9 4 2 VOLUME 29 NUMBER 10 OCTOBER 2011 nature biotechnology
A R T I C L E S
Fluorescent proteins (FPs)
1
whose fluorescence can be revers-
ibly or irreversibly switched by optical irradiation have opened
new opportunities for the imaging of cells. They have facilitated
in vivo protein-tracking schemes
2,3
, applications based on single-
molecule observations
4,5
and fluorescence microscopy with
subdiffraction resolution
6–10
.
Still, photoswitchable proteins have not displayed their full poten-
tial, because proteins that are just photoactivatable
11–13
can be
switched only once, which implies that repeated measurements with
the same molecule are impossible. On the other hand, photochromic
or reversibly switchable fluorescent proteins (RSFPs) can be repeat-
edly photoswitched between a fluorescent and a nonfluorescent state
by irradiation with light of two different wavelengths. However, in
all previously characterized RSFPs, the wavelength used for generat-
ing the fluorescence emission is identical to one of the wavelengths
used for switching the fluorescence on or off. The result is a complex
interlocking of switching and fluorescence readout
14–22
, impeding or
even precluding many applications, including fluorescence nanoscopy
(super-resolution microscopy). Hence, the identification of an RSFP
in which the generation of fluorescence is disentangled from switch-
ing has long been pursued.
RESULTS
Generation of the RSFP Dreiklang
Numerous GFP variants exhibit some degree of (generally undesir-
able) reversible photoswitching
4,23,24
. We found that the fluorescence
of the yellow fluorescent protein Citrine
25,26
, a derivative of GFP, can
be reversibly modulated to a small extent by alternate irradiation with
light of 365 nm (on switching) and 405 nm (off switching), whereas
fluorescence is excited at 515 nm. However, the achievable contrast
was low, especially at pH values >6, rendering the reversible switching
of Citrine unusable (Supplementary Fig. 1).
To further develop this unusual switching behavior, we performed
extensive random mutagenesis as well as directed PCR-mediated
mutagenesis on a plasmid encoding Citrine. We transformed
Escherichia coli with the plasmid, and screened with an automated
home-built fluorescence microscope for bacterial colonies expressing
fluorescent proteins whose fluorescence was excited with green light
(515 nm) and which could be reversibly photoswitched from a fluorescent
state to a long-lived nonfluorescent state by irradiation with near-UV
(405 nm) light and back to a fluorescent state by UV (365 nm) light
(Fig. 1a
). In several consecutive screening rounds ~70,000 individual
clones were analyzed. Finally, we identified a mutant differing from
Citrine at four positions (Citrine-V61L, F64I, Y145H, N146D)
(Supplementary Fig. 2), which can be effectively switched and excited
to fluoresce. We named this switchable fluorescent protein Dreiklang,
the German word for a three-note chord in music.
At thermal equilibrium, Dreiklang adopts the brightly fluorescent
on-state, with a quantum yield of 0.41 and an extinction coefficient
of 83,000 M
−1
cm
−1
at pH 7.5 (Table 1). In the on-state, it exhibits
two absorption bands (peaking at 412 nm and 511 nm), correspond-
ing to the neutral (protonated) and ionized (deprotonated) states of
the chromophore, respectively (Fig. 1b)
1
. The pK
a
of the on-state
chromophore is 7.2, which is 1.5 pH units higher than the pK
a
of
A reversibly photoswitchable GFP-like protein with
fluorescence excitation decoupled from switching
Tanja Brakemann
1,6
, Andre C Stiel
1,6
, Gert Weber
2
, Martin Andresen
1
, Ilaria Testa
1
, Tim Grotjohann
1
,
Marcel Leutenegger
1
, Uwe Plessmann
3
, Henning Urlaub
3,4
, Christian Eggeling
1
, Markus C Wahl
2
,
Stefan W Hell
1
& Stefan Jakobs
1,5
Photoswitchable fluorescent proteins have enabled new approaches for imaging cells, but their utility has been limited either
because they cannot be switched repeatedly or because the wavelengths for switching and fluorescence imaging are strictly
coupled. We report a bright, monomeric, reversibly photoswitchable variant of GFP, Dreiklang, whose fluorescence excitation
spectrum is decoupled from that for optical switching. Reversible on-and-off switching in living cells is accomplished at illumination
wavelengths of ~365 nm and ~405 nm, respectively, whereas fluorescence is elicited at ~515 nm. Mass spectrometry and high-
resolution crystallographic analysis of the same protein crystal in the photoswitched on- and off-states demonstrate that switching
is based on a reversible hydration/dehydration reaction that modifies the chromophore. The switching properties of Dreiklang enable
far-field fluorescence nanoscopy in living mammalian cells using both a coordinate-targeted and a stochastic single molecule
switching approach.
