ThunderSTORM: a comprehensive ImageJ plug-in for PALM and STORM data analysis and super-resolution imaging

Super-resolution microscopy enables imaging of structures smaller than the diffraction limit. Single-molecule localization microscopy methods, such as photoactivation localization microscopy and stochastic optical reconstruction microscopy, reconstruct images by plotting the centroids of fluorescent point sources from a series of frames in which only a few molecules are fluorescing at a time. These approaches require simpler instrumentation than methods that depend on structured illumination and thus are becoming widespread. The functionalized rhodamine derivative reported in this paper spontaneously converts between a bright and dark state due to pH-dependent cyclization. At pH 7, less than 0.5% of the dye molecules are fluorescent at any given time. Blinking occurs on time scales of seconds to minutes and can therefore be used for single-molecule localization microscopy without sample treatment or activation. The dye is bright and straightforward to use, and it is easy to synthesize and functionalize. Thus, it has potential to become a new and powerful addition to the toolset for super-resolution imaging.

In the recent past, single-molecule based localization or photoswitching microscopy methods such as stochastic optical reconstruction microscopy (STORM) or photoactivated localization microscopy (PALM) have been successfully implemented for subdiffraction-resolution fluorescence imaging. However, the computational effort needed to localize numerous fluorophores is tremendous, causing long data processing times and thereby limiting the applicability of the technique. Here we present a new computational scheme for data processing consisting of noise reduction, detection of likely fluorophore positions, high-precision fluorophore localization and subsequent visualization of found fluorophore positions in a super-resolution image. We present and benchmark different algorithms for noise reduction and demonstrate the use of non-maximum suppression to quickly find likely fluorophore positions in high depth and very noisy images. The algorithm is evaluated and compared in terms of speed, accuracy and robustness by means of simulated data. On real biological samples, we find that real-time data processing is possible and that super-resolution imaging with organic fluorophores of cellular structures with approximately 20 nm optical resolution can be completed in less than 10 s.

Within only a few years super-resolution fluorescence imaging based on single-molecule localization and image reconstruction has attracted considerable interest because it offers a comparatively simple way to achieve a substantially improved optical resolution down to ~20 nm in the image plane. Since super-resolution imaging methods such as photoactivated localization microscopy, fluorescence photoactivation localization microscopy, stochastic optical reconstruction microscopy, and direct stochastic optical reconstruction microscopy rely critically on exact fitting of the centre of mass and the shape of the point-spread-function of isolated emitters unaffected by neighbouring fluorophores, controlled photoswitching or photoactivation of fluorophores is the key parameter for resolution improvement. This review will explain the principles and requirements of single-molecule based localization microscopy, and compare different super-resolution imaging concepts and highlight their strengths and limitations with respect to applications in fixed and living cells with high spatio-temporal resolution.

Review: 97 refs.

