Selective identification of cooperatively binding fragments in a high-throughput ligation assay enables development of a picomolar caspase-3 inhibitor

Full text of DOI:10.1002/anie.200901647

Communications
DOI: 10.1002/anie.200901647
Dynamic Ligation Screening
Selective Identification of Cooperatively Binding Fragments in a High-
Throughput Ligation Assay Enables Development of a Picomolar
Caspase-3 Inhibitor**
Marco F. Schmidt, Adeeb El-Dahshan, Sandro Keller, and Jꢀrg Rademann*
Fragment-based lead discovery has become popular over the
past years, allowing for a more efficient sampling of the
chemical space and thus resulting in higher hit rates than that
for the screening of non-fragment chemical libraries.[1] Since
the concept of fragment-based lead discovery was first
proposed,[2] various detection methods have been used to
identify the binding of fragments to protein templates,
including HPLC,[3] NMR spectroscopy,[4] X-ray crystallogra-
phy,[5] and mass spectrometry.[6] Recently, dynamic ligation
screening has been introduced to detect reversibly ligated
fragments in a biochemical assay through their competition
with a fluorogenic enzyme substrate.[7]
method for the selective detection of cooperatively binding
fragments should be powerful in the development of potent
protein ligands.
We envisioned that the selective detection of cooperative
fragments should be possible through the combination of
dynamic ligation screening (DLS)[7] with fluorescence polar-
ization (FP) experiments.[9] Using a fluorescent directing
probe for reversible fragment ligations would allow differ-
entiation between additive, cooperative, and competitive
binding of fragments to a protein target, as detected by
unchanged, enhanced, or reduced FP, respectively (Figure 1).
Reported detection methods for protein-binding frag-
ments, however, do not answer the most challenging question
in fragment-based ligand discovery: How can a primary
ligand as the starting point be extended optimally by a
secondary fragment? While “second-site screening”[8] by
NMR spectroscopy or crystallography delivers fragment hits
for different binding sites, it does not provide information
regarding the optimal, bioactive combination of fragments.
Evidently, for a thermodynamically optimized extension
of a primary ligand (A) by a secondary fragment (B),
cooperative binding of both components is required. Coop-
erative binding in this context results if the binding of A is
amplified by the binding of B. Cooperative binding therefore
is distinct from additive binding (in which A and B bind
independently without mutual influence) and competitive
binding where the secondary fragment B inhibits the binding
of A. Cooperatively binding fragments result in strongly
increased binding affinities if the two components are
connected by a suitable covalent linkage. Consequently, a
Figure 1. Selective detection of cooperatively binding fragments in a
dynamic ligation assay. Binding of fragment A (orange) is determined
by fluorescence polarization. Three alternative cases of fragment
interactions can be distinguished in the assay: Additive binding of
fragment B (green) has no effect on the observed polarization
(case 1). Competitive binding leads to decreased polarization (case 2).
Finally, cooperative binding of fragments A and B results in an
extended correlation time tAB and, thus, an increase in FP (case 3).
[*] Dr. M. F. Schmidt, Dr. A. El-Dahshan, Dr. S. Keller,
Prof. Dr. J. Rademann
Leibniz Institute of Molecular Pharmacology (FMP)
Robert-Rꢀssle-Strasse 10, 13125 Berlin (Germany)
Fax: (+49)30-9406-2901
Caspase-3 (cysteinyl-aspartyl-specific protease 3) was
selected as protein target to test the feasibility of this
approach for the identification of cooperative fragments.
The protein has been identified as the cellular switch towards
for apoptosis.[10] Therefore, it is considered as a drug target for
clinical indications involving overregulated cell death, such as
traumatic brain injury, status epilepticus, amyotrophic lateral
sclerosis (ALS), and Parkinsonꢀs disease,[11] and potent
inhibitors of caspase-3 have been reported.[12]
The fluorescent ligation probe was designed on the basis
of the native substrate consensus sequence DEVD.[12c] An a-
ketoaldehyde peptide was selected, as it enables the clear
separation of the protein-interacting keto functionality from
E-mail: rademann@fmp-berlin.de
Dr. M. F. Schmidt, Dr. A. El-Dahshan, Prof. Dr. J. Rademann
Institute of Chemistry and Biochemistry, Free University Berlin
Takustrasse 3, 14195 Berlin (Germany)
[**] We thank Christoph Erdmann, Franziska Hinterleitner, and Dagmar
Krause for technical support and Dr. Samuel Beligny, Jꢀrn Saupe,
and Dr. Peter Schmieder for helpful discussions. Work at the Leibniz
Institute of Molecular Pharmacology was supported by the DFG
(FOR 806, SFB 765, Ra 895/2-5). J.R. thanks the Fonds der
Chemischen Industrie for continuous support.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.200901647.