1
Max Planck Institute for Biophysical Chemistry, Department of NanoBiophotonics, Göttingen, Germany.
2
Freie Universität Berlin, Institut für Chemie und Biochemie,
AG Strukturbiochemie, Berlin, Germany.
3
Max Planck Institute for Biophysical Chemistry, Bioanalytical Mass Spectrometry, Göttingen, Germany.
4
University Medical
Center Göttingen, Department of Clinical Chemistry, Bioanalytics, Göttingen, Germany.
5
University of Göttingen Medical School, Göttingen, Germany.
6
These authors
contributed equally to this work. Correspondence should be addressed to S.W.H. (shell@gwdg.de) or S.J. (sjakobs@gwdg.de).
Received 12 April; accepted 20 July; published online 11 September 2011; doi:10.1038/nbt.1952
© 2011 Nature America, Inc. All rights reserved.
nature biotechnology VOLUME 29 NUMBER 10 OCTOBER 2011 9 4 3
A R T I C L E S
Citrine (5.7) (Supplementary Fig. 1c,d)
26
. Hence, an important role
of the four mutations that discriminate Dreiklang from Citrine is to
shift the pK
a
of the on-state chromophore.
Excitation with 500 nm light results in fluorescence with an emis-
sion maximum at 529 nm (Fig. 1b). Irradiation at the absorption band
of the neutral state with light of 405 nm, switched the protein to a non-
fluorescent off-state (Fig. 1c,d). Upon photoswitching to the off-state,
a new and unusual absorption band at 340 nm appeared, whereas the
absorption bands corresponding to the on-state disappeared (Fig. 1e).
Subsequent illumination at the 340-nm absorption band of the off-
state switched the protein back into the on-state (Fig. 1c,d). Switching
Dreiklang on at 365 nm and off at 405 nm was reversible (Fig. 1d). As
expected from the pK
a
of the on-state chromophore, the fluorescence
of Dreiklang could be reversibly switched with a good contrast in a
pH range from 6 to 9 (Supplementary Fig. 1).
Effective switching could be performed at all physiological tem-
peratures applied (10–40 °C) and was accelerated with increasing
temperatures, although the maximal fluorescence was concomitantly
reduced at higher temperatures (Supplementary Fig. 3). Likewise, the
thermal relaxation from the optically induced nonfluorescent state into
the equilibrium on-state depends on temperature (Fig. 1f, inset).
Thermal equilibration was not affected by irradiation at 515 nm
(0.82 W/cm
2
), thus, exciting fluorescence, affected neither the
nonfluorescent nor the fluorescent state (Fig. 1f), although noticeable
cross-talk to the absorption band of the neutral state is anticipated
at higher intensities. In living E. coli cells expressing Dreiklang, we
recorded ~160 switching cycles before the fluorescence was reduced to
50% (at 1.4 W/cm
2
, 365 nm; 110 W/cm
2
, 405 nm; 0.82 W/cm
2
, 515 nm)
(Supplementary Fig. 4), demonstrating its high switching endurance.
The on-off contrast of the fluorescence signal depends on the
sample and the light intensities applied. In widefield images of living
mammalian cells, the off-state background signal was <1.4% of the
on-state signal (see below), whereas in thick layers of E. coli cells,
which were used for screening, it was generally 5–10% (Fig. 1d and
Supplementary Fig. 4). In the latter case, we attribute the background
mainly to scattering, rather than to ineffective switching.