The functionality of the human body and with it many of its diseases are based on single cells or even on single cellular components. It is therefore essential to gain insight into the intra- and intercellular processes in order to understand the overall physiological functions and the mechanisms of ailments. Among the many techniques which are available for investigation, taking microscopic images of regions of interest plays a major role.
\nOptical fluorescence microscopy is a powerful tool since it augments the advantages of optical microscopy, which are non-invasive imaging of the inside of sufficiently transparent samples, with the high specificity of molecular fluorescence labeling. This is in contrast to, for example, electron microscopy, which is limited to measuring ultra-thin slices, or atomic force microscopy which provides only information about the sample surface.
\nThe microscope’s optical resolution determines the smallest structure size which can be distinguished in the image. In order to be able to visualize for example small cellular structures like single filaments or record molecular transport processes, a resolution in the range of typically several tens of nanometers or even better is needed. Unfortunately, light of wavelength λ, emitted by a point source and imaged by a lens, is always detected as a blurred spot. Adjacent objects which are closer than d = λ/2n sin α cannot be separated since their images are blurred by diffraction into a single pattern. At this, n is the refractive index of the medium and α is half the opening angle of the objective lens. Accordingly, it is not possible to focus visible light to a spot size smaller than 200 nm laterally and 400 to 700 nm axially.
\nThis diffraction barrier was postulated by Émile Verdet, Ernst Abbe and Lord Rayleigh at the end of the 19th century and limited the resolution for all far-field light microscopes until the 1990s. 4Pi microscopy and I5M improved the resolution in axial direction up to a factor of 7 by using opposing objective lenses coherently. Still, this does not overcome the fundamental limitations due to the wave characteristics of light.
\nIn the last decades, super resolution imaging techniques have been established which overcome this diffraction barrier. A review by Stefan W. Hell gives a comprehensive overview and is recommended for a deeper insight. The key element in order to distinguish fluorescent objects less than 200 nm apart is the on and off switching of their signal such that it can be separated in space and time. This typically requires specific fluorescent molecules which can be transferred resp. switched between a fluorescent on state and a dark off state.
\nThe available switching variants are manifold. Most basic is the switching between a bright singlet S1 and a dark ground S0 electronic state. Alternatively, the molecule can be transferred between an excitable on state and a non-excitable off state, for example by a long lasting electronic triplet state or a state generated by chemical bonding. The mechanism used depends on the actual microscopy concept.
\nIn order to increase the resolution well beyond the diffraction limit there are two complementary approaches. Either the region in which fluorescent molecules are in their on state is actively controlled by targeted switching, or single molecules at random positions are stochastically switched between on and off and their location is determined subsequently.
\nTargeted switching is used in methods such as stimulated emission depletion (STED), saturated pattern excitation microscopy (SPEM), saturated structured illumination microscopy (SSIM) and reversible saturable/switchable optically linear fluorescence transition (RESOLFT). For STED in particular the fluorescence excitation, induced by a diffraction-limited focused beam, is restricted in space by a second overlayed beam that features a central zero intensity area. This second beam de-excites the molecules to the electronic ground state, only at the zero intensity center fluorescence is still allowed. The extent of that defined area scales inversely with the square root of the applied STED intensity and is not limited by the diffraction barrier anymore. Within biological samples resolutions of about 20 nm full width at half maximum (FWHM) in the focal plane can be reached. Stochastic switching is used in the Single Marker Switching (SMS) schemes. Depending on the applied switching mechanism they are referred to as photo-activated localization microscopy (PALM), stochastic optical reconstruction microscopy (STORM), photo-activation localization microscopy with independently running acquisition (PALMIRA), ground state depletion microscopy followed by individual molecule return (GSDIM) and variants thereof. The probability for two fluorophores to be emitted at the same time within a diffraction limited volume has to be negligible. Therefore, the fraction of stochastically switched molecules in the on state needs to be restricted sufficiently. The burst of N photons, before the molecule subsequently transfers to the dark state, is detected as a diffraction limited pattern, spatially separated from the spots of other molecules. The centroid can be localized with a precision much better than the diffraction limit and scales with 1/√N. The hereby determined position is registered in a position histogram.
