Fluorescence as a phenomenon is part of a larger family of related luminescent processes in which a susceptible substance absorbs light, only to reemit light (photons) from electronically excited states after a given time (Figure 1). Photoluminescent processes that are generated through excitation, whether this is via physical, mechanical, or chemical mechanisms, can generally be subdivided into fluorescence and phosphorescence.

Compounds that display fluorescent properties are generally termed fluorescent probes or dyes and de facto the term ‘fluorochrome’ is most appropriate. Often ‘fluorochrome’ and ‘fluorophore’ are used interchangeably. Strictly taken the term ‘fluorophore’ refers to fluorochromes that are conjugated covalently or through adsorption to biological macromolecules, such as nucleic acids, lipids, or proteins. Fluorochromes come in different flavors and include organic molecules (dyes), inorganic ions (e.g., lanthanide ions such as Eu, Tb, Yb, etc.), fluorescent proteins (e.g., green fluorescent protein), and atoms (such as gaseous mercury in glass light tubes). Some commonly used fluorochromes are listed in Table 1. Recently, inorganic luminescent semiconducting nanoparticles, quantum dots, have been introduced as labels for biological assays, bio-imaging applications, and theragnostic purposes—the combination of diagnostic and therapeutic modalities in one and the same particle (see reference [1]).

Fluorescence follows a series of discrete steps of which the outcome is the emittance of a photon with a longer wavelength—a process which can be visualized in more detail via the Jabłoński diagram in Figure 2B. When light of a particular wavelength hits a fluorescent sample, the atoms, ions or molecules therein absorb a specific quantum of light, which pushes a valence electron from the ground state GS₀—this initial state is an electronic singlet in which all electrons have opposite spin and the net spin is 0—into a higher energy level (Figure 1 and Figure 2B), creating an excited state ES_n. This process is fast and in the femtosecond range and requires at least the energy ΔE = E_ESn − E_GS0 to bridge the gap between excited and ground states in order for excitation to occur. The energy of photons involved in fluorescence and generally a quantum of light can be expressed via Planck’s law [2]:

(1)

where E is the quantum’s energy (J), h is Planck's constant (J.s), ν the frequency (s⁻¹), λ is the wavelength of the photon (m), and c is the speed of light (m.s⁻¹). However, there are several excited state sublevels (vibrational levels) and which level is reached primarily depends on the fluorescent species’ properties.

Irradiation with a spectrum of wavelengths generates numerous allowed transitions that populate the various vibrational energy levels of the excited states, some of which have, according to the Franck-Condon principle, a higher probability to occur than others (the better two vibrational wave functions overlap, the higher the probability of transition) and combined form the absorption spectrum of the fluorescent dye (Figure 3B).

After excitation to the higher energy level ES_n, the electron quickly relaxes to the lowest possible excited sublevel, which is in the picosecond range. The energy decay from dropping to a lower vibrational sublevel occurs through intramolecular non-radiative conversions and the converted heat is absorbed via collision of the excited state fluorescent molecule with the solvent molecules. Emission spectra are usually independent of the excitation wavelength because of this rapid relaxation to the lowest vibrational level of the excited state, which is known, as Kasha’s rule [3]. In most cases, absorption and emission transitions involve the same energetic levels as schematically shown in Figure 3B. For this reason and the fact that excitation is instantaneous, involving electrons only and leaving the heavier nuclei in place, many fluorochromes display near mirror image absorption and emission spectra (mirror image rule), although many exceptions exist, as exemplified by the spectrum of cis-parinaric acid in Figure 3. External conversion depletes the excited state through interaction and energy transfer to the solvent and/or solute. Intersystem crossing is the slowest energy dissipation pathway, because the electron has to change spin multiplicity from an excited singlet state to an excited triplet state (Figure 2B). This is essentially a spin forbidden transition, but in some fluorochromes favorable vibrational overlap between the two states makes the transition weakly allowed. Intersystem crossing is mostly observed in molecules containing heavy atoms such as iodine or bromine or in paramagnetic species.

Lanthanide ions are good examples of luminescent probes that show delayed fluorescence—The term “luminescence” is normally used when discussing lanthanide-based emission, because “fluorescence” refers to spin allowed singlet-to-singlet emission, whereas in lanthanide ions, emission is due to intra-configurational f-f transitions (transitions inside the 4f shell) [4]. As a result they have fluorescence lifetimes (the average time in the excited state) in the (sub)microsecond range for Yb(III) and Nd(III), whereas Eu(III), Tb(III) and Sm(III) have even longer lifetimes in the (sub)millisecond range [5]. This is significantly longer than the lifetimes of organic fluorochromes and fluorescent proteins which are generally in the nanosecond range [6]. Because of the large lifetime difference to fluorescence, lanthanide emission should easily be distinguishable from fluorescence emission by time-gating techniques. A critical review on lanthanides as probes can be found in [7].

When the electron finally returns to the lower energy level it originated from, the ground state GS₀, a quantum of light (photon) is emitted with a longer wavelength, which is an allowed transition since the spin is retained. In most fluorochromes, photon emission is produced by a π*→π or π*→n transition, depending on which requires the least energy for the transition to occur. In contrast, σ*→ σ transitions are rare since the required UV light (below 250 nm) is energetic enough to deactivate the excited state electron by predissociation (internal conversion to a GS₀ that is at such a high vibrational level that the bond breaks) or dissociation (the bond breaks at a vibrational excited state).

Photon emittance from singlet states with an average time-scale between 10⁻⁹ and 10⁻⁶ seconds is slow compared with the absorption of photons. This resulting emitted radiation is called fluorescence [8] and in the simplest form of fluorescence microscopy, this emitted light is collected and transported to the detector through the same objective lens used to focus the excitation light onto the sample, as schematically depicted in Figure 2A. Fluorochromes can enter repetitive cycles of excitation and emission as long as no destruction or covalent modification occurs that irreversibly interrupts this process. The transition from the meta-stable triplet state ET₁ to the ground state GS₀ is, because of the necessary spin reversal, forbidden and therefore several orders of magnitude slower than fluorescence, i.e., thus called phosphorescence. Consequently, the emission takes place over long periods of time—in some materials it takes several seconds to even minutes.

Because of the internal energy decay at the excited state levels and since the wavelength varies inversely with the radiative energy (Equation 1), fluorescence emission generally occurs at longer wavelengths and concomitant lower energy than the light used to excite the fluorochrome, provided that a single photon is involved during excitation. It was the British scientist Sir George Stokes who first observed fluorescence when irradiating fluorspar (fluorite) with UV radiation and a red-shift in the resulting emitted light, which he reported in his 1852 publication “on the change of refrangibility of light” [9]. The difference between the emission and excitation maxima is called “Stokes shift” as represented in Figure 3. The Stokes shift varies markedly among different fluorochromes. Additionally, it should be noted that anti-Stokes shifts in which the emission wavelength is smaller than the excitation wavelength are also possible, for instance during photon upconversion or two-photon excitation (discussed below). Anti-Stokes fluorescence also occurs commonly in fluorochromes in which absorption and emission spectra overlap substantially. In summary, fluorochromes not only have characteristic excitation spectra, but also characteristic emission spectra that are dependent on their specific vibronic configuration and properties [10,11].

This paragraph concisely introduces some of the essential characteristics and parameters of fluorescence that are utilized in fluorescence microscopy, such as the fluorescence life time, quantum yield, quenching, photobleaching, energy transfer, and others. Some of these fluorescent properties may be experienced by one researcher as an unwanted “artifact”, e.g., photobleaching or intensity loss via resonance energy transfer, whereas the same feature may be cleverly used by another to solve her/his scientific question, e.g., to study diffusion of molecules via FRAP or molecular interactions via FRET.

Quantum yield: The quantum yield (Ф) basically determines how bright a fluorochrome’s emission is. It is given by the ratio of the number of emitted to absorbed photons, which is determined by the rate constants of emission (Γ) and the sum of all non-radiative decay processes (k_nr) that depopulate the excited state (Figure 4). These decay processes, such as intersystem crossing (k_isc), internal conversion (k_ic), predissociation (k_pd), dissociation (k_d), and external conversion (k_ec), affect the fluorescence outcome, including quantum yield. The fraction of fluorochromes that dissipate the absorbed energy through emission can thus be written as (0 < Ф ≤ 1):

(2)

The quantum yield, as many intrinsic fluorescence parameters, is sensitive to environmental influences, such a solvent polarization, pH, fluorochrome concentration, and the presence of molecules affecting the excited state, such as molecular oxygen. The most reliable method for recording Ф is via the comparative method established by Williams et al. [12], which involves comparing the fluorochrome with well characterized standard samples with known Ф values. Please note that the quantum yield is sometimes incorrectly termed ‘quantum efficiency’, which refers to the efficiency by which photons hitting a photo-reactive surface will produce electron–hole pairs in photo-sensitive devices, such as a charge-coupled device (CCD) or solar cells.

Fluorescence lifetime: The fluorochrome’s fluorescence lifetime (τ) is the average time the electron spends in the excited state before returning to the ground state. In other words, the depopulation of excited state molecules through radiative (Γ) and non-radiative processes (k_nr) follows an exponential decay and the time of this process is basically the fluorescence lifetime:

(3)

The fluorescence intensity decay (I_t) over time (t) following an infinitesimally short excitation (δ-type) may thus be expressed as:

(4)

where I₀ is the initial intensity. During the lifetime, the fluorochrome may undergo conformational changes, diffuse, or interact with surrounding molecules, offering an opportunity to exploit lifetime measurements to probe such actions.