6346
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 6346 –6349

Angewandte
Chemie
the fragment-interacting aldehyde group. a-Ketoaldehydes
have been reported to be fully hydrated in water, and
reactions with amine nucleophiles are disfavored (ender-
gonic) in aqueous solvents, the equilibrium being largely on
the side of the hydrate and free amine.[13]
Reporter probes 6a and 6b were prepared by C-acylation
of polystyrene-supported trimethylsilylethyl(TMSE)-phos-
phoranylidene acetate (Scheme 1).[14] The TMSE protecting
group was removed by employing the mild fluoride reagent
The binding affinity of compound 6a was determined by
FP titration with caspase-3. The FP titration curve indicated a
dissociation constant (KD) of 15 nm. The FP assay was
adapted to 384-well microtiter plates and used to test 7397
fragments including 4019 nucleophilic, primary amines from
the ChemBioNet library. For the high-throughput screening
20 nm of caspase-3, 10 nm of 6a, and 10 mm of each fragment
was incubated in a total volume of 10 mL.
In agreement with the rationale behind our approach
(Figure 1), fragments tested in this dynamic ligation assay
were clustered into three different classes. First, no change in
the FP signal was observed for most fragments, indicating
noncooperative, additive binding of A and B or, more likely,
no binding of B at all (Figure 1, case 1). Second, for 78
fragments a decrease in the FP signal was observed, suggest-
ing a negative effect of B on the binding of A. One reason
could be competitive inhibition of the binding of the
fluorescent reporter probe A by fragment B (Figure 1,
case 2). All 78 fragments were tested in an enzymatic assay
using Ac-DEVD-AMCA 7 (see the Supporting Information)
as the fluorogenic substrate of caspase-3.[12] Indeed, 21 of the
negatively cooperative case 2 ligands were active at a
concentration of 10 mm as competitive inhibitors. Four of
them, 10–13, could be identified as competitive inhibitors with
KI values in the low micromolar range (for structures, see
Scheme 2 in the Supporting Information), making them some
of the best nonpeptidic inhibitors reported to date.[12] Third,
for 176 fragments the FP signal more than 20% stronger than
that of the controls. This observation indicated positively
cooperative binding of A and B to the protein, possibly
through the formation of a ligation product with increased
affinity. Fifty of the cooperatively binding fragments were
validated in the enzyme assay. Fragment 8 (see Figure 4 and
Scheme 3 in the Supporting Information) was identified as the
most potent FP enhancer among the 176 tested compounds
and displayed a KI value of 120 mm.
tris(dimethylamino)sulfonium
difluorotrimethylsilicate
(TAS-F), leading to instantaneous decarboxylation of the
phosphorane on the resin. Oxidative cleavage with dimethyl-
dioxirane (DMD) and acidic removal of the side-chain
protecting groups yielded compounds 6a and 6b.
To better understand the experimental data obtained, we
had to rationalize the binding and interaction of compounds
6a and 8 and quantify the degree of cooperativity (Figure 2).
For this purpose, FP data obtained by titrating 10 nm 6a in the
presence of various concentrations of fragment 8 with
caspase-3 (Figure 3a) were interpreted in terms of a thermo-
dynamic model assuming either merely additive or coopera-
tive binding of 6a and 8 to caspase-3 (Figure 3b). In the first
scenario, we assumed no interaction between 6a and 8 by
taking the ligation equilibrium constant as KC = 0 (Figure 2).
In this purely additive, noncooperative case, the presence of 8
is predicted to have no influence on the binding of 6a
(Figure 3b, black line). Indeed, the experimental data in the
absence of 8 (black symbols) or in its presence at concen-
trations of 100 nm (blue symbols) and 1 mm (green symbols)
were found to be in reasonable agreement with this scenario
(Figure 3b).
