When two or more fluorescent proteins in a cell are less than 200 nm apart, their separate signals will be indistinguishable by the average fluorescent microscope and will appear as one bright blob. The new technique devised by Betzig et al., called photoactivatable localization microscopy (PALM), gets around this fundamental problem using two main tricks.
The first trick is to isolate molecules by viewing just a few at a time. This technique can be likened to asking a few scattered people in a crowded auditorium to stand up briefly, then another few people, and so on, until each person in the room has been momentarily distinguished from the seated mass. The second trick is to take tens of thousands of photos while this is going on and afterwards let the computer work out who was sitting where.
To distinguish individual fluorescently labeled proteins from the crowd inside a cell, the team used photoactivatable fluorescent proteins (PA-FPs) and activated just a few at a time. PA-FPs remain essentially invisible until they are stimulated with a particular wavelength of light (activation light). By delivering a brief, weak dose of activating light to the sample, only a few PA-FPs become fluorescent. These proteins are then viewed using light of a different wavelength (excitation light) until their fluorescence is completely bleached and they are no longer visible. The process is repeated with slowly increasing doses of activation light until all molecules have been activated, viewed, and bleached.
The process takes anywhere from 2–12 hours per sample and generates 104–105 images. The 2D images are then plotted against time. If one imagines the plot as a 3D stack of images, with time = 0 at the top, then the activation and subsequent bleaching of each molecule resembles a separate 3D blob. The computer then turns these blobs into well-defined dots by calculating their center point. If the stack of images were to be flattened into one picture without first converting blobs to dots, the resulting image would appear much like an ordinary, blurry fluorescent microscope image. Instead, the well-defined dots give the flattened final image a clear, highly resolved appearance.
Not all blobs are of equal brightness and clarity, of course, and the computer is able to recapitulate these variations in the final image using some nifty statistical calculations. As Betzig explains, the final image is “a probability density map, where brightness is proportional to the probability that a PA-FP molecule exists at a given location.”
The PALM technique has been made possible thanks to the developments of both PA-FPs and total internal reflection fluorescence (TIRF) microscopy. In TIRF, directing light to a sample at an extremely oblique angle results in the illumination of only very thin sections. The benefit of PALM is far less background noise from out-of-focus light and cellular autofluorescence.
The technique is also possible due to a fortuitous characteristic of gold beads: they glow under the excitation light. Precisely localizing molecules at the nanometer scale is clearly incompatible with movement, but avoiding mechanical or thermal drift at this level over such long periods of time is virtually impossible. To get around this problem, the team added gold beads to their samples. “We just sprinkle them onto the sample like salt,” says Betzig. During the image capturing, this gold-bead seasoning acts as a constantly glowing frame of reference. This simple solution is good news for cell biologists hoping to use the technique, since it circumvents the need for expensive temperature- and vibration-controlled devices.
The authors predict that PALM might eventually be used for localizing proteins in 3D. Currently, however, the technology is lacking. The common method of rendering 3D images of fixed cells or tissues, confocal microscopy, produces large amounts of out-of-focus light, leading to too much peripheral bleaching. “You'd lose all the information you want,” explains Betzig. One other possibility might be to perform PALM on serial cryosections and then construct a 3D image retrospectively. Realigning the sections might be very difficult, however, as gold beads are too tough for microtomes and thus cannot be used as a physical reference.
Although the researchers have so far only looked at single-labeled proteins, the ultimate goal for PALM is to visualize two proteins in two colors at once. But Betzig explains that, presently, “the palette of photoactivated proteins is not all that broad. The success [of two-color PALM] will be dependent on our collaborators' coming up with better labeling techniques. We're close, we're really close,” he says, “but we're not there yet.”