Overcoming compound fluorescence in the FLiK screening assay with red-shifted fluorophores

Article abstract of DOI:10.1021/ja403074j

In the attempt to discover novel chemical scaffolds that can modulate the activity of disease-associated enzymes, such as kinases, biochemical assays are usually deployed in high-throughput screenings. First-line assays, such as activity-based assays, often rely on fluorescent molecules by measuring a change in the total emission intensity, polarization state, or energy transfer to another fluorescent molecule. However, under certain conditions, intrinsic compound fluorescence can lead to difficult data analysis and to false-positive, as well as false-negative, hits. We have reported previously on a powerful direct binding assay called fluorescent labels in kinases ('FLiK'), which enables a sensitive measurement of conformational changes in kinases upon ligand binding. In this assay system, changes in the emission spectrum of the fluorophore acrylodan, induced by the binding of a ligand, are translated into a robust assay readout. However, under the excitation conditions of acrylodan, intrinsic compound fluorescence derived from highly conjugated compounds complicates data analysis. We therefore optimized this method by identifying novel fluorophores that excite in the far red, thereby avoiding compound fluorescence. With this advancement, even rigid compounds with multiple π-conjugated ring systems can now be measured reliably. This study was performed on three different kinase constructs with three different labeling sites, each undergoing distinct conformational changes upon ligand binding. It may therefore serve as a guideline for the establishment of novel fluorescence-based detection assays.

Full text of DOI:10.1021/ja403074j

Article
pubs.acs.org/JACS
Overcoming Compound Fluorescence in the FLiK Screening Assay
with Red-Shifted Fluorophores
Ralf Schneider,†,⊥ Anne Gohla,‡ Jeffrey R. Simard,†,# Dharmendra B. Yadav,§,∇ Zhizhou Fang,‡
Willem A. L. van Otterlo,§,∥ and Daniel Rauh*,†,‡
†Chemical Genomics Centre of the Max-Planck-Society, Otto-Hahn-Strasse 15, 44137 Dortmund, Germany
‡Fakultat Chemie - Chemische Biologie, Technische Universitat Dortmund, Otto-Hahn-Strasse 6, 44227 Dortmund, Germany
̈
̈
§Molecular Sciences Institute, School of Chemistry, University of the Witwatersrand, PO Wits, 2050, Johannesburg, South Africa
∥Department of Chemistry and Polymer Science, Stellenbosch University, Stellenbosch, 7602, Western Cape, South Africa
S
* Supporting Information
ABSTRACT: In the attempt to discover novel chemical scaffolds that can
modulate the activity of disease-associated enzymes, such as kinases, biochemical
assays are usually deployed in high-throughput screenings. First-line assays, such
as activity-based assays, often rely on fluorescent molecules by measuring a
change in the total emission intensity, polarization state, or energy transfer to
another fluorescent molecule. However, under certain conditions, intrinsic
compound fluorescence can lead to difficult data analysis and to false-positive, as
well as false-negative, hits. We have reported previously on a powerful direct
binding assay called fluorescent labels in kinases (‘FLiK’), which enables a
sensitive measurement of conformational changes in kinases upon ligand binding.
In this assay system, changes in the emission spectrum of the fluorophore
acrylodan, induced by the binding of a ligand, are translated into a robust assay
readout. However, under the excitation conditions of acrylodan, intrinsic compound fluorescence derived from highly conjugated
compounds complicates data analysis. We therefore optimized this method by identifying novel fluorophores that excite in the far
red, thereby avoiding compound fluorescence. With this advancement, even rigid compounds with multiple π-conjugated ring
systems can now be measured reliably. This study was performed on three different kinase constructs with three different labeling
sites, each undergoing distinct conformational changes upon ligand binding. It may therefore serve as a guideline for the
establishment of novel fluorescence-based detection assays.
π−π stacking with aromatic amino acid side chains. Thus, drug-
like molecules are composed of a chemical scaffold decorated
with a variety of charged or uncharged moieties designed to
interact with amino acids of the protein target. However, the core
structures of these compounds are often rigid, planar, and
composed of multiple conjugated aromatic moieties, which have
a higher probability of having their own intrinsic fluorescence
properties. This compound fluorescence can cause interference
in fluorescence-based assay readouts, either by contributing
background signal or by participating in unwanted FRET with
the fluorophore(s) involved in the assay readout. Minor
background fluorescence can be accounted for by measuring
the fluorescence of a compound at a given concentration in the
absence of all other assay components and then subtracting it
from the response observed in the actual enzymatic assay.5
However, in cases where the compound fluorescence is more
intense than the fluorophore, which may occur at high
compound concentrations (>10 μM), a simple subtraction of
the signal may not be accurate enough. Moreover, unwanted
1. INTRODUCTION
High-throughput screening (HTS) of large compound libraries
provides an avenue for the rapid early stage discovery of small
molecules which are active against disease-associated targets.