On native gels, Dreiklang behaves as a monomer (Supplementary
Fig. 5). Its maturation half-time is ~1.2 h at 30 °C and ~2 h at
37 °C (Supplementary Fig. 6). At pH = 7.5, Dreiklang is more resist-
ant to photobleaching than its parent Citrine when using the same
light intensities for exciting fluorescence, which may be partly due to
Dreiklang’s lower extinction coefficient (Supplementary Fig. 7).
Light-driven switching of Dreiklang
To determine the molecular basis for the reversible switching of
Dreiklang, we solved the structures of the light-induced off-state (1.7 Å)
and on-state (2.0 Å) using the same protein crystal. To this end, a
Dreiklang protein crystal in a buffer of pH 4.6 was switched at room
temperature (295 K) from the fluorescent equilibrium state into the off-
state by irradiation at 405 nm until fluorescence reached a minimum.
Table 1 Properties of Citrine and Dreiklang
Absorbance
(nm)
Emission
(nm) QY
ε-max
(M
−1
cm
−1
) Equilibrium
Citrine 515 533 0.54 132,000 On
Dreiklang 340
off
/412
on
/511
on
529 0.41 83,000 On
Quantum yield (QY) and extinction coefficient (ε) were measured for the fluorescent
on-state protein at pH 7.5.
Figure 1 Properties of Dreiklang. (a) Scheme
depicting Dreiklang’s switching modality.
(b) Normalized absorbance (solid black line),
fluorescence emission (dashed green line) and
fluorescence excitation (dotted red line) spectra
of the (fluorescent) equilibrium-state Dreiklang
at pH 7.5. (c,d) Switching curves of Dreiklang’s
fluorescence recorded on colonies of living
E. coli. Off- and on-switching was performed
with near-UV (405 nm) and UV light (365 nm),
respectively, and fluorescence read-out with
green light (515 nm). The respective irradiation
scheme is indicated on top of the graphs by
the colored bars. (c) One switching cycle.
Fluorescence was continuously recorded.
(d) Twenty consecutive switching cycles.
Fluorescence was recorded when the cells were
irradiated with green light only. (e) Irradiation-
dependent changes in Dreiklang absorbance.
Absorbance spectra obtained at the indicated
time points during switching of equilibrium-
state Dreiklang (pH 7.5) into the off-state
by irradiation with 405 nm. (f) Irradiation-
independent changes in Dreiklang fluorescence
due to the thermal equilibration from the off-
state into the fluorescent equilibrium state.
After off-switching, fluorescence was recorded
at 25 °C under constant irradiation with 515 nm
(black line) or by consecutive 20-ms pulses
of 515 nm light in 60 s intervals (red dots).
The similar curves demonstrate that 515 nm
light does not photoswitch Dreiklang.
Inset: relaxation half-time from the off- into
the equilibrium-state as a function of temperature. The data were obtained on purified Dreiklang (pH 7.5) (circles) or on living cells expressing
Dreiklang targeted to the ER (squares). Red line: single exponential fit to the data obtained on purified protein.
a
c
d f
e
b
515
1.0
529
ON
405
405
365
515
OFF
365
0.8
0.6
Absorption / fluorescence
0.4
0.2
0
1.6
1.2
0.8
Fluorescence
0.4
0
1.0
0.8
0.6
Fluorescence
0.4
0.2
0
1.0
0.8
0.6
Fluorescence
Relaxation
half-time (s)
0.4
0.2
0
0 500 1,000
1,600
1,200
800
400
0
10
Temperature (°c)
20 30 40
1,500
Time (s)
2,000 2,500
0 5 10
Cycles
15 20
2.5
2.0
1.5
Absorption
1.0
0.5
0
0 5 10 15 20
Time (s)
25 30
300 350 400 450
Wavelength (nm)
500 550 600
300 350 400 450
Wavelength (nm)
500 550 600
Equilibrium
0.5 min
1.0 min
1.5 min
2.0 min
3.0 min
4.0 min
20 min
© 2011 Nature America, Inc. All rights reserved.
