Semisynthesis of fluorescent metabolite sensors on cell surfaces

Article abstract of DOI:10.1021/ja206915m

Progress in understanding signal transduction and metabolic pathways is hampered by a shortage of suitable sensors for tracking metabolites, second messengers, and neurotransmitters in living cells. Here we introduce a class of rationally designed semisyn

Full text of DOI:10.1021/ja206915m

ARTICLE
pubs.acs.org/JACS
Semisynthesis of Fluorescent Metabolite Sensors on Cell Surfaces
Matthias A. Brun, Rudolf Griss, Luc Reymond, Kui-Thong Tan,† Joachim Piguet, Ruud J.R.W. Peters,‡
Horst Vogel, and Kai Johnsson*
Institute of Chemical Sciences and Engineering, Institute of Bioengineering, National Centre of Competence in
ꢀ
Research Chemical Biology, Ecole Polytechnique Fꢀedꢀerale de Lausanne, Lausanne, Switzerland.
S Supporting Information
b
ABSTRACT: Progress in understanding signal transduction and
metabolic pathways is hampered by a shortage of suitable sensors
for tracking metabolites, second messengers, and neurotransmitters in
living cells. Here we introduce a class of rationally designed semisyn-
thetic fluorescent sensor proteins, called Snifits, for measuring
metabolite concentrations on the cell surface of mammalian cells.
Functional Snifits are assembled on living cells through two selective
chemical labeling reactions of a genetically encoded protein scaffold.
Our best Snifit displayed fluorescence intensity ratio changes on living
cells significantly higher than any previously reported cell-surface-
targeted fluorescent sensor protein. This work establishes a generally
applicable and rational strategy for the generation of cell-surface-
targeted fluorescent sensor proteins for metabolites of interest.
’ INTRODUCTION
biosensor.7,8 Although biosensors for various metabolites have
been developed,9À12 for most metabolites of interest no sensors
exist and in particular for applications on cell surfaces very few
biosensors have been reported.10,12 There are different reasons
for the scarcity of FRET-based biosensors for metabolites: First,
only binding proteins that undergo a conformational change
upon ligand binding can be used for the generation of such
biosensors13 and for many metabolites no such binding proteins
are available. Second, biosensors for metabolites often display
relatively small maximum ratio changes.14 When targeted to the
cell surface, fluorescent sensor proteins retained in the secretory
pathway result in background signal that further decreases the
maximum ratio change. Therefore, the construction of suitable
FP-based biosensors often requires the optimization of their
maximum ratio change.10 This mostly involves time-consuming
random screening of sensor protein libraries and this process has
to be repeated for each new biosensor.
The challenges associated with generating FRET-based bio-
sensors call for alternative strategies for the generation of
fluorescent sensors suitable for applications in cells. Here we
introduce a new class of sensor proteins, called Snifits (SNAP-tag
based indicator proteins with a fluorescent intramolecular tether),
for visualizing the concentration of metabolites on the cell surface of
mammalian cells. The approach builds on our previously reported
strategy to generate FRET-based sensor proteins for metabolites
that circumvents the obligatory conformational change of a bind-
ing protein.15 For the application of our Snifit sensor concept on
It is estimated that the primary and secondary metabolism of
an animal cell comprises about 7900 metabolites.1 Furthermore,
the concentration of these metabolites can fluctuate. For exam-
ple, the concentration of neurotransmitters can change within
milliseconds by orders of magnitude, and the concentration of
the central cofactor nicotinamide adenine dinucleotide (NAD+)
has been shown to oscillate with the same frequency as the
circadian rhythm.2 Additionally, these concentration changes are
generally confined to certain compartments within the cell.
Notably, the extracellular cell surface represents an extremely
dynamic microenvironment of the cell, as concentration transi-
ents of neurotransmitters and hormones on the cell surface of
neurons and other specialized cell types regulate important
cellular processes such as chemical neurotransmission. Undoubt-
edly, the determination of the concentration of metabolites with
both high spatial and temporal resolution is a prerequisite for an
in-depth understanding of various biological processes.
Fluorescent-based sensors are an attractive approach for
noninvasive sensing of metabolites in living cells.3 While fluor-
escent sensors for various metabolites exist, many of these
sensors are not suitable for measuring metabolite concentrations
in cells with high temporal and spatial resolution.4,5 Currently,
the majority of sensors used for measuring metabolite concen-
trations in living cells are biosensors based on F€orster resonance
energy transfer (FRET).3,6 The design of FRET-based biosen-
sors for metabolites is usually based on a protein that undergoes a
conformational change upon binding of the analyte of interest.
Sandwiching such a protein between two autofluorescent pro-
teins (FP) can then result in the generation of a FRET-based
Received: July 24, 2011
Published: August 31, 2011
r
2011 American Chemical Society
16235
dx.doi.org/10.1021/ja206915m J. Am. Chem. Soc. 2011, 133, 16235–16242
|

Journal of the American Chemical Society
ARTICLE
Figure 1. Design principle and expression of the semisynthetic sensor proteins on the surface of HEK 293T cells. (A) Semisynthetic sensor proteins.
