Advanced fluorescence microscopy techniques for the life sciences

The development of super-resolved fluorescence microscopy, for which the Nobel Prize was awarded in 2014, has been a topic of interest to physicists and biologists alike. It is inevitable that numerous questions in biomedical research cannot be answered by means other than direct observation. In this review, advances to fluorescence microscopy are covered in a widely accessible fashion to facilitate its use in decisions related to its acquisition and utilization in biomedical research.


INTRODUCTION
The advent of optical microscopy has had a significant impact on the life sciences specially in visualizing the molecular architecture of the cell [1][2][3][4][5][6][7][8][9][10][11] . The development of fluorescent probes has allowed observations at the single molecule level 12 whilst the recent introduction of super resolution microscopy has allowed imaging of an array of nanoscopic biological complexes with unprecedented resolution 1,[5][6][7]9 . In this review, advanced fluorescence microscopy techniques are discussed in an application-oriented fashion.

FLUORESCENCE
A quantum mechanical treatment of an atom reveals the probability of an associated electron, possessing an amount of available energy, of occupying any energy level of the atom. At scales larger than molecular scales, levels group into bands, sometimes referred to as states. Often, electrons occupy the lowest energy state (referred to as the ground state). Electrons could also occupy higher energy states if the ground state is fully occupied or if they are provided with the energy differential sufficient to excite them from the ground state to any unoccupied upper energy state. The energy differential can be provided in the an electron excited (solid black arrow) by an incident photon (blue arrow) from the ground state (S 0 ) to the first state (S 1 ) can be stimulated to relax to S 0 by another photon (orange arrow) of higher wavelength and emitting a photon (red arrow) of higher wavelength. The wavelength of a spontaneously emitted photon is higher than that produced by stimulated emission. after excitation. It is evident that this process is temporary and thus the term blinking. Photobleaching, or permanent loss of fluorescence, can occur due to an alteration to the chemical structure of a fluorescent label following long periods of laser exposure. Typically, photo blinking occurs on the scale of a few seconds, whilst photobleaching occurs after 0.1 -1000 seconds of continuous exposure depending on the label.
Although often implicated as a limiting factor in single particle tracking experiments, photobleaching has been useful in determining the stoichiometry of protein complexes. It was shown that by exciting singly-labeled monomers in an oligomeric complex and recording the overall decrease in intensity caused by photobleaching, the stoichiometry of the complex under study can be deduced by counting the photobleaching steps 15 .
In general, photostability (i.e. photo -blinking and bleaching) determine the fate of a fluorescent label and can sometimes limit its application in biophysical investigations. Whilst photo stability determines the traceability of a fluorescent label, brightness determines its detectability. Quantum yield or the ratio of the number of photons emitted to the number of photons received by a fluorescent label. Higher quantum yields denote higher probabilities of fluorescence detection. There is a limit, however, on how bright a label could be. A large label is typically brighter than a smaller label. Yet, for many biological processes, where the molecule of interest is smaller than the label or its dynamics can be affected by the size of the label, a large label cannot be used ( Figure 2). Other parameters such as labeling schemes, coupling efficiency, and compatibility with live-cell imaging are also considered when choosing an appropriate label (Table 1). A detailed overview of the different labels and their associated parameters can be found in [16][17][18].

Diffraction-limited microscopy techniques
The wave nature of light imposes a fundamental limit on the dimensions of the smallest object that can be observed using light microscopy. The limit was derived by Ernest Abbe who found that any two objects could be resolved if the distance in between is . The shell can be functionalized (i.e. grafted) with a chemical moiety to enable targeted labeling of a protein. Furthermore, the size and composition of the shell can be tuned to improve the photo stability of a quantum dot for single molecule imaging purposes. Though possessing high quantum yields, quantum dots suffer from poor intercellular internalization and uncharacterized size-dependent effects on protein dynamics. (B) Organic dyes are small fluorescent moieties that can be linked to the protein under study via a chemical tag. The small size and high quantum yield of organic dyes render them suitable labels for single molecule imaging. However, their potential phototoxic effects make a large subset, yet, unsuitable for imaging in living cells. Shown in the figure is the chemical structure of the organic dye Rhodamine 6G. (C) Fluorescent proteins are genetically expressible labels which are suitable for imaging in living systems. However, their utility for single molecule imaging is dependent on their proper folding and the size of the protein under study. Shown in the figure is the structure of the Green Fluorescent Protein (GFP). approximately larger than half of the wavelength of the excitation source. This implies that any two fluorescently-labeled entities, observed using visible light (450 -650 nm), cannot be discerned if they are closer than 200 -300 nm ( Figure 3). Fluorescence microscopy techniques which cannot overcome that resolution (diffraction) limit are known as diffraction-limited and are discussed below. Super resolution microscopy techniques which can overcome the diffraction limit of light are explained in the next section.

