Fluorescence resonance energy transfer (FRET) (1), which used to be a biophysical technique strictly for specialists a few decades ago, has become a standard tool to study molecular conformation, interactions, and colocalization at the nanometer scale in cell biology and related fields (2–9). This penetration is exemplified by new editions of cell biology textbooks covering FRET among classical methods such as coimmunoprecipitation or magnetic resonance (MR) (10). The increased popularity of FRET is due to several factors including the availability of brighter, more photostable dyes, quantum dots, and color variants of fluorescent proteins, and the development of new features, add-ons and dedicated hardware and software that are available for commercial microscopes (11) and flow cytometers (12), and facilitate the smooth application of FRET techniques. In addition to the fact that FRET measurements can be applied to detect molecular associations (2–9, 13), the FRET phenomenon can be used to improve spectral characteristics of luminescent probes to be utilized in cytometry (14, 15). In this commentary we will address two concerns frequently encountered in FRET imaging: the need for being able to dissect molecular populations displaying different FRET efficiencies (16), and the surmounting of artifacts that may arise due to molecular diffusion during the FRET experiment (17, 18).
FRET is the nonradiative transfer of excitation energy from an excited, fluorescent donor dye to a (not even necessarily fluorescent) acceptor in the 2–10 nm vicinity of the donor. The probability of the process is characterized by the FRET efficiency (E), which is a measure of the fraction of excitation energy quanta transferred from the donor to the acceptor. Because of the dependence of the rate of FRET on the negative 6th power of the donor–acceptor distance, FRET is also called a spectroscopic ruler (19) allowing sensitive determination of inter- and intramolecular distances under ideal conditions. Because of FRET, the fluorescence intensity and lifetime of the donor dye decrease, whereas its anisotropy increases (provided that the donor and the acceptor are two different dyes and there is no significant back transfer of energy from the acceptor to the donor) (20). At the same time, the fluorescence intensity of the acceptor increases (sensitized emission) and its anisotropy decreases.
Different methods to evaluate the FRET efficiency take advantage of one or more of these spectroscopic changes (3). We can classify them, somewhat arbitrarily, into (a) intensity-based methods measuring donor quenching and/or sensitized emission, and (b) lifetime-based methods determining the fluorescence lifetime components directly or instead, a spectroscopic quantity related to the lifetime (such as the characteristic donor photobleaching rate in a donor photobleaching experiment).
Subcellular inhomogeneity of protein–protein interactions necessitates the resolution of high-FRET and low-FRET molecular populations within the same cell. However, by using classical intensity based methods no distinction can be made between the situation of a single population of donor–acceptor pairs characterized by a single intermediate FRET efficiency value, and a mixture of high-FRET and low-FRET donor–acceptor pairs producing the same average FRET efficiency as in the first case. The reason is that in these measurements photons are collected in an uncorrelated fashion, just the total amount of photons during the measurement time is recorded from a pixel (in microscopy) or a cell (in flow cytometry). To dissect molecular populations, photons should be collected in a time-correlated fashion, i.e. either lifetime-based, or single molecule techniques must be used.
Today's standard lifetime based FRET technique is fluorescence lifetime imaging (FLIM), which is gaining widespread application in biology as a result of commercially available instrumentation. Because the primary effect of FRET is the reduction of the donor lifetime, FLIM is considered to be one of the least artifact-prone FRET methods. FLIM, such as the measurement of fluorescence lifetimes in general, has two main versions: time domain and frequency domain techniques, which are the Fourier transforms of each other. In the first case the flow of photons following a short excitation pulse is detected in a time-correlated fashion, whereas in the latter case the amplitude of the excitation light is modulated and the consequent modulation of the donor fluorescence is analyzed for its amplitude and phase (21–25). Both techniques allow the fitting of models assuming several lifetime components, which may arise from different microenvironmental conditions or various local donor–acceptor distances. Usually for one image either the same model is applied for each pixel, assuming the same number of exponentially decaying components, or a tiresome procedure is applied to test different models in selected regions of interest by hand. To take full advantage of the ability of the technique to distinguish lifetime components, Spriet et al. have introduced an algorithm to select the best fitting model at each pixel independently (16). The principle of the method is the (automated) evaluation of a parameter related to the χ2 of the fit, which helps determine if the addition of a new lifetime component significantly improves the fit. Since the reliability of information gained from FLIM greatly depends on the correct choice of the number of lifetime components (too many components make the fit unstable and meaningless, while too few components yield a loss of information), and the subcellular heterogeneity of interactions can be expected in the majority of biological systems, this improvement can greatly contribute to the seamless application of FLIM. A recent development in FLIM is imaging spectroscopic fluorescence lifetime imaging microscopy (ISFLIM), which often allows unambiguous assignment of FRET (26). Despite some obvious advantages of the FLIM-based FRET approach, its widespread application is hindered by the limited availability of sophisticated fluorescence lifetime microscopic equipments. With the help of the more available standard camera-based or confocal microscopes photobleaching or intensity-based FRET measurements can be performed without the need to modify the already existing systems.
The simplest, technically undemanding method allowing the distinction of multiple lifetime components is donor photobleaching (27–29). In this technique, the donor molecules are exposed to continuous or repeated illumination leading to gradual photobleaching due to an irreversible reaction with oxygen, while the decay of fluorescence is detected. Since bleaching can only occur in the excited state of the dye, the mean donor lifetime is a key determinant of the survival probability and the characteristic photobleaching rate of the donor. In fact, usually even single donor-labeled samples can only be fitted reasonably with two rather than one exponential decay components. The advantage of the method is its simplicity and that there is no need for dedicated instrumentation. Disadvantages include its destructive nature, the need for an external (donor-only) control, and long measurement times leading to blurring of information due to diffusion.