Anisotropy: This refers to the quality of having different properties along different axes: in a pool of randomly oriented fluorochromes, only those fluorochromes with transition dipole moments that are aligned (near) parallel to the polarization direction of the excitation beam (linearly polarized) will be excited (photoselection). Fluorescence polarization is determined by the orientation of the fluorochromes’ transition moment at the instant of emission, which allows the determination of the fluorochromes’ rotation by measuring their anisotropy. Theoretically, the maximal anisotropy is achieved when the emission transition dipole moment is exactly parallel to the absorption transition dipole moment.

Fluorescence quenching: Quenching is a phenomenon by which interaction of a molecule, the quencher, with the fluorochrome reduces the quantum yield or the lifetime (Figure 4). Quenching phenomena can be subdivided into:

Fluorescence intermittency or “blinking”: This phenomenon occurs when the fluorochrome randomly alternates between a fluorescent (“on”) and dark state (“off”) despite continuous excitation illumination. As a result, when tracking single fluorochromes, a stroboscopic effect may interfere with the tracking procedure. Blinking generally follows power law statistics and occurs commonly in luminescent nanoparticles, but also in some organic dyes, and fluorescent proteins. The random nature and power law dynamics generally frustrates and precludes comparison of results between independent experiments, because the “on” and “off” time distributions can significantly differ.

Autofluorescence: Fluorescence that does not originate from the fluorochrome of interest (FOI), but rather from cellular components that have fluorescent properties (background fluorescence), most notably flavins, e.g., FAD⁺ and riboflavin, NAD(P)H, and extracellular matrix components, e.g., elastin and collagen, but also retinol, folic acid, lipofuscin, and chlorophyll. The autofluorescence window approximately encompasses the range from 350 to 600 nm and can effectively be avoided by a number of strategies [13,14]: (i) precise filtering of the FOI signal with narrow bandpass optical filters, (ii) the use of probes that fluoresce outside the autofluorescence window, commonly in the (near) infrared ([N]IR) region, (iii) the use of time-resolved techniques (autofluorescent biofluorochromes have distinctly different lifetimes), (iv) the use of upconverting fluorochromes that allow excitation in the IR; (v) multispectral imaging (recording spectra in every pixel), and (vi) spectral unmixing (mathematical disentangling of mixed emission signals).

Photobleaching: Photobleaching or “fading” is a photochemical process in which the fluorochrome’s ability to enter repetitive excitation/emission cycles is permanently interrupted by destruction or irreversible covalent modification of the fluorochrome by reaction with surrounding (bio)molecules. Photodestructive and photochemical processes occur predominantly when the fluorochrome is in the dark triplet excited state, which is longer lived than singlet states and exhibits a high degree of chemical reactivity (Figure 4). Predominantly, reactions of the triplet state with molecular oxygen, which is a triplet state biradical (parallel spin) in the ground state, causes transformation to a reactive singlet state, i.e., singlet oxygen, formation of the superoxide anion radical and other reactive oxygen species (ROS), These oxidizing ROS not only bleach the fluorochrome, but are also cytotoxic to cells, because they irreversibly modify biomolecules.

Resonance energy transfer: This is a photophysical process in which the excited state energy from a donor fluorochrome is transferred via a non-radiative mechanism to a ground state acceptor chromophore via weak long-range dipole–dipole coupling (Figure 4). The theoretical basis for the molecular interactions involved in resonance energy transfer was first described by Theodor Förster in the 1940s [15,16], and requires that the donor’s emission spectrum overlaps the acceptor’s absorption spectrum and that donor and acceptor are in close proximity. This provides the foundation for FRET microscopy, as discussed in more detail below.

Charge-transfer complexes: These are nanosecond short-lived homodimers (excimer) or heterodimers (exciplex) of two molecules of which at least one is in the excited state that show red-shifted emission compared with the monomer’s emission. Such complexes occur via electrostatic attraction because of partial electronic charge transfer between the individual entities. The individual ground state monomers would normally not associate and the monomers in the complex dissociate and repel each other once relaxation to the ground state occurs. Since these complexes require close proximity, excimer and exciplex formation require high concentrations of the monomers.

Many different fluorescent techniques have since been developed that utilize the specific mechanisms involved in fluorescence, from the initial excitation to the emission of the photon, whether this is by exploiting the differences in fluorescence lifetime, e.g., distinguishing the fluorochrome’s signal from auto-fluorescence background or using the lifetime to determine molecular interactions as used in FLIM-FRET, or to use the fading of the signal as a method for measuring molecular diffusion as in FRAP. This review focuses on photobleaching and energy transfer-based microscopic techniques, which are discussed further below.

The first fluorescence microscopes were developed at the beginning of the 20th century. In 1904, August Köhler constructed an ultraviolet (UV) illumination microscope at Zeiss Optical Works in Jena, Germany, in which he used a cadmium arc lamp as a light source. However, it was Oskar Heimstädt who developed the first working fluorescence microscope in 1911, with which he studied autofluorescence in organic and inorganic compounds [17]. Still, these early fluorescence microscopes suffered from several drawbacks, including the fact that the light sources at that time did not have enough power to excite fluorochromes at high enough rates and that effective separation of the fluorescence signal from the excitation light was difficult to achieve. In order to obtain sufficient signal, Heimstädt had to rely on darkfield illumination, which ensured that only a limited amount of excitation light would enter the objective lens, but this was a technically challenging method and certainly not suited for developing fluorescence microscopy as a mainstream tool. The invention of the epi-fluorescence microscope in 1929 by Philipp Ellinger and August Hirt was a major step forward to achieving this goal.

In this configuration the illumination and detection takes place from one side of the sample (Figure 5), thereby ensuring that only reflected excitatory and emitted light reach the objective, which results in a significantly improved signal to noise ratio. In the 1930s the Austrian Max Haitingen and others developed the technique of ‘secondary fluorescence’ in which samples were systematically stained with fluorescent dyes, initially simply to make the weak autofluorescent biological samples better visible. Haitingen also was the first to introduce the word ‘fluorochrome’ for these fluorescent stains. In the early 1940s, Albert Coons developed the technique of labeling antibodies with fluorescent dyes [18]. This was a breakthrough, as it allowed to specifically label proteins and subcellular structures and as a result made these molecular structures visible in a contrast and resolution never seen before. Since these first experiments, antibody staining with fluorescent secondary markers has become a standard method in biological and biomedical research, and clinical diagnostics for fixed samples, including tissues and single cells. Furthermore, a significant increase in fluorescence signal can be achieved with these fluorochromes over the weak autofluorescent endogenous biological species. The lack of excitation power was resolved with the development of lasers in the 1960s based on Einstein’s theoretical foundations regarding stimulated emission [19] by Gould, Townes, Schawlow, and Maiman [20,21]. Lasers offered what other light sources could not: a high degree of spatial and temporal coherence, which means that the diffraction limited monochromatic and coherent beam can be focused in a tiny spot, achieving a very high local irradiance. Furthermore, it now became possible to effectively separate signals by using suitable filters and dichroic mirrors (also called dichromatic beamsplitters). The latter are specialized interference filters that selectively allow passage of light in a particular wavelength range, while reflecting other wavelengths when placed into the light path at a 45° angle (Figure 5).

The next foremost breakthrough that resulted in an explosion in both instrument and technique development and concurrently biological research was the discovery in the 1960s, sequencing and subsequent development of green fluorescent protein (GFP) as a fluorescent label in the 1990s by Tsien, Chalfie, and Shimomura [22,23,24,25,26,27]. GFP is a 238 amino acid protein that shows bright green fluorescence and was first isolated from the jellyfish Aequorea victoria. The 26.9 kDa wild-type GFP consists of a β-barrel structure (Figure 6A) in which the essential chromophoric moiety, an amino acid triplet of Ser65, Tyr66, and Gly67, lies at the centre. It should be noted that the entire poly-peptide structure is necessary for GFP fluorescence−the chromophore forms autocatalytically−and that the molecular structure surrounding the tripeptide influences its fluorescent properties. Furthermore, the protective β-barrel surrounding the tripeptide ensures stability, makes GFP relatively insensitive to environmental influences, and causes physical separation from species such as molecular oxygen. As a result, GFP’s photobleaching rate is lower than those of conventional fluorochromes. Wild-type GFP (from A. victoria) fluorescence is characterized by a major excitation peak at 395 nm and a minor one at 475 nm (Figure 6C), which results in bright green emission at 509 nm and a quantum yield of 0.77.

From the moment that the crystal structure was elucidated by Remington’s [28] and Phillips’ [29] groups, researchers have modified GFP through directed and random mutagenesis to, amongst others, expand the color spectrum, to narrow the emission peak, to improve photostability, or to enhance the quantum yield for a particular emission wavelength. Additionally, many fluorescent proteins (FPs) from other species, such as the Anthozoan button polyp Zoanthus (ZsYellow), sea anemone Discosoma (DsRed), or Anemonia majano (AmCyan1), have been identified and isolated, which now results in a wide color palette, with various photostabilities, sensing properties, photo-switchability, and useful FRET pairs (see Table 1 and Table 2).

In recent years, many genetic modifications were made on FPs from various sources, which were denoted according to the corresponding fruit colors, such as mCherry or mBanana, the range of which is exemplified in Figure 6A and a selection of these FPs and their properties are listed in Table 1. The development of FP technology was so significant that it opened the doors to completely new ways of performing fluorescence live cell imaging [23], particularly since it was now possible to tailor the fluorescent label’s properties through genetic engineering and to label proteins by expressing fluorescent fusion constructs directly in living cells. Finally, it should be noted that in the literature FPs are often called “autofluorescent proteins”. Although strictly taken this is correct, such classifications should not be confused with the autofluorescence of endogenous cellular biomolecules discussed previously (vide supra).