Epi-fluorescence illumination microscopy
The use of epi-illumination had originally referred to the arrangement of optical components that permits illumination from above the specimen. The term now refers to any orthogonal scheme of illumination whereby the excitation source is directed perpendicular to the plane of the specimen. Epi-fluorescence illumination has, and still is, implemented using xenon arc & mercury vapor lamps, Tungsten-Halogen, and Light Emitting Diodes (LEDs) 19 . However, the strong divergence and incoherence of these excitation sources meant that single-molecules could hardly be observed in thin samples even at substantially-high light levels. To observe single molecules using epi-fluorescence illumination, a laser beam is directed through an objective lens onto a sample and the fluorescence emission is collected using the same objective 12,20 . The emission and excitation are separated, where then, the former can be imaged ( Figure 4A). Although this scheme serves to effectively image thin (< 1 µm) samples, it fails to image dense thick samples where the fluorescence signals from the different imaging planes destructively contribute to the acquired image. The simple illumination scheme has served to uncover the reversible assembly of the 'Tat A' component of the twin-arginine protein transport system 21 .

Confocal microscopy
To ameliorate the aforementioned problem, an upwardly-directed laser beam is focused inside the sample and a micron-sized pinhole is placed in the emission path to restrict the For a list of fluorescent proteins refer to [18] Notes. *Non-exhaustive Figure 3. Schematic diagrams representing the diffraction limit of light. Due to the wave nature of light, a point object observed in the far field is seen as an airy disk. If the distance between two point objects is (A) greater than, or, (B) equal to half of wavelength of the excitation source, the objects become resolvable, and can be precisely located. If the distance between the two point objects is (C) less than half of the wavelength of the excitation source, the objects become unresolvable. This resolution limit is known as the Abbe diffraction limit. Closely located point objects which are not resolvable using diffraction-limited microscopy techniques (D) can be resolved using super-resolution microscopy techniques (E). Unlike diffraction-limited microscopy techniques which operate by detecting the fluorescence from all the molecules in a single image, super-resolution microscopy detects the fluorescence from each molecule, or a subset of molecules, at different times; overlapping the images obtained for distinct fluorophores results in a super-resolved image.
emission to that from a micron-sized spot and eliminate any out-of-focus signal ( Figure 4B).
To image a single plane, the spot is laterally scanned across the sample 22 . Scanning is typically performed by an electrically-expandable material, a set of electronically-controlled mirrors or a multiple-hole spinning disk which deflect the focused beam laterally across the sample. In all cases, acquisition is not set by the detector's frame rate, as in epi-illumination microscopy, but by the scanning speed. Although the fine sectioning of biological samples in confocal microscopy produces high-resolution images, the limitation on the acquisition speed renders confocal microscopy an unsuitable technique for imaging highly dynamic processes.