The simplest intensity-based FRET method is the measurement of donor quenching by comparison of a donor-only labeled sample to a doubly (donor–acceptor) labeled one. Other intensity-based techniques may rely on the irreversible or reversible switching on or off of the acceptor, such as in acceptor photobleaching, photoactivation/photoconversion (3). These methods in their simplest form necessitate only the evaluation of donor fluorescence intensities. On the one hand, these methods are straightforward and simple. On the other hand, due to a need for taking several measurements consecutively on the same sample, they have a poor time resolution. In addition, they have an intrinsic error due to diffusion occurring between recording the images needed for the analysis or, alternatively, they necessitate the use of fixed samples, which might alter the FRET efficiency because of immobilization artifacts (30).
To avoid the blurring effect of molecular diffusion during the FRET measurement, faster intensity-based techniques can be used. A useful approach is the simultaneous detection of three signals (donor, transfer, and acceptor), allowing for a spectral decomposition of the pure donor and acceptor signals as well as the determination of FRET efficiencies on a cell-by-cell or pixel-by-pixel basis. The method can be applied both in flow cytometry ensuring high statistical accuracy and speed (4), and in image cytometry providing subcellular resolution (31). When using a confocal microscope, the excitation wavelengths can be switched line by line, letting little time (a few milliseconds) for molecular diffusion to occur between recording the donor, transfer and acceptor signals. For this reason and because this method is minimally destructive, it can be applied to live, non-fixed samples or used for time-resolved monitoring of physiological processes with confidence. The method necessitates the determination of equivalent amounts of fluorescence arising from the same number of donor and acceptor dyes in the different channels, which can be difficult when using independently expressed fluorescent proteins (GFP or its derivatives) as labels. Recently, a flow cytometric and an image cytometric method have been developed to get around this difficulty (17, 18). In both cases, fusion proteins of donor and acceptor expressed from the same promoter in a single polypeptide (and therefore present in equal quantities) were applied. By the application of these methods the mapping of interactions between CFP- and YFP-labeled membrane proteins, and GFP- or mRFP-labeled transcription factors have been demonstrated. A great advantage of these methods is that they do not need any special equipment, they can be carried out using a standard flow cytometer or a confocal microscope having the appropriate laser lines. FRET efficiency values measured between fluorescent protein pairs have also been used as a decision base of cell separation (32) or to monitor functional assays (33).
Camera-based single molecule FRET techniques have the potential to unify some of the advantages of lifetime-based and intensity-based methods, i.e. to dissect FRET populations and overcome diffusion related problems. Single molecule sensitivity in fluorescence microscopy has been achieved with the advance of sensitive, low noise detectors, and high wavelength dyes emitting in the far red/infrared spectral range, where cellular autofluorescence is low. Now it is possible to gain information on the molecular distribution of FRET efficiencies with the application of time-correlated single-photon counting (TCSPC) techniques, where in the optimal case signals can be detected from single donor–acceptor pairs at a time. In the multiparameter fluorescence imaging approach (34) photon bursts arising from fluorescent particles diffusing through the field of view are analyzed in a versatile way to gain maximum information: lifetime, intensity (brightness), anisotropy, FRET, excitation and emission spectra, spatial position and the dynamics of intensity fluctuations are evaluated, the latter yielding diffusion coefficients via fluorescence correlation spectroscopic (FCS) analysis at each pixel. The technique holds great promise for future cellular applications. To take advantage of the single molecule sensitivity, single molecule conditions are required (very low concentration), which can be achieved in an artificial system, but are often not available in cells having higher protein expression levels. Stochastic bleaching in the whole field of view would reduce the amount of donor–acceptor pairs, thus being unfeasible for FRET. A possible approach is the application of a special bleaching based technique, TOCCSL [Thinning out Clusters while Conserving Stoichiometry of Labeling (35)]. In this technique, fluorescence is completely bleached in a sharply defined area, and intact labeled molecules or molecular aggregates subsequently entering this area by diffusion are detected one by one. This technique, extended to two colors, can be applied to FRET.
An interesting application is when FRET is combined with total internal reflection fluorescence microscopy (TIRFM) conditions (36). The developed imaging system allows screening of large numbers of cells under TIRFM illumination, providing the means to image FRET efficiencies from donor–acceptor labeled membrane-associated proteins with a high signal-to-noise ratio. With this system high throughput analysis of stoichiometric FRET constructs can be performed on live cells (37). High throughput FRET combined with subcellular resolution and the acquisition of morphological information can also be achieved by the application of slide-based cytometry, which is a more informative alternative for flow cytometry (38).
The above mentioned examples demonstrate that a plethora of technical approaches are available to monitor FRET efficiency in cellular systems. To determine which FRET approach to apply, one should consider the characteristics of the experimental systems under investigation, the quality, and quantity of the required information and the availability of appropriate equipment. With the help of FLIM-based FRET measurements lifetime components can be distinguished, and interacting molecular populations assigned to different lifetime components can be dissected. The algorithm introduced by Spriet et al. (16) will help utilize more efficiently this capability of FLIM-based FRET measurements.