Development of a fluorescent-tagged kinase assay system for the detection and characterization of allosteric kinase inhibitors

Article abstract of DOI:10.1021/ja902010p

Kinase disregulation disrupts the intricate network of intracellular signaling pathways and contributes to the onset of diseases such as cancer. Although several kinase inhibitors are on the market, inhibitor selectivity and drug resistance mutations persist as fundamental challenges in the development of effective long-term treatments. Chemical entities binding to less conserved allosteric sites would be expected to offer new opportunities for scaffold development. Because no high-throughput method was previously available, we developed a fluorescence-based kinase binding assay for identifying and characterizing ligands which stabilize the inactive kinase conformation. Here, we present a description of the development and validation of this assay using the serine/threonine kinase p38R. By covalently attaching fluorophores to the activation loop of the kinase, we were able to detect conformational changes and measure the K_d, k_on, and k_off associated with the binding and dissociation of ligands to the allosteric pocket. We report the SAR of a synthesized focused library of pyrazolourea derivatives, a scaffold known to bind with high affinity to the allosteric pocket of p38R. Additionally, we used protein X-ray crystallography together with our assay to examine the binding and dissociation kinetics to characterize potent quinazoline- and quinoline-based type II inhibitors, which also utilize this binding pocket in p38α. Last, we identified the b-Raf inhibitor sorafenib as a potent low nanomolar inhibitor of p38α and used protein X-ray crystallography to confirm a unique binding mode to the inactive kinase conformation.

Full text of DOI:10.1021/ja902010p

Published on Web 06/01/2009
Development of a Fluorescent-Tagged Kinase Assay System
for the Detection and Characterization of Allosteric Kinase
Inhibitors
Jeffrey R. Simard, Mattha¨us Getlik, Christian Gru¨tter, Vijaykumar Pawar,
Sabine Wulfert, Matthias Rabiller, and Daniel Rauh*
Chemical Genomics Centre of the Max Planck Society, Otto-Hahn-Strasse 15,
D-44227 Dortmund, Germany
Received March 15, 2009; E-mail: daniel.rauh@cgc.mpg.de
Abstract: Kinase disregulation disrupts the intricate network of intracellular signaling pathways and
contributes to the onset of diseases such as cancer. Although several kinase inhibitors are on the market,
inhibitor selectivity and drug resistance mutations persist as fundamental challenges in the development
of effective long-term treatments. Chemical entities binding to less conserved allosteric sites would be
expected to offer new opportunities for scaffold development. Because no high-throughput method was
previously available, we developed a fluorescence-based kinase binding assay for identifying and
characterizing ligands which stabilize the inactive kinase conformation. Here, we present a description of
the development and validation of this assay using the serine/threonine kinase p38R. By covalently attaching
fluorophores to the activation loop of the kinase, we were able to detect conformational changes and measure
the Kd, kon, and koff associated with the binding and dissociation of ligands to the allosteric pocket. We
report the SAR of a synthesized focused library of pyrazolourea derivatives, a scaffold known to bind with
high affinity to the allosteric pocket of p38R. Additionally, we used protein X-ray crystallography together
with our assay to examine the binding and dissociation kinetics to characterize potent quinazoline- and
quinoline-based type II inhibitors, which also utilize this binding pocket in p38R. Last, we identified the
b-Raf inhibitor sorafenib as a potent low nanomolar inhibitor of p38R and used protein X-ray crystallography
to confirm a unique binding mode to the inactive kinase conformation.
pocket.3,4 Such compounds would be expected to have improved
selectivity profiles and offer new opportunities for scaffold
development.5 However, approaches that allow for the unam-
biguous identification of such inhibitors fall short.6
1. Introduction
Aberrantly regulated kinases play causative roles in many
disorders such as cancer, diabetes, neurological, immunological,
and infectious diseases. Despite recent successes in kinase
inhibitor drug discovery and the dawn of targeted tumor therapy,
lack of inhibitor selectivity and efficacy, drug target validation,
and the emergence of kinase drug resistance remain key
challenges.1 This can be attributed to the fact that most kinase
inhibitors are ATP-competitive molecules, which form a critical
hydrogen bond with the hinge region of the kinase domain (type
I inhibitors) and often extend into the back of the ATP pocket
in the vicinity of the gatekeeper residue, an amino acid situated
at the back of the ATP pocket that is well-known for influencing
type I inhibitor affinity and selectivity profiles among kinases.2
Classical kinase inhibitors such as staurosporine and dasatinib
are examples of type I inhibitors. The most common mutations
occur at the gatekeeper position in the hinge region in which a
small amino acid side chain (classically Thr) is exchanged for
a larger hydrophobic residue (Ile or Met). Therefore, current
medicinal chemistry endeavors seek to overcome these limita-
tions by identifying new chemical entities that can bind “around”
these mutations or bind to less conserved sites outside the ATP
Kinases are typically found in the active, or DFG-in,
conformation, which was first characterized in Protein Kinase
A (PKA) with the activation loop open and extended, allowing
ATP and substrate molecules to bind.7 Structural changes in
the activation loop, resulting from a 180° flip of the highly
conserved DFG motif, result in an inactive, or DFG-out,
conformation in which the Phe of the DFG motif obstructs ATP
binding while exposing an adjacent allosteric site.5,8 Type II
inhibitors also bind to the hinge region and are ATP-competitive
but take advantage of the DFG-out conformation by additionally
extending into this allosteric site.5 The structural transition
associated with this event has been reported to increase the drug-
(3) Backes, A. C.; Zech, B.; Felber, B.; Klebl, B.; Mu¨ller, G. Expert Opin.
Drug DiscoVery 2008, 3, 1409–1425.
(4) Backes, A. C.; Zech, B.; Felber, B.; Klebl, B.; Mu¨ller, G. Expert Opin.
Drug DiscoVery 2008, 3, 1427–1449.
(5) Liu, Y.; Gray, N. S. Nat. Chem. Biol. 2006, 2, 358–64.
(6) Annis, D. A.; Nazef, N.; Chuang, C. C.; Scott, M. P.; Nash, H. M.
J. Am. Chem. Soc. 2004, 126, 15495–503.
(7) Knighton, D. R.; Zheng, J. H.; Ten Eyck, L. F.; Ashford, V. A.; Xuong,
N. H.; Taylor, S. S.; Sowadski, J. M. Science 1991, 253, 407–14.
(8) Pargellis, C.; Tong, L.; Churchill, L.; Cirillo, P. F.; Gilmore, T.;
Graham, A. G.; Grob, P. M.; Hickey, E. R.; Moss, N.; Pav, S.; Regan,
J. Nat. Struct. Biol. 2002, 9, 268–72.
(1) Zhang, J.; Yang, P. L.; Gray, N. S. Nat. ReV. Cancer 2009, 9, 28–39.
(2) Apsel, B.; Blair, J. A.; Gonzalez, B.; Nazif, T. M.; Feldman, M. E.;
Aizenstein, B.; Hoffman, R.; Williams, R. L.; Shokat, K. M.; Knight,
Z. A. Nat. Chem. Biol. 2008, 4, 691–9.
9
13286 J. AM. CHEM. SOC. 2009, 131, 13286–13296
10.1021/ja902010p CCC: $40.75  2009 American Chemical Society

Fluorescent-Tagged Kinase Assay System
A R T I C L E S
many). Crystallization plates (EasyXtal Tool; 24-well) were
obtained from Qiagen GmbH (Hilden, Germany). Cuvettes and mini
stir bars were obtained from Carl Roth GmbH (Karlsruhe, Ger-
many).
target residence time and provide efficacy at lower drug
concentrations, thereby augmenting the therapeutic window.9
The pyrazolourea BIRB-796, a selective inhibitor of p38R
(developed by Boehringer-Ingelheim)8,10,11 and imatinib
(Gleevec), an Abl/cKit/PDGFR inhibitor,12-14 are well-known
examples of type II inhibitors. Additionally, a number of high
affinity inhibitors that bind exclusively to the allosteric site of
p38R are referred to as type III inhibitors. Such compounds
also stabilize the DFG-out conformation and recently served
as excellent starting points for the design and synthesis of type
II inhibitors, which are able to overcome mutation-associated
drug resistance in cSrc kinase.15
Considering the less-conserved nature of the amino acids
lining the allosteric pocket, one would expect methods for
discriminating between DFG-in and DFG-out binders to become
valuable commodities to guide innovative compound design.
