Cyclin-dependent kinase 4 inhibitors as a treatment for cancer. Part 1: Identification and optimisation of substituted 4,6-Bis anilino pyrimidines

Article abstract of DOI:10.1016/S0960-894X(03)00202-6

Using a high-throughput screening campaign, we identified the 4,6-bis anilino pyrimidines as inhibitors of the cyclin-dependent kinase, CDK4. Herein we describe the further chemical modification and use of X-ray crystallography to develop potent and selective in vitro inhibitors of CDK4.

Full text of DOI:10.1016/S0960-894X(03)00202-6

Bioorganic & Medicinal Chemistry Letters 13 (2003) 2955–2960
Cyclin-Dependent Kinase 4 Inhibitors as a Treatment for Cancer.
Part 1: Identification and Optimisation of Substituted 4,6-Bis
Anilino Pyrimidines
John F. Beattie, Gloria A. Breault,* Rebecca P. A. Ellston, Stephen Green,
Philip J. Jewsbury, Catherine J. Midgley, Russell T. Naven, Claire A. Minshull,
Richard A. Pauptit, Julie A. Tucker and J. Elizabeth Pease*
AstraZeneca, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK
Revised 19 September 2002; accepted 20 December 2002
Abstract—Using a high-throughput screening campaign, we identified the 4,6-bis anilino pyrimidines as inhibitors of the cyclin-
dependent kinase, CDK4. Herein we describe the further chemical modification and use of X-ray crystallography to develop potent
and selective in vitro inhibitors of CDK4.
# 2003 Elsevier Ltd. All rights reserved.
Cancer has longbeen recognized as a disease of aber-
rant cellular proliferation. Traditional cancer therapies
aim to exploit the proliferative machinery, for example,
DNA replication or chromosome segregation. As such,
they demonstrate only partial selectivity for tumour
cells compared with normal proliferatingtissues such as
the gut and bone marrow. More recently, many of the
proteins regulating entry into, progression through and
exit from the cell cycle, have been identified and their
roles elucidated.1 In particular, it is now recognised that
many of these proteins are themselves mutated, deleted
or over-expressed in cancer leadingto the deregulated
growth of the tumour cell. Such proteins, therefore,
represent potential targets to develop more selective
agents that may have an improved therapeutic margin
in the clinic.
then another member of the CDK family to become
active duringcell cycle progression. Inhibitory proteins,
such as members of the p16INK4A family or the
p21CIP/p27KIP family, also regulate CDK activity and
these inhibitors must be either sequestered or destroyed
to allow the CDK enzyme to become fully activated. As
cells respond to the presence of mitogens, the level of
cyclin D1 increases and eventually triggers the activa-
tion of CDK4 and CDK6. These kinases phosphorylate
the retinoblastoma tumour suppressor protein pRb
thereby abrogating its inhibition of members of the E2F
family of transcription factors. This then triggers a
programme of gene expression that results in entry into
the S phase of the cell cycle.
There are now many examples where tumours contain
multiple copies of the cyclin D1 gene or express abnor-
mally high levels of the protein.2 Similarly, many
tumours have been described that contain mutations,
The cyclin-dependent kinases (CDKs) are especially
important in controllingentry into and prorgession
through the cell cycle.1 They are a family of serine/
threonine kinases whose activity is regulated at several
levels. The bindingof each CDK to a specific cyclin
partner protein is required for activity. The synthesis
and degradation of cyclins is tightly controlled such that
their level fluctuates duringthe cell cycle. It is these
fluctuatinglevels of the cyclin that cause first one, and
3,4
deletions or silencingof the p16 or the pRb egne.
Finally, mutations of CDK4 itself have been described
within melanoma patients, where the region of the
enzyme that binds p16 has been mutated makingthe
enzyme resistant to inhibition.5 Together, these findings
imply that the deregulation of the pRb pathway, and
CDK4, is important in cancer progression. The inhibi-
tion of CDK4 may therefore be a valuable approach to
treat tumours that have lost the natural CDK4 inhi-
bitor, p16. In this respect, it is interestingto note that
*Correspondingauthor. E-mail: j.elizabeth.pease@astrazeneca.com
0960-894X/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0960-894X(03)00202-6

