Development of an optimized backbone of FRET biosensors for kinases and GTPases.
ABSTRACT Biosensors based on the principle of Förster (or fluorescence) resonance energy transfer (FRET) have shed new light on the spatiotemporal dynamics of signaling molecules. Among them, intramolecular FRET biosensors have been increasingly used due to their high sensitivity and user-friendliness. Time-consuming optimizations by trial and error, however, obstructed the development of intramolecular FRET biosensors. Here we report an optimized backbone for rapid development of highly sensitive intramolecular FRET biosensors. The key concept is to exclude the "orientation-dependent" FRET and to render the biosensors completely "distance-dependent" with a long, flexible linker. We optimized a pair of fluorescent proteins for distance-dependent biosensors, and then developed a long, flexible linker ranging from 116 to 244 amino acids in length, which reduced the basal FRET signal and thereby increased the gain of the FRET biosensors. Computational simulations provided insight into the mechanisms by which this optimized system was the rational strategy for intramolecular FRET biosensors. With this backbone system, we improved previously reported FRET biosensors of PKA, ERK, JNK, EGFR/Abl, Ras, and Rac1. Furthermore, this backbone enabled us to develop novel FRET biosensors for several kinases of RSK, S6K, Akt, and PKC and to perform quantitative evaluation of kinase inhibitors in living cells.
- SourceAvailable from: Daniel Ritt[Show abstract] [Hide abstract]
ABSTRACT: Protein scaffolds play an important role in signal transduction, functioning to facilitate protein interactions and localize key pathway components to specific signaling sites. Connector enhancer of KSR-2 (CNK2) is a neuronally expressed scaffold recently implicated in nonsyndromic, X-linked intellectual disability (NS-XLID) [1-3]. NS-XLID patients have deficits in cognitive function and their neurons often exhibit dendritic spine abnormalities , suggesting a role for CNK2 in synaptic signaling and/or spine formation. To gain insight regarding how CNK2 might contribute to these processes, we used mass spectrometry to identify proteins that interact with the endogenous CNK2 scaffold. Here, we report that the major binding partner of CNK2 is Vilse/ARHGAP39 and that CNK2 complexes are enriched for proteins involved in Rac/Cdc42 signaling, including Rac1 itself, α-PIX and β-PIX, GIT1 and GIT2, PAK3 and PAK4, and members of the cytohesin family. Binding between CNK2 and Vilse was found to be constitutive, mediated by the WW domains of Vilse and a proline motif in CNK2. Through mutant analysis, protein depletion and rescue experiments, we identify CNK2 as a spatial modulator of Rac cycling during spine morphogenesis and find that the interaction with Vilse is critical for maintaining RacGDP/GTP levels at a balance required for spine formation.Current biology: CB 03/2014; · 10.99 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Integrating biological imaging into early stages of the drug discovery process can provide invaluable readouts of drug activity within complex disease settings, such as cancer. Iterating this approach from initial lead compound identification in vitro to proof-of-principle in vivo analysis represents a key challenge in the drug discovery field. By embracing more complex and informative models in drug discovery, imaging can improve the fidelity and statistical robustness of preclinical cancer studies. In this Review, we highlight how combining advanced imaging with three-dimensional systems and intravital mouse models can provide more informative and disease-relevant platforms for cancer drug discovery.Nature Reviews Cancer 04/2014; · 29.54 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: It has been over 20 years since JUN amino-terminal kinases (JNKs) were identified as protein kinases that are strongly activated by cellular stress and that have a key role in apoptosis. Examination of Jnk-knockout mice and characterization of JNK behaviour in neuronal cells has further revealed the importance of the JNK family in the nervous system. As well as regulating neuronal death, JNKs govern brain morphogenesis and axodendritic architecture during development, and regulate important neuron-specific functions such as synaptic plasticity and memory formation. This Review examines the evidence that the spatial segregation of JNKs in neurons underlies their distinct functions and that compartment-specific targeting of JNKs may offer promising new therapeutic avenues for the treatment of diseases of the nervous system, such as stroke and neurodegenerative disorders.Nature Reviews Neuroscience 04/2014; 15(5):285-99. · 31.38 Impact Factor
Volume 22 December 1, 2011
MBoC | ARTICLE
Development of an optimized backbone of FRET
biosensors for kinases and GTPases
Naoki Komatsua, Kazuhiro Aokia,b, Masashi Yamadac, Hiroko Yukinagac, Yoshihisa Fujitac,
Yuji Kamiokac,d, and Michiyuki Matsudaa,c
aLaboratory of Bioimaging and Cell Signaling, Graduate School of Biostudies, Kyoto University, Kyoto 606-8501,
Japan; bPREST, Japan Science and Technology Agency (JST), Saitama 332-0012, Japan; cDepartment of Pathology
and Biology of Diseases, Graduate School of Medicine, and dInnovative Techno-Hub for Integrated Medical
Bio-Imaging, Kyoto University, Kyoto 606-8501, Japan
ABSTRACT Biosensors based on the principle of Förster (or fluorescence) resonance energy
transfer (FRET) have shed new light on the spatiotemporal dynamics of signaling molecules.
Among them, intramolecular FRET biosensors have been increasingly used due to their high
sensitivity and user-friendliness. Time-consuming optimizations by trial and error, however,
obstructed the development of intramolecular FRET biosensors. Here we report an opti-
mized backbone for rapid development of highly sensitive intramolecular FRET biosensors.
The key concept is to exclude the “orientation-dependent” FRET and to render the biosen-
sors completely “distance-dependent” with a long, flexible linker. We optimized a pair of
fluorescent proteins for distance-dependent biosensors, and then developed a long, flexible
linker ranging from 116 to 244 amino acids in length, which reduced the basal FRET signal
and thereby increased the gain of the FRET biosensors. Computational simulations provided
insight into the mechanisms by which this optimized system was the rational strategy for in-
tramolecular FRET biosensors. With this backbone system, we improved previously reported
FRET biosensors of PKA, ERK, JNK, EGFR/Abl, Ras, and Rac1. Furthermore, this backbone
enabled us to develop novel FRET biosensors for several kinases of RSK, S6K, Akt, and PKC
and to perform quantitative evaluation of kinase inhibitors in living cells.
Förster (or fluorescence) resonance energy transfer (FRET) is a
process of nonradiative energy transfer between donor and ac-
ceptor fluorophores (Jares-Erijman and Jovin, 2003). This process
depends on the proper spectral overlap of the donor emission
and acceptor excitation, the distance between them, and the
relative orientation of the fluorophore’s transition dipole mo-
ments (Miyawaki, 2003). With the advent of a myriad of fluores-
cent proteins (FPs), genetically encoded biosensors based on
FRET (hereafter referred to as FRET biosensors) have been in-
creasingly used to visualize the activities of cellular signaling mol-
ecules such as Ca2+, phospholipids, small GTPases, protein ki-
nases, and so forth (Miyawaki, 2003; Aoki et al., 2008). These
FRET biosensors have contributed to our understanding of the
spatiotemporal dynamics of signaling molecules in living cells,
which could not be adequately investigated using the techniques
of conventional biochemistry.
The genetically encoded FRET biosensors are classified into two
types: intramolecular (or unimolecular) FRET biosensors and inter-
molecular (or bimolecular) FRET biosensors (Miyawaki, 2003). The
former contain both donor and acceptor FPs within a single biosen-
sor, whereas the latter consist of a pair of molecules conjugated with
a donor FP and an acceptor FP, respectively. To date, intramolecular
FRET biosensors have been widely used in cell biology due to the
following advantages: high signal-to-noise ratio, easy loading of the
biosensor into the cells, and simple ratiometric image analysis
(Miyawaki, 2003; Aoki et al., 2008).
University of California,
Received: Jan 26, 2011
Revised: Sep 7, 2011
Accepted: Sep 26, 2011
The authors declare competing financial interest: M.M. filed a patent application
for the reported linker: Japan patent application, 2010.
Address correspondence to: Kazuhiro Aoki (email@example.com).
Abbreviations used: CFP, cyan fluorescent protein; cp, circularly permutated;
dbcAMP, dibutyryl-cyclical AMP; ECFP, enhanced CFP; Eevee, extension for
enhanced visualization by evading extraFRET; FP, fluorescent protein; FRET, Förster
(or fluorescence) resonance energy transfer; GAP, GTPase-activating protein; PH,
Pleckstrin homology, TFP, teal fluorescent protein; TPA, tetradecanoylphorbol
13-acetate; YFP, yellow fluorescent protein.
