Practical Considerations for Super-Resolution Microscopy

Optimize nanoscale imaging at every step

(See a list of the products featured in this article.)

Fluorescence microscopy offers an unparalleled view into the spatial and temporal functioning of cells in an intact system. Until recently, however, the fundamental diffraction properties of light limited the resolution of standard fluorescence microscopy to a few hundred nanometers. In the last few years, concurrent innovations in optics, instrumentation, and software have circumvented this limit, providing nanometer-scale views into the arrangement and functions of intracellular components (Figure 1). These methods include structured illumination microscopy (SIM) and stimulated emission depletion (STED) microscopy, as well as single-molecule localization techniques such as photoactivated localization microscopy (PALM), stochastic optical reconstruction microscopy (STORM), direct STORM (dSTORM), and ground-state depletion followed by individual molecule return (GSDIM).

For an excellent primer on super-resolution microscopy (SRM) techniques, see the article by Galbraith and Galbraith [1]. There are a number of critical considerations when preparing samples and labeling for SRM. Several recent reviews have provided detailed protocols and discussions of these considerations [2–5]; here we will summarize the main points from these protocol guides.

Figure 1. Improvement in resolution with STORM imaging. Widefield (diffraction-limited) and STORM images of a BSC-1 cell expressing a GFP-SEC61B fusion, which labels the endoplasmic reticulum, and stained with an Alexa Fluor™ 647 conjugate of an anti-GFP antibody. Images courtesy of Joshua Vaughan, University of Washington, and acquired at an SRM workshop on a custom-built STORM microscope.

Sample preparation

Optimal sample preparation is a critical step in any fluorescence microscopy workflow. With SRM techniques, any nonspecific binding or perturbation of cellular structure will be even more noticeable. It is imperative to select the optimal fixative for the structure under investigation. Ice-cold methanol, glutaraldehyde, and formaldehyde have all been successfully used as fixatives, but the best fixative choice is dependent on each primary antibody and must be empirically determined [2,5]. Allen and coworkers [2] and Whelan and Bell [5] provide detailed protocols and recipes for each fixation protocol, as well as examples of images resulting from suboptimal fixation [5]. In addition to the fixative choice, the temperature and subsequent wash protocols will also affect image quality and should be optimized for a given label [5].

Nonspecific labeling is especially problematic in SRM, and optimizing the blocking protocols for each antigen is critically important [2,3,5]. A possible protocol addition is a prefixation extraction step [4,5]. During extraction, the plasma membrane is permeabilized, thereby removing cytoplasmic components that may bind nonspecifically to the primary antibody [4]. Note that an extraction step does not improve the fidelity of labeling for every antibody; for example, extraction has been shown to be beneficial for labeling of microtubules but not mitochondria [4].

The final step in sample preparation is the labeling of the structure of choice with a fluorescent dye. When using a secondary antibody conjugated to a fluorophore, it is important to consider that the large size of an IgG can limit your effective resolution [3] and that an additional antibody step introduces another point of nonspecific binding. Wherever possible, use highly cross-adsorbed secondary antibodies to reduce nonspecific binding [4]. The use of antibody fragments, such as Fab and F(ab´)₂ [2] or nanobodies [6], should be considered because these labels are smaller in size than an intact IgG. Furthermore, fluorophores can be coupled directly to primary antibodies using a number of approaches, such as NHS ester conjugation [4,7], eliminating the need for a secondary antibody.

Fluorophore performance

Optimal fluorophore choice will vary with the application and the SRM strategy used. Here are practical considerations for some of the different SRM techniques.

Structured illumination microscopy (SIM)—In SIM, the sample is first illuminated with patterned light, and then the information in the Moiré fringes that lie outside of the normal range of observation is analyzed. Reconstruction software deciphers the images to give a resolution limit of about 100 nm (two-fold higher resolution than the diffraction limit of 250 nm). Unlike techniques such as STED and STORM, SIM does not require fluorophores with specialized photochemical properties. Because of this property, SIM lends itself to multicolor imaging (Figure 2). Given the illumination protocols required for SIM, however, the more photostable the dye, the better the data. Photostability profiles for dyes under standard microscopy techniques are a good guide when choosing dyes for SIM (Figure 2D). Bright, photostable fluorophores that match the excitation and emission filters on the given SIM system should be selected. This recommendation is especially relevant for time-lapse or 3D SIM, where numerous images are acquired in either time or space. The addition of a high–refractive index mounting medium that contains an effective antifade reagent improves photostability and therefore SIM image quality. Also note that SIM is a widefield technique and therefore image quality will decrease in thicker samples; consequently, samples should be thin and mounted as close to the coverslip as possible.