The protein part of the sensor is a fusion protein of SNAP-tag, CLIP-tag, and a binding protein (BP). The active semisynthetic sensor protein is obtained
by labeling SNAP-tag with a molecule containing a fluorophore (red star) and a ligand for the binding protein (gray ball) and by labeling CLIP-tag with a
second fluorophore (green star). In the absence of analyte (pink ball), the intramolecular ligand binds to the binding protein, keeping the sensor protein
in a closed conformation. Donor and acceptor fluorophores are in close proximity, resulting in a high FRET efficiency. In the presence of analyte, the
intramolecular ligand is displaced, and the sensor protein shifts toward an open conformation. Donor and acceptor fluorophore are more distant from
each other than in the closed conformation and FRET efficiency therefore decreases. (B) Design of the sensor protein on the cell surface.
SNAP_CLIP_HCA was fused to a transmembrane domain (truncated PDGF receptor, pDisplay) for its display on the extracellular surface of
mammalian cells. (C) SNAP_CLIP_HCA was labeled on the cell surface of HEK 293T cells with the dyes DY-547 and DY-647 via their corresponding
O2-benzylcytosine (BC)- and O6-benzylguanine (BG)-derivatives. Images were taken using a confocal Zeiss LSM 700 microscope.
cell surfaces, we demonstrate that the protein part of the sensor
can be genetically encoded and the functional Snifit is generated on
the surface of live mammalian cells through two specific labeling
reactions. The use of Snifits on living cells represents an important
advance in the development of this sensor family. Furthermore, the
rational design of Snifits permits the use of simple optimization
guidelines to improve the performance of the sensor. The work
described here therefore establishes a rational strategy for the
generation of cell-surface based fluorescent sensors.
metabolite affinity, and sensing kinetics. We further demonstrate
the versatility with regard to fluorescent dyes and investigate
whether the sensor kinetics can be tuned according to individual
requirements. In the second part, we perform a successful
optimization of our model Snifit sensor protein in vitro and on
the cell surface by the rational design of its peptide linkers. While
there are no immediate applications for a Snifit with HCA as
binding protein, this work establishes the necessary foundation
for applying our concept to other metabolites.
The sensor protein SNAP_CLIP_HCA was successfully ex-
pressed as a fusion to a transmembrane anchor on the surface of
HEK 293T cells, as demonstrated by labeling SNAP- and CLIP-
tag with the two synthetic dyes DY-647 and DY-547, respectively
(Figure 1C). The synthetic precursor for the assembly of the
semisynthetic sensor protein on the cell surface follows the same
design guidelines as in our previous work.15 It contains a O6-
benzylguanine (BG) moiety for reaction with SNAP-tag, an 11-
PEG-unit linker, a Cy5 fluorophore which is an integral part of
the linker, and a terminal para-benzenesulfonamide moiety.
However, we exchanged the nonsulfonated Cy5 fluorophore
by a Cy5 derivative that contains two sulfonate groups (BG-
PEG11-Cy5-paraSA 1, Figure 2A) to render the BG substrate less
hydrophobic and therefore better suitable for specific labeling of
cell surface proteins. In parallel, we synthesized a BG derivative
that contains a terminal meta-benzenesulfonamide instead of the
para-benzenesulfonamide (BG-PEG11-Cy5-metaSA 2). Further,
we also synthesized derivatives of Alexa Fluor 488 (BG-PEG11-
Alexa Fluor 488-SA 4), and of Alexa Fluor 594 (BG-PEG11-Alexa
Fluor 594-SA 5), for the creation of alternative FRET pairs
(Figure 2B,C). For control experiments we synthesized BG-
PEG11-Cy5 3, BG-PEG11-Alexa Fluor 488 6, and BG-PEG11-
Alexa Fluor 594 7 lacking the terminal benzenesulfonamide
moiety (Figure 2AÀC).
’ RESULTS AND DISCUSSIONS
Snifit Sensor Proteins on the Cell Surface of Mammalian
Cells. The Snifit sensor protein concept employs a covalently
attached synthetic fluorescent ligand to distinguish between the
presence and the absence of analyte. The sensor protein is
functionalized with the synthetic ligand using the SNAP-tag
labeling technology.16 In the absence of analyte, the sensor
protein is in the closed conformation, in which the synthetic
ligand binds to the binding protein in an intramolecular fashion;
the presence of analyte displaces the intramolecular ligand from
the binding protein and thereby shifts the sensor protein to an
open conformation (Figure 1A). This competitive displacement
of the intramolecular ligand by free analyte can be detected by a
change in the FRET efficiency of the sensor protein.15 In order to
study the applicability and general properties of our Snifit sensor
proteins on the cell surface, we set out to construct a cell surface
targeted version of our proof-of-concept Snifit sensor protein
using human carbonic anhydrase II (HCA) as the binding
protein (Figure 1B). In the first part of our work we describe
the design and full characterization of the cell-surface bound
sensor. This involves the optimization of its cell-surface expres-
sion level, the response of the sensor at the cell surface, its
16236
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
to verify that the observed ratio change is indeed due to a specific
displacement of the intramolecular ligand from the active site of
HCA, we labeled HEK 293T cells expressing SNAP_CLIP_HCA
on the cell surface with BG-PEG11-Cy5 3, a molecule lacking the
terminal para-benzenesulfonamide moiety (Figure 2A), and with
CLIP-Surface 547. Perfusion of this control sensor with 10 mM
benzenesulfonamide did not induce any intensity ratio change of
DY-547/Cy5 (Figure S1).