Additional labeling innovations came from an entirely different discipline of science, which combines nano- and biotechnology. The rapid developments in bionanotechnology over the past decades resulted in the development of luminescent nanoparticles with exceptional physical and chemical properties, not seen in other fluorochromes. Quantum dots (QDs), for instance, are inorganic semiconducting nanoparticles consisting of a core-shell configuration creating a spectral bandgap, e.g., CdSe/ZnS QDs, as depicted schematically in Figure 6B. The size of this bandgap determines the QD’s fluorescent properties and thus the QD’s emission can be directly tuned by their size (Figure 6B) or better said the physical size of the band gap (the band gap energy is inversely proportional to the square of the size of the quantum dot) [30,31,32]. In practice this means that the smaller the QD, the bluer the light. In general, QDs have a relatively long lifetime, which provides the possibility to correct for background signals from short lived fluorescent species by time-gating techniques [33,34]. In addition, it was recently shown that the size of the QD also determines the lifetime, which increases with size [35]. Because QDs have broad absorption spectra, a single light source can be used to excite multiple QDs with different emission wavelengths simultaneously (Figure 6C). This allows the use of both simple voltaic arc lamps and common commercially available lasers, such as argon-ion, helium-cadmium and krypton-argon, with 405 and 488 nm laser-lines, which readily allow excitation of QDs, albeit with varying degrees of efficiency. It is particularly attractive to exploit excitation in the ultraviolet and violet regions (Figure 6C) with blue diode and diode-pumped solid-state lasers that have spectral lines at 375, 405, 442 and 473 nm. Quantum dots are furthermore characterized by a number of additional unique properties [30,31]: (i) QDs are about 10–100 times brighter than organic fluorogenic dyes; (ii) are 100–1,000 times more resistant to photobleaching, because the shell and various coatings form physical barriers that separate the excited state from surrounding biomolecules and molecular oxygen; and (iii) show narrower and more symmetric emission spectra compared with other fluorochromes (typical full-width at half max (FWHM) of ~25–40 nm [36]). A comprehensive description of QD spectral properties is provided by Alivisatos and Biju [30,31,37] and an excellent review on why the small size makes nanoparticles so different from bulk materials is given by Roduner [38].

Nowadays, fluorescence microscopy is the method of choice for live cell imaging and it is a standard procedure to study normal and pathological cell biological processes in single cells, subcellular compartments or across a population of cells by introducing fluorochromes specifically targeted to the (bio)molecules of interest. Major advantages are that fluorescence microscopy techniques provide information with spatio-temporal resolution and are generally less destructive compared with other imaging techniques, e.g., Electron Microscopy (EM).

So what makes fluorescence microscopy such a great and specific tool for cellular and molecular imaging and analysis? Essentially it is its selectivity and contrast enhancement. While the selectivity is achieved predominantly through the labeling methods mentioned above, the contrast increase is realized in the microscope itself. Modern fluorescence microscopes can maximize the collection of emitted fluorescent light, while minimizing the collection of the incident excitation light. Thus, one of the main advantages of fluorescence microscopy is the dramatic increase in signal from the labeled structures and molecules against a dark background—analogous to the stars in the black night sky that are not visible during the day. Image contrast critically dependents on the ability of the microscope to pass fluorescent light to the detector (i.e., a CCD camera [charge-coupled device] or a photomultiplier tube [PMT]) while blocking the excitation light. Due to its selectivity and contrast, not only fine cellular and subcellular structures, but even single molecules can be made visible in the fluorescence microscope. If they are spatially well separated and thus not too close to each other or do not light up at the same time, localization of individual molecules is feasible, but is diffraction limited as established by Ernst Abbe (see §1.2.2).

Three basic components are present in every light microscope, irrespective of the type: (i) an illumination source; (ii) a magnifying lens; and (iii) an image acquisition device. The classical transmitted wide field light microscope typically consists of a white light bulb or light-emitting diode (LED) that illuminates the sample in toto, a convex lens system, and the human eye as a detector. Conversely, in a wide field fluorescence microscope, the white light source is replaced by a high power lamp (a mercury or xenon source), which excites the fluorochromes in the fluorescently labeled sample and induces fluorescence emittance as shown schematically in Figure 5. Images are typically acquired visually by eye or electronically with a CCD camera. As described above, fluorochromes have characteristic excitation spectra. Thus, an appropriate excitation filter, usually a band-pass filter (BP), is placed between the lamp and the sample to narrow the wavelength range of light reaching the sample to such an extent that the fluorochromes used are excited efficiently, whereas unwanted excitation is minimized. Since the emitted light has a longer wavelength than the excitation light, an emission filter (either a long pass [LP] or BP filter) placed between the sample and the detector effectively blocks the excitation light and prevents perturbation of the final image. Intense light is required for successful fluorescence excitation. Lasers generally produce high intensity light and have proven to be an excellent alternative excitation source to mercury and xenon lamps. Accordingly, lasers are now commonly utilized in confocal and multiphoton laser scanning microscopy to point illuminate the sample. Since lasers are a source of monochromatic light, generally no excitation filter is required. However an emission filter or an alternative spectral selection device is still needed to stop the excitatory laser light from reaching the detector and to tune in on the fluorochrome’s emission signal, especially in multiple labeling experiments. In the case of a confocal laser scanning microscope, the mercury lamp is replaced by a set of appropriate lasers that allow the excitation of various fluorochromes, the excitation filters are removed, and the detecting camera is replaced by a photomultiplier tube [39,40,41]. Lichtman and Conchello recently published a good and concise introduction into fluorescence microscopy [42], whilst more extensive reviews can be found in Pawley’s handbook [39].

In fluorescence microscopy, the optical resolving power determines the amount of detail observed in a specimen, which in turn is determined by a number of physical factors and instrument limitations. If light hits a small object under observation, the direction of the incidence light is changed (diffraction) and this deflection increases with decreasing size of the object. In order to obtain sharp images, the objective must capture as much of the deflected light as possible, which is achieved with wide angular openings (aperture). Ernst Abbe first defined the numerical aperture (N.A.) [43], which determines the objective’s light capturing capacity and can mathematically be written as:

(5)

where n is the refraction index of the medium between the object and the objective (n = 1 for air and for immersion oil n = 1.51), and α is half the objective opening angle (Figure 7A). To maximize capturing the deflected light and to increase the resolution, several strategies can be employed, including the use of a condenser lens for illumination to widen the angle of the ray cone on the illumination side and/or to use an immersion liquid between objective lens and the cover slip, which abrogates reflections that normally diminish the resolving power. In air, theoretically N.A. = 1 can be achieved when α = 90°, but in practice, N.A. values above 0.95 cannot be attained. Conversely, with immersion oil, N.A. values larger than 1 can easily be reached and thus the use of oil immersion objectives is the only way to increase magnification with sufficient resolution and contrast.

So what exactly is resolution or optical resolving power? The introduction of several parameters and concepts are required to delineate the concept of resolution. First of all, it is important to realize that in fluorescence microscopy the resolution is not directly governed by the magnification. Secondly, resolution and contrast are two reciprocally interrelated parameters that are important in fluorescence microscopy and should not be considered as separate entities. Intuitively it becomes clear that when the contrast approaches zero, it would be futile to discuss optical resolution. As stated previously, the individual objects in a specimen cause diffraction of light and as a result, an illuminated point source within the specimen is observed as a bright central spot (Airy disc) with surrounding diffraction rings (Airy pattern) as shown in Figure 7. It was George Biddell Airy who first theoretically described this phenomenon [44] in 1835, although others had previously observed it experimentally. Generally, optical microscopes can be assumed to be linear and shift-invariant, which means that the image of a specimen essentially consists of the linear superposition of all the specimen’s individual elements, which results in the final image, e.g., a cell with all its labeled components and organelles. The intensity point spread function (PSF) characterizes such a system and the Airy pattern is essentially the intensity distribution of the intensity PSF in the focal plane (x–y). Combining both the PSF in lateral (x–y) and axial (x–z) direction (see Figure 7B) creates a complex three dimensional shape, the three dimensional PSF, which characterizes the response of the entire optical system, lenses, mirrors, optical apertures, and imperfections or misalignments in the optical system to the illuminated point source and the diffraction those elements and the object cause.

Essentially, resolution may be defined as the smallest distance between two points in the specimen that can still be discriminated as separate points (Figure 7C) at a particular contrast, which with Equation 5 can be written as:

(6)

Contrast is defined as the difference between the maximum intensity and minimum intensity occurring in the space between two objects of equal intensity (the Airy discs; contrast = 1) [46,47]. When the two point objects are well separated, the contrast minimum between them is near zero and the objects can be discriminated (they are resolved) as shown in Figure 7C (green). However, as the point objects approach and their PSFs start to overlap, the intensity minimum between the two object maxima is reduced until the objects are no longer resolved (Figure 7C, red). Even though this formulation closely follows practical microscopy, commonly the Rayleigh criterion for resolution is used, which states that two points are resolved when the first minimum of one Airy disc is aligned with the central maximum of the second Airy disc. From the above mentioned considerations, it becomes obvious that the smaller the Airy pattern (Figure 7B), the higher the resolution, the more detail can be obtained in an image (compare wide field and confocal microscopy in Figure 8).

It was Ernst Abbe who in 1873 theoretically laid the foundation for describing the diffraction limit that led to the formulation of Equation 6. He described that the smallest resolvable distance between two points using a conventional light microscope cannot be smaller than half the wavelength of the imaging light [48]. Consequently, an increase in resolution can only be achieved if the wavelength of light used is as small as possible. Thus if a wavelength of 400 nm is used, which is still in a range that is usable in biological imaging, an approximate lateral resolution of 200 nm is feasible. However, in comparison to the size of a single animal cell (10–100 µm), viruses (20–200 nm), and organelles (500 nm–10 µm), this still precludes attaining sufficient resolution. Further reduction of the wavelength into the UV causes significant damage to biomolecules and in live cell imaging seriously affects and alters normal cellular function and homeostasis. Consequently, the measured results are perturbed and evaluation of the biological function cannot be performed with sufficient accuracy and thus the results might be scientifically flawed.

