Fluorescence Microscopy PDF

97 3 Fluorescence Microscopy Jurek W. Dobrucki The human eye requires contrast to perceive details of objects. Several ingenious methods of improving the contrast of microscopy images have been designed, and each of them opened new applications of optical microscopy in biology. The simplest and ve...

97 3 Fluorescence Microscopy Jurek W. Dobrucki The human eye requires contrast to perceive details of objects. Several ingenious methods of improving the contrast of microscopy images have been designed, and each of them opened new applications of optical microscopy in biology. The simplest and very effective contrasting method is ‘‘dark field.’’ It exploits the scattering of light on small particles that differ from their environment in refractive index – the phenomenon known in physics as the Tyndall effect. A fixed and largely featureless preparation of tissue may reveal a lot of its structure if stained with a proper dye – a substance that recognizes specifically some tissue or cellular structure and absorbs light of a selected wavelength. This absorption results in a perception of color. The pioneers of histochemical staining of biological samples were Camillo Golgi (an Italian physician) and Santiago Ramon y Cajal (a Spanish pathologist). They received the 1906 Nobel Prize for Physiology or Medicine. A revolutionary step in the development of optical microscopy was the introduction of phase contrast proposed by the Dutch scientist Frits Zernike. This was such an important discovery that Zernike was awarded the 1953 Nobel Prize in Physics. The ability to observe fine subcellular details in an unstained specimen opened new avenues of research in biology. Light that passes through a specimen may change polarization characteristics. This is a phenomenon exploited in polarization microscopy, a technique that has also found numerous applications in material science. The next important step in creating contrast in optical microscopy came with an invention of differential interference contrast by Jerzy (Georges) Nomarski, a Polish scientist who had to leave Poland after World War II and worked in France. In his technique, the light incident on the sample is split into two closely spaced beams of polarized light, where the planes of polarization are perpendicular to each other. Interference of the two beams after they have passed through the specimen results in excellent contrast, which creates an illusion of three-dimensionality of the object. Fluorescence Microscopy: From Principles to Biological Applications, First Edition. Edited by Ulrich Kubitscheck.  2013 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2013 by Wiley-VCH Verlag GmbH & Co. KGaA. 98 3 Fluorescence Microscopy However, by far the most popular contrasting technique now is fluorescence. It requires the use of so-called fluorochromes or fluorophores, which absorb light in a specific wavelength range, and re-emit it with lower energy, that is, shifted to a longer wavelength. Today, a very large number of different dyes with absorption from the UV to the near-infrared region are available, and more fluorophores with new properties are still being developed. The principal advantages of this approach are a very high contrast, sensitivity, specificity, and selectivity. The first dyes used in fluorescence microscopy were not made specifically for research, but were taken from a collection of stains used for coloring fabrics. The use of fluorescently stained antibodies and the introduction of a variety of fluorescent heterocyclic probes synthesized for specific biological applications brought about an unprecedented growth in biological applications of fluorescence microscopy. Introduction of fluorescent proteins sparked a new revolution in microscopy, contributed to the development of a plethora of new microscopy techniques, and enabled the recent enormous growth of optical microscopy and new developments in cell biology. A Japanese scientist, Osamu Shimomura, and two American scientists, Martin Chalfie and Robert Y. Tsien, were awarded a Nobel Prize for Chemistry in 2008 for the discovery and development of green fluorescent protein (GFP) in biological research. The use of fluorophores requires several critical modifications in the illumination and imaging beam paths. Fluorescence excitation requires specific light sources, and their emission is often recorded with advanced electronic light detection devices. However, fluorescence exhibits features that the optical contrast techniques do not show: fluorescent dyes have a limited stability, that is, they photobleach, and then may produce phototoxic substances. This requires special precautions to be taken. All the technical and methodological questions involved are presented and discussed in this chapter. 