Typically, a suite of assays are developed to fully characterize hit
molecules and to elucidate details of their mode of action on the
target. Such an assay suite may include binding assays, enzymatic
assays for the primary target, selectivity assays for secondary
targets, and a variety of other biophysical and analytical methods,
including surface plasmon resonance or mass spectrometry.
Many of the commercially available assays rely on the use of
fluorescent molecules and tags for detecting enzyme reaction
products and/or monitoring the depletion of enzyme substrates.
This is accomplished by measuring changes in fluorescence
intensity, changes in fluorescence polarization, or by combining
fluorophores to enable time-resolved Forster resonance energy
̈
transfer (TR-FRET).1,2
Following Lipinski’s “Rule of 5” drug-like molecules usually
have molecular weights ranging from 400 to 600 Da.3,4 To
achieve selectivity for the target enzyme, the chemical scaffolds of
these molecules must be designed to engage in multiple
interactions, such as hydrogen bonds, van der Waals forces, or
Received: April 8, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
FRET artifacts are much more difficult to identify and to correct,
leading to false-positive or -negative results. Therefore, a cascade
of hit validation steps are performed after each HTS in which
compounds with high intrinsic fluorescence may be lost,
especially when they only have a weak affinity for the target
enzyme. Such molecules, however, may be very valuable for
further lead optimization, as demonstrated in fragment-based
screening approaches.6,7
conditions where neither intrinsic compound fluorescence nor
light scattering occurred. We show that an alternative
fluorophore could be found for all constructs which was able
to report the conformational change as reliably as acrylodan, but
without the above-mentioned complications due to fluorescent
compounds and light scattering. Finally, as a proof of principle,
we show the value of these improvements by screening a focused
library of highly rigid compounds against Abl in the FLiK assay.
Our laboratory focuses on the development of assays for
protein kinases and phosphatases, which are major classes of
therapeutic targets in a variety of diseases. The dynamic,
reversible phosphorylation of a target substrate protein often
changes its state of activity. For the discovery of novel inhibitors
which bind more favorably to inactive enzyme conformations, we
have previously established detection methods called fluorescent
labels in kinases (‘FLiK’; Figure 1) and fluorescent labels in
2. MATERIALS AND METHODS
Recombinant kinase constructs were expressed and purified as described
previously.10−12 The fluorophores were diluted from 10 mM DMSO or
DMF stock solutions to a final concentration of 15 μM in 500 μL ice-
cold labeling buffer (50 mM Hepes, 200 mM NaCl, 10% glycerol, pH
7.3) before adding protein to a final concentration of 10 μM. The
mixture was incubated overnight on ice in the dark. Unbound excess
fluorophore was quenched with 1 mM DTT, and the protein was
washed 4× in 0.5 mL centricons (Amicon Ultra-0.5, 10 kDa, Merck
Millipore, Darmstadt, Germany) by concentrating down to <100 μL
volumes followed by dilution to 500 μL with washing buffer (labeling
buffer + 1 mM DTT). The concentrations of protein−fluorophore
complexes were then determined photometrically. For this, the
extinction coefficient of the protein was calculated using ProtParam
(http://web.expasy.org/protparam, Swiss Institute of Bioinformatics,
Lausanne, Switzerland), while the extinction coefficient of the
fluorophore was determined in a dilution series (plotting the optical
density at 280 nm against the fluorophore concentration). Quantitative
and single labeling for all proteins was confirmed by ESI-MS (Table S1).
FLiK experiments were performed in 384 well plates as described
elsewhere10,11,17 by adding 19 μL of 100 nM protein-fluorophore
conjugates diluted in FLiK buffer (50 mM Hepes, 200 mM NaCl, pH
7.4) to 1 μL of inhibitor (20x in DMSO). The final DMSO
concentration in each well was 5% v/v. Fluorescence was measured
on an Infinite M1000 plate reader (Tecan, Mannedorf, Switzerland) as
̈
described earlier10 and in the main text. Titrations were performed in
quadruplets, Z′ was determined with 8 wells of DMSO as negative
control and 8 wells of 2 μM BIRB-796 + p38α or 25 μM GNF-2 + Abl,
respectively, as positive control. Z′ values were determined using the
formula: Z′ = 1 − [3*(σp + σn)/| μp − μn|], where ‘σ’ and ‘μ’ are the
standard deviations and mean values, respectively, of the positive (p)
and negative (n) controls.