2956
J. F. Beattie et al. / Bioorg. Med. Chem. Lett. 13 (2003) 2955–2960
adenovirus vectors expressingp16 have been shown to
induce tumour regression in animal models.6
Table 1. Structures and enzyme activity for 4,6-bis anilino
pyrimidines
A number of groups have identified CDK inhibitors.7À18
In this paper, we describe a series of novel, potent inhib-
itors of CDK4 with selectivity against CDK2. Such
compounds may prove valuable in treatingtumours
with deregulated CDK4 such as those lacking p16.
A high-throughput screen of the AstraZeneca com-
pound collection usingan in vitro scintillation proxi-
mity assay (SPA) for measuringincorporation of [ g-33-
P]-adenosine triphosphate into GST-Rb,18 identified the
4,6-bis anilino pyrimidine 1 as an inhibitor of CDK4
(IC50 15 mM) possessingsome selectivity against CDK2
(IC50 93 mM). This prompted us to initiate chemistry to
explore the activity of this class of compounds.
Compd
CDK4 IC50 (mM)a
CDK2 IC50 (mM)a
X
1
2
3
4
5
6
7
8
9
15
33
8
18
27
8
4
23
2
93
>100
39
2,4-F2
H
2-Me
2-iPr
2-F
2-Cl
2-Br
2-OMe
2-NO2
24
>100
95
28
85
25
aValues are means of at least two determinations. The assay-to-assay
variation was generally Æ2-fold based on the results of a standard
compound.
ited sub-series it was evident that a range of sub-
stituents are tolerated with the 2-bromo 7 and 2- nitro
9 derivatives beingthe most potent. However, the SAR
was difficult to interpret. Later, it became apparent
from the crystal structures that the B-ringin this series
takes up a range of orientations making interpretation
of activity trends difficult. The 2-chloro 6 and 2-methyl
3 substituents were of interest due to the increased
Our initial goals were to improve potency, demonstrate
evidence of SAR and, ideally, generate compounds with
adequate solubility to allow protein ligand X-ray crys-
tallography studies. The synthetic route to 1 and related
compounds is illustrated in Scheme 1. Reaction of 4,6-
dichloropyrimidine with 4-hydroxyaniline gave the
monochloro intermediate 1-1. This intermediate was
treated with the appropriate aniline to introduce ringB
resultingin intermediate 1-2. The phenol was then
alkylated with epibromohydrin followed by ringopen-
ingof the epoxide with dimethyl amine resultingin the
final compound.
potency. These substituents alongwith the 2-fluoro
5
were chosen for further investigation because the
accessibility of polyfunctionalised anilines would allow
the work to progress to disubstitution of ring B, which
we believed would result in enhancements in potency.
Amongthe disubstituted derivatives that were prepared,
three were of particular interest (Table 2). The 2,5-
dichloro 10 demonstrated that enhanced potency could
be achieved. The 2,6-difluoro 11 and the 2-fluoro, 5-tri-
fluoromethyl 12 were also significant as they combined
moderate potency with suitable solubility for structural
studies19 and led to protein–ligand crystal structures
One aspect of the work concentrated on evaluatingthe
effect of substituents in ringB, such as the 2-sub-
stituents illustrated in Table 1. The unsubstituted com-
pound 2 was less potent than the original hit but still
retained some activity and selectivity. Within this lim-
Table 2. Structures and enzyme activity for disubstituted 4,6-bis-
anilino pyrimides
Compd
CDK4 IC50 (mM)a
CDK2 IC50 (mM)a
X
Y
10
11
12
1
9
6
22
38
35
2-Cl
2-F
2-F
5-Cl
6-F
5-CF3
aValues are means of at least two determinations. The assay-to-assay
variation was generally Æ2-fold based on the results of a standard
compound.
Scheme 1. Synthesis of 4,6-bis anilino pyrimidines.