Determination of the Kd
comp
of the Sensor Protein. It is
critical for the performance of the sensor that the physiological
concentration of the analyte of interest matches the affinity of the
sensor protein for its analyte. The binding isotherm of our sensor
protein reflects a competition between the free inhibitor and the
intramolecularly bound inhibitor.15 Thus, the binding curve is
characterized by a dissociation constant for the free ligand
(Kdcomp,ligand), which equals the amount of free ligand that needs
to be added to displace the intramolecular ligand in 50% of the
sensors. We first determined the Kdcomp,SA for the sensor on the
cell surface of HEK 293T cells by perfusion with increasing
concentrations of benzenesulfonamide and found a value of
Kdccoommpp,,SSAA = 600 ( 250 μM (Figure 4A,B). The here measured
comp,SA
Kd
is in agreement with the Kd
found for soluble
sensor (340 ( 60 μM),15 and indicates that the sensor protein
retains its properties when displayed on the cell surface. We then
also determined the Kdcomp for cells expressing the sensor protein
at low, medium, and high densities to investigate if sensor
proteins interact in an intermolecular fashion at high local sensor
comp
protein concentrations. In the latter case, the measured Kd
would become dependent on the concentration of the sensor
protein. We found that the Kdcomp,SA did not significantly change
for different cells expressing the sensor protein at varying
concentrations on the cell surface (Supporting Information),
indicating that the binding of the attached ligand is indeed
predominantly intramolecular even at higher local sensor
protein concentrations.
Variety of Synthetic Fluorophores Can Be Applied for
Sensor Construction. Synthetic fluorophores provide spectro-
scopic properties, such as high photostability or infrared excita-
tion and emission maxima, which cannot be found in any of the
currently existing FPs.17,18 Since the use of self-labeling tags
permits the inclusion of any synthetic fluorophore into the sensor
proteins, we wanted to demonstrate the versatility of the sensor
concept in terms of the employed fluorophore pair for FRET. For
this purpose, we labeled our sensor protein on the cell surface
with two different fluorophore pairs, Alexa Fluor 488/DY-547
and Alexa Fluor 488/Alexa Fluor 594. Both combinations
showed benzenesulfonamide-dependent donor/acceptor inten-
sity ratio changes. We found a ΔRmax of 1.3 ( 0.1 for the
combination Alexa Fluor 488/DY-547 (Figure S2) and a ΔRmax
of 1.7 ( 0.1 for the Alexa Fluor 488/Alexa Fluor 594 FRET pair
(Figure S3). Control experiments with the corresponding BG
substrates lacking the terminal benzenesulfonamide moiety again
did not exhibit any donor/acceptor intensity ratio change,
demonstrating that the response of the biosensor is due to a
specific interaction between HCA and the intramolecular ligand
(Figures S4, S5).
Figure 2. Substrates for SNAP-tag sensor protein labeling. (A) BG-
PEG11-Cy5-paraSA 1. BG-PEG11-Cy5-metaSA 2. (B) BG-PEG11-Alexa
Fluor 488-SA 4. (C) BG-PEG11-Alexa Fluor 594-SA 5. The molecules 3,
6, and 7 are the corresponding control molecules lacking the terminal
benzenesulfonamide moiety.
To assemble the sensor at the extracellular surface of mammalian
cells, we labeled HEK 293T cells expressing SNAP_CLIP_HCA on
the cell surface with BG-PEG11-Cy5-SA 1 (2 μM) and CLIP-
Surface 547 (10 μM) for 10 min at RT (Figure 3A). CLIP-Surface
547 is a substrate for labeling CLIP-tag with the fluorescent dye DY-
547. To determine the response of the sensor at the cell surface to
added inhibitor, we applied a solution of 10 mM benzenesulfona-
mide in HBSS to the HEK 293T cells by using a perfusion system,
and monitored the intensity ratio of DY-547/Cy5 fluorescence by
microscopy. We found a sulfonamide-dependent DY-547/Cy5
ΔRmax of 1.5 ( 0.1 that was fully reversible upon perfusion of the
HEK 293T cells with HBSS without inhibitor (Figure 3C). In order
The use of two synthetic dyes for the construction of cell
surface biosensors avoids another problem of FP-based biosen-
sors. By employing membrane-impermeable dyes, only sensor
proteins that are properly trafficked to the cell surface are labeled.