“ASCB®,“ “The American Society for Cell Biology®,” and “Molecular Biology of
the Cell®” are registered trademarks of The American Society of Cell Biology.
© 2011 Komatsu et al. This article is distributed by The American Society for Cell
Biology under license from the author(s). Two months after publication it is avail-
able to the public under an Attribution–Noncommercial–Share Alike 3.0 Unported
Creative Commons License (http://creativecommons.org/licenses/by-nc-sa/3.0).
This article was published online ahead of print in MBoC in Press (http://www
.molbiolcell.org/cgi/doi/10.1091/mbc.E11-01-0072) on October 5, 2011.
4648 | N. Komatsu et al. Molecular Biology of the Cell
A serious disadvantage of the intramolecular FRET biosensors is
that it is difficult to render them highly sensitive. Although several
research groups have generated a number of FP variants that are
optimized for FRET applications (Karasawa et al., 2004; Rizzo et al.,
2004; Nguyen and Daugherty, 2005), the development of FP pairs
for FRET is still one of the greatest challenges in FRET biosensor
development (Li et al., 2006). Even if we could choose optimal FP
pairs, there would remain the harder task of designing the composi-
tion of domains to be included in the biosensors. The FRET efficiency
of intramolecular biosensors is influenced primarily by the distance
and the relative orientation of the two fluorophores (Jares-Erijman
and Jovin, 2003; Nagai et al., 2004). Because these two parameters
are hard to predict, the developers are forced to spend large amounts
of time in optimizing the biosensors by trial and error.
To circumvent these problems, we exploited an optimized back-
bone of intramolecular FRET biosensors. We first optimized the FPs
and then developed a long, flexible linker that markedly increased
the gain of FRET signals. This new backbone enabled us to improve
the previously reported FRET biosensors of PKA, ERK, JNK, EGFR/
Abl, Ras, and Rac1 and to develop new FRET biosensors for RSK,
S6K, Akt, and PKC Ser/Thr kinases in a short period of time. Further-
more, by using cells stably expressing a FRET biosensor, we quanti-
tatively and rapidly evaluated the effect of several kinase inhibitors
on ERK activity.
Strategy for rational design of intramolecular
To accelerate the development of the intramolecular FRET biosen-
sors, we attempted to provide an optimized backbone structure.
The prototype FRET biosensor used in this study comprises 1) a sen-
sor domain and a ligand domain connected by a flexible linker, and
2) cyan and yellow FPs (CFP and YFP) serving as a donor and an ac-
ceptor, respectively (Figure 1A). The sensor domain changes its con-
formation upon perception of the signal. This sensitized sensor do-
main interacts with the ligand domain, thereby inducing a global
change of the biosensor conformation and concomitant increase (or
decrease in some cases) of the FRET efficiency from the donor to
The FRET efficiency of the intramolecular FRET biosensor is
dependent primarily on the distance and the relative orientation
of the donor and the acceptor (Miyawaki, 2003). Although the
“orientation-dependent” FRET biosensor may be able to exhibit
higher sensitivity than does the “distance-dependent” FRET biosen-
sor (Jares-Erijman and Jovin, 2003), we could hardly predict and con-
trol the orientation of the donor and the acceptor for the develop-
ment of an optimal FRET biosensor because we usually do not know
the three-dimensional structures of the biosensor of which the sen-
sor domain is bound (“ON”) and not bound (“OFF”) to the ligand
domain. Thus the main job of our intramolecular FRET biosensors is
to eliminate the orientation-dependent FRET and to render the bio-
sensors completely distance-dependent with a long, flexible linker.
Before the evaluation of FRET biosensors described in this study,
we will define the technical terms related to their performance. In the
present FRET biosensors, the CFP and YFP variants are mostly used
as the donor and the acceptor, respectively; therefore we will de-
scribe CFP and YFP as the default donor and the acceptor, respec-
tively. FRET is detected by ratiometry (Jares-Erijman and Jovin, 2003):
Cells are excited at a 440 nm wavelength, and the ratio of fluores-
cence intensity of the YFP channel (FRET) versus fluorescence inten-
sity of the CFP channel (CFP), FRET/CFP, is used to represent the
level of FRET ON state. Here, the “dynamic range” of the FRET bio-
sensor is the theoretical range of FRET/CFP in the ON state biosen-
sor and that in the OFF state (Figure 1C). In practical use, the change
of activity or concentration of the molecule is monitored by the
change of FRET/CFP after stimulation. This “gain” of the FRET signal
is the relative increase or decrease in FRET/CFP after stimulation and
is expressed as a percentage of the FRET/CFP value before stimula-
tion. Therefore the gain of a FRET biosensor expressed in a certain
cell type depends on both the dynamic range of the FRET biosensor
and the increase in the fraction of the ON state after stimulation ver-
sus that before stimulation (Figure 1C). Meanwhile, “sensitivity” of
FRET biosensors denotes a concentration of stimulants that increases
the FRET/CFP value to 50% of the dynamic range (Figure 1C).
Optimization of FRET donor and acceptor pairs
We first optimized the donor and acceptor FPs in the distance-
dependent intramolecular FRET biosensor. The prototype biosensor
is based on the structure of a PKA activity sensor, AKAR3 (Allen and
Zhang, 2006), which contained a consensus peptide of PKA phos-
phorylation (Zhang et al., 2001) and the phosphate binding domain
of FHA1 (Figure 2A). We initially used a 72-amino-acid (a.a.) polyg-
lycine linker used in an ERK activity sensor, EKAR, as a flexible long
linker (Harvey et al., 2008). We tested teal fluorescent protein (TFP)-
and CFP-derived FPs, including enhanced CFP (ECFP), Turquoise-
GL, and CyPet, as donor FPs, and we examined YFP-derived FPs,
including Venus, circularly permutated Venus mutants, mCitrine,
and YPet (see Materials and Methods and Supplemental Table S1),
as acceptor FPs. The gain of the biosensors was quantified in HeLa
FIGURE 1: Optimized backbone of an intramolecular FRET biosensor.
(A) Mode of action of the intramolecular FRET biosensor. (B) Structure
of the DNA encoding an optimized intramolecular FRET biosensor.
Shown are the unique restriction enzyme sites used to exchange each
domain for the development of the biosensor. (C) Schematic
representation of the titration curve of FRET/CFP ratio in
intramolecular FRET biosensors.
530 nm530 nm
Volume 22 December 1, 2011 An optimized backbone of FRET biosensors | 4649
cells stimulated with dibutyryl-cyclical AMP (dbcAMP), a membrane-
permeable cAMP analogue. Except for the biosensor containing
cp50Venus as the acceptor, the FRET/CFP ratio was increased upon
dbcAMP stimulation in all biosensors. Among them, the FRET bio-
sensors containing ECFP/YPet and Turquoise-GL/YPet exhibited the
largest gain in FRET/CFP (Figure 2B). A substantial amount of the
FRET biosensor with CyPet/YPet was cleaved at the linker region via
a currently unknown mechanism (Supplemental Figure S1). Notably,
YPet did not show any superiority to Venus when mTFP, an FP de-
rived from coral (Ai et al., 2006), was used as the donor (Figure 2, A
and C), suggesting that a pair of dimerization-prone FPs is suitable
for a distance-dependent FRET biosensor (see Discussion). Taken
together, these results led us to conclude that the ECFP/YPet or
Turquoise-GL/YPet pair was suitable for the donor and acceptor pair
of the distance-dependent intramolecular FRET biosensor.
Optimization of the length of flexible linkers
Previously we showed that the basal GTP/GDP ratio of the Ras FRET
biosensor was markedly larger than that of the endogenous Ras pro-
tein, implicating that the close proximity of the sensor and ligand
domains could increase the proportion of FRET biosensors in the
ON state (Mochizuki et al., 2001), probably because the ligand do-
main competitively inhibits access of GTPase-activating proteins
(GAPs) to the biosensor. This observation prompted us to lengthen
the linker to reduce the proportion of ON state FRET biosensors
(Figure 3A). In a preliminary experiment, linkers consisting of
(SAGG)n, where n indicates the number of repeats, served better
than the 72 a.a. polyglycine linkers (Levskaya et al., 2009; data not
shown). Thus we prepared (SAGG)n linkers (n = 13–61) and inserted
them into the prototype PKA biosensor. As expected, FRET/CFP in
the absence of stimulation correlated inversely with the length of the
linker (Figure 3B and Supplemental Figure S2). The gain of the FRET
biosensors upon dbcAMP stimulation correlated with the length of
the linker owing to the decrease in FRET/CFP in the absence of db-
cAMP (Figure 3C).