Total Internal Reflection Fluorescence Microscopy (TIRF-M)
Light incident at a boundary between two different media changes its direction of propagation; this process is known as refraction. If the angle of incidence exceeds the critical angle dictated by the ratio of the refractive indices of the two mediums, light undergoes total internal reflection. To satisfy boundary conditions, an evanescent (decaying) field propagates parallel to the plane of the boundary. The depth of penetration of the evanescent field is dictated by the angle of incidence and the ratio of the medium's refractive indices; however, it can be adjusted to a few hundred nanometers. The evanescent field exclusively excites fluorophores at the vicinity of the boundary. This illumination scheme is known as Total Internal Reflection Fluorescence Microscopy (TIRF-M) ( Figure 4C) 23 . Due to its simple implementation, TIRF-M found numerous applications in biology, especially where it relates to imaging lipid membranes and its constituents. The superior In Epi-illumination (A), an upwardly-directed laser beam, under filling a lens objective, is directed perpendicular to the plane of the fluorescent sample and the fluorescence, from a large portion of the sample, is collected using the same objective. For large samples, the acquired image is blurred as a result of the overlapping of the fluorescence signal obtained from the different imaging planes. In confocal microscopy (B), an upwardly-directed laser beam is focused at specific planes of a sample to excite fluorophores inside the focusing spot. By scanning the beam laterally across the sample, a full image of a specific plane can be acquired. Acquisition rates in confocal microscopy are limited by the scanning speed of the beam rendering image acquisition slow. To resolve that problem, light is directed at an angle, above the critical angle, to undergo total internal reflection at the interface (C). To satisfy the boundary conditions at the interface, an evanescent (decaying) wave propagates for a few hundred nanometers above the boundary exciting fluorophores at the vicinity of the interface. The acquired image is not blurred as a result of the emission being limited to a thin imaging plane. Furthermore, acquisition rates in TIRF microscopy is only limited by the camera frame rate rendering image acquisition fast. In light sheet microscopy (D) a laser beam is directed though a cylindrical objective, or a twodimensional photonic lattice, creating a light sheet that excites a section through a fluorescent sample. The fluorescence from the excited plane is orthogonally collected. LSFM combines fine optical-sectioning with fast acquisition rates. I i is the incident beam, I r is the reflected beam, and, I f is the fluorescence emission.
signal-to-noise ratio of TIRF-M has been configured for real time measurements on numerous transporters and channels at the single molecule level [24][25][26][27][28][29][30][31] (Figures 5 and  6). Although lipid membranes and their constituents can be easily imaged in TIRF-M, the richness of intra-cellular biology is not accessible to this technique. The gating and alternating action of an ion channel, or transporter, can also be detected by single-molecule FRET. In FRET, energy is transferred between a donor (blue) and acceptor (orange) when they are less than 10 nm apart. The ratio of the emission of the acceptor to the emission of the donor, known as the FRET ratio, is a direct measure of how far both fluorophores are. By labeling ion channels, and other biomolecules, at closely located sites, using suitable FRET pairs, and monitoring the FRET ratio, it is possible to observe proteins dynamics in real time. Reproduced with permission from [63].

Light Sheet Fluorescence Microscopy (LSFM)
Light sheet microscopy was developed to overcome the limitations imposed by confocal and TIRF microscopies 32 . Light sheet microscopy relies on an orthogonal illumination scheme whereby a couple of microns' thick sheet of light, produced by directing a Gaussian-shaped beam though a cylindrical lens objective, illuminates a section of a fluorescently labeled sample. The fluorescence signal from the section is collected by another objective placed perpendicular to the excitation beam ( Figure 4D).
Although this scheme enables video rate imaging and effectively reduces photo bleaching, the thickness of the light sheet does not permit fine sectioning, and, therefore images suffer from reduced contrast. Lattice light sheet microscopy was recently reported by Chen and coworkers where the cylindrical-lens system was replaced by a twodimensional optical lattice 33 . This configuration produces a hundred nanometers' ultra-thin light sheets that can be used to illuminate thin sections of a sample. This improvement permits 3D intra-cellular imaging at sub second temporal resolutions combined with unprecedented axial and lateral spatial resolutions (Figure 7). Light sheet microscopy has been instrumental in uncovering the organization of RNA polymerase II in vivo 10 , tracking single molecules on the apical membrane of living cells 34 and imaging mouse embryos during the first development stages 35 .