Because insights into the stabilization of the DFG-out confor-
mation by type II/III inhibitors were only demonstrated by
protein X-ray crystallography nearly a decade ago with the
development of imatinib,14,16 currently available methods are
few in number.6,17 Additionally, there does not appear to be
any high-throughput screening (HTS) method available, which
can screen exclusively for DFG-out binders, that is rapid, robust,
and easy to analyze. Therefore, we set out to develop a novel
fluorescence-based kinase binding assay using p38R as a model
for kinases regulated by the equilibrium between DFG-in and
DFG-out conformations.
Here, we present our efforts to establish fluorescent-labeled
p38R as a highly sensitive and reliable assay system for
screening and identifying new DFG-out stabilizing scaffolds.
More specifically, we describe how kinetic measurements such
as kon and koff and direct measurements of Kd are achieved for
DFG-out binders (types II and III) using this approach.
Facilitated by this system, we were able to identify sorafenib
as a high affinity binder of p38R and confirmed a unique type
II inhibitor binding mode using protein X-ray crystallography,
further highlighting the power of this approach for screening
and discriminating compounds, which take advantage of the less-
conserved allosteric binding pocket in some kinases.
2.2. Protein Expression, Purification, and Labeling. The
human p38R MAP kinase construct containing the mutations
required for specific labeling (C119S/C162S/A172C) was cloned
into a pOPINF vector and was transformed as an N-terminal His-
tag construct with PreScission Protease cleavage site into BL21(DE3)
E. coli. Cultures were grown at 37 °C until an OD600 of 0.6, cooled
in 30 min to room temperature, and then induced with 1 mM IPTG
for overnight (∼20 h) expression at 18 °C while shaking at 160
rpm. Cells were lysed in buffer A (50 mM Tris pH 8.0, 500 mM
NaCl + 5% glycerol + 25 mM imidazole) and loaded onto a 30
mL Ni-column (self-packed), washed with 3 CV of Ni buffer A,
and then eluted with a 0-50% linear gradient using Ni buffer B
(Ni buffer A + 500 mM imidazole) over 2 CV. The protein was
cleaved by incubating with PreScission Protease (50 µg/mL final
concentration) in a 12-30 mL capacity 10k-MWCO dialysis
cassette (Thermo Scientific) overnight at 4 °C in dialysis buffer
(50 mM Tris pH 7.5, 5% glycerol, 150 mM NaCl, 1 mM EDTA,
1 mM DTT). The protein was then centrifuged for 15 min at
∼13 000 rpm to remove any precipitate that may have formed
during the cleavage step. The supernatant was then taken and diluted
4-fold in anion buffer A (50 mM Tris pH 7.4, 5% glycerol, 50
mM NaCl, 1 mM DTT), loaded onto a 1 mL Sepharose Q FF
column (GE Healthcare) and washed with 10 CV of anion buffer
A. The protein was eluted with a 0-100% linear gradient of anion
buffer B (anion buffer A + 600 mM NaCl) over 20 CV. The protein
was pooled and concentrated down to 2 mL and passed through a
Sephadex HiLoad 26/60 Superdex 75 column equilibrated with size
exclusion buffer (20 mM Tris pH 7.4, 5% glycerol, 200 mM NaCl,
1 mM DTT) at a rate of 2 mL/min. The eluted protein was then
concentrated to ∼10 mg/mL, aliquoted, and frozen at -80 °C.
Protein and a thiol-reactive fluorophore of choice (dissolved in
DMF or DMSO) were combined in 20 mM HEPES buffer (pH
7.0) at a 1:1.5 ratio, respectively, and allowed to react in the dark
overnight at 4 °C. The % v/v DMF was kept <0.5% during the
conjugation. Conjugated protein was subsequently concentrated,
washed three times in a 10k-MWCO Centricon using 50 mM
HEPES (pH 7.45) + 200 mM NaCl, aliquoted, and frozen at -80
°C. Monolabeling of the protein was verified by ESI-MS and mass
spectroscopic analysis of tryptic fragments (Figure S1).
2.3. Fluorescence Characterization of Fluorescent p38r
Conjugates. Fluorophore-labeled p38R conjugates were excited at
the known wavelength of each fluorophore, and emission spectra
were recorded every 30 s in the absence and presence of saturating
concentrations of BIRB-796 (DFG-out binder), to determine how
the emission changes as the kinase shifts from the DFG-in to the
DFG-out conformation. These measurements also revealed the most
sensitive wavelength to monitor fluorescence changes in real-time
(kinetics) and to assess the feasibility of using ratiometric fluores-
cence to plot binding curves (Kd). The emission spectra of each
fluorescent-p38R conjugate (unbound and DFG-out saturated) were
used to calculate the fluorescence parameters ∆Istd and ∆Rmax as
described elsewhere.18 For most fluorophores, excitation and
emission slits were set to 3 and 5 nm, respectively. In the case of
pyrene, excitation and emission slits were set to 3 and 10 nm,
respectively. For NBD, slits were set to 5 and 20 nm, respectively.
2.4. End Point and Kinetic Measurements. Fluorescent-labeled
p38R (50 nM) and various concentrations of inhibitor (1 nM to 20
µM) were incubated in the dark at 4 °C overnight before end point
fluorescence measurements were carried out in either polystyrene
cuvettes or 96-well plates to determine the Kd of each compound.
The same buffer that was used to wash and store the kinase was
also used for all measurements. The long incubation times were
2. Materials and Methods
2.1. Materials. The fluorophores N-((2-(iodoacetoxy)ethyl)-N-
methyl)amino-7-nitrobenz-2-oxa-1,3-diazole (IANBD ester), 5-io-
doacetamidofluorescein (5-IAF), N-(1-pyrene)iodoacetamide,
5-((((2-iodoacetyl)amino)ethyl)amino)naphthalene-1-sulfonic acid
(1,5-IAEDANS), and 6-acryloyl-2-dimethylaminonaphthalene (acry-
lodan) were purchased from Invitrogen GmbH (Karlsruhe, Ger-
(9) Copeland, R. A.; Pompliano, D. L.; Meek, T. D. Nat. ReV. Drug
DiscoVery 2006, 5, 730–9.
(10) Regan, J. J. Med. Chem. 2002, 45, 2994–3008.
(11) Regan, J.; Pargellis, C. A.; Cirillo, P. F.; Gilmore, T.; Hickey, E. R.;
Peet, G. W.; Proto, A.; Swinamer, A.; Moss, N. Bioorg. Med. Chem.
Lett. 2003, 13, 3101–4.
(12) Atwell, S. J. Biol. Chem. 2004, 279, 55827–32.
(13) Capdeville, R.; Buchdunger, E.; Zimmermann, J.; Matter, A. Nat. ReV.
Drug DiscoVery 2002, 1, 493–502.
(14) Nagar, B.; Bornmann, W. G.; Pellicena, P.; Schindler, T.; Veach, D. R.;
Miller, W. T.; Clarkson, B.; Kuriyan, J. Cancer Res. 2002, 62, 4236–
43.
(15) Getlik, M.; Gru¨tter, C.; Simard, J. R.; Klu¨ter, S.; Rabiller, M.; Rode,
H. B.; Robubi, A.; Rauh, D. J. Med. Chem. in press.
(16) Schindler, T.; Bornmann, W.; Pellicena, P.; Miller, W. T.; Clarkson,
B.; Kuriyan, J. Science 2000, 289, 1938–42.
