Putting into practice a microscopy scheme proposed a decade ago, researchers have imaged fluorescent proteins within cells with 2- to 25-nm resolution. That is far below the classical resolution limit of half a wavelength, which translates, for example, to a few hundred nanometers in the visible. The investigators, led by Eric Betzig and Harald F. Hess of Howard Hughes Medical Institute’s Janelia Farm Research Campus in Ashburn, Va., developed the technique, dubbed photoactivated localization microscopy. The key to achieving the resolution is that the imaged fluorophores, which are a small subset of the total fluorescent molecules present, are scattered across the sample, and the molecular signal is much larger than any background noise. Thus, in every frame, the fluorescent molecules that are visible are far enough from each other that their diffraction-limited images, which are about 200 nm wide, do not overlap. These isolated approximations of the ideal point spread function of the fluorescence originating from a particular molecule allow the center of each spot to be determined more precisely, with the error decreasing by the square root of the number of photons captured. Because well more than 10,000 photons can be detected from a single fluorophore before it photobleaches and disappears from view, the location uncertainty for a molecule can be cut a hundredfold with this technique. Besides offering higher resolution than that possible with other optical methods, the approach offers contrast-related benefits when compared with electron microscopy. To perform the technique, the researchers used photoactivated fluorescent proteins that were expressed by the cell that they studied. Thus, there was no need for the preparation required for electron microscopy, which can sometimes disturb the sample. In addition, Betzig noted, the method is more discriminating in its labeling than that used in electron microscopy. “The [photoactivated fluorescent proteins] are produced within the cell bound to only a single target protein of interest and, therefore, give a very detailed map of the spatial distribution of only that protein,” he said. As described in the Aug. 10 edition of Sciencexpress, the researchers serially activated the fluorescent proteins using a brief pulse from a Coherent Inc. diode laser at 405 nm. They chose the intensity and duration of the pulse so that only sparse fields of individually resolvable single molecules were activated. They then used a 561-nm laser from Lasos Lasertechnik GmbH to image each fluorescent protein in turn until the proteins photobleached. They captured the images using a cooled electron-multiplying CCD camera from Andor Scientific, fitting the observed profile at each location to the ideal 200-nm-wide point spread function. In building up a single image of a thin section of a cell or of the interface between a whole cell and the substrate on which it rested, they repeated the process of activation and photobleaching as many as 100,000 times. Capturing a frame typically took between a half second and a second, which meant that acquiring a complete image stack took from two to 12 hours. Mitochondria (red) as revealed by photoactivated localization microscopy are overlaid on a transmission electron microscopy image of the same region. Worried about nanometer-scale shifts during these long data runs, the scientists built a custom microscope head designed to reduce thermal and mechanical drift by minimizing the optical path length between the sample and the objective lens. They added 50-nm gold beads to the sample and localized their position in every frame to compensate for any drift, an approach that Betzig said worked well. “These fiducial markers were so effective that we can now contemplate applying photoactivated localization microscopy to commercial microscopes of lower mechanical and thermal stability,” he said. In demonstrations, the researchers imaged specific target proteins in lysosomes and mitochondria. They visualized the former via expression of the lysosomal transmembrane CD63, while they imaged the latter using a fluorophore-tagged oxidase import sequence. They compared both with transmission electron microscopy images, which showed a high degree of correlation to those acquired with photoactivated localization microscopy. They also imaged various targets in fixed whole cells, which again showed that the technique compared well with electron microscopy. Resolution limitations of conventional optical microscopes lead to blurry images of small cellular components such as mitochondria (top level). However, if specific proteins are tagged with fluorescent molecules that can be activated one at a time (between top and second level), the positions of the molecules be can determined with much greater precision. Their positions then can be used to assemble a photoactivated localization microscopy image (second level) at a resolution comparable to that of an electron microscope (bottom level), with the added benefit of being specific to the target protein. Downsides of the approach are that it suffers from long data collection times and that the photoactivated proteins do not all produce the same number of photons. As a result, the localization precision varies and, therefore, the highest-resolution image is the one with the fewest number of labeling fluorophores. Betzig noted that data collection might be sped up by boosting the laser excitation power, lessening the time spent per single molecule, or perhaps by increasing the number of activated molecules per frame. It is unclear how much they can reduce the total time, but they hope to get as low as five minutes for a single image. As for the localization problem, oxygen-scavenging techniques might increase the number of photons per molecule. There also is the possibility of using more photostable activatable labels or other approaches that could amplify the photon flow. Betzig said that he saw photoactivated localization microscopy as a research tool in cell biology, in particular, nanoscale protein interactions that drive cellular functions. However, he added that more work is required before it can fulfill that role. That is one reason why the investigators are planning to develop a multilabel version of their technique as well as to find ways to increase the data acquisition speed. Another goal is to improve the ease of use, turning the system into one that biologists could use routinely. “If we get to that point, commercialization is a definite possibility,” Betzig said.