CDK2/cyclinA inhibitors: Targeting the cyclinA recruitment site with small molecules derived from peptide leads

Article abstract of DOI:10.1016/j.bmcl.2005.12.004

The syntheses of potent small molecule inhibitors of the CDK2/cyclinA recruitment site are described. Structure-activity trends of nanomolar octapeptides were examined through amino-acid substitution and truncation of the sequence resulting in the identif

Full text of DOI:10.1016/j.bmcl.2005.12.004

Bioorganic & Medicinal Chemistry Letters 16 (2006) 1716–1720
CDK2/cyclinA inhibitors: Targeting the cyclinA recruitment
site with small molecules derived from peptide leads
Georgette Castanedo,a Kevin Clark,a Shumei Wang,a Vickie Tsui,a Mengling Wong,a
John Nicholas,a Dineli Wickramasinghe,b James C. Marsters, Jr.a and Daniel Sutherlina,*
aDepartment of Medicinal Chemistry, Genentech Inc., 1 DNA Way, South San Francisco, CA 94080, USA
bDepartment of Molecular Oncology, Genentech Inc., 1 DNA Way, South San Francisco, CA 94080, USA
Received 11 October 2005; revised 23 November 2005; accepted 1 December 2005
Available online 27 December 2005
Abstract—The syntheses of potent small molecule inhibitors of the CDK2/cyclinA recruitment site are described. Structure–activity
trends of nanomolar octapeptides were examined through amino-acid substitution and truncation of the sequence resulting in the
identification of a smaller, albeit significantly less potent, tetrapeptide lead. These losses in affinity were recovered by side-chain opti-
mization and by rigidification of the peptide backbone using a combination of solid-phase parallel synthesis and structure-based
design. Finally, two guanidine functionalities were replaced to improve drug-like properties, resulting in neutral small molecules
equal in activity to that of the peptide lead.
Ó 2005 Elsevier Ltd. All rights reserved.
Cell-cycle progression requires the coordinated interac-
tion and activation of cyclins and cyclin-dependent ki-
nases (CDKs). During S-phase, cyclinA (CycA) binds
to CDK2 and activates the kinase as well as recruits sub-
strates and inhibitors to the complex through a hydro-
act as a selective anti-cancer agent. The few reported
studies of recruitment site inhibitors have primarily fo-
cused on the SAR of endogenous peptide substrates or
inhibitors of cyclinA as a guide for future small molecule
design.4 This communication describes the optimization
˚
phobic groove that is approximately 50 A away from
of a peptide lead to a potent small molecule
(IC50 = 6 nM).
the CDK2 active site.1 Blockage of this CycA recruit-
ment site would allow a more selective approach to
CDK inhibition than the standard active-site antago-
nists. The transcription factor E2F is a substrate for
the CDK2/CycA complex and is phosphorylated as a
prelude to S-phase exit. Cellular response to CDK2/
CycA inhibition is thought to be greater in tumor versus
normal tissue since E2F is commonly deregulated in tu-
mor cells and has been reported to result in cell death in
combination with CDK2/CycA inactivation.2 Therefore,
selective inhibition of this phosphorylation event should
block S-phase exit and sensitize cancer cells to apopto-
sis. We have tested the consequence of cyclinA inhibi-
tion and validated this approach by using cell
penetrating peptides that block the recruitment site on
cyclinA and selectively kill cancer cells in vitro and in vi-
vo.3 A small molecule derived from these proof-of-con-
cept peptides with improved drug-like properties could
Endogenous substrates and inhibitors of the CDK2/
CycA recruitment site have the consensus sequence
RXL and require a second C-terminal hydrophobic ami-
no acid either directly after the leucine or separated by
an additional residue (p27-RNLF, E2F1-RRLDL).
Crystal structures of peptides derived from these
sequences bound in this site illustrate a shallow pro-
tein–protein interface.5 Peptide 1 (PVKRRLFG, Table
1)6 was used as our starting point and was tested in a
CDK2/CycA ELISA using the retinoblastoma protein
(Rb) as a substrate (IC50 = 12 nM).7 A peptide library
was prepared to evaluate the contribution of each amino
acid side chain. Alanine scanning of peptide 1 showed
the most significant loss in activity for leucine and phen-
ylalanine, and arginine replacement (peptides 5, 7, and
8) consistent with the consensus sequence. The remain-
ing residues had little or no effect on binding.
Keywords: CDK2/cyclinA; Protein–protein interaction inhibitor;
Guanidine replacement.