(17) Vogtherr, M.; Saxena, K.; Hoelder, S.; Grimme, S.; Betz, M.;
Schieborr, U.; Pescatore, B.; Robin, M.; Delarbre, L.; Langer, T.;
Wendt, K. U.; Schwalbe, H. Angew. Chem., Int. Ed. 2006, 45, 993–7.
(18) de Lorimier, R. M.; Smith, J. J.; Dwyer, M. A.; Looger, L. L.; Sali,
K. M.; Paavola, C. D.; Rizk, S. S.; Sadigov, S.; Conrad, D. W.; Loew,
L.; Hellinga, H. W. Protein Sci. 2002, 11, 2655–75.
9
J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009 13287

A R T I C L E S
Simard et al.
required to avoid the Kd time dependence of type II inhibitors
binding to p38R.8 In the cuvette format, a series of cuvettes
containing different amounts of inhibitor were prepared using
inhibitor stocks (0.01, 0.1, 1.0, and 10.0 mM) prepared in DMSO.
The % v/v DMSO did not exceed 0.2% in cuvettes. All measure-
ments of the cuvettes were made with a JASCO FP-6500
fluorescence spectrophotometer (JASCO GmbH, Groꢀ-Umstadt,
Germany). A Tecan Safire2 (Tecan Deutschland GmbH, Crailsheim,
Germany) was used to measure the fluorescence read-out in the
96-well plate format.
For acrylodan- and IAEDANS-labeled p38R, ratiometric fluo-
rescence values enabled reliable binding curves to be plotted to
directly determine the Kd of ligand binding to p38R. Where
indicated, ratiometric fluorescence values obtained for unsaturated
and p38R saturated in the DFG-out conformation were plotted as
%p38R bound by the ligand. The %p38R bound is calculated as
follows:
different inhibitor concentrations was fit to a Hill four-parameter
equation to determine IC50 values. Each reaction was performed in
duplicate, and at least three independent determinations of each
IC50 were made. To determine Km values for ATP, initial velocities
of kinase reaction were determined for eight different ATP
concentrations ranging from 15 to 0 mM using 3-fold dilution steps
andfourtimepoints.ThesevelocitieswerefittedtotheMichaelis-Menten
equation (V ) Vmax*S/(Km + S)) using the XLfit add-on for
Microsoft Excel to yield the values for Km and Vmax, the maximal
velocity for the reaction catalyzed by the given amount of kinase.
2.6. Compound Synthesis. Details on the synthesis and char-
acterization of a focused library of type III pyrazolourea and type
II quinoline- and quinazoline-pyrazolourea hybrid inhibitors are
provided in the Supporting Information.
2.7. Protein Crystallography. Various inhibitors were cocrys-
tallized with wild-type p38R using conditions similar to those
previously reported for unmodified p38R.20 Briefly, protein-inhibitor
complexes were prepared by mixing 30 µL of p38R (10 mg/mL)
with 0.3 µL of inhibitor (100 mM in DMSO) and incubating the
mixture for 1-2 h on ice. Samples were centrifuged at 13 000 rpm
for 5 min to remove excess inhibitor. Crystals were grown in 24-
well crystallization plates using the hanging drop vapor diffusion
method and by mixing 1.5 µL of protein-inhibitor solution with
0.5 µL of reservoir (100 mM MES pH 5.6-6.2, 20-30% PEG4000,
and 50 mM n-octyl-ꢀ-D-glucopyranoside).
% bound ) ((R -Runsat)/Rsat’d) × 100
where R is the ratiometric fluorescence at a given concentration of
ligand, and Runsat and Rsat’d are the ratiometric fluorescence values
obtained for p38R in the absence or presence of a saturating
concentration of the same ligand, respectively.
Rate constants for the association and dissociation of allosteric
binders from p38R were determined using the cuvette format while
monitoring fluorescence changes in real-time. A mini stir bar was
placed in the bottom of each cuvette to ensure rapid mixing as
inhibitor was delivered through the injection port located above
the cuvette. The dissociation rate of allosteric binders was measured
by adding a 10-fold excess of unlabeled p38R to a suspension of
fluorophore-labeled p38R (0.1 µM) complexed with an equimolar
amount of inhibitor while monitoring the fluorescence in real-time.
The addition of excess unlabeled kinase causes the inhibitor to
redistribute, resulting in a net dissociation of inhibitor from labeled-
p38R, which is representative of the true koff of the ligand.19 The
association rate of allosteric binders was measured by delivering
increasing molar equivalents of compound (1-4:1 compound:
protein), as demonstrated elsewhere,19 to a suspension of fluoro-
phore-labeled p38R (0.1 µM) while monitoring fluorescence
changes in real-time. In the case of both association and dissociation,
fluorescence changes were fit to a first-order function. The observed
rate constant (kobs) of the fluorescence change for the binding of a
specific dose of compound was plotted against the inhibitor
concentration, and the result is a straight line, which was fit linearly
(R2 > 0.99) with a slope equal to the kon of that particular ligand.
For all fluorescent p38R conjugates, association and dissociation
rate constants could be used to calculate the equilibrium constant
2.8. Data Deposition. All crystal structures described were
deposited into the Protein Data Bank (PDB codes: 3GCP, 3GCQ,
3GCS, 3GCU, 3GCV).
3. Results
3.1. Selection of the Labeling Position on the Activation
Loop. High-resolution crystal structures of human p38R in
complex with type I, type II, or type III inhibitors have provided
a wealth of structural detail about both the active and the inactive
conformation, led to the deposition of 90 complex structures
into the protein data bank (PDB), and have spurred the design
of diverse inhibitor classes. This wealth of structural knowledge
also made p38R an ideal candidate for the development of a
fluorescence-based direct binding assay that allows for the
detection and characterization of inhibitors, which bind to the
DFG-out conformation. The principle of the fluorescent-labeled
p38R assay for allosteric binders is highlighted in Figure 1a.
The fluorophores we employed for labeling are commonly used
to generate fluorescent protein conjugates, are relatively small
in size, and are sensitive to polarity and/or conformational
changes. Acrylodan is known to produce a typically robust
response18,21 and was expected to be especially sensitive to
movements of the activation loop upon binding of DFG-out
binders.
Prior to labeling the kinase, we first analyzed several crystal
structures of p38R in the DFG-in and DFG-out conformations
to identify the most suitable fluorophore attachment site in the
activation loop. Care was taken to choose an amino acid position
that exhibits significant movement upon ligand binding and is
solvent-exposed to enable the attachment of a fluorophore added
in solution. Sequence alignments of the activation loop of several
human kinases (Figure 1b) were also used to avoid mutating
common phosphorylation sites (and adjacent or nearby residues)
or highly conserved regions or positions known to be involved
(Keq) for the binding and dissociation of 100 nM BIRB-796 (Keq
koff/kon).
)
2.5. ATP Km and IC50 Determination with an in Vitro
Activity-Based Assay. IC50 determinations for p38R kinase were
measured with the HTRF KinEASE assay from Cisbio according
to the manufacturer’s instructions in which a GST-tagged ATF-2
peptide was phosphorylated by p38R. After completion of the
reaction, an anti[phospho-ATF-2] antibody labeled with Europium
Cryptate and an anti-GST antibody labeled with the fluorophore
D2 were added. The FRET between Europium Cryptate and D2
was measured to quantify the phosphorylation of ATF-2. ATP
concentrations were set at the Km value for wild type, mutated/
unlabeled, and the mutated/labeled p38R. Kinase and inhibitor were
preincubated for 2 h before the reaction was started by addition of
ATP and ATF-2. A Tecan Safire2 plate reader was used to measure
the fluorescence of the samples at 620 nm (Eu-labeled antibody)
and 665 nm (D2-labeled antibody) 60 µs after excitation at 317
nm. The quotient of both intensities for reactions made with eight
(20) Bukhtiyarova, M.; Northrop, K.; Chai, X.; Casper, D.; Karpusas, M.;
Springman, E. Protein Expression Purif. 2004, 37, 154–61.
(21) Richieri, G. V.; Ogata, R. T.; Kleinfeld, A. M. Mol. Cell. Biochem.
1999, 192, 87–94.
