In the past two decades, super-resolution microscopy techniques have been used in biological research . Amongst the many developed super-resolution techniques, STED microscopy has been adopted as a powerful super-resolution extension to confocal microscopy . Having the potential to resolve structures at the <30nm range in live samples, STED is well suited for live-cell research on the single-protein scale. STED microscopy has been developed and improved extensively over the past decade, yet it has remained challenging to accurately address and measure the three-dimensional Point-Spread-Function (PSF) of STED systems.
In STED microscopy the shape of the PSF does not only depend on the optics of the microscope, but also on specific factors such as the STED laser power and the fluorescent behavior of the specimen itself . These parameters are estimated from the instrument settings, and from typical fluorophore parameters. The parameters and estimations stored in the meta-data of the measurements can be used to calculate a theoretical PSF. The shape of the PSF (see figure 1) determines the achievable spatial resolution of the microscope and it is a necessary prerequisite to perform deconvolution.
While the theoretical estimation of the PSF of the STED system is a successful approach for performing deconvolution, it is important to remember that it is still based on approximations and on an ideal, typical acquisition conditions rather than on an actual experimental situation. Predicting the STED PSF for each specific fluorophore, for each sample and imaging conditions with absolute accuracy - as desired for perfect image restoration - is close to impossible if only the theoretical approach is used. Therefore, to investigate the performance of the STED system, and potentially improve image restoration with deconvolution, it can be advantageous to determine the PSF experimentally.
The ideal tool to measure the PSF would be a point‑like light emitter with a high brightness density, stable with respect to photobleaching, with known spectral properties and with a well-defined and sharp distribution of brightness and size. So far, the samples of choice for STED microscopy have been fluorescent beads with small diameters of around 40 nm. These common beads show a broad distribution of size and brightness, a relatively low brightness density and spectral properties different from the dyes used in the specific experiments. Due to these disadvantages it has been very challenging to acquire an accurate three-dimensional PSF for STED microscopes. GATTAquant’s research has been focusing to solve the problems, disadvantages and imperfections of common beads and presents now a revolutionary new type of fluorescent beads, the GATTA-Beads, which are ideal for application in the field of super-resolution microscopy e.g. for measuring experimentally the PSF of STED microscopes.
In this article it is shown that GATTA-Beads are a good match in combination with STED microscopes from Leica Microsystems and Huygens software from SVI (Scientific Volume Imaging B. V.), which is specialized in deconvolution, restoration, visualization and analysis of microscopy images.
Fluorescent beads based on the DNA origami technology
GATTAquant is a company which develops products for the microscopy market and is best known for their nanoruler samples which allow a precise test of the achievable spatial resolution of fluorescence and especially of super-resolution microscopes [4,5]. Using the so called DNA origami technology, GATTAquant recently developed a new type of fluorescent bead, called GATTA-Bead. With DNA origami technology it is possible to produce billions of identical objects on the nanoscale level, which can be used as molecular breadboards for organic dye molecules. Therefore, the organic dye molecules can be positioned with nanometer precision and with defined distances, in order to prevent quenching which would lower the effective brightness of the samples. In this way, the GATTA-Beads were developed with very high brightness density per volume. Compared to conventional fluorescent beads, the new type of fluorescent beads offered by GATTAquant show some advantageous properties:
- The GATTA-Beads are optimized to provide a large number of interaction free dye molecules per volume. Addition of further dye molecules would lower the effective brightness of the beads due to quenching of the dye molecules. The absence of quenching is also indicated by the narrow distribution of fluorescent lifetime of the dye molecules as shown in figure 2.
- The use of the DNA origami technology results in billions of essentially identical structures with respect to size and shape, which is of special importance for measuring the PSF of STED microscopes. The monodisperse particle size leads to fluorescent beads with a very narrow distribution of brightness compared to conventional fluorescent beads, as shown in figure 2.
- GATTA-Beads are labeled with common organic dye molecules which are best known in biological applications, e.g. Cyanine-, Alexa- or ATTO-dyes. Using the same sort of dyes and thus the same spectral properties as in the later measurements of biological samples yields relevant and realistic results regarding deconvolution of STED measurements and measurement of PSF in general. Furthermore, GATTAquant offers to customize the fluorescent beads. The beads can be labeled with essentially any dye molecule and eventually for every arbitrary combination, for multicolor beads.
- GATTA‑Beads can be labeled in addition to the organic dye molecules with linker molecules like biotin. This offers the possibility to immobilize the beads to a great variety of surfaces, as to standard coverslips or glass slides.
- GATTA‑Beads are available either in solution, for a very flexible use of the samples, or as ready‑to‑use beads, immobilized and embedded on a coverslip. Embedding media which support photostability are used, in order to preserve the initial brightness significantly. Because of the enhanced photostability in combination with high brightness density and large number of dye molecules per structure offered by the GATTA-Beads, it is possible to measure three‑dimensional PSFs of STED microscopes.