To illustrate this, we constructed a cell surface-targeted version of
the sensor protein SNAP_mCherry_HCA, in which CLIP-tag is
16237
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
Figure 3. SNAP_CLIP_HCA as a sensor for sulfonamides on the surface of HEK 293T cells. (A) Donor channel (DY-547), FRET channel (Cy5), and
transmission channel of the labeled sensor protein on HEK 293T cells. Images were taken using a Leica LAS AF 7000 wide-field microscope. (B) Ratio
images of the labeled sensor protein in absence of free benzenesulfonamide (t = 0 s, t = 1200 s) and in the presence of 10 mM free benzenesulfonamide
(t = 600 s). (C) Time course of sensor opening (upon addition of 10 mM benzenesulfonamide) and of sensor closing (upon removal of the free
benzenesulfonamide). The red bar indicates the time span of perfusion with benzenesulfonamide.
replaced by the red FP mCherry (Figure S6A). SNAP_mCherry_
HCA and SNAP_CLIP_HCA perform very similarly as purified
sensors in vitro.15 As expected, SNAP_mCherry_HCA can also
be labeled on the cell surface via SNAP-tag. However, it also
exhibitsa strong fluorescent signal in the secretory pathway due to
the inherent fluorescence of mCherry (Figure S6B). Evidently,
this intracellular background fluorescence makes the analysis of
the FRET sensors on the cell surface more demanding, especially
in cases where the location of the sensor protein cannot clearly
be resolved by confocal microscopy, such as for the analysis of
cells in living tissue or for high-throughput screening applica-
tions in microtiterplates.
Kinetic Characterization of the Sensor Proteins. The temporal
resolution of our sensor proteins is determined by the rate of
sensor opening and sensor closing. Ideally, these rates should
exceed the rate of the small molecule’s concentration change. The
kinetic requirements of the sensor protein heavily depend on the
type of molecule; e.g., the concentration change of most hor-
mones takes places within minutes. On the other hand, neuro-
transmitter release and clearance occurs on a millisecond time
scale. For our sensor protein, sensor opening and sensor closing
are composed of two separate steps, i.e., the unbinding of the
intramolecular ligand followed by the binding of the free ligand
for sensor opening, and the unbinding of the free ligand followed
by the binding of the intramolecular ligand for sensor closing
(Figure S7). As the temporal resolution of a biosensor is
important for biological applications, we wanted to characterize
the kinetics mechanisms of opening and closing of our sensor
protein. For that purpose, we measured the rates of sensor
opening and closing for our sensor protein upon addition and
removal of benzenesulfonamide and then also for two additional
inhibitors of HCA, i.e., methazolamide, and ethoxzolamide. All
three inhibitors possess significantly different kon and koff values
(Table S1)19 and should therefore also generate different opening
and closing rates of the sensor protein. HEK 293T cells expressing
the sensor protein were consecutively perfused with 10 mM
benzenesulfonamide, 1 mM methazolamide, and 10 μM ethox-
zolamide (Figure 4C). All concentrations were chosen to set the
sensor protein at saturating (open state) conditions. For the
perfusion with 10 mM benzenesulfonamide, we found best-fit
parameters of t1/2,open = 60 ( 5 s for sensor opening and of
t1/2,close = 40 ( 10 s for sensor closing by fitting the DY-547/Cy5
ratio time course to a single exponential function (Figure S8).
Analogously, we measured t1/2,open =65(5sandt1/2,close =65(15 s
for the perfusion with 1 mM methazolamide, and t1/2,open = 80 (
15 s and t1/2,close = 150 ( 15 s for the perfusion with 10 μM
ethoxzolamide. The t1/2,close value for ethoxzolamide was signifi-
cantly larger than the t1/2,close value for benzenesulfonamide. This
is in agreement with its slower koff values compared to benzene-
sulfonamide (Tables S1, S2) and suggests that the rate of sensor
closing mainly depends on the unbinding of the free ligand and
not on the (re)binding of the intramolecular ligand because the
latter step is the same for all three inhibitors. In contrast, we found
that the t1/2,open value for sensor opening did not change
significantly for the three inhibitors. This indicates that the rate
of sensor opening mainly depends on the unbinding of the
intramolecular ligand. In agreement with this hypothesis, we found
that by changing the intramolecular para-substituted benzenesul-
fonamide to a meta-substituted benzenesulfonamide, which ex-
hibits a faster koff value toward HCA,19 the sensor opens about 10
times faster upon perfusion with 500 μM benzenesulfonamide
(t1/2,open = 7 ( 2 s) (Figure 4D, Figures S9, S10). The t1/2,close for
sensor closing does not change significantly compared to the para-
substituted benzenesulfonamide (t1/2,close = 25 ( 5 s) and
confirms that the rate of sensor closing mainly depends on the
unbinding of the free ligand. These experiments demonstrate how
it is possible to tune the kinetics of sensor opening by choosing
an appropriate intramolecular ligand. The rate of sensor closing
16238
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
Figure 4. Kinetic and thermodynamic characterization of the sensor protein SNAP_CLIP_HCA on the surface of HEK 293T cells. (A) Time course of
perfusion of labeled SNAP_CLIP_HCA with increasing concentrations of benzenesulfonamide (a = 1 μM, b = 10 μM, c = 50 μM, d = 100 μM, e = 500
μM, f = 1 mM, g = 10 mM). (B) Benzenesulfonamide titration curve of SNAP_CLIP_HCA on the extracellular surface of HEK 293T cells. (C) Time
course of sensor opening and closing upon perfusion with 10 mM benzenesulfonamide (a), with 1 mM methazolamide (b), and with 10 μM
ethoxzolamide (c). (D) Comparison of the opening rate between a sensor protein labeled with the terminal para-benzenesulfonamide (solid line) and a
sensor protein labeled with the terminal meta-benzenesulfonamide (dotted line) upon addition of 10 mM benzenesulfonamide.