(19) Hibbs, R. E.; Talley, T. T.; Taylor, P. J. Biol. Chem. 2004, 279, 28483–
(22) Thompson, J. D.; Gibson, T. J.; Higgins, D. G. Curr. Protoc.
Bioinformatics 2002, 23.
91.
9
13288 J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009

Fluorescent-Tagged Kinase Assay System
A R T I C L E S
Figure 1. Schematic diagram of the fluorescent kinase binding assay. (a) Kinases such as p38R exist in both inactive (left) and active (right) states depending
on the orientation of the DFG motif (orange sticks) and the activation loop (red). The binding of type II/III (represented by blue surface) and type I ligands
(represented by yellow surface) can stabilize either the inactive DFG-out or the active DFG-in conformations, respectively, and induce conformational
changes, which can be detected by different fluorophores (differently colored spheres). (b) Schematic representation of the binding mode of the type II
inhibitor BIRB-796 in p38R (PDB code: 1kv2) highlights the different regions of the binding site utilized by different inhibitor types. (c) Sequence alignment
of several kinase activation loops guides labeling of the DFG+1 and DFG+2 positions as the most suitable for labeling the kinase (red text). Regions that
are highly conserved or crucial to kinase structural stability or enzymatic activity are shown (boxed regions). Alignments were performed using Clustal W.22
in key structural interactions that are crucial to kinase regulation.
The activation loop sequence is bookended by the highly
conserved DFG (box 1) and APE motifs (box 6). The DFG+3
position is commonly a basic amino acid, which interacts
directly with the primary phosphorylation site of the activation
loop.23 The DFG+3 through DFG+5 (box 2) serves as a
hydrophobic anchor point with other structural features of the
C-lobe in tyrosine kinases.24 This is followed by a variable
length segment (box 3-4) contains a high incidence of Tyr,
Ser, and Thr residues, which can be phosphorylated. The
C-terminal end of the activation loop (box 5) forms several
interactions with the C-lobe of the kinase and is important in
substrate binding.
We found that residues immediately following the DFG motif
(DFG+1 and DFG+2) at the N-terminal end of the activation
loop exhibit significant movement with conformational changes
and are typically not associated with disease-related genetic
alterations known to influence kinase activity.25 Additionally,
sequence alignments of all human kinases reveal that at least
47 kinases have a naturally occurring Cys in these two positions,
suggesting that mutation of the labeling site residue to Cys
would likely be tolerated. Therefore, Ala172 of p38R (DFG+2
position) was subsequently mutated into a Cys for specific
reaction with thiol-reactive fluorophores. We also used available
structural information to identify solvent-exposed Cys, which
could undesirably react with the fluorophore. We determined
that only two of the four Cys in p38R were solvent-exposed
(Cys119 and Cys162) and subsequently mutated these into Ser
to increase the probability that the kinase would be singly
labeled only on the activation loop, which was ultimately
confirmed by mass spectrometry methods (Figure S1). In the
case where no structures are available for a kinase of interest,
a variety of online modeling tools can be used to generate DFG-
in and DFG-out conformational models for the same amino acid
sequence (Supporting Information). Additionally, the use of
published screening data of various inhibitor types against a
panel of kinases26 may be a useful tool for assessing whether
or not the kinase of interest is capable of adopting the DFG-
out conformation, which is necessary for the assay to succeed.
3.2. Fluorescence Characterization of Fluorescent-p38r
Conjugates. The newly introduced Cys at position 172 of p38R
was labeled with various thiol-reactive fluorophores reported
to be sensitive to the polarity of their surrounding microenvi-
ronment. The fluorescent properties of each labeled kinase were
characterized by inducing the DFG-out conformation using
the well-known p38R type II inhibitor, BIRB-796. Using the
fluorescence spectra obtained for labeled p38R in the presence
and absence of a saturating concentration of BIRB-796, we
calculated ∆Istd, the normalized intensity change as compared
to average intensity, and ∆Rmax, the maximal standard intensity
change at two wavelengths (Supporting Information). These
parameters are two of the most important criteria for identifying
ideal fluorophore-protein conjugates for fluorescence studies.
(23) Nolen, B.; Taylor, S.; Ghosh, G. Mol. Cell 2004, 15, 661–75.
(24) Levinson, N. M.; Seeliger, M. A.; Cole, P. A.; Kuriyan, J. Cell 2008,
134, 124–34.
(25) Torkamani, A.; Kannan, N.; Taylor, S. S.; Schork, N. J. Proc. Natl.
Acad. Sci. U.S.A. 2008, 105, 9011–6.
(26) Karaman, M. W. Nat. Biotechnol. 2008, 26, 127–32.
9
J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009 13289

A R T I C L E S
Simard et al.
constructs were observed for each inhibitor (Table S1). Thus,
although the ATP Km was increased significantly in the mutated
kinase constructs used for labeling, these changes did not
correlate with significant changes to inhibitor activity or affinity.
3.3. End Point and Real-Time Measurements: Determina-
tion of Kd, kon, and koff. End point measurements of ac-p38R
using a dual-wavelength ratio (R ) 514 nm/468 nm) allowed
for direct and reliable Kd determination of type III and type II
DFG-out binders. At each inhibitor concentration, the emission
spectrum was measured (Figure 3a), and R was either plotted
to show the saturation of ac-p38R in the inactive state (Figure
3b) or plotted on a logarithmic scale to directly determine the
Kd (Figure 3c). Similar types of measurements were also possible
with IAEDANS-p38R (Figure S2).
Figure 2. Thiol-reactive fluorophores tested in the fluorescent kinase assay.
Several fluorophores were conjugated to A172C of p38R, and their changing
fluorescence properties were examined upon binding of BIRB-796, a known
DFG-out binder of p38R. All values for ∆Rmax and ∆Istd that meet the criteria
deemed ideal fluorophore-protein conjugates18 appear in bold. The superior
∆Rmax of acrylodan is the result of a ∼45 nm shift in emission maxima in
the DFG-out conformation. IAEDANS, a structural analogue of acrylodan,
does not exhibit a large emission shift, but there is an increase in emission
at ∼515 nm relative to ∼470 nm, allowing reliable binding curves to be
measured despite the suboptimal ∆Rmax. Pyrene and fluorescein are
considerably more bulky than the other fluorophores and appear to enhance
BIRB-796 dissociation rates, resulting in higher calculated equilibrium
constants (Keq) for 100 nM BIRB-796 under these experimental conditions.
[Note: The chemical structure of Atto680 has not been released by the
manufacturer (http://www.innovabiosciences.com).]
A significant change of emission at 468 nm also allowed for
the possibility of studying the kinetics of dissociation (koff) and
association (kobs) in real-time for different concentrations of
ligand. Following addition of inhibitor, the fluorescence emission
at 468 nm decreased in a dose-dependent manner and was nicely
fit to a first-order kinetic (Figure 3d). These types of experiments
yielded rate constants (kobs) that were then plotted and fit linearly
to obtain the kon for the ligand (Figure 3e), which we found to
be (4.3 ( 0.8) × 103 M-1 s-1 (n ) 3) for BIRB-796.
Alternatively, addition of a 10-fold molar excess of unlabeled
p38R to ac-p38R prebound with ligand resulted in extraction
of BIRB-796, a first-order increase in fluorescence, and direct
determination of koff (Figure 3f), which we found to be (5.1 (
0.5) × 10-5 s-1 (n ) 3). Both kinetic values differ by a factor
of 10 from those published elsewhere for BIRB-796 using
alternative methods.8 Determination of kon and koff also allows
for the indirect/calculated determination of Kd values (Kd ) koff/
kon) and sheds light on the factors contributing to different ligand
affinities. Such rate measurements were possible with all
fluorescent-p38R conjugates tested in this study (Figure S3).
Last, these measurements demonstrate the reversibility of the
fluorescence response and demonstrate the changing equilibrium
that exists between the DFG-in and DFG-out conformations.
