Identification of Potent Ras Signaling Inhibitors by Pathway-Selective Phenotype-Based Screening

Article abstract of DOI:10.1002/anie.200352587

Blocking the path to cancer: Application of efficient strategy for a phenotype-based screening of compound libraries to a library of sulindac-derived compounds (see scheme) yielded new inhibitors of the tumor-relevant Ras signal transduction pathway with high fidelity (as shown by microscopy). The results underline the advantage of using biologically prevalidated compound classes in chemical biology research.

Full text of DOI:10.1002/anie.200352587

Communications
Drug Discovery I
Identification of Potent Ras Signaling Inhibitors
by Pathway-Selective Phenotype-Based
Screening**
Oliver Müller, Eleni Gourzoulidou,
Mercedes Carpintero, Ioanna-Maria Karaguni,
Anette Langerak, Christian Herrmann, Tarik Möröy,
Ludger Klein-Hitpaß, and Herbert Waldmann*
Phenotype-based screening of compound libraries utilizes
chemistry and biology together at a very early phase of a
chemical biological research program and identifies both
potential biological targets and hit compounds in one step.[1]
The power of the combination of this approach with modern
chemistry and biology research techniques was recently
demonstrated in the context of chemical genetics research.[2]
[*] Dipl.-Chem. E. Gourzoulidou, Dr. M. Carpintero, C. Herrmann,
Prof. Dr. H. Waldmann
Max-Planck-Institut für molekulare Physiologie
Abteilung Chemische Biologie
Otto-Hahn-Strasse 11, 44227 Dortmund (Germany)
and
Universität Dortmund
Fachbereich 3, Organische Chemie
Fax: (+49)231-133-2499
E-mail: herbert.waldmann@mpi-dortmund.mpg.de
Priv.-Doz. Dr. Dr. O. Müller, I.-M. Karaguni, A. Langerak
Max-Planck-Institut für molekulare Physiologie
Abteilung Strukturelle Biologie, Dortmund (Germany)
Prof. Dr. T. Möröy, Priv.-Doz. Dr. L. Klein-Hitpaß
Institut für Zellbiologie (Tumorforschung)
Universitätsklinikum Essen
Virchowstrasse 173, 45122 Essen (Germany)
[**] We thankAlfred Wittinghofer for his support. This study was
supported by the Fonds der Chemischen Industrie and the Euro-
pean Union (Marie Curie Fellowship to M.C.).
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
450
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
DOI: 10.1002/anie.200352587
Angew. Chem. Int. Ed. 2004, 43, 450 –454

Angewandte
Chemie
The use of phenotypic change in entire organisms or
unbiased cellular systems as a readout may render the
identification of the biological target a complicated and
laborious endeavor. The study of the reversal or change of the
phenotype of a well-defined cellular system carrying known
genomic lesions that directly correlate with a particular
change in the microscopic phenotype should significantly
improve the process of target identification. This improve-
ment should be particularly evident if the genomic lesion
affects a protein that acts at a prominent position of a
multifaceted and branched signal transduction pathway.
Pathway selectivity can be ensured by counterscreening two
cell lines carrying lesions in different signaling pathways. In
addition, an increase in efficiency can be expected if the
compound class to be screened can be regarded as biologically
prevalidated and if links from this compound class to the
pathways of interest are known. The particular organism or
cell type chosen should allow conclusive identification of the
biological target while leaving room for identification of
additional target proteins. Ideally, the biologically prevali-
dated compound classes should be based on “privileged
structures,” that is, structures that mediate binding to multiple
biological targets.[3] If these prerequisites are fulfilled, one can
expect that even a fairly small compound library will deliver
hits with an appreciable rate.[4] Hits from such screens are
expected to be of high quality because they have already been
through a first round of biological validation. In addition, the
library might yield hits for more than one target in a given
pathway of interest, and possibly for targets in all pathways
screened.