mainly depends on the unbinding of the analyte and is therefore a
characteristic of the binding protein.
of the FRET sensor10,23 in combination with a suitable screen-
ing technique to analyze sensor mutant libraries. However,
while there are an impressive number of successful examples
for the optimization of FRET-based sensors, there is still no
general way that allows the design and improvement of a
FRET sensor protein in a rational way.
The key to rationally optimize a FRET sensor is to understand
and exploit its molecular mechanism. On the basis of our sensor
model we hypothesized that shortening the linker between
CLIP-tag and HCA would increase FRET efficiency in the closed
state while increasing the linker length between SNAP-tag and
CLIP-tag should decrease FRET efficiency in the open state.15 In
both cases the result should be an increase of the performance of
the sensor protein. Our original sensor protein SNAP_CLIP_H-
CA contains a flexible linker of 18 amino acids between SNAP-
tag and CLIP-tag and a flexible linker of six amino acids between
CLIP-tag and HCA, respectively. Despite their frequent use, flexible
peptide linkers are not a very efficient means to gain distance
between two proteins because these linkers can occupy a wide range
of conformations in space.23,24 Therefore, we pursued a different
strategy to alter the effective linker length. Poly-L-proline linkers
have been used for a long time as precise molecular rulers due to
Rational Optimization of the Semisynthetic Sensor Pro-
tein Scaffold. After having demonstrated the successful imple-
mentation of our Snifit sensor proteins on the cell surface, we set
out to investigate whether it is possible to optimize the sensor
protein scaffold regarding the dynamic range of its signal change.
The maximum ratio change of a FRET-based sensor is one of the
key parameters for its efficient application in a cellular setting or
in a whole organism. It determines whether the sensor exhibits a
sufficient signal over noise and can thus be used to measure a
reliable signal of the sensor response. Although there is no clear
threshold above which the dynamic range of a certain FRET
sensor is said to be sufficiently high, it is generally acknowledged
that the higher the dynamic range of a sensor is, the more
reliably its signal can be interpreted.20 Since the majority of
FRET sensors based on FPs still suffer from a rather small
dynamic range, the optimization of FRET sensors in terms of
their dynamic range is an area of active research.20 Strategies
for the optimization of FRET sensors include the use of
improved fluorescent proteins,21 different FRET pairs,22 and
the mutations of the peptide linkers between the protein units
16239
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
Figure 5. Rational optimization of the semisynthetic sensor protein for sulfonamides. (A) The insertion of rigid polyproline linkers should lead to an
increased distance between the two fluorophores in the open state of the sensor protein and thereby also to less FRET. The distance of the two
fluorophores in the closed state of the sensor protein should not be significantly affected. Consequently, the maximum ratio change should increase. (B)
Emission spectra of SNAP_PP0_CLIP_HCA at a low and a high concentration of benzenesulfonamide (100 pM, black; 10 mM, white). The addition of
benzenesulfonamide leads to an increase in the Cy3/Cy5 emission ratio with a maximum ratio change of 1.9 ( 0.1. (C) Emission spectra of
SNAP_PP30_CLIP_HCA at a low and a high concentration of benzenesulfonamide (100 pM, black; 10 mM, white). The addition of
benzenesulfonamide leads to an increase in the Cy3/Cy5 emission ratio with a maximum ratio change of 4.9 ( 0.3. (D) Fluorescence titration curves
of the optimized sensor proteins. Shown is the ratio of fluorescence of donor (570 nm) and acceptor emission (670 nm) obtained by titrating the
corresponding labeled sensor mutant with benzenesulfonamide. The data are represented as the mean ( standard deviation of triplicates. The sensor
data are fitted according to a single-site binding isotherm. (E) Bar graph of the measured maximum ratio changes for each sensor mutant. For
comparison, the maximum ratio change of the original sensor protein15 is also displayed. All maximum ratio changes for the different sensor mutants
were obtained by dividing the Fmax value by the F0 value obtained after fitting the sensor mutant’s benzenesulfonamide titration data (Supporting
Information). The data are represented as the mean ( standard deviation of triplicates.
their well-defined property of forming a stable and rigid helical
structure (the polyproline II helix) with a pitch of 3.1 Å per
residue in aqueous solution.25À27 We constructed seven sensor
mutants in which the peptide linkers between CLIP-tag and
HCA were removed, and polyproline linkers of varying length
(0, 6, 9, 12, 15, 30, 60) wereinserted betweenSNAP-and CLIP-tag
to replace the original peptide linker. According to our model,
this should increase the distance between the fluorophores mainly
in the open state of the sensor and should thus result in an
increased maximum ratio change (Figure 5A).
