Jeffrey Perkel has been a scientific writer and editor since 2000. He holds a PhD in Cell and Molecular Biology from the University of Pennsylvania, and did postdoctoral work at the University of Pennsylvania and at Harvard Medical School.
When it comes to divining the secret lives of cells, researchers often turn to microscopy. Using fluorescence confocal or TIRF (total internal reflection fluorescence) imaging, these scientists can probe the structure and organization of cells with startling clarity, and the resulting images provide dramatic swatches of color on otherwise drab pages of the biological literature. Beautiful as they are, those images still suffer from one significant limitation: the diffraction limit.
A consequence of the physical properties of microscope optics and the wavelength of light passing through them, the diffraction limit restricts how close two objects can be and still be resolved. The actual limit in any given microscope setup depends on the particular objective and wavelength of light being used. But in practice, says Christopher O’Connell, super-resolution system products manager at Nikon Instruments, “If two objects are closer than about 200 to 250 nanometers apart, you will not be able to distinguish them.”
Consequently, two macromolecular assemblies that are only, say, 150 nm apart will appear to be a single particle rather than two, blurring the image and suggesting colocalization where none in fact exists.
Recently, microscopists at the technology’s leading edge have developed work-arounds to this problem, producing so-called “super-resolution” microscopes with resolutions from twice as good to nearly 10 times better than the diffraction limit. They certainly aren’t inexpensive, but for researchers with the need, and who are willing and able to accommodate the constraints, these instruments deliver unprecedented clarity.
PALM, STORM and GSDIM
At the National High Magnetic Field Laboratory at Florida State University, microscopy expert Michael Davidson runs a lab equipped with some 30 microscopes, including seven confocal instruments. He also has one super-resolution microscope, a Nikon Instruments N-STORM, which he uses to study DNA replication foci in mammalian cells. He clearly likes the instrument, calling it “dynamite” and “a dreamboat.”
STORM, or “stochastic optical reconstruction microscopy,” is a kind of localization microscopy, a technique that produces super-resolution images not via enhanced optics per se but essentially using a mathematical trick. Suppose you have a sample with a single fluorophore. The spot that that fluorophore produces on the detection CCD is bigger than the molecule itself—like everything else in the microscope, it is diffraction-limited—but if you zoom in enough, you can see that the spot also is not uniformly bright. Instead, the pixel intensities resemble something of a three-dimensional Gaussian peak. By identifying the center of that peak, you can position a single molecule with subpixel accuracy to exceed the diffraction limit.
Unfortunately, most samples contain many more than a single fluorophore. If you allow them all to light up, this peak mapping becomes problematic—it simply isn’t possible to localize every point. It would be better if you could turn on just a few fluorophores, map their positions, turn on a few more fluors and so on. That, essentially, is what STORM and other localization technologies, including PALM (photoactivated localization microscopy) and GSDIM (ground state depletion with individual molecule return), do.
Developed by Harvard researcher Xiaowei Zhuang, STORM traditionally involves labeling samples with pairs of fluorophores: an activator and a reporter (such as Cy3 and Cy5). Initially, the reporter dye is dark. Illumination of the activator with a specific wavelength of light causes it to turn on the reporter dye, priming it to fluoresce when the sample is hit with another (excitation) wavelength. By illuminating the sample in such a way that only a few dye pairs are activated at a time, researchers can cause just a few molecules to light up at once. These are then imaged until they photobleach, removing them from the equation. Then the next set of dyes is activated, imaged and bleached, and so on, until thousands of different localization events are collected and mapped, producing an image with resolution on the order of 20 to 50 nm.
PALM, commercialized by Carl Zeiss Microscopy in the ELYRA P.1, works essentially the same way, except using genetically encoded, photoactivatable fluorophores, says Duncan McMillan, director of biosciences product marketing at Carl Zeiss Microscopy. Many such fluorescent proteins exist, including PA-GFP. Normally, green fluorescent proteins fluoresce when excited with blue (488-nm) light. But PA-GFP is different; it must first be activated by a higher-energy wavelength, such as 405 nm. Once activated, the molecule is, again, primed for excitation and imaging.
Similarly, GSDIM, available from Leica Microsystems as the Leica SR GSD, keeps fluorophores in a kind of dark “triplet” state with a high-powered light source, says marketing manager Chris Vega. The triplet state is distinct from the usual ground and excited states of fluorophores, and molecules in this state tend to stay that way for a relatively long time, Vega explains, until they stochastically decay into an “on” configuration in which they can be imaged.
The exceptional resolution of localization approaches depends on the number of photons collected and thus comes at the expense of speed and data-set bloat. It might take 18,000 frames to collect a complete data set for just one image, McMillan says: 600 seconds (10 minutes) multiplied by 30 frames per second produces 18,000 frames, each measuring one megabyte in size. As a result, localization technologies are typically limited to imaging fixed (i.e., dead) samples. (Though Xiaowei Zhuang did publish a Nature Methods paper in 2011 illustrating three-dimensional, live-cell PALM, collecting 2D images at up to 0.5 seconds and 3D images at 1 to 2 seconds.)