9 4 4 VOLUME 29 NUMBER 10 OCTOBER 2011 nature biotechnology
A R T I C L E S
After the off-state diffraction data was recorded at 100 K, we warmed
the very same crystal of Dreiklang back to 295 K, switched it by irra-
diation with 365 nm light until the fluorescence reached a maximum
and recorded the on-state diffraction data at 100 K. In addition, we
solved the X-ray structure of Dreiklang in the fluorescent equilib-
rium-state to a resolution of 1.9 Å (Fig. 2). Notably, the kinetics of
the thermal equilibration of the fluorescence signal after switch-
ing off was comparable for Dreiklang in solution and in the crystal
(Supplementary Fig. 8), indicating that the crystal lattice did not have
major effects on the switching behavior.
The overall structure of Dreiklang resembles that of GFP and related
proteins (Fig. 2b). The chromophore, autocatalytically formed from
the Gly65-Tyr66-Gly67 tripeptide, resides in an alpha-helical seg-
ment, enclosed by an 11-stranded beta-barrel. As expected from the
similar spectroscopic properties, the on-state structure was practically
superimposable on the fluorescent equilibrium-state structure. The
on-state chromophore consists of an imidazolinone-ring, connected
by a methine bridge to a p-hydroxyphenyl ring. The two rings of the
chromophoric systems were largely co-planar, facilitating a conju-
gated pi-electron system and hence supporting fluorescence.
In the off-state, the p-hydroxyphenyl ring lies largely in plane with the
C
α66
-C
β66
bond, as well as with the C
α66
-C
66
and the C
α66
-N
66
bonds,
indicating that the methine bridge connecting the two rings is
maintained. However, in the off-state structure, the planarity of the
five-membered ring was markedly distorted with the chromophoric
C
65
exhibiting a tetrahedral geometry indicative of an sp
3
hybridi-
zation. A clear signal in the electron density map indicates a new
hydroxyl group at the C
65
atom, suggesting that the imidazolinone
ring was converted into a 2-hydroxyimidazolidinone ring (Fig. 2a
and Supplementary Fig. 9). We propose that the hydration of the
imidazolinone ring shortens the chromophoric pi-electron system,
resulting in the new absorption band at 340 nm and the simultaneous
disappearance of the absorption bands at 412 and 511 nm (Fig. 1e).
To further confirm this light-induced chemical modification, we
carried out electrospray ionization mass spectrometry (ESI-MS).
To this end, switching of Dreiklang in solution (pH 6.9; 295 K) was
monitored by measuring the fluorescence signal; the proteins in the
respective states were immediately analyzed by ESI-MS under native
conditions (in 18% acetonitrile). We found a mass difference of 18 ±
0.3 Da between the nonfluorescent state and the light-induced on-
state or the equilibrium state, respectively (Fig. 2a and Supplementary
Figs. 10 and 11). This strongly indicates the reversible covalent addi-
tion of a water molecule that occurred parallel to changes in the fluo-
rescence signal. Hence, the ESI-MS data are in full agreement with the
X-ray data, supporting the view of a reversible light-induced covalent
chemical modification, that is, a hydration-dehydration reaction of
the chromophoric five-membered ring as the underlying molecular
mechanism of switching in Dreiklang.
A similar reversible hydration reaction was postulated, although
controversially discussed, to occur during the chromophore forma-
tion of GFP
27–29
. This might suggest that the light-induced reversible
switching of Dreiklang is based on a molecular reaction that is pos-
sibly occurring during chromophore maturation of some fluorescent
proteins. Hence, we propose that Dreiklang may be used as a scaffold
for further engineering and that this switching mechanism may be
transferred to other fluorescent proteins.