Truncation of the peptide from the N-terminus to the
consensus sequence caused a 120-fold reduction in
potency (peptides 10–12). Truncation of the C-terminal
*
Corresponding author. Tel.: +1 650 225 3171; fax: +1 650 225
2061; e-mail: dans@gene.com
0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.12.004

G. Castanedo et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1716–1720
1717
Table 1. Peptide sequence length and side-chain contributions to
binding
Compound Sequence
IC50 (nM) Fold diff.
1
2
PVKRRLFG
AVKRRLFG
PAKRRLFG
PVARRLFG
12
16
—
1.3
0.8
5.2
3
10
62
4
5
PVKARLFG
PVKRALFG
1,600
130
22,000
8200
10
130
6
11
1800
680
7
PVKRRAFG
PVKRRLAG
PVKRRLFA
Scheme 1. Reagents and conditions: To prepare compound 17—Rink
resin, (a) HOBt, HBTU, DIPEA, DMF, 4 h; then 20% pip/DMF,
10 min; (b) N,N-bis(t-butoxycarbonyl)-1H-pyrazole-1-carboxamidine,
DIPEA, DMF; (c) TFA; (d) TMSNCO, DIPEA. Pbf is 2,2,4,6,7-
pentamethyldihydrobenzofuran-5-sulfonyl.
8
9
0.8
10
11
12
13
14
15
16
VKRRLFG
KRRLFG
RRLFG
120
770
1,400
14
10
64
120
PVKRRLF
1.2
2.3
PVKRRLF-NH2
PVKRRL-(3-Cl)F-NH2
(desNH2)-RRL-(3-Cl)F-NH2
28
0.6
3000
0.05
250
glycine, resulting in either the phenylalanine acid 13 or
amide 14, was well tolerated. Peptide libraries were pre-
pared using a variety of templates with natural, unnatu-
ral, L, and D amino acids to probe the binding surface.
Most analogs resulted in loss of activity to varying de-
grees, while substitution on the aryl ring of phenylala-
nine met with the greatest success. The substitution of
phenylalanine with 3-chlorophenylalanine, illustrated
in peptide 15, resulted in a 10-fold improvement in
potency relative to peptide 14. These data are consistent
with previously reported SAR with RXLIF peptides,
where 4-halophenylalanines occupy a similar location
in the binding pocket.5c The lead peptide 16 resulted
from a combination of C- and N-terminal truncations
along with the 3-chlorophenyl alanine substitution and
lacking the N-terminal amine (des-amino arginine for
arginine). This 3 lM peptide was significantly less po-
tent than the lead peptide 1 but represented a more pro-
ductive starting point for medicinal chemistry
development due to its lower molecular weight and
smaller size.
Figure 1. Compounds 19. N-terminal modifications and CDK2/CycA
binding.
aryl-carbonyl rotational barrier (compound 19e), or by
changing the benzoic acid to a phenyl acetic acid (com-
pound 19f). When molecular models and, later, crystal
structures developed from compounds similar to 19a
were examined, it became apparent that the guanidine
group in the GMBA was not likely to contact cyclinA
in the same location as the N-terminal guanidine of
the full-length peptides. Rather, the para-substituted
benzoic acid directs the guanidine to a separate region
in the binding site that also has negative electrostatic po-
tential, which was previously occupied by the backbone
of the full-length peptides.
