1 Introduction

The importance of superresolution fluorescence microscopy is beyond any doubts, as notably acknowledged by the recent assignment of the Noble Prize 2014 in chemistry to Eric Betzig, Stefan W. Hell, and William E. Moerner [1].

Since its invention in the 1990s [2], imaging beyond the diffraction limit matured from proof-of-principle experiments [3, 4] to widespread applications in the biological, medical and material sciences. Nowadays, it has found its way into many labs which are - unlike a few years ago - not specialized in optics and developments in microscopy, but working on other questions, using microscopy as a tool. This development has been particularly fostered by the commercial availability of turnkey systems. Nowadays, superresolution imaging in two or three dimensions with high speed even of living systems has become feasible for a wide community [5–8].

Stimulated Emission Depletion (STED) microscopy, which was the first successful implementation of superresolution microscopy [4], is particularly useful for examining nanoscale processes, as it offers a high temporal resolution as well as the intrinsic facility of optical sectioning since it is based on a confocal microscope.

However, generally speaking, the higher the demands, the higher the complexity of the microscope needed. This is especially true for multicolor STED microscopy. Extreme care has to be taken that the high intensity STED beam does not interfere with the second color. 2color STED microscopy has been realized by using two independent channels with one excitation and one STED laser for each channel [9], which gives rise to a rather complex and high-maintenance setup, or by using long Stokes-shift dyes, requiring only one common STED laser [10]. Although favorable in terms of price and complexity of the system, the choice of fluorophores is in this case very limited. Multiplexing was successfully extended to 3color imaging using the fluorescence lifetime as a second parameter to distinguish two fluorophores with otherwise similar spectral properties [11]. Unfortunately, the high temporal resolution as well as the high emission intensities needed to reduce crosstalk between the channels further complicate the experimental setup.

Recently 4color STED was demonstrated by exploiting different bleaching behavior of fluorescent proteins [12]. After acquisition of a STED image of two fluorophores with emission in the same spectral window, a bleaching step preferentially bleaching only one of the fluorophores was performed. The STED image after the bleaching step shows mainly the fluorophore with higher photostability, therefore subtraction of the two images reveals the discrete spatial distribution of both fluorophores. Combination with a conventional 2color STED setup yielded 4color STED images. However, bleaching in STED experiments is a difficult-to-control feature, therefore extreme care has to be taken to obviate crosstalk. Additionally, the choice of fluorophores is once again limited and the achievable resolution is reduced relying on photolabile fluorophores [13].

To circumvent these problems we developed a general approach for STED multiplexing that can be extended to an arbitrary number of colors. We use transient DNA binding of a dye labeled DNA strand (imager strand) to a structure that was labeled with the complementary DNA sequence (docking strand). The imager strand can be easily replaced by washing and adding the next imager carrying the same dye but having a different DNA sequence. Varying the sequence of the docking and imager strand, a theoretically unlimited degree of multiplexing can be reached.

With this DNA PAINT (point accumulation for imaging in nanoscale topography) [14] type of approach multicolor STED is realized using the most fundamental STED setup with only a single excitation and a single STED wavelength. Advantageously, this approach can be performed with any STED microscope, no matter what configuration was chosen (wavelength, continuous wave (cw) or pulsed, gatedSTED, …). The free choice of fluorophore allows easy adaption to the available system and optimized resolution: Working with a single spectral color, chromatic aberrations are avoided altogether. As the duration of the binding events are controlled by the length of the complementary DNA sequences [14], complete removal of the imager strands is possible and therefore crosstalk is prevented. This in principle allows for multiplexing to an arbitrary level with minimized bleaching by continuous dye replacement. We demonstrate the potential of multiplexing with transient binding in STED microscopy using DNA nanostructures that enable a defined number of binding sites in a programmed arrangement.

