Fluorescence microscopy is ubiquitous these days. Open a life science journal, and you're bound to see photomicrographs in shades of red, yellow, green, and blue.
Most of this work involves intensity analysis, inferring the abundance of a protein or magnitude of an event from the strength of the color. But fluorescent labels generate two other types of signals, too. One is the lifetime of the fluorophore's excited state; the other is the wavelength of its emitted light.
Microscopists and microscope manufacturers have developed tools and techniques that allow researchers to probe these signals, enabling them to add depth and nuance to their studies. As a result, multiprotein associations can now be monitored, while autofluorescence worries are a thing of the past. But such improvements aren't cheap; lifetime-measuring equipment costs between €22,500 and €55,000, while spectral-imaging microscopes could cost $300,000 or more.
FLUORESCENCE LIFETIME IMAGING MICROSCOPY
By measuring the time a...
THE TWO TYPES OF FLIM
Courtesy of C. Peukert, K. Konig, F. Schiller
Optical sections of human skin showing fluorescence lifetime data (color scale at left).
The two major FLIM variants are frequency-domain and time-domain FLIM. The frequency-domain method modulates both the source (a continuous laser beam or high-power LED) and the detector, shifting the phase of the emitted beam relative to the excitation beam. As the tangent of the phase shift is proportional to the lifetime of the excited fluorophore, the latter can be calculated.
Time-domain FLIM (properly called time-correlated single-photon counting) uses a pulsed laser beam to deliver short pulses of photons and a photon-counting detector to measure lifetimes directly. "The frequency-domain method is much faster but has lower efficiency than single-photon counting," says Enrico Gratton, principal investigator of the Laboratory for Fluorescence Dynamics at the University of Illinois, Urbana-Champaign. "So people use one or the other, depending on the type of measurements they want to make."
Either technique can be performed with a one-photon, multiphoton, wide-field, or confocal microscope. But FLIM is more difficult to learn than other types of microscopy, Periasamy says, because users need to understand the biophysics as well as the biology involved. Once the finer points are mastered, however, the data are easier to interpret than those obtained by traditional microscopy, which sometimes requires mathematical modeling to remove spectral bleedthrough when imaging, e.g., protein-protein interactions.
Courtesy of Alison North, Rockefeller University
An example of separating (unmixing) three red fluorescent dyes (Rhodamine-phalloidin, Rhodamine-Red-X, ToPro3). The red image represents a conventional bandpass filter image without separation.
Most commercial equipment is designed for time-domain FLIM, and several companies sell the components: Becker & Hickl and PicoQuant in Germany; TauTec and Olympus in the United States; Lambert Instruments and Nikon Europe in the Netherlands; and (for detectors) Hamamatsu Photonics in Japan and PerkinElmer in the United States.
According to Giovanni Biscotti in scientific customer support at Becker & Hickl in Berlin, an upgrade kit for a microscope starts at €22,500 and includes a single-photon counting card, a high-speed detector suitable for photon counting, and a control card. A system equipped for the best time and image resolution, including analysis software, setup, and support, can reach €55,000. A user might also need to install an extra microscope port, and must have a pulsing laser with a high repetition rate (20–80 MHz).
Periasamy says FLIM is unnecessary for investigators who simply want to determine whether a protein-protein interaction occurs, but it is invaluable for making quantitative measurements or studying interactions among more than two proteins. Eliceiri says potential users should be well grounded in imaging before attempting FLIM and that trying out various types of equipment and software are essential first steps. Both Periasamy and Gratton offer training to potential users.
Spectroscopy with a confocal microscope measures the third type of signal from an excited fluorophore, its wavelength. The technique is particularly useful when multiple fluorescent proteins need to be detected in a single experiment, because conventional confocal microscopy cannot cope with more than four fluorophores at one time. Spectral imaging is also valuable when autofluorescence is likely to confound imaging results.
Unlike FLIM, spectral imaging requires a dedicated microscope. Newest to market is the Nikon Eclipse C1si. Unveiled last December, it costs about $245,000, including imaging software. According to Jeff Larson, a product manager at Nikon, it is the only microscope on the market that can acquire as many as 32 10-nm channels simultaneously, within two seconds. "With the others, you have to make multiple passes, which can cause photobleaching and change the shape of the spectrum," Larson says.