\nThe succession of switching on, emitting/detecting photons and switching off of random fluorescent molecules needs to be repeated a sufficient number of times in order to achieve a detailed image of the sample. Typical resolutions are in the range of several 10 nm FWHM in the focal plane.
\nBoth the targeted and stochastic approach need adaptations in the optics and/or light sources of the microscope setups. An alternative strategy to overcome the diffraction barrier is to analyze the independent stochastic intensity fluctuations of fluorescent emitters in the super-resolution optical fluctuation imaging (SOFI) concept. For this, just a short video of the sample with labels switching repeatedly and independently between a fluorescent and a non-fluorescent state is required. The cumulant of the original pixel time series, related to the correlation function, gives the pixel value of the final SOFI image, calculable up to different orders. Non-correlated fluctuations cancel each other out whereby only highly correlated fluctuations remain. The resolution improvement depends on the order of cumulant reached which again requires i.a. high signal intensities. A 5-fold improvement in spatial resolution beyond the diffraction barrier can typically be achieved. Since cells in their natural environment have a distinct spatial extension, resolution increases not only in two, but in all three spatial dimensions is indispensable to super-resolve their three-dimensional (3D) structure. The SOFI concept is intrinsically three-dimensional by taking a video z-stack and the STED technique of depletion can equally be extended to the axial direction. In contrast the expansion of the SMS-based techniques to the third dimension is typically realized by breaking the axial symmetry of the detection point spread function (PSF). By using two opposing objective lenses in a 4Pi like geometry the detection efficiency can be increased twofold, improving the resolution by a factor of √2. Interference between the signals detected through both lenses increases the axial resolution even further, resulting in an overall resolution of about 6 nm FWHM in the axial and 8–22 nm in the lateral direction. However, such methods are restricted to thin layers which are in the range of about 0.25 µm and 1 µm for the interferometric PALM (iPALM) and the 4Pi-SMS implementation respectively. The restriction is caused by the limited focal length of the required high numerical objective lenses. Recently with the whole-cell 4Pi single-molecule switching nanoscopy (W-4PiSMSN) setup the 4Pi-SMS scheme is optimized i.a. by deformable mirrors such that whole cells along a 10 µm axial range can be imaged with isotropic resolution. Since the focal depth of the high numerical aperture (NA) objective lens is still limited to about 1.2 µm the concatenation of optical slices is necessary.
\nA new concept for super resolution imaging is needed which provides a much greater axial range, preferably fully isotropic. In this dissertation a stereo three-dimensional Single Marker Switching (Stereo 3D-SMS) microscope is presented which is capable to image large sample volumes.
\nMultiple objective lenses image the same emitter from different perspectives which are not on the same optical axis. Similar to the concept of stereo view the spatial position of the emitter can be calculated from the respective two-dimensional (2D) detection patterns. In order to optimize the detection efficiency and to achieve an isotropic resolution over a great volume four objective lenses are used simultaneously, arranged in a tetrahedron like manner. This stereo SMS concept applies the basic principle of localizing the detection patterns even for the expansion to the third dimension and has no need for any PSF modifications.
\nThe dissertation covers the whole development process from the plain idea towards the first applications with the following main points:
\n•\tConceiving an implementation of the stereo view procedure.
\n•\tSimulation of the expected capabilities in terms of resolution and spatial volume.
\n•\tComputer aided design of the setup.
\n•\tProgramming the control of the electronic devices.
\n•\tDevelopment of the concept as well as the related algorithms to generate a 3D image from the measured raw data.
\nThe thesis in hand starts with the theoretical background of fluorescent imaging and super resolution. Then, the implementation of the experimental setup is presented. Beside the technical compone