In experiments analogous to those shown in Figure 3d-f for
BIRB-796, we used ac-p38R to determine the kon and koff for
the smaller type III pyrazolourea analogue 1 (structure shown
in Figure 4) to be (6.6 ( 1.2) × 103 M-1 s-1 (n ) 3) and (7.1
( 3.2) × 10-3 s-1 (n ) 3), respectively. Both values are different
from previously published values by 2 orders of magnitude.8
The differences in measured kinetic values for BIRB-796 and
1 are likely a consequence of the different assay systems
employed in each study. In the studies reported by Pargellis et
al., the dissociation of these compounds from p38R was
measured by the binding of an ATP-competitive inhibitor of
p38R, SKF86002, which becomes highly fluorescent upon
binding. The binding of SKF86002 shifts the kinase more toward
the DFG-in conformation, resulting in the dissociation of
pyrazolourea compounds from the adjacent allosteric site as the
DFG Phe flips back toward the allosteric pocket. Conversely,
the displacement of prebound SKF86002 was used to measure
the binding of pyrazoloureas. However, it is possible that the
displacement of low nanomolar affinity ligands such as BIRB-
796 (Kd ) 0.1 nM8) by the weaker inhibitor SKF86002 (Kd )
180 nM8) is not an optimal approach and may result in slower
apparent koff values and lower apparent Kd values. Thus, Kd
values determined in this manner may be more subject to the
experimental conditions under which the rate constants are
determined.
Ideally, values for ∆Istd should be >0.25 and values for ∆Rmax
should be >1.25.18 According to these criteria, acrylodan-labeled
p38R (ac-p38R) is the most ideal probe for a fluorescence-based
assay for detecting DFG-out binders (Figure 2).
Ac-p38R allows for ratiometric measurements because DFG-
out binders induce a shift in the emission maximum from 468
to 514 nm, indicative of the movement of acrylodan from a
less polar to a more polar environment.19,27 This feature is
advantageous because ratiometric fluorescence measurements
correct for dilution and pipetting errors between different
samples in a titration series and produce reliable end point
measurements. Although all other fluorophores examined showed
a change in fluorescence upon ligand binding, they did not
produce reliable binding curves (IAEDANS was an exception).
However, all fluorescent kinase conjugates were acceptable with
respect to ∆Istd and were suitable for making sensitive kinetic
measurements of ligand binding and dissociation (Supporting
Information and Figures S2 and S3).
In a final step, we characterized the kinetic properties of the
wild type, mutated/unlabeled, and mutated/acrylodan-labeled
p38R to determine the effects of the mutations and labeling on
the activity of the kinase. In general, we found a 25-fold
reduction in ATP Km in the mutated kinase with no additional
reduction in the ATP Km with the attachment of an acrylodan,
suggesting that the mutations do disrupt the affinity of p38R.
Although it is not surprising that alterations to the activation
loop would have some effect on kinase activity, these manipula-
tions did not result in a kinetically dead kinase as we
demonstrated previously for cSrc kinase.28 IC50 values were then
determined for type I and type II inhibitors in an activity-based
assay, and only minor differences between the different kinase
(27) Richieri, G. V.; Ogata, R. T.; Kleinfeld, A. M. J. Biol. Chem. 1992,
267, 23495–501.
(28) Simard, J. R.; Kluter, S.; Grutter, C.; Getlik, M.; Rabiller, M.; Rode,
H. B.; Rauh, D. Nat Chem Biol. 2009, 5, 394-6.
9
13290 J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009

Fluorescent-Tagged Kinase Assay System
A R T I C L E S
Figure 3. Real-time and end point fluorescence measurements using ac-p38R labeled on the activation loop. (a) Acrylodan emission at 468 nm decreases
upon binding of BIRB-796, a known DFG-out binder of p38R, resulting in a red-shift of the maximum emission wavelength in the bound state. End point
equilibrium measurements can be made to directly obtain the Kd. (b) Ratiometric fluorescence data (R ) 514 nm/468 nm) were plotted to show the saturation
of the inactive state or (c) on a logarithmic scale of inhibitor concentration to obtain the Kd. Fluorescence traces can also be measured in real-time at a single
wavelength (468 nm) to determine various kinetic rate constants. (d) The fluorescence decay resulting from the addition of different amounts of BIRB-796
(large arrow) was fit to a first-order decay function to obtain kobs. (e) Experimentally determined kobs values can be plotted to determine kon for BIRB-796.
(f) Extraction of BIRB-796 from ac-p38R using an excess of unlabeled p38R allowed direct determination of koff, which was also fit to a first-order function.
The data presented above are representative of a typical set of experiments carried out for BIRB-796 and other DFG-out binders using ac-p38R.
The optimal method for accurately measuring the affinity of
any ligand to a protein is to directly measure the formation of
the ligand-protein complex as demonstrated with our approach.
However, we were still able to observe that the Kd values of
different pyrazoloureas were more influenced by koff rather than
kon, an effect that is well-documented in the case of p38R.8
Furthermore, we were able to validate the conditions used to
measure our kinetic rate constants by using them to calculate
Kd values of 11.9 ( 1.3 nM and 1.079 ( 0.347 µM for BIRB-
796 and 1, respectively. These values compare well with the
direct Kd values determined directly using end point measure-
ments in which values of 7.5 ( 1.9 nM (n ) 3) and 1.19 (
0.14 µM (n ) 3) for BIRB-796 and 1 were measured,
respectively.
3.4. SAR of a Focused Pyrazolourea Compound Library. We
designed and generated a focused library of pyrazoloureas, a
class of compounds whose pharmacophore and binding mode
are known in p38R, and used several new derivatives of this
scaffold to examine structure-activity relationships (SAR) and
characterize the fluorescence response of ac-p38R. Pyrazoloureas
represent one of the prototypes for type III and type II kinase
inhibitors. Type III pyrazoloureas not only stimulated the
development of the former clinical candidate BIRB-79610 but
also provided a wealth of structural, binding, and dissociation
data, allowing for comparison of Kd values determined here
using ac-p38R (Figure 4) with other approaches.8,11,29 Several
of these known compounds were also synthesized to serve as a
measuring stick for our fluorescent-tagged kinase binding assay.
As expected, the Kd of 6 (BIRB-796) was time-dependent
(Figure S4) as reported elsewhere.8 All type III inhibitors also
showed this time dependence, but required less time (2-4 h)
to reach equilibrium with p38R. Thus, long preincubation times
are necessary to ensure complete binding of type II and type
III ligands before making end point measurements. In general,
we found excellent agreement between our Kd values and those
reported elsewhere for known compounds using other ap-
proaches, with the largest differences occurring for compounds
with a published Kd of <10 nM. As discussed above for BIRB-
796, the affinity of DFG-out binders to p38R is dictated
primarily by koff. Therefore, we believe that the calculated Kd
values reported in the literature for the strongest binders are
more subject to the experimental conditions under which koff
was measured and have been shown to vary significantly
depending on the methods used.29 However, certain trends in
(29) Kroe, R. R.; Regan, J.; Proto, A.; Peet, G. W.; Roy, T.; Landro, L. D.;
Fuschetto, N. G.; Pargellis, C. A.; Ingraham, R. H. J. Med. Chem.
2003, 46, 4669–75.
9
J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009 13291

A R T I C L E S
Simard et al.
Figure 4. Measured Kd values and SAR of a focused library of pyrazoloureas. The synthesis and characterization of all pyrazolourea derivatives are described
in the Supporting Information. All compounds were titrated against ac-p38R (50 nM) over a concentration range of 1 nM to 20 µM to generate binding
curves using the ratiometric fluorescence change (R ) 514 nm/468 nm) observed upon binding to the DFG-out conformation of p38R. Kd values for each
compound were then directly obtained from the binding curves. All reported values represent the mean ( s.d. from at least three independent titrations.
the SAR are always maintained with respect to aryl moiety I
(substituted phenyl attached to the urea) and aryl moiety II
(substituted phenyl attached to the pyrazole) regardless of the
assay system used to make the measurements.
approaches.10 Given the hydrophobic characteristics of this
subpocket, it is also not surprising to see a significant loss
of affinity for the more polar 4-aminophenyl 7 or 3-ami-
nophenyl 8 type III derivatives.