J. F. Beattie et al. / Bioorg. Med. Chem. Lett. 13 (2003) 2955–2960
2957
that allowed us to improve our understandingof how
the compounds were binding.
In these structural studies, inactive CDK2 monomer
was used as a structural surrogate for the CDK4 target.
This approach is convenient since the crystal structure
of CDK2 is known,20 and the crystals are particularly
robust, often diffract to high resolution and can readily
be soaked in inhibitor solutions enablingrapid experi-
mental support for molecular design strategies. The
bindingconformations obtained with CDK2 were
interpreted assumingthey were relevant to the true
CDK4 target. While the use of these structural surro-
gates for molecular design will be limited where the
details of the protein–ligand interactions differ, our
studies of a range of inhibitor series bound to a range of
kinase targets indicate that, in general, the bioactive
conformation of the series is conserved, even between
kinases with distant sequence relationships (unpublished
work).
Crystal structures19 of CDK2 complexed with 11 and 12
revealed the bindingmode of this series: the pyrimidine
N1 acts as a hydrogen bond acceptor with the 6-aniline
NH actingas a hydrogen bond donor ( Fig. 1). The for-
mer interaction, with the Leu 83 amide NH, is also seen
in the CDK2–ATP complex;20 the second interaction is
with the Leu 83 amide carbonyl oxygen. Both these
interactions are consistent with the accepted kinase
purine-mimetic pharmacophore: a central hydrogen
bond acceptor flanked by two hydrogen bond donating
groups. Purine-mimetic kinase inhibitors reported to
date have one, two or all three of these interactions.20,33
Figure 1. Crystal structures of CDK2 complexed with 11 and 12: (a) ribbon diagram showing the location of 12 in the purine pocket between the N-
terminal and C-terminal domains of CDK2; (b) final 2Fo-Fc electron density for the protein (green, 1.3s level) and for the inhibitor 12 (red, 1.3s
level) calculated in the vicinity of the ATP-bindingsite of CDK2. The trifluoromethyl substituent is clearly visible definingthe orientation of ring B.
The bridging water molecule between the 4-anilino NH and Asp145 is well defined, whereas Lys33 becomes mobile. There is typically very little
density for the solubilisinggroup (which is modelled as a racemic mixture) indicatingthis part of the inhibitor is disordered; (c) superposition of the
bindingmodes of 11 and 12. The racemic solubilisinggroup is also poorly defined in 11. RingB in 11 adopts two conformations: in both, one of the
fluoro substituents displaces the bridging water. In both, the orientation is different to 12. The figures were prepared using Bobscript and
Raster3D.30À32

2958
J. F. Beattie et al. / Bioorg. Med. Chem. Lett. 13 (2003) 2955–2960
Given the symmetry of the 4,6-bis anilino pyrimidine
substructure, the orientation within the bindingsite is
determined by the substitution pattern on the aniline
rings. In this series, ring A was para-substituted by a
‘solubilising’ group, and is oriented towards solvent at
the edge of the ATP binding site. Ring B, with lipophilic
substituents, lies inside the bindingsite. Comparison of
the two structures, however, revealed that while the
position of the hydrogen bonding groups was con-
served, the orientation of ringB was variable. In
particular, in the complex with 12 a water molecule
bridges between the 4-aniline NH and Asp145, disrupt-
ingthe Lys33-Asp145 salt bridge. This water molecule is
displaced by ringB in the complex with 11, with the
2-fluoro near the water position (Fig. 1).
Since these initial structures indicated that this water
molecule could be readily displaced, and that ringB
could occupy a range of orientations, we decided to
synthesise compounds that could exploit the space
available in both ringB bindingmodes simultaneously.
By N-alkylatingthe 4-aniline NH, it was proposed that
ringB would orientate in a manner similar to that seen
in the complex with 12 (Fig. 1), and the N-alkyl sub-
stituent would displace the water molecule.
As earlier work had suggested the 2,5-dichloro sub-
stitution pattern to be potent we decided to investigate
the effects of N-alkylation within this series. The 4,6-
dichloropyrimidine was reacted with 2,5-dichloroaniline
followed by reaction with 4-hydroxyaniline. The phenol
was then reacted with epibromohydrin. The resulting
epoxide 2-1 acted as a protectingrgoup duringthe
alkylation. The aniline nitrogen was then alkylated and
the epoxide opened with dimethyl amine to yield the
final product (Scheme 2).
Scheme 2. Synthesis of alkylated 4,6-bis anilino pyrimidines.
Table 3. Structures and enzyme activity for alkylated 4,6-bis anilino
pyrimides
A selection of compounds that were alkylated on the
4-aniline nitrogen is shown in Table 3. In general, the
potency of the compounds improved with alkylation.
By far the most advantageous N-substituent was cya-
nomethyl, 13 that showed a 20-fold improvement in
potency over the unsubsituted, 10. Lengthening of the
carbon chain by one 14 reduced the activity by an order
of magnitude. The cyano group appeared to make spe-
cific interactions, as changing to an acetylene group, 15,
resulted in a 10-fold decrease in potency, despite an
increase in the lipophilicity. Once the N-substituent
became bulky 16, potency was substantially reduced,
however, favourable effects on selectivity versus CDK2
were still observed.
Compd
CDK4 IC50 (mM)a
CDK2 IC50 (mM)a
R
10
13
14
15
16
1
22
1
5
19
H
CH2CN
CH2CH2CN
CH2CCH
CH2Ph
0.05
0.5
0.8
5
>100
aValues are means of at least two determinations. The assay-to-assay
variation was generally Æ2-fold based on the results of a standard
compound.
chain. The loss of this electrostatic complimentarity
may account for the drop in activity found for 15. Bulky
alkyl chains 16 can no longer be accommodated by this
bindingmode, presumably causingan alternative orien-
tation with consequent loss in activity.
The structure of an analogue of 13, the 2-bromo, 4-
methyl 17 (CDK4 IC50 0.1 mM, CDK2 IC50 3 mM)
bound to CDK2 was solved revealinga bindingmode
consistent with the design hypothesis: the B-ring orien-
tation was similar to that found for 12 but with the
methylenenitrile displacingthe bridigngwater. The
nitrile C lies within 1.3 A of the water, consistent with a
drop of activity on extendingthe alkyl chain, 14 (Fig. 2).
The nitrile triple bond is stacked against the p cloud of
the Phe80 side chain, with its polarity complimentary to
the local electrostatic potential caused by the Lys33 side