Recent advances in optical microscopy have enabled biological imaging beyond the diffraction limit at nanometer resolution. A general feature of most of the techniques based on photoactivated localization microscopy (PALM) or stochastic optical reconstruction microscopy (STORM) has been the use of thin biological samples in combination with total internal reflection, thus limiting the imaging depth to a fraction of an optical wavelength. However, to study whole cells or organelles that are typically up to 15 microm deep into the cell, the extension of these methods to a three-dimensional (3D) super resolution technique is required. Here, we report an advance in optical microscopy that enables imaging of protein distributions in cells with a lateral localization precision better than 50 nm at multiple imaging planes deep in biological samples. The approach is based on combining the lateral super resolution provided by PALM with two-photon temporal focusing that provides optical sectioning. We have generated super-resolution images over an axial range of approximately 10 microm in both mitochondrially labeled fixed cells, and in the membranes of living S2 Drosophila cells.

Fluorescence microscopy has become one of the most widely used tools in modern cell biology. By selectively labelling molecules of interest with fluorescent tags, researchers can visualize the structural organization, distribution, and clustering of proteins within their native cellular environment. Single-molecule localization microscopy (SMLM) has extended fluorescence imaging by enabling spatial resolution far beyond the diffraction limit of light. Among SMLM techniques, fixed Photoactivated Localization Microscopy (PALM), Stochastic Optical Reconstruction Microscopy (STORM), and Direct Stochastic Optical Reconstruction Microscopy (dSTORM) provideaccess to subcellular organization below the diffraction limit that is otherwise difficult to achieve through conventional or even intermediate super-resolution methods, such as Structured Illumination Microscopy (SIM) or Stimulated Emission Depletion microscopy (STED). These techniques attain super-resolution by temporally separating fluorophore emission events, localizing individual molecules with high precision, and assembling their positions into a reconstructed image. Despite the recent advances of newer super-resolution modalities, fixed PALM and (d)STORM remainwidely used, due to their molecular specificity, accessibility, and broad applicability across diverse biological systems. This chapter briefly introduces the core principles of SMLM, situates PALM and (d)STORM within the evolution of super-resolution microscopy, and provides detailed protocols for the image acquisition and data analysis of fixed-cell PALM and dSTORM imaging in monolayer cultures, such as PC12 cells and primary hippocampal neurons. The workflow described here is compatible with standard SMLM platforms and uses Zeiss ZEN Black 2012 and Abbelight NEO software for reconstruction and analysis.

Optical imaging techniques provide much important information in understanding life science especially cellular structure and morphology because “seeing is believing”. However, the resolution of optical imaging is limited by the diffraction limit, which is discovered by Ernst Abbe, i.e. λ/2(NA) (NA is the numerical aperture of the objective lens). Fluorescence super-resolution microscopic techniques such as Stimulated emission depletion microscopy (STED), Photoactivated localization microscopy (PALM), and Stochastic optical reconstruction microscopy (STORM) are invented to have the capability of seeing biological entities down to molecular level that are smaller than the diffraction limit (around 200-nm in lateral resolution). These techniques do not physically violate the Abbe limit of resolution but exploit the photoluminescence properties and labelling specificity of fluorescence molecules to achieve super-resolution imaging. However, these super-resolution techniques limit most of their applications to the 2D imaging of fixed or dead samples due to the high laser power needed or slow speed for the localization process. Extended from 2D imaging, light sheet microscopy has been proven to have a lot of applications on 3D imaging at much better spatiotemporal resolutions due to its intrinsic optical sectioning and high imaging speed. Herein, we combine the advantage of localization microscopy and light-sheet microscopy to have super-resolved cellular imaging in 3D across large field of view. With high-density labeled spontaneous blinking fluorophore and wide-field detection of light-sheet microscopy, these allow us to construct 3D super-resolution multi-cellular imaging at high speed (~minutes) by light-sheet single-molecule localization microscopy.

Superresolution imaging techniques based on the precise localization of single molecules, such as photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM), achieve high resolution by fitting images of single fluorescent molecules with a theoretical Gaussian to localize them with a precision on the order of tens of nanometers. PALM/STORM rely on photoactivated proteins or photoswitching dyes, respectively, which makes them technically challenging. We present a simple and practical way of producing point localization-based superresolution images that does not require photoactivatable or photoswitching probes. Called bleaching/blinking assisted localization microscopy (BaLM), the technique relies on the intrinsic bleaching and blinking behaviors characteristic of all commonly used fluorescent probes. To detect single fluorophores, we simply acquire a stream of fluorescence images. Fluorophore bleach or blink-off events are detected by subtracting from each image of the series the subsequent image. Similarly, blink-on events are detected by subtracting from each frame the previous one. After image subtractions, fluorescence emission signals from single fluorophores are identified and the localizations are determined by fitting the fluorescence intensity distribution with a theoretical Gaussian. We also show that BaLM works with a spectrum of fluorescent molecules in the same sample. Thus, BaLM extends single molecule-based superresolution localization to samples labeled with multiple conventional fluorescent probes.

