Multicolor Super-Resolution Fluorescence Imaging
Using Stochastic Individual Molecule Blinking
- Multicolor Super-Resolution Fluorescence Imaging - Using Stochastic Individual Molecule Blinking
- Fig. 1: Spectra of the special quadband fluorescence filter for multicolor super-resolution imaging using the stochastic individual molecule blinking of conventional fluorescent dyes on a STORM microscope. a) Spectrum of the quadline beamsplitter.
- Fig. 1: Spectra of the special quadband fluorescence filter for multicolor super-resolution imaging using the stochastic individual molecule blinking of conventional fluorescent dyes on a STORM microscope. b) Spectrum of the quadband emission filter.
- Fig. 2: Acquisition settings for multicolor super-resolution imaging using the stochastic individual molecule blinking of conventional fluorescent dyes on the Nikon N-STORM microscope.
- Fig. 3: Application of super-resolution fluorescence imaging using stochastic individual molecule blinking (images from Rösch et al. , modified). a) Super-resolution imaging of lipid droplets stained with Lipidtox Red.
- Fig. 3: Application of super-resolution fluorescence imaging using stochastic individual molecule blinking (images from Rösch et al. , modified). b) Two color superresolution image of lipids droplets (blue) and adipose differentiation-related protein (ADRP, red) in human hepatocytes. Lipid droplets were stained with LipidTox Red, ADRP was stained with Alexa Fluor 647.
- Fig. 3: Application of super-resolution fluorescence imaging using stochastic individual molecule blinking (images from Rösch et al. , modified). c) 3 color 3D super-resolution images of lipid droplets stained with Lipidtox Red (blue) and two different viral proteins of hepatitis C virus stained with Alexa Fluor 647 (red) and Alexa Fluor 488 (green) localizing around the lipid droplet. Images rendered in Bitplane Imaris.
Stochastic optical reconstruction microscopy (STORM) has the potential to increase the resolution in fluorescence light microscopy up to tenfold. This helps scientists to get new insights into biological processes and structural details. For STORM fluorophores are utilized that show a blinking behavior  switching between a fluorescent and a nonfluorescent state. This was initially described for pairs of cyanine dyes, one activator dye and one reporter dye  in close proximity to each other. For multicolor STORM the channels are separated by using different activator dyes while the reporter dye is the same in each channel .
Many fluorophores show an intrinsic blinking behavior without activator dyes under defined conditions . Here a way of multicolor super-resolution imaging on a standard STORM microscope using the stochastic individual molecule blinking of conventional fluorescent dyes without activator fluorophores is reported. A standard STORM microscope can be used by adding a specifically designed quadband fluorescence filter.
In optical microscopy the maximum achievable resolution is limited to approximately half of the wavelength of the light by the optical properties of the microscopes and the wave-like nature of light .
This diffraction limit has been overcome by recently introduced super-resolution light microscopic techniques like STORM , PALM , SIM  and STED .
For STORM first a time-series of raw images is acquired with the microscope. During this time-series the fluorescent molecules must show a blinking behavior. Due to the stochastic blinking there is only a minor probability that fluorescent signals of two neighboring molecules overlap in a single time frame. In this case the signals from single fluorescent molecules should appear as diffraction limited spots. The positions of the fluorophores can be computed with nanometer precision by applying an algorithm to find the centroid position of the non-overlapping spots. The sum of all positions computed for all images of the time series results in the final super-resolution image.
In the initial STORM approach the blinking is achieved using a pair of cyanine dyes, one activator dye and one reporter dye that are covalently bound to the same target molecule (e.g. an antibody) . Excitation of the activator dye with low power laser light is used to stochastically turn a subpopulation of the reporter dye from the dark state to the fluorescent state. Upon the following excitation of the reporter dye with high power laser light the fluorescent signal of the reporter dye is read out and the dye is transferred to the dark state again. This signal is used for the computation of the molecule positions. For multicolor STORM different activator dyes and the same reporter dye are used .
STORM is not limited to these pairs of cyanine dyes. Any fluorescent molecules showing two distinguishable states can be used for STORM. Many different commercially available fluorophores are reported to show a blinking behavior without an activator dye under defined conditions like special imaging buffers .
Multicolor super-resolution fluorescence imaging using the stochastic individual molecule blinking of conventional fluorescent dyes can be performed on a standard STORM microscope (here Nikon N-STORM). Only one special quadband fluorescence filter (Multi-STORM filter) has to be installed. The filter is composed of a quadline beamsplitter (zt405/488/561/640rpc TIRF, Chroma, spectrum see fig. 1a) and a quadband emission filter (brightline HC 446, 523, 600, 677, Semrock, spectrum see fig. 1b). This filter allows excitation of up to four different dyes and recording of its specified (blinking) emission by simply switching the excitation laser wavelength and without the change of any filtercube during the experiment. The acquisition is done with a single Andor iXon DU 897 EMCCD camera.