Herein we report the validation of this approach by
employment of a phenotypic screen with a cell line activated
in the Ras pathway for primary screening, and a cell line
activated in the Wnt pathway for counterscreening (Figure 1).
The mutationally activated proto-oncoprotein Ras plays a
significant role in the establishment of more than 30% of all
human tumors. The development of small molecules that
interfere with Ras signaling activity is a promising approach
that is being pursued intensively in anticancer drug research.[5]
Antagonization of the interaction between Ras and its
downstream effector Raf in the Ras signaling pathway poses a
particularly challenging and difficult problem since this
prototypical protein–protein interaction covers a large sur-
face area.[5] Despite the attempts made in various intensive
screening studies (including high-throughput screening
approaches), inhibitors of the Ras–Raf interaction could
only be identified in two cases.[6,7] In one of these cases, the
metabolite 2 of the nonsteroidal antiinflammatory drug
(NSAID) sulindac (1; see Table 1) was shown to be a Ras–
Raf inhibitor. A compound derived from 1 was found to
interfere with the Ras pathway as well, although its molecular
target remained unidentified.[6]
Figure 1. Schematic representation of the Ras (left) and the Wnt
(right) pathways. Both pathways start outside the cell with the binding
of a signaling protein to a transmembrane receptor. In the case of Ras,
this binding leads, through a so-called exchange factor, to the activa-
tion of the Ras protein, which itself activates several kinases through
the effector Raf kinase. The activation of the Wnt receptor leads to the
activation, that is, the negative inhibition of the key signaling protein
b-catenin through the action of several kinases and the tumor suppres-
sor protein adenomatous polyposis coli (APC). The two pathways are
formed by different proteins and are based in part on different bio-
chemical principles. Nevertheless, both pathways lead to similar cellu-
lar effects: activation of target gene expression and transformation,
that is, the transition from a normal cell into a tumor cell. By using
suitable cell-culture models such as those used in the study described
herein, cell transformation can be visualized by light microscopy.
compound library with the underlying structure of this
NSAID as a guiding scaffold. We analyzed the library for
potential inhibitors of the Ras signal transduction pathway by
using the phenotype-based strategy delineated above as the
primary screen.
The sulindac-derived compound library was synthesized
on a solid support as shown in Scheme 1. Variously sub-
stituted indenylacetic acids 3[9] were attached to 2-chlorotrityl
chloride resin (obtained from Calbiochem–Novabiochem,
loading 1.08 mmolgÀ1, or from CBL–Patras, loading
1.6 mmolgÀ1). Application of the Wang linker led to unde-
sired cleavage of the indenylacetic acids under the basic
conditions of the subsequent steps (see below). Resin-bound
intermediates 4 were subjected to a Knoevenagel condensa-
tion with aromatic aldehydes to yield 188 sulindac analogues
5 in overall yields of 47–98% (calculated based on the initial
loading of the resin) after release from the resin under acidic
conditions. Critical to the successful execution of the Knoe-
venagel reaction is the choice of the base, which must be
strong enough to mediate the condensation yet must not
induce cleavage of the ester. Use of (iPr)2NEt or piperidine in
DMF or toluene at up to 1008C did not induce condensation.
In the presence of sodium or potassium tert-butoxide the
condensation proceeded, but the ester was also cleaved. DBU
was used as the base of choice. In the presence of 10 equiv-
alents DBU in DMF or toluene at 608C the desired products
are formed without unwanted cleavage of the ester. The
library was further expanded by subjecting immobilized
indenylacetic acids 4 containing a bromine or an iodine
atom in the aromatic ring to a Heck, Suzuki, or Sonogashira
coupling, followed by the Knoevenagel condensation under
Sulindac and its metabolites also influence various other
biological phenomena, such as the inflammation-relevant
cyclooxygenase pathway, an apoptosis-inducing pathway, and
the tumor-relevant Wnt pathway (Figure 1).[8]
The established link between sulindac and the Ras
pathway, and the ability of sulindac-derived compounds to
interact with different proteins led us to synthesize a
Angew. Chem. Int. Ed. 2004, 43, 450 –454
www.angewandte.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
451

Communications
Table 1: Summary of the major results of the MTT assay and of analysis
of cell biological parameters.[a] Data for sulindac (1) and its physiological
sulfide metabolite 2 are shown for comparison.[6,12]
No. Compound
Effects on MDCK- IC50
IC50 ratio IC50 ratio
F3 cells
MDCK- MDCK/ C57MG-
F3 [mm] MDCK- Wnt-1/
F3
MDCK-F3
1
2
no effect
320
200
1/1
n.a.