Scheme 1. Synthesis of the peptidyl a-ketoaldehydes CF-DEVD-CHO
(6a) and Ac-DEVD-CHO (6b) from triphenylphosphane polystyrene,
which was alkylated and acylated as reported earlier.[14] Reaction
conditions: a) Trimethylsilylethyl bromoacetate (5 equiv), toluene,
15 min, microwave, 1008C; b) Et3N (5 equiv), in CH2Cl2, 2 h, RT;
c) Fmoc-Asp-(OtBu)-OH (5 equiv), MSNT (5 equiv), lutidine
(4.9 equiv) in CH2Cl2, 12 h, RT; d) 20% piperidine/DMF, 6 min.
e) Fmoc-AA-OH (5 equiv), DIC (5 equiv), HOBt (5 equiv) in DMF, 3 h
(steps (d) and (e) were repeated n times); f) 5,6-carboxyfluorescein
(10 equiv), DIC (10 equiv), HOBt (10 equiv) in DMF, 3 h, RT or Ac2O
(4 equiv) in DMF, 20 min, RT (two times); g) TAS-F (3 equiv), in DMF,
3 h, RT; h) DMD (3–4 equiv)/acetone, in CH2Cl2, 30 min, 08C; i) TFA/
CH2Cl2/H2O (50:45:5, v:v:v). Fmoc=9-fluorenylmethyloxycarbonyl,
MSNT=1-(2-mesitylenesulfonyl)-3-nitro-1H-1,2,4-triazole, DIC=N,N’-
diisopropylcarbodiimide, HOBt=1-hydroxy-1H-benzotriazole, TFA=tri-
fluoroacetic acid.
The FP data at 10 mm 8 (red symbols), however, revealed
that the stronger binding of 6a is not in agreement with the
noncooperative model. Therefore, in the second scenario, we
explicitly implemented cooperative binding by allowing the
formation of the ligation product from protein-bound 6a and
8, as reflected by KC > 0 (Figure 2). Using this model, the FP
Angew. Chem. Int. Ed. 2009, 48, 6346 –6349
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
6347

Communications
protein-protein interactions. The information gained from
this assay can be utilized directly for the construction of
improved protein ligands by a rational approach.
Experimental Section
Synthetic procedures and analytical data of novel compounds are
provided in the Supporting Information.
Fluorescence polarization (FP) assays were performed in
untreated black 384-well microtiter plates (Corning B.V. Life
Sciences, Schipol-Rijk, Netherlands) using the microplate reader
Genios Pro, (Tecan, Crailsheim, Germany) (buffer conditions:
50 mm HEPES pH 7.4, 100 mm NaCl, 0.5% CHAPS, and 1 mm
EDTA). The peptidyl a-ketoaldehyde 6a was titrated in a
concentration of 10 nm against various concentrations of cas-
pase-3 (0–100 nm) in a total volume of 10 mL. The excitation
wavelength was 485 nm, and the emission wavelength was 535 nm.
The measured FP data were analyzed in GraphPad Prism 4 for
Windows (GraphPad, La Jolla, USA) by nonlinear regression
(curve fitting).
Figure 2. A model of additive and cooperative binding of molecules A
(orange) and B (green) to a protein (blue). In the case of additive binding,
A and B bind independently (KAB =KAKB) and do not interact with each
other (KC =0). Cooperative binding resulting from protein-templated liga-
tion can be conceptualized as consisting of two stages: 1) additive binding
of the two fragments as given by KAB =KAKB (upper part) and 2) formation
of the ligation product as characterized by KC >0 (lower part).