2 Materials and methods

2.1 Microscope

Measurements were performed on a home-built STED microscope. The excitation laser (491 nm, max. 20 mW; Cobolt, Sweden) and the STED laser (592 nm, max. 1500 mW, MPBC, Canada) both working in cw mode were joined by a dichroic mirror (z 590 sprdc; AHF, Germany) and focused into the sample through a 1.4 NA objective lens (Olympus, Germany). Fluorescence was collected by the same lens, separated from the excitation by a dichroic mirror (zq 491 RDC; AHF, Germany) and detected by an APD (Perkin Elmer, Germany) with a multimode optical fiber (diameter 62.5 µm) serving as confocal pinhole.

Laser scanning was performed using a galvanic scanhead (YANUS IV Scan Head; TILL Photonics, Germany) and data was acquired using the Imspector® software [15]. The different channels were overlaid subsequently using custom Labview software (National Instruments, Germany).

2.2 DNA Origami

Unmodified and fluorescently labeled oligonucleotides were purchased from MWG Eurofins at HPSF or HPLC grade, respectively. All other chemicals were purchased from Sigma Aldrich.

12 helix bundle (12HB) DNA origamis were folded using a 5fold excess of unmodified and an 10fold excess of modified staple strands in a folding buffer containing TRIS-acetate-EDTA (TAE) and 16 mM MgCl₂ using a fast folding program at 47 °C for 2 hours.

Rectangular DNA origamis were folded using a 5fold excess of unmodified and a 10fold excess of modified staple strands in a folding buffer containing TAE and MgCl₂ (12.5 mM) using a temperature gradient form 95 to 4 °C over 2 hours.

All DNA origami samples were purified at 4 °C using an Amicon® filter system (100k) according to the manufacturer’s instructions.

2.3 Measurements

Experiments were performed in Labteks® (Nunc) on a BSA/biotin/NeutrAvidin®-surface: Chambers were incubated with 0.5 mg/mL bovine serum albumin (BSA)/biotin (1 mg/mL) in phosphate buffered saline (PBS) for 8 h at 4 °C, washed with PBS, incubated with 0.25 mg/mL NeutrAvidin® in PBS for 30 min at room temperature and washed 3 times with PBS. To immobilize DNA origamis on the surface, the freshly purified samples were diluted in PBS with 100 mM MgCl₂ to an approximate concentration of 1 nM, incubated for a few minutes at room temperature and washed with PBS. Measurements were carried out in PBS with 2 mM Trolox/Troloxquinone [16] and 50 mM MgCl₂.

The first imager was added to a final concentration of 50 nM. Imaging was performed at 0.05 ms/pixel and a pixel size of 20x20 nm², two consecutive STED images were recorded and summed up to increase S/N. After recording image 1, the imager solution was removed, buffer was added and removed for one washing step and the second imager solution was added. The same imager concentrations and imaging conditions were used for all imagers and images.

For the 2color experiment, excitation laser power was 3 µW, for the 4color experiment, excitation laser power was 6 µW, STED laser power was 350 mW for both (all laser powers measured in the aperture).

2.4 Analysis

The raw images were exported from Imspector® as ASCII and analyzed using custom-written Labview software. The data was transformed into false-color images (red and green for the 2color experiments, red, green, blue and white for the 4color experiments) and overlaid. To correct for small lateral drift, the images were shifted in x and y with respect to each other until all DNA origamis on the field of view matched the expected structure. This was done using 5 nm steps.

3 Results

As a model sample, we chose a DNA-origami-based platform [17]. DNA origami has the benefit of being a very reproducible platform for superresolution imaging because distances and number of fluorophores are precisely controlled [18–21]. Furthermore, the attachment of specific DNA sequences for transient binding is straightforward [14].

As a first proof-of-principle, we utilized the 12HB DNA origami with a cylindrical shape based on a 12helix-bundle-motif [22] (see scheme in Fig. 1). The cylinder has a length of about 200 nm and a diameter of 14 nm, consisting of an 8064 nucleotide long scaffold and about 200 short staple strands. Four staples, equally distributed over one long side of the 12 HB DNA origami, are modified at their 5′ terminus with biotin, which are used for attaching the DNA origami to a streptavidin-coated surface. For fluorescence labeling, staple strands are elongated at their 5′ end by a specific, 12 nucleotide long sequence. These extensions protrude from the DNA origami, yielding docking strands for the fluorescently labeled complementary imager strand, which is added to the solution for the imaging process.

