A near-infrared fluorophore for live-cell super-
resolution microscopy of cellular proteins
Graz
ˇ
vydas Lukinavic
ˇ
ius
1
†
, Keitaro Umezawa
1
†
, Nicolas Olivier
2
, Alf Honigmann
3
, Guoying Yang
4
,
Tilman Plass
5
,VeronikaMueller
3
, Luc Reymond
1
,IvanR.Corre
ˆ
aJr
6
, Zhen-Ge Luo
7
, Carsten Schultz
5
,
Edward A. Lemke
5
, Paul Heppenstall
4
, Christian Eggeling
3,8
, Suliana Manle y
2
and Kai Johnsson
1
*
The ideal fluorescent probe for bioimaging is bright, absorbs at long wavelengths and can be implemented flexibly in living
cells and in vivo. However, the design of synthetic fluorophores that combine all of these properties has proved to be
extremely difficult. Here, we introduce a biocompatible near-infrared silicon–rhodamine probe that can be coupled
specifically to proteins using different labelling techniques. Importantly, its high permeability and fluorogenic character
permit the imaging of proteins in living cells and tissues, and its brightness and photostability make it ideally suited for
live-cell super-resolution microscopy. The excellent spectroscopic properties of the probe combined with its ease of use in
live-cell applications make it a powerful new tool for bioimaging.
F
lexible and specific methods to couple synthetic fluorescent
probes to proteins in living cells are established, yet these
methods are not generally compatible with the fluorophores
best suited for live-cell imaging
1
. Fluorescent labels that are
excited and emit in the near-infrared are especially biocompatible,
because they avoid the use of light, which may cause phototoxicity
or unwanted autofluorescent background
2
. Although numerous
synthetic near-infrared fluorophores with exceptional photostability
and brightness exist, they tend to be membrane impermeable and
show unspecific binding to cellular components (Supplementary
Table S1 and Fig. S1)
3
. Thus, invasive approaches, such as electro-
poration or glass-bead loading, are required to introduce these
probes into cells and/or cumbersome additional manipulations
are needed to reduce the background signal from unspecific
binding
4–6
. Here, we introduce a highly permeable and biocompati-
ble near-infrared silicon–rhodamine (SiR) fluorophore that can be
coupled specifically to intracellular proteins in live cells and
tissues using different labelling techniques. The fluorogenic charac-
ter of the probe and its high brightness permit live-cell imaging
experiments without washing steps. In addition, the fluorophore
proved to be ideally suited for live-cell super-resolution microscopy.
Results and discussion
To develop near-infrared fluorophores suitable for the specific lab-
elling of proteins inside living cells we turned our attention to a
recently introduced class of fluorophores based on silicon-containing
rhodamine derivatives such as SiR-methyl (Fig. 1a)
7–9
. SiR deriva-
tives were shown to possess excellent spectroscopic properties,
and in certain cases to be membrane permeable
8
. To achieve selec-
tive coupling of SiR derivatives to the proteins of interest, we envi-
sioned exploiting self-labelling protein tags such as SNAP-tag
1
.
SNAP-tag fusion proteins can be labelled specifically with molecular
probes using benzylguanine (BG) derivatives
10,11
. However, imaging
experiments with a BG derivative of SiR-methyl revealed an unac-
ceptably high background signal from unspecific binding of the
probe to different cellular structures (Supplementary Fig. S2). We
speculated that exchange of the methyl group at the 2-position of
the phenyl ring by a carboxyl group would reduce the hydrophobi-
city and unspecific binding of the fluorophore. At the same time the
carboxyl group would permit the molecule to be in equilibrium with
the uncharged spirolactone and could increase membrane per-
meability (Fig. 1a). Recently, spirolactames and spirolactones of
SiR derivatives were described for the development of an irreversible
turn-on sensor for Hg
þ
(ref. 12).
