20) Staab, H. A. Syntheses with heterocyclic amides (azolides). Angew.
Chem. 1962, 74, 407-23.
(21) Mazzanti, A.; Lunazzi, L.; Minzoni, M.; Anderson, J. E. Rotation in
biphenyls with a single ortho-substituent. J. Org. Chem. 2006, 71,
5474-5481.
JM061061W