The properties described above make the GATTA‑Beads to the ideal tool for various applications in fluorescence microscopy. The next section will describe a very prominent application, the measurement of the three-dimensional PSF of a state‑of‑the‑art commercial STED microscope of Leica Microsystems for further use in deconvolution with the Huygens software.
Measuring the 3D STED-PSF
Measuring the PSF of STED microscopes provides a direct feedback about the correct calibration of the system. Furthermore, deconvolution of the measured data can be improved by using the measured PSF of the microscope.
The Leica TCS SP8 STED 3X is the latest version of Leica’s STED system. The system offers three‑dimensional, multicolor, super-resolution fluorescence microscopy and is widely used in research labs all over the world. Countless, high quality peer-reviewed studies have been published using the system, showing its quality and user friendliness.
In order to measure the 3D-PSF of the Leica TCS SP8 STED 3X, GATTA‑Beads with a diameter of 23 nm and labeled with 80±10 dye molecules were used. The beads were immobilized on high-precision coverslips (#1.5H, (170 ± 5) µm) and embedded in a polymer including anti-fading reagents. GATTA‑Beads with ATTO647N dye molecules were measured with 775 nm STED depletion laser at 50% of max. power and GATTA‑Beads with ATTO542 were imaged with the 660 nm STED laser at 13% of max. power. For excitation, a white light laser source was used and the excitation wavelengths were 646 nm for ATTO647N and 542 nm for ATTO 542. The laser line was selected with AOTF/AOBS and adjusted to the respective excitation maximum of the dyes.
Images were acquired with pixel sizes of 15 or 20 nm, and with 150 nm z-step size.
Images were acquired with 8x line accumulation at 600 Hz (unidirectional) in photon counting mode with the Leica Hybrid Detector (HyD), which allowed for the best trade-off in SNR and bleaching.
PSF distilling using Huygens Software
The acquired bead images are an indirect measure for the PSF of the system. In order to obtain the PSF of the system, an additional PSF distilling process is required [6,7].
Image formation in fluorescence microscopy can be described by a convolution process with the inclusion of (Poisson) noise. With deconvolution, the aim is to restore the object (unknown) from the measured image through knowledge of the PSF and noise. In the PSF distilling, the object is known, and is modeled as a sphere with a diameter of 20 nm. The PSF is the unknown function in this inverse problem. A non-linear maximum likelihood estimation method is applied, analogue to deconvolution algorithms, with the role of the object function and PSF reversed . With the PSF distillation the aim is to find the most likely PSF that generated the measured image. This process is illustrated in figure 3.
The Huygens PSF distiller (Huygens Professional, Scientific Volume Imaging B.V.) was used to distill the PSF from the acquired bead images (figure 4). The Huygens PSF Distiller is a wizard-based software tool that assists in distilling a PSF from bead images. The data consists of multiple beads recorded as separate images, multiple beads in the same image, or a combination of both. The PSF distiller uses a non-linear iterative method to extract the PSF from the bead images. The distiller wizard automatically selects the most suitable beads from the image, centers the beads, and accumulates them to average out noise, as illustrated in figure 5.
The automatic bead selection process is based on the size of the objects in the raw image and how well the bead is separated from other beads. A large object size, compared to the size of the theoretical PSF, might indicate a cluster of beads rather than individual beads. These large objects are discarded in the bead accumulation process such that overestimation of the size of the final PSF is prevented.
In a typical acquired image (with dimensions of 512x512x11 voxels), the field of view contained approximately 50-100 GATTA‑Beads. The Huygens PSF distiller was able to select around 5-10 beads per image that were suitable for the distilling process. This increases the effective signal by 5-10 times, which is sufficient for noisy confocal and STED images.
After accumulating the best beads, the non-linear iterative restoration method is used to optimally distill the PSF. After distilling the PSF, the Huygens PSF distiller offers the possibility for automatic estimation of the STED parameters. This estimation is performed by iteratively fitting different STED parameters, such as STED saturation factor, immunity fraction and fill-factor, such that the theoretical PSF with these settings matches the measured PSF as best as possible. This estimation could also help to extrapolate theoretical STED parameters when different microscope settings (such as STED laser power) are being used.
By using GATTA‑Beads and the SVI Huygens software, the three-dimensional PSF of the Leica TCS SP8 STED 3X super-resolution microscope was measured and distilled for two spectral regions. Figure 6 shows the measured three-dimensional PSFs. To the best of our knowledge, this represents the first accurate measurement of the three-dimensional PSF of a STED microscope with fluorescent beads, having a diameter smaller than 30 nm.