The thermodynamic model used to distinguish between
cooperative and additive binding is reported in detail in the
Supporting Information. In short, raw FP and enzyme inhibition
data (Figure 3a) were converted into binding isotherms (Fig-
data measured at all concentrations of 8 could be reproduced
satisfactorily (Figure 3b, blue, green, and red lines). The best
agreement with the experimental data was obtained assuming
a ligation constant of KC = 20. This indicates that the
equilibrium concentration of the protein-bound ligation
product is 20 times higher than that of protein carrying
unligated 6a and 8, corresponding to a Gibbs free energy of
ligation of DG8 = ÀRTlnKC = À7 kJmolÀ1. Thus, the protein-
templated ligation reaction is exergonic, which starkly con-
trasts with the endergonic ligation reaction in aqueous
solution, where no ligation product of a-ketoaldehyde and 8
is detectable.[13]
To confirm the cooperative binding of the reversibly
formed ligation product, the covalent and irreversible ligation
product of a-ketoaldehyde 6b and the most active fragment 8
was synthesized by reductive amination (Figure 4). The
obtained ketone 9 was tested in the functional caspase-3
assay. A KI value of 80 pm was determined (Figure 3), making
9 the most active caspase-3 inhibitor reported to date.[12] Thus,
compound 9 inhibits the enzyme more than 300 times more
effectively than the a-ketoaldehyde precursor 6b (KI =
25 nm).
In summary we have developed and implemented the
differential detection of cooperative and competitive frag-
ments in one FP experiment operable in high-throughput
format. Cooperative fragment binding can be verified and
distinguished from merely additive effects. By employing a
model for the ligation equilibrium we could estimate the
ligation constant and the Gibbs free energy of the reversible
ligation reaction on the protein surface. A chemically stable
combination of the best fragment hits yielded the most potent
inhibitor of caspase-3 reported to date. Thus, the described
dynamic method for cooperativity screening can be used to
dissect and analyze binding contributions of single fragments
in a ligation system. As it is based on FP detection, the
method is broadly applicable to enzyme targets as well as to
Figure 3. Titration experiments. a) Raw data (symbols) and sigmoidal
fits (lines) according to Equation (1). FP was recorded upon adding
caspase-3 (component P, lower abscissa) to 10 nm 6a alone (black) or
in the presence of 8 at concentrations of 100 nm (blue), 1 mm (green),
and 10 mm (red). I is the FP signal intensity (left ordinate). Enzyme
inhibition was performed by adding 9 (upper abscissa) to 3 nm
caspase-3 in the presence of 5 mm Ac-DEVD-AMCA (magenta). v0 is
the initial rate of the enzyme reaction (right ordinate). b) Normalized
titration data as obtained from the experimental FP data by employing
Equation (2) (symbols) and best fits derived from Equations (3–9)
(lines). The ordinates give the degree of binding, that is, the concen-
tration of protein–ligand complex divided by the total concentration of
6a (FP data, left ordinate) or caspase-3 (enzyme inhibition data, right
ordinate). See the Supporting Information for equations and details.
6348 www.angewandte.org
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 6346 –6349

Angewandte
Chemie
2007, 6, 211 – 219; c) J. Rademann, Angew. Chem. 2004, 116,
4654 – 4656; Angew. Chem. Int. Ed. 2004, 43, 4554 – 4556; d) P. T.
Corbett, J. Leclaire, L. Vial, K. R. West, J. L. Wietor, J. K. M.
Sanders, S. Otto, Chem. Rev. 2006, 106, 3652 – 3711; e) D. A.
Erlanson, Curr. Opin. Biotechnol. 2006, 17, 643 – 652.
[2] W. P. Jencks, Proc. Natl. Acad. Sci. USA 1981, 78, 4046 – 4050.
[3] a) I. Huc, J. M. Lehn, Proc. Natl. Acad. Sci. USA 1997, 94, 2106-
2110; b) J. M. Lehn, A. Eliseev, Science 2001, 291, 2331 – 2332.
[4] a) S. B. Shuker, P. J. Hajduk, R. P. Meadows, S. W. Fesik, Science
1996, 274, 1531- 1534; b) V. L. Nienhaber, P. L. Richardson, V. J.
Klinghofer, J. J. Bousha, J. Greer, Nat. Biotechnol. 2000, 18,
1105 – 1108; c) P. Vongvilai, M. Angelin, R. Larrson, O. Ram-
strꢁm, Angew. Chem. 2007, 119, 966 – 968; Angew. Chem. Int.
Ed. 2007, 46, 948 – 950.
