Detection of ligands from a dynamic combinatorial library by X-ray crystallography

Article abstract of DOI:10.1002/anie.200351951

A dynamic combinatorial library of ligands derived from hydrazines A and isatins B was assembled in the presence of crystals of the target protein, cyclin-dependent kinase 2 (CDK2). Unambiguous detection of potent ligands, formed as part of the library, was successfully achieved by X-ray crystallographic analysis.

Full text of DOI:10.1002/anie.200351951

Angewandte
Chemie
Dynamic Combinatorial Chemistry
DCL, and its detailed binding mode is defined from the
electron-density maps. Furthermore, only very small amounts
of protein are required for each individual DCX experiment
compared with previously reported DCC protocols.
Detection of Ligands from a Dynamic
Combinatorial Library by X-ray
Crystallography**
It has been established with both X-ray crystallography
and NMR spectroscopic techniques that very small molecules
or “fragments” (MW = 100–200) are capable of binding to
proteins in a reproducible and specific manner, even though
their intrinsic potency, as determined with an in vitro bio-
logical assay, is very weak(IC 50 mm–mm).[10–14] We postulated
that fragments bound to adjacent pockets within the active
site of a protein could, in principle, self-assemble to generate
larger, more potent ligands if they had complimentary
chemical reactivities.[15] Such a method would have key
advantages, as the time-consuming and expensive practice
of conventional synthesis, purification, and testing of combi-
natorial libraries designed to discover potent inhibitors might
be obviated.
Miles S. Congreve,* Deborah J. Davis, Lindsay Devine,
Cesare Granata, Marc O’Reilly, Paul G. Wyatt, and
Harren Jhoti
Dynamic combinatorial chemistry (DCC) is an approach to
molecular recognition in which specific members of a
combinatorial library are selected and amplified with the
use of a template.[1–6] The principle difference between DCC
and traditional combinatorial chemistry is that the reaction
linking the building blocks together in DCC is reversible and
there is an ongoing interchange between the different
members of the dynamic combinatorial library (DCL)
under thermodynamic control.[7] A DCL is therefore able to
respond to molecular recognition events owing to the
presence of a template, such as a protein, which can stabilize
a particular member of the library and
To explore the potential of DCX, we used cyclin-depend-
ent kinase 2 (CDK2) for an initial proof-of-principle study.
This target has been the subject of intense investigation by
many groups, with the aim of developing inhibitors for the
induce a shift in the equilibrium, favoring
the formation of the selected species. A
drawbackof the method is that it usually
requires excess quantities of protein for an
effect on the equilibrium to be observed.
Also, the effect can only be detected by
comparison of identical libraries, generated
with and without the protein component
present, using either mass spectrometric
analysis or HPLC chromatograms as a
fingerprint. We now report a complimen-
tary approach in which ligands are observed
directly by X-ray crystallography by inter-
pretation of electron-density maps from
crystals exposed to a dynamic combinatorial
library mixture. We call this technology
dynamic combinatorial X-ray crystallogra-
phy or DCX. This approach was used to
detect rapidly potent inhibitors of the
cyclin-dependent kinase 2 protein. DCX
has key advantages over previously
reported DCC technologies[8,9] in that
direct identification of the ligand is possible
from the mixture of components in the
Table 1: Combinatorial array of oxindole compounds synthesized by reaction of hydrazines A and
isatins B.[a]
A”B
A1
–
A2
A3
A4
A5
A6
[b]
R1 =Cl
R2 =Cl
R3 =Cl
R3 =SO2NH2
R1 =Cl; R3 =SO2Me
B1
B2
B3
B4
B5
R5 =NO2
R5 =Cl
10–25
60–95
10–25
30–50
30–50
60–95
60–95
60–95
60–95
60–95
60–95
60–95
30–50
60–95
60–95
30–50
60–95
10–25
60–95
60–95
60–95
60–95
60–95
60–95
60–95
30–50
30–50
30–50
30–50
10–25
R5 =SO3H
R7 =CF3
R5 =OCF3
[a] Values indicate the extent to which the reaction occurred in aqueous DMSOafter 48 h at room
temperature as assessed by percentage purity by peak area of the product by LC/MS (10–25%, 30–50%,
or 60–95% of total peaks excluding solvent front). [b] R groups=H unless indicated otherwise.
DMSO=dimethyl sulfoxide.
[*] Dr. M. S. Congreve, Dr. D. J. Davis, L. Devine, C. Granata,
Dr. M. O’Reilly, Dr. P. G. Wyatt, Dr. H. Jhoti
Astex Technology
treatment of a number of human cancers.[16–17] We chose an
inhibitor series developed for CDK2 based on an oxindole
template.[18] These inhibitors (Table 1) present substituents in
adjacent lipophilic binding pockets within the ATP-binding
site of the kinase and can be disconnected to reactive
fragments (termed monomers) of approximately equal size
and complexity (hydrazines A and isatins B; Table 1). Before
attempting to perform DCX with crystals of CDK2, we first
established the suitability of the chemistry for synthesizing a
436 Cambridge Science Park
Milton Road, Cambridge CB4 0QA (UK)
Fax: (+44)1223-226-201
E-mail: m.congreve@astex-technology.com
[**] The authors thank Professor Chris Abell, Dr. Robin Carr, Dr. David
Rees, and Dr. Chris Murray for useful suggestions and discussions.
DCX is a Trade Mark of Astex Technology, Ltd.
Supporting information for this article is available on the WWW
under http://www.angewandte.org or from the author.
Angew. Chem. Int. Ed. 2003, 42, 4479 –4482
DOI: 10.1002/anie.200351951
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4479

