Modern-day optical microscopists are breaking through the diffraction barrier, the long-held rule—imposed by the very nature of light waves—that limits how closely the cell’s inner workings can be scrutinized. Using clever tricks to sidestep the boundaries set by light’s innate tendency to spread, these scientists are zooming in to view the interior of organelles previously accessible only using labor-intensive electron microscopy. “It opens a new world,” says Stefan Hell, director of the Max Planck Institute for Biophysical Chemistry in Göttingen, Germany.
Conventional fluorescence microscopy cannot resolve a region smaller than some 200 nanometers across. That means that two objects closer than 200 nm will appear to be merged together. A single fluorescent macromolecule will also appear as a spread-out blur, so a 4-nanometer green fluorescent protein (GFP) will look like a 400-nanometer-wide smear, says Jennifer Lippincott-Schwartz, chief of the organelle biology section at the National Institute of Child Health and...
The crucial trick is to flip fluorophores on and off so the microscope only “sees” some of the sample at any given time. With several new super-resolution microscopes entering the market, biologists can now distinguish structures a few dozen nanometers in size. Here, The Scientist offers a taste of some top super-resolution methods.
PALM and FPALM: Fluorescence Photoactivation Localization Microscopy
Resolution: 20–30 nm
User: Petr Jezek, head of the Institute of Physiology, Academy of Sciences, Prague, Czech Republic
Project: Jezek studies the structure and organization of mitochondria. At about 250 nm wide, these tubular-shaped organelles are just at the resolution limit of conventional light microscopy. He collaborated with Joerg Bewersdorf, a biophysicist at Yale University, to look more closely at mitochondria with FPALM microscopy. Now, Jezek can see inside mitochondria to localize the many DNA-containing nucleoids. He hopes to image cristae, infoldings of the inner mitochondrial membrane, as well.
Technique: A pointillist painter assembles a picture out of tiny dots. Microscopists can apply the same principle if they arrange for only a few fluorophores to glow at a given time. PALM and FPALM rely on fluorescent proteins, such as PA-GFP and PA-mCherry, which remain dark until activated by ultraviolet light. If the activating light is very dim, only a few proteins will switch on. The system then uses a second laser to image those fluorophores. Each will look like a blurry spot, but a computer calculates the spot’s center—the molecule’s probable location. Once imaged, the fluorophores switch off permanently. With thousands of cycles of activation and imaging, researchers can assemble a pointillist picture based on individual fluorophores. The same principle works with antibody stains (see STORM).
• Since the fluorophores are only counted once, you can quantify how many proteins are in different parts of the sample.
• Researchers can track single particles in live samples to follow movement, with imaging times of 2 to 10 seconds, says Sam Hess, who developed FPALM at the University of Maine in Orono,
• Images might take a long time to acquire and process, ranging from a few seconds to 5 minutes, Hess says, or even up to half an hour, according to Lippincott-Schwartz, one of the developers of PALM.
• It takes several tries to get photoactivated proteins expressed at just the right levels for PALM, according to Bewersdorf. Fixed samples require careful preparation, akin to EM protocols, to preserve nanoscale structure.
Cost: Most labs that do imaging already have most of what they need to do FPALM, and could build a system from scratch for around $100,000, Hess says. “It is just a fluorescence microscope with two light sources and then an unusual choice of probes.” Researchers will also need software—which Hess is willing to share—to turn the thousands of individual blurs into a single crisp image. Zeiss plans to start shipping its PALM scopes this spring, starting at just under $500,000.
STED: Stimulated Emission Depletion
Resolution: 30–80 nm
User: Valentin Nägerl, Professor of Neuroscience and Bioimaging at the University of Bordeaux 2, France
Project: Nägerl studies the synapse, the electrical and chemical junction between two neurons. On the receiving cell, dendritic spines take a mushroom form, with a bulbous head attached via a thin neck. The neck’s diameter, it is thought, may regulate how signals are transmitted. The necks, however, are between 40- and 500-nm wide, too fine to resolve with regular microscopy. Nägerl teamed up with Hell, the inventor of STED microscopy, to zoom in on synapses. (PNAS, 105:18982–87, 2008)
Technique: Imagine shining a flashlight in the dark; the light spreads out from the lamp to illuminate a wide circle. If you wanted to illuminate a smaller area, you could place a doughnut in front of the flashlight, effectively blocking visibility for the objects on the edges. STED uses a similar principle, employing two scanning laser beams. The first beam activates fluorescent molecules. The second is the doughnut—a ring of red-shifted light that naturally quenches fluorophores activated by the first laser. By turning off the fluorophores at the edges, STED can minimize the area where molecules are actually emitting light, down to an 8-nm spot.
• STED requires no special image processing; what you see is what you get.
