Triazolo[1,5-a]pyrimidines as novel CDK2 inhibitors: Protein structure-guided design and SAR

Article abstract of DOI:10.1016/j.bmcl.2005.11.048

Crystallographic and modelling data, in conjunction with a medicinal chemistry template-hopping approach, led to the identification of a series of novel and potent inhibitors of human cyclin-dependent kinase 2 (CDK2), with selectivity over glycogen syntha

Full text of DOI:10.1016/j.bmcl.2005.11.048

Bioorganic & Medicinal Chemistry Letters 16 (2006) 1353–1357
Triazolo[1,5-a]pyrimidines as novel CDK2 inhibitors:
Protein structure-guided design and SAR
Christine M. Richardson,* Douglas S. Williamson, Martin J. Parratt, Jenifer Borgognoni,
Andrew D. Cansfield, Pawel Dokurno, Geraint L. Francis, Rob Howes,
Jonathan D. Moore, James B. Murray, Alan Robertson,
Allan E. Surgenor and Christopher J. Torrance
Vernalis (R&D) Ltd., Granta Park, Great Abington, Cambridge CB1 6GB, UK
Received 10 October 2005; revised 11 November 2005; accepted 14 November 2005
Available online 1 December 2005
Communicated by Stephen Neidle
Abstract—Crystallographic and modelling data, in conjunction with a medicinal chemistry template-hopping approach, led to the
identification of a series of novel and potent inhibitors of human cyclin-dependent kinase 2 (CDK2), with selectivity over glycogen
synthase kinase-3b (GSK-3b). One example had a CDK2 IC50 of 120 nM and showed selectivity over GSK-3b of 167-fold.
Ó 2005 Elsevier Ltd. All rights reserved.
Cyclin-dependent kinases (CDKs) are important for cell
cycle control. When activated by phosphorylation and
bound to the appropriate cyclin partner, they promote
progress from G1 to S and from G2 to M (CDK inacti-
vation is required for exit from mitosis). The central role
of CDKs in driving cell cycle progression supports the
hypothesis that CDK inhibitors will be useful in anti-
cancer treatment. One compound, roscovitine 1,1 is in
advanced clinical trials for this indication. None of the
CDK inhibitors disclosed to date are entirely specific
for a particular CDK. As the results from knock-out
mice imply substantial functional redundancy in the
CDK family, some degree of inhibitor promiscuity is
probably desirable. Structure-guided drug design
against CDK targets has largely focussed on CDK2,
as its crystal structure has been available since 1993.2
CDK2 is now thought to be dispensable for tumour for-
mation and maintenance.3,4 Fortunately, sequence
alignment and homology modelling indicate a high de-
gree of similarity between the active site of CDK2 and
that of CDK1, the kinase most likely to replace missing
CDK2 activity.5 Only two of the 28 residues bounding
the ligand binding site are not strictly conserved (H84S
and Q85M) and neither of these side-chain switches
would directly modify the active site, which implies that
identification of compounds targeting CDK1 as well as
CDK2 is likely. Therefore, we considered that using a
structure-guided strategy based on CDK2 was an appro-
priate means to generate CDK inhibitors that might
prove useful for the therapy of proliferative disorders.
Ph
NH
N
N
N
HN
N
OH
1
A medium-throughput screening campaign against a li-
brary of 3341 commercial compounds identified the pyr-
azolo[1,5-a]pyrimidine 2 (CDK2 IC50 1.8 lM, HCT116
GI50 8 lM).6–9 Lead optimisation of the resulting series
of pyrazolo[1,5-a]pyrimidine CDK2 inhibitors has been
previously reported.10 Given the intensely competitive
nature of the CDK2 arena, however, there was a need
to consider more novel scaffolds, and so a structure-
guided medicinal chemistry strategy was employed to
Keywords: Structure-guided drug design; CDK2; Triazolo [1,5-a]-
pyrimidines; SAR; Potency; Selectivity.