Super resolution microscopy techniques
Since many biological molecules perform their functions cooperatively, in nanoscopic molecular complexes (5 nm -100 nm), diffraction-limited fluorescence microscopy techniques fail to uncover the architecture, and eventually dynamics, of those biological machineries. To overcome the diffraction limit, a set of fluorescence microscopy techniques, known as super resolution microscopies, were developed since 1994. Super resolution microscopies rely on localizing individual fluorophores by sequentially modulating their fluorescence and constructing a super-resolved image by super position of the exclusively localized fluorophores. Methods for fluorescence modulation are figuratively summarized in Figure 8 and explained in [36].

Photoactivated Localization Microscopy (PALM)
The engineering of photoactivatable fluorescent proteins has prompted the development of PALM in 2006 37 . In PALM, a weak short-wavelength laser beam randomly activates a subset of inactive fluorophores. The activation beam has a low intensity enough to ensure that fluorophores in the same diffraction limited region are unlikely to be all activated at the same time. Another beam, of longer wavelength and higher intensity, is used to excite the activated fluorophores and the emission is subsequently collected. Since the distance between the majorities of the active fluorophore pairs is larger than the diffraction limit, the centers of the associated airy disks can be determined and the position of the fluorophores can be precisely located.
Following localization, the active fluorophores are photobleached and another subset of inactive fluorophores is activated, excited, localized and photobleached. To ensure the localization of the majority of the fluorophores residing in the desired optical section, the aforementioned steps are repeated and the images of the localized fluorophores are overlapped until a satisfactory complete super-resolved image is produced ( Figure 9A).
Since multiple fluorophores can be localized at the same time, PALM is considered to be a wide-field imaging technique. Although acquisition rates in wide-field imaging techniques (e.g. LSFM and TIRF) are typically considered to be limited by the detector's frame rate, acquisition in PALM is relatively slow; limited by the total number of frames needed to acquire a complete super-resolved image. The affordable engineering of photoactivatable fluorophores and the applicability of PALM in TIRF and confocal (fPALM) 38 configurations have encouraged its use in visualizing the dynamic architecture of the endocytic machinery 1 , and, the nanoscopic oligomerization of hemagglutinin in biological membranes 39 .

Stochastic Optical Reconstruction Microscopy (STORM)
The photobleaching of active fluorophores, in PALM, has been traditionally seen as an additional time consuming step, which if eliminated, would enable faster imaging. PALM and STORM share common operation principles, in particular, the random activation and localization of a subset of fluorophores. Nonetheless, the use of STORM is distinguished by the use of photoswitchable fluorophores which can be activated (turned on) and deactivated (turned off) at will ( Figure 9B) 40 . This improvement discards the requirement for successively photobleaching and activating fluorophores, thereby shortening the total acquisition time. STORM was first demonstrated using single fluorophores, however, the use of fluorophore-pairs was recently reported 41,42 where, the emission of an activator fluorophore, following excitation, was shown to activate a photoswitchable reporter fluorophore; this process, when iterated, yields super-resolved images. STORM has been recently used in elucidating the mechanism of actions of the Nucleoid-associated proteins Figure 9. Schematic diagrams demonstrating the mechanism of operation of PALM, STORM and STED. In PALM (A), a randomly selected subset of a large number of closely-located inactive fluorophores (empty black circles) is activated using a weak short wavelength laser beam. The activated fluorophores (empty orange circles) are excited using a longer wavelength laser beam and the fluorescence (solid orange circles) is collected. Since the distance between the exclusively excited fluorophores is larger than the diffraction limit, they can be localized (black dots). The fluorophores are photo bleached under continuous excitation and another set of fluorophores is activated, excited, localized and photo bleached. This process is repeated until a complete super resolved image is obtained. In STORM (B), a randomly selected subset of a large number of closely located inactive switchable fluorophores (empty black circles) is switched using a weak short wavelength laser beam. The activated fluorophores (empty orange circles) are excited using a longer wavelength laser beam and the fluorescence (solid orange circles) is collected and localized. Instead of undergoing photo bleaching, the activated fluorophores are deactivated, or switched off, to a dark state by the application of a weak short wavelength laser beam, which in turn randomly activates (switches on) another subset of inactivated fluorophores. This is repeated until a complete super resolved image is obtained. In STED (C) a laser beam (large blue circle) excites a large number of active fluorophores (orange solid circles) in a diffraction limited region. A doughnut shaped beam (large orange circle), centered on the center of the excitation beam, de-excites all the excited fluorophores (red solid circles) except for the very few which are located at the center (orange solid circles). (continued on next page. . . )