3.1 Features of Fluorescence Microscopy 3.1.1 Image Contrast In full daylight, it is almost impossible to spot a firefly against the background of grasses, bushes, and trees at the edge of a forest. At night, however, the same firefly glows and thus becomes visible to an observer, while the plants and numerous other insects that might be active in the same area are completely undetectable. This example parallels, to some degree, the observation of a selected molecule under a fluorescence microscope. To be useful for an investigator, optical microscopy requires high contrast. This condition is fulfilled well by fluorescence microscopy, where selected molecules are incited to emit light and stand out against a black background (Box 3.1), which embraces a countless number of other molecules in the investigated cell (Figure 3.1).  3.1 Features of Fluorescence Microscopy 99 Box 3.1 Discovery of Fluorescence Fluorescence was first observed by an English mathematician and astronomer, Sir John Frederick William Herschel, probably around 1825. He observed blue light emitted from the surface of a solution of quinine. Sir John F.W. Herschel was a man of many talents. He made important contributions to mathematics, astronomy, and chemistry. He also made a contribution to photography. He studied and published on the photographic process and was the first one to use the fundamental terms of analog photography – ‘‘a negative’’ and ‘‘a positive.’’ He is thought to have influenced Charles Darwin. Sir John F.W. Herschel described this observation in a letter to The Royal Society in London, in 1845: A certain variety of fluor spar, of a green colour, from Alston Moor, is well known to mineralogists by its curious property of exhibiting a superficial colour, differing much from its transmitted tint, being a fine blue of a peculiar and delicate aspect like the bloom on a plumh... He found a similar property in a solution of quinine sulfate:... Though perfectly transparent and colourless when held between the eye and the light, or a white object, it yet exhibits in certain aspects, and under certain incidences of the light, an extremely vivid and beautiful celestial blue color, which, from the circumstances of its occurrence, would seem to originate in those strata which the light first penetrates in entering the liquid... (Herschel, 1845a). In the next report, he referred to this phenomenon as epipolic dispersion of light (Herschel, 1845b). Herschel envisaged the phenomenon he saw in fluorspar and quinine solution as a type of dispersion of light of a selected color. In 1846, Sir David Brewster, a Scottish physicist, mathematician, astronomer, well known for his contributions to optics, and numerous inventions, used the term internal dispersion in relation to this phenomenon. In 1852, Sir John Gabriel Stokes, born in Ireland, a Cambridge University graduate and professor, published a 100 page long treatise ‘‘On the Change of Refrangibility of Light’’ (Stokes, 1852) about his findings related to the phenomenon described by Sir John Herschel. He refers to it as ‘‘dispersive reflection.’’ The text includes a short footnote, in which Stokes said: I confess I do not like this term. I am almost inclined to coin a word, and call the appearance fluorescence, from fluor-spar, as the analogous term opalescence is derived from the name of a mineral It was known at that time that when opal was held against light, it appeared yellowish red; however, when viewed from the side, it appeared bluish. Thus, the phenomenon was similar to the one observed by Herschel in fluorspar. We now know that opalescence is a phenomenon related to light scattering, while the phenomenon seen in a solution of quinine was not exactly a dispersion of 100 3 Fluorescence Microscopy light of a selected wavelength, but rather the emission of a light of a different color. Sir J.G. Stokes is remembered for his important contributions to physics, chemistry, and engineering. In fluorescence, the distance, in wavelength, between the maximum of excitation and emission is known as Stokes shift. The physical and molecular basis for the theory of fluorescence emission was formulated by a Polish physicist Aleksander Jabłoński (1945). Hence, the energy diagram that describes the process of excitation and emission of fluorescence is