Structure-based design of potent non-peptide MDM2 inhibitors

Article abstract of DOI:10.1021/ja051147z

A successful structure-based design of a class of non-peptide small-molecule MDM2 inhibitors targeting the p53-MDM2 protein-protein interaction is reported. The most potent compound 1d binds to MDM2 protein with a K_i value of 86 nM and is 18 ti

Full text of DOI:10.1021/ja051147z

Published on Web 06/30/2005
Structure-Based Design of Potent Non-Peptide MDM2 Inhibitors
Ke Ding,† Yipin Lu,† Zaneta Nikolovska-Coleska,† Su Qiu,† Yousong Ding,† Wei Gao,†
Jeanne Stuckey,‡ Krzysztof Krajewski,§ Peter P. Roller,§ York Tomita,| Damon A. Parrish,
Jeffrey R. Deschamps, and Shaomeng Wang*,†
Departments of Internal Medicine and Medicinal Chemistry and ComprehensiVe Cancer Center, and
Life Sciences Institute, UniVersity of Michigan, 1500 East Medical Center DriVe, Ann Arbor, Michigan 48109,
Laboratory of Medicinal Chemistry, National Cancer Institute-Frederick, National Institutes of Health, Frederick,
Maryland 21702, Lombardi Cancer Center, Georgetown UniVersity Medical Center, Washington, D.C. 20007,
The Laboratory for the Structure of Matter, NaVal Research Laboratory, 4555 OVerlook AVenue,
Washington, D.C. 20375
Received February 23, 2005; E-mail: shaomeng@umich.edu
The p53 tumor suppressor plays a central role in controlling cell
cycle progression and apoptosis1,2 and is an attractive cancer thera-
peutic target because its tumor suppressor activity can be stimulated
to eradicate tumor cells.1-3 A new approach to stimulation of the
activity of p53 is through blocking its interaction with the MDM2
oncoprotein, an endogenous cellular inhibitor of p53, using non-pep-
tide small-molecule MDM2 inhibitors.3,4 The design of non-peptide
small-molecule MDM2 inhibitors to block the p53-MDM2 interac-
tion is being intensively pursued as a new strategy for anti-cancer
drug design.3,4
Figure 1. Structure-based strategy to design a new class of MDM2
inhibitors.
Although potent, peptide-based small-molecule MDM2 inhibitors
have been designed to block the p53-MDM2 interaction,5 very
few non-peptide, small-molecule MDM2 inhibitors have been re-
ported to date.3 Most of them have weak or modest binding affin-
ities,3 highlighting the challenge in designing potent, non-peptide,
small-molecule MDM2 inhibitors. Recently, synthetic cis-imida-
zoline analogues (termed Nutlins) were reported as a class
of potent, non-peptide, small-molecule MDM2 inhibitors.4 In this
communication, we wish to report a successful structure-based
design of a new class of non-peptide MDM2 inhibitors based upon
the spiro-oxindole core structure.
Figure 2. Predicted binding model using computational docking for initial
lead compound 1a and for an optimized compound 1d. For 1a and 1d, car-
bon atoms are colored in light magenta, nitrogen atoms in blue, chloride atoms
in green, and oxygen atoms in red. For MDM2 binding site, carbon atoms
are colored in gray, nitrogen atoms in blue, oxygen atoms in red, and sulfur
atoms in yellow. Hydrogen bonds are depicted with a dashed yellow line.
The structural basis of the p53-MDM2 interaction has been
established by X-ray crystallography.6 The crystal structure shows
that the interaction between p53 and MDM2 is primarily mediated
by three hydrophobic residues (Phe19, Trp23, and Leu26) of p53
and a small but deep hydrophobic cleft in MDM2. This hydrophobic
cleft is an ideal site for designing small-molecule MDM2 inhibitors
that can block the p53-MDM2 interaction and is the target site in
our MDM2 inhibitor design.
Since the indole ring of Trp23 residue of p53 is buried deeply
inside a hydrophobic cavity in MDM2 and its NH group forms a
hydrogen bond with the backbone carbonyl in MDM2, Trp23
appears to be most critical for binding of p53 to MDM2. We have
therefore searched for chemical moieties that can mimic the
interaction of Trp23 with MDM2. In addition to the indole ring
itself, we have found that oxindole can perfectly mimic the side
chain of Trp23 for interaction with MDM2 (Figure 1).
a spiro-oxindole core structure (Figure 1). Our modeling studies
showed that although these compounds fit poorly into the MDM2
cleft due to steric hindrance, the spiro(oxindole-3,3′-pyrrolidine)
core structure may be used as the starting point for the design of a
new class of MDM2 inhibitors (Figure 1). The oxindole can closely
mimic the Trp23 side chain in p53 in both hydrogen-bonding forma-
tion and hydrophobic interactions with MDM2, and the spiropyr-
rolidine ring provides a rigid scaffold from which two hydrophobic
groups can be projected to mimic the side chain of Phe19 and
Leu26. We have designed candidate compounds using different R1,
R2, and R3 groups with different configurations and docked them
into the MDM2 binding cleft using the GOLD program.7 Our
docking studies showed that compound 1a in Figure 1 closely
mimics p53 in its interaction with MDM2 (Supporting Information)
and predicted that 1a should have a good affinity to MDM2. The
6-chloro substituent on the oxindole ring in 1a occupies a smaller
hydrophobic cavity in MDM2 and was shown to improve the
binding affinity of peptide-based MDM2 inhibitors.5
Since many anticancer drugs are natural products or derivatives
of natural products, we used substructure search techniques to
identify natural products that contain an oxindole ring. Among the
natural products we have identified, there were a number of natural
alkaloids such as spirotryprostatin A and Alstonisine which contain
† Departments of Internal Medicine and Medicinal Chemistry and Comprehen-
sive Cancer Center, University of Michigan.
‡ Life Sciences Institute, University of Michigan.
§ National Institutes of Health.
The designed compound 1a was synthesized using an asymmetric
1,3-dipolar reaction8 as the key step (Scheme 1). The absolute
| Georgetown University Medical Center.
Naval Research Laboratory.
9
10130
J. AM. CHEM. SOC. 2005, 127, 10130-10131
10.1021/ja051147z CCC: $30.25 © 2005 American Chemical Society