In the case of aryl moiety I, we observed a noticeable
affinity ranking such that 3 (phenyl) < 2e (4-chlorophenyl)
<4b(naphthyl), aSARtrendthatiswell-documented.8,10,11,29,30
Moiety I fits into a small hydrophobic subpocket behind the
gatekeeper residue of p38R, and the bulkier naphthyl moiety
forms better lipophilic interactions as a result of its ability
to penetrate deeper into this subpocket. It could also be
argued that more solvation entropy is gained (release of more
water molecules) upon burial of the bulkier naphthyl in this
subpocket.31 Thus, a phenyl moiety alone at this position
does not contribute significantly to the affinity of the
compound, explaining the much higher Kd of 3 in comparison
to 2e and 4b. This is further confirmed by 5, which has a
central phenyl in place of the bulkier naphthyl of 6 (BIRB-
796), resulting in a 50-fold higher Kd. Similar findings for
these compounds were also observed elsewhere using other
The addition of a monosubstituted aryl moiety II to the N1
of the pyrazole of 1 extends the ligand into the allosteric pocket
and results in a significant increase in the affinity. This phenyl
ring shows a distinct preference for substituents in the para and
meta positions in our assay, a SAR trend that is also well-
documented.10 To further investigate the importance of the
substitution pattern of moiety II, we used the crystal structure
of p38R in complex with BIRB-796 (PDB-code: 1kv2) together
with in silico modeling to design and synthesize a bulkier
inhibitor 9 (RL19). We hypothesized that the combination of a
4-trifluoromethyl and 2,6-dichloro substitution pattern on the
phenyl ring would prevent the compound from fitting into the
allosteric pocket, which was confirmed by the complete lack
of fluorescence response from ac-p38R.
Recently, we reported the development of several type II
quinazoline-pyrazolourea hybrid inhibitors of cSrc kinase.15
In the case of cSrc, smaller type III pyrazoloureas were found
to inhibit cSrc with mid micromolar IC50 values. In a manner
similar to the development of BIRB-796 from 1,10 the fusion
of these compounds with a quinazoline scaffold, which are also
(30) Regan, J.; Capolino, A.; Cirillo, P. F.; Gilmore, T.; Graham, A. G.;
Hickey, E.; Kroe, R. R.; Madwed, J.; Moriak, M.; Nelson, R.; Pargellis,
C. A.; Swinamer, A.; Torcellini, C.; Tsang, M.; Moss, N. J. Med.
Chem. 2003, 46, 4676–86.
(31) Lafont, V.; Armstrong, A. A.; Ohtaka, H.; Kiso, Y.; Mario Amzel,
L.; Freire, E. Chem. Biol. Drug. Des. 2007, 69, 413–22.
9
13292 J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009

Fluorescent-Tagged Kinase Assay System
A R T I C L E S
micromolar inhibitors of cSrc,32 resulted in potent low nano-
molar type II inhibitors that extend into the ATP binding site
to interact with the hinge region. Furthermore, we showed that
the 1,4-para fused hybrids 10a, 10b, and 10d were better binders
to cSrc while using molecular modeling to predict that p38R
would better accommodate the 1,3-meta fused inhibitors 10g
and 10h. Using ac-p38R, we were able to confirm this
hypothesis and found that 10g and 10h have a 6-fold higher
affinity over 10a and 10b. We solved the crystal structure of
10h (RL62) and 10b (RL45) each in complex with p38R and
observed additional stabilizing interactions, which explain why
1,3-meta hybrids have a higher affinity for p38R (Figure S5).
We also performed an extensive analysis of the kinetic rate
constants for these compounds using ac-p38R (Table S2) and
found that the dissociation rate of 10h is slower than that of
10b, while both compounds exhibit similar kobs for binding
((1.40 ( 0.25) × 10-3 s-1 (n ) 3) for 10h and (1.36 ( 0.13)
× 10-3 s-1 (n ) 3) for 10b, respectively), which may partially
account for the higher affinity of 10h. Last, these detailed kinetic
and structural characterizations of 1,3- and 1,4-fused hybrid
compounds led us to design and synthesize several pyra-
zolourea-quinoline analogues as compounds that are potent type
II inhibitors of p38R. Further details regarding the kinetic and
structural characterization of these and additional hybrid
compounds can be found in the Supporting Information.
3.5. Binding of Alternative Drug Scaffolds. Following our
SAR studies of several pyrazolourea compounds, we investi-
gated a number of commercially available type I and type II
kinase inhibitors to confirm that our new assay system was not
dependent on the pyrazolourea scaffold. Marketed drugs such
as lapatinib (Tykerb) and imatinib (Gleevec), selective potent
type II inhibitors of HER2 and Abl kinases, respectively, and
the promiscuous type I inhibitor staurosporine did not trigger a
significant fluorescence change (Figure S6A). Further screening
revealed that the addition of sorafenib (Nexavar), a well-known
b-Raf and VEGFR2 inhibitor, produced a strong fluorescence
response and binds to ac-p38R with a Kd of 56 ( 5 nM (n )
3), which is within range of the published Kd against its intended
kinase targets.33 The Kd of sorafenib was also found to be time-
dependent similar to other type II inhibitors. In a recent
publication where 38 known kinase inhibitors were screened
against a panel of 317 kinases,26 the authors found that lapatinib,
imatinib, and staurosporine do not bind appreciably to p38R
(Kd >10 µM), in strong agreement with our data. Interestingly,
they found that sorafenib binds to p38R with a Kd ≈ 370 nM,
nearly 6-fold higher than our determined Kd value. Because we
observed a time dependence for sorafenib binding to p38R over
a period of 4 h, their higher Kd value may be due to the shorter
1 h preincubation used in that large screen.
1uwh34) (Figure S6C). Further structural details are described
in the Supporting Information.
3.6. Detection of ATP-Competitive Inhibitors/Identifying
False Hits in Screens. In addition to several DFG-out binders,
we also tested whether several known ATP-competitive inhibi-
tors of p38R could be detected by ac-p38R and, if detected, to
determine whether these compounds could be discriminated
from DFG-out binders. The majority of compounds tested were
not detected up to concentrations of 10 µM. However, a few
imidazole derivatives, SKF86002 and SB203580, were detected
with varying degrees of sensitivity. Through the use of
fluorescence spectroscopy and protein X-ray crystallography,
we were able to explain the detection of such compounds and
found that real-time kinetic measurements of binding made with
ac-p38R can be used to easily discriminate type I inhibitors
should they be detected (Figure 5).
Using our ac-p38R assay, the type I inhibitor SKF86002 gave
a binding curve with a Kd of 721 ( 36 nM (Figure 5A), which
is significantly higher than the published value (Kd ≈ 180 nM8),
suggesting that ac-p38R is relatively insensitive to type I binders.
We designed a simple experiment to confirm that ac-p38R loses
the ability to accurately sense binding of type I ATP-competitive
compounds. We used the intrinsic fluorescence of the inhibitor
described above, SKF86002, to measure its binding to ac-p38R.
It is known that this compound exhibits a very large increase
in fluorescence emission upon binding to p38R.8 Because the
emission maxima of SKF86002 (∼420 nm) and acrylodan
(∼470 nm) do not overlap, we added this compound to ac-
p38R and made end point measurements to obtain a binding
curve with a Kd of 78 ( 9 nM, which is 10-fold lower under
the same experimental conditions than when measuring the
fluorescence of acrylodan. Therefore, it is clear that the
acrylodan label itself is insensitive to most type I binders, most
likely due to the fact that the fluorescent-tagged activation loop
is not expected to change conformations upon binding of these
types of ligands. However, the binding of some type I inhibitors
such as SKF86002 induce unexpected conformational changes
involving the activation loop and/or a reorientation of the N-lobe
relative to the C-lobe upon binding to the kinase hinge region.