Communications
DCL based on these oxindole inhibitors. Hydrazone chemis-
electron density that was consistent with ligand binding only
in the latter case (Table 2). Furthermore, the electron density
was consistent with the expected potent ligand A5B2
(Figure 1e). Next, the degeneracy of the DCL was increased
to 30 possible ligands using a reaction cocktail that contained
hydrazines (A1–6) and isatins (B1–5). Difference electron
density consistent with the reaction product A5B2 was again
observed, clearly demonstrating that the method is useful for
the identification of ligands from a DCL mixture (Figure 1 f).
The detection of ligands formed in this study does not rely
on an observed perturbation of the reaction equilibrium in the
way that is conventionally used in the screening of dynamic
combinatorial libraries. Indeed, since the total amount of
protein in the crystals is very small compared to the total
amount of the monomers present in solution, one would not
expect to be able to measure any gross changes in the
outcome of the reaction caused by binding of product ligands
to the crystals. However, it is still possible that the protein
influences the outcome of the chemical reactions occurring
directly in the crystals. In this study the reaction products are
formed reversibly, and each product might be formed in situ
within the protein crystal, under thermodynamic control. A
consideration of the environment within the protein crystal
would lead one to suggest that the monomers may tend to
compete for occupancy of the ATP-binding groove, allowing
“templating” to occur and helping to drive selection of the
most potent ligands in the active site. Indeed, this effect has
been previously shown in studies of reactions that occur in the
active site of a protein, but under kinetic control.[26] However,
it cannot be discounted that the effect observed in this case
might be driven by equilibration of the ligands themselves
into the CDK2 crystals, preformed by reaction of monomers
present in the aqueous solution. Experiments designed to
distinguish between these two possible situations are under-
way in our laboratories.
try, used for assembly of the library products, has previously
been shown to be highly suitable for dynamic combinatorial
library syntheses.[19–20] A range of hydrazines (A1–6) and
isatins (B1–5) were selected that would present a variety of
functional groups to the lipophilic pockets in the ATP-binding
site. The selection of these monomers was aided by knowl-
edge of the features of the active site. A trial of each of the 30
possible reactions was then carried out under aqueous
conditions in the presence of 20% of the cosolvent dime-
thylsulfoxide (DMSO) over a 48-hour period, and the results
showed that each product was successfully synthesized
(Table 1). Further monomer competition studies were then
carried out and these indicated that all the possible products
could be formed as part of the DCL.[21]
Once the suitability of the chemistry was established, the
reactions were then carried out in the presence of protein
crystals. A summary of these experiments is given in Table 2.
Table 2: Summary of X-ray crystallography experiments and biological
assay results.[a]
Mixture composition
Product
IC50 [nm]
A5 + B1
A5 + B2
A5 + B3
YES
YES
YES
30
30
30
A5 + B4
A5 + B5
NOinactive
YES
30
B2 + (A1,A2,A3,A4)
B2 + (A1,A2,A3,A4,A5,A6)
(B1,B2,B3,B4,B5) +
(A1,A2,A3,A4,A5,A6)
NON/A
YES (A5B2)
YES (A5B2)
N/A
N/A
[a] Table indicates if electron density consistent with a reaction product
from mixtures of isatins and hydrazines was detected upon exposure to
CDK2 protein crystals. The biological activity subsequently measured of
the corresponding purified product of the reaction is also given when
synthesized. Full details of the X-ray data collected and biological assay
conditions are described in the Supporting Information.
In conclusion, we have demonstrated that X-ray crystal-
lography can be used to detect small-molecule ligands
generated in situ and bound to a target protein. These
findings provide the basis for a new drug discovery technology
in which dynamic combinatorial libraries can be used to
identify novel, potent ligands. We believe this approach could
be broadly applicable to the discovery of inhibitors of
therapeutically useful proteins within drug discovery pro-
grams.
The first studies were performed with individual crystals of
CDK2[22–24] soaked in reaction solutions containing monomer
A5 and each of the isatin monomers (B1–5) in turn. In all but
one case (A5 + B4) the resulting electron density in the ATP
pocket indicated that the corresponding ligand had bound
(Table 2, Figures 1a–d). To correlate these findings with the
biological activity, each of these ligands was then synthesized
as purified single compounds, and the activity determined in
an assay for CDK2. Each of the ligands that had bound in the
protein crystal was shown to be a potent enzyme inhibitor.
Analogue A5B4, which did not bind in X-ray studies, was
biologically inactive (Table 2).[25] These data are broadly
consistent with the published structure–activity relationships
for this series of inhibitors.[18]
Studies were then carried out to investigate the use of our
approach to detect ligands from a dynamic combinatorial
library mixture. Firstly, two “cocktails” of reaction solutions
were added to CDK2 crystals to explore the potential of the
method to discover potent inhibitors from a mixture of
nonbinders. Comparison of the cocktail containing (A1–4) +
B2 with the cocktail that contained (A1–6) + B2 revealed
Received: May 22, 2003 [Z51951]
Keywords: analytical methods · combinatorial chemistry ·
.
medicinal chemistry · proteins · X-ray diffraction
[1] S. Otto, R. L. E. Furlan, J. K. M. Sanders, Drug Discovery Today
2002, 117 (Combinatorial Chemistry: A Supplement).
[2] J.-M Lehn, A. V. Eliseev, Science 2001, 291, 2331.
[3] C. Karan, B. L. Miller, Drug Discovery Today 2000, 5, 67.
[4] O. Ramström, J.-M. Lehn, Nat. Rev. Drug Discovery 2002, 1, 26.
[5] D. A. Erlanson, J. W. Lam, C. Wiesmann, T. N. Luong, R. L.
Simmons, W. L. DeLano, I. C. Choong, M. T. Burdett, W. M.
Flanagan, D. Lee, E. M. Gordon, T. O'Brien, Nat. Biotechnol.
2003, 21, 308.
4480
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2003, 42, 4479 –4482