All seven sensor mutants could be successfully expressed and
purified from E. coli (Figure S11). After labeling with BG-PEG11-
Cy5-SA and with CLIP-Surface 547, the mutants were tested for
their ability to sense sulfonamide in vitro. We found that increasing
16240
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
In a next step, we wanted to test the performance of
our two best sensor mutants SNAP_PP15_CLIP_HCA and
SNAP_PP30_CLIP_HCA on the cell surface of mammalian
cells. Both sensor proteins could be successfully expressed and
labeled on the surface of HEK 293T cells. The expression level of
the sensor mutants was found to be very similar to the one of
the original SNAP_CLIP_HCA sensor protein (Figure S15);
thus, the introduction of long polyproline stretches into the
sensor protein does not seem to influence the expression and
the trafficking of the sensor protein to the cell surface. Upon
perfusion with 10 mM benzenesulfonamide, we measured a
maximum ratio change of 2.1 ( 0.1 for the sensor protein
SNAP_PP15_CLIP_HCA (Figure S16), and a maximum
ratio change of 3.3 ( 0.4 for the most optimized sensor
protein SNAP_PP30_CLIP_HCA (Figure 6). This signifies a
major improvement compared to the original sensor protein
(ΔRmax of 1.5) and proves that our sensor proteins can be
rationally optimized for their application on the cell surface.
Further, the ΔRmax of our best sensor protein is clearly superior
than PBP-based FRET reporters displayed on cell surfaces, which
have not yet achieved a ΔRmax of more than 1.5.10
Figure 6. Optimized sensor protein SNAP_PP30_CLIP_HCA on the
surface of HEK 293T cells. (A) Time course of the donor channel of the
sensor protein SNAP_PP30_CLIP_HCA upon addition and removal of
10 mM benzenesulfonamide. The addition of 10 mM benzenesulfona-
mide leads to an increased donor emission. (B) Time course of the
acceptor channel of the sensor protein SNAP_PP30_CLIP_HCA upon
addition and removal of 10 mM benzenesulfonamide. The addition of
10 mM benzenesulfonamide leads to a decreased acceptor emission. (C)
Time course of the intensity ratio of donor emission vs acceptor
emission upon addition and removal of 10 mM benzenesulfonamide.
The red bar indicates the time span of perfusion with benzenesulfonamide.
’ CONCLUSIONS
In summary, we have demonstrated the successful implemen-
tation of our Snifit sensor proteins on the surface of mammalian
cells. This is an important accomplishment as the adaption of
sensors to applications in living cells represents a major challenge
in the field. Further, a thorough characterization of the Snifit
sensor proteins has been carried out including the description of
its thermodynamic and kinetic properties. We have also success-
fully validated our proposed optimization guidelines thereby
proving that it is possible to rationally optimize Snifit sensor
proteins. The rational engineering of the peptide linkers of the
Snifit sensor protein allowed an improvement of the sensor’s
dynamic range from 1.8 to 4.9 in vitro. On the cell surface, the
maximum ratio change could be improved from 1.5 to 3.3,
thereby producing FRET sensor proteins for the cell surface
with unprecedented maximum ratio changes.
The use of self-labeling tags instead of FPs for Snifit sensor
construction offers several advantages: (i) a large choice of
fluorophores with tailor-made properties is available for the
construction of the sensor protein starting from a single sensor
construct; (ii) the application of membrane-impermeable dyes
allows an exclusive labeling of the subpopulation of sensor
proteins that are properly trafficked to the cell surface; and
(iii) the sensor response rate upon emergence of the metabolite
of interest can be tuned by choosing an appropriate intramole-
cular ligand. Finally, since our sensor concept circumvents the
need for a conformational change of the binding protein, the
herein presented work provides the necessary basis for the
generation of sensor proteins for the detection of previously
inaccessible metabolites on the cell surface of mammalian cells.
the length of the polyproline linker between SNAP- and CLIP-
tag also increases the maximum ratio change of the sensor mutant
increases. The best sensor mutant SNAP_PP60_CLIP_HCA
displayed a maximum ratio change of 6.9 ( 0.7 (Figure S12);
comp
however, its Kd
value was lower than those of the other
constructs, which showed no significant differences (Figure S13).
For all further studies we therefore focused on the mutant SNAP-
PP30-CLIP-HCA, which displays a maximum ratio change of
4.9 ( 0.3 (Figure 5). Our findings are in accordance with our
optimization predictions and prove that increasing the linker
length between SNAP- and CLIP-tag indeed leads to improved
sensor mutants. Further, the similar Kdcomp values measured for
our sensor mutants up to a linker length of 30 prolines indicate
that the insertion of these peptide linkers does not significantly
change the effective molarity of the intramolecular ligand; i.e.,
they do not seem to induce an intramolecular aggregation or to
hinder sensor closing.