Our mutagenesis studies showed that the amino acid residues Y203
and E222 as well as the chromophore building G65 are crucial for the
unusual switching characteristics of Dreiklang. The amino acids G65
and Y203 facilitate the positioning of the side chain of E222 close to
the imidazolinone ring (Fig. 2c). In the fluorescent-state, Y203 and
E222 form hydrogen bonds to a water molecule (Wat
a
) and thereby
stabilize it in close vicinity to the C
65
of the chromophore (Fig. 2d),
a situation that is different in the nonswitchable GFP (avGFP-S65T)
25
(Fig. 2d, inset). In GFP, a water molecule corresponding to Wat
a
is
stabilized by water-mediated H-bonds only. We propose that Wat
a
is
Figure 2 Molecular basis of Dreiklang
photoswitching. (a) Dreiklang in the fluorescent
equilibrium-state (top), the nonfluorescent
off-state (middle) and the fluorescent on-state
(bottom). Left, top: representative Dreiklang
protein crystal. Left, bottom: proposed chemical
structure of the chromophore. Central: details
of the X-ray structures (PDB IDs: 3ST2, 3ST3,
3ST4, respectively, top to bottom). Shown is the
chromophore (carbon, magenta/gray; oxygen,
red; nitrogen, blue). In the equilibrium-state
and the on-state, water Wat
a
(magenta sphere)
is additionally displayed. Final 2F
o
–F
c
electron
densities are contoured at the 1σ level. The
off-state and the on-state structures have been
successively recorded on the same protein
crystal. Right: representative deconvoluted
ESI-MS spectra of Dreiklang photoswitched in
solution and measured under native conditions.
(b) Overall Dreiklang ribbon structure displayed
in two orthogonal views. (c) Chromophore and
immediate surrounding of on-state Dreiklang
(magenta) and GFP (PDB: 1EMA
25
) (cyan).
The Van-der-Waals’ radii of important atoms
are indicated by spheres to highlight structural
restraints. The chromophores are depicted
as ball and stick whereas the surrounding
amino acid residues are shown in the stick
representation. (d) Superimposed representations of the Dreiklang hydrogen bond network in the (fluorescent) equilibrium- and the off-states.
Equilibrium-state carbons, magenta; off-state carbons, gray; oxygen, red; nitrogen, blue. Important water molecules are shown as magenta
(equilibrium-state) and gray (off-state) spheres. Inset: hydrogen bond network in GFP.
a b
d
c
17.0
28,510.4 Da
hv 405 nm
T203
GFP
E222
GFP
GYG
off
Wat
c
Wat
b
Wat
a
GYG
eq
H145
Y203
S205
E222
H148
T203
TYG
Q69
S205
E222
GFP
TYG
GFP
Y203
Drei
E222
Drei
GYG
Drei
hv 365 nm
14.0
10.0
Intensity (a.u.)
Intensity (a.u.)
6.9
3.5
85
68
51
34
17
0
28,400 28,500 28,600
28,400 28,500
Mass (Da)
Mass (Da)
28,528.2 Da
Intensity (a.u.)
7.8
6.3
4.7
3.1
1.6
0
28,400 28,500 28,600
Mass (Da)
28,510.5 Da
28,600
0
EquilibriumOff-stateOn-state
© 2011 Nature America, Inc. All rights reserved.
nature biotechnology VOLUME 29 NUMBER 10 OCTOBER 2011 9 4 5
A R T I C L E S
important for the light-driven off-switching in Dreiklang, because it
appears to be in a suitable position for a nucleophilic addition across
the C=N bond of the imidazolinone ring.
The off-state structure reveals a new water molecule (Wat
b
) dis-
placed by 1.02 ± 0.09 Å from the position of Wat
a
. It is held in this
position by a differently configured hydrogen bonding network. We
found during occupancy refinement of Wat
b
that the occupancy was
significantly smaller than 1.0 (e.g., chain C: 0.35), strongly indicating
that this position is not always occupied by a water molecule
in the off-state protein. Hence we assume that Wat
b
is taken up from
the environment after Wat
a
has been used for the light-induced
hydration of the imidazolinone ring upon switching from the
on- to the off-state.