We therefore synthesized SiR-carboxyl in five steps from com-
mercially available material in an overall yield of 22% (for details
see the Supplementary Information). SiR-carboxyl is excitable at
around 640–650 nm and emits at around 660–670 nm, and the exci-
tation and emission maxima were found to be 645 nm and 661 nm,
respectively (Supplementary Fig. S3a). It possesses an extinction
coefficient of 100,000 M
21
cm
21
and a fluorescence quantum
yield of 0.39 in aqueous solution (Supplementary Table S2). An
important difference between SiR-carboxyl and tetramethylrhoda-
mine (TMR, Fig. 1a) is the higher propensity of SiR-carboxyl to
form the non-fluorescent spirolactone in solvents of low dielectric
constant 1, which we investigated by using dioxane–water mix-
tures
13
. SiR-carboxyl exists predominantly in its spirolactone
form in dioxane–water mixtures of dielectric constant less than
30, conditions under which TMR remains predominantly in its
open form (Fig. 1b and Supplementary Fig. S3b). As expected,
SiR-methyl is not sensitive to changes in solvent polarity as it
cannot form a spirolactone (Fig. 1b). Next, we synthesized deriva-
tives of SiR-carboxyl for the covalent labelling of the SNAP-tag,
as well as for CLIP-tag and Halo-tag fusion proteins (dubbed
1
Ecole Polytechnique Fe
´
de
´
rale de Lausanne, Institute of Chemical Sciences and Engineering (ISIC), Institute of Bioengineering, National Centre of
Competence in Research (NCCR) in Chemical Biology, 1015 Lausanne, Switzerland,
2
Ecole Polytechnique Fe
´
de
´
rale de Lausanne, Laboratory of Experimental
Biophysics, NCCR in Chemical Biology, 1015 Lausanne, Switzerland,
3
Max-Planck-Institute for Biophysical Chemistry, Department NanoBiophotonics, Am
Fassberg 11, 37077 Go
¨
ttingen, Germany,
4
European Molecular Biology Laboratory, Mouse Biology Unit, via Ramarini 32, 00015 Monterotondo (RM), Italy,
5
European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, Germany,
6
New England Biolabs Inc., 240 County Road, Ipswich,
Massachusetts 01938, USA,
7
Institute of Neuroscience and State Key Laboratory of Neuroscience, Shanghai Institutes for Biological Sciences, Chinese
Academy of Sciences, Shanghai 200031, China,
8
Medical Research Council Human Immunology Unit, Weatherall Institute of Molecular Medicine,
University of Oxford, UK;
†
These authors contributed equally to this work.
*
e-mail: kai.johnsson@epfl.ch
ARTICLES
PUBLISHED ONLINE: 6 JANUARY 2013 | DOI: 10.1038/NCHEM.1546
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry132
© 2013 Macmillan Publishers Limited. All rights reserved.
SiR-SNAP, SiR-CLIP and SiR-Halo; Fig. 1a). CLIP-tag and Halo-
tag are two other popular protein tags utilized for the covalent
labelling of proteins in living cells
14,15
. All three tags can be labelled
readily with their corresponding substrates, with rate constants
comparable to those reported for other substrates (Supplementary
Table S2). Furthermore, the relative fluorescence quantum yield of
the protein-conjugated dye is comparable to that of SiR-carboxyl
(Supplementary Table S2). SiR-labelled proteins also possess excel-
lent photostability: the bleaching rate of the SiR-SNAP-labelled
SNAP-tag is identical to those of Atto647N-labelled proteins
(Atto647N is a dye frequently used in single-molecule applications)
and much lower than the bleaching rate of Alexa647-labelled
proteins (Supplementary Fig. S3c).
Live-cell imaging with SiR-carboxyl derivatives. We applied
SiR-SNAP, SiR-CLIP and SiR-Halo for live-cell imaging. To
demonstrate the potential of SiR-labelled proteins for multicolour
imaging, these experiments were performed in a HeLa cell line
that stably expressed a green fluorescence protein (GFP) fusion
with a -tubulin (GFP-a-tubulin) and a mCherry fusion with
histone H2B (H2B-mCherry) as markers
16
. We observed a specific
fluorescence labelling of fusion proteins of SNAP-, CLIP- and
Halo-tags in different organelles of l iving cells within 30–60
minutes, with no significant background si gnal (Fig. 2), which
demonstr ates that SiR-SNAP, SiR-CLIP and SiR-Halo readily pass
through the plasma membrane as well as the int ernal membranes.
An import ant featur e of the dye for cellular ap plications is that the
reactions of SiR-SNAP , SiR-CLI P and SiR-Halo in a cell culture
medium (DMEM) containing 10% fetal bovine serum (FBS) with
their cognate protein-tag results in a more than fivefold inc rease in
the fluorescence signal (Supplementary Fig. S5a). Consequently,
fluorescence imaging of SiR-labelled proteins can be performed,
albeit with some background signal, even without remo ving excess
substrate through a washing step (Supplementary Fig. S4). The
highly efficient and selective fluorescence labelling of intra cellular
proteins with SiR-SNAP, SiR -CLIP and SiR-Halo sets the dye apart
from al l the other near-infrar ed fluorophor es that we tested
(Supplementary Table S1).
