Researchers would like to obtain wide-field images of biological processes with nanometer resolution without destroying the sample. Although methods that produce this level of resolution do exist, none meet all these conditions. However, postdoctoral researcher Alexey Sharonov and physical chemistry professor Robin M. Hochstrasser of the University of Pennsylvania in Philadelphia have now devised a way to perform rapid and relatively simple wide-field imaging at a nanometer scale.Their technique combines concepts used in other approaches: photobleaching, point-spread function measurements and intermittency caused by bimolecular collisions. The method depends upon a constant stream of fluorescent probes. As two molecules collide, they create a spike in the fluorescence signal that allows the location of each probe to be determined with nanometer resolution. These points map out the surface. The researchers’ earlier work — in which they investigated the dynamics of probes colliding with surfaces — led them to consider using probe flux for imaging. They also had identified the contours of nanoscale objects by accumulating single-molecule images. The new technique combines these ideas. In standard optical methods, resolution is limited by diffraction to about a half wavelength of the illumination source. However, the location of single molecules can be determined with greater precision if the molecules are fluorescent and act as a point source of light. Researchers then can capture the emission intensity distribution and fit it to the two-dimensional parameters of a point-spread function from an ideal point source. The result is improved accuracy related to the square root of the number of photons. Emitters have been located with nanometer precision using this method.The new method exploits the fact that, as fluorescent probes strike an object and become immobilized, they create a spot that fluoresces until the probe is photobleached or released. The density of probes on the surface must be controlled so that the probability of finding two probes within the diffraction-limited spot is low. Then each one can be treated as a point source. Achieving this condition can be accomplished by adjusting such parameters as the probe flux, the laser power or the frequency with which images are captured. The researchers used a setup from earlier studies to demonstrate the technique. “One of the beautiful things about this experiment is that it can be applied using existing microscope technologies,” Hochstrasser said. “We just needed the idea plus the design of software.”He noted also that it was vital to understand the diffusion-controlled collision dynamics in solutions. That knowledge allowed the researchers to accurately predict how many probes would end up at the target’s surface and thereby not exceed the density threshold.Capturing fluorescenceThey used an inverted microscope from Olympus of Melville, N.Y., with a krypton-ion laser operating at 568 nm. They directed the beam into the sample and conducted experiments in total internal reflection fluorescence microscopy mode with a bandpass filter and dichroic mirror from Chroma Technologies of Rockingham, Vt., filtering and steering the laser, respectively. After passing through bandpass filters and beamsplitters from Omega Optical of Brattleboro, Vt., the fluorescence was captured by a CCD camera with multiplication-on-chip capability from Roper Scientific of Tucson, Ariz.To localize the fluorescent molecules, the researchers used image processing algorithms developed in-house. The processing involved several steps designed to clean up noise, discriminate between fluorescent spots and the background and determine the location of the spot. The precision of that determination varied with conditions but was estimated to be about 50 nm, well below the diffraction limit. Locations were stored and later used to assemble a synthetic image. The average speed for this first step of processing was just under 120 points per second.Then, the researchers further refined the molecular locations determined in the first step by discarding fits that had too much noise. They fixed the remaining locations with an accuracy that could be as good as 1 nm, depending on the photon count from the point, and they formed another synthetic image. The greater accuracy that was achieved by this step came with a trade-off: almost a 10-fold drop in the number of points that could be processed per second.In a conventional Nile red fluorescence image (left) of a supported bilayer on glass, diffraction blurs the picture and obscures most detail. In a high-resolution synthetic image (right) obtained by locating 2778 single Nile red probes collected in 4095 frames, the outline of the bilayer is clear and details distinct. Images courtesy of Alexey Sharonov, University of Pennsylvania.In a series of experiments, the researchers imaged a variety of objects, for example, large unilamellar vesicles with the probe Nile red, choosing it in part because its collision kinetics with the vesicles are well known and because it emits when bound to a lipid but not when in water. They adjusted the concentration to achieve one to two bright spots per square micron, collecting 4095 frames of a 40-μm-square area at 50 frames per second. The resulting synthetic images resolved single vesicles less than 200 nm apart and other details not distinguishable with standard optics.A standard fluorescence image (left) and a synthetic high-resolution image (right) show vesicles, or cellular compartments, attached to a glass surface. The synthetic image, derived from the localization of individual Nile red probes mapping out the surface, shows vesicles with a center-to-center distance below the diffraction limit. The center bottom two, for example, are separated by about 200 nm.They also imaged a supported lipid bilayer using Nile red. Collecting images over a 188-square-μm area, they resolved features not visible using conventional optical methods. The work is detailed in the Dec. 12 issue of PNAS.The researchers are looking to apply the technique in a number of areas. Plans call for obtaining accurate shapes of large organic assemblies such as human plasma fibrinogen. Other projects involve DNA, where subdiffraction imaging of a large number of strands and rapid location of template structures are possible. To fully exploit the technique, some additional work must be done. “We need to explore properties of new probes, such as fluorescently labeled proteins as probes, and to find the protocols for the experiment in cells and cell membranes,” Hochstrasser said.