Altogether, these findings lend support to a reversible light-induced
hydration/dehydration reaction of the five-membered ring. In the
on-state, the chromophore exists in the protonated and the depro-
tonated form, resulting in absorption bands at 412 nm and 511 nm,
respectively. Irradiation at the 511-nm band induces fluorescence,
whereas irradiation at the 412-nm band induces a covalent modifica-
tion (hydration) of the five-membered ring, resulting in a nonfluores-
cent chromophore absorbing at 340 nm. Subsequent irradiation at this
band results in a dehydration of the off-state chromophore convert-
ing it back into the on-state chromophore. Although the reversible
water addition/elimination reaction appears to be the key factor in
the unusual switching behavior of Dreiklang, it is possible that addi-
tional short-lived intramolecular rearrangements may occur, possibly
including a cis-trans isomerization, structural flexibility or a strong
bending of the chromophore.
Use of Dreiklang for fluorescence recovery after switching
To evaluate the properties of Dreiklang for imaging purposes, we
prepared a layer of purified Dreiklang molecules and wrote complex
patterns
6,30,31
into this layer showing that Dreiklang can be exploited
for reversibly recording and reading information (Fig. 3a and
Supplementary Movie 1). Next, we generated several fusion proteins,
namely Dreiklang-Map2, Dreiklang-α
-tubulin, Dreiklang-histone2B,
keratin19-Dreiklang, vimentin-Dreiklang, β-actin-Dreiklang and
Dreiklang targeted to the mitochondrial matrix (Mito-Dreiklang)
and expressed these fusion proteins in cultivated mammalian cells
(Fig. 3b and Supplementary Fig. 12). The fact that Dreiklang could
be functionally fused to α-tubulin and histone2B demonstrates that
Dreiklang behaves as a monomer also in vivo. Targeted to the different
cellular structures, Dreiklang could be switched on and off in living
cells repeatedly (Fig. 3b).
Because switching Dreiklang on and off requires UV and near-
UV light, we were concerned that this irradiation might negatively
affect the viability of the cells. To investigate this issue, we expressed
vimentin-Dreiklang in PtK2 cells. The Dreiklang fluorescence was
switched 100 times on and off in whole living cells with widefield
illumination, displaying a contrast in the fluorescence signal of around
75:1 in each cycle (Supplementary Fig. 13 and Supplementary
Movie 2). No major alterations in the labeled structures were
observed during this period (~10 min). Next, we targeted Dreiklang
to the endoplasmic reticulum (ER) of PtK2 cells and switched the
ability to fluoresce 5 or 20 times on and off (Supplementary Fig. 14).
We found that five switching cycles had no detectable effect on the
viability of the cells after 2 h compared to the control where switching
was omitted. After 20 cycles, ~16% of the cells were dead after 2 h
(n = 300 cells, six independent experiments) and a larger and highly
variable fraction of cells showed changes in their shapes. This finding
demonstrates that repeated irradiation with UV light is unfavorable
but does not necessarily induce cell death. However, most practical
applications will require either fewer switching cycles and/or much
smaller regions exposed to UV light for switching, which thus would
reduce the overall light dose by at least an order of magnitude and the
effects on the cellular viability accordingly.
As a demonstration that Dreiklang’s properties can be exploited
for repeated measurements of protein dynamics in individual living
cells, we expressed Dreiklang targeted to the ER in Vero cells (Fig. 3c).
Using a commercial confocal microscope, we selectively photo-
switched (with 405 nm) ER-Dreiklang in one region of the living cell
from the fluorescent into the nonfluorescent state and followed the
movement of nonswitched on-state molecules into the switched-off
region by probing fluorescence with 515 nm. After the measurement,
the fluorophores in the whole cell were switched back to the initial
fluorescent state with 360 nm using wide-field illumination; the
Figure 3 Applications of Dreiklang. (a) ‘PacMan’
movie. Thirty-three individual images were
written successively at the same position
of a polyacrylamide-Dreiklang layer. Before
writing each new frame, all molecules were
photoswitched to the on-state. Shown are
the first and the thirty-third frame (see also
Supplementary Movie 1). Scale bar, 100 µm.