To achieve optimal switching behavior of the fluorophores an imaging buffer containing 100 mM MEA and an enzymatic oxygen scavenger system  is used.
The raw time series are acquired using the N-STORM acquisition program module (v. 220.127.116.11). First the laser settings for the different color channels have to be adjusted. This is done in the advanced settings of the acquisition program module (fig. 2). The conventional organic fluorophores used in this imaging approach can switch spontaneously from a non-fluorescent dark state to the fluorescent state. To enhance the switching speed from the dark state to the fluorescent state the 405 nm laser can be used as "activation" channel.
All color channels are imaged using the Multi-STORM filter. Therefore no filter changing i.e. mechanical movements are necessary. This avoids spatial signal shifts and leads to shorter acquisition times. The excitation spectra of distinct fluorophores (e.g. Cy5 and Alexa Fluor 488) should overlap as little as possible to avoid cross excitation.
To get super-resolution images from the raw time series the computation of the fluorophore positions is performed in the N-STORM analysis program module (v. 18.104.22.168). Identification settings have to be entered manually. As the software was designed for experiments with pairs of cyanine dyes, multicolor experiments with conventional fluorescent dyes need a workaround: In this case the computation process has to be done for each color channel separately. First the computation of the fluorophore positions for the first color channel (e.g. Cy5) is performed using suitable identification settings. After the computation the list with the positions of the fluorophores is saved as a "molecule list". Then the computation of the fluorophore positions for the next color channel (e.g. Alexa Fluor 488) has to be performed using appropriate settings for this channel. Again the molecule list has to be saved afterwards (under a different file name).
Then each "molecule list"-file must be loaded again. After loading one file the software displays a super-resolution image where all color channels are present. Only for one of these color channels the STORM image was computed using the right identification settings (e.g. for the Cy5 channel). For this channel the visualization settings (Gaussian rendering etc.) should then be adjusted and the "crosstalk correction" tool should be applied to correct for minor crosstalk. Afterwards the image can be exported to NIS Elements. The exported image again contains all color channels of which only the channel calculated with the right settings (in this case Cy5) can be used. For the next color channels the steps are the same as for the first channel. All exported images must have the same size and pixel size. Then they can be merged resulting in the final super-resolution image.
For a 3D image the steps are basically the same. Additionally for each channel an appropriate 3D calibration curve has to be loaded before the computation of the molecule positions. To get an image stack the "z stepping filter" in the analysis program module has to be activated. Then every z-slice has to be exported as a single image. This has to be performed for all color channels. These z-slices can be merged to a 3D image stack in NIS Elements.
Due to the 3D resolution improvement chromatic aberration in range of tens of nanometers becomes observable and has to be corrected. This can be done by attaching fluorescent beads (with the same spectra as the fluorophores) to the coverglass. Using these beads an axial offset between the different channels can be measured. This offset can be corrected by adding or removing slices in the one color 3D super-resolution images before merging them to a multicolor super-resolution 3D image.
Super-resolution fluorescence imaging using stochastic individual molecule blinking can be easily applied at standard STORM systems and provides more flexibility in experiments and applications. It is especially useful in cases where the creation of a pair of activator and reporter dye is not possible or very difficult. Examples are lipophilic dyes like LipidTox Red (Invitrogen) or intercalating dyes like Pico Green (Invitrogen). These fluorophores are not conjugated to antibodies and so the chemical coupling of an activator fluorophore is not possible. Figure 3 shows some applications in virus research.
The author would like to thank Kathrin Rösch and Eva Herker for the preparation of the hepatitis C samples and Rudolph Reimer for the critical review of the manuscript.
 Rust M. J. et al.: Nature Methods 3, 793-795 (2006)
 Bates M. et al.: Science 317, 1749-1753 (2007)
 Heilemann M. et al.: Angew. Chem. Int. Ed. 47, 6172-6176 (2008)
 Abbe, E.: Beiträge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung, Arch. F. Mikroskop. Anat. 9, 413-420 (1873)
 Betzig E. et al.: Science 313, 1642-1645 (2006)
 Gustafsson M. G. L.: J. Mircosc. 198, 82-87 (2000)
Hell S. W. and Wichmann J.: Optics Letters 19, 780-782 (1994)
 Rösch K. et al.: Three-color super resolution microscopy of core, NS5A, E2 and lipid droplets in HCV cc-infected cells. Poster at the "20th International Symposium on Hepatitis C Virus and Related Viruses", Melbourne, Australia (2013)
Dennis Eggert (corresponding author via e-mail request button below)