called the Jabłoński diagram. Aleksander Jabłoński was a gifted musician; during his PhD years, he played the first violin at the Warsaw Opera. In postwar years, he organized a Physics Department at the University of Toruń in Poland and worked there on various problems of fluorescence. (a) (b) (c) (d) (e) (f) (g) Figure 3.1 Importance of image contrast plasma and mitochondrial membranes and in microscopy. (a) Four dark-gray dots are is accumulated inside active mitochon- almost undetectable against a light-gray dria. JC-1 monomers emit green fluores- background. (b) Four white dots, in the cence. At high concentrations (above 0.1 same positions as in (a), are easily dis- µM in solution), JC-1 forms the so-called cerned against a black background. (c) A J-aggregates emitting red luminescence. A transmitted light image of a live cell in cul- gradient of electric potential is responsible ture – almost no internal features can be for the passage of JC-1 molecules across distinguished when no contrasting tech- the membranes into active mitochondria. nique is used. (d) Nuclear DNA stained Mitochondria characterized by a high mem- with DAPI; DAPI is a heterocyclic molecule brane potential accumulate JC-1 and the with affinity for DNA. The major mode of dye reaches the concentration that is suf- binding to DNA is thought to be dependent ficiently high to form J-aggregates (Smiley on positioning a DAPI molecule in a minor et al., 1991; Cossarizza et al., 1993). (g) Flu- groove of a double helix. (e, f) Low (green, orescence signals from mitochondria and e) and high (red, f) potential mitochondria nucleus overlaid in one image. (Images fluorescently labeled with JC-1. JC-1 is a c-g courtesy of Dr A. Waligórska, Jagiellonian carbocyanine dye, which readily crosses University). The range of applications of fluorescence microscopy was originally underes- timated. Fluorescence microscopy was seen as just a method of obtaining nice, colorful images of selected structures in tissues and cells. The most attractive feature of contemporary fluorescence microscopy and many modern imaging and analytical techniques, which grew out of the original idea, is the ability to image and study quantitatively not only the structure but also the function, that is, physiology of intact cells in vitro and in situ.  3.1 Features of Fluorescence Microscopy 101 Figure 3.1 illustrates the fact that modern fluorescence microscopy should not be perceived merely as a technique of showing enlarged images of cells and subcellular structures but as a way of studying cellular functions. Figure 3.1e–g show some selected structures within a cell (mitochondria and the cell nucleus) and demon- strate the ability to selectively convert a physiological (functional) parameter – in this case, mitochondrial potential – into a specific fluorescence signal. 3.1.2 Specificity of Fluorescence Labeling Let us expand the analogy of the firefly. Although one does not see any features of the firefly – its size, shape, or color remain unknown – the informed observer knows that the tiny light speckle that reveales the position of a male firefly. It is most likely a European common glowworm, Lampyris noctiluca. No other insects are expected to emit light while flying in this region at this time of the year. Although there may be hundreds of insects hovering in this place, they blend into the black background and therefore are not visible. Thus, a tiny light label is characteristic of a firefly: it is specific. Specificity of fluorescence labeling of selected molecules of interest as well as specificity of translating some selected physiological and functional parameters such as membrane potential or enzyme activity into specific signals in a cell is another important advantage of fluorescence microscopy (Figures 3.1 and 3.2). Hundreds of fluorescent molecules, both small heterocyclic molecules and proteins, are available to be used as specific labels and tags in fixed and live cells. Also, a host of methods have been optimized for attaching a fluorescent tag to a molecule of interest; they are discussed in Chapter 4. Specificity of labeling is an important advantage of fluorescence microscopy. However, this specificity is never ideal and should not be taken for granted. Let us consider the example given in Figure 3.2. DRAQ5 binds DNA fairly specifically; staining of RNA is negligible owing to either a low binding constant or a low fluorescence intensity of the DRAQ5 complex with RNA, as long as the concentra- tion of DRAQ5 is low. At high DRAQ5 concentrations, when all