Why are SiR-carboxyl derivatives so exceptionally well suited for
live-cell labelling? A possible explanation is that the reactions of
SiR-SNAP, SiR-CLIP and SiR-Halo with their respective protein
tags keeps the fluorophores in their fluorescent zwitterionic form,
and the aggregation of unreacted dye and unspecific binding to
hydrophobic structures, two sources of background signal when
using synthetic fluorophores in live cells, results in the formation
of non-fluorescent spirolactone. This hypothesis is based on the fol-
lowing observations: addition of the surfactant SDS to aqueous sol-
utions of SiR-SNAP, SiR-CLIP and SiR-Halo increased their
fluorescence intensity more than threefold, which indicates their
reversible aggregation and the concomitant formation of the non-
fluorescent spirolactone, even at submicromolar concentrations
(Supplementary Fig. S5b). The formation of the non-fluorescent
lactone on aggregation becomes apparent from an analysis of the
ultraviolet absorbance spectra of SiR-SNAP (Fig. 1c). In the
absence of SDS the ultraviolet band at 290 nm, characteristic for
the lactone, is the major band, whereas the zwitterion with an absor-
bance maximum at 650 nm becomes predominant in the presence
of SDS (Supplementary Fig. S5c). The formation of non-fluorescent
spirolactones by SiR-carboxyl derivatives in the hydrophobic
environment of aggregates is in agreement with their propensity
to form spirolactones in solvents of relatively low dielectric constant.
For example, SiR-SNAP is present predominantly as its spirolactone
0.0
0.2
0.4
0.6
0.8
1.0
0204060
SiR-Me
SiR-COOH
SiR-SNAP
TMR-COOH
BG-TMR
Normalized area
b
80
Dielectric constant
c
0.0
0.1
0.2
0.3
250 350 450 550 650 750
Absorbance (a.u.)
Wavelen
g
th
(
nm
)
TBS
TBS + 0.1% SDS
Ethanol
Si N
+
+
+
N
O
HO
Si NN
O
R
O
O
Si NN
SiR-carboxyl
SiR-CLIP
SiR-Halo
SiR-methyl
SiR-tetrazine
O
NN
O
HO
O
O
TMR
SiR-SNAP
O
O
R
O
N
H
N
N
N
NH
2
H
N
NN
NH
2
O
H
N
Cl
O
3
O
H
N
N
N
N
N
H
N
R = OH
R R= =
R =
R =
a
–
–
O
Figure 1 | SiR dyes used for SNAP-, CLIP-, Halo-tag and tetrazine labe lling. a, Structur es of SiR dyes, TMR and the formation of spirolactone of SiR-carboxyl.
b, Normali z ed integral of absorpti on spectra of the zwitte rion region of SiR and TMR deriv a tiv es in water–dioxane mixtures as a fun ction of dielectric
constant. Note tha t absorban ce at 1 ¼ 80 (0% of dioxane) is affected by th e aggregation of fluorophores. BG-TMR repr esents TMR coupled to BG
(Supplementary Fig. S2)
11
. c, Absorption spectra of 2.5
m
M SiR-SNAP measur ed in ethanol (red), Tris- buffer ed saline (TBS) buffer with (dashed bla ck) and
without (s olid black) 0.1% sodium dodecyl sulfate (SDS).
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1546
ARTICLES
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry 133
© 2013 Macmillan Publishers Limited. All rights reserved.
in ethanol (1 ¼ 25, Fig. 1c), conditions under which the correspond-
ing TMR derivative exists predominantly as a fluorescent zwitterion
(Supplementary Fig. S5d). SiR-SNAP furthermore shows the same
high propensity as SiR-carboxyl to form the spirolactone in
dioxane–water mixtures (Fig. 1b and Supplementary Fig. S5e).
However, a difference between SiR-SNAP and SiR-carboxyl is
that only SiR-SNAP aggregates and forms the spirolactone in
pure water. Finally, a comparison of absorbance spectra of
SiR-SNAP in the presence of SDS or after coupling to the
SNAP-tag in the absence of SDS indicates that about 65% of SiR
is present in its fluorescent zwitterionic form after covalent coupling
to the SNAP-tag (Supplementary Fig. S5c). Overall, these experi-
ments support a scenario in which coupling of SiR-carboxyl deriva-
tives to protein tags favours formation of the fluorescent zwitterion,
whereas aggregation of unreacted dye or its unspecific binding to
hydrophobic surfaces favours formation of the non-fluorescent
spirolactone, which thereby greatly facilitates imaging.
SiR-carboxyl derivatives thus permit imaging experiments with
high sensitivity at a wavelength that shows little cellular autofluores-
cence and phototoxicity. An experiment that epitomizes the differ-
ent attractive features of SiR-SNAP is a long-term imaging
experiment in which cells that express SNAP fusions are grown in
the presence of the probe and are imaged continuously over 48
hours (Supplementary Movie S1).
Another interesting application of SiR-carboxyl derivatives is
their use for the labelling of intracellular proteins in more
complex biological samples. We therefore investigated labelling of
SNAP-tag in cortical neurons in rat-brain sections. For these exper-
iments, SNAP-tag was expressed in cortical neurons through
in utero electroporation of plasmids that encoded SNAP and GFP
as markers (Fig. 3a). Three days after electroporation, brain sections
were prepared, cultured and then incubated with SiR-SNAP. After
removal of the excess SiR-SNAP through washing, the sample
was fixed with paraformaldehyde and imaged (Fig. 3b). The
images show specific labelling of cortical neurons that express
SNAP-tag with no significant background fluorescence.
Live-cell super-resolution microscopy using SiR-SNAP. We also
investigated the use of SiR-SNAP for live-cell super-resolution
microscopy or nanoscopy of biological structures. All nanoscopy
approaches rely on switching between the fluorescent and dark
states of dyes
17
; thus, in such experiments, the photophysical
properties of the fluorophore critically affect the attainable
resolution
18
. Furthermore, ideally nanoscopy of biological
structures is performed on living cells, as this avoids the
introduction of structural artefacts during fixation of the cells and
permits characterization of the dynamic processes
19
.
One approach to nanoscopy relies on the stochastic conversion
of isolated dyes into a fluorescent state, which thereby allows the
subsequent reconstruction of super-resolved images from the pos-
itions of precisely localized single molecules
20,21
. Stochastic blinking
of conventional fluorophores can be achieved by transiently trans-
forming them into long-lived dark states, as realized in GSDIM
(ground-state depletion followed by individual molecule return)
22
,
direct stochastic optical reconstruction microscopy ((d)STORM)
23
and blinking microscopy
24
. To evaluate the potential of
SiR-carboxyl derivatives for nanoscopy based on their stochastic
blinking (dubbed as GSDIM/STORM here), we performed proof-
of-principle experiments with a SNAP-tag fusion protein of
histone H2B. H2B-SNAP was expressed in U2OS cells, incubated
with SiR-SNAP and imaged with a 640 nm laser power of
1kWcm
22
. Under these conditions, a large percentage of the
dye was maintained in a dark state, and a sparse population in the
fluorescent state emitted, on average, 630 detectable photons,
which allowed a high localization precision in each frame (Fig. 4).
Blinking was very stable, with the number of molecules quickly sta-
bilizing as the bleaching and recovery rates equilibrated
(Supplementary Fig. S6). This allowed us to localize 140,000 mol-
ecules in 10,000 frames, at a frame rate of 50 Hz, to yield a super-
resolved image in just over three minutes. The dye’s excellent photo-
physical properties permit GSDIM/STORM imaging to be repeated
several times, as illustrated in Fig. 4c, where the same cell is imaged
twice with a 10 min interval. Furthermore, SiR-labelled proteins can
also be utilized for GSDIM/STORM on fixed cells (Supplementary
Fig. S7). These experiments demonstrate that SiR-carboxyl deriva-
tives are superior probes for live-cell GSDIM/STORM.
The excellent fluorescence properties of SiR-labelled proteins
also make them attractive candidates for stimulated emission
a
c
b
d
CLIP-H2B CLIP-Cox8A
SNAP-actin Halo-actin
Figure 2 | Three-colour confocal fluorescence microscopy of the tagged proteins. a–d,SNAP(redinc), CLIP (red in a,b) and Halo-tagged proteins
(red in d) in living HeLa cells expressing EGFP-a-tubulin (green) and H2B-mCherry (blue)
16
. The characteristic staining of the fusion proteins demonstrates
the suitability of SiR-carboxyl derivatives for live-cell imaging. Z-stacks of images were deconvolved using the Huygens Essentials package and presented as
MIPs. Scale bar, 10
m
m.
ARTICLES
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1546
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry134
© 2013 Macmillan Publishers Limited. All rights reserved.
–
–
+
–
–
–
–
–
–
+
+
+
+
+
GFP SNAP
Hoechst
GFP SNAP Hoechst
200 m
50 m
Merge
Merge
b
a
c
Micropipette
Electrodes
DNA injection site
+
+
Figure 3 | Ex vivo labe lling of a rat brain with SiR-SNAP. a, Scheme of in utero electropora tion. Pl asmid DNA is injected into an E16 rat embryo in utero
through a micropipette and then electropora ted with electrodes. The red square corresponds to the region shown in (b). b, SNAP and GFP plas mids
(ratio 1:1) were introduced into a subset of neural progenitors at E16 by in utero electroporation. At E19, brains were sectioned and stained with SiR-SNAP and
Hoechst. Scale bar, 200
m
m. c, Images of the electroporated cortical neurons at a higher magnification (yellow box in b). The excellent overlap of the GFP
and SiR-SNAP signals demonstrates the specificity of the labelling. Scale bar, 50
m
m.
t
0
t
0
+ 0 min 1
a
b
c
cʹ
Figure 4 | Live-cell GSDIM/STORM imaging of nuclear localized H2B-SNAP-SiR. a, Wide-field image of H2B-SNAP-SiR does not allow the detection of
substructures. b, Single frame image showing the stochastic flu or escence (blinking) of individual molecules. c,GSDIM/S TORM images reconstructed from
10,000 raw images (c
′
was ta ken ten minutes after c). The enhancement in resolution permitted the detection of substructures. Scale bar, 5
m
M
(inset, 500 nm).
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1546
ARTICLES
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry 135
© 2013 Macmillan Publishers Limited. All rights reserved.
depletion (STED) microscopy
25
. In STED, the excitation laser is
overlaid by a second STED laser that features at least one zero-inten-
sity point and that restricts fluorescence emission, for example, only
to the very focal centre. Scanning of such a reduced observation
volume produces a fluorescence image with superior spatial resol-
ution. STED microscopy is well suited for (dynamic) imaging in
live cells
26
and even in vivo
27
. Apart from a fluorescent protein
with a limited fluorescence quantum yield and photostability
28
,
no appropriate near-infrared fluorophores for STED microscopy
of intracellular proteins in living cells are available currently.
Therefore we investigated the potential of SiR-SNAP for live-cell
STED microscopy of the centrosomal protein Cep41. The
0 200 400 600 800 1,000
0.4
0.6
0.8
1.0
Confocal
STED FWHM = 77 ± 3 nm
FWHM = 248 ± 8 nm
Distance (nm)
~190 nm
0 200 400 600 800 1,000
0.0
0.2
0.4
0.6
0.8
1.0
Distance (nm)
Confocal FWHM = 350 ± 8 nm
STED FWHM = 85 ± 12 nm
SNAP-Cep41/Hoechst 33342
SNAP-Cep41SNAP-Cep41
Confocal STED
Distal end
Proximal end
~200 nm
~400 nm
Centriole
Procentriole
Confocal STED
Cep41-SNAPCep41-SNAP
Normalized averaged
fluorescence intensity
0.2
Normalized fluorescence intensity
a
c
d
b
Figure 5 | Confocal an d STED imag ing of Cep41 protein loc aliza tion in living U2OS ce lls. a, Schema tic presentation of the centrosome structure. b,Confocal
two-colour imaging of SNAP-Cep41 (red)-expr essing cells stained with SiR-SNAP. Nuclear DNA was stained with Hoechst 33342 (blue). Scale bar, 10
m
m.
c, Comparison of confocal (left) and STED microscopy (middle) images of Cep41-SNAP bound to microtubules, along with an intensity line profile (righ t)
obtained by aver aging the profiles of seven differ ent microtubule sections in the image. Scale bar, 500 nm. d, Comparison of confocal (left) and STED
(middle) microscopy images of SNAP-Cep41 localiz ed at the centrosome with an intensity line profile (right) along the white dotted line mar ked in the
images. The full width at half maximum (FWHM) of the imag ed structures was obta ined by fitting fluorescence-intensity profiles to Gauss or Lorentz
distributions (OriginPro 8.1 , http://www.originlab.com/). Two separated Lorentz distributions are indica ted by gre y da shed lines for the STED profile fitting.
Distance between the peaks of the double Lorentz fitting was taken as the diameter of the structure. The diffuse signal visible in the top-le ft cor ner is the
second centriole, which is located outside the focal plane. A corresponding two-colour im age of SNAP -C ep41 and the centrosomal marker GFP-Centrin2 is
presented in Supplementary Fig. S8. Scale bar, 500 nm. Num bers are presented as the fitted value+standard error of the fit.
ARTICLES
NATURE CHEMISTRY DOI: 10.1038/NCHEM.1546
NATURE CHEMISTRY | VOL 5 | FEBRUARY 2013 | www.nature.com/naturechemistry136
© 2013 Macmillan Publishers Limited. All rights reserved.