The observed decrease of FRET/CFP before stimulation could
be caused by two mechanisms. First, the long linker might reduce
FRET efficiency simply by increasing the distance between CFP and
YFP in the OFF state. Second, the long linker might reduce the frac-
tion of ON state FRET biosensors in the basal state. To clarify which
mechanism pertained, we separated the phosphorylated FRET bio-
sensor from the nonphosphorylated one in HeLa cell lysates with
SDS polyacrylamide gels containing a phosphorylated amino acid
chelator, Phos-tag (Kinoshita et al., 2006; Figure 3D). The 116 a.a.
linker markedly reduced the basal phosphorylation level in
FIGURE 2: Optimization of pairs of FPs for distance-dependent FRET
biosensors. (A) Scheme of the PKA or ERK activity sensor consisting
of YFPs (donor), a FHA1 or WW phosphopeptide binding domain
(ligand domain), a 72-Gly linker, a PKA or ERK substrate (sensor
domain), CFPs (acceptor), and nuclear export signal (NES). (B) HeLa
cells expressing AKAR3 with various pairs of FPs as indicated were
stimulated with 1 mM dbcAMP for 10 min. The gain in FRET/CFP is
represented with the SD (n > 5). (C) HeLa cells expressing EKAR with
various pairs of FPs as indicated were stimulated with 10 ng/ml EGF
for 10 min. The gain in FRET/CFP is represented with the SD (n > 5).
Venus / mTFP1
ERK-72-Gly linkerPKA-72-Gly linker
mTFP, ECFP, CyPet,
Venus, mCitrine, YPet,
PKA or ERK
Venus / mTFP1
Venus / ECFP
mCitrine / ECFP
YPet / ECFP
cp50Venus / ECFP
cp157Venus / ECFPcp172Venus / ECFPcp195Venus / ECFPcp229Venus / ECFP
YPet / Turquoise-GL
FIGURE 3: Effect of long linkers on the FRET gain. (A) Scheme of the
PKA activity sensor consisting of YPet and ECFP. (B) HeLa cells
expressing AKAR3 with a linker of various lengths were imaged by
FRET microscopy to obtain the basal FRET/CFP. Each dot corresponds
to the value from a single cell (n > 5). Horizontal bars are the mean
values. (C) HeLa cells expressing a PKA sensor as indicated were
stimulated with 1 mM dbcAMP for 10 min. The gain in FRET/CFP is
represented with the SD (n > 5). (D) HeLa cells expressing AKAR3 with
52, 84, or 116 a.a. length of linker were stimulated with 1 mM
dbcAMP or 1 mM dbcAMP and 50 nM Calyculin A for 10 min. Top,
cell lysates were subjected to Phos-tag immunoblotting analysis with
an anti-GFP antibody and a fluorescence-tagged secondary antibody.
Tositions of phosphorylated (p) and nonphosphorylated biosensors
(np) are indicated on the right of the representative gel image.
Bottom, average values of the fraction of phosphorylated biosensors
are shown with SD for three independent experiments. P value was
calculated by a one-tailed paired t test.
52 aa84 aa
180 aa 244 aa
52 aa84 aa
180 aa 244 aa
Phos-tag Western blotting
p = 0.04
52 aa 84 aa 116 aa
52 a.a. 84 a.a.116 a.a.
4650 | N. Komatsu et al. Molecular Biology of the Cell
comparison with the 52 a.a. linker. Because the reduction of the
phosphorylated biosensor in the presence of dbcAMP was much
lower, the gain of the biosensor containing the 116 a.a. linker was
larger than that of the biosensors containing shorter linkers (Figure
3D). The levels of phosphorylated biosensors in the presence of the
Ser/Thr phosphatase inhibitor, Calyculin A, were almost equal
among the three biosensors. These results indicated that the prefer-
able effect of the long linker was brought about by reducing the
fraction of ON state FRET biosensors in the basal state (Figure 3D).
Through these analyses, we established the optimized backbone
of the distance-dependent intramolecular FRET biosensors, consist-
ing of a long, flexible linker and an optimized pair of FPs. This sys-
tem was designated as the extension for enhanced visualization by
evading extraFRET (Eevee). We also named the long linkers “EV
linkers.” Because the increase in the gain reached a zenith at 116
a.a. (Figure 3C), we used this linker in the following study.
Mathematical validation of the Eevee backbone system
To understand the mode of action of the Eevee backbone, we
built and simulated a mathematical model of distance-dependent
intramolecular FRET biosensors (see Supplemental Information).
Computer simulations suggested that linker length and FP di-
merization exerted distinct effects of FRET increase: sensitivity and
dynamic range (Supplemental Figure S3, A–C). More importantly,
these numerical simulations predicted the presence of an optimal
length of linker to obtain the maximal gain of FRET, depending on
the strength of FP dimerization (Supplemental Figure S3D). In line
with this prediction, the dimerization-prone FP pair (i.e., YPet/
ECFP) demonstrated a higher gain for longer length of linker in
comparison to the Venus/ECFP pair (Supplemental Figure S3E).
Convincingly, a gain of AKAR3 with Venus/ECFP was saturated at
the linker length of 84 a.a., showing that the gain of the biosensor
reached its highest point (Supplemental Figure S3E). Thus this
simple simulation provided a plausible model to understand the
mechanism by which the combination of EV linker and the di-
merization-prone FPs increased the gain of distance-dependent
intramolecular FRET biosensor.
FRET biosensors of kinases containing the Eevee backbone
To demonstrate proof of concept of the Eevee backbone, we ap-
plied the backbone to the previously reported FRET biosensors of
protein kinases, an ERK sensor EKAR (Harvey et al., 2008) and a JNK
sensor JNKAR1 (Fosbrink et al., 2010). The new FRET biosensors of
PKA, ERK, and JNK were named AKAR3EV, EKAREV, and JNKAR1EV,
respectively. HeLa cells expressing FRET biosensors were time
lapse–imaged and stimulated (Figure 4). The gain of AKAR3EV was
approximately sixfold and threefold larger than that of the original
AKAR3 and AKAR4, respectively (Allen and Zhang, 2006; Depry
et al., 2011; Figure 4A and Supplemental Video S1). Similarly, the
gains of EKAREV and JNKAR1EV were larger than those of EKAR
and JNKAR1 by four- and threefold, respectively (Figure 4, B and
C, and Supplemental Videos S2 and S3). We also found that, in
EKAREV, either the FHA1 or WW domain was equally used as the
phosphopeptide binding (ligand) domain (data not shown).
The Eevee backbone was also applied to Picchu, an EGFR and
Abl tyrosine kinase activity sensor (Kurokawa et al., 2001). In Picchu
with an EV linker, the SH2–SH3 region and the substrate peptide
(a.a. residues 217–225) of human CrkII were used as the ligand do-
main and sensor domain, respectively. The resulting PicchuEV ex-
hibited a twofold increase in the gain in EGF-stimulated HeLa cells
in comparison to the original Picchu (Figure 4D and Supplemental
Video S4). Notably, the time course of FRET/CFP was similar be-
tween Picchu and PicchuEV (Figure 4D), suggesting that PicchuEV
was also capable of monitoring the rapid kinetics of regulation in the
protein kinases and phosphatases like the prototype Picchu.