*
Corresponding author. Tel.: +44 1223 895434; fax: +44 1223
895556; e-mail: c.richardson@vernalis.com
0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.11.048

1354
C. M. Richardson et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1353–1357
address this. One aspect of this strategy was to consider
core templates which maintained similar interactions to
the pyrazolo[1,5-a]pyrimidine but which were free from
blocking IP; a so-called template-hopping strategy. The
work reported here concerns SAR around one novel ser-
ies of CDK2 inhibitors which resulted from this
approach.
Compound 6 also showed good activity against
CDK1, with an IC50 of 140 nM. The proposed binding
modes, obtained using rDock,11 were verified by X-ray
crystallography. Figure 1 shows the overlay of com-
pound 6 with NU6102 (4), as determined from overlays
of the two protein–ligand complexes, and with interac-
tions to the central templates highlighted. The crystal
structure of compound 6 in human CDK2 shows that
the aryl sulfonamide packs against the backbones of
His84 and Glu85, with the sulfonamide oxygens inter-
Early modelling work on the pyrazolo[1,5-a]pyrimidine
2, using the proprietary code rDock (formerly called
RiboDock),11 subsequently verified by X-ray crystallog-
raphy, suggested a binding mode where the bicyclic sys-
tem was located in the ATP binding site, making polar
interactions with the kinase binding motif. This com-
prises the backbone NH and carbonyl of Leu83 and
the backbone carbonyl of Glu81, with the interaction
between the Glu backbone and the ligand being a C–
HÁ Á ÁO hydrogen bond, which is now well recognized12
as an important type of intermolecular interaction, par-
ticularly in the kinase field. Replacement of the ligand
C3 atom with nitrogen results in a triazolo[1,5-a]pyrim-
idine scaffold, where the key interactions can be main-
tained, and which is amenable to substitution at the
desired positions. Compound 3, the triazolo[1,5-a]py-
rimidine variant of 2, was synthesised. This had a
CDK2 IC50 of 26 lM, CDK1 IC50 8.8 lM and
HCT116 GI50 >80 lM, with conservation of binding
mode verified by X-ray crystallography.
˚
acting with the backbone NH of Asp86 (3.12 A) and
˚
the side chain of Lys89 (3.2 A) and the nitrogen of the
sulfonamide being in proximity to the side chain of
˚
Asp86 (3.4 A), although with a relatively poor geometry
for a hydrogen bonding interaction. The cyclohexyl
group of compound 6 packs against a relatively hydro-
phobic surface involving Ile10, Val18 and Gln131 and
the backbones of Gly11 and Glu12, where the polar
atoms are orientated orthogonal to the ligand. There is
an additional interaction, not seen in the NU6102 (4)
˚
complex, from Lys33 (3.06 A) to the triazolo[1,5-a]py-
rimidine ring. It is interesting to note that, whilst the
interactions to the core template are conserved and the
aryl sulfonamide binds to a similar region of the active
site, when comparing the complexes of compound 6
and NU6102 (4) with CDK2, the O-cyclohexyl units
are slightly offset in the crystal structures. The compar-
ison in Figure 1 shows NU6102 as obtained from the
activated CDK2-cyclin A structure, whilst compound
6 is taken from the complex with monomeric CDK2.
This overlay is depicted, as it best shows the similarity
in chain length between the substituents in compound
6 and NU6102. Comparisons have also been undertaken
between compound 6 in the active form of CDK2 and
NU6102 (4) in the active (1h1s derived) form, and be-
tween the complexes of compound 6 and NU6102 (4)
in the inactive form of CDK2 as obtained in-house.
Whilst the binding modes for the core and the aryl sul-
fonamide moieties are conserved, the cyclohexyl group
shows some differences in orientation between the differ-
ent structures. This is consistent with results from dock-
ing work on compound 6, which suggested that the
linker between the triazolo[1-5a]pyrimidine core and
the cyclohexyl group was flexible, and that the cyclohex-
yl moiety could make favourable interactions with a
number of residues in the active site.