Single-molecule localization microscopy (SMLM), such as stochastic optical reconstruction microscopy and (fluorescence) photoactivated localization microscopy, has enabled superresolution microscopy beyond the diffraction limit. However, the temporal resolution of SMLM is limited by the time needed to acquire sufficient sparse single-molecule activation events to successfully construct a superresolution image. Here, a novel fast SMLM technique is developed to achieve superresolution imaging within a much shortened duration. This technique does not require a faster switching rate or a higher activation density, which may cause signal degradation or photodamage/bleaching, but relies on computational algorithms to reconstruct a high-density superresolution image from a low-density one using the concept of blind image inpainting. Our results demonstrate that the technique reduces the acquisition time by up to two orders of magnitude compared to the conventional method while achieving the same high resolution. We anticipate our technique to enable future real-time live cell imaging with even higher resolution.

Localization-based superresolution microscopy techniques such as Photoactivated Localization Microscopy (PALM) and Stochastic Optical Reconstruction Microscopy (STORM) have allowed investigations of cellular structures with unprecedented optical resolutions. One major obstacle to interpreting superresolution images, however, is the overcounting of molecule numbers caused by fluorophore photoblinking. Using both experimental and simulated images, we determined the effects of photoblinking on the accurate reconstruction of superresolution images and on quantitative measurements of structural dimension and molecule density made from those images. We found that structural dimension and relative density measurements can be made reliably from images that contain photoblinking-related overcounting, but accurate absolute density measurements, and consequently faithful representations of molecule counts and positions in cellular structures, require the application of a clustering algorithm to group localizations that originate from the same molecule. We analyzed how applying a simple algorithm with different clustering thresholds (tThresh and dThresh) affects the accuracy of reconstructed images, and developed an easy method to select optimal thresholds. We also identified an empirical criterion to evaluate whether an imaging condition is appropriate for accurate superresolution image reconstruction with the clustering algorithm. Both the threshold selection method and imaging condition criterion are easy to implement within existing PALM clustering algorithms and experimental conditions. The main advantage of our method is that it generates a superresolution image and molecule position list that faithfully represents molecule counts and positions within a cellular structure, rather than only summarizing structural properties into ensemble parameters. This feature makes it particularly useful for cellular structures of heterogeneous densities and irregular geometries, and allows a variety of quantitative measurements tailored to specific needs of different biological systems.

Super-resolution (SR) imaging with high-throughput is invaluable to fast and highprecision profiling in a wide range of biomedical applications. However, prevalent SR methods require sophisticated acquisition devices and specific imaging control, and may cost a fairly long time on a single field-of-view. These essentially increase the construction difficulty, including challenges in imaging throughput, system establishment, and automation. Using the natural photophysics of fluorescence, fluctuation-based microscopy techniques can routinely break the diffraction limit with no need for additional optical components, but its long acquisition time still poses a challenge for high-throughput imaging or visualizing transient organelle dynamics. Here, we propose an SR method based on the Auto-Correlation with two-step Deconvolution (SACD) that reduces the number of frames required by maximizing the detectable fluorescence fluctuation behavior in each measurement, with further removal of tunable parameters by a Fourier ring correlation analysis. It only needs 20 frames for twofold lateral and axial resolution improvements, while the SR optical fluctuation imaging (SOFI) needs more than 1000 frames. By capturing raw images for ~10 minutes, we record an SR image with ~128 nm resolution that contains 2.4 gigapixels covering an area of ~2.0 mm × 1.4 mm, including more than 2,000 cells. Beyond that, by applying continuity and sparsity joint constraint, the Sparse deconvolution-assisted SACD enables 4D live-cell SR imaging of events such as mitochondrial fission and fusion. Overall, as an open-sourced module, we anticipate SACD can offer direct access to SR, which may facilitate the biology studies of cells and organisms with high-throughput and low-cost. In quantitative biology, the microscopy-based high-throughput screening is used to monitor the variability of biological systems as well as examine the heterogeneity 1 , and the advances in three-dimensional (3D) resolution can contribute to the minimization of the uncertainties in analyzing noisy biological processes. In parallel, super-resolution (SR) microscopy techniques relying on the blinking of single fluorophores have been developed to break the diffraction limit, including photoactivated localization microscopy (PALM) 2 and stochastic optical reconstruction microscopy (STORM) 3 . However, precisely localizing individual fluorophores requires tens of thousands of frames to accumulate one final SR view, which inherently limits the throughput and leads to requirements for purpose-built experimental settings 4 . Therefore, although livecell PALM/STORM has been reported 5-7 , excessive illumination power (~10 kW/cm 2 ) 8 , long exposures (>2 s) 9 , and the particular photochemical environment that promotes long dark states to reduce bleaching 9 prevent it from being a general live-cell imaging method 10 with high-throughput.