Figure 1. Principle of the FLiK assay. A thiol-reactive fluorophore is
introduced at a site of the kinase which undergoes a significant
conformational change upon ligand binding, thereby changing the
solvent shell of the fluorophore (colored sphere), resulting in alterations
in its emission spectrum. The inactive state of the kinase (left) is
stabilized by a type III inhibitor (blue). Classic ATP competitive
inhibitor (yellow) bound to the active kinase (right). Definition of type
I−IV kinase inhibitors.14−16
3. RESULTS AND DISCUSSION
3.1. Compound Fluorescence. The initial discovery,
development, and application of the FLiK approach to screening
kinases in high throughput has relied so far on the fluorophore
acrylodan, which has been shown to provide a very robust assay
readout9−11,13 due to the appearance of two emission maxima
(λem,max) at ∼470 and ∼510 nm. When ligands bind to the labeled
kinase and thereby introduce conformational changes (Figure 1),
the total intensity of each of the two acrylodan emission maxima
changes relative to one another, thereby enabling a highly
desirable ratiometric fluorescence readout that provides a higher
reliability and reproducibility when compared to monochromic
readouts of intensity.12
This type of readout is also popular in many of the
commercially available TR-FRET activity-based assays for
kinases. When characterizing the labeled kinase using FLiK,
the emission spectra were recorded in the presence and absence
of selected kinase inhibitors which are known to induce different
kinase conformations.
phosphatases (‘FLiP’),8 which serve as conformation-specific
binding assays for kinases and phosphatase. The FLiK assay has
been established for a variety of kinases, where the environ-
mentally sensitive fluorophore acrylodan was attached to the
target kinase at a specific site which undergoes a significant
conformational change upon ligand binding.9−13 The resulting
alteration of the microenvironment of acrylodan is reported by a
significant change in its emission spectrum.
To date, we have used FLiK in several HTS campaigns
performed both in-house13,17 or in collaboration with
pharmaceutical industry partners. In these campaigns, we have
observed two challenges associated with the exclusive use of
acrylodan in these assays: (i) intrinsic compound fluorescence as
described above and (ii) other randomly occurring signals of high
intensity (“spikes”) derived from light scattering by lab dust
particles. While the latter issue can be minimized by performing
the HTS in a cleanroom, the problem with intrinsic compound
fluorescence needed to be addressed to further improve the FLiK
technology.
To determine whether compounds alone can contribute
fluorescence signals at these wavelengths, thereby complicating
analysis of the ratiometric readout derived from the labeled
kinase, we measured the emission of compounds alone in buffer
(no labeled kinase) (Figure 2a) at an initial concentration of 10
To do so, we conjugated eight thiol-reactive fluorophores
(Table 1) to three different kinase constructs and searched for
B
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 1. Tested Fluorophores
a
b
Excitation and emission maxima as provided by the supplier. Neither structure nor exact molecular mass of Alexa Fluor 660 was provided by the
supplier. ΔM: Mass increase after reaction with protein. QY of acrylodan-mercaptoethanol adduct in water.18 cTnC mutant labeled with IAANS
c
d
e
at Cys-84; QY in EGTA buffer using quinine sulfate in 0.1 N H2SO4 as standard (0.52 at λex = 365 nm and 0.47 at λex = 325 nm).19 QY in water (pH
f
1.0) using quinine sulfate in 5 M H2SO4 as standard (QY = 0.55).20 Texas Red-transferrin conjugate in aqueous solution using sulforhodamine 101
g
as standard.21 Determined in aqueous buffer solution as provided by the supplier.
h
μM, which is a typical compound concentration used in HTS
campaigns.1 Compounds with extensive π−electron systems,
such as staurosporine and especially DBY-14 (a highly rigid
analog of the allosteric Abl ligand GNF-2 synthesized in-house),
displayed strong intrinsic emission up to 450 nm when excited at
the excitation maximum for acrylodan (386 nm). Next, we
C
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
ring conjugated system, such as erlotinib or BIRB-796 were
excitable up to ∼350 nm. Compounds with multiple conjugated
ring systems (>2 rings), such as staurosporine, were excitable at
much higher wavelengths. Notably, above 450 nm, none of the
chosen compounds were excitable. We therefore conclude that
fluorophores with an absorption (λexc) at >450 nm would likely
reduce the occurrence of intrinsic compound fluorescence and
might be chosen as alternatives for acrylodan to develop a far-red
FLiK assay.
Thus, a mixture of five commercially available fluorophores
(Atto 565, Texas Red, Atto 610, Dy-647, and Alexa Fluor 660)
was selected, which all have excitation maxima well above 450
nm. Additionally, we investigated PyMPO, which has an
excitation maximum between 400 and 450 nm, and compared
their performance to the fluorophores IAANS and acrylodan,
which excite at lower wavelengths (<400 nm) (Table 1).