J. F. Beattie et al. / Bioorg. Med. Chem. Lett. 13 (2003) 2955–2960
2959
Figure 2. Superposition of the bindingmodes of 12 and 17. The nitrile C lies within 1.3 A of the position occupied by the bridging water molecule in
the complex between CDK2 and 12. The racemic solubilisinggroup is also poorly defined in 17. RingB in 17 adopts an orientation similar to 12.
The figures were prepared using Bobscript and Raster3D.30À32
Keller, P. R.; Wu, Z.; Dobrusin, E.; Leopold, W. R.; Fattaey,
A.; Garrett, M. D. J. Biol. Chem. 2001, 276, 16617.
13. Soni, R.; O’Reilly, T.; Furet, P.; Muller, L.; Stephan, C.;
Zumstein-Mecker, S.; Fretz, H.; Fabbro, D.; Chaudhuri, B. J.
Natl. Cancer Inst. 2001, 93, 436.
14. Honma, T.; Yoshizumi, T.; Hashimoto, N.; Hayashi, K.;
Kawanishi, N.; Fukasawa, K.; Takaki, T.; Ikeura, C.; Ikuta,
M.; Suzuki-Takahashi, I.; Takashi, H.; Nishimura, S.; Mor-
ishima, H. J. Med. Chem. 2001, 44, 4628.
15. Soni, R.; Muller, L.; Furet, P.; Schoepfer, J.; Stephan, C.;
Zumstein-Mecker, S.; Fretz, H.; Chaudhuri, B. Biochem. Bio-
phys. Res. Commun. 2000, 275, 877.
16. Jeong, H.-W.; Kim, M.-R.; Son, K.-H.; Young Han, M.;
Ha, J.-H.; Garnier, M.; Meijer, L.; Kwon, B.-M. Bioorg. Med.
Chem. Lett. 2000, 10, 1819.
In conclusion, we identified a novel series of 4,6-bis
anilino pyrimidines as CDK4 inhibitors usinghihg
throughput screening. Through modifications to ring B
an increase in potency (10-fold) and improved solubility
were achieved, allowingprotein–ligand structures (with
CDK2) to be obtained. These revealed that two binding
modes of the compounds are possible and the insight
provided by the structural work led us to introduce
specific substituents on the 4-aniline NH. The resulting
2,5-disubstituted N-alkylated 4,6-bis anilino pyrimidines
represent highly potent and selective inhibitors of
CDK4 that are the subject of further investigations.
17. Ryu, C.-K.; Kang, H.-Y.; Lee, S. K.; Nam, K. A.; Hong,
C. Y.; Ko, W.-G.; Lee, B.-H. Bioorg. Med. Chem. Lett. 2000,
10, 461.
18. Breault, G. A., Pease, J. E. PCT Int. Application WO
0012485 A1 20000309.
Acknowledgements
We are grateful for the excellent support we have
received from Michael Block and would also like to
thank Sandra Oakes for repeat enzyme measurements.
19. Protein and crystals were obtained accordingto estab-
lished procedures21,22 Diffraction data were collected on
beamline 9.6 at SRS, Daresbury, at 100 K. Data processing,
data reduction and structure solution by molecular replace-
ment were carried out usingprorgams from the CCP4
References and Notes
suite.23 Compounds 11, 12, and 17 were modelled as racemic
24
1. Sherr, C. J. Science 1996, 274, 1672.
mixtures of the chiral alcohol usingInsihgtII,
elled into electron density usingQUANTA.
complex model was refined usingCNX
and mod-
The protein
25
2. Barnes, D. M.; Gillet, C. E. Breast Cancer Res. Treat. 1998,
52, 1.
3. Roussel, M. F. Oncogene 1999, 18, 5311.
4. Nevins, J. R. Hum. Mol. Genet. 2001, 10, 699.
5. Tsao, H.; Benoit, E.; Sober, A. J.; Thiele, C.; Haluska, F. G.
Cancer Res. 1998, 58, 109.
and Buster,27 and
26
final structures28,29,30 have been deposited in the Protein
Data Bank with deposition codes 1h00, 1h06 and 1h07
together with structure factors and detailed experimental
conditions.
6. Frizelle, S. P.; Grim, J.; Zhou, J.; Gupta, P.; Curiel, D. T.;
Geradts, J.; Kratzke, R. A. Oncogene 1998, 16, 3087.
7. Webster, K. R.; Kimball, S. D. Emerg. Drugs 2000, 5, 45.
8. Kimball, S. D.; Webster, K. R. In Annual Reports in Med-
icinal Chemistry; Doherty, A. M., Ed.; Academic: San Diego,
2001; Vol. 36, p 139 and references therein.
9. Ikuta, M.; Kamata, K.; Fukasawa, K.; Honma, T.;
Machida, T.; Hirai, H.; Suzuki-Takahashi, I.; Hayama, T.;
Nishimura, S. J. Biol. Chem. 2001, 276, 27548.
20. Schulze-Gahmen, U.; DeBondt, H. L.; Kim, S.-H. J. Med.
Chem. 1996, 39, 4540.
21. Lawrie, A. M.; Noble, M. E.; Tunnah, P.; Brown, N. R.;
Johnson, L. N.; Endicott, J. A. Nat. Struct. Biol. 1997, 4, 796.
22. Legraverend, M.; Tunnah, P.; Noble, M.; Ducrot, P.;
Ludwig, O.; Grierson, D. S.; Leost, M.; Meijer, L.; Endicott,
J. J. Med. Chem. 2000, 43, 1282.
23. CCP4 Acta Crystallogr. 1994, D50, 760.
24. InsightII, Accelrys.
10. Carini, D. J.; Kaltenbach, R. F.; Liu, J.; Benfield, P. A.;
Boylan, J.; Boisclair, M.; Brizuela, L.; Burton, C. R.; Cox, S.;
Grafstrom, R.; Harrison, B. A.; Harrison, K.; Akamike, E.;
Markwalder, J. A.; Nakano, Y.; Seitz, S. P.; Sharp, D. M.;
Trainor, G. L.; Sielecki, T. M. Bioorg. Med. Chem. Lett. 2001,
11, 2209.
11. Honma, T.; Hayashi, K.; Aoyama, T.; Hashimoto, N.;
Machida, T.; Fukasawa, K.; Iwama, T.; Ikeura, C.; Suzuki-
Takahashi, I. J. Med. Chem. 2001, 44, 4615.
25. Quanta2000, Accelrys.
26. CNX version 2000.1, Accelrys.
27. Buster, Global Phasing.
28. Crystallographic statistics for 11 are: space group P212121,
unit cell 54.3, 72.2, 72.1 A, resolution 1.6 A, 35229 unique
reflections from 106,279 observations give 93.6% complete-
ness with Rmerge=3.6% and mean I/s(I) of 21.The final model
containing2175 protein, 200 water and 44 inhibitor atoms has
an R-factor of 21% (Rfree using5% of data is 25%). Mean
temperature factor for protein is 26.4 and for ligand is 43.2 A2.
12. Fry, D. W.; Bedford, D. C.; Harvey, P. H.; Fritsch, A.;