Becker & Hickl
PerkinElmer Life & Analytical Sciences
The C1si provides a bandwidth of 320 nm, which can be placed anywhere between 400 and 750 nm. Alternatively, it can provide 5-nm or 2.5-nm channels and 160-nm or 80-nm bandwidths. "A very narrow channel gives you the ability to unmix multiple, closely overlapping probes," Larson explains. "A wider channel gives you an increased signal-to-noise ratio." Nikon provides standard confocal detectors to allow the instrument to perform as a regular three-channel confocal microscope.
Leica produces two versions of its spectral-imaging microscope. The TCS SL starts at about $160,000; the upgradeable TCS SP2 starts at about $220,000. A unique feature of these microscopes is the scan head's acousto-optical beam splitter, which uses proprietary technology to direct light from the laser to the specimen and emissions from the specimen to the detector.
This beam splitter, which can be programmed for simultaneous use with up to eight laser lines, produces sharp bands and neutral transmission across the visible spectrum. "You could even excite green fluorescent protein with 488 nm and yellow fluorescent protein with 514 nm," says Frank Lie, applications and sales specialist at Leica Microsystems. "With conventional beam splitters, you can get certain combinations such as green and red excitation, but not all the different kinds."
Carl Zeiss' spectral imaging offering, the LSM 510 META, starts at about $300,000 but can also be obtained as a multiphoton microscope. It acquires all spectral information about a single pixel at once instead of sweeping across the specimen several times. Product manager Sebastian Tille says another advantage is the system's software, which can unmix overlapping emission spectra by several methods, including a process dubbed emission fingerprinting.
By comparing emission spectra with reference spectra from fluorescent dyes scanned with the same microscope, emission fingerprinting eliminates errors resulting from an instrument's own spectral properties. The software also excludes potential sources of error, such as autofluorescence and reflection. A user can select this method for postprocessing analysis or for obtaining an immediate result during acquisition.
Jeremy Skepper, director of the Multi-Imaging Centre at the University of Cambridge, UK, is familiar with the Leica and Zeiss instruments. Both do the same job, he says, but he uses a Leica because its capacity for postacquisition spectral unmixing, coupled with some computer wizardry by his group, enables him to cope with tens of gigabytes of data.
Courtesy of Ammasi Periasamy
Lifetime FRET imaging follows the changes in the fluorescent lifetime of a fluorescent donor molecule (CFP) in the presence (CFP-YFP-C/EBPa; tmean = 2.0 ns) and in the absence (CFP-C/EBPa; tmean = 2.65 ns) of the FRET acceptor (YFP). In contrast to confocal FRET imaging, no spectral bleedthrough correction is required.
Michael Davidson, research faculty member in optical microscopy at the National High Magnetic Field Laboratory at Florida State University in Tallahassee, says the biggest drawback with most spectral-imaging microscopes is the slow speed of the wavelength scanning systems, which can make live-cell imaging difficult. "For fixed samples, however, spectral imaging is by far the superior method, because you have far more control over your detector windows than with a barrier filter," he adds.
Davidson, who creates educational Web sites on all forms of optical microscopy, says the software used in some spectral imaging systems is still rather crude. "But as these instruments start to mature," he adds, "the linear unmixing software will get better, and advanced detection methods coupled to more efficient scanning will allow real-time imaging with live cells and fluorescent proteins."
Eliceiri envisions a microscope that could simultaneously record spectral, intensity, and lifetime information from a fluorophore in real time. The main challenge, he says, would be to store and analyze the many gigabytes of information such an instrument would collect. As part of a five-laboratory consortium, the Open Microscopy Environment, Eliceiri is trying to improve ways of storing four-dimensional data and eventually hopes to add spectral and lifetime data. "But it will be quite a few years before combined spectral-lifetime imaging becomes a practical technique," he predicts.
Meanwhile, Skepper advises biologists to hang onto their old microscopes. "FLIM and spectral imaging are going to be tools in addition to what we already have," he says. "They are very good for specific purposes, but are certainly not going to replace all of the intensity-based microscopes just yet."