 

Fig. 1 a) Principle of DNA based transient binding for STED: The structure of interest (e.g. a 12HB) was equipped with 19 docking strands with the sequence 1 (red) at each end and with 19 docking strands with sequence 2 (green) in the middle of the 12HB. By adding imager 1 complementary to sequence 1, the marks at the ends are exclusively and transiently labeled and imaged. After washing, imager 2 complementary to sequence 2 was added and imaged likewise. b) Schematic superresolution images corresponding to the labeling steps as described in 1A and overlay (middle panel). c) Fluorescence emission spectra of imager strands labeled with two OregonGreen488 dyes before and after binding the complementary DNA. An approx. 5 fold fluorescence increase is observed as self-quenching is avoided after hybridization.

Download Full Size | PDF

At both ends of the 12HB, two fluorescent marks were created opposite to the biotin-modifications by elongating 19 staples each with a specific sequence 1 (see Table 1) [19]. The distance between these two marks is ~150 nm, which is well below the diffraction limit. A third mark was created in the very center of the DNA origami by inserting 19 docking strands with a specific orthogonal sequence 2, leading to a fluorescently labeled feature between the two terminal spots. 19 docking strands per mark were used to ensure high fluorophore density.

Table 1. Sequences used for fluorescence labeling (italic: spacer between DNA origami and docking strand; OG488: OregonGreen488, the fluorophore used in all experiments)

View Table

Depending on the length of the complementary DNA sequence, the binding times of the imager strand to the docking strand can be adapted to the specific experimental needs and are in the upper millisecond to second range for 9 and 10 complementary nucleotides, respectively [14]. We chose a complementary sequence length of 10 nucleotides, meaning that the dye-labeled counter strands are easily washed away after imaging. This approach has recently been shown to work for multicolor DNA-PAINT experiments which are based on the stochastic binding that creates blinking events for localization based superresolution [23, 24]. To exploit the DNA labeling for STED, the concentration of dye-labeled imager strands in solution has to be higher by about one order of magnitude to ensure that most binding sites are occupied. This increases background noise and therefore complicates the measurements. Hence, we used a double-labeled imager strand: Many fluorophores are known to interact with each other when the local concentration is high enough, forming a dimer-complex with different spectral properties, e.g. reduced fluorescence intensity at the given excitation wavelength [25]. A high local concentration was achieved by attaching two fluorophores at the ends of a flexible molecule such as a short oligonucleotide. Figure 1(c) demonstrates this effect for OregonGreen488, the dye used in all following experiments: The fluorescence intensity of a 10nt-oligonucleotide labeled with OregonGreen488 both on the 3′ and the 5′ end in solution is decreased by a factor of about 5 compared to a solution where the unlabeled complementary sequence is added. Binding to the complementary sequence results in a rather stiff, double-stranded DNA where the interaction of the fluorophores at the 3′ and 5′ end is excluded. Interaction between different oligonucleotides can be neglected since the concentration is not high enough to form a substantial amount of quenched dimers. This double-labeling reduces the background fluorescence substantially and allows for successful STED imaging. Additionally, because each staple is labeled with two fluorophores, the signal is enhanced compared to common labeling with one fluorescent label per DNA strand.

Figures 1(a) and 1(b) show the principle of the fluorophore binding to the DNA origami as well as corresponding fluorescence images. After binding the biotinylated 12HB DNA origami to a neutrAvidin®-surface, imager 1 complementary to sequence 1 is added and binds to the elongated staples at the two ends of the DNA origami, which gives rise to a superresolution image as schematically shown in Fig. 1(c), left. After a washing step, where imager 1 is unbound from the DNA origami and removed (3rd panel of Fig. 1(a)), imager 2 is added, solely labeling sequence 2 in the center of the DNA origami (Fig. 1(b), right panel).

To validate whether this approach can be applied for STED imaging, we examined the 12HB DNA origami as described with our home-built cw-STED microscope, featuring a single channel with excitation at 491 nm and a STED wavelength of 592 nm.