no effect
1/1
n.a.
phenotype rever-
5a
5b
5c
5d
5e
5 f
sion, induction of 50
apoptosis
10/1
3/1
6/1
2/1
4/1
6/1
28/1
6/1
6/1
9/1
phenotype rever-
sion, induction of 90
apoptosis
Scheme 1. a) 2-Chlorotrityl chloride resin, CH2Cl2, (iPr)2NEt, 2 h, 218C;
b) DBU, DMF or toluene, 608C, 16–48 h, aromatic aldehyde R3CHO;
c) 2% CF3COOH in CH2Cl2; d) alkene (5 equiv), [Pd2(dba)3]
(0.5 equiv), P(o-tol)3 (2 equiv), Et3N/dioxane (1:1), Bu4NBr (5 equiv),
858C, 20 h; e) boronic acid (10 equiv), [Pd(PPh3)4] (0.35 equiv);
K3PO4·H2O (20 equiv), DMF, 828C, 22 h; f) alkyne (20 equiv), CuI
(0.3 equiv), DMF/Et3N (1:1), PPh3 (0.5 equiv), [Pd(PPh3)4] (0.5 equiv),
858C, 20 h. DBU=1,8-diazabicyclo[5.4.0]undec-7-ene, DMF=dimethyl-
formamide; dba=trans,trans-dibenzylideneacetone.
phenotype rever-
sion
50
2/1
phenotype rever-
sion, induction of 50
apoptosis
4/1
phenotype rever-
sion
Heck reaction proceeded best if [Pd2(dba)3] was employed
together with P(o-tol)3 and nBu4NBr as the catalyst system
and the reaction was carried out in NEt3/dioxane (1:1) at
858C. The Suzuki reaction proceeded efficiently at 828C in
DMF in the presence of [Pd(PPh3)4], and the Sonogashira
reaction gave satisfactory results if [Pd(PPh3)4] was used as
the catalyst in DMF/NEt3 (1:1) at 858C. It was necessary to
use an alkene, alkyne, or boronic acid in 5–20-fold excess and
to run the reactions for approximately 20 h to obtain accept-
able results.
Knoevenagel condensation of the coupling products with
different aromatic aldehydes followed by release from the
solid support by treatment with acid led to sulindac analogues
6–8, which have an aromatic substituent, an alkene or an
alkyne, attached to the former indene benzene ring
(Scheme 1). The desired sulindac analogues 6–8 were
obtained in overall yields ranging from around 50% to over
90% by using these reaction sequences. The compounds had a
purity in most cases of over 80% immediately after cleavage.
Crude products with a purity of less than 80% were purified
by flash chromatography on silica gel with ethyl acetate/
cyclohexane (1:12(v/v)) and 1% acetic acid as eluent to
increase the purity to more than 80%. In total, 239 com-
pounds were prepared.
10
30/1
2.5/1
3/1
phenotype rever-
sion, induction of 40
apoptosis
phenotype rever-
sion, induction of 50
apoptosis
5g
5h
phenotype rever-
sion
50
6/1
[a] Micromolar concentrations leading to a 50% decrease of the number
of living cells are shown. The cell biological effects of 100 mm of each
substance are shown. MTT=3-(4,5-dimethylthiazol-2-yl)-2,5-diphenylte-
trazolium bromide; n.a.=not applicable.
the conditions described above (Scheme 1). The conditions of
the Pd-catalyzed reactions had to be optimized to obtain high
yields. After substantial experimentation it was found that the
452
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2004, 43, 450 –454

Angewandte
Chemie
NMR spectroscopic investigation with NOE techniques
revealed that, in general, Z isomers were predominantly
formed in the Knoevenagel reactions, with Z/E ratios of more
than 9/1. The corresponding ethylene glycol esters of selected
compounds were prepared (see the Supporting Information).