Figure 7 shows the intensity profile along X, Y and Z for the theoretical PSF (calculated based on STED parameters present in the meta-data of the file), and the measured (distilled) PSF for the ATTO 647N GATTA‑Beads. The distilled PSF shows a narrower base profile, and a smaller FWHM in X/Y compared to the theoretical PSF calculated based on the LIF file meta data. This indicates that some of the estimated STED parameters in the meta-data (such as STED saturation factor), are underestimated for the given situation, i.e. the 775 nm STED laser is able to deplete the ATTO 647N fluorophore more effectively than assumed for an average fluorophore at this wavelength by the Leica acquisition software.
In order to test if the deconvolution results were improved by using a measured PSF, different regions in the beads sample were imaged. The images were deconvolved with the theoretical PSF and with the measured (distilled) PSF using the GPU accelerated Good's roughness Maximum Likelihood Estimation (GMLE)  algorithm in Huygens Professional. An example of the deconvolution results is shown in figure 8. In this example, it was possible to separate two closely spaced beads at ~65 nm after deconvolution with the measured PSF, while this was not possible by using the theoretical PSF based on meta-data information. This example shows that the STED deconvolution result can be improved when a measured PSF is used.
In figure 9, intensity profiles along the X, Y and Z axis for the ATTO 542 PSF are shown for both the theoretical and measured (distilled) PSF. When comparing both the intensity profiles and the FWHM, it was observed that there is little difference between the theoretical and measured PSF, indicating that the STED efficiency in the meta-data is well estimated for this sample and imaging condition. Note that instrument settings were chosen to obtain a good data set to distill the STED PSF, not to show maximal resolution achievable with the instrument.
A different region of the ATTO542 bead sample was deconvolved with both the measured PSF and theoretical PSF. The comparison is shown in figure 10 (c) and (d). The intensity profile along the direction indicated with the arrow is shown in figure 10 (e). It was possible to clearly separate two beads that were spaced 170 nm apart, both with the theoretical and measured PSF. The deconvolution result with the measured PSF shows a slightly better contrast compared to the theoretical PSF deconvolution, indicating that STED deconvolution can benefit from using a measured PSF even when the measured PSF closely matches the theoretical PSF.
GATTAquant offers a new type of fluorescent bead with improved properties for measuring the three-dimensional PSF of state-of-the-art STED microscopes. In combination with the SVI Huygens Software, the 3D STED PSF could be efficiently distilled and characterized. The distilled PSF shows a narrower base and shows either a smaller or comparable FWHM compared to the theoretical PSF, which is calculated based on image meta-data. Deconvolution was performed on bead images, and the results show that by using a measured PSF with STED deconvolution in Huygens, a higher image contrast and an improved two-point resolution can be obtained compared to using a theoretical PSF.
This shows that the 3D STED PSF, as acquired by imaging GATTA-beads and distilled in the Huygens Software, can be used to extract more information and resolution from STED data via image restoration methods readily available in the Huygens Software.
Images were acquired on the Leica Microsystems TCS SP8 3X STED system at the VUmc, (Amsterdam, the Netherlands). The authors of this article would like to thank Lino Miltenburg, Jeroen Kole and René Musters for their help with acquiring images on this system. We thank Carsten Forthmann for proofreading, analysing data and preparing figures.
- Blom, H. & Brismar, H. STED microscopy: increased resolution for medical research? Journal of internal medicine 276, 560–578 (2014).
- Hell, S. W. & Wichmann, J., Breaking the diffraction resolution limit by stimulated emission. Stimulated-emission-depletion fluorescence microscopy. Opt. Lett. 19, 780 (1994).
- Schoonderwoert, V., Dijkstra, R., Luckinavicius, G., Kobler, O. & van der Voort, H. Huygens, STED Deconvolution Increases Signal-to-Noise and Image Resolution towards 22 nm. Micros. Today 21, 38–44 (2013).
- Schmied, J. J. et al., Fluorescence and super-resolution standards based on DNA origami. Nature methods 9, 1133–1134 (2012).
- Straube, T., Schmied, J. J. & Forthmann, C. Quantifying the Resolution of a Leica SR GSD 3D Localization Microscopy System with 2D and 3D Nanorulers (2016).
- van der Voort H.T.M., Strasters K.C., Restoration of confocal images for quantitative image analysis, J. of Microscopy 178 (2). 165.181 (1995).
- van Kempen, G.M.P., van der Voort, H.T.M., Bauman, J.G.J., Straster, K.C., Comparing maximum likelihood estimation and constrained Tikhonov-Miller restoration. IEEE Engineering in Medicine and Biology Magazine, 15, 76-83 (1996).
- Marcel Lauterbach. STED figure. Available at https://de.wikipedia.org/wiki/STED-Mikroskop#/media/File:STED_Mikroskop_PSFs.jpg.
- Verveer, P. J. & Jovin T. M., Image restoration based on Good's roughness penalty with application to fluorescence microscopy, J. Optical Society of America 15, 1077-1083 (1998).