binding sites for the dye on DNA are saturated, some binding to RNA can be detected. A similar phenomenon is observed with most, if not all, fluorescent labels that exhibit affinity for DNA. Thus, experimental conditions, especially the ratio between the available dye and the available binding sites, have to be optimized in order to fully exploit the advantages of the DNA-labeling techniques. As for Col-F, this small molecule binds to two major components of the extracellular matrix; in this respect, it is less specific than antibodies directed against selected epitopes on collagen or elastin in a selected species. The advantage of Col-F is the simplicity of labeling, deep tissue penetration, and low level of nonspecific staining. The advantage of immunoflu- orescence in the detection of collagen or elastin is its very high specificity, but it comes with limitations: shallow penetration into tissue and some nonspecific bind- ing of the fluorescently labeled secondary antibody used in the labeling method. An experimenter has to make a choice between the advantages and limitations of the low molecular weight label and the immunofluorescence approach. Similarly, 102 3 Fluorescence Microscopy (a) (b) Figure 3.2 Specificity of fluorescence la- (EdU), was incorporated into nascent DNA beling. (a) Three fluorescent probes were during a short period of S-phase in the divi- used to stain different classes of molecules sion cycle, in cells exposed to an inhibitor of and structures in a fragment of live con- topoisomerase 1, camptothecin. Cells were nective tissue – DNA in cell nuclei (DRAQ5, fixed and the EdU was labeled fluorescently a deep-red emitting dye, shown here as using ‘click chemistry’ (Salic and Mitchi- blue), fibers of the extracellular matrix (Col- son, 2008). Newly synthesized DNA is thus F, green) and active mitochondria (tetram- marked in green. In the same cell, γ H2AX, a ethylrhodamine abbreviated as TMRE, red). phosphorylated form of histone H2AX, which Scale bar, 50 µm. DRAQ5 is an anthra- is considered a marker of DNA double- cycline derivative, which has high affinity strand breaks (DSBs) (Sedelnikova et al., for DNA. The dye readily crosses plasma 2002), was labeled using a specific antibody. and nuclear membranes and binds to DNA Thus, DNA regions with DSBs are immuno- in live cells. It is optimally excited by red labeled in red. When images of numerous light, emitting in deep red (Smith et al., foci representing DNA replication and DNA 1999). Here, it is shown as blue to be dis- damage are overlaid, it becomes apparent tinguished from TMRE. Col-F is a dye that that most damage occurred in replicating binds to collagen and elastin fibers (excita- DNA (replication and γH2AX foci show large tion – blue, emission – green) (Biela et al., yellow areas of overlapping green and red 2013). TMRE enters cells and is accumulated signals) (Berniak et al., 2013). Scale bar, by active mitochondria (excitation – green, 5 µm. (Images a and b provided by emission – red), (Hiramoto et al., 1958; J. Dobrucki and P. Rybak, Jagiellonian Bernas and Dobrucki, 2002). (b) A DNA University, Kraków.) precursor analog, ethylenedeoxyuridine Figure 3.2b shows two labeling methods that differ to some degree in their speci- ficity. Incorporating precursors ethylenedeoxyuridine (EdU) into nascent DNA and subsequently labeling the incorporated molecules (the Click reaction) leads to a very specific labeling of newly synthesized DNA, with very low or even no background staining. Labeling the phosphorylated moieties of histone H2AX with specific anti- bodies is very specific, but some nonspecific binding by a secondary antibody cannot be avoided. Consequently, a low-level fluorescence background is usually present. 3.1.3 Sensitivity of Detection Contemporary fluorescence microscopy offers highly sensitive detection of fluo- rescent species. The advent of highly sensitive light detectors and cameras made it possible to detect even single molecules in the specimen (Figure 3.3). Thus, the observation of single molecule behavior, such as blinking, or the detection of  3.2 A Fluorescence Microscope 103 Figure 3.3 Single molecules of a fluorescent lipid tracer, TopFluor-PC, in lipid bilayers, imaged with an electron-multiplying CCD (EMCCD) camera. Each molecule produces a diffraction-limited signal as discussed in Chapter 2 (field of view, 10 × 10 µm). (Image courtesy of Katharina Scherer, Bonn University.) interactions between individual molecules by Förster resonance energy transfer (FRET) has become possible. The analogy between watching a fluorescent molecule under a microscope and a firefly at night still holds when one thinks of the size and shape of a firefly. A small spot of fluorescent light seen through a microscope may represent one molecule. Although the observer can identify the position of the molecule in space and can watch its movement, the shape and size of this molecule remain unknown (Figure 3.3; see sections below and Chapter 2). 