2960
J. F. Beattie et al. / Bioorg. Med. Chem. Lett. 13 (2003) 2955–2960
29. Crystallographic statistics for 12 are: space group
P212121, unit cell 54.3, 72.9, 73.8 A, resolution 2.3 A, 11954
unique reflections from 30352 observations give 90.6% com-
pleteness with Rmerge=7.8% and mean I/s(I) of 13. The final
model containing2181 protein, 88 water and 33 inhibitor
atoms has an R-factor of 22% (Rfree using5% of data is
26%). Mean temperature factor for protein is 23.5 and for
ligand is 46.8 A2.
unique reflections from 53,065 observations give 87.7% com-
pleteness with Rmerge=9.6% and mean I/s(I) of 10. The final
model containing2285 protein, 200 water and 39 inhibitor
atoms has an R-factor of 21% (Rfree using5% of data is
23.5%). Mean temperature factor for protein is 23 and for
ligand is 44 A2.
31. Esnouf, R. J. Mol. Graph. 1997, 156, 132.
32. Merritt, E. A.; Murphy, M. E. P. Acta Crustallogr. 1994,
D50, 869.
30. Crystallographic statistics for 17 are: space group
P212121, unit cell 53.3, 71.5, 72.6 A, resolution 1.85 A, 21,455
33. Bacon, D.; Anderson, W. F. J. Mol. Graph. 1988, 6, 219.