A large number of crystal structures are available on hu-
man CDK2 in complex with small ligands which bind
deeply within the ATP site, and which interact with
the kinase motif. An initial strategy involved exploration
of the extent to which side chains present in other ligand
series may be transferable to the new template. Of par-
ticular interest was NU6102 (4),13 the CDK2 complex
of which (pdb code: 1h1s) is available from the PDB.14
A structural overlay of the bound pyrazolo[1,5-a]pyrim-
idine 2 with the structure of the NU6102 (4) complex
suggested that the ligand templates did not directly
coincide, but were offset in order to satisfy a common
interaction pattern with the protein backbone. A direct
side-chain replacement strategy might not, therefore,
provide the optimal gain in potency.
SO2NH2
N
SO2NH2
Although the crystal structures of compounds 5 and 6
were prepared by soaking into crystals of apo-CDK2,2
some protein flexibility can be seen, with both ligands
appearing to mould the dynamic, glycine-rich loop (res-
idues Ile10 to Tyr19) which borders the active site, as
illustrated in Figure 2.16
HN
N
H
N
H
N
O
NH
N
N
N
N
N
N
N
N
O
X
X
N
Br
Optimisation of the triazolo[1,5-a]pyrimidine series was
undertaken, with the aim of improving potency for
CDK2 and selectivity over glycogen synthase kinase-
3b (GSK-3b), the main counterscreen. This was selected,
since CDK2 and GSK-3b have therapeutically antago-
nistic effects, but also exhibit some similarities17 in their
small molecule inhibition profiles. Optimisation was
achieved using the synthetic approach outlined in Scheme
1. The 7-Cl of 5,7-dichloro[1,2,4]-triazolo[1,5-a]pyrimidine
2 X = CH
3 X = N
4
5 X = CH2
6 X = [CH2]0
Compound 5 was synthesised with the directly equiva-
lent side chains, whilst compound 6 was made with a
shorter side chain, as suggested by the structural over-
lays. The CDK2 assay data (Table 1) clearly show the
shorter side chain to be preferred, consistent with the
hypothesis, with an increase in potency of 32-fold.

C. M. Richardson et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1353–1357
1355
Table 1. Enzyme activity (CDK2, GSK-3b and CDK1), cell-based inhibition in HCT116 (colon) cell line and calculated SlogP values for
triazolo[1,5-a]pyrimidines 5, 6 and 11–17
R1
HN
N
N
R2
N
N
Compound R1
R2
CDK2 IC50 (lM)a,b GSK-3b IC50 (lM)c CDK1 IC50 (lM)b HCT116 GI50 (lM) SlogP d
O
5
6
–SO2NH2
11
29
—
11
25
2.5
2.2
O
–SO2NH2
–SO2NH2
–SO2NH2
0.35
0.19
0.14
NH
11
0.73
0.97
—
—
31
1.8
NH
12
13
14
7.4
—
39
23
0.77
1.2
HO
–SO2NH2 Et2N
0.40
0.27
11
3.4
NH
NH
–SO2NH2
0.78
—
>80
0.74
H N
15
16
17
–SO2NMe2
0.12
0.25
0.26
20
43
50
5.6
>80
>80
>80
1.3
1.5
2.9
H N
NH
NH
–SO2Me
16
H N
–SO2Ph
4.3
H N
a All IC50 and GI50 values are means of at least two determinations and are rounded to two significant figures where appropriate.
b [ATP] 100 lM (Km 50 lM).
c [ATP] 10 lM (Km 10 lM).
d Calculated with MOE.24,25
718 underwent facile displacement with a range of
amines to afford compounds of type 8. The 5-Cl then
underwent Suzuki coupling with boronic acids in the
presence of sodium carbonate and tetrakis(triphenyl-
phosphine)palladium(0) in 1,4-dioxane and water in a
Smith SynthesizerTM at 150 °C, to afford compounds of
type 9 (R2 = Ar). Alternatively, ethers of type 10
(X = O) could be synthesised by treating the chloride 8
in 1,4-dioxane and acetonitrile with an alcohol in the
presence of sodium hydride and then heating the result-
ing mixture in a microwave reactor at 150 °C. The 5-Cl
could also be displaced with amines by heating in a
microwave reactor at 150 °C in ethanol, resulting in
compounds of type 10 (X = NR3).