As mentioned above, the FLiK assays developed to date by our
laboratory were at risk for a higher rate of detection of false
positives due to intrinsic compound fluorescence. Many of these
false positives were due to additive fluorescence signal coming
from the compounds at the emission wavelengths being
monitored in a particular assay. We were able to minimize the
number of such false-positives by subtracting their intrinsic
fluorescence (as measured in buffer) from the total fluorescence
intensity (obtained when the compound is placed into solution
with the labeled kinase).5 However, FRET-based fluorescence
artifacts due to energy transfer between compound and
fluorophore are not additive and can still lead to false-positives
or false-negatives. Data derived from such compounds are
significantly more difficult to analyze.
In an attempt to avoid such unwanted interactions between
the chosen fluorophore and these compounds, it became
necessary to investigate the performance of fluorophores which
excite at longer wavelengths. As done for acrylodan, we measured
10 μM solutions of the reference inhibitors in buffer using the
excitation and emission wavelengths (λexc, λem, respectively)
provided by the supplier for the different fluorophores (Figure
2c). When using λexc and λem reported for IAANS, we measured
significant intrinsic compound fluorescence for staurosporine,
DBY-14, and BIRB-796, and, with much reduced intensity,
dasatinib, nilotinib, and GNF-2. Under the acrylodan conditions,
less rigid molecules, such as BIRB-796 or GNF-2, did not show
significant intrinsic fluorescence any more as expected from
Figure 2b, but DBY-14 showed an extremely high intrinsic
fluorescence. When excited at 412 nm (PyMPO), only DBY-14
was fluorescent, while all other inhibitors were not detected.
Notably, when excited at >450 nm, no intrinsic fluorescence
could be observed for any molecule, confirming that the use of
far-red fluorophores allows for the measurement of highly
conjugated molecules.
We then simulated a dusty lab environment by striking the
assay plates on a standard white cotton lab coat. Several spikes
suddenly appeared (Figure 2d) under the acrylodan settings.
Similar effects were observed under the IAANS settings but with
much reduced intensity. Notably, these added spikes did not
appear under any other conditions tested, suggesting that the lab
dust particles would not have an effect when using fluorophores
that absorb at >400 nm.
Finally, to investigate how these alternative fluorophores
would perform in an assay setting, we analyzed data obtained
from three recombinant kinase constructs labeled on three
different structural features of the kinase domain that undergo
significant conformational changes upon ligand binding. In these
Figure 2. Effects of compound fluorescence. (a) Emission spectra of 10
μM compounds in FLiK buffer (λexc = 386 nm). (b) Excitation spectra
(λem = 500 nm). (c) Emission of 10 μM compounds in FLiK buffer with
λexc and λem as provided by the supplier. Measuring under the settings for
IAANS (purple), acrylodan (dark blue), and PyMPO (light blue)
resulted in intrinsic fluorescence of rigid inhibitors. No intrinsic
compound fluorescence is observed under the settings for the remaining
fluorophores. (d) Rubbing the plate once on a lab coat to simulate a
dusty lab environment generated new fluorescent signals with high
intensity under the acrylodan settings (arrows), demonstrating
sensitivity to lab dust. (e) Emission spectra of p38α labeled with
IAANS at the activation loop (filled circles) in the apo form (blue) and
supplemented with 1 μM BIRB-796 (red). Most of the fluorescence is
derived from the compound (open circles). (f) Ratiometric fluorescence
(R = I402 nm/I462 nm) without (filled circles) and with subtraction of the
background fluorescence (filled circles) plotted against the inhibitor
concentration. RFI is relative fluorescence intensity, and a.u. is arbitrary
units.
recorded the excitation spectra of these inhibitors to determine
the shortest wavelength at which these compounds will not
produce an emission signal (Figure 2b). Less complex
compounds, such as imitinib or GNF-2, which only have single
aromatic ring systems, were not significantly excitable at
wavelengths above ∼300 nm, while compounds with a two-
D
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
constructs, the fluorophores were fused by nucleophilic
substitution (IAANS) or Michael addition (all other fluoro-
phores) to (i) the activation loop (‘a-loop’) of p38α, (ii) the
glycine-rich loop (‘p-loop’) of p38α, and (iii) the myristate
pocket of Abl, as described previously10−12 and in Materials and
Methods.