FRET biosensors of small GTPases based
on the Eevee backbone
We further applied the Eevee backbone to the intramolecular FRET
biosensors of small GTPases. As examples, we adduced the FRET
biosensors for Ras and Rac1, Raichu-Ras and Raichu-Rac1, respec-
tively (Mochizuki et al., 2001; Itoh et al., 2002). The new FRET bio-
sensors, RaichuEV-Ras and RaichuEV-Rac1, exhibited two- to three-
fold larger gains than the original FRET biosensors (Figure 5 and
Supplemental Videos S5 and S6). Similarly to AKAR3 with the long
linker (Figure 3C), the basal FRET/CFP ratio of Raichu-Ras with the
EV linker was considerably reduced by the EV linker (Figure 5C). As
expected, the basal FRET/CFP of RaichuEV-Rac1 was also markedly
lower than that of the original Raichu-Rac1 (Figure 5F). We further
confirmed that the EV linker reduced the ratio of GTP versus GDP
bound to biosensors by thin-layer chromatography with the 32Pi-la-
beled, unstimulated HeLa cells (Supplemental Figure S4). Thus, in
FIGURE 4: Improvement of FRET biosensors by the Eevee backbone.
At the top of each panel, the structure of novel biosensors based on
the Eevee backbone is shown. In the substrate peptide sequences,
red letters indicate the phosphorylation site. Blue letters indicate
amino-acid substitutions to increase the affinity to either the FHA1 or
WW domain. Green letters indicate the docking site of the kinases.
(A) HeLa cells expressing AKAR3EV, AKAR3, or AKAR4 were
stimulated with 1 mM dbcAMP and time lapse–imaged by FRET
microscopy (Supplemental Video S1). The FRET/CFP ratio of each cell
was normalized by dividing by the averaged FRET/CFP value before
stimulation. The mean and SD from at least 10 cells are plotted
against time. (B) HeLa cells expressing EKAREV or EKAR were
stimulated with 10 ng/ml EGF (Supplemental Video S2). The average
of normalized FRET/CFP ratio is shown with SD (n > 10). (C) HeLa cells
expressing JNKAR1EV or JNKAR1 were stimulated with 1 μg/ml
Anisomycin (Supplemental Video S3). The average of normalized
FRET/CFP ratio is shown with SD (n > 10). (D) HeLa cells expressing
PicchuEV, an EGFR/Abl kinase sensor, or Picchu were stimulated with
25 ng/ml EGF (Supplemental Video S4). The average of normalized
FRET/CFP ratio is shown with SD (n > 10).
Elapsed time (min)
-100 1020 30
Elapsed time (min)
0 10 20 30 40 50
Elapsed time (min)
Elapsed time (min)
Volume 22 December 1, 2011 An optimized backbone of FRET biosensors | 4651
analogy to AKAR3EV, the EV linker seemed to facilitate the access of
GAP, leading to a reduction in the GTP-bound ON state of Raichu
under the basal condition.
Rapid development of new FRET biosensors
by the Eevee backbone
We tried to develop novel Ser/Thr kinase activity sensors for RSK,
S6K, Akt, and classical PKC by using the Eevee backbone (Figure 6).
Because the orientation-dependent FRET could be neglected in the
Eevee backbone, we simply needed to determine the optimal
substrate peptide specific to each kinase and the appropriate phos-
phate-binding domain for that peptide sequence. For example,
we chose substrate peptides encompassing Ser-1798 of TSC2 (Roux
et al., 2004) and Thr-1135 of Rictor (Dibble et al., 2009) for the de-
velopment of Eevee-RSK and -S6K, respectively. HeLa cells express-
ing these FRET biosensors were stimulated with EGF and, after 10
or 30 min, inhibited with either BI-D1780 or rapamycin (Figure 6,
A–F, and Supplemental Videos S7 and S8). We observed the appar-
ent increase in FRET/CFP upon EGF stimulation with Eevee-RSK
and -S6K. The effect of the inhibitors was also clearly demonstrated.
This result indicated that the Eevee backbone provided an easy and
rapid method for the development of FRET biosensors.
Such simple application of the Eevee backbone might not always
guarantee a large gain, however, as we show in the case of FRET
biosensors of Akt and PKCβ developed by the same approach. The
substrate peptides were from previously reported consensus se-
quences of Akt (Obata et al., 2000) and MARCKS, respectively
(Violin et al., 2003; Kunkel et al., 2005). The gains of Akt and PKC
FRET sensors, Eevee-Akt-cyt and Eevee-PKC-cyt, were 8 and 4% in
EGF-stimulated Cos7 cells and tetradecanoylphorbol 13-acetate
(TPA)-stimulated HeLa cells, respectively (Supplemental Figure S5).
These gains were markedly lower than those of the other Eevee
biosensors. To increase the fraction of the
phosphorylated FRET biosensor, we added
the Pleckstrin homology (PH) domain of Akt
to the N terminus of the FRET biosensors of
Akt (Figure 6G). The PH domain binds to
and serves as a plasma membrane–target-
ing signal. Hence, addition of the PH do-
main to the FRET biosensors of Akt could
increase the effective concentration of the
FRET biosensor around the active Akt at the
plasma membrane. As expected, cells ex-
pressing this FRET biosensor, which we
named Eevee-Akt, exhibited a rapid and ro-
bust increase in FRET/CFP (ca. 25%) at the
plasma membrane (Figure 6, H and I, and
Supplemental Video S9). As a control, we
prepared Eevee-Akt-TA, in which Thr in the
substrate peptide was replaced with Ala.
Unexpectedly, we found that Eevee-Akt-TA
also exhibited an 18% increase in FRET/CFP
ratio upon stimulation (Figure 6H). This in-
crease in FRET/CFP was probably caused by
the bystander FRET (i.e., local increase in
the concentration of the FRET biosensors at
the plasma membrane could result in the
stochastic intermolecular FRET; Supplemen-
tal Figure S6; Chiu et al., 2002).
In a similar manner, we constructed
Eevee-PKCβ, in which the C1 domain of
PKCβ was fused at the N terminus of the FRET biosensor (Figure 6J).
The C1 domain binds to diacylglycerol and serves as another plasma
membrane–targeting signal. Again, cells expressing Eevee-PKC ex-
hibited a rapid and marked increase in FRET/CFP (Figure 6L and
Supplemental Video S10). Similarly to Eevee-Akt-TA, Eevee-PKC-TA
also showed membrane translocation (Supplemental Figure S6) and
a slight, but not ignorable, increase in FRET/CFP ratio (Figure 6K).
Taken together, Eevee-PKC and Eevee-Akt monitor not only kinase
activity of PKC and Akt, but also stimulation-dependent membrane
translocation of the kinases. Furthermore, we constructed the Akt
and PKC biosensors that were localized to plasma membrane by
lipid modification of the K-Ras4B C-terminus region (Eevee-Akt-pm
and Eevee-PKC-pm; Supplemental Figure 5). TA mutants of these
plasma membrane–targeted FRET biosensors did not respond to
the stimulation. Notably, FRET gains of Eevee-Akt-pm and Eevee-
PKC-pm upon stimulation corresponded to the difference of FRET
gains between Eevee-Akt and Eevee-Akt-TA and between Eevee-
PKC and Eevee-PKC-TA (Supplemental Figure S7). Thus Eevee-
PKC-pm and Eevee-Akt-pm are applicable to visualize only the ki-
nase activity of PKC and Akt at the plasma membrane, respectively.
Quantitative evaluation of kinase inhibitors by Eevee
biosensor-expressing cell lines
With FRET biosensors having large gains in hand, we developed a
quantitative and rapid assay to evaluate the effects of kinase inhibi-
tor on Ser/Thr kinases in living cells. First, we established HeLa cell
lines stably expressing Eevee biosensors. We previously failed to
establish such stable cell lines with retrovirus/lentivirus-mediated
gene transfer or by the transfection of linearized plasmid DNAs be-
cause of frequent recombination between YFP and CFP (data not
shown). We recently found that this problem could be readily over-
come by the use of a piggyBac transposase system (Yusa et al.,
FIGURE 5: FRET biosensors of small GTPases based on the Eevee backbone. Structures of
FRET biosensors based on the Eevee backbone, RaichuEV, are shown at the top of panels (A)
and (D). RafRBD and PAK CRIB denote the Ras-binding domain of Raf1 and the Cdc42/
Rac-interactive binding domain, respectively. Cos7 cells expressing RaichuEV or the prototype
Raichu were stimulated with 50 ng/ml EGF and time lapse–imaged (Supplemental Videos S5 and
S6). Representative FRET/CFP ratio images are shown in the intensity-modulated display mode.