Confocal laser scanning microscopy (CLSM) is a technique, which combines high-resolution optical imaging with depth selectivity [39]. The original technique used a stage-scanning confocal optical system and was invented by Marvin Minsky in 1957 [49]. Essentially, the CLSM is based on a conventional optical microscope in which instead of a lamp, a laser beam is focused onto the sample and an image is built up pixel-by-pixel by collecting the emitted photons, usually with a PMT. Thus, CLSM combines point-by-point illumination with simultaneous point-by-point detection (Figure 9) and illumination and detection are restricted to a single diffraction-limited point. A key feature of CLSM is its ability to acquire well focused images from various depths within the sample; a process called “optical sectioning”. The images of a mouse intestine section in Figure 8, noticeably illustrate the gain in resolution in CLSM imaging over conventional wide field imaging.

This is achieved by placing a small pinhole aperture before the detector (Figure 9B), which prevents emitted out-of-focus light from the planes above and under the focal plane, as well as stray light from reaching the detector.

In conventional wide field microscopy, such signals cause glare, distortion and blurriness in the image and these artifacts are collectively called “convolution”. Even though a small improvement in both axial and lateral resolution (notice the central maxima in Figure 7B) is achieved over wide field techniques, the abrogation of interference of out-of-focus and stray light with the in-focus signal and a significant reduction in the diffraction pattern (Figure 7B) cause a considerable increase in resolving power. The focal point in the sample and the pinhole lie in conjugate planes, as shown in Figure 9B, and this optical arrangement of the focal points is called ‘confocal’. Notice the difference in the light pathways over a wide field fluorescence microscope in Figure 9. The smaller the pinhole, the less light from out-of-focus areas within the specimen reaches the detector, the lower the intensity of the image. During scanning, with a defined focusing along the z-axis (axial direction) and lateral movement (x and y axis), the confocal volume element is moved through the specimen by a succession of object planes. It is thus possible to obtain optical sections of the specimen and reconstruct its 3D structure.

In biological imaging applications, may that be single cells, tissues, or intact model organisms, CLSM has proven its power over the past thirty years and concomitantly produced spectacular new scientific insights, particularly because of its optical sectioning capability and the possibility for 3D reconstruction from a stack of individual sections. Furthermore, the advantage of this technique comes from its non-linear behavior in that the technique is sensitive to the square of the light intensity and not just the light intensity. This combined with the increase in contrast by cutting off unwanted signals from out-of-focus planes, which results in sharper images with better z-resolution (Figure 8), clearly offers an advantage over the wide field microscope [50,51].

Multiphoton fluorescence microscopy resembles CLSM in that both use focused laser beams to scan a specimen in a raster pattern (point-by-point) to generate images, and both have an optical sectioning capability. Unlike confocal microscopes where the optical sectioning capability is generated in the emission light path, in a multiphoton microscope the sectioning capabilities are caused on the excitation side because of the extraordinary way that the fluorochromes are excited. The concept of multiple photon excitation (MPE) was first theoretically described by Maria Göppert-Mayer in 1931 in her doctoral thesis [52] and subsequently observed experimentally in 1961 by Kaiser and Garret when they detected blue fluorescence in CaF₂ : Eu²⁺ crystals in response to a red 0.5 ms pulsed ruby laser [53]. However, it was Winfried Denk who developed two photon imaging for use in living cells and tissues in the lab of Watt Webb [54].

Multi-/two-photon excitation (TPE) is based on the simultaneous absorption of two (or multi) photons by the fluorochromes in a sub-femtoliter volume at the focus (Figure 9C) in which the energy for exciting the electron from the ground state to the excited state is provided by two photons with approximately half the energy (Figure 10B). Since energy scales inversely with the wavelength according to Planck’s law (Equation 1), half the energy means twice the wavelength of an excitory photon in single photon excitation (compare Figures 10A,B). Since the statistical probability of the concomitant absorption of photons is extremely low, high local photon fluxes are required (MW/cm² to GW/cm² [55]), which only became possible with the introduction of femtosecond mode-locked pulsed lasers. Two-photon excitation is a non-linear process, as the absorption rate increases with the second power of the excitation light intensity. As a consequence, even when using high power pulsed lasers, the excitation is restricted to a small volume in the focal plane of the specimen (see schematically in Figure 9C and in the fluorescein solution in Figure 10), where the photon density is high enough. The practical outcome of this phenomenon is an optical sectioning without the need for a pinhole to block fluorescence from out-of-focus locations (Figure 9C). By subsequently scanning this excitation volume through a sample, z-stacks of 2-dimensional images can be collected and 3-dimensional images can be reconstructed.

An additional benefit of limiting the excitation of the fluorochromes to such a small volume in the plane of focus is a significant reduction in the overall photobleaching. This is exemplified in Figure 10 when a fluorescein-stained formvar film is illuminated in either a CLSM or a TPE mode. The bleaching pattern in the film clearly shows significant bleaching above and below the focal plane in CLSM, whereas bleaching is exclusively restricted to the focal plane in TPE microscopy. Overall, this significantly reduces the photodamage and cytotoxicity normally associated with fluorescence microscopic imaging experiments and ensures that near-normal cellular homeostasis is maintained. Consequently, cells may be observed for longer periods of time with fewer toxic effects. Another important advantage of this technology is that, since most fluorochromes are excited in the range from 350–550 nm, deep red or infra-red excitation light can be used in TPE, which penetrates much deeper into the specimen due to reduced scattering and absorption by endogenous chromophores. A variant, two-photon autofluorescence microscopy (2PAM), exploits the autofluorescence of endogenous biomolecules to assess the disease states in tissues based on changes in morphological, spectral, and lifetime parameters. Since, for instance cell layer thickness can be used as an indicator of dysplasia and carcinoma, the imaging depth capabilities of TPE are particularly important in 2PAM. Durr et al. determined that ex-vivo 2PAM imaging in a human tongue biopsy was feasible to a depth of 370 μm [56]. In normal laser scanning devices, only a penetration depth of about 250 μm can be reached. Some researchers have used high-pulse power regenerative amplifier systems [57,58] to increase imaging depth and Denk et al. reported imaging at a depth of 1,000 µm in living mouse brains by use of a Ti:Al₂O₃ regenerative amplifier [58].

The increased achievable imaging depth is a significant advantage of TPE microscopy over conventional CLSM methods. Other non-linear methods, such as second harmonic-based microscopy or the use of fluorochromes that show upconverting properties have also been developed for a myriad of biological applications. What all these methods have in common is that their excitation and emission follow anti-Stokes shifts, i.e., the excitation wavelength is larger than the emission wavelength.

Second-harmonic imaging microscopy (SHIM) is based on the nonlinear optical effect of second-harmonic generation (SHG) in which laser light is focused on a sample to generate frequency-doubled light (two photons with wavelength λ are extinguished to generate a single photon with wavelength 0.5λ [60,61]). Unlike in TPE, SHG does not involve an excited state and therefore no energy is lost during relaxation of the excited state as in the case of TPE. Furthermore, SHG and SHIM offer several advantages in live cell imaging: (i) SHG is energy conserving; (ii) conserves laser coherence; (iii) requires no labeling; (iv) since no excited state is involved, photobleaching and phototoxicity are virtually absent; and finally (v) has sectioning capabilities because in SHG the amplitude is proportional to the square of the incidence light intensity as in TPE [60,61].

In contrast to TPE, in photon upconversion, the chromophore’s properties allow the sequential absorption of long wavelength photons. The fundamental processes involved in upconversion are complex and consist of several competing processes, including sequential energy transfer, excited state absorption, and phonon interaction (a quantum of vibrational energy that arises from oscillations within a crystal) [62,63,64]. Generally, rare earth metals (d- and f-block elements) such as lanthanides are involved as luminescent complexes with organic enhancers, or as dopants in luminescent nanoparticles. Less effective upconversion is achieved when using actinides or transition metal ions [62,63]. Materials with upconverting properties are often referred to as “upconverting phosphors”, which is potentially a confusing term. As illustrated in Figure 10C, the occurrence of metastable excited states with long lifetimes in the ms range abolish the need for simultaneous arrival of the photons and thus high local photon fluxes. For upconverting chromophores, a normal fluorescence microscopic setup with low power lasers–intense anti-Stokes emission is already possible below 1 W/cm² [64,65]–inexpensive detectors, such as a PMT for photon counting, a standard long-pass excitation filter, and a narrow band-pass emission filter with adequate infrared blocking capabilities generally suffices as instrumentation. The extraordinary photoluminescent characteristics in upconverting materials allow: (i) imaging against dark backgrounds, since upconversion does not occur in endogenous cellular biomolecules; (ii) with N(IR) excitation, excellent imaging depths can be achieved; (iii) the use of conventional imaging systems keeps investments and costs low, (iv) their photoluminescent properties are relatively insensitive to environmental changes, such as pH or solvent polarizability; (v) virtually no photobleaching occurs, and (vi) such materials have long shelf-lives. More in-depth deliberations on upconversion can be found in references [13,62,63,64,66].