Phenotype-based screening of the library for in vivo
effects was performed with the Madine–Darby canine
kidney (MDCK) cell line for primary screening. MDCK
cells are immortalized epithelial cells with a large and round
phenotype. They grow in a monolayer, attached to each other
by regular cell-to-cell contacts (Figure 2). MDCK cells can be
each compound on cell proliferation and cell survival was
analyzed. These parameters were quantified by using a
colorimetric proliferation assay with 3-(4,5-dimethylthiazol-
2-yl)-2,5-diphenyltetrazolium bromide (MTT) as an indicator
of the number of living cells.[12,13] Cells were also screened for
signs of apoptosis. Swollen and vesicle-forming cell bodies
and fragmented nuclei were taken as signs of apoptosis. To
confirm the apoptotic effects of the compounds, cells were
stained with annexin V, which binds to phosphatidylserine on
the outer plasma membrane leaflet of cells in the early phases
of apoptosis.[14,15]
In total, 189 compounds were subjected to the primary
screen and analyzed for their ability to return the phenotype
of MDCK-F3 cells to the phenotype of MDCK cells
(Figure 2) and to induce apoptosis, as well as for their effect
on the survival of Ras-transformed MDCK-F3 cells and
C57MG/Wnt-1 cells in the MTT assay relative to the survival
of the untransformed parent cell lines.
Compounds were scored as positive hits and selected for
further analysis if they a) induced the reversion of the
transformed phenotype of MDCK-F3 cells to the phenotype
of MDCK cells or induced apoptosis, b) decreased the
number of living cells by 50% at a concentration at least
two–threefold lower when applied to MDCK-F3 cells than
that required for MDCK cells, and c) showed at least two–
threefold lower cytoxicity with both C57MG/Wnt-1 and
C57MG cells than with MDCK and MDCK-F3 cells
(Table 1; Figure 3).
Figure 2. Representative results of the cell biological phenotype
assays. Cellular morphology was visualized microscopically by immu-
nostaining of intracellular F-actin fibers (red) and by staining nuclei
with the DNA dye 4,6-diamidino-2-phenylindole (blue). A) MDCK cells
have a round and regular phenotype. These cells grow attached to
each other in epithelium-like, monolayer accumulations. B) MDCK-F3
cells show a spindle-shaped, longitudinal phenotype without regular
cell contacts. C) MDCK-F3 cells treated with compound 5e display a
round morphology resembling that of the MDCK mother cells, but the
treated cells form cell contacts. This change can be assigned to the
reversion of the H-Ras-induced transformation. D) MDCK-F3 cells
treated with compound 5b show morphological signs of apoptosis,
such as swollen, round cell bodies, condensed DNA, and apoptotic
vesicles (arrow).
Figure 3. Representative cell survival and proliferation assay. The
curves show the dependence of the percentage of living cells on the
concentration of compound 5h. H-Ras-transformed MDCK-F3 cells
(filled circles) are more sensitive to compound 5h than untransformed
MDCK cells (open circles), which indicates that the compound affects
the H-Ras pathway. C57MG-Wnt1 (filled squares) and C57MG cells
(open squares), which were used for counterscreening in a control
experiment, showed even lower sensitivity to 5h.