Stimulated emission depletion (STED) microscopy—STED microscopy uses two laser pulses to localize fluorescence at each focal spot; the first pulse is used to excite a fluorophore to its fluorescent state, and the second pulse is a modified beam used to de-excite (through stimulated emission depletion) any fluorophores surrounding the excitation focal spot. STED is a universal property of all fluorescent dyes and proteins—the challenge is to find those dyes photostable enough to withstand the high-intensity illumination required while also matching the excitation and depletion laser lines available on commercial STED systems. Despite these criteria, many fluorophores work well for STED, including several Molecular Probe™ dyes—Oregon Green™ 488, Alexa Fluor™ 488, Alexa Fluor 532 (Figure 3), Alexa Fluor 568, and Alexa Fluor 594 dyes—as well as GFP. For dual-color STED, it is often beneficial to pair a dye exhibiting a standard Stokes shift with a dye exhibiting a long Stokes shift (e.g., Alexa Fluor 488 and Pacific Orange™ dyes [8]).

Single-molecule localization microscopy (STORM, dSTORM, and GSDIM)—Single-molecule localization techniques place the most stringent demands on fluorophores. Dyes need to be induced to “blink” from a dark state to an on state, thereby enabling the precise localization of the actual position of each fluorophore from its point-spread function. The SRM method STORM utilizes stochastic activation and time-resolved localization of photoswitchable fluorophores to generate high-resolution images. Photoswitching dyes must have high photon outputs per switch, coupled with a low duty cycle (i.e., they are in a non-emitting state longer than in an emitting state). With appropriate dye–buffer combinations, an optimized STORM system can generate images with 17 nm resolution [9].

Two forms of STORM exist. In the first method, two fluorescent dyes are coupled to the antibody of choice—an “activator dye” to induce switching and a “reporter dye” from which emission is detected. A good starting point, according to Bates and coworkers [7], is a degree of labeling (DOL) between 2 and 4 for the activator and 0.1 and 1 for the reporter (see “Applications for therapeutic monoclonal antibodies” for a summary of antibody labeling methods). For the activator–reporter STORM technique, many different Alexa Fluor dyes have been used as effective activators and both Alexa Fluor 647 dye and Alexa Fluor 750 dye have been shown to be reliable reporters. The availability of multiple activator–reporter pairs facilitates multicolor STORM applications.

The second method, commonly known as dSTORM or GSDIM, relies upon direct switching of a single fluorophore through specific excitation parameters. When using dSTORM or GSDIM, it is possible to use commercially available secondary antibodies with a DOL between 2 and 4 as a starting point [3]. Lower DOLs prove beneficial for densely labeled specimens; however, if the DOL (or indeed the labeling density with the primary antibody) is too low, then the continuity of the structure may be lost [3]. The accuracy of localizing individual fluorophores is critically dependent on the blinking properties. Many recent studies have characterized optimal dye properties and buffer conditions for single-molecule detection, and these are succinctly summarized by Allen and coworkers [3]. For single-molecule localization techniques, STORM, dSTORM, and GSDIM imaging of most dyes occurs in an oxygen-buffering system, such as a glucose oxidase and catalase with a primary thiol (e.g., β-mercaptoethylamine (MEA)) [3,4,9]. Figure 4 shows single-molecule GSDIM imaging using direct switching of either Alexa Fluor 568, Alexa Fluor 660, or Alexa Fluor 680 dye.