From the above considerations it quickly becomes clear that multiphoton fluorescence microscopy is a powerful tool in biomedical research that offers reduced photobleaching and low photo-toxicity, higher penetration depths, and higher spatial resolution than other in vivo imaging modalities. With the continuous development of new fluorescent proteins, equally multiphoton microscopy is subject to constant improvement. Furthermore, it can be deduced from the literature that in life science research, predominantly two-photon techniques are now commonly applied to resolve scientific questions. A comprehensive guide to choosing the right fluorescent protein and excitation wavelength for two-photon applications and a review on the two-photon spectral properties of fluorescent proteins was recently presented by Drobizhev et al. [67]. Excellent reviews on two and multiphoton microscopy, including a deliberation of the advantages and disadvantages may be found in [45,47,54,55,59,68,69,70,71].

In a typical FRAP experiment (Figure 11A), fluorescent molecules are irreversibly photobleached in a small area of the cell by high intensity illumination with a focused laser beam. Subsequently, diffusion of the surrounding non-bleached fluorescent molecules into the bleached area leads to recovery of fluorescence with a particular velocity, which is recorded at low laser power.

As previously described, photobleaching in most fluorochromes requires excitation to an excited state and the presence of molecular oxygen. During FRAP, the high light intensity in the presence of molecular oxygen causes irreversible damage to the fluorochrome (Figure 4), thereby permanently interrupting the cycle of repetitive excitation and photon emission. Ultimately, those fluorochrome molecules that are permanently damaged no longer contribute to the recovery of fluorescence in the bleached area. The bleached fluorochromes are replaced by unbleached ones, a process that occurs as the result of the diffusional exchange between them.

The fraction of fluorescent molecules that can participate in this exchange is referred to as the mobile fraction (M_f), whereas the fraction that cannot exchange between bleached and non-bleached regions is called the immobile fraction (I_f), as shown in Figure 13A. Therefore, FRAP provides important insights into the properties and interactions of molecules within the cellular environment.

FRAP can also be used to measure the dynamics of 2D or 3D molecular mobility, e.g., in diffusion, transport, or any other kind of movement of fluorescently labeled molecules in living cells. A representative example is given in Figure 12, which shows that monomeric GFP-Myosin III can easily traverse the nuclear envelope membrane. Myosins are motor proteins that together with kinesins and dyneins are responsible for a wide range of movement and transport processes. Class III myosins play critical roles in the vertebrate retina and inner ear function and show an exceptionally high affinity for actin [111]. The nucleus is bleached with high intensity (~500 ms; >30 mW) with a 488 nm laser. Subsequently, the nucleus is devoid of green fluorescence. Over time the fluorescence recovers and reaches a plateau. Notice by comparing Figures 12A,D that the total fluorescence intensity decreases, because a significant number of fluorochromes were irreversibly bleached.

In FRAP experiments, the images are analyzed and processed to generate a kinetic plot of photobleaching by displaying the temporal fluorescence changes in the bleached region of the cell. From this plot, the mobile and immobile fractions can be determined by calculating the ratios of the final to the initial fluorescence intensity (see Figure 13 and the equations therein). By convention, the speed of recovery to half the plateau intensity (I_∞) is called ‘half maximum’ or ‘half life’ (τ_½). The shorter the half life, the faster the fluorescence recovery occurred and the higher the diffusion. Furthermore, the half life of recovery is proportional to the bleach area size if the recovery is diffusion limited.

A more absolute way of obtaining the half life and immobile/mobile fractions, which is also suitable for automation, is through non-linear curve fitting of the experimental data points using a simple exponential equation:

(7)

Subsequently, the fitted coefficients can be used to extract the required information from the FRAP curve with:

Mobile fraction= M_f (8)

Immobile fraction (IM_f ) = 1 – M_f (9)

Substitution of I_t with ½ M_f results in an expression for the half life:

(10)

From Equation 10, the half life can be calculated from τ, which provides information on the diffusion of the fluorochrome. This is just a concise and simple example on how to extract information from FRAP curves via modeling. Many methods are currently available to analyze FRAP data curves ranging from biochemical binding models to molecular transport modeling and are often included in the software provided by the microscope manufacturer. An in-depth overview of analytical methods, including potential pitfalls in data analysis is provided in references [102,112,113,114,115,116,117,118,119].

Different profiles of the temporal fluorescence recovery intensity plot provide information about the protein’s mobility, which can be classified as high, intermediate or immobile (Figures 13B–D).

Changes in the mobile fraction may also give clues about various intracellular processes and their temporal outcomes, e.g., interaction of a protein of interest with, for instance, other proteins or (bio)molecules. The mobile fraction can also be markedly affected by cellular membrane barriers and micro-domains within the membrane. These discontinuities can prevent, or temporarily restrict, the free diffusion of molecules through various cellular compartments or within the membrane itself. Conversely, active transport via coated vesicles or motor proteins, such as the aforementioned myosins (Figure 12) that use actin filaments to ATP-dependently transport cargo over large distances, can cause a significantly higher mobility compared with diffusion limited processes. Such information can easily be extracted from FRAP data curves.

In any FRAP experiment, a series of fluorescence images is initially collected to give a baseline value for the intensity in both the ROI and the surrounding labeled cellular environment. Following this, a defined region of the sample is illuminated with high intensity light causing photobleaching of the fluorochromes within this region. This creates a darker, bleached region within the sample. Photobleached molecules are subsequently replaced by non-bleached molecules over time, leading to an increase in fluorescence intensity in the bleached region, as described previously (Figure 11A and Figure 12).

The intensity of the scanning laser applied during the acquisition of typical confocal images is often sufficient to produce significant bleaching of the entire sample within the field of view during the course of image acquisition. Each point in the scanned sample receives the same total light intensity, resulting in uniform bleaching of the whole sample. One way of correcting for this acquisition bleaching is through mathematical corrections (see Equation 13). Additionally, a number of fundamental and instrumental approaches can be employed to minimize this background bleaching ab initio. For example, applying line scans instead of 2D scans, using fluorescent probes that are less susceptible to photobleaching, decreasing the laser power or the pixel resolution by zooming out or using faster scans. In these cases, a compromise between temporal and spatial resolution in a time course experiment needs to be take into consideration. For example, if images are acquired at high speed, the spatial information is often lost, but if the images are acquired slowly in order to increase spatial resolution or gain a better signal-to-noise ratio, the information about the dynamic processes occurring in the biological sample may be lost. Consequently, in quantitative FRAP experiments the crispness of the image itself is often less important and therefore the spatial resolution is sacrificed for the benefit of maximizing the temporal resolution. In practice, the FRAP data (fluorescence intensity values) are acquired quickly in order to successfully record the recovery of the bleached region. However, this results in a significant decrease in the signal-to-noise ratio, which can be partly compensated for by increasing the aperture of the confocal pinhole.

Although undesirable, a small degree of background bleaching can be tolerated and corrected for. Because background bleaching is effectively constant throughout the field of view, the fluorescence intensities in a region of the cell some distance from the bleached region, or in another cell in the field of view, can be used to correct the images for this effect. In practice, curve fitting is performed with an exponential equation that contains an extra term to correct for the bleaching that occurs during image acquisition. Many methods can be employed, ranging from recording an image acquisition decay curve (light blue curve in Figure 13A) and correcting the FRAP curve in every single time point [120] to expansion of equation 7 with a term that assumes that the acquisition bleaching follows a simple exponential decay, such as used in the “back multiplication method”. The normalized, uncorrected FRAP curve is fitted according to:

(13)

The values for τ₂, y₀, which is the minimum plateau that the acquisition decay curve approaches as t→∞, and B are obtained from fitting the decay curve recorded either by measuring the decay in the whole cell, an adjacent cell (both whole cell ROI), or a reference region (reference ROI). Furthermore, a myriad of modeling methods based on the kinetics of the molecular phenomenon are available, as described in references [102,112,113,114,115,116,117,118,119].

The ideal fluorescent probe for use in photobleaching studies should be highly fluorescent (high quantum yield), but only moderately susceptible to photobleaching. This permits bleaching within realistic time frames, but limits bleaching during image acquisition. Green fluorescent protein is the most widely used fluorescent probe for cellular studies because of its stability, low cytotoxicity, because it does not bleach significantly at low light intensities, does not seem to be damaging to the cell after undergoing irreversible photobleaching (see § 1.2), it can be readily expressed in several cell types where it is fused to a particular protein, and in many cases tagging a protein of interest with GFP has no significant influence on the function and localization of the protein under investigation [85,121,122,123,124]. Because of these characteristics, GFP behaves more like a “non-invasive” extrinsic fluorochrome, with a stability that is higher than the commonly used organic dye fluorescein isothiocyanate (FITC) [125]. FITC is neither the ideal probe nor a good choice for performing quantitative FRAP-studies with the CLSM, exactly because of its high susceptibility to background bleaching during image acquisition. A good alternative for use in FRAP studies are the Alexa series of dyes, because of their superior properties, such as their high fluorescence quantum yields and relatively low propensity to bleaching during image acquisition, but still allow efficient bleaching in the ROI with an adequate high dose of laser light [126].

A modified version of the FRAP method, called inverse FRAP (iFRAP) was initially developed to study the mobility of molecules in small areas of the nucleus and their exchange with the surrounding nucleoplasm [98,127]. Inverse FRAP, as schematically depicted in Figure 11B, was initially developed by Misteli et al. [98] and from its setup is particularly useful to study the residency time of molecules in small organelles. In iFRAP, the entire population of fluorochromes in the cell is bleached, except the accumulated fluorochromes in a small part of the organelle. Subsequently, the loss in fluorescence in the accumulation is recorded over time (Figure 11B), from which the rate of exchange with the surroundings can be calculated. Since this loss in intensity directly reflects the releasing process between molecules, no further complicated analytical methods are required to determine the rate of exchange [122]. One of the main limitations of iFRAP lies in the long time needed to photobleach the entire cell (this can take several seconds), which renders iFRAP unsuitable to detect fast translocations. Therefore, iFRAP is mostly useful for analyzing the dissociation kinetics of molecules bound to an immobile intracellular structure.

