Live-Cell Super-Resolution with dSTORM
Fluorescence Nanoscopy in Living Cells with Photoswitchable Dyes
- Fig. 1: The composite image shows the superior spatial resolution of dSTORM imaging. The dSTORM principle combines photoswitching of organic fluorophores (upper right) with stochastic activation and single-molecule localization (lower right). Photoswitching occurs if a fluorophore transits from the fluorescent (or “bright”) state into a metastable dark state, e.g. via reduction to a radical anion or a further reduced state by reaction with thiols. In d STORM imaging, the majority of the fluorophores is switched into a metastable dark state, whereas only a subset of fluorophores resides in its fluorescent state and is recorded at a given time. The activation of single fluorophores occurs stochastically and such that they are separated by > µ/2. The position is determined by approximating the emission profile with a 2D Gaussian function with a precision of a few nanometers. The whole procedure is repeated sufficiently many times, and a high resolved image is reconstructed from the sum of all single-molecule positions (“localizations”).
- Fig. 2: (A) Widefield and (B) dSTORM images of microtubules labeled with the fluorophore ATTO 520, and (C, D) widefield and (E) dSTORM images of mitochondria labeled with Alexa 647 in COS7 cells.
- Fig. 3: dSTORM image of the core histone H2B in a living HeLa cell labeled with (A) TMP-ATTO 655, (B) SNAP-Cell TMR-Star and (C) SNAP-Cell 505.
- Fig. 4: (A) Absorption spectrum showing the formation of a blue-shifted radical anion of the rhodamine fluorophore Alexa 488 after irradiation with 488 nm in presence of a thiol-containing reducing agent. (B) Single-molecule photoswitching of Alexa 488 phalloidin, labeled to the actin skeleton of a U373 cell. Direct excitation of the radical anion with very low light intensities at 405 nm substantially increases the number of dyes returning to the fluorescent state.
- Fig. 5: dSTORM imaging of the microtubule network in a COS-7 cell labeled with Alexa 647. A sufficient number of single-molecule localizations have to be recorded in order to obtain a representative image of the labeled structure.
Live-Cell Super-Resolution with dSTORM: A detailed microscopic characterization of cellular structures is important to understand cellular function. Conventional microscopy in some cases is limited by the achievable spatial resolution of about 200 nm in the imaging plane, which is not sufficient to reveal details at the near-molecular level. This is important if the organization of proteins in small organelles, clusters or machineries are studied. Here, advanced microscopic techniques that are capable to bypass the resolution limit can offer additional and valuable information and pave the way to a new level of understanding. Methods that employ stochastic activation and read out of photoactivatable or photoswitchable fluorophores are in particular interesting because they can achieve near-molecular resolution, can be operated with moderate irradiation intensities and even applied inside live cells.
Super-Resolution Microscopy of Cellular Structures
The limit in light microscopy is to about 200 nm, such that objects which are closer cannot be discerned. In the past years, a variety of super-resolution fluorescence imaging techniques have been developed which can bypass the resolution limit in microscopy . These technologies have made the important step from pure demonstration towards real-life applications, with a particular focus on cellular structures. Among the different super-resolution methods available today, so-called stochastic methods benefit from a simple experimental realization and provide near-molecular spatial resolution. The key of these techniques are photoactivatable or photoswitchable fluorophores: a controllable number of only a very few fluorophores is activated stochastically e.g. by light, the fluorescence signal is read-out with sensitive detectors (such as single-photon sensitive EMCCD cameras), and the position of individual fluorophores is determined to a precision of a few nanometers. This procedure is repeated until a sufficient number of single-molecule coordinates were collected, from which a "pointillistic" image is finally reconstructed (fig. 1).
Direct Stochastic Optical Reconstruction Microscopy (dSTORM)
A very simple approach for stochastic super-resolution imaging is direct stochastic optical reconstruction microscopy [2, 3]. dSTORM operates a large variety of unmodified and widely used organic fluorophores as photoswitches, on the basis of light-induced photophysical transitions and redox-chemical processes that are inherent to these probes.