[5] a) T. L. Blundell, H. Jhoti, C. Abell, Nat. Rev. Drug Discovery
2002, 1, 45 – 54; b) A. Sharff, H. Jhoti, Curr. Opin. Chem. Biol.
2003, 7, 340 – 345; c) W. J. Sanders et al., J. Med. Chem. 2004, 47,
1709 – 1718.
[6] a) D. A. Erlanson, J. A. Wells, A. C. Braisted, Annu. Rev.
Biophys. Biomol. Struct. 2004, 33, 199 – 223; b) 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. OꢀBrien, Nat. Biotechnol. 2003, 21, 308 – 314.
[7] M. F. Schmidt, A. Isidro-Llobet, M. Lisurek, A. El-Dahshan, J.
Tan, R. Hilgenfeld, J. Rademann, Angew. Chem. 2008, 120,
3319 – 3323; Angew. Chem. Int. Ed. 2008, 47, 3275 – 3278.
[8] A. Whitty, Nat. Chem. Biol. 2008, 4, 435 – 439.
Figure 4. The a-ketoaldehyde inhibitor 6b (component A in Figures 1
and 2; orange) was ligated with the active amine 8 (component B in
Figures 1 and 2; green) by reductive amination, yielding the picomolar
inhibitor 9 (red). Reaction conditions: a) SiCl3H (5 equiv), DMF/
MeOH 1:1 (v/v) with 1% AcOH, 6 h, RT over molecular sieves (4 A).
´
˚
[9] V. Uryga-Polowy, D. Kosslick, C. Freund, J. Rademann, Chem-
BioChem 2008, 9, 2452 – 2462.
[10] M. O. Hengartner, Nature 2000, 407, 770 – 776.
Inhibition constants (KI values) of 6b, 8, and 9 relative to caspase-3
were determined in a functional enzyme assay.
[11] D. W. Nicholson, Nature 2000, 407, 810 – 816.
[12] a) M. Garcia-Calvo, E. P. Peterson, B. Leiting, R. Ruel, D. W.
Nicholson, N. A. Thornberry, J. Biol. Chem. 1998, 273, 32608 –
32613; b) B. H. Han et al., J. Biol. Chem. 2002, 277, 30128 –
30136; c) D. Lee et al., J. Biol. Chem. 2000, 275, 16007 – 16014;
d) Y. Han et al., Bioorg. Med.Chem. Lett. 2005, 15, 1173 – 1180;
e) J. Sakai, A. Yoshimori, Y. Nose, A. Mizoroki, N. Okita, R.
Takasawa, S, Tanuma, Bioorg. Med. Chem. 2008, 16, 4854 –
4859; f) D. R. Goode, A. K. Sharma, J. Hergenrother, Org. Lett.
2005, 7, 3529 – 3532.
ure 3b) by using Equations (1) and (2) (for all equations, see the
Supporting Information). Binding data for various concentrations of 8
were then fitted with the aid of a set of four equilibrium equations
[Eqs. (3)–(6)] and three equations of mass conservation [Eqs. (7)–(9)]
in order to obtain the ligation constant, KC.
Received: March 25, 2009
Published online: July 23, 2009
[13] C. Godoy-Alcꢂntar, A. K. Yatsimirsky, J.-M. Lehn, J. Phys. Org.
Chem. 2005, 18, 979 – 985.
Keywords: caspase-3 · drug discovery ·
[14] a) A. El-Dahshan, S. Weik, J. Rademann, Org. Lett. 2007, 9, 949 –
952; b) S. Weik, T. Luksch, A. Evers, J. Bꢁttcher, A. Hasilik, H.-
G. Lꢁffler, G. Klebe, J. Rademann, ChemMedChem 2006, 1,
445 – 457; c) S. Weik, J. Rademann, Angew. Chem. 2003, 115,
2595 – 2598; Angew. Chem. Int. Ed. 2003, 42, 2491 – 2494.
.
fluorescence polarization · high-throughput screening ·
protease inhibitors
[1] Reviews: a) D. Fattori, A. Squarcia, S. Bartoli, Drugs R&D 2008,
9, 217 – 227; b) P. J. Hajduk, J. Greer, Nat. Rev. Drug Discovery
Angew. Chem. Int. Ed. 2009, 48, 6346 –6349
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 6349