Angewandte
Chemie
Figure 1. (Fo–Fc) and (2Fo–Fc) maps for the singlet and cocktail reactions. All (Fo–Fc) maps are colored red and trimmed to within 2.5 ꢀ of the
ligand. The difference maps were generated after segmented rigid body, positional, and isotropic B-factor refinements of the protein, in the
absence of any ligand. The final refined protein–ligand structures and corresponding final (2Fo–Fc) maps (contoured at 1.0s and colored gray) are
shown in all six figures for reference. Elements in the ligands are colored as follows: yellow (C), blue (N), red (O), cyan (Cl), pink (F), and orange
(S). Atoms in the protein are colored similarly, except that carbon atoms are green. a) 2.5s (Fo–Fc) map for the reaction (A5 + B2); b) 2.0s
(Fo–Fc) map for the reaction (A5 + B1); c) 2.5s (Fo–Fc) map for the reaction (A5 + B3). d) 2.0s (Fo–Fc) map for the reaction (A5 + B5). e) 2.5s
(Fo–Fc) map for the reaction B2 + (A1–6). f) 2.5s (Fo–Fc) map for reaction (A1–6) + (B1–5). Maps for the reaction (A5 + B4) are provided in
the supplementary data. All figures were generated with AESOP (M.E.M. Noble, Oxford, unpublished results).
Angew. Chem. Int. Ed. 2003, 42, 4479 –4482
www.angewandte.org
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
4481