Scale bars are 10 μm. (B and E) The FRET/CFP ratio of each cell was normalized by dividing by
the averaged FRET/CFP value before stimulation. The mean and SD from at least 10 cells are
plotted against time. (C and F) Basal FRET/CFP ratios of RaichuEV and Raichu are plotted. Each
dot corresponds to the value from a single cell, and at least six cells were analyzed. The
horizontal bar indicates the mean.
Elapsed time (min)
-10 0 10 20 30 40 50 60
Elapsed time (min)
4652 | N. Komatsu et al. Molecular Biology of the Cell
2009). After single-cell cloning, the cells were seeded onto a 96-well
glass base plate and treated with stimulants and kinase inhibitors
(Figure 7A). For example, cells expressing EKAREV-nuc, which local-
ized to nucleus, were stimulated with EGF in the presence of de-
creasing concentrations of various inhibitors and FRET-imaged by
an automated epifluorescence microscope (Figure 7B and Supple-
mental Figure S8). By computer-assisted processing of the FRET/
CFP ratio of each cell, we could obtain the IC50 values of kinase in-
hibitors (Figure 7C). Intriguingly, this single cell–based assay also
revealed unexpected differences in terms of the mode of action of
the inhibitors. EGF receptor inhibitors such as AG1478 and
PD153035 inhibited ERK activity as a bistable (all-or-nothing) re-
sponse, whereas the MEK inhibitor PD184352 inhibited ERK with a
graded response (Figure 7, D and E, and Supplemental Figure S9).
Taken together, these results demonstrated that the Eevee system
combined with piggyBac transposase enabled rapid and quantita-
tive evaluation of the effect of drugs in living cells.
FIGURE 6: Novel Ser/Thr kinase FRET biosensors based on the Eevee
backbone. (A, D, G, and J) Structures of Eevee-RSK, Eevee-S6K,
Eevee-Akt, and Eevee-PKC, which are FRET biosensors of RSK, S6K,
Akt, and classical PKC activities, respectively. Red and blue letters in
the substrate peptide sequences denote the phosphorylation sites
and amino-acid substitutions to increase the affinity to FHA1. (B, E, H,
and K) HeLa cells expressing Eevee-RSK (B), Eevee-S6K (E), Eevee-
PKC (K), or Cos7 cells expressing Eevee-Akt (H) were time lapse–
imaged and stimulated with 10 ng/ml EGF (B and E), 50 ng/ml EGF
(H), or 1 μM TPA (K). Cells expressing Eevee-RSK and Eevee-S6K were
further treated with 10 nM BI-D1870 (B) and 100 nM rapamycin (E),
respectively, at 30 min after EGF stimulation. The FRET/CFP ratio of
each cell was normalized by dividing by the averaged FRET/CFP value
before stimulation (Supplemental Videos S7–S10). The mean and SD
from at least 10 cells are plotted against time. (C, F, I, and L)
Representative FRET/CFP ratio images with Eevee-RSK (C), Eevee-
S6K (F), Eevee-Akt (I), and Eevee-PKC (L) are shown in the intensity-
modulated display mode. Scale bars are 10 μm.
Elapsed time (min)
Elapsed time (min)
-10 0 10 20 30 40 50
Elapsed time (min)
Elapsed time (min)
FHA1 AktPH NES
FIGURE 7: Quantitative evaluation of kinase inhibitors with Eevee-
expressing cell lines. (A) Schematic view of the experimental design.
Cells expressing EKAREV-nuc were seeded, starved, and treated with
stimulant in the presence of decreasing concentrations of the
indicated kinase inhibitors. After 30 min, FRET images were acquired
and processed for quantification. (B) Shown here are the
representative FRET/CFP ratio images of HeLa cells stably expressing
EKAREV-nuc and treated with 25 ng/ml EGF and the indicated
concentrations of an EGFR inhibitor, PD153035. Scale bar is 50 μm.
(C) Averaged FRET/CFP ratios are plotted against the concentrations
of kinase inhibitors and fitted with curves by the four-parameter
logistic model. (D and E) FRET/CFP ratios in each cell at the indicated
concentration of PD153035 (D) or a MEK inhibitor, PD184352 (E),
were quantified and represented in histograms (n > 80 cells).
+EGF +EGF+EGF+EGF +EGF
+ + + + + + ++++ +-
JNK inhibitor VIII
JNK inhibitor VIII
ERK activity (FRET/CFP)
ERK activity (FRET/CFP)
Volume 22 December 1, 2011 An optimized backbone of FRET biosensors | 4653
We have developed an optimized backbone, Eevee, which allows
us to quickly develop FRET biosensors. The flexible long linker and
the optimized FP pairs cooperatively served to increase the gain of
the FRET biosensors. The Eevee backbone was used to improve
FRET biosensors of PKA, ERK, JNK, EGFR/Abl, Ras, and Rac1
(Figures 4 and 5) and to develop FRET biosensors of RSK, S6K, Akt,
and PKC (Figure 6).
The key technology of the Eevee backbone is the flexible
long linker EV, which renders FRET biosensors mostly distance-
dependent. It has been reported that circularly permutated (cp)
FPs improved FRET biosensors of calcium and PKA (Nagai et al.,
2004; Allen and Zhang, 2006), which would seem to provide evi-
dence that orientation has a critical impact on the FRET efficiency
of the intramolecular FRET biosensors. Consistent with this idea,
the FRET efficiency of orientation-dependent biosensors has
been reported to be drastically influenced by the addition or de-
letion of one to several amino acids at the C terminus of FPs
(Miyawaki et al., 1997; Horikawa et al., 2010). The developer,
however, could not predict whether the FRET of a newly con-
structed, orientation-dependent biosensor would increase or de-
crease upon the perception of the signal (Violin et al., 2003;
Kunkel et al., 2005). For this reason, we decided to provide an
optimized backbone for the distance-dependent type of FRET
biosensor. The evidence that FRET of the Eevee backbone is
mostly distance-dependent is as follows: 1) All of the FRET bio-
sensors based on the Eevee backbone were associated with an
increase in FRET/CFP in the ON state; 2) the basal fraction of
OFF-state biosensors was inversely correlated with the length of
the linker (Figure 3C); and 3) a series of cpVenus variants did not
improve the gain (Figure 2B). Of note, there existed optimal linker
lengths, depending on the pair of FPs, which were qualitatively
explained by the mathematical model (Supplemental Figure S3).
To quantitatively predict an optimal length of linker, we need to
measure and/or estimate quantitative parameters and implement
a model with those parameters.
An intramolecular FRET biosensor often suffers from a high
basal FRET level, which could be caused by two mechanisms.
First, the distance between the donor and acceptor FPs may not
be sufficient to eliminate the basal FRET. Second, and more likely,
the ligand domain may increase the fraction of ON-state biosen-
sors by masking the sensor domain from negative regulators
such as phosphatases and GAPs. Consistent with this idea, the
basal GTP/GDP ratio on the Raichu-RhoA biosensor varies sig-
nificantly according to the affinity of RhoA binding domains for
RhoA (Yoshizaki et al., 2003). We presume that a flexible long
linker minimizes this masking effect by reducing the effective
concentration of the ligand domain around the sensor domain.
There have been several reports of a linker peptide improving
the FRET biosensor. For example, a “flip-flop” linker composed
of a rigid α-helical linker with a flexible diglycine motif (Sato
et al., 2003) and an “elastic” linker derived from spider silk pro-
tein flagelliform (Grashoff et al., 2010) have been shown to in-
crease the gain of the FRET biosensor. It is unknown, however,
whether these linkers also serve to reduce the masking effect of
the ligand domain.
We showed that two FP pairs, ECFP/YPet and Turquoise-GL/
YPet, were used preferably to the distance-dependent intramolecu-
lar FRET biosensors (Ouyang et al., 2008). The wild-type Aequorea
FP forms a dimer with congeneric Aequorea FP at high concentra-
tion, of which the dissociation constant (Kd) is 110 μM (Zacharias
et al., 2002). Recently Kotera et al. (2010) have verified that an
Aequorea FP pair possessing reversible dimerization property en-
hances FRET of intramolecular FRET biosensors. Originally, YPet
was reported to enhance FRET without inducing heterodimerization
between YPet and CyPet in an intermolecular FRET biosensor
(Nguyen and Daugherty, 2005). Recent reports suggested, however,
that the increase in FRET gain by using YPet as an acceptor seems
to be attributable to an enhanced dimerization with the congeneric
FPs (Ohashi et al., 2007; Kotera et al., 2010). In agreement with this
report, we have found that YPet did not improve FRET when a TFP,
which was derived from Clavularia FP (Ai et al., 2006), was used as a
donor (Figure 2C). It should be noted that, beyond a certain thresh-
old, the increasing donor–acceptor FP binding affinity will yield a
substantial fraction of FRET biosensors locked to the closed form
even in the absence of stimulation.