C O M M U N I C A T I O N S
studies predicted that both 1e and 1f should be less potent than 1d,
and in fact, FP-based binding experiments showed that 1e and 1f
with Ki values of 0.65 and 0.39 µM, respectively, are substantially
less potent than 1d.
One major advantage of non-peptide MDM2 inhibitors over
peptide-based inhibitors is their superior cell permeability.3,4 Based
upon their mode of action,1,3,4 it is predicted that a potent, non-
peptide small-molecule MDM2 inhibitor will be effective in inhibi-
tion of cell growth in cancer cells with wild-type p53 and will have
a weaker activity in cancer cells with either mutated or deleted
p53.4 We evaluated our designed MDM2 inhibitors in p53 wild-
type LNCaP human prostate cancer cells9 for their ability to inhibit
cell growth and found that, as predicted, the potent MDM2 inhibitor
1d is a highly effective inhibitor of cell growth, with an IC50 value
of 0.83 µM. Compounds 1a, 1b, 1c, 1e, and 1f inhibit cell growth
with IC50 values of 9.7, 2.1, 6.7, 2.7, and 1.9 µM, respectively. It
is significant that their activities in inhibition of cell growth in
LNCaP cells correlate almost perfectly with their binding affinities
to MDM2.
Figure 3. Competitive binding curves and Ki values of inhibitors to MDM2
as determined using a FP-based binding assay.
Scheme 1. Synthesis of Compound 1a and Other Designed
MDM2 Inhibitorsa
a Reagents and conditions: (a) 4 Å molecular sieves, toluene, 70 °C;
(b) dimethylamine/THF, room temperature; (c) Pd(OAc)4, CH2Cl2-MeOH
(1:1), 0 °C.
The cellular selectivity of these compounds was evaluated in
human prostate cancer PC-3 cells with a deleted p53.9 Consistent
with our predictions, these MDM2 inhibitors are much less potent
in PC-3 cells than in LnCaP cells (Supporting Information). For
example, 1d has an IC50 value of 22.5 µM in PC-3 cells, 27 times
less than that in LNCaP cells.
One potential concern in the development of MDM2 inhibitors
as new anti-cancer drugs is that such inhibitors may be equally
toxic to normal cells with wild-type p53. We thus evaluated 1d in
normal human prostate epithelial cells9 with wild-type p53 and
determined that 1d has an IC50 value of 10.5 µM in inhibition of
cell growth, 13 times less toxic than to LNCaP cancer cells. This
shows that our designed MDM2 inhibitors have good selectivity
between cancer and normal cells with wild-type p53.
stereochemistry of 1a and other designed analogues was determined
by X-ray crystallographic analysis of one of the key intermediates
5 (Supporting Information).
To determine the ability of 1a to disrupt the interaction between
MDM2 and p53, we have established a fluorescence polarization-
based (FP-based) binding assay using a recombinant human MDM2
protein and a p53-based peptide5 labeled with a fluorescent tag
(Supporting Information). This fluorescently labeled p53-based
peptide has a Kd value of 1 nM to MDM2 (Supporting Information),
consistent with its previously reported high-affinity for MDM2.5
A natural p53 peptide (residues 16-27) was determined to have a
Ki value of 1.52 µM (Figure 3) in this binding assay. Compound
1a has a Ki value of 8.46 µM (Figure 3) in our FP-based assay and
thus represents a novel lead compound.
In summary, we have described a successful structure-based
design of a novel class of potent, non-peptide small-molecule
MDM2 inhibitors to target the p53-MDM2 interaction. Our studies
provide a convincing example that a structure-based strategy can
be employed to design highly potent, non-peptide, small-molecule