3.2. Labeling at the Activation Loop of p38α. The
negative effect of intrinsic compound fluorescence on the assay
readout of FLiK could easily be demonstrated using p38α labeled
with IAANS at the a-loop. BIRB-796, a potent type II inhibitor of
p38α, displayed moderate levels of intrinsic compound
fluorescence in FLiK buffer when excited at 326 nm (Figure
2c), the excitation maximum for IAANS. The emission spectra of
IAANS-labeled p38α (a-loop) obtained in the presence and
absence of 1 μM BIRB-796 (Figure 2e) showed a significant left-
shift of the emission maximum from ∼460 nm (apo) to 402 nm
(1 μM BIRB-796). However, by overlaying spectra of BIRB-796
recorded with (filled circles) or without (open circles) labeled
protein, it is clear that most of the recorded fluorescence at 402
nm is derived from the compound rather than the attached
fluorophore. Nevertheless, plotting the ratio of emission
intensities at 402 and 462 nm measured with protein against
the logarithmic inhibitor concentration (Figure 2f) generates a
nearly sigmoidal regression curve. However, the standard
deviations became very large at high compound concentrations,
further suggesting the influence of background fluorescence on
these ratiometric values. Performing a background correction (by
subtracting the intensity measured in buffer from the one
measured with protein at each inhibitor concentration) leads to a
significantly different regression curve, further demonstrating
that this fluorophore is not suitable for FLiK.
Using a-loop labeled p38α, the same procedure was used to
investigate the suitability of the remaining fluorophores, in this
case using acrylodan as the reference fluorophore having the
lowest excitation maximum of the collection (Figure 3).
Interestingly, no intrinsic fluorescence was observed up to a
concentration of 10 μM BIRB-796 for all of the remaining
fluorophores tested (see Figure 2c), suggesting that background
correction would not be necessary when using these
fluorophores.
For each kinase-fluorophore conjugate, the emission spectrum
of p38α labeled at the a-loop with each different fluorophore was
recorded in its unbound (apo) state to identify the emission
maxima suitable for calculating a ratiometric readout for
detecting the binding of BIRB-796 (Figure 3a). Spectra recorded
under BIRB-saturated conditions were then overlaid to identify
secondary emission wavelengths to calculate ratiometric read-
outs with an internal calibration. This secondary emission
wavelength was easily identified in the spectrum of acrylodan,
which exhibits different maxima in the bound and unbound
states.11 Also, PyMPO showed a significant bathochromic shift,
however smaller as compared to acrylodan. For all other
fluorophores tested, we selected a secondary wavelength at a
shoulder of the main peak (Texas Red, Alexa Fluor 660, Dy-647,
Atto 565) and/or where the emission was stable and did not
change emission intensity significantly upon ligand binding (Dy-
647, Atto 610). In either case, one wavelength changes
significantly relative to another, resulting in a ratiometric
fluorescence change associated with ligand binding.
Figure 3. Emission spectra and titration curves of p38α labeled at the a-
loop. (a) Emission spectra recorded in the absence (blue) and presence
(red) of 1.7 μM BIRB-796. (b) Titration curves of the protein-
fluorophore conjugates. The ratiometric assay readout for each
fluorophore is summarized in the table below. (c) Calculated Z′ factors
(n = 3). Arrow: best alternative for acrylodan.
Plotting the ratiometric fluorescence against the logarithmic
concentration of BIRB-796 resulted in sigmoidal regression
curves with all fluorophores except Atto 565, where no change in
the ratio was observed under the chosen conditions. This
E
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
fluorophore is therefore unable to report the conformational
change associated with movement of the activation loop of p38α.
The Kd values for BIRB-796 determined from the regression
curves of the other fluorophores were within 2-fold of our
previously reported Kd value for BIRB-796 (30 nM).17 Although
most fluorophores could report affinity, the assay window
(maximal ratio change) and reproducibility varied. Calculating
the Z′, a commonly used factor describing the robustness of an
assay,2 we found that PyMPO and Dy-647 worked as well as the
reference fluorophore (acrylodan). Notably, Dy-647 displayed
the lowest standard deviation (n = 3) (Figure 3c) and, in contrast
to PyMPO, no sensitivity toward interference from compound
fluorescence (Figure 2b,c).
These findings suggest that the use of Dy-647 may enable
more straightforward detection of highly conjugated and rigid
compounds (Figure 2a,c) which have high intrinsic fluorescence
when excited at lower wavelenths. Thus, to reduce the number of
false hits picked up in HTS FLiK screens, we propose that Dy-
647 would serve as a suitable alternative to acrylodan. In fact,
when applied to the a-loop of p38α, selectivity for type II and III
inhibitors was observed for this fluorophore (Figure S1a), as
previously described with acrylodan.12 Additionally, type IV
allosteric inhibitors binding to the remote p38α MAP insert
pocket9 were not detected, further suggesting that Dy-647 is
amenable for HTS campaigns and retains the expected selective
detection of ligands which trigger conformational changes in the
a-loop.