To develop a novel FRET biosensor of Ser/Thr kinase with the
Eevee backbone, we need to find a substrate peptide that is phos-
phorylated by the said Ser/Thr kinase with high efficiency and spec-
ificity. We will adduce S6K as an example. Because the two major
substrates of S6K are S6 and Rictor (Ferrari et al., 1991; Dibble et al.,
2009), naturally we chose a substrate peptide encompassing Ser-
240 of S6 and Thr-1135 of Rictor. Interestingly, the FRET biosensor
containing the substrate peptide derived from Rictor exhibited a
significant response to EGF and rapamycin (Figure 6), whereas the
other FRET biosensor containing a substrate peptide derived from
S6 did not respond to EGF due to inefficient phosphorylation (data
not shown). Notably, newly developed biosensors, Eevee-RSK, Ee-
vee-S6K, Eevee-Akt, and Eevee-PKC, did not achieve gains as large
as did AKAR3EV, EKAREV, and JNKAR1EV. This result is possibly
due to the lower phosphorylation efficiency. Hence, screening of an
optimal substrate peptide that is efficiently phosphorylated by the
kinase of interest remains an important step in the development of
FRET biosensors of a kinase. This problem of low phosphorylation
efficiency may be overcome by two options. First, by the addition of
a subcellular localization domain, the Eevee biosensor could be
concentrated at the specific intracellular compartment where the
said kinase is activated. Successful examples are Eevee-Akt and
Eevee-PKC (Figure 6). Second, we may use a motif that facilitates
the association of the substrate with the target kinase as exemplified
in EKAREV and JNKAR1EV (Figure 4; Harvey et al., 2008; Fosbrink
et al., 2010).
In summary, we described an optimized backbone for intramo-
lecular FRET biosensors. The Eevee biosensor still needs a number
of steps and careful characterization to create a sensitive and spe-
cific FRET biosensor, such as a choice of efficiently phosphorylated
peptide and intracellular targeting. This simple and versatile system
should prove useful for developing FRET biosensors with high pro-
ductivity and for accelerating our understanding of spatiotemporal
MATERIALS AND METHODS
FRET biosensor construction
The cDNAs of YFPs and CFPs were amplified by PCR. Fluorescence
proteins used as acceptor or donor were Venus (Nagai et al., 2002),
cpVenus (cp50Venus, cp157Venus, cp173Venus, cp195Venus,
cp229Venus; Nagai et al., 2004), mCitrine (Griesbeck et al., 2001),
YPet, CyPet (Nguyen and Daugherty, 2005), mTFP1 (Ai et al., 2006),
mTurquoise, mTurquoise-GL (Goedhart et al., 2010), ECFP, and
SECFP (a brighter version of ECFP developed by A. Miyawaki
[RIKEN, Saitama, Japan]). The cDNA of SECFP contains additional
mutations of K26R, D130G, N165H, and S176G based on the cDNA
of ECFP. mTurquoise and mTurquoise-GL were the gifts of T. W. J.
Gadella, Jr. (Swammerdam Institute for Life Sciences, Amsterdam,
4654 | N. Komatsu et al. Molecular Biology of the Cell
The Netherlands; Goedhart et al., 2010). The mutation of Ala at the
206 position in mTurquoise and mTurquoise-GL was restored to Lys
with a two-step overlap PCR, generating Turquoise and Turquoise-
GL (Goedhart et al., 2010). Amino acid substitutions in YFP and CFP
variants used in this study are summarized in Supplemental Table
S1. These cDNAs of fluorophores were inserted into the eukaryotic
expression vectors pCAGGS (Niwa et al., 1991) and/or pPBbsr,
which harbors blasticidin S–resistant gene (Yusa et al., 2009). The
unique restriction enzyme sites are shown in Figure 1B. EKAR was
obtained from Addgene (Cambridge, MA; http://www.addgene.
org/). AKAR3, AKAR4, and JNKAR1 were gifts from J. Zhang (Johns
Hopkins University, Baltimore, MD).
The flexible linkers collectively designated EV consisted of vari-
ous numbers of the 20 a.a. peptide, SAGGSAGGSAGGSAGGSAGG
(Levskaya et al., 2009). The cDNAs of the EV were introduced into
an expression plasmid by a combination of standard subcloning and
annealed DNA duplex ligation. The DNA of the 20 a.a. EV linker was
generated by annealing the sense-oligonucleotide (5’-GTACCAGT-
GGTAGTGCTGGTGGTT-3’) and antisense-oligonucleotide (5’-CCG-
CCAGCACTACCACCAGCACTG-3’) and inserted into an expression
vector of the PKA FRET biosensor, generating pAKAR3EV-20.
pEevee-20 was cleaved with EcoRI/Aor13HI or EcoRI/Asp718I.
Fragments containing the EV linker coding sequence were ligated
with the following DNA, which was obtained by annealing (5’-CCG-
GCAGTGCTGGTGGTAGTGCTGGTGGTA-3’) and (5’-GTACTAC-
CACCAGCACTACCACCAGCACTG-3’). The resulting plasmid was
named pAKAR3EV-52, which contained a 52 a.a. linker. By repeat-
ing this step, we obtained expression plasmids containing EV linkers
up to 244 a.a. in length.
The Ser/Thr kinase FRET biosensors with long linkers were
collectively designated Eevee, and the plasmids thereof, pEevee,
were constructed essentially in the manner of pRaichu-Ras
(Mochizuki et al., 2001). From the N terminus, Eevee consists of
YPet (a.a. 1–238), a spacer (Leu-Glu), the FHA1 domain of yeast
Rad53 (a.a. 241–382) used as the ligand domain, a spacer (Gly-
Thr), an EV linker, a spacer (Ser-Gly), a Ser/Thr kinase substrate
peptide used as the sensor domain, a spacer (Gly-Gly-Arg), ECFP
(a.a. 513–750), a spacer (Ser-Arg), and the nuclear export sequence
of the HIV–1 rev protein (LQLPPLERLTLD) or the nuclear localiza-
tion signal of the SV40 large T antigen (GGPPKKKPKVEDP). Ee-
vee–Akt and Eevee-PKC included the PH domain of human Akt1
(a.a. 5–152) and the C1 domain of human PKCβ (a.a. 5–121) at the
N terminus of YPet, respectively. In EKAREV, the FHA1 domain was
replaced with the WW domain of human Pin1 (a.a. 241–295). cD-
NAs of the substrate peptides were synthesized so that the amino
acid residue at the phosphoacceptor was Thr and the amino acid
at the +3 position from said Thr was Asp, rendering the substrate
sequence optimal for the binding to the FHA1 domain. Similarly,
the amino acid residue at the +1 position from the phosphoaccep-
tor amino acid was substituted for Pro when the WW domain was
used as the ligand domain. The synthesized oligonucleotides were
annealed and inserted into the expression vector. The substrate
peptide sequences used in this study are listed in Supplemental
A plasmid encoding PicchuEV was generated as follows. The
DNA fragment of human CrkII encompassing the SH2 domain and
the SH3 domain (a.a. 1–204) was used as the ligand domain. The
substrate peptide of human CrkII (a.a. 217–225) was used as the
sensor domain. pRaichu-Ras and Raichu-Rac1 containing an EV
linker were constructed by replacing the linker region of the original
biosensors with an EV linker (Mochizuki et al., 2001; Itoh et al.,
Cells, reagents, and antibodies
HeLa cells were purchased from the Human Science Research Re-
sources Bank (Sennanshi, Japan). The Cos7 cells used were Cos7/
E3, a subclone of Cos7 cells established by Y. Fukui (National Re-
search Institute of Health, Taiwan, Republic of China). HeLa cells and
Cos7 cells were maintained in DMEM (Sigma-Aldrich, St. Louis, MO)
supplemented with 10% FBS. The cells were plated on 35-mm glass
base dishes or 96-well glass base plates (Asahi Techno Glass, Tokyo,
Japan), which were coated with collagen type I (Nitta Gelatin, Osaka,
Japan). Plasmids encoding FRET biosensors were transfected into
HeLa cells and Cos7 cells by 293fectin or Lipofectamine 2000, ac-
cording to the manufacturer’s instructions (Invitrogen, San Diego,
CA), respectively. EGF was purchased from Sigma-Aldrich. dbcAMP,
TPA, Calyculin A, Anisomycin, PD153035, and JNK inhibitor VIII
were purchased from Calbiochem (La Jolla, CA). PD184352 was
obtained from Toronto Research Chemicals (Ontario, Canada). BI-
D1870 was purchased from Symansis (Shanghai, China). Rapamycin
was obtained from LC Laboratories (Woburn, MA). PLX-4720 was
purchased from Selleck Chemicals (Houston, TX). The expression
vector of piggyBac transposase was provided by A. Bradley (Well-
come Trust Sanger Institute, Cambridge, UK; Yusa et al., 2009).