New technologies are ready to revolutionize glomerular imaging and significantly improve or replace immunofluorescence and electron microscopy, which have driven research and diagnosis of glomerular diseases for over 50 years. Advanced forms of transmission and scanning electron microscopy have revealed the detailed spatial relationships of the glomerular basement membrane, podocytes, and endothelial cells. These may be overshadowed by super resolution microscopy (SRM), which combines the advantages of immunofluorescence and electron microscopy, offers high resolution identification of specific molecules, and images large, physiologically relevant volumes of the glomerulus. Rapidity, ease of use and low cost with some types of SRM make them potentially suitable for routine diagnosis. SRM visualizes structures below the classical diffraction limit of conventional light microscopy by adding a time variable to either the illumination of the specimen, or to the fluorescence signal emitted by it. Ensemble techniques vary illumination and include Structured Illumination Microscopy (SIM) and Stimulation Emission Depletion Microscopy (STED). Single molecule localization techniques vary the light emission by fluorescence labels in the specimen, and include Photoactivated Localization Microscopy (PALM) and Stochastic Optical Reconstruction Microscopy (STORM). Technologies such as expansion microscopy and genetic labeling can also create effective super resolution imaging by non-optical, specialized preparation techniques. All technologies require dark field fluorescence and some require computer image analysis and reconstruction. Replicating successful application in other areas of biology, SIM, STED, and STORM have visualized normal and nephrotic disease podocytes, and have confirmed their appearances to be similar to those seen by electron microscopy, but with added new information on cell configuration and protein localization. STORM has also localized podocyte cytoskeleton and adhesion proteins, and glomerular basement membrane proteins at a resolution never before possible. These pioneering efforts show the promise of super resolution microscopy, and lay the groundwork for future study and new diagnostic tools for glomerular diseases.

PALM and STORM: Into large fields and high-throughput microscopy with sCMOS detectors

In super-resolution localization microscopy, e.g., stochastic optical reconstruction microscopy or photoactivated localization microscopy, a long acquisition time is required because of stochastic imaging nature, which limits its application in dynamic imaging for live cell. To overcome the limitation, one approach based on compressed sensing (CS) has been used in the previous reports. However, the imaging performance obtained by this method may be affected due to the use of interior point method (IPM). To address the problem, in this paper, we introduce an alternative CS reconstruction method and apply the recently developed YALL1 (your algorithm for L1 norm problems) method to super-resolution imaging model. Two types of numerical simulation experiments were performed to evaluate the performance of the proposed method. In case 1, the microscopy data from a single frame was simulated, which was used to evaluate the performance of YALL1 in single-emitter detection. In case 2, the dynamic microscopy data from a series of time points was generated, which was used to evaluate the performance of YALL1 in resolving fine structures. The results show that compared with the previous reported IPM method, the localization accuracy of super-resolution is improved by the proposed YALL1 method, even if there is high emitter density and noise in measurement data. In addition, the imaging time can also be reduced, because fewer imaging cycles are required for reconstructing the final super-resolution image by YALL1 method. Hence, the technique provides the potential in imaging fast cellular processes.