According to Vega, the Leica SR GSD features a novel stage design specifically to simplify and accelerate data processing. Because localization images are built from thousands of frames, these systems require that the sample remain immobile during imaging. They often use “fiduciary markers,” such as gold microbeads, that help the software align images even if they “drift” due to thermal changes. Leica circumvented that issue in the GSD with its SuMo stage design.
“We’ve eliminated the need for [fiduciary markers] by decoupling the objective from the stand itself and coupling it to the stage, so that the sample, stage holder and objective are one independent unit,” Vega says.
Because of that, he says, the GSD can image-process on the fly, allowing researchers to see the image during acquisition rather than waiting until after all the data has been collected.
STED
Leica has commercialized another super-resolution approach based on a completely different strategy. Stimulated Emission Depletion (STED) microscopy uses two overlapping lasers to limit the size of the sample area capable of fluorescing, thus effectively increasing resolution.
The Leica TCS STED CW, explains Vega, is based on a point-scanning confocal system. (In contrast, localization methods typically are built on TIRF platforms.) At each pixel, an excitation laser illuminates a diffraction-limited spot, as in normal confocal microscopy. But a second laser, shaped like a donut, effectively deactivates the fluors at the periphery of the excitation spot (like a photomask), reducing the fluorescing spot to a super-resolution point as small as 50 nm.
SIM
Commercialized by Nikon (N-SIM), Zeiss (ELYRA S.1) and Applied Precision (DeltaVision OMX Blaze), structured illumination microscopy (SIM) is perhaps the most complicated super-resolution approach. According to O’Connell, SIM actually involves optical improvements to generate its super-resolution images.
SIM, he explains, is a widefield approach in which the specimen is repeatedly imaged with specially “patterned” excitation light in a minimum of three orientations at three different phases, for nine (or more) total images. In so doing, the system produces a kind of ghost image (a Moiré pattern) that is directly related to the original image; the SIM software then back-calculates to determine what the original image looked like. (For a good visual representation of this, visit this SIM tutorial or check out Davidson’s excellent Molecular Expressions website.)
SIM typically produces images with about twice the resolution of confocal microscopy—about 100 nm or so, which is somewhat poorer than localization techniques. But SIM is far faster and thus amenable to live-cell imaging.
“If you are doing a lot of confocal stuff and want to double the resolution, then SIM is a great way to go, especially with fixed cells,” says Davidson, who will be adding a Zeiss ELYRA PS.1, capable of both PALM and SIM, to his microscopy lab shortly.
Hari Shroff, chief of the Section on High Resolution Optical Imaging at the National Institute of Biomedical Imaging and Bioengineering, doesn’t buy commercial instruments; he builds his own (including a 3D PALM system now in use at the National Institutes of Health). Shroff says the advantages of SIM are the wide range of fluorescent dyes compatible with the technique and its speed.
“With PALM or STORM, it’s not uncommon to spend minutes imaging one plane. With SIM, we can acquire a super-resolution image in a second, meaning you can follow live processes in 4D,” Shroff says—albeit not quite as quickly as with conventional imaging techniques.
Purchasing considerations
When making a purchasing decision, resolution is obviously a key factor. But it isn’t the only one.
According to O’Connell, a major consideration when selecting a super-resolution microscopy system is: Do you plan to image only fixed cells, or live cells as well? In the former case, speed is irrelevant. But for live-cell imaging, speed counts, at least if the biological process being imaged is highly dynamic. “People have published PALM images of live cells,” McMillan says, “but you cannot image rapid dynamics…. Live-cell and rapid dynamics are not necessarily the same.”
Another consideration is your choice of fluorophores. Does the system have the appropriate illumination sources for your fluorophores or fluorescent proteins of interest? More to the point, is the technique compatible with the fluors you regularly use, or will you need to switch to a new set of dyes?
According to Vega, for instance, GSDIM works with standard fluorophores, whereas PALM requires genetically engineering the cell with photoactivatable fluorescent proteins, and STORM traditionally requires “nonstandard” photoactivatable fluorophore pairs. (That said, the N-STORM also supports a newer variant called “direct STORM,” or dSTORM, which does use standard dyes, O’Connell says; the ELYRA P.1 supports dSTORM, as well.)
Are there other imaging modalities you hope to couple with super-resolution microscopy? Nikon offers a highly modular system, O’Connell says, meaning researchers can purchase systems combining SIM and STORM, confocal and STORM, and confocal and SIM. “So you can take images with all these modalities,” he notes.
And what about 3D imaging? If you only plan to collect two-dimensional images, any system will work (assuming the lateral resolution is acceptable). But the systems have different constraints on the how deep into a sample they can scan, and how extensive a “z-stack” they can acquire. Localization techniques, for instance, are typically built on TIRF systems, which limit imaging to within a few 100 nm of the coverslip, whereas SIM can image out to about 20 microns, according to McMillan.
Ultimately, when preparing to spend several hundreds of thousands of dollars, it pays to do your homework. See if you can take the systems out for a test drive using your own samples. In the meantime, for excellent interactive tutorials and in-depth explanations on super-resolution microscopy, check out this 2010 review, "A guide to super-resolution fluorescence microscopy", in the Journal of Cell Biology, as well as Molecular Expressions’ pages on both Nikon and Zeiss super-resolution solutions.