3.2. Labeling at the Glycine-Rich Loop of p38α. We took
a similar approach to study p38α labeled on the glycine-rich loop,
a FLiK assay designed to detect the binding of type I (DFG-in
conformation) as well as types II and III (DFG-out) inhibitors
without relying on activity-based assays.11 Similar experiments
were carried out as described above for the a-loop FLiK assay. In
this case, three fluorophores (Texas Red, Dy-647, and Atto565)
were unable to report the conformational change upon binding
of the reference inhibitor BIRB-796 (Figure 4a,b) under the
chosen conditions. Besides acrylodan, PyMPO, and Atto 610
provided the best assay performance, as judged by the Z′ factor
(Figure 4c). The spectral changes were the greatest in the case of
PyMPO, but again due to the possibility of inducing unwanted
compound fluorescence at wavelengths used to excite PyMPO
(see Figure 2b,c), Atto 610 was chosen as the best alternative for
acrylodan in this protein construct. With Atto 610, a clear
bathochromic shift (arrow in Figure 4a) was observed in the
emission spectra when a saturating amount of inhibitor was
added. This shift indicates a change in the solvent shell of the
fluorophore22 due to an increased exposure to the buffer as a
result of the conformational change around the labeling site. As
expected for the p-loop as labeling site,11 types I−III, but not type
IV, inhibitors were detected using this fluorophore (Figure S1b).
3.3. Labeling at the Myristate Pocket of Abl. Finally, we
tested alternative fluorophores with Abl labeled at the myristate
pocket as a representative FLiK assay for a remote allosteric
binding site residing outside of the ATP binding cleft. We labeled
Abl in a manner which allows detection of conformational
changes in helix I at the C-terminal end of the kinase domain.10
Upon addition of 30 μM GNF-2, a known type IV Abl inhibitor,
only minor changes in the spectra of Abl kinase labeled with
PyMPO, Texas Red, or Dy-647 were observed (Figure 5a). This
translated into a weak or unmeasurable Z′ for these three
fluorophore-Abl conjugates (Figure 5c). By stark contrast, Atto
565 exhibited the largest bathochromic shift in the emission
spectra upon binding of GNF-2 when compared to the remaining
Figure 4. Emission spectra and titration curves of p38α labeled at the
glycine-rich loop. (a−c) Description as in Figure 3.
fluorophores. With Atto 565, an excellent signal-to-noise ratio
was obtained together with small standard deviations of each data
point (Figure 5b), resulting in the strongest Z′ of this collection
of fluorophores. The quality of this FLiK assay was comparable to
F
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
The assay window could be further increased by calculating the
ratio of the fluorescence measured to the right of the maximum
(610 nm) and left to the maximum (576 nm) (Figure S1c), since
the intensities at these two wavelengths moved in opposite
directions upon binding (Figure 5a). Titration of Abl-Atto 565
with allosteric activators (DPH) and inhibitors (GNF-2) gave
sigmoidal regression curves with Kd values in the expected
range,10,23 while types I and II (i.e., ATP competitive) inhibitors
were not detected (Figure S1c), further suggesting this FLiK
assay as a robust tool for the discovery of novel allosteric ligands
of Abl.
Having identified a far-red-shifted fluorophore that was
compatible with a FLiK assay for the myristate pocket of Abl,
it was possible to perform screens under conditions which would
minimize the detection of highly fluorescent GNF-2 analogs. We
synthesized a small library of rigid GNF-2 analogs which we
tested with the old (acrylodan) and new (Atto 565) Abl FLiK
assay. We first recorded the emission spectra of these compounds
alone (10 μM) using the emission and excitation wavelengths for
either acrylodan or Atto 565 (Figure 6a). When excited at 386
nm (λexc of acrylodan), severe compound fluorescence was
measured that saturated the detector for several of these
compounds (DBY-10−16, 18−21, 23), therefore disabling
them from ratiometric measurement. For the remaining
compounds, intrinsic fluorescence was observed, but it did not
reach saturated levels.
When using acrylodan, uncorrected and background corrected
ratiometric data are significantly different (Figure 6b), suggesting
that the majority of the detected signal is derived from the
compound alone. Additionally, standard deviations (n = 2) were
high in the background corrected data set, which indicates low
reliability of the assay readout. In our previous screens, we
excluded such compounds with background fluorescence >3×
the signal for DMSO alone, meaning that all DBY compounds
would have been excluded during the hit evaluation despite being
close analogs of GNF-2, a known ligand of the allosteric Abl site.
These issues were all avoided by switching to Abl-Atto 565,
demonstrating the great advantage of using this far-red-shifted
fluorophore. The lack of compound interference (Figure 6a)
becomes more apparent when comparing corrected and
uncorrected fluorescence data (Figure 6c). Unlike with
acrylodan-labeled Abl, the corrected and uncorrected values
obtained with Abl-Atto 565 were practically identical, and the
ratiometric data for all compounds could be calculated reliably.