The main advantage of dSTORM is that it can easily be combined with well-established labeling techniques such as immunofluorescence to visualize cellular structures (fig. 2). Additionally, small and specific markers for cellular structures can be used, such as the fluorophore-labeled heptapeptide phalloidin to label actin structures. Organic fluorophores are brighter than fluorescent proteins, which leads to a better spatial resolution reaching 20 nm in the imaging plane. Additionally, organic fluorophores can be cycled many times between a fluorescent and a non-fluorescent state and imaged repeatedly without irreversible photobleaching. This fact is in particular helpful for studies of dynamic processes, which in contrast to static samples require more than one read out of a fluorophore to follow. Likewise, the photoswitching kinetics of organic fluorophores can readily be controlled, which enables fast imaging  that is essential for the observation of faster dynamics. For example, a spatial resolution of 30 nm together with a temporal resolution of 1 Hz has been used to study actin filament dynamics on a myosin surface .
dSTORM in Live Cells
The challenge for all super-resolution technologies is their application in living cells under moderate excitation conditions and reasonable acquisition times. Here, stochastic methods have the advantage of applying comparably low irradiation intensities on the one hand, and providing a spatial resolution of about 20 nm on the other hand. A first demonstration of live cell imaging used photoactivatable fluorescent proteins which were genetically attached to a protein of interest . However, fluorescent proteins are less bright than organic fluorophores, and in most cases can only be read out once and with relatively low imaging speed. These parameters limit both, the achievable resolution (which is directly related to the number of photons emitted by a single fluorophore) as well as the temporal resolution and thus their suitability to observe dynamic processes. At this point, organic fluorophores represent a promising alternative for stochastic imaging inside living cells [3, 7], making use of their inherent photophysical and photochemical properties in reducing experimental conditions as they are found with e.g. the tripeptide glutathione in the cytosol and nucleus in all kinds of cells (fig. 3).
Labeling Live Cells with Organic Fluorophores
Next to fluorophores with appropriate photoswitching properties, live-cell dSTORM requires approaches to label target molecules with high specificity. Recently, we have demonstrated live-cell dSTORM imaging of core histone H2B that was fused to a chemical tag (eDHFR) to introduce the oxazine fluorophore ATTO655  (fig. 3A). Other tags, such as the SNAP- and CLIP-tag, and other organic fluorophores, such as a tetramethylrhodamine (TMR), can readily be used  (figs. 3B, 3C).
The Mechanisms behind Photoswitching of Organic Fluorophores
Among the various stochastic technologies introduced in the recent past, dSTORM has proven to be the most flexible and versatile approach. A main reason for this is the refined understanding of intrinsic or induced photophysical and photochemical properties of organic fluorophores, which have finally paved the way to fast imaging [4, 5], multi-color  and live-cell applications [7, 8]. Briefly, stochastic super-resolution methods require that from the ensemble of fluorophores in a sample, only a very few are activated at a given time (and their position determined), whereas the large majority remains in a stable non-fluorescent state with a lifetime of seconds . These non-fluorescent states very recently were found and characterized for many organic fluorophores to be of radical nature or further reduced states  (fig. 4A). This finding is surprising and important at the same time: first, the radical states of organic fluorophores are remarkably stable in aqueous solution and inside living cells, thus ideal for stochastic super-resolution; and second, radical ions exhibit a blue-shifted absorption band that can be excited e.g. at 405 nm for the rhodamine fluorophores Alexa 488 and tetramethylrhodamine (TMR), which adds another means to control the number of fluorophores that are stochastically activated (fig. 4B). Additionally, the described photo-induced processes are not a side reaction, yet, occur for most if not nearly all fluorophores with high reliability and minimal photobleaching. This fact is crucial and often underestimated for a precise and quantitative description of cellular structure with near-molecular resolution (fig. 5).
The authors are thankful for funding by the Biophotonics and the Systems Biology Initiative (FORSYS) of the German Ministry of Research and Education (BMBF, grants 13N9234 and 0315262).
 Heilemann M., J. Biotech. 149, 243-251 (2010)
 Heilemann M. et al., Angew. Chem. Int. Ed. 47, 6172-6176 (2008)
 Heilemann M. et al., Angew. Chem. Int. Ed. 48, 6903-6908 (2009)
 Wolter S. et al., J. Micr. 237, 12-22 (2010)
 Endesfelder U. et al., Chem. Phys. Chem. 11, 836-840 (2010)
 Shroff H. et al., Nat. Meth. 5, 417-423 (2008)
 Wombacher R. et al., Nat. Meth. 7, 717-719 (2010)
 Klein T. et al., Nat. Meth. 8, 7-9 (2011)
 van de Linde S. et al., J. Biotech. 149, 260-266 (2010)
 van de Linde S. et al., Photochem. Photobiol. Sci. DOI: 10.1039/C0PP00317D (2011)