Modification of the C5 side chain of compound 6 was
investigated, as shown in Table 1. All the 5-position sub-
stituents comprising a heteroatom and a cyclohexyl ring
were tolerated, with the exception of compound 12,
where the presence of a trans-4-OH group resulted in a
10-fold loss in CDK2 potency. This is in line with
SAR published around NU6102 and its parent,
NU2058.19,20 The N,N-diethyl C5 substituent of com-
pound 13 was also well tolerated. The trans-4-amin-
ocyclohexylamino moiety, as found in H717 (18),21
a
potent and selective CDK2 inhibitor, was found to be
particularly beneficial for CDK2 potency. The sensitivi-
ty of the primary sulfonamide to modification was also
studied. N-Alkylation of the primary sulfonamide or

1356
C. M. Richardson et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1353–1357
Lys89
removal of the nitrogen to form a sulfone was tolerated,
with no significant loss in CDK2 potency, as can be seen
by comparing compound 14 with 15, 16 and 17. This is
consistent with the observation that the ligand sulfon-
amide moiety is interacting with the backbone of
Asp86 and the interaction from the sulfonamide nitro-
gen to the side chain of this residue is poor. Compound
15 was the most potent and selective example, with a
CDK2 IC50 of 120 nM and a GSK-3b IC50 of 20 lM,
a selectivity of 167-fold. The GSK-3b selectivity is med-
iated, in part, by the Asp86 to Thr138 switch, which
influences interaction to the ligand sulfonamide group.
In addition, comparison of the CDK2 complex with a
crystal structure of GSK-3b22 shows that there is a dif-
ferent conformational preference in the loops bounding
the cyclohexylamine binding site, which would necessi-
tate a different ligand conformation in this region and
disrupt the recognition of the side-chain amine.
Leu83
Asp86
Glu81
Lys33
Figure 1. Overlay of NU-6102 (4) in green with compound 6 in orange.
Residues involved in interactions to the templates (Lys89, Asp86,
Lys33, Glu81 and Leu83) shown in stick representation.15
O
N
N
N
HN
N
NH
18
A subset of compounds showed selectivity against Chk1
and PDK1, unlike NU6102 (4), which hits Chk1 and
PDK1 with IC50 <500 nM, and which is similarly potent
against GSK-3b (IC50 40 nM) and CDK2 (80 nM).23
Compound 14 was highly selective, with a Chk1 IC50
of 93 lM and PDK1 IC50 of 43 lM, as was compound
15, which had a Chk1 IC50 of 21 lM and PDK1 IC50
>200 lM. Compound 17 was less discriminatory, with
selectivity over CDK2 of 8.5-fold for Chk1 and 54-fold
for PDK1. Only compound 6 was more potent against
CDK1, but all compounds do show some inhibition,
although at a lower level than might be anticipated from
a comparison of the protein structure alone. These re-
sults broadly support the hypothesis that a structure-
guided strategy against CDK2 can give rise to inhibitors
with a degree of activity against other members of the
CDK family.
Figure 2. Overlaid complexes with compound 5 (carbon atoms in
magenta), and compound 6 (carbons in orange) with the native
structure with carbon atoms in grey. Backbone from Ile10 to Gly13
shown, with Ile10 side chain.15
Unlike the corresponding pyrazolo[1,5-a]pyrimidines,10
the triazolo[1,5-a]pyrimidines showed disappointing
activity against human colon tumour cells, with only 5
of the 10 examples in Table 1 having an HCT116 GI50
<80 lM. The most potent of these was compound 5,
which had a GI50 of 11 lM against HCT116 cells. This
value was extremely similar to its CDK2 IC50 (also
11 lM), however, which suggests that a different mode
of action may be responsible for its cytotoxic effect. Ef-
forts were made to improve the cellular potency of the
series by increasing logP, but unfortunately this did
not correlate with an improvement in HCT116 GI50.