Thus, using Atto 565 circumvents the problem of intrinsic
fluorescence with this particular class of compounds and would
enable their characterization in the Abl FLiK assay.
4. DISCUSSION AND SUMMARY
We have shown the importance of using fluorophores that are
excitable at >450 nm to avoid intrinsic compound fluorescence.
Light scattering is also avoided, further reducing the number of
putative false-positive and -negative hits picked up during
screening scenarios.
We showed the general applicability of red-shifted fluoro-
phores by fusing eight different fluorophores to three different
kinase constructs, each labeled on structural elements that
undergo significant and distinct conformational changes. For all
protein-fluorophore conjugates tested, alternative fluorophores
were identified which can report this conformational change as
reliably as acrylodan, the original fluorophore used in the
development of these various FLiK assays. With these red-shifted
fluorophores, even highly rigid compounds with high intrinsic
Figure 5. (a) Emission spectra (blue: apo; red: +30 μM GNF2) and (b)
titration curves of Abl labeled at the myristate pocket. (c) Calculated Z′.
Due to limited solubility of GNF-2 in buffer,24 the titrations did not
exceed 30 μM.
the one previously developed with acrylodan-labeled Abl (Figure
5c).
G
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
seen with p38α labeled with Atto 610 at the a-loop, can translate
into a robust assay as judged by the Z′ factor.
These results represent a major improvement of our FLiK
technology that will enable a more straightforward discovery of
complex enzyme inhibitors and reduce the number of
fluorescence artifacts and false positives and negatives.
Furthermore, it is a valuable starting point for the development
of more red-shifted fluorescence-based assays detecting con-
formational changes in a target enzyme without relying on
activity measurements. This study may therefore be especially
interesting to be applied in assays for ligands of allosteric binding
sites which are modulated by conformational changes and where
conventional assays fall short.
ASSOCIATED CONTENT
* Supporting Information
Table S1 and Figure S1. This material is available free of charge
via the Internet at http://pubs.acs.org.
■
S
AUTHOR INFORMATION
Corresponding Author
daniel.rauh@tu-dortmund.de
■
Present Addresses
⊥Sartorius Stedim Biotech GmbH, August-Spindler-Str. 11,
37079 Gottingen, Germany.
̈
#Amgen Inc., 360 Binney St., Cambridge, MA 02142, U.S.A.
∇Curadev Pharma Pvt. Ltd., B-87, Sector 83, Noida, India.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Prof. Charles B. de Koning (School of Chemistry,
University of the Witwatersrand, Johannesburg) for synthetic
advice, Ms. Simone Eppmann (Dortmund) for protein
́
preparation, and Mr. Andre Richters (Dortmund) for helpful
discussions. This work was cofunded by the German federal state
North Rhine Westphalia (NRW) and the European Union
(European Regional Development Fund: Investing In Your
Future), the German Federal Ministry for Education and
Research (NGFNPlus) (grant no. BMBF 01GS08104) all to
D.R. D.B.Y. and W.A.L.v.O acknowledge the National Research
Foundation, Pretoria, South Africa for research funding and the
University of the Witwatersrand Research Office for postdoctoral
funding for D.B.Y.
Figure 6. Screen of rigid GNF-2 analogs. (a) High intrinsic fluorescence
(10 μM DBY compounds in FLiK buffer) is observed under the
acrylodan settings (blue), while no interference was observed under the
Atto 565 settings (green). (b,c) Ratiometric assay readout of Abl labeled
with acrylodan (b) and Atto 565 (c) supplemented with 10 μM DBY
compounds (n = 2). (d) Structures of the analyzed compounds (see also
Figure 2).
REFERENCES
■
(1) Macarron, R.; Hertzberg, R. P. Mol. Biotechnol. 2009, 47, 270.
(2) Inglese, J.; Johnson, R. L.; Simeonov, A.; Xia, M.; Zheng, W.;
Austin, C. P.; Auld, D. S. Nat. Chem. Biol. 2007, 3, 466.
(3) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J. Adv.
Drug Delivery Rev. 2001, 46, 3.
(4) Anastassiadis, T.; Deacon, S. W.; Devarajan, K.; Ma, H.; Peterson, J.
R. Nat. Biotechnol. 2011, 29, 1039.
(5) Simard, J. R.; Rauh, D. Methods Mol. Biol. 2012, 800, 95.
(6) Murray, C. W.; Rees, D. C. Nat. Chem. 2009, 1, 187.
(7) Erlanson, D. A. Top. Curr. Chem. 2012, 317, 1.
compound fluorescence (as recorded at lower wavelengths)
could now be reliably measured.