Communications
[6] P. A. Brady, R. P. Bonar-Law, S. J. Rowan, C. J. Suckling, J. K. M.
Sanders, Chem. Commun. 1996, 3, 319.
[7] P. Seneci, Solid-Phase Synthesis and Combinatorial Technolo-
gies, Wiley-Interscience, New York, 2000.
[8] M. Hochgürtel, H. Kroth, D. Piecha, M. W. Hofmann, C.
Nicolau, S. Krause, O. Schaaf, G. Sonnenmoser, A. V. M. Eliseev,
Proc. Natl. Acad. Sci. USA 2002, 99, 3382.
[9] W. G. Lewis, L. G. Green, F. Grynszpan, Z. Radic, P. R. Carlier,
P. Taylor, M. G. Finn, K. B. Sharpless, Angew. Chem. 2002, 114,
1095; Angew. Chem. Int. Ed. 2002, 41, 1053.
[10] T. L. Blundell, C. Abell, A. Cleasby, M. J. Hartshorn, I. J. Tickle,
E. Parasini, H. Jhoti, Drug Des. Discovery 2002, 279, 53 (Special
Publication: Proceedings of the Royal Society of Chemistry
Meeting—Cutting Edge Approaches to Drug Design, March
2001, Ed.: D. Flower, Royal Society of Chemistry, London).
[11] R. Carr, H. Jhoti, Drug Discovery Today 2002, 7, 522.
[12] T. L. Blundell, H. Jhoti, C. Abell, Nat. Rev. Drug Discovery 2002,
1, 45.
[13] V. L. Nienaber, P. L. Richardson, V. Klighofer, J. J. Bouska, V. L.
Giranda, J. Greer, Nat. Biotechnol. 2000, 18, 1105.
[14] P. J. Hajduk, R. P. Meadows, S. W. Fesik, Q. Rev. Biophys. 1999,
32, 211.
[15] B. A. Katz, J. Finer-Moore, R. Mortezaei, D. H. Rich, R. M.
Stroud, Biochemistry 1995, 34, 8264.
[16] M. Knockaert, P. Greengard, L. Meijer, Trends Pharmacol. Sci.
2002, 23, 417.
[17] P. M. Fischer, D. P. Lane, Curr. Med. Chem. 2000, 7, 1213.
[18] H. N. Bramson, J. Corona, S. T. Davis, S. H. Dickerson, M.
Edelstein, S. V. Frye, R. T. Gampe, P. A. Harris, A. Hassell,
W. D. Holmes, R. N. Hunter, K. E. Lackey, B. Lovejoy, M. J.
Luzzio, V. Montana, W. J. Rocque, D. Rusnak, L. Shewchuk,
J. M. Veal, D. H. Walker, L. F. Kuyper, J. Med. Chem. 2001, 44,
4339.
[19] J. K. M. Sanders, Pure Appl. Chem. 2000, 72, 2265.
[20] V. Berl, I. Huc, J.-M. Lehn, A. DeCian, J. Fischer, Eur. J. Org.
Chem. 1999, 11, 3089.
[21] Further details of the experiments carried out to confirm the
suitability of the chemical process used are described in the
Supporting Information.
[22] J. Rosenblatt, H. De Bondt, J. Jancarik, D. O. Morgan, S. H. Kim,
J. Mol. Biol. 1993, 230, 1317.
[23] A. M. Lawrie, M. E. M. Noble, P. Tunnah, N. R. Brown, L. N.
Johnson, J. A. Endicott, Nat. Struct. Biol. 1997, 4, 796.
[24] H. L. De Bondt, J. Rosenblatt, J. Jancarik, H. D. Jones, D. O.
Morgan, S. H. Kim, Nature 1993, 363, 595.
[25] Details of the biological assay used and further biological results
on additional library members are described in the Supporting
Information.
[26] R. Nguyen, I. Huc, Angew. Chem. 2001, 113, 1824; Angew.
Chem. Int. Ed. 2001, 40, 1774.
4482
ꢀ 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
Angew. Chem. Int. Ed. 2003, 42, 4479 –4482