transformed by transfection with the H-Ras oncogene. The
resulting F3 cells are long and spindlelike and grow in
multiple layers without regular cell-to-cell contacts.[10] Since
the H-Ras oncogene causes such a significant morphological
alteration, cellular morphology can be used as a parameter for
measuring the effects of a compound on the activated H-Ras
pathway.[6] The specificity of the results obtained in the
MDCK/MDCK-F3 assays was evaluated by counterscreening
for the effects of all compounds on two cell lines. C57MG cells
are murine epithelial mammary gland cells with a round and
large phenotype. The transfection of C57MG cells with the
Wnt-1 oncogene leads to similar morphological changes as
the H-Ras transfection of MDCK cells described above:
C57MG cells turn into spindle-shaped C57MG/Wnt-1 cells
that grow without any contact inhibition.[11] Three parameters
were investigated after incubation of cells with the com-
pounds: In addition to detailed comparative analysis of the
morphologies of treated and untreated cells, the influence of
Of all the tested compounds, 23 fulfilled the above
criteria. The most promising compounds were 5a–5h (see
Table 1 and Figure 3). Closer inspection of the screening
results reveals an initial structure–activity relationship for the
compounds that fulfil the biological criteria. The most active
compounds, 5a–5h, carry a small substituent in the 5- or 6-
position of the indenylacetic acid part of the molecule and a
methyl group on the five-membered ring. The arylidene
substituent should preferably be a five-membered electron-
rich aromatic ring. Only carboxylic acids were active.
Taken together, the results demonstrate that phenotype-
based pathway-selective screening provides new small-mole-
Angew. Chem. Int. Ed. 2004, 43, 450 –454
www.angewandte.org
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
453

Communications
[8] a) M. J. Thun, S. J. Henley, C. Patrono, J. Natl. Cancer Inst. 2002,
94, 252 – 266; b) E. T. Goluboff, Expert Opin. Invest. Drugs 2001,
10, 1875 – 1882; c) I.-M. Karaguni, M. Carpintero, E. Gourzou-
lidou, L. Klein-Hitpaß, T. Mꢀrꢀy, H. Waldmann, O. Müller,
unpublished work.
[9] a) T.-Y. Shen, B. E. Witzel, H. Jones, B. O. Linn, R. B.
Greenwald (Merck and Co.), DE 2039426, 1971 [Chem. Abstr.
1971, 74, 141379]; b) A. Bhattacharya, B. Segmuller, A. Ybarra,
Synth. Commun. 1996, 26, 1775 – 1784.
cule inhibitors of tumor-relevant pathways with high fidelity.
Only a fairly small compound collection, one screen, and one
counterscreen were employed in our investigation. The cell-
biological screening and analysis of the sulindac-derived
compound library yielded eight compounds with the desired
profile. This result amounts to a hit rate of above 4%, a value
that is entirely acceptable. The most active compound 5e had
an IC50 value of 10 mm in the cellular cytotoxicity assay and
was 30 times more active than the parent molecule sulindac.
The results recorded for the phenotypic screen certainly
reflect the fact that the underlying structure chosen for the
compound library was based on that of sulindac, a compound
whose influence on the Ras pathway had already been
established and that can be regarded as biologically preva-
lidated. These results reinforce the notion forwarded above
that biologically prevalidated compound classes should be
used preferentially in chemical biology investigations in
general, and in phenotype-based screens in particular, to
obtain high hit rates.
[10] J. Behrens, M. M. Mareel, F. M. Van Roy, W. Birchmeier, J. Cell
Biol. 1989, 108, 2435 – 2447.
[11] H. Shimizu, M. A. Julius, M. Giarre, Z. Zheng, A. M. Brown, J.
Kitajewski, Cell Growth Differ. 1997, 8, 1349 – 1358.
[12] I. M. Karaguni, K. H. Glusenkamp, A. Langerak, C. Geisen, V.
Ullrich, G. Winde, T. Mꢀrꢀy, O. Müller, Bioorg. Med. Chem.
Lett. 2002, 12, 709 – 713.
[13] F. Denizot, R. Lang, J. Immunol. Methods 1986, 89, 271 – 277.
[14] S. J. Martin, C. P. Reutelingsperger, A. J. McGahon, J. A. Rader,
R. C. van Schie, D. M. LaFace, D. R. Green, J. Exp. Med. 1995,
182, 1545 – 1556.
[15] M. van Engeland, L. J. Nieland, F. C. Ramaekers, B. Schutte,
C. P. Reutelingsperger, Cytometry 1998, 31, 1 – 9.
In addition to high hit rates and hit compounds with
significantly higher cellular activity than is often obtained
with other methods, our results prove that phenotype screening
based on a branched signaling pathway can deliver compounds
that affect the pathway by acting on different targets. In the
following paper,[16] we demonstrate that the majority of the
compounds shown in Table 1 interfere with the Ras–Raf
interaction. Such a screen may additionally yield compounds
that target the pathway employed for counterscreening,[17]
which renders the entire library synthesis, as well as the
biological and biochemical investigations, very efficient.
[16] H. Waldmann, I. M. Karaguni, M. Carpintero, E. Gourzoulidou,
C. Herrmann, C. Brockmann, H. Oschkinat, O. Müller, Angew.
Chem. 2004, 116, 460; Angew. Chem. Int. Ed. 2004, 43, 454
(following article in this issue).
[17] H. Waldmann, O. Müller, unpublished results.
Received: August 7, 2003 [Z52587]
Keywords: bioorganic chemistry · combinatorial chemistry ·
.
library screening · medicinal chemistry · signal transduction
[1] Review: B. R. Stockwell, Neuron 2002, 36, 559 – 562.
[2] See, for example, a) T. U. Mayer, T. M. Kapoor, S. J. Haggarty,
R. W. King, S. L. Schreiber, T. J. Mitchison, Science 1999, 286,
971 – 974; b) S. Hotha, J. C. Yarrow, J. G. Yang, S. Garrett, K. V.
Renduchintala, T. U. Mayer, T. M. Kappor, Angew. Chem. 2003,
115, 2481 – 2484; Angew. Chem. Int. Ed. 2003, 42, 2379 – 2382.
[3] B. E. Evans, K. E. Rittle, M. G. Bock, R. M. DiPardo, R. M.
Freidinger, W. L. Whitter, G. F. Lundell, D. F. Veber, P. S.
Anderson, R. S. L. Chang, V. J. Lorri, D. J. Cerino, T. B. Chen,
P. J. Kling, K. A. Kunkel, J. P. Sprinter, J. Hirshfield, J. Med.
Chem. 1988, 31, 2235 – 2246.
[4] R. Breinbauer, I. R. Vetter, H. Waldmann, Angew. Chem. 2002,
114, 3002– 3015; Angew. Chem. Int. Ed. 2002, 41, 2879 – 2890.
[5] Reviews: a) H. Waldmann, A. Wittinghofer, Angew. Chem.
2000, 112, 4360 – 4383; Angew. Chem. Int. Ed. 2000, 39, 4192–
4214; b) J. Downward, Nat. Rev. Cancer 2002, 3, 11 – 22.
[6] a) C. Herrmann, C. Block, C. Geisen, K. Haas, C. Weber, G.
Winde, T. Moroy, O. Müller, Oncogene 1998, 17, 1769 – 1776;
b) I. M. Karaguni, P. Herter, P. Debruyne, S. Chtarbova, A.
Kasprzynski, U. Herbrand, M. R. Ahmadian, K. H. Glusen-
kamp, G. Winde, M. Mareel, Cancer Res. 2002, 62, 1718 – 1723.
[7] J. Kato-Stankiewicz, I. Hakimi, G. Zhi, J. Zhang, I. Serebriiskii,
L. Guo, H. Edamatsu, H. Koide, S. Menon, R. Eckl, S. Sakamuri,
Y. Lu, Q.-Z. Chen, S. Agarwal, W. R. Baumbach, E. A. Golemis,
F. Tamanoi, V. Khazak, Proc. Natl. Acad. Sci. USA 2002, 99,
14398 – 14403.
454
ꢀ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2004, 43, 450 –454