In general, the best assay performance was achieved when a
bathochromic shift was detected upon ligand binding, as seen in
the case of Abl labeled with Atto 565. Calculating a ratio of
intensities measured left and right to the shifting maximum
allows for a straightforward detection of ligand binding. A simple
intensity decrease of the emission maxima without a bath-
ochromic shift, as, for example, seen with p38α labeled with
Texas Red at the a-loop, translated into poor assay performance.
Even a minor bathochromic shift of only a few nanometers, as
(8) Schneider, R.; Beumer, C.; Simard, J. R.; Grutter, C.; Rauh, D. J.
̈
Am. Chem. Soc. 2013, 135, 6838.
(9) Getlik, M.; Simard, J. R.; Termathe, M.; Grutter, C.; Rabiller, M.;
̈
van Otterlo, W. A.; Rauh, D. PloS one 2012, 7, e39713.
(10) Schneider, R.; Becker, C.; Simard, J. R.; Getlik, M.; Bohlke, N.;
Janning, P.; Rauh, D. J. Am. Chem. Soc. 2012, 134, 9138.
(11) Simard, J. R.; Getlik, M.; Grutter, C.; Schneider, R.; Wulfert, S.;
̈
Rauh, D. J. Am. Chem. Soc. 2010, 132, 4152.
H
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(12) Simard, J. R.; Getlik, M.; Grutter, C.; Pawar, V.; Wulfert, S.;
̈
Rabiller, M.; Rauh, D. J. Am. Chem. Soc. 2009, 131, 13286.
(13) Simard, J. R.; Kluter, S.; Grutter, C.; Getlik, M.; Rabiller, M.;
̈
̈
Rode, H. B.; Rauh, D. Nat. Chem. Biol 2009, 5, 394.
(14) Rabiller, M.; Getlik, M.; Kluter, S.; Richters, A.; Tuckmantel, S.;
̈
̈
Simard, J. R.; Rauh, D. Arch. Pharm. 2010, 343, 193.
(15) Fang, Z.; Grutter, C.; Rauh, D. ACS Chem. Biol. 2013, 8, 58.
̈
(16) Dar, A. C.; Shokat, K. M. Annu. Rev. Biochem. 2011, 80, 769.
(17) Simard, J. R.; Grutter, C.; Pawar, V.; Aust, B.; Wolf, A.; Rabiller,
̈
M.; Wulfert, S.; Robubi, A.; Kluter, S.; Ottmann, C.; Rauh, D. J. Am.
̈
Chem. Soc. 2009, 131, 18478.
(18) Prendergast, F. G.; Meyer, M.; Carlson, G. L.; Iida, S.; Potter, J. D.
J. Biol. Chem. 1983, 258, 7541.
(19) Dong, W. J.; Wang, C. K.; Gordon, A. M.; Cheung, H. C. Biophys.
J. 1997, 72, 850.
(20) Diwu, Z.; Chen, C. S.; Zhang, C.; Klaubert, D. H.; Haugland, R. P.
Chem. Biol. 1999, 6, 411.
(21) Panchuk-Voloshina, N.; Haugland, R. P.; Bishop-Stewart, J.;
Bhalgat, M. K.; Millard, P. J.; Mao, F.; Leung, W. Y. J. Histochem.
Cytochem. 1999, 47, 1179.
(22) Lakowicz, J. R. Principles of Fluorescence Spectroscopy; Springer:
New York, 2006; Vol. 3.
(23) Yang, J.; Campobasso, N.; Biju, M. P.; Fisher, K.; Pan, X. Q.;
Cottom, J.; Galbraith, S.; Ho, T.; Zhang, H.; Hong, X.; Ward, P.;
Hofmann, G.; Siegfried, B.; Zappacosta, F.; Washio, Y.; Cao, P.; Qu, J.;
Bertrand, S.; Wang, D. Y.; Head, M. S.; Li, H.; Moores, S.; Lai, Z.;
Johanson, K.; Burton, G.; Erickson-Miller, C.; Simpson, G.; Tummino,
P.; Copeland, R. A.; Oliff, A. Chem. Biol. 2011, 18, 177.
(24) Zhang, J.; Adrian, F. J.; Jahnke, W.; Cowan-Jacob, S. W.; Li, A. G.;
Iacob, R. E.; Sim, T.; Powers, J.; Dierks, C.; Sun, F.; Guo, G. R.; Ding, Q.;
Okram, B.; Choi, Y.; Wojciechowski, A.; Deng, X.; Liu, G.; Fendrich, G.;
Strauss, A.; Vajpai, N.; Grzesiek, S.; Tuntland, T.; Liu, Y.; Bursulaya, B.;
Azam, M.; Manley, P. W.; Engen, J. R.; Daley, G. Q.; Warmuth, M.;
Gray, N. S. Nature 2010, 463, 501.
I
dx.doi.org/10.1021/ja403074j | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX