ARTICLE · super-resolution, the number of switching cycles afforded by the fluorophore assumes a vital role. Because they are able to generate an image with a single on–off cycle3,6,21,

ARTICLEdoi:10.1038/nature10497

Diffraction-unlimited all-optical imagingand writing with a photochromic GFPTim Grotjohann1*, Ilaria Testa1*, Marcel Leutenegger1*, Hannes Bock1, Nicolai T. Urban1, Flavie Lavoie-Cardinal1, Katrin I. Willig1,Christian Eggeling1, Stefan Jakobs1,2 & Stefan W. Hell1

Lens-based optical microscopy failed to discern fluorescent features closer than 200 nm for decades, but the recentbreaking of the diffraction resolution barrier by sequentially switching the fluorescence capability of adjacent featureson and off is making nanoscale imaging routine. Reported fluorescence nanoscopy variants switch these features eitherwith intense beams at defined positions or randomly, molecule by molecule. Here we demonstrate an optical nanoscopythat records raw data images from living cells and tissues with low levels of light. This advance has been facilitated by thegeneration of reversibly switchable enhanced green fluorescent protein (rsEGFP), a fluorescent protein that can bereversibly photoswitched more than a thousand times. Distributions of functional rsEGFP-fusion proteins in livingbacteria and mammalian cells are imaged at ,40-nanometre resolution. Dendritic spines in living brain slices aresuper-resolved with about a million times lower light intensities than before. The reversible switching also enablesall-optical writing of features with subdiffraction size and spacings, which can be used for data storage.

In a fluorescence microscope, diffraction prevents (excitation) lightbeing focused more sharply than l/(2NA), with l being the wavelengthof light and NA the numerical aperture of the lens. Thus, as they areilluminated together, features residing any closer together than this dis-tance also fluoresce together and appear in the image as a single blur. Thediffraction resolution barrier can be overcome by forcing such nearbyfeatures to fluoresce sequentially, but this strategy clearly requires amechanism for keeping fluorophores that are exposed to excitation lightnon-fluorescent1–3.

In stimulated emission depletion (STED) microscopy1,4, this isaccomplished by the so-called STED beam, which turns the fluorescencecapability of fluorophores off by a photon-induced de-excitation.Because at least a single de-exciting photon must be available withinthe lifetime (t < 1–5 ns) of the fluorescent molecular state, the intensityof the focal STED beam must exceed the threshold Is 5 Ct21 with Caccounting for the probability of a STED beam photon to interact withthe fluorophore1,4. The STED beam, usually formed as a doughnut over-laid with the excitation beam, features a central point of zero intensity atwhich the fluorophores can still assume the fluorescent state. As thispoint can be positioned with arbitrary precision in space, the coordinateof the emitting (on-state) fluorophores is known at any instant: it is theposition of zero intensity3,5,6 and its immediate vicinity, where the STEDbeam is still weaker than Is. The diameter of this area is given byd < l/[2NA3 (1 1 Im/Is)

1/2], with Im (typically ? Is) denoting theintensity at the doughnut crest. Hence, features that are (just slightly)more apart than d= l/(2NA) cannot fluoresce at the same time evenwhen simultaneously illuminated by excitation light6. Scanning thebeams across the sample and recording the fluorescence yields imagesof subdiffraction resolution d automatically and irrespective of the fluor-ophore concentration in the sample.

De-excitation by stimulated emission is the most basic and generalmechanism for modulating the fluorescence ability of a molecule.However, by requiring light intensities .Is < 1–10 MW cm22, attain-ing high resolutions by this mechanism necessitates large Im values. For

example, d , 40 nm typically entails Im 5 100–500 MW cm22 (ref. 6).Although intensities of this order have been demonstrated to be live-cell compatible4,7–10, all-optical nanoscopy methods operating at fun-damentally lower light levels are highly in demand2,5,11–13, because theyallow larger fields of view5,14 and can avoid photodamage.

A route to low light level operation is to replace STED with afluorescence switching mechanism having a lower threshold Is (refs2, 5, 11–13). Following the equation for Is, this can be realized byexploiting transitions between fluorophore states of longer lifetimet? 1 ms (refs 2, 5, 11). Hence, it has been suggested that fluorescencecan be switched by transferring the fluorophores transiently to ageneric metastable dark (triplet) state of t < 1023–100 ms (refs 2,15). A more attractive option is to use fluorophores that can be expli-citly ‘photoswitched’5,11, for example, by photoisomerization. Hence,in 2003 it was proposed to implement a STED-like microscope withSTED being replaced by a reversible on–off switch as encountered inorganic photochromic fluorophores and reversibly photoswitchablefluorescent proteins (RSFPs)5,11.

In fact, this strategy is even more general because any reversible trans-ition between a signalling and a non-signalling state can be used forbreaking the diffraction barrier15. Therefore, all concepts that switchthe fluorescence capability of molecules at sample coordinates predefinedby patterns of light have been generalized under the name RESOLFT6,12,which stands for reversible saturable optical (fluorescence) transitionbetween two states. Note that a photoswitch is a perfect saturabletransition. Concomitantly, the concept was extended to subdiffractionwriting11,13 and data storage, in which case the on-state is a reactivestate from which the molecule can be made permanent whereas the off-state serves as a temporary ‘mask’ defining the structure to be written.

Super-resolution by switching RSFPs was shown in 200516, but thisstudy relied on asFP59517, a tetrameric protein with low fluorescencequantum yield. Moreover, when translating the light pattern acrossthe sample, the proteins faded after a few cycles, implying that featuresthat had been turned off could not be turned on again in order to be

1Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, Am Fassberg 11, 37077 Gottingen, Germany. 2University of Gottingen Medical School, Robert-Koch-Str. 40, 37075Gottingen, Germany.*These authors contributed equally to this work.

2 0 4 | N A T U R E | V O L 4 7 8 | 1 3 O C T O B E R 2 0 1 1

Macmillan Publishers Limited. All rights reserved©2011

www.nature.com/doifinder/10.1038/nature10497

read out. Biological imaging therefore remained unviable18. Otherstudies using a variant of the RSFP called dronpa19 faced the samechallenge20. As a rule of thumb, an m-fold resolution improvementalong a certain direction requires ,m switching cycles, meaning thatm 5 10 along the x- and y-axes entails ,m2 5 100 cycles, whereas,1,000 cycles are required for x, y and z (ref. 6). Thus, for RESOLFTsuper-resolution, the number of switching cycles afforded by thefluorophore assumes a vital role.

Because they are able to generate an image with a single on–offcycle3,6,21, the super-resolution concepts called (F)PALM22,23 andSTORM21,24, which have emerged in the interim, have successfullyharnessed the switching between metastable states for gaining sub-diffraction resolution. However, these methods rely on the imagingand computation-aided localization of individual fluorophoresamidst the scattering and autofluorescence background common in(living) cells and tissues. Moreover, rapid localization of a sufficientlylarge number of fluorophores requires the excitation light to beintense22,23. In contrast, a RESOLFT approach is able to instantlyrecord the emission from all fluorophores attached to the nanosizedfeature of interest6, and can be easily combined with confocal micro-scopy for three-dimensional imaging and background suppression.Yet again, because all RSFPs, conventional fluorescent proteins25 andphotochromic rhodamines26 seemed unsuitable (Supplementary Fig. 1),an all-optical nanoscopy approach operating at low light levels appearedunviable.

Similarly, although STED/RESOLFT-inspired optical writing withphotochromic compounds has been shown to yield structures ,l/(2NA), writing such structures with spacings ,l/(2NA) remainedchallenging27–30, again the impediment being the requirement ofmany on–off cycles before the structure is made permanent. Herewe introduce a RSFP enabling both low-light-level all-optical nano-scopy of living cells and tissues, and far-field optical writing andreading of patterns of subdiffraction size and density.

Generating a reversibly switchable GFPAll fluorescent proteins have a similar fold, namely an 11-strandedb-barrel with a central helix containing the chromophore, which istypically in a cis-configuration31. Light-driven switching of RSFPsgenerally involves an isomerization of the chromophore, frequentlycoupled with a change of its protonation state32–36. We started fromEGFP37 and identified, using its X-ray structure31, amino acid residuesthe exchange of which was expected to facilitate isomerization. Weexpressed numerous EGFP variants in Escherichia coli and screenedfor colonies expressing an RSFP with an automated microscope. Tothis end, we alternated site-directed and error-prone mutagenesiswhile maintaining the key amino acids of EGFP (that is, F64L andS65T)37; we concomitantly introduced A206K to ensure that the proteinremained a monomer38.

The amino acid exchange Q69L was sufficient to makeEGFP(A206K) reversibly switchable, but the resulting on–off contrastwas low. Although it makes the protein switchable39, we avoided themutation E222Q because it seemed to reduce the number of cycles.After analysing ,30,000 clones, we identified EGFP(Q69L/V150A/V163S/S205N/A206K) (Supplementary Fig. 2) that could be reversiblyswitched on at l 5 405 nm and off at 491 nm, and named it reversiblyswitchable EGFP (rsEGFP).

At equilibrium, rsEGFP adopts a bright on-state (fluorescencequantum yield WFL 5 0.36; extinction coefficient e 5 47,000 M21 cm21

(Supplementary Table 1)). In the on-state, rsEGFP exhibits a singleabsorption band peaking at 491 nm (Fig. 1a), corresponding to theionized state of the phenolic hydroxyl of the chromophore40. The pKa

of the chromophore is 6.5 (Supplementary Fig. 3). Absorption at 490 nmyields fluorescence peaking at 510 nm and, in a competing process,switches rsEGFP off (Figs 1a–c). Prolonged irradiation of a pH 7.5 solu-tion of purified rsEGFP at ,490 nm reduces the rsEGFP fluorescence to1–2% of its initial value. The off-state exhibits a single absorption band

at 396 nm, corresponding to the neutral state of the chromophore(Fig. 1b). Excitation at this band switches the protein back to the on-state. At room temperature rsEGFP converts spontaneously from theoff- into the on-state with a half-time of ,23 min (Fig. 1d).

We compared the properties of rsEGFP with that of the well-knownRSFP dronpa19. With the proteins embedded in a 12.5% polyacrylamide(PAA) layer and using light of 491 nm (0.6 kW cm22) and 405 nm(2 kW cm22), a complete on–off cycle took 250 ms for dronpa and20 ms for rsEGFP (Fig. 1c). Dronpa went through ,10 cycles beforeits fluorescence was reduced to 50%, whereas rsEGFP went through,1,200 cycles under the same conditions (Fig. 1e). To compare bleach-ing, dronpa and rsEGFP were kept in the on-state by continuous irra-diation at 405 nm (1 kW cm22) while fluorescence was generated byirradiation at 491 nm (3 kW cm22). Whereas dronpa fluorescence wasreduced to 50% within tK< 30 s, for rsEGFP we measured tK < 800 s(Fig. 1f). The rsEGFP chromophore maturated with a half-time of ,3 hat 37 uC (Supplementary Fig. 4). The protein behaved as a monomer invitro (Supplementary Fig. 5), could be fused to various proteins,including a-tubulin and histone H2B (Supplementary Fig. 6), andwas repeatedly switchable in living cells (Supplementary Fig. 7).

Wavelength (nm)350 400 450 500 550

0.2

0.4

0.6

0

Ab

so

rptio

n

0.1 0.2 0.3 0.4 0.5Time (s)

00

0.5

10

0.5

1

Switching cycle100 200150500

0.5

1

600200 40000

0.5

1

Time (s)

rsEGFPDronpa

rsEGFPDronpa

Equilibrium1 min2 min3 min4 min5 min 7 min 10 min 15 min (off-state)

Flu

ore

scen

ce (n

orm

.)

Flu

ore

scen

ce (n

orm

.)

Wavelength (nm)400 450 500 550 650600

AbsorptionExcitationEmission

0.5

1

No

rmaliz

ed

a b

c

e f

0.5

0.7

0.3

0.1

Time (s)1,000 2,000 3,000

t1/2 ~23 min

d

00

rsEGFPDronpa

Flu

ore

scence (no

rm.)

Flu

ore

scence (no

rm.)

Figure 1 | Properties of rsEGFP. a, Absorption (red dashed line), excitation(solid black line) and fluorescence (dotted green line) spectrum of rsEGFP inthe fluorescent equilibrium state at pH 7.5. b, Absorption spectra obtained atdifferent time points during irradiation with 488-nm light. c, Switching curvesof dronpa (blue) and rsEGFP (red) immobilized in PAA using the sameintensities. Switching was performed by alternating irradiation at 405 nm(2 kW cm22) and at 491 nm (0.6 kW cm22). The duration of off-switching at491 nm was chosen such that the fluorescence reached a minimum; irradiationwith 405 nm was chosen so that the proteins were fully switched. d, Relaxationof rsEGFP embedded in PAA from the off-state into the fluorescent equilibriumstate at 22 uC. The black line is a stretched exponential fit with a stretching factorof ,0.6 accounting for inhomogeneous spectral broadening or the involvementof multiple dark states. e, Fluorescence per switching cycle normalized to theinitial fluorescence, with the same light intensities and switching durations as inc. f, Photobleaching: rsEGFP and dronpa embedded in a PAA layer were kept intheir on-states by continuous irradiation at 405 nm (1 kW cm22), whilefluorescence was probed by irradiation at 491 nm (3 kW cm22).

ARTICLE RESEARCH

1 3 O C T O B E R 2 0 1 1 | V O L 4 7 8 | N A T U R E | 2 0 5


Rewritable data storageTo analyse whether immobilized rsEGFP could be used for repeatedshort-term data storage41, we coated a microscope slide with a ,1-mmthin layer of rsEGFP (,0.03 mM) in PAA. Switching and reading byillumination at 405 nm and 491 nm in a scanning confocal set-up pro-vided an on–off contrast of ,50:1. We translated the text of 25 Grimm’sfairy stories (http://www.gutenberg.org/files/11027/11027.txt) into7-bit binary ASCII code (‘0’: off; ‘1’: on) and wrote and read the,270,000 letters into a 17mm 3 17mm region in 6,596 frames, eachcomprising 41 letters (287 bits) (Fig. 2). Individual bits were ,0.5mm indiameter with 1mm centre-to-centre spacing, corresponding to a DVDstorage density. Discriminating ‘0’ from ‘1’ by a simple thresholdentailed 7 bit errors within the entire data set. After ,6,600 read/writecycles in the same region, the average fluorescence of the ‘1’ was reducedby ,35% (Supplementary Fig. 8). Hence, the same rsEGFP layer can beused for ,15,000 read/write processes.

RESOLFT nanoscopy of living samplesNext, we implemented a scanning confocal set-up with a 405 nm(ultraviolet) beam for switching the rsEGFP on, a 491 nm (blue) beamfor eliciting fluorescence, and a doughnut-shaped 491 nm beam forthe off-switching (Supplementary Fig. 9). We fused rsEGFP to theamino-terminus of the bacterial actin homologue MreB42 andexpressed the fusion protein in E. coli bacteria. Living bacteria onagar-coated slides were recorded by first irradiating each pixel for100 ms with ultraviolet light (1 kW cm22), thus activating most ofthe rsEGFP in the focal volume. Then the doughnut-shaped bluebeam (Im < 1 kW cm22) was applied for 10–20 ms to switch all thersEGFP molecules off, except those located within d/2 distance fromthe doughnut centre. Lastly, the rsEGFP fluorescence was read out for1–2 ms by the 491-nm beam (,1 kW cm22). The sequence wasrepeated for each sample pixel.

The double-helical cytoskeletal structure of rsEGFP–MreB is moreclearly revealed by RESOLFT than by its confocal counterpart

(Fig. 3a). The RESOLFT image of a typical filament showed a full-width half-maximum (FWHM) of ,70 nm. Because this valueseemed to be determined by the thickness of the filament itself, a moreaccurate upper limit for the resolution d is obtained by imaging thefiner keratin-19–rsEGFP intermediate filament network in livingmammalian cells (Fig. 3b, c). Line profiles from recorded data gaved , 40 nm corresponding to a 5–6-fold all-optical resolutionimprovement over confocal microscopy (Fig. 3c).

To investigate its applicability to living brain tissue, we locallyinjected viral particles carrying a lifeact–rsEGFP construct into acultured organotypical hippocampal brain slice. Lifeact is a 17-amino-acid-long peptide with high affinity to filamentous actin43.RESOLFT revealed fine morphological differences between the spinesprotruding from a dendrite (Fig. 3d). A profile through a spine neckshowed a FWHM of ,80 nm. Electron microscopy of similar samplesdemonstrated that this value is close to the actual size of the spinenecks themselves44, suggesting a resolution d substantially ,80 nm.Repeated imaging revealed dynamic changes over 5 min (Fig. 3d).Altogether, the resolution is comparable to that provided by STED

Frame 1 1,500 3,000

4,500 6,000

Frame 6,596

10

60

120

180

5 10 15

Binary code

Spot

0 0 0 0 0 0 0 01 1 1 1 1 1 1 1 1

Threshold

Mean c

ounts

18 μ

m

18 μm

Figure 2 | Rewritable data storage. The text of 25 Grimm’s fairy stories(ASCII code; 1.9 Mbits) consecutively written and read on a 17 3 17mm area ofa PAA layer containing rsEGFP, with bits written as spots (representativeframes shown). The white dots mark spots that were recognized as set bits (‘1’s).The graph shows an intensity profile along the indicated area, averaged overthree pixels along the y-axis. The blue line indicates the threshold used to assignread spots to ‘0’s or ‘1’s.

0

0 1,000 2,000 (nm)

500 1,000 1,500 (nm)

a Confocal 1

0

0.2

0.4

0.6

0.8

5000 (nm)

~80 nm

d Confocal I RESOLFT

Confocal II RESOLFT

Confocal III RESOLFT 1 2 3 4 5

III

III

Confocalb RESOLFT1

0

0.2

0.4

0.6

0.8

~70 nm

Confocal

c

20

40

60

Ph

oto

n c

ou

nts

200 40000

(nm)

RESOLFT

20

40

60

Ph

oto

n c

ou

nts

100 20000

(nm)300

~36 nm~41 nm1

1 2

2

RESOLFT

Figure 3 | RESOLFT nanoscopy of living cells. a, E. coli bacterium expressingrsEGFP–MreB: confocal (left) and corresponding RESOLFT (middle) image.b, Mammalian (PtK2) cell expressing keratin-19–rsEGFP imaged in the confocal(left) and the RESOLFT (middle) mode. a, b, Graphs show the normalizedfluorescence profiles between the two white markers with the white arrowheadindicating the direction (solid red, RESOLFT; dashed blue, confocal).c, RESOLFT image (left) of keratin-19–rsEGFP filaments in a PtK2 cell recordedwith a pixel size of 10 nm 3 10 nm; smoothed with a low-pass Gaussian filter of1.2 pixel width. Graphs 1 and 2 extracted from the image as indicated revealresolution d , 40 nm. d, Dendrite within a living organotypic hippocampal sliceexpressing lifeact–rsEGFP. Main image: confocal overview. I–III: three spines, asindicated on the main image, each imaged in the confocal (left) and theRESOLFT mode (right). Spine III was repeatedly imaged in the RESOLFT modewithin 5 min, demonstrating the changes over time. Graph: normalized profileacross a spine neck as imaged in the RESOLFT (solid red) or the confocal mode(dashed blue) between the two white markers. Scale bars, 1mm.

RESEARCH ARTICLE



http://www.gutenberg.org/files/11027/11027.txt

on similar structures10, but here it is obtained with light intensitieslower by about a million times.

RESOLFT optical data storageFor investigating subdiffraction resolution writing, an rsEGFP layer wasprepared as previously outlined. The writing entailed (1) an ultravioletbeam (405 nm, 1 kW cm22) applied for 100ms to switch rsEGFP on, (2)a 2-ms break for equilibration, (3) a doughnut-shaped blue beam(491 nm, 0.5 kW cm22) lasting 20 ms confining the on-state withind/2 around the doughnut centre, and (4) an ,2-ms 532 nm beam(,900 kW cm22) for transferring on-state rsEGFP to a permanent off(bleached) state (Fig. 4a) (Supplementary Fig. 10a). Lastly, the rsEGFPmolecules located outside this region were switched back on, which iscritical for writing another feature within subdiffraction proximity.

We wrote nine patterns of 3 3 3 bit fields in an rsEGFP layer, with250 nm centre-to-centre separation between individual bits (Fig. 4b),both in the conventional and in the RESOLFT mode. Whereas con-ventional writing and/or confocal reading blurred the data, the bitswere fully discernible when both writing and reading were performedby RESOLFT. We wrote and read the data down to distances of 200 nmbetween the individual bits (Supplementary Fig. 10b). Hence thisscheme allowed storing and reading out bits ,4 times more denselythan by regular focusing. The structures could be read 5–10 times.

Discussion and conclusionThe many-switching cycles afforded by the fluorescence proteinrsEGFP reported here has facilitated live-cell RESOLFT microscopy,a super-resolution microscopy that is similar to STED microscopy inusability but operates at ,106 times lower levels of light. Multiphoton-induced optical damage45 can therefore be virtually excluded. Thefundamental reduction in optical intensity required for the on–off

switching stems from the fact that the fluorescence capability of themolecule is not modulated by disallowing the population of its nano-second fluorescent state, but rather by toggling it between two long-lived ground states, one in which the fluorophore remains dark whenexposed to the excitation light.

RESOLFT is readily combined with confocal imaging, whichincreases its use in scattering living samples. In fact, the imaging ofneuronal spines in living organotypical brain slices testifies this poten-tial. Although the recording time reported here is still of the order ofmost other super-resolution techniques3,6,21 and slower than the fastestbiological STED recordings7, by gathering the signal from typicallymany molecules located at predefined positions, RESOLFT has all theprerequisites for fast imaging. Scanning with arrays of doughnuts orzero-intensity lines (so-called structured illumination5,14,18,46) anddetection by a camera will reduce the number of scanning steps requiredto cover large fields of view and facilitate low-intensity video-rateimaging. The maximum recording speed is determined by the time ittakes to establish the disparity of (on–off) states in space, that is, by theswitching kinetics, which probably can be improved by further muta-genesis. Note that the switching is not restricted to changes in brightness(on–off) only. Other reversible transitions between disparate states mayalso prove suitable for RESOLFT imaging, such as states yielding differ-ences in emission wavelengths, lifetime or polarization.

Photoswitching between long-lived states also poses challenges,because in the process the molecule can assume transient (dark)states, such as triplet states, which depend on the molecular micro-environment. In this regard, STED maintains a unique advantagebecause it entails just basic optical transitions between the groundand the fluorescent state; no atom relocation, spin flip or change inchemical bond is required to switch the fluorescence capability of themolecule — just light. Therefore, switching fluorescence by STED isnearly universal and instantaneous.

The switching stamina of rsEGFP also enabled writing and readingof patterns of both subdiffraction size and spacing d, which has so farbeen difficult for direct far-field optical writing. In our study, thesmallest obtainable structure size was co-determined by the fact thatthe 532-nm light moderately bleached the off-state proteins too, thusreducing the writing contrast. However, this initial demonstrationshould spur on new advancements in this field, because currentnanowriting efforts are dominated by concepts that resort to muchshorter wavelengths of electromagnetic radiation at which focusingbecomes exceedingly difficult. In fact, RESOLFT and related conceptsare unique for creating materials that are nanostructured in threedimensions30. To maximize the resolution along the optical axis (z),RESOLFT imaging and writing can also be combined with 4Pi micro-scopy47, in which case three-dimensional resolution of ,10 nmshould become possible at ultralow light levels.

The resolution demonstrated here is similar or even exceeds theresolution attained until now by STED in living cells8,10. Although inboth methods the resolution can be continually increased by increasingIm/Is, in STED microscopy this strategy will reach practical limits dueto the intensities required. Using a threshold intensity Is that is lower bymany orders of magnitude, switching between long-lived states over-comes these limits and, as we have demonstrated here, offers a pathwayto lens-based optical imaging and writing at molecular dimensions.

METHODS SUMMARYProtein generation and screening. Site-directed mutagenesis was performedwith the QuikChange Site Directed Mutagenesis Kit (Stratagene) or a multiple-site approach using several degenerative primers. The proteins were expressedfrom the high-copy expression vector pQE31 (Qiagen) and expressed in E. coli.Viral transfection. A modified Semliki Forest Virus containing the pSCA-Lifeact-rsEGFP vector construct was injected into the slice cultures using a patchpipette. Imaging was performed within 16–48 h after incubation.Data storage. A layer containing immobilized rsEGFP was prepared by mixing24.5 ml purified proteins (0.09 mM) with 17.5 ml Tris-HCl pH 7.5, 30ml acryla-mide (Rotiphorese Gel 30, Roth), 0.75 ml 10% ammonium persulfate and 1ml 10%

RESOLFT

writing x

Scanning

Diffraction-

limited

writing

Scanning

491 nm 532 nm 405 nm405 nm

States Permanent offOff On

28 832

bDiffraction-limited write + read

4,702123

a

RESOLFT write + read

0

(μm)

xxxx

x x x x

0 0.50.25 0.75

‘1’ ‘0’

Δx

Δx

Fluorescence

(norm.)1

0.5

1

Figure 4 | Subdiffraction-resolution writing and reading using rsEGFP andvisible light. a, Top, schematic of RESOLFT writing: rsEGFP molecules areswitched off at 491 nm using a doughnut-shaped focal intensity (dashed blueline) so that the on-state is confined to a subdiffraction-sized region around thedoughnut centre. Subsequent irradiation with 532-nm light makes the on-statemolecules permanent by bleaching. Irradiation at 405 nm switches the off-statemolecules back into the on-state, allowing the writing of another feature insubdiffraction proximity. Bottom, schematic of diffraction-limited writing.b, Conventional (left) and subdiffraction RESOLFT (middle) joined writingand reading in a layer of immobilized rsEGFP. The outlines of thecorresponding 3 3 3 bit patterns were identical. The distance between twobleached spots was 250 nm in each case. Right, normalized line profiles of thefluorescence signal between the two arrows (solid red, RESOLFT; dashed blue,confocal). Scale bars, 1mm.

ARTICLE RESEARCH

1 3 O C T O B E R 2 0 1 1 | V O L 4 7 8 | N A T U R E | 2 0 7


TEMED. About 10 ml of this solution was placed on a glass slide and a cover slipwas pressed onto the sample to attain a thin layer. Custom MATLAB (TheMathWorks) programs allowed automated generation of the voltages and signalsfor moving the sample and for generating the desired laser pulses. Images werealso taken using the software Imspector (http://www.imspector.de).RESOLFT set-up. We implemented a home-built confocal microscope with anormally focused beam for generating fluorescence plus a doughnut-shapedbeam for switching rsEGFP off (both at 491 nm wavelength). The beams werecircularly polarized, superimposed in the focal plane and applied sequentially.The 405 nm beam for switching rsEGFP on was also circularly polarized. Thefluorescence emitted between 500–560 nm was imaged on the opening of a multi-mode fibre and detected by a counting avalanche photodiode. The same set-upwas used for writing, which was most specific at 532 nm.

Full details of methods used are available in Supplementary Information.

Received 23 May; accepted 24 August 2011.

Published online 11 September 2011.

1. Hell, S. W. & Wichmann, J. Breaking the diffraction resolution limit by stimulatedemission: stimulated-emission-depletion fluorescence microscopy. Opt. Lett. 19,780–782 (1994).

2. Hell, S. W. & Kroug, M. Ground-state depletion fluorescence microscopy, a conceptfor breaking the diffraction resolution limit. Appl. Phys. B 60, 495–497 (1995).

3. Hell, S. W. Far-field optical nanoscopy. Science 316, 1153–1158 (2007).4. Klar, T. A., Jakobs, S., Dyba, M., Egner, A. & Hell, S. W. Fluorescence microscopywith

diffraction resolution barrier broken by stimulated emission. Proc. Natl Acad. Sci.USA 97, 8206–8210 (2000).

5. Hell, S. W. Toward fluorescence nanoscopy. Nature Biotechnol. 21, 1347–1355(2003).

6. Hell, S. W. Microscopy and its focal switch. Nature Methods 6, 24–32 (2009).7. Westphal, V. et al. Video-rate far-field optical nanoscopy dissects synaptic vesicle

movement. Science 320, 246–249 (2008).8. Hein, B., Willig, K. I. & Hell, S. W. Stimulated emission depletion (STED) nanoscopy

of a fluorescent protein-labeled organelle inside a living cell. Proc. Natl Acad. Sci.USA 105, 14271–14276 (2008).

9. Eggeling, C. et al. Direct observation of the nanoscale dynamics of membranelipids in a living cell. Nature 457, 1159–1162 (2009).

10. Nagerl, U. V., Willig, K. I., Hein, B., Hell, S. W. & Bonhoeffer, T. Live-cell imaging ofdendritic spines by STED microscopy. Proc. Natl Acad. Sci. USA 105,18982–18987 (2008).

11. Hell, S. W., Jakobs, S. & Kastrup, L. Imaging and writing at the nanoscale withfocused visible light through saturable optical transitions. Appl. Phys., A Mater. Sci.Process. 77, 859–860 (2003).

12. Hell, S. W., Dyba, M. & Jakobs, S. Concepts for nanoscale resolution in fluorescencemicroscopy. Curr. Opin. Neurobiol. 14, 599–609 (2004).

13. Hell, S.W.Strategy for far-fieldoptical imaging andwriting without diffraction limit.Phys. Lett. 326, 140–145 (2004).

14. Gustafsson, M. G. L. Nonlinear structured-illumination microscopy: wide-fieldfluorescence imaging with theoretically unlimited resolution. Proc. Natl Acad. Sci.USA 102, 13081–13086 (2005).

15. Hell, S. W. in Topics in Fluorescence Spectroscopy Vol. 5 (ed. Lakowicz, J. R.)361–422 (Plenum, 1997).

16. Hofmann, M., Eggeling, C., Jakobs, S. & Hell, S. W. Breaking the diffraction barrier influorescence microscopy at low light intensities by using reversiblyphotoswitchable proteins. Proc. Natl Acad. Sci. USA 102, 17565–17569 (2005).

17. Lukyanov, K. A. et al. Natural animal coloration can be determined by anonfluorescent green fluorescent protein homolog. J. Biol. Chem. 275,25879–25882 (2000).

18. Schwentker, M. A. et al. Wide-field subdiffraction RESOLFT microscopy usingfluorescent protein photoswitching. Microsc. Res. Tech. 70, 269–280 (2007).

19. Ando, R., Mizuno, H. & Miyawaki, A. Regulated fast nucleocytoplasmic shuttlingobserved by reversible protein highlighting. Science 306, 1370–1373 (2004).

20. Dedecker, P. et al. Subdiffraction imaging through the selective donut-modedepletion of thermally stable photoswitchable fluorophores: numerical analysisand application to the fluorescent protein dronpa. J. Am. Chem. Soc. 129,16132–16141 (2007).

21. Huang, B., Babcock, H. & Zhuang, X. Breaking the diffraction barrier: super-resolution imaging of cells. Cell 143, 1047–1058 (2010).

22. Betzig, E. et al. Imaging intracellular fluorescent proteins at nanometer resolution.Science 313, 1642–1645 (2006).

23. Hess, S. T., Girirajan, T. P. K. & Mason, M. D. Ultra-high resolution imaging byfluorescence photoactivation localization microscopy. Biophys. J. 91, 4258–4272(2006).

24. Rust, M. J., Bates, M. & Zhuang, X. Sub-diffraction-limit imaging by stochasticoptical reconstruction microscopy (STORM). Nature Methods 3, 793–796 (2006).

25. Dickson, R. M., Cubitt, A. B., Tsien, R. Y. & Moerner, W. E. On/off blinking andswitching behaviour of single molecules of green fluorescent protein. Nature 388,355–358 (1997).

26. Bossi, M., Foelling, J., Dyba, M., Westphal, V. & Hell, S. W. Breaking the diffractionresolutionbarrier in far-field microscopy bymolecular optical bistability. N. J. Phys.8, 275 (2006).

27. Scott, T. F., Kowalski, B. A., Sullivan, A. C., Bowman, C. N. & McLeod, R. R. Two-colorsingle-photon photoinitiation and photoinhibition for subdiffractionphotolithography. Science 324, 913–917 (2009).

28. Li, L., Gattass, R. R., Gershgoren, E., Hwang, H. & Fourkas, J. T. Achieving l/20resolution by one-color initiation and deactivation of polymerization. Science 324,910–913 (2009).

29. Andrew, T. L., Tsai, H. Y. & Menon, R. Confining light to deep subwavelengthdimensions to enable optical nanopatterning. Science 324, 917–921 (2009).

30. Fischer, J., Freymann, G. & Wegener, M. The materials challenge in diffraction-unlimited direct-laser-writing optical lithography. Adv. Mater. 22, 3578–3582(2010).

31. Ormo, M. et al. Crystal structure of the Aequorea victoria green fluorescent protein.Science 273, 1392–1395 (1996).

32. Andresen, M. et al. Structure and mechanism of the reversible photoswitch of afluorescent protein. Proc. Natl Acad. Sci. USA 102, 13070–13074 (2005).

33. Andresen, M. et al. Structural basis for reversible photoswitching in Dronpa. Proc.Natl Acad. Sci. USA 104, 13005–13009 (2007).

34. Henderson, J. N., Ai, H. W., Campbell, R. E. & Remington, S. J. Structural basis forreversible photobleaching of a green fluorescent protein homologue. Proc. NatlAcad. Sci. USA 104, 6672–6677 (2007).

35. Adam, V. et al. Structural characterization of IrisFP, an optical highlighterundergoing multiple photo-induced transformations. Proc. Natl Acad. Sci. USA105, 18343–18348 (2008).

36. Brakemann, T. et al. Molecular basis of the light-driven switching of thephotochromic fluorescent protein Padron. J. Biol. Chem. 285, 14603–14609(2010).

37. Patterson, G. H., Knobel, S. M., Sharif, W. D., Kain, S. R. & Piston, D. W. Use of thegreen fluorescent protein and its mutants in quantitative fluorescencemicroscopy. Biophys. J. 73, 2782–2790 (1997).

38. Zacharias, D. A., Violin, J. D., Newton, A. C. & Tsien, R. Y. Partitioning of lipid-modified monomeric GFPs into membrane microdomains of live cells. Science296, 913–916 (2002).

39. Bizzarri, R. et al. Single amino acid replacement makes Aequorea victoriafluorescent proteins reversibly photoswitchable. J. Am. Chem. Soc. 132, 85–95(2010).

40. Tsien, R. Y. The green fluorescent protein. Annu. Rev. Biochem. 67, 509–544(1998).

41. Adam, V. et al. Data storage based on photochromic and photoconvertiblefluorescent proteins. J. Biotechnol. 149, 289–298 (2010).

42. Vats,P.&Rothfield, L.Duplicationandsegregationof theactin (MreB) cytoskeletonduring the prokaryotic cell cycle. Proc. Natl Acad. Sci. USA 104, 17795–17800(2007).

43. Riedl, J. et al. Lifeact: a versatile marker to visualize F-actin. Nature Methods 5,605–607 (2008).

44. Harris, K. M. & Kater, S. B. Dendritic spines: cellular specializations imparting bothstability and flexibility to synaptic function. Annu. Rev. Neurosci. 17, 341–371(1994).

45. Hopt, A. & Neher, E. Highly nonlinear photodamage in two-photon fluorescencemicroscopy. Biophys. J. 80, 2029–2036 (2001).

46. Heintzmann, R., Jovin, T. M. & Cremer, C. Saturated patterned excitationmicroscopy—a concept for optical resolution improvement. JOSA A 19,1599–1609 (2002).

47. Hell, S. W., Schmidt, R. & Egner, A. Diffraction-unlimited three-dimensional opticalnanoscopy with opposing lenses. Nature Photon. 3, 381–387 (2009).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank J. Jethwa for careful reading and M. Andresen,T. Brakemann, S. Lobermann, R. Schmitz-Salue and A. C. Stiel for discussions andsupport, as well as T. Gilat and F. Voss (MPI of Neurobiology, Munich) for help with theslice culture preparation and A. Schonle for adapting the software Imspector. We thankThe Project Gutenberg for making Grimm’s Fairy Tales available in electronic format,L. Rothfield (University of Connecticut Health Center) for providing the plasmid pLE7,R. Wedlich-Soldner (MPI of Biochemistry, Munich) for the lifeact–YFP construct andV. Stein (MPI of Neurobiology, Munich) for the virus protocol. This work was supportedby the Deutsche Forschungsgemeinschaft (DFG) through the DFG-Research Center forMolecular Physiology of the Brain (to S.J.) and by a Gottfried-Wilhelm-Leibniz prize ofthe DFG (to S.W.H.).

Author Contributions T.G., I.T., M.L., H.B., F.L.-C. performed research, I.T., M.L., T.G., H.B.,C.E. set up the microscopes, N.T.U., K.I.W. prepared samples, M.L., T.G., I.T., K.I.W., S.J.,S.W.H. analysed data, S.J., C.E., S.W.H. designed research. S.J., M.L., S.W.H. wrote thepaper. All authors discussed the data and commented on the manuscript.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of this article atwww.nature.com/nature. Correspondence and requests for materials should beaddressed to S.J. ([email protected]) or S.W.H. ([email protected]).

RESEARCH ARTICLE



http://www.imspector.de

www.nature.com/nature

www.nature.com/reprints

www.nature.com/nature

mailto:[email protected]


Diffraction-unlimited optical imaging and writing with a photochromic GFP

Tim Grotjohann1, Ilaria Testa1, Marcel Leutenegger1, Hannes Bock, Nicolai T Urban, Flavie

Lavoie-Cardinal, Katrin I Willig, Christian Eggeling, Stefan Jakobs2 & Stefan W Hell2

1These authors contributed equally 2To whom correspondence should be addressed. E-mail: [email protected] (SJ), [email protected] (SWH)

Mutagenesis

For site-directed mutagenesis, the QuikChange Site Directed Mutagenesis Kit (Stratagene, La

Jolla, CA) or a multiple-site mutagenesis approach using several degenerative primers were

used1. Error-prone random mutagenesis was essentially performed as described2. For

expression in E. coli, the coding sequences were cloned into the high-copy expression vector

pQE31 (Qiagen, Hilden, Germany).

Protein expression and purification

Proteins were expressed in the E. coli strain BL21-CP-RIL and purified by Ni-NTA affinity

chromatography (His SpinTrap, GE Healthcare, Freiburg, Germany), according to the

manufacturer’s instructions. The purified proteins were concentrated by ultrafiltration and

taken up in 100 mM Tris-HCl, 150 mM NaCl, pH 7.5.

Protein characterization

For the determination of the absorption, excitation and emission spectra of rsEGFP (Fig. 1a,

b), a protein solution (pH 7.5) was analyzed with a photospectrometer (Varian Cary 4000

UV/VIS, Varian, Palo Alto, CA, USA) and a fluorescence spectrometer (Varian Cary

Eclipse), respectively. To determine the fluorescence excitation spectra, the fluorescence was

recorded at 540 nm. For the emission spectra, rsEGFP was irradiated at 450 nm. The

fluorescence quantum yields and the molar extinction coefficients at the respective absorption

maximum were determined relative to the reported values of EGFP (quantum yield ΦFL =

WWW.NATURE.COM/NATURE | 1

SUPPLEMENTARY INFORMATIONdoi:10.1038/nature10497



0.60, molar extinction coefficient at 489 nm ε = 53,000 M-1cm-1)3. Irradiation-dependent

changes in the absorption were quantified by illuminating the protein solution in a cuvette

with a fiber coupled mercury lamp (Leica) equipped with a (488 ± 5) nm excitation filter. For

each measurement of the spectrum the irradiation was briefly interrupted.

For the measurement of the thermal relaxation from the off state into the fluorescent

equilibrium state, rsEGFP was embedded in a PAA layer. Using a custom built confocal

scanning microscope, the rsEGFP was switched on with 405 nm light in an area of 12×12 µm2

and subsequently the central 6×6 µm2 were turned off with 491 nm light. The level of thermal

relaxation from the off state into the fluorescent equilibrium state was determined by

irradiation with light of 491 nm after a defined time. For each data point the experiment was

repeated. The ratio of the average intensity in the central square (initially off) versus the

average intensity of the surrounding frame (always on) represents the fraction of thermally

relaxed rsEGFP. The intensity and time of the 491 nm light for probing of the thermally

relaxed fluorescence was kept very low to avoid switching-off during this period and thus to

avoid an overestimation of the thermal relaxation time.

Switching curves were recorded on rsEGFP or Dronpa embedded in PAA by alternate

irradiation with light of 405 nm (2 kW/cm2) and 491 nm (0.6 kW/cm2) using a 100× 1.4 NA

oil immersion objective lens (Leica). The duration of the on/off switching cycles was adjusted

individually for each protein by determining the minimum time to reach the

maximum/minimum in fluorescence (Fig. 1c, e). Photo-bleaching from the on state was

measured by constant irradiation with 405 nm (1 kW/cm2) and 491 nm (3 kW/cm2) laser light

(Fig. 1f).

For semi-native polyacrylamide gel electrophoresis, 20µg of each protein were taken

up in a concentrated sucrose solution to a final concentration of 10% (w/v) sucrose and loaded

on a 12.5 % polyacrylamide gel containing 0.1% sodiumdodecyl sulphate.

Size separation chromatography was performed with a Superdex 200 column

(Pharmacia, Uppsala, Schweden) and an Äkta prime plus system (GE Healthcare, Freiburg,

Germany). Prior to chromatography, proteins were concentrated to 3 mg/ml (100 mM Tris-

HCl, 150 mM NaCl, pH 7.5). The flow rat was set to 4 ml/min. Protein absorption was

monitored at 280nm. All measurements were performed at 4°C.

To determine the chromophore maturation time of rsEGFP, TOP10 E. coli cells

(Invitrogen) were transformed with a pBad-rsEGFP plasmid and grown overnight at 37 °C in

LB-Amp medium. The overnight culture was used to inoculate 200 ml LB-Amp growth

medium and grown to an OD600 of 0.5 at 37 °C. After adding arabinose to a final


SUPPLEMENTARY INFORMATIONRESEARCHdoi:10.1038/nature10497

concentration of 0.2 % to induce protein expression, bacteria were further incubated at 37 °C

for 4 hours. Finally, cells were opened up and pelleted. rsEGFP was purified immediately

from the supernatant using a His SpinTrap column (GE Healthcare). The proteins were

diluted in buffer (final concentration: 20 mM NaH2PO4, 500 mM NaCl, 30 mM imidazol, pH

7.5). Care was taken that all preparation steps took place at 4 °C. Finally, fluorescence

emission spectra of rsEGFP were taken at several time points using a fluorescence

spectrometer (Varian Cary Eclipse) while incubating the protein solution at 37°C.

Determination of the single-molecule brightness

The single-molecule brightness of rsEGFP, EGFP and Dronpa were determined using

fluorescence fluctuation spectroscopy, specifically fluorescence correlation spectroscopy

(FCS)4-6 and fluorescence intensity distribution analysis7, 8. Both methods analyze

characteristic fluctuations δF(t) in the fluorescence signal F(t) in time t about an average

value F(t) = <F(t)> + δF(t) by either calculating the second-order auto-correlation function

G(tc) (FCS, with correlation time tc) or by building up a frequency histogram P(n, ΔT) of

photon counts detected per time window ΔT (FIDA, with number of photons n). Fluctuations

in F arise for example from diffusion of the fluorescent proteins in and out of the confocal

detection volume or by transitions into and out of a dark state such as the triplet, other

metastable dark or the switch-off state.

FCS and FIDA data were analyzed using common theory8-10. As outlined in detail

previously11, the analysis most importantly resulted in two characteristic molecular

parameters of the fluorescent proteins: the single-molecule brightness (or count-rate per

particle) q and the observation time τobs. Without saturating the excitation, the brightness q ~

ΦFL ε scales with the fluorescence quantum yield ΦFL and the extinction coefficient ε

(compare Supplementary Table 1). For EGFP, the observation time τobs is given by its average

transit time through the focal spot, while for rsEGFP and Dronpa it is given by both the transit

time and – if faster - the average switch-off time11.

The fluorescence fluctuation data were recorded at the same microscope setup as was

used for RESOLFT imaging using only the 491 nm laser line, a water immersion objective

(60× UPLSAPO, NA 1.2, Olympus, Japan), two single-photon avalanche photodiodes

(SPCM-AQR-13-FC, Perkin Elmer Optoelectronics, Fremont, CA) coupled to a 50:50 fiber

splitter (FONT, Surrey, Canada) serving as the confocal pinhole (~1.4 times the Airy disc of

the image laser spot), and a hardware correlator (Flex02-01D, Correlator.com, NJ) for further



processing of the fluorescence counts for FCS and FIDA (ΔT = 62.5 µs). Data was recorded

for different powers of the 491 nm excitation laser and for the fluorescent proteins in aqueous

solution (100 mM Tris-HCl, 150 mM NaCl, pH 7.5).

The Supplementary Table 1 shows the single-molecule brightness values q for rsEGFP

and Dronpa relative to that of EGFP (q = 1). Dronpa is by a factor of 1.5 brighter and rsEGFP

by a factor of 0.6 dimmer than EGFP, which corresponds to the relative products q ~ ΦFL ε.

Note, that we have measured similar differences in total count-rates from an ensemble of

equally concentrated rsEGFP, Dronpa and EGFP solutions. This consistency in single-

molecule and ensemble brightness indicates a complete maturation of the proteins.

The observation times τobs were ~160 µs for EGFP (in accordance to the expected

focal transit time), while those of rsEGFP and Dronpa were shorter, reaching a value of ~40

µs at excitation powers > 5 µW for rsEGFP and ~55-65 µs at excitation powers > 20 µW for

Dronpa. The shorter observation times in the case of rsEGFP and Dronpa result from a fast

population of >160µs-lived dark states (for details see time11). As a consequence, the number

of photons detectable from a single rsEGFP molecule per single on-time (which is

proportional to the factor qτobs) is rather low and significantly reduced compared to Dronpa

and EGFP, presumably impeding the use of rsEGFP in single-molecule based nanoscopy

approaches.

Preparation of the protein polyacrylamide (PAA) layer

24.5 µl purified rsEGFP (0.09 mM) was mixed with 17.5 µl Tris-HCl pH 7.5, 30 µl

acrylamide (Rotiphorese Gel 30, Roth, Karlsruhe, Germany), 0.75 µl 10% ammonium

persulfate and 1µl 10% TEMED. About 10 µl of this solution was placed on a glass slide and

a cover slip was pressed onto the sample to ensure an appropriate thin layer. After complete

polymerization, the sample was sealed with silicon-based glue (Picodent twinsil, Picodent,

Wipperfürth, Germany) to prevent drying-out.

Cloning

The plasmid prsEGFP-MreB was generated by amplifying rsEGFP (forward primer:

TACGAATTCAAGGAGATATACATATGGTGAGCAAGGGCGAGGAG / reverse primer:

CATTCTAGACTTGTACAGCTCGTCCATG). The PCR product was used to replace the

YFP sequence in the pLE7 plasmid12 using the EcoRI and XbaI restriction sites. To create

fusion constructs of rsEGFP with Keratin19, with the peroxisomal membrane protein Pex16,

with Vimentin, or with the histone H2B, rsEGFP was amplified (forward primer:



GATCCACCGGTCGCGGCGTGAGCAAGGGCGAGGAGCTG / reverse primer:

ACAACTTAAGAACAACAATTGTTACTTGTACAGCTCGTCCATGCC). The PCR

fragment was cloned into the gateway destination vector pMD-tdEosFP-N13 using the

restriction sites AgeI and AflII, thereby replacing the tdEosFP coding sequence against the

rsEGFP sequence. The final plasmids pMD-Ker19-rsEGFP, pMD-Pex16-rsEGFP, pMD-Vim-

rsEGFP, and pMD-H2B-rsEGFP were constructed by gateway vector conversion (Invitrogen,

Carlsbad, CA) using the donor vectors pDONR223-Krt19, pDONR223-Pex16, pDONR223-

Vim, and pDONR223-Hist1H2BN14. The plasmid pMD-rsEGFP-Map2, coding for a rsEGFP-

Map2 fusion protein, was generated by exchanging the α-tubulin sequence of pEGFP-Tub

(Clontech) against the Map2 sequence, PCR-amplified from pDONR223-Map214 using the

restriction sites XhoI and BamHI. Subsequently, the EGFP sequence was exchanged against

the PCR amplified sequence of rsEGFP using the restriction sites NheI and BglII. To generate

pMD-rsEGFP-α-Tubulin, the coding sequence of rsEGFP was amplified (forward primer:

GATCCGCTAGCGCTAATGGTGAGCAAGGGCGAGGAG / reverse primer:

CACTCGAGATCTGAGTCCGGACTTGTACAGCTCGTCCATGCC) and exchanged

against the EGFP sequence in pEGFP-Tub (Clontech) using the NheI and BglII restriction

sites. The plasmid pMD-Mito-rsEGFP, coding for a mitochondrial targeted rsEGFP, was

created by exchanging the DsRed1 sequence of pDsRed1-Mito (Clontech) against a PCR-

amplified rsEGFP sequence (forward primer:

TCCACCGGTCGCCACCATGGTGAGCAAGGGCGAGGAG / reverse primer:

GTCGCGGCCGTTACTTGTACAGCTCGTCCAT). The restriction sites AgeI and NotI were

used. The plasmid pSCA3-Lifeact-rsEGFP was cloned by replacing the YFP sequence of

pSCA3-Lifeact-YFP against rsEGFP. To this end, rsEGFP was amplified (forward primer:

TCCACCGGTCGCCACCATGGTGAGCAAGGGCGAGGAGCTGTTC / reverse primer:

TACCCTGCGGCCGCTTTACTTGTACAGCTCGTCCATGCC) and cloned into the AgeI

and NotI restriction sites.



Preparation of E. coli samples for imaging

Samples were prepared as described previously15. In brief, powdered agar was washed twice

in water and taken up to 2 % w/v in water. The suspension was heated to dissolve the agar.

Approximately 1 ml of the agar solution was spread uniformly on a clean microscope slide

and dried over night. 10 µl of an E. coli culture was placed onto one coated slide, covered

with a cover slip and sealed with silicone-based glue (Picodent twinsil, Picodent,

Wipperfürth, Germany)

Expression of the rsEGFP-MreB fusion protein in E. coli cells

E. coli cells (strain SURE or MC1000) were transformed with the plasmid prsEGFP-MreB

and grown on LB agar plates containing ampicillin over night at 37 °C. A single colony was

picked and transferred to 5 ml LB-Amp medium. Cells were grown over night at 30 °C on a

shaker. Then, the cell suspension was again diluted and the cells were gown to an OD600 of

~0.6 at 30 °C. Protein expression was induced by adding 20 µM – 100 µM Isopropyl-β-D-

thiogalactopyranosid. After induction, the cells were grown at 25 °C or 30 °C under agitation

for 1-2 days.

Mammalian cell culture

PtK2 (Potorous tridactylis) cells were cultured under constant conditions at 37 °C and 5 %

CO2 in DMEM (Invitrogen, Carlsbad, California) containing 5% FCS (PAA, Pasching,

Austria), 100 units/ml streptomycin, 100 µg/ml penicillin (all Biochrom, Berlin, Germany),

and 1 mM pyruvate (Sigma, St. Louis, USA ). For transfection, cells were seeded on cover

glasses in 6-well plates. At the next day, cells were transfected with plasmid DNA using

Nanofectin (PAA, Pasching, Austria) according to the manufacturer’s instructions. After 24

hours the growth medium was replaced. Cells were imaged 24 - 72 h after transfection.

Preparation of organotypic hippocampal slice cultures

Hippocampal slices (350 µm thick) were prepared from postnatal day 5–7 wild-type C57BL/6

mice, embedded in a plasma clot on 0.14 mm thick glass cover slips and incubated in a roller

incubator at 35 °C, according to the Gähwiler method16. The age of the slice cultures used in

the experiments ranged from 12 to 42 days after the preparation. For imaging, slice cultures

were transferred to a chamber at room temperature and sustained in artificial cerebrospinal

fluid, containing 126 mM NaCl, 2.5 mM KCl, 2.5 CaCl2, 1.3 mM MgCl2, 30 mM glucose and

27 mM HEPES; the pH was adjusted with NaOH to 7.4.



Viral transfection

For transfection, a modified Semliki Forest Virus containing the pSCA-Lifeact-rsEGFP

vector construct was injected into the CA1 area of the slice cultures using a patch pipette

connected to a pressure-generator (Picospritzer, Parker, Pine Brook, NJ)17. The cultures were

then incubated for at least 16 hours and imaged within 48 hours after transfection.

Diffraction limited data storage

A custom-built confocal epi-fluorescence microscope (Supplementary Fig. 9) was used for

writing and reading. The 491 nm excitation light was provided by a solid-state laser (50 mW,

Calypso 50, Cobolt, Stockholm, Sweden), whose output power was reduced by a laser power

controller (LPC-VIS, Brockton Electro-Optics, Brockton, MA) and modulated by an acousto-

optic modulator (AOTF.nC/TN, AA Opto-Electronic, Orsay, Cedex, France). The 405 nm UV

light was generated by a diode pumped crystal laser (30 mW, BCL-030-405-S, CrystaLaser,

Reno, NV). Its power was reduced by a neutral density filter and modulated by an acousto-

optic modulator (MTS 130A3, AA Opto-Electronic). Shutters blocked the beams during long

non-operational periods of time. Each laser beam was cleaned up by a laser line clean-up filter

(z491/10x and z405/10x, AHF Analysentechnik, Tübingen, Germany). The coupled out

beams were collimated with achromatic lenses of 50 mm focal length (Linos Photonics,

Göttingen, Germany). A dichroic mirror (z450dcxr, AHF) combined the laser beams and a

dichroic mirror (z500dcxr, AHF) reflected them into the aperture of the 100× 1.4 NA oil

immersion objective (HCX PC APO 100×/1.40-0.70 oil, Leica Microsystems, Wetzlar,

Germany). A three-axis piezo stage (NanoMAX311 with BPC203, Thorlabs, München,

Germany) was used to position and scan the sample. The induced fluorescence from the

excited focal region was focused by an achromatic tube lens of 300 mm focal length (Linos)

onto a gradient-index multi-mode fiber with 62.5 µm core diameter (GIF625, Thorlabs) acting

as confocal pinhole. A band pass filter (HQ532/70m, AHF) rejected back-reflected and/or

stray laser light. An avalanche photo-diode (APD) (SPCM-AQR-13-FC, PerkinElmer, Salem,

MA) detected the fluorescence photons, which were finally counted and recorded on a

personal computer. A data acquisition card (NI-DAQ PCI-6731, National Instruments, Austin,

TX) was used to generate the scan voltages and the AOM drive voltages as well as the

blanking and synchronization signals. Custom MATLAB (The MathWorks, Natick, MA)

programs allowed automated generation of the voltages and signals for moving the sample to

the desired coordinates and for generating the desired laser pulses, as well as for recording the



fluorescence images. The laser beams were mutually aligned by recording the laser light

back-scattered by a silver nanobead (60 nm silver colloid, BBInternational, Cardiff, UK) with

a photo-multiplier tube (PerkinElmer). Images were taken using the software Imspector

(www.imspector.de).

We wrote Grimm’s Fairy Stories (http://www.gutenberg.org/files/11027/11027.txt),

encompassing 25 tales with ~270,000 letters, in a 17 µm × 17 µm region (Figure 2). Because

this large data set could not be written all at once into this tiny region, we first translated the

text into 7-bit binary ASCII code (0: off; 1: on) and wrote and read the whole text frame-by-

frame in 6596 frames of 41 letters (287 bits) each. The confocal setup was used to read, erase,

and write the data into the layer. We switched the rsEGFP in individual spots of ~0.5 µm

diameter with a center-to-center spacing of 1 µm, corresponding to a DVD standard (0.74 µm

pits) storage density.

For analyzing the data, we averaged the read-out fluorescence signal over an area of

0.6 µm × 0.4 µm (3×2 pixels) at the center of the respective data spot resulting in the average

spot brightness B for assigning either 0 or 1. To set the threshold signal B0, discriminating

between these values, we plotted all values in a histogram displaying a sharp and a broader

peak representing the 0’s and the 1’s, respectively (Supplementary Fig. 8). With this simple

brightness threshold to discriminate ‘0’ from ‘1’, the bit error rate was 3.7⋅10-6, corresponding

to 7 bit errors within the entire set. We expect that improving the algorithm further reduces

the error rate by >10-fold. After ~6,600 read/write cycles in the same region, the average

fluorescence of the set bits was reduced by ~35 % (Supplementary Fig. 8).

Prior to every storage cycle, we set all bits to 0’s by scanning over the entire frame

with 491 nm light (0.6 kW/cm2; 1.2 ms per step of 500 nm), followed by writing the 1’s at

405 nm (~40 W/cm2 for 10 ms per bit). Afterwards the data was read by scanning the frame at

491 nm (0.2 kW/cm2; 1 ms dwell time; step size 200 nm). Each storage cycle took ~22 s per

frame, including overhead. Writing and reading of the entire data required ~40 hours in total.

Switching fluorescence off by STED In order to turn off the fluorescence capability of a fluorophore in the focal plane of the

microscope, the focal intensity of the STED-beam must exceed the threshold Is =Cτ−1

=(hvσ−1)τ−1 ≈ 1-10 MW/cm2, with C=hvσ−1 denoting the photon energy divided by the

probability cross-section for de-excitation.



http://www.imspector.de/

RESOLFT microscope

For RESOLFT imaging with rsEGFP, we extended our confocal microscope for providing

two alternating foci at 491 nm (see also Supplementary Fig. 9): (i) a normal diffraction-

limited focus for reading out the fluorescence of rsEGFP and (ii) a focus with a central

intensity minimum (“zero”) for switching rsEGFP off at the focal periphery. In place of the

spatial filtering with the single-mode fiber, the following set-up was inserted to get the

desired beam profiles. A half wave plate allowed adjusting the polarization of the 491 nm

laser beam to match the axis of a Glan-Thompson polarizer (B. Halle). An electro-optic

modulator (LM 0202 5W 400–650nm with pulse amplifier LIV 20, Linos) rotated the linear

polarization by 90° for alternating between horizontal and vertical polarization. The vertically

polarized beam was reflected by a polarizing beam splitter (PTW5 450–650nm, B. Halle) and

spatially filtered by a Kepler telescope consisting of a 40 mm achromatic doublet, a 10 μm

pinhole (Linos) in the common focal plane, and a 100 mm achromatic doublet. The

horizontally polarized beam went straight through the polarizing beam splitter and was sent

through an identical spatial filter. It passed then through a 2π phase ramp (463 nm mask,

vortex plate VPP-A, RPC Photonics, Rochester, NY) for creating a lateral doughnut. A

second polarizing beam splitter (PWT5) recombined both beams thereafter. Alternating the

polarization allowed selecting either of the two 491 nm beam paths with a mutual contrast of

better than 2000:1. Finally, we introduced a quarter wave and a half wave plate (B. Halle) in

front of the objective for circularizing the polarization of the 491 nm doughnut beam.

We set up two independent spatial filters mainly for minimizing the wave front aberrations, in

particular those introduced by the polarizing beam splitters, as well as for independently

adjusting the depths of the switching and the reading foci. Furthermore, the collimation lens

for coupling out the 405 nm beam was replaced by an achromatic doublet with 100 mm focal

length to better fill the objective aperture. The linear polarization of the 405 nm beam was

roughly circularized by a quarter wave plate. For the mutual alignment of the beams, we

flipped in a pellicle (~8% reflectivity, BP108, Thorlabs) in between the main dichroic mirror

and the wave plates, which prevented any change of the polarization of the 491 nm beam. For

finding live cells that expressed rsEGFP well, we inserted a lens ahead of the main dichroic

mirror, which resulted in a wide-field illumination of an area of 30–40 µm in diameter. We

looked at the cells using an ocular in a separate beam path that was addressable instead of the

confocal pinhole.



RESOLFT pulse sequence and scanning scheme

We applied the following six-step scanning scheme. (i) The scan stage moved to the next

position. (ii) A break of 1 ms duration warranted that the stage settled at the target position.

(iii) The rsEGFP were turned on by a short pulse of focused light at 405 nm wavelength (1

kW/cm2 for 100 μs). (iv) A break of 2 ms duration let the activated proteins equilibrate. (v) At

the focal periphery, the proteins were then turned off by light at 491 nm wavelength focused

into a doughnut shape (~1 kW/cm2, 10–20 ms). (vi) The remaining bright proteins at the focal

center were probed by focused light at 491 nm wavelength (~1 kW/cm2, 1–2 ms). We chose

theses light intensities because we observed non-linear bleaching effects at substantially

higher light intensities. For gaining the next pixel, the beams were targeted further away on

the sample than the size of the focal spot. Thus we scanned the sample in several passes by

recording only every N-th pixel per pass. N was chosen such that rsEGFP could equilibrate

into the bright state before it underwent another on/off cycle. The pulse sequence was

repeated for each target position until the entire image was recorded.

The same microscope was used for subdiffraction all-optical writing. For subdiffraction

RESOLFT-type writing, the highest writing contrast was obtained by repeating the writing

procedure at a single point for 50 times. The on-state molecules were most selectively made

permanent (i.e. bleached) at 532 nm (Supplementary Fig. 10a).