. continued)
The stimulation of the de-excited fluorophores results in the emission of photons having a wavelength longer than that emitted spontaneously from the center of the doughnut shaped beam. Both emissions are filtered and the fluorophores at the center of the beam are localized (black dots) accordingly. By scanning the beam and repeating the excitation and stimulation processes, a complete super-resolved image is obtained.
in chromosome organization 11 , visualizing the architecture of apoptotic pores 43 (Figure 10), and, the periodic arrangement of actin in axons 44 (Figure 11).

Ground State Depletion followed by Individual return Microscopy (GSDIM)
GSDIM relies in its operation on depleting the ground state by exciting all fluorophores in a field of view to an upper state 45,46 . The excited electrons relax to a long-lived dark ''off'' state by irradiative emission. Individual molecules return back, stochastically, from the dark state to the ground state by emitting a photon. The stochastically-emitted fluorescence is localized and a complete super-resolved image of a sample is constructed following the  return, and localization, of all the excited fluorophores. Similar to other super resolution microscopy techniques, GSDIM has been instrumental in facilitating numerous discoveries as, for instance, visualizing the structural organization of the nuclear pore complex 8 ( Figure 12).

Stimulated Emission by Depletion (STED) microscopy
As opposed to stochastic functional techniques like STORM and PALM, where a random subset of fluorophores is imaged at any one time, STED is a spatially-targeted technique denoting that a few fluorophores at a certain position are chosen to be imaged at any one time. STED exploits the difference in the wavelength between spontaneous and stimulated emission 47 . A collimated beam excites a large number of fluorophores in an optical section and another, doughnut shaped beam, of larger wavelength, centered at the excitation beam forcefully depresses the majority of the fluorophores to the ground state. The size of the annulus of the doughnut shaped beam can be adjusted down to molecular scales. The fluorophores in the middle of the doughnut-shaped beam, emit spontaneously whilst the forcefully depleted fluorophores emit stimulatingly.
Since both emissions possess different wavelengths, they can be separated and the spontaneously emitted fluorescence signal is localized. Scanning the excitation, and stimulation, beams laterally across the sample yields a complete super-resolved image of the optical section ( Figure 9C). It can be shown that the obtained resolution using STED is inversely proportional to the intensity of the doughnut shaped beam. Therefore, the use of STED requires the engineering of special fluorophores 48 that can withstand high laser powers and that the entire sample is not damaged at such high laser intensities.
STED has been used in the study of a number of biological systems including the clustering of the envelope protein during the budding of HIV-1 particles 7 , the periodic organization of neural cytoskeleton 6 and the transient formation of cholesterol-mediated nanoscopic complexes in biological membranes 49 . STED is the first super resolution microscopy technique to be successfully tested on living animals 50 ( Figure 13). Images obtained using diffraction-limited and super-resolution microscopies are shown in Figure 14.

Structured Illumination Microscopy (SIM)
As opposed to STED, where the true likeness of a sample is obtained after imaging, SIM is another super resolution technique that exploits the ability of periodic illumination to produce negative imprints of a sample. SIM employs a grid-like light pattern to illuminate a sample [51][52][53] . The illumination pattern is periodic with high intensity regions a wavelengthseparated from low intensity regions. The high intensity regions excite the underlying fluorophores leaving the fluorophores underlying the low intensity regions in the ground state. By rotating the illumination pattern, an image of dark regions on a bright background (negative imprint) can be obtained.
To obtain a true super-resolved image, the negative imprints undergo advanced image processing, therefore, SIM requires that the sample be immobilized during pattern rotation and on other hand, non-trivial computation expertise to deconvolve the obtained images. The speed of acquisition is fast but is generally limited by the speed of pattern rotation and the total number of images acquired during this process. Recently, however, SIM's pattern illumination scheme has inspired the development of a 100,000 folds faster variant  of STED microscopy 54 . The reported scheme relies on orthogonal illumination obtained by passing two STED beams across two perpendicularly aligned diffraction gratings that split the beams onto periodic positions across the sample plane. The recently developed technique has been demonstrated on the imaging of 100-squared microns of GFP-labeled Keratin at less than a second temporal resolution and the dynamics of filopodia-like structures at relevant time scales (Figure 15).