The utilization of iFRAP to study the dynamics of mRNAs at speckles (subnuclear domains localized at the interchromatin region, containing pre-mRNA splicing factors) exemplifies both the procedure and potential of iFRAP. This work mainly addressed the question on the physiological meaning of the accumulation of transcribed mRNA in these speckles. To answer this highly relevant biological question, fluorescently labeled Drosophila fushi tarazu (ftz) pre-mRNAs were micro-injected into the nuclei of Cos7 cells and the association and dissociation kinetics of pre-mRNAs from these speckles were analyzed using iFRAP as shown in Figure 14. The authors showed that some pre-mRNAs can be shuttled between speckles and the nucleoplasm, suggesting that pre-mRNAs repeatedly associated with and dissociated from speckles until introns were removed (Figure 14) [127].

The regeneration time (half-recovery period) is a measure for the speed of protein movement.

Apart from erroneous data modeling and processing or operator errors, there are several potential and important complications associated with FRAP [122], some of which are unexpected and should be taken into account:

A complementary technique to FRAP, termed fluorescence loss in photobleaching (FLIP), has been used to reveal the connectivity between different compartments in the cell or the mobility of a molecule within the whole compartment [113,131]. FLIP experiments differ from FRAP and iFRAP by the repetitive bleaching of the same region in the specimen (Figure 15A), thereby preventing recovery of fluorescence in that region.

In FLIP experiments the repetitive bleaching occurs adjacent to the unbleached ROI (Figure 15A). The loss in fluorescence in the ROI defines the mobile fraction of the fluorescently labeled protein. Conversely, the incomplete loss in fluorescence defines the immobile fraction of fluorescently-labeled protein that does not move into the continuously photo-bleached area. The observation that molecules do not become bleached suggests that they are isolated (immobilized) in distinct cellular compartments. FLIP experiments are very useful to demonstrate the connectivity and fluxes between different regions of the cell and thus is an ideal and direct method for studying the exchange of molecules between two compartments (e.g., compartments that are separated by lipid bilayers). The continuity of cellular structures, such as the Golgi apparatus, the endoplasmic reticulum, the protein traffic between the nucleus and cytoplasm, the nucleolus and splicing factor compartments, and the nucleolus and nucleoplasm have all been studied using FLIP [131,133,134,135]. FLIP is often used in combination with FRAP experiments to obtain combined information regarding active or passive transport. In fact, FLIP can be used as a control for FRAP experiments.

For example, FRAP and FLIP were used in conjunction to determine the mobility of GFP-tagged proteins involved in various steps of ribosome biogenesis in living cell [132]. The comparative FRAP and FLIP analysis of the protein dynamics data (Figure 16) revealed that the nucleolar proteins, upstream-binding factor-1 (UBF1), nucleolin, fibrillarin, Rpp29 (a human RNase P subunit), and B23 (multifunctional nucleolar phosphoprotein) move much faster between the nucleolus and nucleoplasm than the ribosomal proteins S5 and L9. The FLIP-FRAP investigations by Chen and Huang [132] suggest that a new level of regulation for rRNA synthesis exists. Furthermore, by combining FRAP and FLIP, the different dynamical properties of the proteins involved in the various steps of ribosome biogenesis could be determined and discriminated (Figure 16B). The results imply that the proteins’ nucleolar association is likely due to their specific functional activities rather than specific nucleolar-targeting events.

The complications and complexities that need to be considered in a FRAP experiment apply in a similar way to FLIP experiments and need to be taken into account. These were described in more detail before (vide supra).

An alternate way to measure FRET is by determining the sensitized emission. In this method only the donor molecule is excited, which transfers energy to the acceptor causing it to enter an excited state instead, and changes in the fluorescence signal are subsequently measured in the acceptor channel only. This is the simplest way to measure FRET and would be the ideal method if the donor and acceptor channels would be fully separated and no cross-talk would occur. Unfortunately, in practice this approach necessitates particular precautions with respect to cross-talk or “bleed-through” of the detection channels and consequently sensitized-emission FRET requires high quality, suitable and specific filters and the appropriate set of controls. In order to obtain accurate FRET results, three samples have to be prepared: (i) a sample that consists of the donor-only; (ii) the second sample consists of the acceptor-only; and (iii) the third is the actual FRET experiment sample. These samples are imaged with the respective image channel settings for the donor, acceptor, and for the final sensitized emission measurement, i.e., donor excitation and acceptor detection. From these three images, the final FRET signal can readily be calculated, but adequate quantization remains nonetheless challenging [201].

There are different methods to calculate the final FRET signal (F ^c). All these methods have in common that correction factors have to be determined.

In the Youvan method [202] the corrected FRET signal (F ^c) is calculated according to:

(20)

F ^c is calculated by subtracting the corrected donor contribution (Donor_corr) and the corrected acceptor (Acceptor_corr) contribution from the measured FRET signal. The correction values are determined in detail by:

(21)

where D, A, and F represent a monochrome image acquired through the donor, acceptor or FRET channel. Superscript ^b indicates that the monochrome image has been background subtracted and the individual pixels from the donor and acceptor fluorochrome are denoted with subscript _d or _a. Therefore, represents a background-subtracted donor image taken in the FRET channel; is a background-subtracted donor image taken using the donor channel. The same notation is used for the acceptor fluorochrome and the ratios in Equation 21 are used to correct the FRET channel intensities pixel-by-pixel. In this method the F ^c is corrected for the donor and acceptor contribution to the final signal measurement with the FRET settings, but is not normalized for donor and acceptor concentrations. The method proposed by Gordon et al. [203] extends the Youvan method and corrects for the concentration of the donor and acceptor. The Gordon FRET correction can therefore be expressed as:

(22)

The Gordon method works best in cases where low concentrations of donor and acceptor are used. Alternatively, the method by Xia et al. [204] normalizes the FRET values for donor and acceptor concentrations as follows and produces even better results:

(23)

Sensitized emission FRET measurements can be performed on wide field as well as on confocal microscopes with respective dichroics and filter sets. Major drawbacks of the sensitized emission method are: (i) the method is relatively laborious and requires extensive image processing; and (ii) the need for image processing increases all the noise in the images and might ultimately exceed the FRET signal in experiments where the FRET signal is weak to begin with. Thus, measuring and calculating the controls, applying the correction factors and calculating the final FRET signals may lead to substantial errors. The experiments become particularly tricky when this FRET approach is carried out on independently labeled fluorescent partners, i.e., where the stoichiometry is not known or cannot be controlled and the concentrations of donor and acceptor differ significantly or even change during the experiments. Therefore, it is much easier to carry out such FRET experiments on intramolecular structural changes where the fluorescent labels are within the same molecule and the stoichiometry is well defined, e.g., with Ca²⁺ sensors such as the Cameleons [166,205]. An excellent comparison of the methods used to quantify FRET in sensitized emission FRET was made by Berney and Danuser [206], and the interested reader is referred to that review. Nevertheless, despite these limitations, if no FLIM microscope is available, sensitized emission is a suitable and relatively inexpensive method to carry out FRET in dynamic live cell experiments, provided that the FRET signal is intense enough.

A major improvement of sensitized emission FRET, reducing the risk of cross-talk, noise and other unwanted effects to a minimum, is a further development of the aforementioned method and is named Spectral Imaging FRET. In this method, instead of detecting the donor and acceptor fluorescence intensities around λ_max, the entire emission spectrum of both donor and acceptor is acquired upon donor excitation [207,208] in virtually the same way as in a fluorimeter. The advantage of this approach is obvious: by taking the distinctive shapes of the spectra into account, and any changes therein, information on FRET can be obtained with a much higher certainty and virtually devoid of any contamination from other channels. Spectral imaging therefore permits the researcher to directly evaluate the level of cross-talk because of direct excitation of the acceptor, and facilitates the accurate measurement and calculation of the true FRET signal [207].

Because the interpretation of intensity-based FRET measurements is challenging and limited by experimental conditions and possible artefacts, such as signal cross-contamination, the concentration and labelling efficiency of the fluorochromes, variations in excitation intensity and exposure duration, and photobleaching, researchers sought ways to overcome these limitation. Since the fluorescence lifetime τ is affected by energy transfer, but essentially insensitive to the aforementioned limitations, measuring the lifetime via FLIM provides essential information on FRET and ameliorates many limitations associated with intensity-based FRET. The first FLIM instrument was described as early as 1959 and was based on a frequency-domain microscope setup that only allowed single point measurements [209]. The first true FLIM imaging was reported by Wang et al. in 1989 [210]. FLIM was further developed independently by various groups, but Kusumi and co-workers were among the first to perform time-resolved fluorescence imaging in single cells [211].

FLIM is a technique that maps the spatial distribution of the lifetimes within microscopic images of fixed as well as living cells. As stated previously, fluorochromes are not only characterized by their excitation and emission spectra, but also by their unique lifetime. The fluorescence lifetime is the exponential decay in emission after the excitation of a fluorescent probe (see Equations 3 and 4). In the physical sense, the fluorescence lifetime τ is the time needed for the fluorescence intensity to decrease to 1/e (=1/2.71) of its initial value I₀ and can generally be considered as the average time that the fluorochromes resides in the excited state. The fluorescence lifetime does not change upon intensity variations provided that sufficient signal is available to be detected. Furthermore, lifetime measurements are not dependent on the transmission efficiency of the microscope, the local concentration of the fluorochromes, the local excitation light intensity, or on the local fluorescence detection efficiency [11] and are generally insensitive to moderate levels of photobleaching [212]. However, even though the fluorescence lifetime is largely insensitive to the aforementioned factors, the fluorochrome’s lifetime is sensitive to its environment such as changes in pH, polarity, refractive index of the medium, temperature, to name but a few. Even though these phenomena need to be taken into account in some experiments, in others they provide the basis for mapping spatial variations of the lifetime in response to changes in the environment and thus may sensitively provide information on the biomolecule’s action or state in vivo.