Supplemental Figures and Tables

Supplementary Fig. 1 � Switching fatigue of RSFPs. Plotted are the respective normalized fluorescence intensities in the on-state (solid lines) and the off-state (dashed lines) against the cycle number. Measurements were performed on living E. coli colonies expressing the indicated RSFP. The duration of the switching cycles were adjusted to the spectroscopic properties of the RSFPs. Care was taken to ensure that all proteins were completely switched and at the same time the irradiation times were kept as short as possible. For switching and fluorescence readout, colonies of E. coli cells expressing rsEGFP, rsFastLime, Padron and Dronpa were constantly irradiated with light of 491 nm (~20 kW/cm²). For switching in the other direction, light of 405 nm (30 kW/cm²) was applied. In case of rsCherryRev, light of 532 nm (~20 kW/cm²) instead of 491 nm was used for switching into the off-state and fluorescence excitation. Note, that the increase of the fluorescence intensity of rsCherryRev with an increasing number of switching cycles appears to be due to a photoinduced structural change of the protein18. Inset: Residual fluorescence intensity in the off-state and the number of switching cycles until the fluorescence is bleached to 50 %. Theses values are extracted from the figure. Note, that both values depend on the light intensities used and on the specific experimental conditions.



Supplementary Fig. 2 � Amino acid sequence alignment of EGFP and rsEGFP. Differences are highlighted.



Supplementary Fig. 3 � Absorption spectra of equilibrium-state rsEGFP at different pH values. Purified rsEGFP protein in various buffers (pH 5: 0.1 M citrate buffer, pH 6, 7: 0.1 M Tris-HCl buffer, pH 9, 10: 0.1 M glycine-NaOH buffer).



Supplementary Fig. 4 � Maturation of the rsEGFP chromophore at 37 °C.



Supplementary Fig. 5 � rsEGFP is a monomer. a, Semi-native polyacrylamide gel electrophoresis of rsEGFP. Purified monomeric EGFP, dimeric dTomato, tetrameric DsRed, and rsEGFP were separated on a semi-native gel. Images were taken with a custom built gel documentation system. To detect green fluorescence (EGFP and rsEGFP) the gel was irradiated with light of 470 ± 5 nm and fluorescence was detected at 525 ± 30 nm. To detect red fluorescence (dTomato and DsRed) the gel was irradiated with light of 545 ± 10 nm and fluorescence was detected at 617 ± 37. Both images were overlaid and are represented in false colours. b, Size separation chromatography at 4 °C. Shown are chromatography runs of rsEGFP, monomeric EGFP, dimeric dTomato, and tetrameric DsRed.



Supplementary Fig. 6 � Expression of various functional rsEGFP fusion proteins in mammalian PtK2 cells. (a) Mito-rsEGFP, (b) Vimentin-rsEGFP, (c) Pex16-rsEGFP, (d) histone H2B-rsEGFP, (e) Keratin19-rsEGFP, (f) rsEGFP-MAP2, and (g) rsEGFP-α-Tubulin. Shown are single confocal sections of living cells. Fluorescence was excited by simultaneous irradiation with light of 488 nm and 405 nm. Scale bars: 20µm.



Supplementary Fig. 7 � Switching of Keratin19-rsEGFP fusion proteins expressed in living PtK2 cells. The proteins were switched 10 times from the on-state to the off-state and back. Images of the on-state were recorded by simultaneous irradiation with light of 488 nm and 405 nm. Off-state images were recorded with light of 488 nm only. Switching into the on-state was performed by scanning with 488 nm and 405 nm. Switching into the off-state was performed by scanning with light of 488 nm. Shown are single confocal sections. Scale bar: 25µm.



Supplementary Fig. 8 � Histogram of the data spot fluorescence intensities. Plotted are the detected fluorescence intensities during the optical data storage experiment as shown in Fig. 2. Shown are the fluorescence intensities of all spots of the first 100 frames (blue squares), the last 100 frames (black points) and the whole experiment (red bars). The shift in the mean counts of the on-state spots represents the photo-bleaching during the recording of the ~6500 frames.



Supplementary Fig. 9 � Schematic of the RESOLFT microscope used. LPC: laser power controller; AOTF, AOM: acousto-optic modulators; GTP: Glan-Thompson polarizer; EOM: electro-optic modulator (polarization rotator); PBS: polarizing beam splitter; PH: pinhole (10 µm in diameter); PM: phase mask (vortex plate or phase disc).



Supplementary Fig. 10 � High-density optical data storage in a thin rsEGFP layer. a, Comparison of on-state rsEGFP bleaching (red curve) with bleaching of off-state rsEGFP (blue curve). The total exposure times and applied intensities were the same in both cases, only the sequence of irradiation was different. For on-state bleaching the sequence was: 405 nm, 532 nm, 491 nm. For off-state bleaching the sequence was: 405 nm, 491 nm, 532 nm. Plotted are the respective fluorescence intensities recorded at the beginning of irradiation with 491 nm against the number of the switching cycle. For the RESOLFT writing we chose 50 cycles (green line) because it provided the highest contrast. b, RESOLFT writing and reading of nine 3x3 bit patterns in a layer of rsEGFP. The set bits had a distance of 200 nm. The displayed image represents an average of four recordings. Scale bar: 1 µm.



Supplementary Table 1 � Properties of EGFP, rsEGFP and Dronpa



Supplemental References 1. Sawano, A. & Miyawaki, A. Directed evolution of green fluorescent protein by a new

versatile PCR strategy for site-directed and semi-random mutagenesis. Nucleic. Acids

Res. 28, E78 (2000).

2. Leung, D. W., Chen, E. & Goeddel, D. V. A method for random mutagenesis of a

defined DNA segment using a modified polymerase chain reaction. Technique 1, 11-

15 (1989).

3. Patterson, G. H., Knobel, S. M., Sharif, W. D., Kain, S.R. & Piston, D. W. Use of the

green fluorescent protein and its mutants in quantitative fluorescence microscopy.

Biophys. J. 73, 2782-2790 (1997).

4. Magde, D., Elson, E. & Webb, W. W. Thermodynamic fluctuations in a reacting

system: measurement by fluorescence correlation spectroscopy. Phys. Rev. Lett. 29,

705-708 (1972).

5. Ehrenberg, M. & Rigler, R. Rotational brownian motion and fluorescence intensity

fluctuations. Chem. Phys. 4, 390-401 (1974).

6. Haustein, E. & Schwille, P. Ultrasensitive investigations of biological systems by

fluorescence correlation spectroscopy. Methods 29, 153-166 (2003).

7. Chen, Y., Muller, J. D., So, P. T. & Gratton, E. The photon counting histogram in

fluorescence fluctuation spectroscopy. Biophys. J. 77, 553-567 (1999).

8. Kask, P., Palo, K., Ullmann, D. & Gall, K. Fluorescence-intensity distribution analysis

and its application in biomolecular detection technology. Proc. Natl. Acad. Sci. USA

96, 13756-13761 (1999).

9. Widengren, J., Mets, U. & Rigler, R. Fluorescence correlation spectroscopy of triplet-

states in solution - a theoretical and experimental-study. J. Phys. Chem. 99, 13368-

13379 (1995).

10. Eggeling, C., Widengren, J., Rigler, R. & Seidel, C. A. M. Photobleaching of

fluorescent dyes under conditions used for single-molecule detection: evidence of two-

step photolysis. Anal. Chem. 70, 2651-2659 (1998).

11. Eggeling, C. et al. Reversible photoswitching enables single-molecule fluorescence

fluctuation spectroscopy at high molecular concentration. Microsc. Res. Tech. 70,

1003-1009 (2007).



12. Vats, P. & Rothfield, L. Duplication and segregation of the actin (MreB) cytoskeleton

during the prokaryotic cell cycle. Proc. Natl. Acad. Sci. USA 104, 17795-17800

(2007).

13. Testa, I. et al. Nanoscale separation of molecular species based on their rotational

mobility. Opt. Express 16, 21093-21104 (2008).

14. Lamesch, P. et al. hORFeome v3.1: a resource of human open reading frames

representing over 10,000 human genes. Genomics 89, 307-315 (2007).

15. Pfennig, N. & Wagener, S. An improved method of preparing wet mounts for

photomicrographs of microorganisms. J. Microbiol. Meth. 4, 303-306 (1986).

16. Gähwiler, B. H., Capogna, M., Debanne, D., McKinney, R. A. & Thompson, S. M.

Organotypic slice cultures: a technique has come of age. Trends Neurosci. 20, 471-

477 (1997).

17. DiCiommo, D. P. & Bremner, R. Rapid, high level protein production using DNA-

based Semliki Forest virus vectors. J. Biol. Chem. 273, 18060-18066 (1998).

18. Stiel, A. C. et al. Generation of monomeric reversibly switchable red fluorescent

proteins for far-field fluorescence nanoscopy. Biophys. J. 95, 2989-2997 (2008).



Documents

ARTICLE · super-resolution, the number of switching cycles afforded by the fluorophore assumes a vital role. Because they are able to generate an image with a single on–off cycle3,6,21,