Alexa Fluor 647 dye is used extensively in dSTORM (Figure 1) because it exhibits extremely good photoswitching properties—namely, high photon output, cycle number, and survival fraction, along with a low duty cycle; it is the most forgiving dye for the novice user [4] and has been used in test samples to optimize STORM systems [10]. Moreover, even when the fluorescent label used is not ideal for single-molecule detection, such as occurs with a GFP chimera, the use of an anti-GFP antibody conjugated to Alexa Fluor 647 dye enables imaging of the GFP chimera by dSTORM or GSDIM (Figure 1). For two-color dSTORM, the best resolution is reportedly achieved with Alexa Fluor 647 and Alexa Fluor 750 dyes under phosphine quenching conditions [11]. It is also possible to perform two-color dSTORM with other Alexa Fluor dyes: for example, Alexa Fluor 568 and Alexa Fluor 647 dyes [12] or Alexa Fluor 532 and Alexa Fluor 647 dyes [13].

Figure 2. Multicolor SIM imaging.(A) Drosophila S2 cells labeled with Alexa Fluor™ 488 Phalloidin (blue), anti-tubulin antibody followed by a secondary antibody conjugated to Alexa Fluor 568 (magenta), and MitoTracker™ Deep Red FM (gold). (B) BPAE cell labeled with DAPI (red), Alexa Fluor 488 Phalloidin (purple), and MitoTracker Red FM (orange). (C) SIM imaging of HeLa cells labeled with anti–alpha-tubulin antibody (clone DM1A) conjugated to orange-red Alexa Fluor dyes (Alexa Fluor 546, Alexa Fluor 555, Alexa Fluor 568, or Alexa Fluor 594), with (D) bleach curves over 100 frames of SIM imaging (15 images per frame). Images courtesy of (A) William Voss and Catherine Galbraith, Oregon Health & Science University, and acquired on a Zeiss™ Elyra SR SIM system, and (B–D) Talley Lambert, Harvard Medical School, and acquired on a DeltaVision OMX™ 3D-SIM system.

Figure 3. 3D STED projection. COS-7 cells were fixed and labeled with anti-tubulin antibody followed by F(ab´)2 fragments conjugated to Alexa Fluor™ 532 dye (depletion wavelength, 660 nm). Image courtesy of Jana Doehner, University of Zurich, and acquired on a Leica™ TCS SP8 STED 3X system.

Figure 4. Use of Alexa Fluor dyes in single-molecule localization microscopy. COS-7 cells were fixed and labeled with anti-tubulin antibody followed by a secondary antibody conjugated to either Alexa Fluor™ 568, Alexa Fluor 660, or Alexa Fluor 680 dye, and imaged in the presence of GLOX and MEA (see Allen et al. [3] and Halpern et al. [4] for protocols). Widefield images (red) are overlaid with ground-state depletion followed by individual molecule return (GSDIM) images (green). Images courtesy of Jana Doehner, University of Zurich, and acquired on a Leica™ SR GSD microscope.

References

1. Galbraith CG, Galbraith JA (2011) J Cell Sci 124:1607–1611.
2. Allen JR, Ross ST, Davidson MW (2013) Phys Chem Chem Phys 15:18771–18783.
3. Allen JR, Ross ST, Davidson MW (2013) J Opt 15:094001.
4. Halpern AR, Howard MD, Vaughan JC (2015) Curr Protoc Chem Biol 7:103–120.
5. Whelan DR, Bell TD (2015) Sci Rep 5:7924.
6. Mace EM, Orange JS (2012) Commun Integr Biol 5:184–186.
7. Helma J, Cardoso MC, Muyldermans S et al. (2015) J Cell Biol 209:633–644.
8. Bates M, Jones SA, Zhuang X (2013) Cold Spring Harb Protoc 2013:540–541.
9. Dempsey GT (2013) Methods Cell Biol 114:561–592.
10. Metcalf DJ, Edwards R, Kumarswami N et al. (2013) J Vis Exp 79:e50579.
11. Vaughan JC, Dempsey GT, Sun E et al. (2013) J Am Chem Soc 135:1197–1200.
12. Agullo-Pascual E, Lin X, Leo-Macias A et al. (2014) Cardiovasc Res 104:371–381.
13. Furstenberg A, Heilemann M (2013) Phys Chem Chem Phys 15:14919–14930..