Figure 2 shows 2color confocal and STED images (raw data) acquired with DNA PAINT type labeling. First, confocal and STED images of the 12HB DNA origami were recorded after addition of imager 1 complementary to sequence 1. The two spots shown in red with a distance of 150 nm are clearly resolved in the STED image, which is not possible in the confocal image (see Fig. 2(c)). The full-width-half-maximum (FWHM) is about 50 nm, which is in accordance with the expected resolution of the microscope for the applied settings [18]. After washing and adding the sequence complementary to sequence 2, a single spot (green) is obtained for each DNA origami structure.

 

Fig. 2 a) Confocal (top) and STED (bottom) images of 12HBs imaged as described in Fig. 1.The marks at the ends of the 12HB are shown in red (corresponding to sequence 1), the mark in the center is shown in green (corresponding to sequence 2). Imaging was performed using OregonGreen488. b) Magnified view of one of the structures: The upper panel shows the confocal image, which does not reveal any structural features. The STED image (lower panel) clearly reveals a red double spot and a green spot in between. c) The intensity profiles of the red and the green channel overlaid in the same fashion as the images. The two red marks exhibit distances of 65 and 85 nm from the green matching the designed distances of 75 and 75 nm within measurement uncertainty.

Download Full Size | PDF

To improve the signal-to-noise (S/N) ratio, two consecutive STED images per color were recorded and summed up without further drift correction. Recording of a single image takes approximately 10 seconds, meaning that between the two images, statistically all fluorophores should be exchanged and potentially bleached fluorophores should be removed and substituted by unaffected ones

For overlaying the images, we adjusted the images for small translational and rotational drift that might have occurred between the two recordings. The overlaying algorithm exploits the knowledge about the relative mark positions on the DNA origami so that no additional fiducial markers are required. Notably, no adjustment beyond translational and rotational drift is necessary as we always measured in the same spectral range and avoid any other such as chromatic aberrations. The 12HB DNA origami with its three binding sites in two colors was then reconstituted for every single structure independent of the orientation (Fig. 2, lower panels). This shows that the transient binding approach is feasible for multicolor STED imaging. In particular, there is no crosstalk between the imaging sequences, the unbinding of the labeled DNA strands is fast enough, and the washing step sufficient to get rid of any residuals of imager 1, which would lead to crosstalk between the channels. Remarkably, all DNA origamis show three spots, meaning that the labeling efficiency is high. The profiles in Fig. 2(c) show that the distance between the red spots is 150 nm, as designed, and the centered green spot has a distance of 65 and 85 nm, respectively, to the red spots, well in accordance with the design of the DNA origami.

To further extend this principle of multicolor STED, we used a rectangular DNA origami with dimensions of approximately 70 nm x 100 nm [26, 27]. All four corners were labeled by extending 15 staples per corner with four different sequences (Table 1). These docking strands were addressed sequentially with the corresponding imager strands. This yields a rectangle with distances of 47 nm (short sides), 57 nm (long sides) and 73 nm (diagonals) estimated from the centers of the labeled regions. Imaging was performed as before, both confocal and STED images were recorded for each imager sequentially. Each STED image was constructed by summing up two consecutive scans to increase the S/N ratio. The overlay shows that one can obtain a 4color STED image where all four corners are clearly resolved (see Fig. 3).

 

Fig. 3 Rectangular DNA origamis were equipped with specific DNA docking strands in each of the four corners and then imaged sequentially with the complementary imager strands. a) Overlaid confocal scans show a high yield of 4fold labeled structures but no structure. b) Overlaid STED scans of the same region verify the rectangular design of the four-color labeled rectangular DNA origami. c) A zoom of one of the structures and the corresponding scheme of the design show that the structure is visualized as designed. d) Intensity profiles verify the expected distances between the corners of 73 nm (diagonals, red-white, green-blue), 57 nm (long sides, white-blue, green-red) and 47 nm (short sides, blue-red, white-green), respectively. The intensity profiles are color-coded according to their color in the image, the white channel is shown in black. Images were taken with a pixel size of 40 nm for the confocal and 20 nm for the STED images.