(b) Switching of various Dreiklang fusion
proteins in living Vero cells (from left to right,
on, off, etc.). From top to bottom: Dreiklang-
MAP2, Dreiklang-α-tubulin, mito-Dreiklang,
Dreiklang-Histone2B. Fluorescence was excited
with green light (495 nm); switch-off (light
blue arrowheads): near-UV (420 nm); switch-on
(violet arrowheads): UV-light (360 nm).
(c) FRAS with Dreiklang. Dreiklang was targeted
to the ER in living Vero cells. Images from top
to bottom: overview; before switching;
immediately after switching Dreiklang off in
region of interest (ROI)-i; and at the end of the measurement (65 s after switching). Graphs on the right: the plotted fluorescence signals were collected
within the indicated ROIs during 20 repetitions of the same FRAS experiment on a single cell (ROI-i: in this region Dreiklang was selectively switched off
prior to the measurement; ROI-ii: in the same cell, showing the flow of switched-off Dreiklang molecules into this region; ROI-iii: in the neighboring cell).
Shown are raw data. The red line marks the mean values at each time point, demonstrating the reduction of statistical noise. Scale bars in b and c, 10 µm.
a
b
c
33
switching
cycles
iii
ii
i
iii
ii
i
iii
ii
i
End Post-switch Pre-switch
1.2
1.0
0.8
0.6
0.4
0.2
0
0
Fluorescence
i
10 20
Time (s)
30 40 50 60 70
1.2
1.0
0.8
0.6
0.4
0.2
0
Fluorescence
ii
Time (s)
0 10 20 30 40 50 60 70
1.2
iii
1.0
0.8
0.6
0.4
0.2
0
Fluorescence
Time (s)
0 10 20 30 40 50 60 70
© 2011 Nature America, Inc. All rights reserved.
9 4 6 VOLUME 29 NUMBER 10 OCTOBER 2011 nature biotechnology
A R T I C L E S
measurement, which was substantially shorter (65 s) than the ther-
mal equilibration of Dreiklang to the on-state, was repeated several
tens of times. We denominate this approach as fluorescence recovery
after switching (FRAS). Individual measurements of cellular protein
movements often exhibit statistical noise
32–34
, which was strongly
reduced in FRAS by averaging over many switching cycles. Analogous
approaches have previously been done using conventional RSFPs or
photoactivatable proteins
16,35
.
Nanoscopy with Dreiklang
To overcome the diffraction barrier, all super-resolution (nanoscopy)
concepts use a molecular mechanism to sequentially inhibit the
fluorescence of adjacent features
36
. Two families of approaches rest
on this principle. In the first family, which has been termed RESOLFT
(reversible saturable optical (fluorescence) transitions between two
states), the sample coordinates at which the fluorophores are on and
off are predefined by a pattern of light
6,8,36
. In the second family
encompassing (F)PALM, STORM and others
9,10,36,37
, individual
fluorophores are switched on and off stochastically, whereby the
coordinate of the emitting molecule is found using the photons
detected on a camera. We reasoned that the decoupling of fluores-
cence switching from excitation in Dreiklang would enable additional
possibilities for super-resolution microscopy.
In the stochastic methods, the number of emitted photons and the
emission rate of a molecule in the on-state together determine the locali-
zation precision and the recording time. Although these parameters are
difficult to control in conventional fluorescent proteins or RSFPs, in
Dreiklang the on-time can be influenced by adapting the light intensities
used for switching on and off, eventually even interactively in response
to the local fluorophore densities. We fused Dreiklang to Map2 and
expressed the fusion construct in PtK2 cells. The cells were initially
irradiated for 10 s at 405 nm (0.2 kW/cm
2
) to switch the majority of
Dreiklang proteins off. Subsequently we recorded 8,000 image frames
(10 ms each) with irradiation at 491 nm (2 kW/cm
2
). Between two
frames the cells were irradiated for 0.1–0.5 µs with light of 405 nm
and/or 355 nm (2 W/cm
2
) to adjust the number of on-state fluorophores
to less than one per diffraction area. We obtained an average localization
precision of ~15 nm using ~700 detected photons, resulting in images
that were clearly superior to the conventional counterparts (Fig. 4a).