According to the sensor model, decreasing the linker between
CLIP-tag and HCA should lead to a shorter distance between the
two fluorophores in the closed state and therefore to an opti-
mized sensor mutant. In our polyproline sensor mutants we have
already removed the six amino acid linker between CLIP-tag and
HCA. A further deletion of amino acids from the N-terminus of
HCA results in a reduced stability of the protein.28,29 Thus, to
test our mechanistic predictions, we decided to rather construct a
sensor mutant with an increased linker length between CLIP-tag
and HCA. According to our optimization guidelines this sensor
protein should then display a decreased maximum ratio change
compared to the original sensor protein. Insertion of 15 proline
residues between CLIP-tag and HCA resulted in sensor protein
SNAP_CLIP_PP15_HCA. In accordance with our predic-
tions, we found that the insertion of 15 proline residues
between CLIP-tag and HCA almost completely eliminated
the signal change of the sensor protein upon addition of ben-
zenesulfonamide (Figure S14).
’ METHODS
Synthesis. Detailed synthetic procedures and characterizations are
described in the Supporting Information.
Cell Culture and Transfection. HEK 293T cells were grown on
polylysine coated glass coverslips (Ø 15 mm) in DMEM Glutamax
medium (Lonza) supplemented with 10% fetal bovine serum (Lonza)
and transiently transfected by using Lipofectamine (Invitrogen) accord-
ing to the manufacturer’s protocol.
16241
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242

Journal of the American Chemical Society
ARTICLE
SNAP- and CLIP-Tag Labeling on the Cell Surface of HEK
293T Cells. Twenty-four hours after transfection, HEK 293T cells were
labeled with a solution of 2 μM of the corresponding BG derivative and
10 μM of the corresponding O2-benzylcytosine (BC) derivative in
Hank’s buffered salt solution (HBSS) complemented with 10 mg/mL
BSA for 10 min at room temperature. After being labeled, cells were
washed four times with HBSS.
Confocal Microscopy. Images of HEK 293T cells labeled with
CLIP-Surface 547 and with SNAP-Surface 647 were taken using a Zeiss
LSM 700 confocal microscope equipped with a 40Â plan Apochromat
1.3 numerical aperture (NA) oil immersion objective lens. The imaging
was performed using a 555 nm laser line for excitation of DY-547 and a
639 nm laser line for excitation of DY-647. Fluorescence was collected at
400À620 nm and at 620À700 nm for DY-547 and DY-647, respectively.
The settings for scanning were: Â2 zoom, image format 1024 Â 1024
pixels, pinhole 1Airy unit (AU), average 16 frames.
Wide-Field Microscopy and Live Cell FRET Imaging. Glass
coverslips with labeled HEK 293T cells were transferred to a Warner
imaging chamber (RC-20). Perfusion of the chamber was performed
gravity-fed at a flow rate of 0.5 mL/min. Time-course experiments of
sensor imaging were performed using a Leica LAS AF 7000 wide-field
microscope equipped with a 40Â plan Apochromat 1.25 NA oil
immersion objective lens. A xenon arc lamp was used for imaging of
the HEK 293T cells. For each frame, the two channels (donor and
FRET) were measured consecutively with an interval of 30 ms between
the two emission channels. The following filter sets were used for the
FRET ratio imaging: for DY-547/Cy5, excitation at 530 nm (bandwidth
35 nm), emission at 580 nm (bandwidth 40 nm) (DY-547) and at
700 nm (bandwidth 72 nm) (Cy5); for Alexa Fluor 488/DY-547, excitation
at 470 nm (bandwidth 40 nm), emission at 520 nm (bandwidth 40 nm)
(Alexa Fluor 488) and at 605 (bandwidth 70 nm) (DY-547); for Alexa
Fluor 488/Alexa Fluor 594, excitation at 470 (bandwidth 40 nm),
emission at 520 nm (bandwidth 40 nm) (Alexa Fluor 488) and at
632 nm (bandwidth 60 nm) (Alexa Fluor 594). If not indicated
otherwise, the image size was 293 μM Â 293 μM, and an average of 5
cells per image were analyzed for the intensity ratio plots.
’ REFERENCES
(1) Wishart, D. S.;; et al. Nucleic Acids Res. 2009, 37, D603–D610.
(2) Sassone-Corsi, P.; Nakahata, Y.; Sahar, S.; Astarita, G.; Kaluzova,
M. Science 2009, 324 (5927), 654–657.
(3) Zhang, J.; Newman, R. H.; Fosbrink, M. D. Chem. Rev. 2011,
111 (5), 3614–3666.
(4) Breaker, R. R. Nature 2004, 432 (7019), 838–845.