In the latter case, these localized conformational changes could
modulate the polar environment in the vicinity of the fluorophore
without movement of the activation loop, resulting in false hits
when screening for DFG-out binders.
In another example, we identified SB203580 as a potent low
nanomolar type I p38R inhibitor and close structural analogue
of SKF86002. Surprisingly, the fluorescence change observed
in ac-p38R upon binding of SB203580 was robust and allowed
binding curves to be generated (Figure 5B), which gave Kd
values that were the same as those previously published using
other methods (15 ( 2 nM).10 To better understand why this
compound in particular triggered such a sensitive response, we
cocrystallized it with p38R (Figure 5C). The structure was solved
to a resolution of 2.3 Å with positive difference density for the
inhibitor clearly visible in the ATP binding pocket. Although
the pyridinyl group of the inhibitor forms a hydrogen bond
to the hinge region, which qualifies SB203580 as a type I
inhibitor, the kinase adopted the DFG-out conformation. The
plane of the methylsulfinyl substituted phenyl is sandwiched
between the DFG motif and the P-loop and forms π-π
stacking interactions with the side chains of Tyr35 (glycine-
To confirm that our new approach was correctly reporting
the binding of sorafenib to the DFG-out conformation, we
cocrystallized it with wild type p38R and solved the structure
to a resolution of 2.1 Å (Figure S6B). Although the kinase is
found in the DFG-out conformation and the inhibitor resides in
the typical type II orientation, bridging the allosteric site and
the hinge region of the kinase, the interaction of the ligand
with the gatekeeper Thr and the hinge region of p38R differs
from the previously reported sorafenib b-Raf complex (pdb code:
(32) Michalczyk, A.; Kluter, S.; Rode, H. B.; Simard, J. R.; Grutter, C.;
Rabiller, M.; Rauh, D. Bioorg. Med. Chem. 2008, 16, 3482–8.
(33) Wilhelm, S.; Carter, C.; Lynch, M.; Lowinger, T.; Dumas, J.; Smith,
R. A.; Schwartz, B.; Simantov, R.; Kelley, S. Nat. ReV. Drug DiscoVery
2006, 5, 835–44.
(34) Wan, P. T.; Garnett, M. J.; Roe, S. M.; Lee, S.; Niculescu-Duvaz, D.;
Good, V. M.; Jones, C. M.; Marshall, C. J.; Springer, C. J.; Barford,
D.; Marais, R. Cell 2004, 116, 855–67.
9
J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009 13293

A R T I C L E S
Simard et al.
Figure 5. Detection of the binding of imidazole derivatives SB203580 and SKF86002 by ac-p38R. The high affinity ATP-competitive inhibitor of p38R,
SKF86002, binds with a Kd ≈ 78 nM (0) according to changes in the fluorescence emission of the inhibitor itself upon binding (left axis), while the
ratiometric fluorescence of acrylodan (right axis) reports a Kd of 721 nM (9) (A). SB203580 shares structural moieties (red) similar to SKF86002 and
produced a binding curve with a Kd of 15 nM when added to ac-p38R (B). We solved the crystal structure of SB203580 to 2.3 Å, and electron density maps
of SB203580 (red) and p38R (gray) are contoured to 1σ. Possible hydrogen-bonding interactions are highlighted by dotted lines. The pyridine ring of the
inhibitor forms an essential hydrogen bond to the hinge region (Met109) of the kinase. The proximal phenyl substituent is perfectly sandwiched between the
side chain of Phe169 of the DFG-motif (orange) and the side chain of Tyr35 of the glycine-rich loop (green) respectively. The methylsulfinyl substituted
phenyl is electron rich and forms energetically favorable π-π interactions with the side chains of Tyr35 and Phe169 and most likely stabilizes the DFG-
motif in the “out” conformation. In addition, water molecule W1 bridges a hydrogen bond between N3 of the imidazole of the inhibitor and the backbone
carbonyl (red) of Leu167 and is likely to contribute to stabilization of the DFG-out conformation (C). For real-time kinetic measurements, a single dose of
BIRB-796 or SB203580 (10 µM) was added to ac-p38R (0.1 µM). ATP-competitive inhibitors always produce an instantaneous change in fluorescence (blue
trace), while slower binding DFG-out stabilizing inhibitors (black trace) typically take several minutes to reach equilibrium (D).
rich loop) and Phe169 (DFG motif). The N3 of the imidazole
moiety hydrogen bonds via a water molecule to the backbone
of Leu167 located at the N-terminal end of the DFG motif. The
net result of these interactions is the stabilization of p38R in
the DFG-out conformation despite the type I binding mode of
SB203580. Although the compound is not bound within the
allosteric site, the assay detected the compound due to this
unique binding mode.
Interestingly, SB203580 has been analyzed extensively by
both protein X-ray crystallography and NMR techniques (PDB
codes: 2ewa17 and 1a9u35). While one study reported the binding
of SB203580 to the DFG-in conformation,35 the other group
reported that this inhibitor can in fact bind to both DFG-
conformations (∼50% inhibitor occupancy in each conforma-
tion) and further confirmed this finding using 2D-NMR
experiments.
subpocket behind the gatekeeper as observed for SB203580.
The core Y-shaped structure of these compounds is exactly the
same, with the exception of the additional phenyl moiety of
SB203580, which is responsible for forming the π-π stacking
interactions with Phe169 and Tyr35 of p38R to stabilize the
DFG-out conformation. This ring is not present in SKF86002.
Assuming the binding mode is similar, this structural difference
may reduce the ability of SKF86002 to stabilize the DFG-out
conformation, thus explaining the observed insensitivity of the
acrylodan-labeled activation loop to its binding, in contrast to
its highly sensitive response to SB203580. Regardless of this
insensitivity to most type I inhibitors, the assay appears to be
very sensitive to type I binders that interact and modulate the
conformation of the DFG motif and, thus, the activation loop.
It is known that the kinase activation loop and helix C, which
forms the “roof” of the allosteric pocket, are both in cross-talk
with the glycine-rich loop,34,36-38 a structural element that serves
as a regulatory flap above the ATP binding site. This loop can
interact directly with some type I inhibitors, often contributing
We can postulate on the basis of these observations that
SKF86002 may bind in a similar manner to SB203580. Both
compounds share structural similarities, in particular the 4-fluo-
rophenyl moiety, which likely extends back into the hydrophobic
(36) Hanks, S. K.; Hunter, T. FASEB J. 1995, 9, 576–96.
(37) Mapelli, M.; Massimiliano, L.; Crovace, C.; Seeliger, M. A.; Tsai,
L. H.; Meijer, L.; Musacchio, A. J. Med. Chem. 2005, 48, 671–9.
(38) Wong, L. J. Mol. Biol. 2005, 351, 131–43.
(35) Wang, Z.; Canagarajah, B. J.; Boehm, J. C.; Kassisa, S.; Cobb, M. H.;
Young, P. R.; Abdel-Meguid, S.; Adams, J. L.; Goldsmith, E. J.
Structure 1998, 6, 1117–28.
9
13294 J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009

Fluorescent-Tagged Kinase Assay System
A R T I C L E S
to their increased affinity, as demonstrated for the potent Abl
kinase inhibitor nilotinib.39 Thus, we expect that the interaction
of this structural element with some type I inhibitors will change
the environment of the acrylodan without actually inducing a
movement of the activation loop. Acrylodan responds to any
change in the local environment, whether induced directly by
the movement of the activation loop or indirectly by the
movement of nearby elements such as helix C or the glycine-
rich loop.
As a result of these findings, it is likely that some type I
inhibitors may register as false hits in future large screening
campaigns for DFG-out binders. Therefore, it was necessary to
determine whether or not these ligands could be discriminated
from DFG-out binders using our assay. Using the cuvette
method, we were able to accomplish this by looking at the
kinetics of the fluorescence changes in real time. In the case of
DFG-out binders such as BIRB-796, the kinetic is relatively
slow (Figure 5D, Figure 3E), and the fluorescence change takes
several minutes to reach completion. However, in the case of
ATP-competitive inhibitors such as SB203580, the induced
fluorescence change is instantaneous (2-4 s). This is not
surprising because the ATP binding site is relatively easy to
access in comparison to the allosteric site, which only becomes
available when the kinase samples the DFG-out conformation.