3.2 A Fluorescence Microscope 3.2.1 Principle of Operation The analogy between a firefly and a fluorescent object that was useful in introducing basic concepts ends, when one considers the instrumentation required to detect fluorescence in a microscope. While a firefly emits light on its own by a biochemical process called bioluminescence, which uses energy from ATP, but does not require light to be initiated, the fluorescence in a microscopic object has to be excited by incident light of a shorter wavelength. Thus, a fluorescence microscope has to be constructed in a way that allows excitation of fluorescence, subsequent separation of the relatively weak emission from the strong exciting light, and finally, the detection of the fluorescence. An efficient separation of the exciting from the fluorescence light, which eventually reaches the observer’s eye or the electronic detector, is mandatory for obtaining high image contrast. A sketch of a standard widefield fluorescence microscope is shown in Figure 3.4. 104 3 Fluorescence Microscopy SP HL EXC OC SP OB EXL CL EXF DM EM DC (a) Specimen Excitation filter Eyepiece Emission filter Quartz collector lens Carbon Arc lamp (b) Objective Condenser 0 50 (c) Figure 3.4 Fluorescence microscope – prin- excitation filter; DM, dichroic mirror; EM, ciple of operation. (a) A schematic diagram emission filter; EXC, exciting light incident of an inverted fluorescence microscope (epi- on the specimen. (b) A schematic diagram fluorescence). This type of a microscope of an early fluorescence microscope, with enables studies of live cells maintained in dia-illumination. (c) Fluorescence images of standard growth medium, in tissue culture microtubules (green) and actin fibers (red) vessels, such as Petri dishes. HL, halogen and an image of the same cell in reflected lamp; SP, specimen; OB, objective lens; OC, light (Box 3.2), demonstrating focal contacts ocular (eyepiece); DC, digital camera; EXL, (black). (Images by J. Dobrucki, Jagiellonian exciting light source; CL, collector lens; EXF, University, Kraków.)  3.2 A Fluorescence Microscope 105 Box 3.2 Reflected Light Imaging The epifluorescence design makes it possible to detect reflected light. In this mode of observation, a dichroic mirror has to be replaced with a silver spattered mirror that reflects and transmits the incident light in desired proportions; also the emission filter has to be selected to allow the reflected exciting light to enter the detector (additional optical components are needed to minimize the effects of light interference). Imaging reflected light adds another dimension to fluorescence microscopy. Figure 3.4c shows an example of reflected light imaging – visualization of focal contacts in a fibroblast attached to a glass surface. The images were collected in a confocal fluorescence microscope that was set up for two-channel fluorescence and reflected light imaging. Fluorescence is excited by light that is emitted by a mercury lamp. The exciting light is reflected toward the sample by a dichroic (two-color) mirror (Figure 3.4 and Figure 3.5). This special mirror is positioned at 45◦ angle toward the incoming light and reflects photons of a selected wavelength but allows light of longer wavelengths Cell Cover glass Excitation light source Dichroic mirror Excitation filter Emission Fluorescence filter emission (a) Detector (b) Figure 3.5 A fluorescence microscope fil- microscope filter block, which reflects green ter block. (a) A schematic diagram of a but transmits red light. The excitation filter typical filter block containing an excita- is facing the observer, the dichroic mirror is tion filter, a dichroic mirror, and an emis- mounted inside the block, and the emission sion filter. The exciting light emitted by filter (on the left side of the block) cannot a light source is reflected by the dichroic be seen in this shot. Light beams emitted by mirror toward the specimen; the fluores- laser pointers were used; in a fluorescence cence emission has a longer wavelength microscope, the intensity of the red emission than the exciting light and is transmitted would be significantly lower than the inten- by a dichroic mirror toward the ocular or sity of the exciting green light. (Photograph a light detector. (b) A photograph of a by J. Dobrucki). 