Phos-tag was obtained from the Phos-tag Consortium (Hiroshima,
Japan; www.phos-tag.com). Anti-green fluorescence protein (GFP)
sera were prepared in our laboratory. LI-COR (Lincoln, NE) blocking
buffer and the IRDye680- and IRDye800-conjugated anti–rabbit and
anti–mouse immunoglobulin G secondary antibodies were obtained
Phospho-affinity gel electrophoresis was performed essentially as
described previously (Kinoshita et al., 2006). Conventional SDS-
polyacrylamide separation gels were supplemented with 50 μM
Phos-tag and 100 μM MnCl2, according to the manufacturer’s proto-
col. Proteins were detected and quantified by using an Odyssey In-
frared Imaging System (LI-COR).
Time-lapse FRET imaging
FRET images were obtained and processed using essentially the
same conditions and procedures as previously reported (Aoki and
Matsuda, 2009). Briefly, HeLa cells or Cos7 cells expressing FRET
biosensors were starved for 6–12 h with phenol red–free DMEM/
F12 medium or Medium 199 (Invitrogen) containing 0.1% bovine
serum albumin (BSA) or phenol red–free M199 (Invitrogen) with
20 mM HEPES and 0.1% BSA. Starved cells were treated with stimu-
lus, followed by the addition of inhibitors if necessary. Cells were
imaged with an inverted microscope (IX71 or IX81; Olympus, Tokyo,
Japan) equipped with a 60× objective lens (Olympus), a cooled
CCD camera (CoolSNAP HQ or CoolSNAP K4; Roper Scientific,
Tucson, AZ), an LED illumination system (CoolLED precisExcite;
Molecular Devices, Sunnyvale, CA), an IX2-ZDC laser-based autofo-
cusing system (Olympus), and an MD-XY30100T-Meta automatically
programmable XY stage (SIGMA KOKI, Tokyo, Japan). The follow-
ing filters used for the dual-emission imaging studies were obtained
from Omega Optical (Brattleboro, VT): an XF1071 (440AF21) excita-
tion filter, an XF2034 (455DRLP) dichroic mirror, and two emission
filters (XF3075 480AF30 for CFP and XF3079 535AF26 for YFP). Af-
ter background subtraction, FRET/CFP ratio images were created
with MetaMorph software (Universal Imaging, West Chester, PA),
and represented by the intensity-modulated display mode. In the
Volume 22 December 1, 2011 An optimized backbone of FRET biosensors | 4655
to the members of the Matsuda Laboratory for their helpful input.
K.A. was supported by a Grant-in-Aid for Scientific Research on Pri-
ority Areas and by the JST PRESTO program. M.M. was supported
by the Research Program of Innovative Cell Biology by Innovative
Technology (Cell Innovation) from the Ministry of Education, Culture,
Sports, and Science (MEXT), Japan.
Ai HW, Henderson JN, Remington SJ, Campbell RE (2006). Directed
evolution of a monomeric, bright and photostable version of Clavularia
cyan fluorescent protein: structural characterization and applications in
fluorescence imaging. Biochem J 400, 531–540.
Allen MD, Zhang J (2006). Subcellular dynamics of protein kinase A activity
visualized by FRET-based reporters. Biochem Biophys Res Commun 348,
Aoki K, Kiyokawa E, Nakamura T, Matsuda M (2008). Visualization of growth
signal transduction cascades in living cells with genetically encoded
probes based on Forster resonance energy transfer. Philos Trans R Soc
Lond B Biol Sci 363, 2143–2151.
Aoki K, Matsuda M (2009). Visualization of small GTPase activity with
fluorescence resonance energy transfer-based biosensors. Nat Protoc 4,
Chiu VK, Bivona T, Hach A, Sajous JB, Silletti J, Wiener H, Johnson RL, 2nd,
Cox AD, Philips MR (2002). Ras signalling on the endoplasmic reticulum
and the Golgi. Nat Cell Biol 4, 343–350.
Depry C, Allen MD, Zhang J (2011). Visualization of PKA activity in plasma
membrane microdomains. Mol Biosyst 7, 52–58.
Dibble CC, Asara JM, Manning BD (2009). Characterization of Rictor phos-
phorylation sites reveals direct regulation of mTOR complex 2 by S6K1.
Mol Cell Biol 29, 5657–5670.
Ferrari S, Bandi HR, Hofsteenge J, Bussian BM, Thomas G (1991). Mitogen-
activated 70K S6 kinaseIdentification of in vitro 40 S ribosomal S6
phosphorylation sites. J Biol Chem 266, 22770–22775.
Fosbrink M, Aye-Han NN, Cheong R, Levchenko A, Zhang J (2010).
Visualization of JNK activity dynamics with a genetically encoded fluo-
rescent biosensor. Proc Natl Acad Sci USA 107, 5459–5464.
Goedhart J, van Weeren L, Hink MA, Vischer NO, Jalink K, Gadella TW, Jr
(2010). Bright cyan fluorescent protein variants identified by fluores-
cence lifetime screening. Nat Methods 7, 137–139.
Gotoh T, Niino Y, Tokuda M, Hatase O, Nakamura S, Matsuda M, Hattori S
(1997). Activation of R-Ras by Ras-guanine nucleotide-releasing factor. J
Biol Chem 272, 18602–18607.
Grashoff C et al. (2010). Measuring mechanical tension across vinculin
reveals regulation of focal adhesion dynamics. Nature 466, 263–266.
Griesbeck O, Baird GS, Campbell RE, Zacharias DA, Tsien RY (2001). Reduc-
ing the environmental sensitivity of yellow fluorescent protein. Mecha-
nism and applications. J Biol Chem 276, 29188–29194.
Harvey CD, Ehrhardt AG, Cellurale C, Zhong H, Yasuda R, Davis RJ,
Svoboda K (2008). A genetically encoded fluorescent sensor of ERK
activity. Proc Natl Acad Sci USA 105, 19264–19269.
Horikawa K, Yamada Y, Matsuda T, Kobayashi K, Hashimoto M, Matsu-Ura
T, Miyawaki A, Michikawa T, Mikoshiba K, Nagai T (2010). Spontaneous
network activity visualized by ultrasensitive Ca(2+) indicators, yellow
Cameleon-Nano. Nat Methods 7, 729–732.
Itoh RE, Kurokawa K, Ohba Y, Yoshizaki H, Mochizuki N, Matsuda M (2002).
Activation of rac and cdc42 video imaged by fluorescent resonance
energy transfer-based single-molecule probes in the membrane of living
cells. Mol Cell Biol 22, 6582–6591.
Jares-Erijman EA, Jovin TM (2003). FRET imaging. Nat Biotechnol 21,
Karasawa S, Araki T, Nagai T, Mizuno H, Miyawaki A (2004). Cyan-emitting
and orange-emitting fluorescent proteins as a donor/acceptor pair for
fluorescence resonance energy transfer. Biochem J 381, 307–312.
Kinoshita E, Kinoshita-Kikuta E, Takiyama K, Koike T (2006). Phosphate-
binding tag, a new tool to visualize phosphorylated proteins. Mol Cell
Proteomics 5, 749–757.
Kotera I, Iwasaki T, Imamura H, Noji H, Nagai T (2010). Reversible dimeriza-
tion of Aequorea victoria fluorescent proteins increases the dynamic
range of FRET-based indicators. ACS Chem Biol 5, 215–222.