• STED might also be useful for photobleaching or photoactivation of fluorescent molecules, so researchers can watch the movement of a lit-up or darkened spot. Nägerl suggests that STED lasers could bleach or activate a smaller spot than such studies normally employ, allowing finer resolution of molecular behavior.
• Not all fluorophores work for STED; they must be both activated and deactivated by commercially available lasers and able to withstand repeated on/off cycles without photobleaching.
• The high-powered lasers might damage live samples.
Cost: A STED scope is just a confocal with an extra doughnut laser. Hell says it’s possible to make a do-it-yourself STED microscope for less than $40,000. Leica’s pulsed-laser STED system costs approximately $1 million, and a new continuous-wave laser system starts at $500,000. Leica will also upgrade a microscope for STED at a price of $250,000–$400,000.
SIM: Structured Illumination Microscopy
Resolution: 100 nm
User: Karl Oparka, head of the Institute for Molecular Plant Sciences at the University of Edinburgh, United Kingdom
Project: Oparka studies plasmodesmata, the small pores that connect plant cells. At 100 nm, “they’re tantalizingly below the limit of conventional imaging,” he says. Oparka used SIM to collect images of plasmodesmata as well finer threads of membrane running between plasmodesmata. “We weren’t quite aware of how intricate the network was,” Oparka says. (Plant Physiol, 153:1453–63, 2010)
Technique: From afar, your eyes cannot resolve the pattern on the grating of a screened window. But if two screens were overlaid, slightly offset, their patterns would combine to form a pattern you could see—lighter where the empty squares coincide, and darker where wires block the empty squares. This new, coarser pattern is called a moiré fringe. Now suppose you knew the pattern on one screen, but not the other; you could deduce the unknown screen’s pattern from the moiré fringes of the combined screens. SIM uses the same concept. “It makes otherwise unresolvable high-resolution information visible in the observed image, in the form of low-resolution moiré fringes,” says Mats Gustafsson of the Howard Hughes Medical Institute’s Janelia Farm in Ashburn, Virginia., one of the developers of this technology. In this case, the “unknown screen” is the sample, and the “known screen” is a pattern of light and dark lines in the illuminating light. A computer program can back-calculate, from the moiré fringes, what the unknown sample looks like.
• SIM works with any fluorophore.
• Because SIM can assemble a picture from only a few sets of fringes, imaging is fast—less than a second for a simple, two-dimensional sample, Gustafsson says.
• The system is prone to creating imaging artifacts. “You have to know your specimen extremely well,” Oparka says, to distinguish real structures from artifactual ones.
• The multiple image cycles needed to assemble an image may photobleach or damage a sample.
Cost: You can modify a fluorescence microscope for SIM with the right laser and a moveable grating. The key element is software that turns the coarse fringes into an image. Would-be users can write their own, or contact a SIM user willing to share software. Several companies offer SIM systems: The version from Applied Precision, Inc., starts at $500,000. Nikon Instrument’s N-SIM costs $500,000–$600,000, and the company can also modify recently purchased microscopes for SIM at a cost of approximately $400,000. Carl Zeiss Microimaging, Inc., expects to start shipping SIM scopes this spring, for about $600,000.
STORM: Stochastic Optical Reconstruction Microscopy
. Mennella, B. Huang and D.A. AgardResolution: 20–30 nm
User: David Agard, HHMI Investigator and Professor of Biochemistry and Biophysics at University of California, San Francisco
Project: Agard is interested in centrioles, which measure 250 nm across—just a spot on a regular microscope. He uses SIM to see some of the structures, but to really look closely, he collaborates with Bo Huang, a biophysicist at UCSF who was involved in developing STORM. “Now we can really resolve different layers and different substructures,” Agard says.
Technique: STORM works just like PALM, putting together an image from thousands of tiny points of light. STORM was developed using organic fluorescent dyes, such as Cy3 and Cy5, and PALM using genetically encoded fluorescence proteins, but the distinction is historical. “There’s really not much difference,” Huang says. The optics are the same, and PALM or STORM will work with either kind of tagging.
• Since organic dyes like Cy5 work with antibodies, it isn’t necessary to clone a new fluorophore onto the gene of interest.
• These dyes tend to be brighter than fluorescent proteins.
• Fluorescent molecules on antibodies will be 10 or 20 nm away from their epitope, whereas GFP or other tags are closer. At super-resolution, that difference may make antibody localization less precise.
• Dyes like Cy5 turn on and off repeatedly, and there may be multiple fluorophores on one antibody, so the antibody approach may count the same molecule several times. That makes quantification impossible.
Cost: Nikon’s N-STORM costs between $300,000 and $400,000; some recent Nikon microscopes are also eligible for a STORM upgrade for approximately $250,000.