106 3 Fluorescence Microscopy Excitation Emission wavelength detection band 100 Excitation spectrum 80 Spectral intensity (AU) Emission spectrum 60 40 20 0 350 400 450 500 550 600 650 700 750 (a) Wavelength (nm) Excitation band Transmission (reflected by of dichroic dichroic mirror) mirror 100 80 Transmission (%) 60 40 Emission band detection 20 0 350 400 450 500 550 600 650 700 750 (b) Wavelength (nm) Figure 3.6 Spectral properties of a typical fluorescent label, fluorescein (a), and the char- acteristics of a filter set suitable for selecting the excitation band and detecting fluorescence emission (b). to pass through. The dichroic mirror is selected for a given application, that is, it is made to reflect the chosen exciting wavelength and allow the passage of the expected fluorescence (Figure 3.6). It is important to realize that the efficiency of converting exciting light into fluorescence is usually low, that is, only one out of many exciting photons is converted into a photon of a longer wavelength and subsequently detected as fluorescence. Moreover, the fluorescence is emitted by the sample in all directions, but only a selected light cone, that is, a fraction of this fluorescence, is collected by the objective lens. Consequently, fluorescence is weak in comparison with the exciting light and has to be efficiently separated and detected.  3.2 A Fluorescence Microscope 107 High-quality optical filters are used to select exclusively the desired exciting wavelength (excitation filter) and the fluorescence emission bands (emission filter) (Figure 3.5). It is worth noting that the arrangement described here, called epifluorescence (which resembles the reflected light microscope that is used in studies of metal surfaces), makes it relatively easy to separate the fluorescence from the exciting light. It is also safer for the operator than an original dia-illumination system. In the early days of fluorescence microscopy, a direct (dia-)illumination of the sample was used (Figure 3.4b). The exciting light was prevented from reaching the observer only by virtue of placing an efficient emission filter in the light path. In the original fluorescence microscope design, removing the excitation filter from the light path allowed the intense exciting light into the eyepiece and the eyes of the observer. Even in the case where the objective lens does not allow UV light to pass through, this would be very dangerous. In the epifluorescence design, there is no danger of sending the exciting light directly into the eyes of the observer. Even when the dichroic and excitation filters are removed, the exciting light will not be incident directly onto the ocular. This does not mean, however, that it is safe to look through the microscope with the exciting light source turned on, and all the filter blocks removed out of the light path. When filters are removed, the exciting light is still sufficiently reflected and scattered to pose a hazard to the eyes of the observer. To protect one’s lab mates, if filter blocks are removed, it is advisable to leave an appropriate note to warn others who might come to ‘‘only have a brief look at their sample’’. 3.2.2 Sources of Exciting Light A typical fluorescence microscope contains two sources of light. One, usually a halogen lamp, is used for initial viewing of a specimen in a transmitted light mode. Another, often a mercury arc lamp, is used for exciting fluorescence. Halogen lamps belong to a class of incandescent (or ‘‘candescent’’) light sources, where the emission of photons occurs through the heating of a tungsten filament. High-pressure mercury vapor arc-discharge lamps (HBO) are a popular source of exciting light in standard fluorescence microscopes. The intensity of light emitted by a mercury arc lamp is up to 100 times greater than that of a halogen lamp. The spectrum of emission, shown in Figure 3.7a, extends from the UV to the infrared. The spectrum depends to some degree on the pressure of the mercury vapor and on the type of lamp. The desired excitation spectral band is selected by using an appropriate excitation filter. As the spectrum consists of numerous sharp maxima, the intensity of the exciting light strongly depends on the selected wavelength. Mercury arc lamps are very convenient sources of excitation for a number of typical fluorophores. Another popular source of exciting light is a xenon arc