In FLIM-FRET, the presence of the acceptor causes energy transfer and as a result the donor looses its excited state energy faster than in the absence of the acceptor. Consequently, the donor fluorescence lifetime decreases concomitantly. Combination of FLIM and FRET allows the measurement of lifetime dynamics pixel-by-pixel and thus mapping of its spatial distribution to indirectly measure biomolecule concentrations, interactions between biomolecules, and conformational changes with a much higher accuracy than conventional FRET methods. The FRET efficiency (E_FRET) may be calculated according to:

(24)

where τ_DA and τ_D are the excited-state lifetimes of the donor in the presence and absence of the acceptor, respectively.

FLIM has been used in several applications, such as the analysis of protein-protein interactions with high temporal specificity, in ion concentration imaging as well as in measuring oxygen concentration and in various medical applications [213,214]. Thus, FLIM is an example of using fluorescence beyond emission intensity measurements and determining the spatial location or distribution of a molecule or protein, but rather FLIM allows examination of the protein micro-environment with high resolution. Although the FLIM instrumentation setup is more complex and expensive compared with other FRET detection methods, FLIM is the most accurate method for FRET measurements in live cell experiments, but equally cannot be considered a main-stream and trivial methodology. FLIM can be used in scanning confocal, multi-photon, or in wide field microscopes, and can be implemented in two ways: (i) either in the frequency domain, using sinusoidally modulated excitation light; or (ii) in the time domain, using pulsed excitation sources. Traditionally such pulsed lasers sources have been employed in multiphoton imaging and deliver discrete and consistent bursts of light to the sample. The lifetime of most fluorescence markers is in the order of a few nanoseconds, and most commercially available pulsed lasers deliver light in 12.5 nanosecond pulses, which provides a good window for lifetime imaging. Nonetheless, most fluorescence sources can be adapted for FLIM imaging provided that the light that reaches the sample is effectively modulated [215,216]. Generally, single-photon-counting is widely used in scanning FLIM, either time-correlated or time- and space-correlated, but a number of innovations have recently been introduced that allow faster data acquisition (discussed in [183]). In wide field FLIM, gated or modulated image intensifiers are generally used with CCD camera detection. This form of FLIM can be preformed in the time-domain mode, with pulsed excitation sources and gated or time-correlated single-photon counting or in the frequency domain with intensity modulating excitation and homodyne or heterodyne phase-sensitive detection [183]. Major advantages of wide field FLIM are that inexpensive light-emitting diodes (LEDs) can be used as excitation source, and the high acquisition speed, since no scanning is performed and consequently all pixels are acquired simultaneously. On the down side, because the emission is only sampled briefly, the maximum photon count is not reached, which translates in a limited accuracy in the lifetime measurement. Nonetheless, wide field FLIM might simplify the technique and reduce the costs to such an extent that FLIM can be routinely applied in various laboratories.

The combination of FLIM-FRET and TPE microscopy offers the same capability for life time measurements, but has the significant advantage of producing less scattering, increased spatial resolution and depth-sectioning. Other recent advances, including multi-parameter FLIM-FRET, acceptor fluorescence rise-time FLIM-FRET, and single molecule FLIM are discussed elsewhere [183].

FLIM-FRET has been used to address questions about protein conformational changes. For example, to study integrin-effector binding using α4 and β1 integrin in a quantitative manner using a GFP-mRFP FRET pair [218]. In this study FLIM-FRET was used to obtain information on how the conformational state of the receptor may determine the activity of anti-integrin molecules, by using assays capable of detecting integrin effector binding. Another example is the measurement of conformational changes in elongin C when co-expressed with elongin B, which were determined by the small increase in the intramolecular FRET efficiency using Cerulean and Citrine [219]. A ratiometric chloride indicator Clomeleon based on CFP and Topaz, a variant of YFP, was used for FLIM-FRET measurements as part of a study of neuronal development by monitoring intracellular chloride concentrations. It was found that the FRET signal correlated well with the neuronal development, with the Clomeleon lifetime indicating the concentration of chloride in the neurons [220]. Figure 23 shows an example of a FLIM-FRET experiment, with which Llères et al. demonstrate the effect of the linker peptide length between GFP and mCherry on the lifetime [217]. Notice that a seven amino acids (AA) linker displays a larger change in mean fluorescence lifetime (τ_m) compared with the 17 AA linker. Furthermore, the FRET efficiency is significantly higher compared with the 17 AA linker (Figure 23B,C). In contrast, the negative control of free moving mCherry and GFP in Figure 23A show long lifetimes and zero FRET efficiencies, as expected. The differences in the FRET results are largely due to alterations in geometry and the increased distance between donor and acceptor.

Besides combining FLIM with FRET to study protein-protein interactions in living cells, FLIM is also effectively combined with various other techniques, including super-resolution microscopy using stimulated emission depletion (STED) [221], as further discussed below, in virus, yeast and bacteria detection [6], lifetime-based imaging of fingerprints in forensics [222], in microfluidic [223,224] and lab-on-chip devices [189], and other novel biosensors. Excellent reviews on the technical, theoretical, and biological aspects of FLIM can be found in [6,11,39,40,180,182,183,213,216,217,223,225,226,227]. Guides to easy quantitative FLIM analysis based on phasor analysis, in which the life time data is sine (S) and cosine (G) transformed into a spatial coordinate system, are given in references [228,229] and a modified form to allow time-gated fluorescence lifetime image analysis was recently developed in Hans Gerritsen’s group [230].

Polarization anisotropy imaging is a method based on the measurement of fluorescence polarization [231] as introduced previously in §1.1.2. These measurements of fluorescence polarization offer particular advantages for high-contrast discrimination of FRET with FPs. The concept is based on the fact that excitation with polarized light is only possible in that part of the fluorochrome population that have absorption vectors aligned parallel to the polarization vector of the excitation light (Figure 24A), which is a process called “photoselection”. If the fluorochrome’s transition moment does not change, i.e., the fluorochrome does not rotate, the major part of the fluorescence emission remains parallel to the excitation direction so that the fluorescence can be considered anisotropic in terms of polarization.

The fluorescence anisotropy (r) is defined as the intensity-corrected difference between the emission parallel (I_par) and perpendicular (I_per) to the excitation polarization direction [8]:

(25)

The anisotropy will disappear if the molecules rotate during the nanosecond fluorescent lifetime. However, because of the relatively large size of FPs and their slow rotation, they do not display high fluorescence depolarization during the measurement. If FRET occurs between two FPs that are slightly misaligned, then the polarized fluorescence emission will emerge at a different angle (from the excitation vector), which is a projection of the FP’s rotation. The principal strength of this approach is that measuring fluorescence polarization parallel and perpendicular to the excitation is straightforward with high signal-to-noise ratios. Due to the fact that data can be acquired rapidly and minimal image processing is needed, this approach is well suited for applications in high-content screening.

Polarization anisotropy measurements are not without disadvantages. Any direct excitation of the acceptor must be carefully avoided, because it decreases the donor signal and reduces the signal-to-noise ratio of the measurement. Anisotropy measurements require polarizers, which cause a significant reduction in the emission signal, and consequently higher concentrations of fluorochrome are needed compared with intensity-based measurements to obtain accurate results. In addition, although the polarization FRET technique is excellent in discriminating between the presence and absence of FRET, it is not a good approach for differentiating between strong and weak FRET. This approach is also susceptible to any polarization artifacts caused by the optical system used, e.g., elements in the optical path such as beam-splitters, filters, mirrors. The polarization can be degraded in high numerical aperture objectives, so polarized FRET experiments should be limited to imaging with objectives with numerical apertures of 1.0 or less. As a consequence, polarization anisotropy imaging should only be applied after careful examination of the microscopic setup and control measurements to determine to what extent the optics influence the polarization. A successful application example of this technique is the ability to distinguish FRET between linked and unlinked Cerulean and Venus fluorescent proteins in living cells with a larger dynamic range than other approaches [169].

In upconverting-FRET (UC-FRET), the donor commonly consists of a rare earth metal-containing upconverting chromophore (UCC) or nanoparticle (UCN) (see also §1.2.4) and a normal acceptor fluorochrome (“downconverting”). As in other FRET applications, an overlap requirement between the emission spectrum of the UCC and the excitation spectrum of the acceptor exists. A major advantage is that, because the UCC’s excitation spectrum is well separated from the excitation spectrum of the acceptor due to the large anti-Stoke shift, direct excitation of the acceptor does not occur (see Figure 24B). Further advantages include, sensitive acceptor emission detection, the excitation in the NIR/IR abolishes autofluorescence of endogenous fluorochromes, and the narrow donor emission band can be readily differentiated from the acceptor emission. All these features result in the detection of the sensitized emission against a dark background. Prolonged FRET sensing is particularly feasible with UCN, since their biocompatible coating acts as a shield, which results in low toxicity, high photostability and minimum photobleaching and photodamage. A major disadvantage is that such UCC cannot be genetically encoded or engineered and thus various methods that introduce the UCC or UCN into the cell will lead to some disruption of the cellular homeostasis.