It was hypothesized that the binding energy of 16 could
be increased through optimization of favorable interac-
tions with the protein surface and the preorganization of
the compound to the bound conformation. Resin-bound
3-chlorophenylalanine 17 was used in standard peptide
synthesis techniques to prepare analogs 19a–g (Scheme
1). The N-terminal flexible alkyl chain of the arginine
mimetic 16 could be replaced with a more rigid aryl
group to give compound 19a. The 4-fold increase in
activity upon incorporation of the para-guanidinometh-
ylbenzoic acid (GMBA) in 19a was consistently ob-
served throughout several scaffolds (Fig. 1). Placing
the guanidinomethyl substituent in the meta position,
compound 19b, was less effective but was equipotent
with the straight chain compound 16. The importance
of the guanidine functionality in 19a was illustrated by
the urea 19c, which was more than 27-fold less potent,
and the benzyl amine 19d, 5-fold less potent. No
improvement in activity was obtained through ortho
substitution of the aryl ring, designed to perturb the
Having had some success in improving binding affinity
by reducing conformational freedom, a similar ap-
proach was undertaken at the C-terminus. In order to
prepare analogs, the protected intermediate 21 was pre-
pared on chlorotrityl resin starting from resin-bound
leucine 20 (Scheme 2). This advanced intermediate was
then used in amide bond couplings with a variety of
amines to prepare analogs 22. The C-terminal amide
in compound 16 could easily be replaced with hydrogen,
illustrated by phenethylamine 22a (Fig. 2). Replacement

1718
G. Castanedo et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1716–1720
Table 2. Structure–activity relationships for compounds 26
Compound
R
X
IC50 (nM)
Scheme 2. Reagents and conditions: to prepare compound 20 use 2-
chlorotrityl resin, Fmoc-leucine, DIPEA, CH2Cl2, 4 h; then 20% pip/
DMF, 10 min; (a) HOBt, HBTU, DIPEA, NMP, 4 h; then 20% pip/
DMF, 10 min; (b) N,N-bis(t-butoxycarbonyl)-1H-pyrazole-1-carbox-
amidine, DIPEA, DMF; (c) 10% AcOH, 10% CF3CH2OH, CH2Cl2,
2 h; (d) i. polystyrene-carbodiimide, polystyrene-HOBt, 10% DMF in
CH2Cl2, amine, 18 h; ii. TFA.
26a
26b
26c
26d
26e
Guanidine
Guanidine
Guanidine
Guanidine
Guanidine
Arg
Nle
Ala
dAla
Pro
47
31
58
31,000
3,000
26f
Ala
Ala
251
790
26g
26h
26i
Ala
21
Abu
6.0
Figure 2. Compounds 22. C-terminal modifications of the truncated
peptide and CDK2/CycA binding.
of the chlorine in 22a with either a fluorine or trifluoro-
menthyl group was not effective, giving compounds with
IC50 values of 18 and 42 lM, respectively (structures not
shown). Potency was improved 20-fold by locking the
chloro-phenyl group and the amide functionality into
a gauche arrangement by replacing the phenethylamine
with a trans-2-arylcyclohexyl amine, giving compound
22b. The dihedral angle between substituents of the
1,2-trans-disubstituted cyclohexane in 22b is identical
to the N-Ca-Cb-Ph dihedral angle in reported structures
of full-length RXLF peptides. The importance of the
absolute stereochemistry is illustrated by diastereomer
22c, which was more than 45-fold less potent. In order
to further improve potency of compounds, combina-
tions of these C-terminal and N-terminal modifications
were pursued while also exploring the SAR of the inter-
nal arginine residue.
Scheme 3. Reagents and conditions: (a) i. NaHCO3, EtOH, 4 h, 99%;
ii. Zn dust, AcOH, MeOH, 100%; iii. EDC, Boc-leucine, CH2Cl2; iv.
HCl; (b) i. EDC, Boc amino acid X, CH2Cl2; ii. HCl; (c) i. EDC, Boc-
4-(aminomethyl)benzoic acid; ii. HCl; iii. N,N-bis(t-butoxycarbonyl)-
1H-pyrazole-1-carboxamidine, DIPEA, DMF; (d) i. EDC, [4-(N-Boc-
piperidinyl)methyl]-4-benzoic acid, CH2Cl2; ii. HCl (e) i. EDC,
a-bromo-para-toluic acid, CH2Cl2; ii. 1-methyl-3-aminopyrazole or
2-aminothiazole, DMSO, 5 min, 130 °C; (f) NBS, benzoyl peroxide,
5% AcOH in benzene, 130 °C, 5 min, 54%; (g) i. EDC, CH2Cl2; ii.
2-aminothiazole, DMSO, 5 min, 130 °C.
to give dipeptide 24. This intermediate was further elabo-
rated with one variable amino acid (indicated as X in
Scheme 3 and Table 2) to give compounds 25. From this
point the synthetic routes diverged in to access com-
pounds containing guanidines 26a–e or other functional-
ities 26f–i.