Binding-Activation Localization Microscopy (BALM)
Dissimilar to PALM, STORM, STED and SSIM which can only provide positional information, BALM can also provide binding kinetics. BALM was developed with the aim of visualizing the organization of the bacterial chromosomes in E coli 55 . The DNA binding dyes, YOYO-1 and PicoGreen, were used to visualize the organization of DNA strands at a resolution of 10 nm (Figure 16). The fluorescence of the DNA-binding dyes increases up to 800-fold upon binding. A complete super-resolved image of the sample under study is obtained by the repeated localization and photobleaching of the sequentially binding fluorophores. BALM was also used to visualize the structural organization of amyloid fibrils 9 .

Scanning Near field Optical Microscopy (SNOM)
SNOM was developed to break the diffraction limit of light by abolishing the far field, lens-based, imaging system using a tip brought close to a sample, in the near field, to allow imaging at sub-diffraction limited resolutions 56 . SNOM is commonly implemented in a TIRF configuration, whereby the evanescent field resulting from the total internal reflection of an incident beam can be detected and imaged using a fiber probe placed at a distance less than its penetration depth. The use of a scanning probe to construct a super resolved image of a sample implies that the acquisition rates using SNOM are limited by the probe scanning speed and that the sample, and probe, needs to be stabilized for good quality imaging. These requirements limit the use of SNOM in biological investigations.

CONCLUDING REMARKS
The choice of a microscope is dependent on the spatial organization of the biological entity under study, its temporal dynamics and susceptibility to phototoxicity (Table 2). Generally, super-resolution microscopy techniques offer high spatial resolutions at the expense of slow acquisition rates. This rule does not apply for parallelized STED which affords super resolution imaging at high temporal resolutions but with toxic high light levels.
The use of high light levels to perform super resolution imaging has strongly hampered the progress of live-cell imaging. Combining super resolution microscopy techniques is a popular approach to mitigate this problem. For example, multiple off-state transitions (MOST), also known as protected STED, combines the working principle of stochastic localization microscopy techniques, such as STORM or PALM, with STED to offer 2D coordinate-targeted fluorescence nanoscopy with low light levels 57 .
The combination of STORM, SIM and LSFM was also reported to exclusively allow 3D live-cell super-resolution imaging using non-toxic low light levels at biologically-relevant timescales 58 (Figure 17). Despite these advances, an unfortunate fact remains that the combined fluorescence microscopy techniques are, as yet, incapable of accessing the wide spectrum of time and lengths scales on which biological processes occur. Simultaneous non-fluorescent-and fluorescent-based imaging tools can probe the extremes of those scales. Correlative Light and Electron Microscopy (CLEM) is one example in which the angstrom spatial resolution of electron tomography is combined with the intermediate temporal resolution of fluorescence microscopy to construct functional models for the dynamics of protein assemblies 59,60 .
Simultaneous label-free scattering and fluorescence imaging has also emerged as a powerful tool blending the micro-second temporal resolution of scattering microscopy with the molecular specificity of fluorescence microscopy to image highly dynamic membrane processes 61 . These, and more, developments, which represent a summit in biophotonic research, have formed a pathway that should lead to a satisfactory futuristic understanding of the operating principles of the tiny biological machineries, however, a comprehensive bottom-up molecular-based understanding of the functions performed by tissues and organs is still lacking. The next era should witness combined efforts to develop more advanced microscopy techniques capable of deep-tissue imaging for longer and larger,   Notes. *Non-exhaustive ** Approximate value *** Using parallelized STED beams Reproduced with permission from [58].
smaller and faster 62 , to dissect the network of interactions of the simple biological building blocks and elucidate its complex role in directing and regulating life.