For example, replacement of the methyl group of com-
pound 16 with a phenyl ring afforded compound 17,
which had an SlogP24,25 >1 unit higher and had enzyme
potency similar to that of the methyl sulfone 16, but
both compounds were also equally inactive in cells.
R1
NH
N
N
R1
(ii)
Cl
R2
N
NH
N
N
N
(i)
N
N
N
9
N
R1
Cl
N
(iii)
Cl
N
NH
7
8
N
N
R2
N
X
N
10
Scheme 1. Reagents and conditions: (i) R1NH2, MeOH, rt, 60–81%;
(ii) R2B(OH)2, Na2CO3, Pd(PPh3)4, 1,4-dioxane, H2O, microwave at
150 °C, 30 min, 12–35%; (iii) (for X = O) R2OH, NaH, 1,4-dioxane,
MeCN, microwave at 150 °C, 40 min, 9–12%; (for X = NR3) R2R3NH,
EtOH, microwave at 150 °C, 30 min, 4–64%.

C. M. Richardson et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1353–1357
1357
Lockie, A. M.; Murray, J. B.; Nunns, C. L.; Powles, J.;
Robertson, A.; Surgenor, A. E.; Torrance, C. J. Bioorg.
Med. Chem. Lett. 2005, 15, 863.
In conclusion, a template-hopping approach guided by
modelling and crystallographic data led to the rapid
identification and initial optimisation of a novel series
of CDK2 inhibitors. The relative activities of com-
pounds 5 and 6 clearly demonstrate the role which pro-
tein structure-guided approaches can perform in aiding
lead optimisation. Further optimisation afforded com-
pound 15, with a CDK2 IC50 of 120 nM and selectivity
over GSK3b of 167-fold.
11. Morley, S. D.; Afshar, M. J. Comput. Aided Mol. Des.
2004, 18, 189.
12. Pierce, A. C.; Sandretto, K. L.; Bemis, G. W. Proteins
2002, 49, 567.
13. Davies, T. G.; Bentley, J.; Arris, C. E.; Boyle, F. T.;
Curtin, N. J.; Endicott, J. A.; Gibson, A. E.; Golding, B.
T.; Griffin, R. J.; Hardcastle, I. R.; Jewsbury, P.; Johnson,
L. N.; Mesguiche, V.; Newell, D. R.; Noble, M. E. M.;
Tucker, J. A.; Wang, L.; Whitfield, H. J. Nat. Struct. Biol.
2002, 9, 745.
Acknowledgments
14. Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.;
Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E.
Nucleic Acids Res. 2000, 28, 235.
15. Humphrey, W.; Dalke, A.; Schulten, K. J. Mol. Graph.
1996, 14, 33, Available from http://www.ks.uiuc.edu/
Research/vmd/.
The authors thank Justin Bower for the re-synthesis of
NU-6102 (4), Peter Kierstan, Kate Grant, Joanne
Wayne and Angela Wheatley for running enzyme and
cell-based assays, Steven Athwal and Mhairi Donohoe
for production of CDK2, Adam Hold and Heather
Simmonite for analytical support and James Davidson
for his assistance in the preparation of this manuscript.
16. PDB entry codes for deposited structures: Compound 2,
˚
2c68 (1.95 A); compound 3, 2c69 (2.1 A); compound 4,
˚
˚
2c6o (2.1 A); compound 5, 2c6i (1.8 A); compound 6,
˚
˚
2c6m (1.9 A); compound 6 (cyclin A complex), 2c6t
˚
˚
(2.6 A); compound 11, 2c6k (1.9 A); compound 14, 2c6l
˚
(2.3 A).