Download Full Size | PDF

The overlay of the confocal scans shows a high yield of four-color labeled spots but cannot reveal structural details. The background signal stems from imager strands in solution (50 nM). However, due to the quenching of the double-labeled imager stand in solution we obtain a reasonable S/N ratio. To achieve the same S/N for STED images which are usually dimmer (because the fluorescence signal does not stem from a perfect point source and due to a non-perfect zero intensity in the very center of the donut-shaped STED beam), the summation of two consecutive images is sufficient (compare the color scales of 3(a) and (b)) so that the STED image reveals a high yield of four-color rectangles (see Fig. 3(b)). The magnified view of one of the DNA origamis clearly reflects the expected design as depicted in Fig. 3(c). In addition, Gaussian functions were fitted to the intensity profiles to estimate the distances. All six distances are within the range of uncertainty of the designed distances. Altogether, this clearly shows that using DNA based transient binding, four color STED images are easily obtained.

4 Discussion

In this work, we present a straightforward way to obtain multicolor STED images using a basic, one-channel STED microscope by utilizing transient binding of DNA imager strands to DNA docking strands.

Advantageously, multiplexing is achieved using a single spectral detection channel, excluding chromatic aberrations. This is very favorable for the overlay of the different colors as the correction of chromatic aberrations become very difficult for superresolution imaging. Additionally, the resolving capacity of the microscope is the same for all channels. STED with DNA PAINT labeling also allows for the free choice of fluorophores. Preferably, the dye shows a high degree of self-quenching to reduce background. We used OregonGreen488 in this case because it is well suited to our microscope settings [18]. However, any dye or even ultimately stable fluorophores like nanodiamonds can be used for targeted labeling, as long as they are attachable to an oligonucleotide.

In principle, this multiplexing can be even further extended to an arbitrary number of channels. However, care has to be taken that the background does not increase too much, probably due to unspecific binding to the surface. Also displacement of the sample during the replacement of the imager potentially increases with every replacement performed. We expect this practical limitation for multiplexing to be greater than 10, as estimated by similar Exchange-PAINT-experiments [23]. Additionally, this method can also be used in combination with almost any of the existing multiplexing methods, such as two-channel microscopes or fluorescence lifetime analysis. Also, bleaching the fluorophores instead of washing is possible. As no post-acquisition processing except for lateral and rotational adjustment has to be carried out, it is also simple and fast to use: Chromatic corrections, fluorescence lifetime analysis or localizing of spots as in stochastic superresolution are unnecessary.

By labeling antibodies with DNA, this technique can also be used for multiplexing STED imaging in cells. For assays in living cells, common live-cell compatible tagging approaches like specific tagging proteins such as the SnapTag [28] or the HaloTag [29] might be used. For example, CoenzymeA-modified DNA can be used as a substrate for the ACP-Tag protein [30]. Imaging with an imager specific to the sequence will allow for STED microscopy in living cells. By using orthogonal tagging proteins, this facilitates easy-to-use multicolor live cell STED imaging. However, for monitoring fast processes, care has to be taken that the time needed for washing and labeling does not interfere with structural changes. Usage of a microfluidic device can reduce the time needed for imaging vastly, making it possible to study even dynamic processes.

Since the whole multiplexing approach only relies on modification of the labeling approach, this technique can be easily and cost-efficiently applied by any user without extensive knowledge about optics, as would be required for the rather complex adjustments to expand a home-built or commercially acquired STED microscope into a multicolor system.

In summary, we present a versatile method for extending a simple, one-channel STED microscope with any technical specification to a system capable of multiplexing. As the multiplexing relies solely on the labeling of the sample, there is no need for complex setup adaptions. Since it uses a single dye for all channels, no chromatic corrections have to be considered either. Crosstalk is effectively suppressed, and in principle any fluorophore can be used. Therefore we are certain that this labeling approach is an easy-to-implement enhancement for STED microscopy in daily use.