(5) Navani, N. K.; Li, Y. F. Curr. Opin. Chem. Biol. 2006, 10 (3),
272–281.
(6) Ibraheem, A.; Campbell, R. E. Curr. Opin. Chem. Biol. 2010, 14 (1),
30–36.
(7) Wiechert, W.; Schweissgut, O.; Takanaga, H.; Frommer, W.
Curr. Opin. Plant Biol. 2007, 10 (3), 323–330.
(8) Okumoto, S. Curr. Opin. Biotechnol. 2010, 21 (1), 45–54.
(9) Fehr, M.; Takanaga, H.; Ehrhardt, D.; Frommer, W. Mol. Cell.
Biol. 2005, 25 (24), 11102–11112.
(10) Hires, S.; Zhu, Y.; Tsien, R. Proc. Natl. Acad. Sci. U.S.A. 2008,
105 (11), 4411–4416.
(11) Lager, I.; Looger, L.; Hilpert, M.; Lalonde, S.; Frommer, W.
J. Biol. Chem. 2006, 281 (41), 30875–30883.
(12) Okumoto, S.; Looger, L.; Micheva, K.; Reimer, R.; Smith, S.;
Frommer, W. Proc. Natl. Acad. Sci. U.S.A. 2005, 102 (24), 8740–8745.
(13) Medintz, I. Trends Biotechnol. 2006, 24 (12), 539–542.
(14) Deuschle, K.; Fehr, M.; Hilpert, M.; Lager, I.; Lalonde, S.;
Looger, L.; Okumoto, S.; Persson, J.; Schmidt, A.; Frommer, W.
Cytometry, Part A 2005, 64A (1), 3–9.
(15) Brun, M. A.; Tan, K. T.; Nakata, E.; Hinner, M. J.; Johnsson, K.
J. Am. Chem. Soc. 2009, 131 (16), 5873–5884.
(16) Keppler, A.; Gendreizig, S.; Gronemeyer, T.; Pick, H.; Vogel,
H.; Johnsson, K. Nat. Biotechnol. 2003, 21 (1), 86–89.
(17) Keppler, A.; Arrivoli, C.; Sironi, L.; Ellenberg, J. Biotechniques
2006, 41, (2), 167.
(18) Lavis, L.; Raines, R. ACS Chem. Biol. 2008, 3 (3), 142–155.
(19) Krishnamurthy, V. M.; Kaufman, G. K.; Urbach, A. R.; Gitlin, I.;
Gudiksen, K. L.; Weibel, D. B.; Whitesides, G. M. Chem. Rev. 2008, 108
(3), 946–1051.
(20) Campbell, R. E. Anal. Chem. 2009, 81 (15), 5972–5979.
(21) Nguyen, A. W.; Daugherty, P. S. Nat. Biotechnol. 2005, 23 (3),
355–360.
(22) Ouyang, M. X.; Sun, J.; Chien, S.; Wang, Y. X. Proc. Natl. Acad.
Sci. U.S.A. 2008, 105 (38), 14353–14358.
(23) Golynskiy, M. V.; Rurup, W. F.; Merkx, M. ChemBioChem
2010, 11 (16), 2264–2267.
(24) Evers, T.; van Dongen, E.; Faesen, A.; Meijer, E.; Merkx, M.
Biochemistry 2006, 45 (44), 13183–13192.
(25) Arora, P. S.; Ansari, A. Z.; Best, T. P.; Ptashne, M.; Dervan, P. B.
J. Am. Chem. Soc. 2002, 124 (44), 13067–13071.
(26) Stryer, L.; Haugland, R. P. Proc. Natl. Acad. Sci. U.S.A. 1967, 58,
(2), 719.
’ ASSOCIATED CONTENT
S
Supporting Information. Synthetic procedures, charac-
b
terizations of the sensor proteins on the cell surface and in vitro,
and complete ref 1. Additional figures and tables. This material is
available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
(27) Schuler, B.; Lipman, E. A.; Steinbach, P. J.; Kumke, M.; Eaton,
W. A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102 (8), 2754–2759.
(28) Aronsson, G.; Martensson, L. G.; Carlsson, U.; Jonsson, B. H.
Biochemistry 1995, 34 (7), 2153–2162.
(29) Vince, J. W.; Carlsson, U.; Reithmeier, R. A. F. Biochemistry
2000, 39 (44), 13344–13349.
Corresponding Author
kai.johnsson@epfl.ch
Present Addresses
†Department of Chemistry, National Tsing Hua University,
Hsinchu 30013, Taiwan.
‡Institute for Molecules and Materials, Department of Organic
Chemistry, Radboud University Nijmegen, 6525 AJ Nijmegen,
The Netherlands.
’ ACKNOWLEDGMENT
The authors thank Dr. M.J. Hinner, Dr. G. Lukinavicius, and
A. Schena for helpful discussions and Dr. D. Maurel for technical
assistance. This work was supported by the Swiss National
Science Foundation.
16242
dx.doi.org/10.1021/ja206915m |J. Am. Chem. Soc. 2011, 133, 16235–16242