Kunkel MT, Ni Q, Tsien RY, Zhang J, Newton AC (2005). Spatio-temporal
dynamics of protein kinase B/Akt signaling revealed by a genetically
encoded fluorescent reporter. J Biol Chem 280, 5581–5587.
Kurokawa K, Mochizuki N, Ohba Y, Mizuno H, Miyawaki A, Matsuda M
(2001). A pair of fluorescent resonance energy transfer-based probes
intensity-modulated display mode, eight colors from red to blue are
used to represent the FRET/CFP ratio, with the intensity of each
color indicating the mean intensity of FRET and CFP. For the quan-
tification, the FRET and CFP intensities were averaged over the
whole cell area, and the results were exported to Excel software
(Microsoft Corporation, Redmond, WA). In some experiments, the
FRET/CFP value from before 10 min to the time of stimulation was
averaged and used as the reference. The ratio of raw FRET/CFP
value versus the reference value was defined as the normalized
Multiwell FRET imaging
A HeLa cell line stably expressing EKAREV-nls, which contained the
nuclear localization signal, was established according to Yusa et al.
(2009). The cells were seeded on a 96-well glass base plate at a cell
density of 1.5 × 104 cells/well. One day after seeding, the cells were
serum starved for 6 h, followed by treatment with stimulant and ki-
nase inhibitors for 15 min. Then, the 96-well plate was imaged by an
inverted microscope as described earlier in text, except that an 20×
objective lens was used. FRET and CFP images were obtained in
one position for every well of the 96-well plate.
Spectroscopy by confocal microscopy
Twenty-four hours after transfection, HeLa cells expressing FRET
probes were starved for 3–6 h. Fluorescence spectra were acquired
by using a FV-1000 confocal imaging system (Olympus) in the lambda
scanning mode upon excitation of CFP at a wavelength of 405 nm.
Quantification of guanine nucleotide bound to GTPases
Guanine nucleotides bound to Raichu biosensors were analyzed es-
sentially as described previously (Gotoh et al., 1997). Briefly, HeLa
cells were transfected with expression vectors for Raichu-Ras, Rai-
chuEV-Ras, Raichu-Rac1, and RaichuEV-Rac1. After 36 h, the cells
were metabolically labeled with 32Pi orthophosphate for 2 h and
then lysed. The cell lysates were clarified by centrifugation, and Rai-
chu biosensors were immunoprecipitated by using anti-GFP antise-
rum. Guanine nucleotides bound to Raichu biosensors were sepa-
rated by thin-layer chromatography and quantitated with a BAS-1000
image analyzer (Fujifilm, Tokyo, Japan).
Mathematical and statistical analysis
Simulation was implemented by Mathematica software (Wolfram
Research, Champaign, IL). Details are described in the Supplemen-
Phosphorylation levels (Figure 3) or GTP/(GTP+GDP) levels (Sup-
plemental Figure S4) at the basal state varied from day to day; the
reason for this variation is not clear. Thus we could not combine all
data obtained in 3–4 d and we had to handle the data obtained in
one experiment as one data set. In this case, a paired t test needed
to be applied to examine a statistical significance. The paired t test
demonstrated whether a long linker decreased the basal phospho-
rylation and GTP/(GTP+GDP) levels. In addition, a long linker was
involved in a decrease of basal FRET level, suggesting that the long
linker decreased basal phosphorylation level (Figure 3) and basal
GTP/(GTP+GDP) level (Supplemental Figure S4). Therefore a one-
tailed t test was applied to the analysis.
We thank A. Miyawaki, T. Akagi, J. Miyazaki, K. Yusa, A. Bradley,
J. Zhang, K. Svoboda, and T. W. J. Gadella, Jr., for the plasmids.
K. Morita, Y. Inaoka, K. Hirano, R. Sakai, N. Nonaka, and A. Kawagishi
are also to be thanked for their technical assistance. We are grateful
4656 | N. Komatsu et al. Molecular Biology of the Cell
Ohashi T, Galiacy SD, Briscoe G, Erickson HP (2007). An experimental
study of GFP-based FRET, with application to intrinsically unstructured
proteins. Protein Sci 16, 1429–1438.
Ouyang M, Sun J, Chien S, Wang Y (2008). Determination of hierarchical
relationship of Src and Rac at subcellular locations with FRET biosensors.
Proc Natl Acad Sci USA 105, 14353–14358.
Rizzo MA, Springer GH, Granada B, Piston DW (2004). An improved cyan
fluorescent protein variant useful for FRET. Nat Biotechnol 22, 445–449.
Roux PP, Ballif BA, Anjum R, Gygi SP, Blenis J (2004). Tumor-promoting
phorbol esters and activated Ras inactivate the tuberous sclerosis tumor
suppressor complex via p90 ribosomal S6 kinase. Proc Natl Acad Sci
USA 101, 13489–13494.
Sato M, Ueda Y, Takagi T, Umezawa Y (2003). Production of PtdInsP3 at
endomembranes is triggered by receptor endocytosis. Nat Cell Biol 5,
Violin JD, Zhang J, Tsien RY, Newton AC (2003). A genetically encoded fluo-
rescent reporter reveals oscillatory phosphorylation by protein kinase C.
J Cell Biol 161, 899–909.
Yoshizaki H, Ohba Y, Kurokawa K, Itoh RE, Nakamura T, Mochizuki N,
Nagashima K, Matsuda M (2003). Activity of Rho-family GTPases dur-
ing cell division as visualized with FRET-based probes. J Cell Biol 162,
Yusa K, Rad R, Takeda J, Bradley A (2009). Generation of transgene-free
induced pluripotent mouse stem cells by the piggyBac transposon. Nat
Methods 6, 363–369.
Zacharias DA, Violin JD, Newton AC, Tsien RY (2002). Partitioning of lipid-
modified monomeric GFPs into membrane microdomains of live cells.
Science 296, 913–916.
Zhang J, Ma Y, Taylor SS, Tsien RY (2001). Genetically encoded reporters of
protein kinase A activity reveal impact of substrate tethering. Proc Natl
Acad Sci USA 98, 14997–15002.
for tyrosine phosphorylation of the CrkII adaptor protein in vivo. J Biol
Chem 276, 31305–31310.
Levskaya A, Weiner OD, Lim WA, Voigt CA (2009). Spatiotemporal control
of cell signalling using a light-switchable protein interaction. Nature 461,
Li IT, Pham E, Truong K (2006). Protein biosensors based on the principle of
fluorescence resonance energy transfer for monitoring cellular dynamics.
Biotechnol Lett 28, 1971–1982.
Miyawaki A (2003). Visualization of the spatial and temporal dynamics of
intracellular signaling. Dev Cell 4, 295–305.
Miyawaki A, Llopis J, Heim R, McCaffery JM, Adams JA, Ikura M, Tsien RY
(1997). Fluorescent indicators for Ca2+ based on green fluorescent
proteins and calmodulin. Nature 388, 882–887.
Mochizuki N, Yamashita S, Kurokawa K, Ohba Y, Nagai T, Miyawaki A,
Matsuda M (2001). Spatio-temporal images of growth-factor-induced
activation of Ras and Rap1. Nature 411, 1065–1068.
Nagai T, Ibata K, Park ES, Kubota M, Mikoshiba K, Miyawaki A (2002). A
variant of yellow fluorescent protein with fast and efficient maturation for
cell-biological applications. Nat Biotechnol 20, 87–90.
Nagai T, Yamada S, Tominaga T, Ichikawa M, Miyawaki A (2004). Ex-
panded dynamic range of fluorescent indicators for Ca(2+) by circularly
permuted yellow fluorescent proteins. Proc Natl Acad Sci USA 101,
Nguyen AW, Daugherty PS (2005). Evolutionary optimization of fluorescent
proteins for intracellular FRET. Nat Biotechnol 23, 355–360.
Niwa H, Yamamura K, Miyazaki J (1991). Efficient selection for high-expression
transfectants with a novel eukaryotic vector. Gene 108, 193–199.
Obata T, Yaffe MB, Leparc GG, Piro ET, Maegawa H, Kashiwagi A,
Kikkawa R, Cantley LC (2000). Peptide and protein library screen-
ing defines optimal substrate motifs for AKT/PKB. J Biol Chem 275,