This instantaneous fluorescence response was also observed for
SKF86002 (data not shown).
additional bottleneck, the current lack of suitable high-
throughput methods that can discriminate between DFG-in and
DFG-out binders. To date, no method exists for predicting
whether a kinase of interest can even be inhibited in this manner.
To address these needs, we developed a novel assay system
to monitor transitions in kinase conformation induced and
stabilized by the binding of certain ligand types. Our assay
substantially differs from kinase assays that are widely employed
in pharmaceutical and chemical biology research, which exclu-
sively monitor kinase activity in a classical sense via phospho-
rylation of various substrates. We utilized our new assay for
detailed kinetic characterization of a diverse collection of known
and unknown kinase inhibitors and found that the system is
highly sensitive to type II/III inhibitors and insensitive to most
ATP-competitive inhibitors in the case of p38R. However, we
demonstrated that some unique type I binders such as SB203580,
which unexpectedly induce and stabilize the DFG-out confor-
mation, are sensitively detected as well but can easily be
discriminated from type II or III binders by looking at the
binding kinetics. Thus, the labeling of kinases that are regulated
by a DFG-in/DFG-out conformational equilibrium with a
suitable fluorophore will not only allow for the detection of the
ligand-induced transition between kinase conformations, but also
provides a unique handle to easily discriminate between different
inhibitor types.
We recently used this approach to identify the first known
type III ligands of cSrc kinase and successfully optimized these
scaffolds into potent low nanomolar type II inhibitors, some of
which remained potently active against common drug resistance
mutations.15,28 The findings provide proof that this direct binding
assay extends well into emerging medicinal chemistry research.
In the past, the detection of ligand binding was often evaluated
by protein X-ray crystallography to unravel the binding mode
of the ligand. This is a standard procedure in modern structure-
based drug design. Along these lines, we demonstrated that the
b-Raf/VEGFR inhibitor sorafenib is detected by our assay with
remarkable affinity and adopts a type II binding mode in which
a new kind of interaction is formed between the inhibitor, the
gatekeeper and the hinge region of p38R, features which could
prove valuable to direct further compound design.
4. Discussion
Although kinase inhibitor drug discovery has progressed
dramatically in the past decade and resulted in almost a dozen
marketed drugs,3,4 several challenges such as inhibitor selectivity
of classical ATP-competitive inhibitors and kinase drug resis-
tance remain. A high number of known type I kinase inhibitors
are more sensitive to drug resistance, and limited selectivity
may be a direct consequence of their discovery using enzymatic
assays to screen large compound libraries.1 Because these assays
rely on the kinase to be phosphorylated and in the active
conformation, its ability to adopt the inactive conformation is
reduced, resulting in an enrichment of hits of type I scaffolds.
Because most of the chemical scaffolds that qualify as ATP-
competitive ligands have already been discovered, classical
kinase activity assays are becoming more obsolete for the
discovery of more specific novel compounds to serve as lead
structures for future medicinal chemistry and chemical biology
research. This points toward current medicinal chemistry
endeavors to identify and develop inhibitors that target alterna-
tivebindingsitesandstabilizeinactivekinaseconformations.1,40-43
This is also of benefit to the pharmaceutical industry, which is
beginning to struggle to identify new patentable drug scaffolds
for kinases. However, the shifting focus has resulted in an
5. Conclusions
The development of inhibitors that stabilize inactive (DFG-
out) kinase conformations is moving to the forefront of kinase
inhibitor research. To allow for the sensitive and reliable
identification of DFG-out stabilizers from compound libraries,
we developed a novel fluorescence-based binding assay using
the serine-threonine kinase p38R as a model system. We
utilized known kinase inhibitors as well as newly designed
compounds to validate this system and confirmed our results
with protein X-ray crystallography. In doing so, we were able
to identify the b-raf inhibitor sorafenib as a high affinity binder
of p38R and also observed a unique type II inhibitor binding
mode. Research for the identification of new lead compounds
is mainly driven by the screening of large compound collections,
where up to millions of chemical entities form a single library.
Because the method presented herein detects direct binding of
inhibitors, it does not rely on catalytically active kinase
preparations, costly reagents such as kinase substrates, or
radioactivity. It may therefore offer an inexpensive alternative
to current kinase inhibitor screening methods used by both
academia and industry. Ongoing work is focused on adapting
this approach to a 384-well format to screen large compound
(39) Weisberg, E.; Manley, P. W.; Breitenstein, W.; Bruggen, J.; Cowan-
Jacob, S. W.; Ray, A.; Huntly, B.; Fabbro, D.; Fendrich, G.; Hall-
Meyers, E.; Kung, A. L.; Mestan, J.; Daley, G. Q.; Callahan, L.; Catley,
L.; Cavazza, C.; Azam, M.; Neuberg, D.; Wright, R. D.; Gilliland,
D. G.; Griffin, J. D. Cancer Cell 2005, 7, 129–41.
(40) Adrian, F. J.; Ding, Q.; Sim, T.; Velentza, A.; Sloan, C.; Liu, Y.;
Zhang, G.; Hur, W.; Ding, S.; Manley, P.; Mestan, J.; Fabbro, D.;
Gray, N. S. Nat. Chem. Biol. 2006, 2, 95–102.
(41) Calleja, V.; Laguerre, M.; Parker, P. J.; Larijani, B. PLoS Biol. 2009,
7, 189–200.
(42) Fischmann, T.; Smith, C.; Mayhood, T.; Myers, J.; Reichert, P.;
Mannarino, A.; Carr, D.; Zhu, H.; Wong, J.; Yang, R. S.; Le, H.;
Madison, V. Biochemistry, doi:10.1021/bi801898e.
(43) Kirkland, L. O.; McInnes, C. Biochem. Pharmacol. 2009, 77, 1561–
71.
9
J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009 13295

A R T I C L E S
Simard et al.
J.R.S. was funded by the Alexander von Humboldt Foundation.
This work was supported by the German Federal Ministry for
Education and Research through the German National Genome
ResearchNetwork-Plus(NGFN-Plus)(GrantNo.BMBF01GS08102).
Schering Plough, Bayer-Schering Pharma, Merck-Serono, and Bayer
CropScience are thanked for financial support.
libraries. The application of this new assay system to drug targets
such as EGFR, Abl, PDGFR, and Kit, as we applied to cSrc
kinase,28 may eventually spur the development of new chemical
entities, which take advantage of allosteric binding sites and
lead to more specific drugs, some of which may be able to
overcome the emerging problem of kinase drug resistance
mutations. Its application will be particularly fostered by first
answering the question of which kinases can adopt targetable
inactive conformations.
Supporting Information Available: Experimental details for
the chemical synthesis of the focused library of pyrazolourea-
based type II and type III ligands, data collection and refinement
statistics for p38R cocrystallized with several ligands, several
additional figures, complete refs 10, 12, 26, and 38, as well as
further discussions regarding the characterization of p38R
conjugated with additional fluorophores, the SAR of type II
hybrid inhibitors, and the validation of sorafenib as a potent
type II p38R inhibitor. This material is available free of charge
via the Internet at http://pubs.acs.org.
Acknowledgment. We thank the Dortmund Protein Facility for
cloning, expressing, and purifying the human p38R mutant variants
for labeling. We thank Michael Weyand, Eckhard Hofmann, Ingrid
Vetter, and beamline scientists at X10SA for expert assistance
during data collection. We thank Petra Janning for her assistance
with the HPLC and MS analysis of the fluorescent-labeled and
unlabeled p38R, Sabine Klüter and Haridas Rode for expert
assistance during kinetic studies and organic synthesis. We also
would like to thank Herbert Waldmann for helpful discussions.
JA902010P
9
13296 J. AM. CHEM. SOC. VOL. 131, NO. 37, 2009