Given the success seen with the 3-chlorophenylcyclohexyl
amine, all newly synthesized compounds were prepared
using this functionality. Compounds (26a–i, Table 2)
were prepared by base equilibration of 1-chloro-3-
((1S,2S)-2-nitrocyclohexyl)benzene 238 (Scheme 3) to
the trans stereochemistry followed by reduction of the ni-
tro group using zinc and acetic acid (conditions used to
preserve the chloro-arene functionality). The resultant
amine was coupled to Boc-leucine and then deprotected
When the constrained 3-chlorophenylcyclohexyl amine
was used along with the N-terminal GMBA, a 5-fold
improvement in binding was observed versus 22b to give
47 nM compound 26a (Table 2). The internal amino

G. Castanedo et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1716–1720
1719
acid ‘‘X’’ mainly served a structural purpose, illustrated
by the substitution of the arginine in equipotent
compound 26a to nor-leucine 26b or even the compact
alanine 26c. This observation was in contrast to the
10-fold loss in affinity observed when the arginine to
alanine substitution was effected in the full-length
peptide. The importance of the stereochemistry at this
residue is demonstrated by compound 26d, where the
natural L-alanine has been exchanged with a D-alanine,
resulting in a >500-fold loss in affinity. Substitution with
proline (compound 26e) also resulted in a compound
with poor affinity.
carbonyl and the protein surface. We reasoned that a
ketone would be an ideal substitution for this amide
and searched for available starting materials that would
position a carboxamide and an aryl ketone in the
appropriate geometry. trans-2-(4-Methylbenzoyl)cyclo-
pentane-1-carboxylic acid 27 (Scheme 3) was incorpo-
rated into the optimized scaffold to yield the 40 nM
compound 29 (Scheme 3, Fig. 3). This inhibitor was very
close in potency to the alanine derivative 26h, in contrast
to the inactive proline compound 26e. A model of this
compound superimposed onto a crystal structure of
the peptide fragment RNLF (Fig. 3) shows the ketone
oxygen in the appropriate geometry to accept a hydro-
gen bond from glutamine 254 in CycA.9 The synthesis
of 29 (Scheme 3) was initiated by bromination of
racemic acid 27 and then coupling the resultant bromo-
acid 28 with peptide 24. It was at this point that the
amino-thiazole was installed and the diastereomers were
separated by reversed-phase HPLC. The diastereomer of
29 was inactive in the kinase ELISA. The stereochemical
assignment of these diastereomers was inferred from the
biochemical activities and through a comparison of both
compounds modeled into the active site.
Having replaced the internal arginine with the neutral
and smaller alanine, the remaining guanidine functional-
ity at the N-terminus was targeted for replacement. Sub-
stitutions retaining the positive charge in the general
vicinity of the guanidine were explored and led to piper-
idine 26f. We also surveyed weakly basic heterocycles
that could be protonated once in proximity to the anion-
ic region on the protein surface. This strategy was espe-
cially successful when an aminothiazole was used in
place of the guanidine, yielding a 21 nM analog 26h.
The activity of this compound was further improved 3-
fold via an effective one-carbon homologation of the
alanine to a-aminobutyric acid giving compound 26i.
Many of the compounds detailed in Table 2 demon-
strate that it is possible to obtain lower molecular
weight compounds with equivalent or better affinity
for cyclinA than natural octapeptides, which have pre-
viously been shown to selectively kill tumor cells when
combined with cell penetrating sequences. The rigidifi-
cation and preorganization of the side chains as well as
improving hydrophobic contacts in the binding site
were responsible for the dramatic increases in potency
when compared to the simple truncated peptide leads.
Namely, the meta-chlorophenyl, cyclohexylamine, and
the GMBA moieties accounted for significant improve-
ments in activity.
Compound 26i represents a significant advance over the
truncated lead peptide 16 in that it is 500-fold more po-
tent, contains fewer rotatable bonds, and is neutral at
physiological pH. Anticipating that the peptidic nature
of the compound may affect metabolism and oral
adsorption, we sought to remove amide bonds from
the molecule. Examination of crystal structures indicat-
ed a key hydrogen bond between the N-terminal amide
Acknowledgments
The authors thank Jenny Stamos for protein production
and purification and Christian Wiesmann for
crystallography.
References and notes
1. Russo, A. A.; Jeffrey, P. D.; Patten, A. K.; Massague, J.;
Pavletich, N. P. Nature 1996, 382, 325.
2. Lees, J. A.; Weinberg, R. A. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 4221.
3. (a) Chen, Y.-N. P.; Sharma, S. K.; Ramsey, T. M.; Jiang,
L.; Martin, M. S.; Baker, K.; Adams, P. D.; Bair, K. W.;
Kaelin, W. G. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 4325;
(b) Mendoza, N.; Fong, S.; Marsters, J.; Koeppen, H.;
Schwall, R.; Wickramasinghe, D. Cancer Res. 2003, 63,
1020.