We also realized an implementation based on spatially targeted
switching of the fluorophores (RESOLFT)
6,8
. We imaged living PtK2
cells expressing keratin19-Dreiklang by first switching Dreiklang on
at 355 nm (1 ms, 160 W/cm
2
) using a regular shaped, diffraction-
limited focus. Then, a doughnut-shaped focus of light of 405 nm
(40 ms, 110 W/cm
2
) with a central intensity minimum (zero) was
used for switching Dreiklang off at the focal periphery only. Finally,
fluorescence was probed by 491 nm excitation (4 kW/cm
2
) for 10 ms.
The image was generated by scanning over the sample using this irra-
diation scheme at every pixel (Fig. 4b). The Dreiklang images taken in
the RESOLFT mode reveal many details with a resolution of down to
~35 nm (Supplementary Fig. 15
) that are fully blurred in the corres-
ponding confocal data. Note that the presented image data are raw.
DISCUSSION
The photoswitching of Dreiklang is based on a light-driven reversible
hydration-dehydration reaction modifying the chromophore, which
uniquely decouples the switching spectra from their excitation
counterpart. Thus Dreiklang allows fine-tuning of the duration of
the chromophore states without interference by the fluorescence
excitation light.
This feature provides benefits for super-resolution microscopy tech-
niques applying stochastic single-molecule switching (Fig. 4a) and spa-
tially targeted switching (Fig. 4b), but should also enable applications
that would be difficult or even impossible with conventional switch-
able fluorescent proteins. Among others, this includes FRAS (
Fig. 3c),
allowing multiple bleaching or photoactivation measurements in
single cells, potentially enabling a new generation of single cell–based
screening approaches. One may also envision sophisticated Förster
resonance energy transfer (FRET) measurements with multiple
fluorophores using Dreiklang as an additional switchable component.
Dreiklang may also provide new avenues to multicolor applications in
combination with other RSFPs or conventional fluorescent proteins.
In this case, contrast may be obtained either by switching
19
, or by
different colors or by dissimilar fluorescence lifetimes. In addition to
applications in life sciences, the decoupling of switching from fluo-
rescence excitation should also offer new options in subdiffraction
reading and writing in protein layers.
Long-term protein tracking will require Dreiklang variants with
longer lifetimes of the metastable states. Likewise, it is conceivable to
design proteins offering even more switching cycles and operating
with longer wavelengths. With its X-ray structure available, we antici-
pate the specific generation of further Dreiklang derivatives forming
a new class of switchable fluorescent proteins.
METHODS
Methods and any associated references are available in the online version
of the paper at http://www.nature.com/naturebiotechnology/.
Accession code. The atomic coordinates and structure factors have
been deposited in the Protein Data Bank, http://www.pdb.org (PDB
ID codes 3ST2, 3ST3 and 3ST4).
Note: Supplementary information is available on the Nature Biotechnology website.
ACKNOWLEDGMENTS
We acknowledge access to beamline BL14.2 of the BESSY II storage ring (Berlin)
through the Joint Berlin MX-Laboratory sponsored by the Helmholtz Zentrum
Berlin für Materialien und Energie, the Freie Universität Berlin, the Humboldt-
Universität zu Berlin, the Max-Delbrück Centrum and the Leibniz-Institut für
i
i
ii
ii
iii
iii
iv
iv
a
b
Figure 4 Super-resolution microscopy of living PtK2 cells using Dreiklang.
(a) Cells expressing Dreiklang-Map2 imaged both conventionally (left)
and by super-resolution microscopy based on single-molecule stochastic
switching (center). (b) Keratin19-Dreiklang expressed in living cells and
imaged both confocally (left) and in the RESOLFT mode (spatially targeted
switching) (center). Right: magnifications of the regions indicated in the
main images. Scale bars, 1 µm (middle, left), 250 nm (right).
© 2011 Nature America, Inc. All rights reserved.