UC-FRET as a technique is relatively new, but despite this a number of reports have appeared over the past few years that include applications in bio-assays, such as nucleic acid hybridization [236], ligand binding [237,238,239], enzyme activity [240], and competitive immunoassays [241]. Although UCN per se are now increasingly being developed for use in cellular and in small animal imaging (reviewed in [13]), FRET-based imaging assays are still scarce, not in the last place because in vivo labeling remains challenging with this technology and more work is required to develop UC-FRET for in vivo applications. Nonetheless, Jiang and Zhang reported an UC-FRET-based method based on attaching a siRNA-BOBO-3 complex to the surface of amino-group-modified silica/NaYF₄:Yb,Er UCN to study intracellular release and biostability of siRNA in live cells [242].

“Frustrated energy transfer” or “exciton blockade” was introduced and examined theoretically by Stefan Hell and co-workers [243]. The basic idea behind this multi-photon fluorescence process is that acceptors brought into the excited state through FRET cannot accept energy from a second excited state donor. Therefore, FRET can deliberately be abrogated, with recovery of the donor emission, by forcing direct excitation of the acceptor, thus “frustrating” FRET. In practice, such FRET abrogation would be most efficiently observed when periodically saturating the acceptor with a modulated light source and concomitant detection of the donor emission with phase-sensitive detectors, such as lock-in amplifiers [182]. Major advantages of this method include that, in contrast to non-resonant multiphoton absorption, this method does not require high local light intensities and a significant increase in resolution is predicted [243]. However, reports of applications are rare in the literature, most likely because recent research has shown that deviations from the theoretical deliberations occur in a number of fluorochromes–FRET has been shown to occur to acceptors that already reside in the excited state [244,245,246], which complicates matters. Energy transfer to ground and excited state acceptors is Förster allowed, because the acceptor does not change its spin and consequently singlet–singlet and singlet–triplet annihilation occurs. In common fluorochromes, such as cyanine- and rhodamine-dyes, which show considerable extinction in the visible spectrum, such annihilation processes cannot be completely avoided [247] and consequently, an energy-transfer blockade remains incomplete.

However, Tinnefeld et al. [248] recently showed that under certain circumstances, such a complete blockade or FRET frustration is possible when using radical anion states of fluorochromes as saturable dark states in “energy transfer blockade probes” (ETBPs) that consist of a donor surrounded by one to n acceptors (n = 0, 1, 2, ...). The authors furthermore demonstrate that these ETBPs exhibit a superlinear donor response because of acceptor saturation and this directly translates into resolution enhancement in confocal microscopy in all three dimensions. Despite these first positive results, FRET frustration as a technology remains immature and has partially been surpassed by a competing technology based on photochromic dyes and proteins.

One of the main limitations in FRET measurements involves the determination if FRET can be lifted, for instance by photobleaching the acceptor (see §3.4.1). Besides being irreversible, photobleaching-based FRET is generally neither suited for live cell experiments not for prolonged imaging. Other FRET techniques suffer from disadvantages such as variable donor to acceptor stochiometries (ratiometric FRET), expensive and elaborate instrumentation (FLIM, spectral, and polarization anisotropy imaging), phototoxicity, or the requirement for a large number of control images. The newly developed photo-switchable FPs conveniently allow a form of FRET called photochromic-FRET (pcFRET) in which a FRET-competent and incompetent state of the acceptor can be reversibly induced by illumination with a particular wavelength. pcFRET was first demonstrated using photo-switchable dyes [249,250] and subsequently further developed based on photo-switchable fluorescent proteins [250,251]. The major advantage of this method is the fact that it allows the continuous and prolonged measurement of FRET in living cells and the determination of FRET efficiencies with high accuracy. The sensitivity in pcFRET can significantly be enhanced by utilizing periodic switching with lock-in detection [250], which would be much more effective compared with the aforementioned FRET frustration.

Equal to FRET frustration, pcFRET aims to take the acceptor out of the equation , but in contrast to FRET frustration, pcFRET provides a more solid and reversible mechanism to switch the acceptor on and off, with fewer interference from alternate decay or transfer pathways. Therefore, it may be anticipated that the application of FRET frustration as a technique will remain limited, whilst pcFRET will become more widely used, especially because of the ease of use and the lack of elaborate equipment.

What traditional FRET techniques have in common are that they report the average behavior of a bulk population of molecules, albeit with temporal and spatial resolution. Events such as the dimerization of proteins to form a cell surface receptor or the measurement of the step size of a DNA helicase cannot be directly visualized or measured using these FRET techniques. In the 1990s, Taekjip Ha developed a technique called single molecule FRET (smFRET) in Simon Weiss’ lab. He used near-field scanning optical microscopy (NSOM) to simultaneously obtain dual color images and emission spectra from donor and acceptor fluorochromes linked by a short DNA molecule and thus provided proof of principle that conformational changes, such as rotations or distance changes, within single biomolecules may be detected with FRET involving a single pair [252]. Both donor and acceptor photobleaching were used to confirm the presence of FRET and to calculate the energy transfer efficiency.

The main advantage of smFRET is the fact that events that are normally obscured in ensemble measurements, because they are canceled out due to random averaging, are now clearly visible and can be measured with high accuracy. A recent study by Biju and co-workers exemplifies this. This study showed that dimers of the epidermal growth factor receptor (EGFR) are continuously formed in cell membranes through reversible association of hetero-dimers [EGF(EGFR)₂] and that the lateral propagation of EGFR activation takes place through transient association of a heterodimer with predimers [(EGFR)₂] [253]. These results extent the previous findings by Gadella and Jovin using FLIM-FRET, which showed that in A431 cells a microclustering of EGFR occurs and that a subclass of receptors are present in a predimerized or oligomerized state [254]. Without smFRET imaging and by using bioconjugated quantum dots, this reversible receptor dimerization in the lateral activation of EGFR would have remained obscured. Nanoparticle-based smFRET (or better said single particle FRET) is increasingly being used for biological studies—predominantly with quantum dots, as exemplified by the study described above. It is even possible to produce smFRET between negatively charged nitrogen-vacancy centers in diamond nanoparticles as donors and organic dyes as acceptor [255], which has the additional advantage that nanodiamonds do not show any intermittent behavior.

smFRET has matured and become more reproducible over the past few years, predominantly because of advances in detector technology, novel fluorochromes, such as the aforementioned nanoparticles, and it has been the subject of intense research, especially to overcome some of the disadvantages associated with smFRET. A major advantage, however, is the fact that smFRET-capable systems can readily and cost-effectively be constructed with off-the-shelf components and either a confocal microscopy, or rather a total internal reflection (TIR) microscope, since smFRET temporal-spatial trajectories are predominantly acquired by imaging surface immobilized molecules. To obtain significantly more information, especially when multiple biomolecules are involved in a complex or multiple steps determine the sequential conformational change in a biomolecule during its normal biological activity, multiple labeling and consequently multiple smFRET would be required. Indeed, Ha and Hohng extended smFRET by using four fluorochromes to concomitantly observe the correlated motions of four arms of the Holliday junction, and to investigate the correlation of RecA-mediated strand exchange events at both ends of a synaptic complex via alternating laser excitation (ALEX) [256]. However, the complexity of the data processing scales with the number of concomitantly used fluorochromes, and therefore there is a limit to what is feasible and practical in multiple smFRET. Furthermore, smFRET in vivo is inherently difficult because of autofluorescence and because FP-labels do not have advantageous photophysical properties for smFRET. Kapanidis et al. aimed to resolve some of these issues and turned to photo-switchable FPs, which allows FRET between a single donor and multiple, spectrally identical photo-switchable acceptors [257]. This “switchable-FRET” method reduces both the experimental and analytical complexity and most likely introduces scalability with regard to monitoring multiple distances simultaneously.

A good practical guide to smFRET can be found in reference [258] and a general overview of single molecule techniques in reference [259]. A general and systematic catalog of FRET techniques, dyes and fluorescent proteins, adapted to various imaging systems, including some new approaches for implementation can be found in more specialized reviews [39,40,121,124,138,182,225,226,227,260,261,262,263,264].

Finally, an alternate approach to measuring protein-protein interactions requires a brief introduction. Bimolecular Fluorescence Complementation (BiFC) is based on the reassembly of a functional fluorescent protein label from non-fluorescent fragments of FPs fused to proteins of interest whose interaction facilitate the label’s reassembly (Figure 24C). Biochemical complementation of fragments has been known for subtilisin-cleaved bovine pancreatic ribonuclease since 1958 [265], but truly conditional fluorescence complementation was first shown by Kerppola et al. with YFP fragments fused to either the basic-region leucine zipper (bZIP)-domain or Rel-family proteins [266].

BiFC offers some advantages over classical FRET-based assays, since these assays only report on interactions by a small number of interacting proteins and therefore require overexpression of interacting partners to obtain sufficient sensitivity. Such overexpression is no reflection of the real cellular state and might even disrupt normal cellular processes or cause formation of non-native complexes; all factors that might perturb results. Furthermore, since in BiFC a functional fluorochrome is formed from non-fluorescent fragments, BiFC is less prone to interference from changes in the fluorescence intensity or lifetime due to interaction unrelated environmental factors than FRET-based assays. In addition, BiFC allows the simultaneous study of multiple interaction partners for a given protein of interest with multiple combinations of fusion proteins.

Conversely, the major disadvantages of BiFC are that: (i) BiFC cannot visualize protein interactions in real time, because of the high stability of the protein complex and the long time necessary for fluorochrome maturation (~8 h for the reconstitution of YFP [266]); and (ii) because of the high complex stability, transient or dynamic interactions in which partners associate for a particular time and subsequently dissociate cannot be imaged. In this sense, FRET is more versatile and allows the imaging of dynamic processes. Detailed technical reviews on BiFC can be found in references [267,268,269,270], whilst excellent overviews were recently provided by Kerppola [271,272].