inhibitors to target protein-protein interaction, which remains a
very challenging area in chemical biology and drug design. Further
optimization of this class of promising MDM2 inhibitors may
ultimately lead to the development of an entirely new type of
anticancer drugs.
Supporting Information Available: An Experimental Section
including information on the synthesis and chemical data for 1a-1f,
molecular modeling methods and results for 1a-1f, experimental
procedure for the fluorescence polarization-based binding assay, and
details on the cellular growth inhibition assay and results. This material
is available free of charge via the Internet at http://pubs.acs.org.
Analysis of the predicted binding model of 1a to MDM2
(Supporting Information) suggests that 1a can be further optimized
for its interaction with MDM2. For example, its phenyl ring binds
to the hydrophobic binding pocket occupied by the side chain of
Phe19, but there is additional room available. Similarly, its isobutyl
group fills in the hydrophobic binding pocket occupied by Leu26,
but a slightly larger hydrophobic group could be accommodated.
We have therefore designed new analogues of 1a to optimize further
the interactions at these two hydrophobic binding sites.
Modeling studies predicted that introduction of a chlorine atom
at the meta-position to the phenyl ring in 1a can occupy the
additional room available at this binding site and effectively improve
the hydrophobic interaction, while a chlorine atom at either the
para- or the ortho-position of the phenyl ring in 1a is less optimal
(Supporting Information). Compound 1b with an m-Cl substituent
was synthesized (Scheme 1) and found to have a Ki value of 300
nM, 28 times more potent than 1a. To further confirm our
prediction, 1c with a p-Cl substituent was synthesized and found
to be 26 times less potent than 1b with a Ki value of 7.68 µM.
As indicated above, the isobutyl group in 1a is less than optimal,
and we further optimized the hydrophobic interaction at this site
using 1b as the template. Modeling studies predicted that a 2,2-
dimethylpropyl group should enhance the hydrophobic interaction
(Figure 2 and Supporting Information). The resulting compound
1d containing such a 2,2-dimethylpropyl group has a Ki value of
86 nM in our FP-based assay and is thus 98 times more potent
than the initial lead compound 1a.
References
(1) Vogelstein B.; Lane, D.; Levine, A. J. Nature 2000, 408, 307.
(2) Wu, X.; Bayle, J. H.; Olson, D.; Levine, A. J. Genes DeV. 1993, 7, 1126.
(3) Chene, P. Nat. ReV. Cancer 2003, 3, 102.
(4) Vassilev, L. T.; Vu, B. T.; Graves, B.; Carvajal, D.; Podlaski, F.; Filipovic,
Z.; Kong, N.; Kammlott, U.; Lukacs, C.; Klein, C.; Fotouhi, N.; Liu, E.
A. Science 2004, 303, 844.
(5) Garcia-Echeverria, C.; Chene, P.; Blommers, M. J. J.; Furet, P. J. Med.
Chem. 2000, 43, 3205.
(6) Hussie, P. H.; Gorina, S.; Marechal, V.; Elenbaas, B.; Moreau, J.; Levine,
A. J.; Pavletich, N. P. Science 1996, 274, 948.
(7) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R. J. Mol. Biol.
1997, 267, 727.
(8) Sebahar, P. R.; Williams, R. M. J. Am Chem. Soc. 2000, 122, 5666.
To further confirm the importance of the hydrophobic interaction
at this site, we have synthesized 1e and 1f with, respectively, a
hydrophobic group smaller or larger than that in 1d. Modeling
(9) Fan, R.; Kumaravel, T. S.; Jalali, F.; Marrano, P.; Squire, J. A.; Bristow,
R. G. Cancer Res. 2004, 64, 8526.
JA051147Z
9
J. AM. CHEM. SOC. VOL. 127, NO. 29, 2005 10131