4. (a) Sharma, S. K.; Ramsey, T. M.; Chen, Y.-N. P.; Chen,
W.; Martin, M. S.; Clune, K.; Sabio, M.; Blair, K. W.
Bioorg. Med. Chem. Lett. 2001, 11, 2449; (b) Atkinson, G.
A.; Cowan, A.; McInnes, C.; Zheleva, D. I.; Fisher, P. M.;
Chan, W. C. Bioorg. Med. Chem. Lett. 2002, 12, 2501; (c)
McInnes, C.; Andrews, M. J. I.; Zheleva, D. I.; Lane, D. P.;
Figure 3. Molecular model of small molecule 29 in green bound to
cyclinA and overlapped onto the crystal structure of a peptide
containing the sequence RNLF in gray (PDB ID:1H27, Ref. 5b).

1720
G. Castanedo et al. / Bioorg. Med. Chem. Lett. 16 (2006) 1716–1720
Fisher, P. M. Curr. Med. Chem. Anti-Cancer Agents 2003,
kinase reaction buffer (25 mM Tris-buffered saline [pH 7.5]
with 10 mM MgCl2, 5 mM b-glycerophosphate, 2 mM
DTT, 0.1 mM sodium orthovanadate, and 0.005% Nonidet
P-40). Final concentrations of ATP and CDK2 were
200 lM and 300 pM, respectively. Compounds were diluted
in 10% DMSO in kinase reaction buffer. ATP was added to
each well of the kinase reaction plate, followed in sequence
by GST-Rb and test compound serial dilutions. The kinase
reaction was started with the addition of CDK2/CycA to
the kinase reaction plate and stopped with kinase stop
buffer (50 mM EDTA in ELISA buffer). The ELISA plates
were washed with ELISA wash buffer (50 mM Tris-buffered
saline [pH 7.5] with 0.05% Tween 20 and 1 mM MnCl2).
The reactions were diluted 1:10 into kinase stop buffer in
the ELISA plate. Phosphorylated Rb was detected with
rabbit anti-phospho-Rb (Ser795) followed by (HRP)-linked
IgG anti-rabbit antibody. The plate was developed with
tetramethylbenzidine as substrate and the absorbance was
measured at 450 nm.
3, 57; (d) Zheleva, D. I.; McInnes, C.; Gavine, A.-L.;
Zhelev, N. Z.; Fisher, P. M.; Lane, D. P. J. Peptide Res.
2002, 60, 257.
5. (a) Brown, N. R.; Noble, M. E. M.; Endicott, J. A.;
Johnson, L. N. Nat. Cell Biol. 1999, 1, 438; (b) Lowe, E. D.;
Tews, I.; Cheng, K. Y.; Brown, N. R.; Gul, S.; Noble, M. E.
M.; Gamblin, S. J.; Johnson, L. N. Biochemistry 2002, 41,
15634; (c) Kontopidis, G.; Andrews, M. J. I.; Cowan, A.;
Powers, H.; Innes, L.; Plater, A.; Griffiths, G.; Paterson, D.;
Zheleva, D. I.; Lane, D. P.; Green, S.; Walkinshaw, M. D.;
Fisher, P. M. Structure 2003, 11, 1537.
6. (a) Adams, P. D.; Sellers, W. R.; Sharma, S. K.; Wu, A. D.;
Nalin, C. M.; Kaelin, W. G. Mol. Cell. Biol. 1996, 16, 6623;
(b) Takeda, D. Y.; Wohlschlegel, J. A.; Dutta, A. J. Biol.
Chem. 2001, 276, 1993; (c) Wohlschlegel, J. A.; Dwyer, B.
T.; Takeda, D. Y.; Dutta, A. Mol. Cell. Biol. 2001, 21, 4868.
7. 96-well high-binding microtiter ELISA plates were coated
overnight at 4 °C with mouse anti-GST in PBS. The plates
were decanted and blocked with ELISA buffer (0.05%
bovine c-globulins [BGG] in 50 mM Tris-buffered saline
[pH 7.5] with 0.05% Tween 20 and 0.1 mM sodium
orthovanadate). The kinase reaction was performed in 96-
well non-binding microtiter plates. ATP, CDK2/CycA (HIS
tagged and phospho T160), and GST-Rb were diluted in
8. Hayashi, T.; Senda, T.; Ogasawara, M. J. Am. Chem. Soc.
2000, 122, 10716.
9. The model was built from co-crystal structures of other
peptides and peptidomimetics, and minimized in the
CyclinA binding site using Schrodinger’s Macromodel
program.