References and notes
17. Knight, Z. A.; Shokat, K. M. Chem. Biol. 2005, 12, 621.
18. Makisumi, Y.; Watanabe, H.; Tori, K. Chem. Pharm. Bull.
1964, 12, 204.
1. Meijer, L.; Raymond, E. Acc. Chem. Res. 2003, 36, 417.
2. De Bondt, H. L.; Rosenblatt, J.; Jancarik, J.; Jones, H. D.;
Morgan, D. O.; Kim, S.-H. Nature 1993, 363, 595.
19. Gibson, A. E.; Arris, C. E.; Bentley, J.; Boyle, F. T.;
Curtin, N. J.; Davies, T. G.; Endicott, J. A.; Golding, B.
T.; Grant, S.; Griffin, R. J.; Jewsbury, P.; Johnson, L. N.;
Mesguiche, V.; Newell, D. R.; Noble, M. E. M.; Tucker, J.
A.; Whitfield, H. J. J. Med. Chem. 2002, 45, 3381.
20. Hardcastle, I. R.; Arris, C. E.; Bentley, J.; Boyle, T. F.;
Chen, Y.; Curtin, N. J.; Endicott, J. A.; Gibson, A. E.;
Golding, B. T.; Griffin, R. J.; Jewsbury, P.; Menyerol, J.;
Mesguiche, V.; Newell, D. R.; Noble, M. E. M.; Pratt, D.
J.; Wang, L.-Z.; Whitfield, H. J. J. Med Chem. 2004, 47,
3710.
´
3. Mendez, J. Cell 2003, 114, 398.
4. Pagano, M.; Jackson, P. K. Cell 2004, 118, 535.
5. Martin, A.; Odajima, J.; Hunt, S. L.; Dubus, P.; Ortega,
S.; Malumbres, M.; Barbacid, M. Cancer Cell 2005, 7, 591.
6. All kinases from Upstate Ltd.; http://www.upstate.com.
7. Kinase assays were adapted from the following generic
protocol, with ATP concentrations stated in the footnotes
to Table 1; Lane, M. E.; Yu, B.; Rice, A.; Lipson, K. E.;
Liang, C.; Sun, L.; Tang, C.; McMahon, G.; Pestell, R.
G.; Wadler, S. Cancer Res. 2001, 61, 6170.
21. Dreyer, M. K.; Borcherding, D. R.; Dumont, J. A.; Peet,
N. P.; Tsay, J. T.; Wright, P. S.; Bitonti, A. J.; Shen, J.;
Kim, S.-H. J. Med. Chem. 2001, 44, 524.
22. ter Haar, E.; Coll, J. T.; Austen, D. A.; Hsiao, H.-M.;
Swenson, L.; Jain, J. Nat. Struct. Biol. 2001, 8, 593, PDB
code 1I09.
23. In-house data.
24. MOE is available from the Chemical Computing Group;
http://www.chemcomp.com/.
25. Wildman, S. A.; Crippen, G. M. J. Chem. Inf. Comput.
Sci. 1999, 39, 868.
8. Skehan, P.; Storeng, R.; Scudiero, D.; Monks, A.;
McMahon, J.; Vistica, D.; Warren, J. T.; Bokesch, H.;
Kenney, S.; Boyd, M. R. J. Natl. Cancer Inst. 1990, 82,
1107.
9. GI50 is the inhibitor concentration which reduced tumour
cell growth by 50% over 72 h. HCT116 is a colon
carcinoma cell line, which is wild type with respect to
p53 and Rb.
10. Williamson, D. S.; Parratt, M. J.; Bower, J. F.; Moore, J.
D.; Richardson, C. M.; Dokurno, P.; Cansfield, A. D.;
Francis, G. L.; Hebdon, R. J.; Howes, R.; Jackson, P. S.;