lamp. It emits an almost continuous spectrum of emission in the whole range of visible wavelengths (Figure 3.7b). Metal halide lamps are a modification of mercury vapor lamps, characterized by higher levels of emission between the major mercury arc 108 3 Fluorescence Microscopy MERCURY ARC XENON 0.10 365 436 465 0.15 827 885 Spectrum intensity 546 Spectral intensity 0.08 312-313 0.1 0.06 919 579 980 297-302 334 992 0.04 254 0.05 0.02 475 0 0 300 400 500 600 700 800 300 400 500 600 700 800 900 10001100 (a) Wavelength (nm) (b) Wavelength (nm) METAL HALIDE 0.10 0.08 Spectrum intensity 546 436 365 405 0.06 579 0.04 0.02 0 300 400 500 600 700 800 (c) Wavelength (nm) Figure 3.7 Schematic diagrams of the emission spectra of a (a) mercury arc, (b) xenon, and (c) metal halide lamp. (Source: Figure by M. Davidson, Microscopy Primer, University of Florida, redrawn.) spectral lines (Figure 3.7c) and by a much longer lifetime. Here, the spectrum is dependent on the metal used for doping. The stream of photons emitted by the filament in the bulb is not ideally uniform in space or constant in time. It has been demonstrated that a hot light source emits groups (bunches) of photons, rather than individually independent photons. This intriguing phenomenon (photon bunching) was observed in a classical physics experiment performed by Hanbury Brown and Twiss (for a discussion of photon bunching and the relation of this phenomenon to the corpuscular and wave nature of light, see (Hanbury Brown and Twiss, 1957). There are also more trivial sources of spatial and temporal inhomogeneity of emitted light, including fluctuations in the temperature of the filament, instability of the electric current, and the influence of external electromagnetic fields. Consequently, the popular mercury arc lamps pose some problems for quantitative microscopy. The illumination of the field of view is not uniform, the intensity of light fluctuates on a short time scale that is comparable with the times needed to record pixels and images, and it diminishes over days of the lamp use, because the electrodes are subject to erosion. Thus, quantitative fluorescence microscopy studies are hampered by the lack of short- and long-term stability of the exciting light. A higher stability of emission is offered by light-emitting diodes (LEDs) and laser light sources. A typical HBO burner needs to be placed exactly at the focal point of the collector lens (when exchanging HBO burners, the quartz bulb must not be touched by  3.2 A Fluorescence Microscope 109 fingers). Centering is important to ensure optimal illumination, that is, the highest attainable and symmetric illumination of the field of view. After being turned off, HBO burners should be allowed to cool before switching them on again. They should be exchanged upon reaching the lifetime specified by the producer (usually 200 h). Using them beyond this time not only leads to a significantly less intensive illumination but also increases the risk of a bulb explosion. Although relatively rare, such an event may be costly, as the quartz collector lens located in front of the burner may be shattered. HBO burners need to be properly disposed of to avoid contamination of the environment with mercury. Today’s metal halide lamps are usually precentered and fixed in the optimal position in the lamp housing. Lasers emit light of discrete wavelengths, characterized by high stability both spatially (the direction in which the beam is propagated is fixed; so-called beam- pointing stability), and temporally – both on a short and long timescale. In the early days of fluorescence microscopy, lasers did not exist; subsequently, their use was limited to laser scanning microscopes, which were more advanced and more expensive than standard widefield fluorescence microscopes (Chapter 5). The stability of the light beams and the ability to focus them to a diffraction- limited spot was an important advantage of lasers over mercury arc lamps. The disadvantage, however, was their high cost combined with a limited number of usable emission lines for exciting popular fluorophores. For instance, the popular 25 mW argon ion laser provided the 488 and 514 nm lines, which were not optimal for working with the most popular (at that time) pair of fluorescent dyes – fluorescein and rhodamine. Other gas lasers that are often used in confocal fluorescence microscopy include krypton–argon (488, 568, and 647 nm) and helium–cadmium (442 nm). Solid-state lasers, including a frequency doubled neodymium-doped yttrium aluminum garnet (Nd:YAG, 532 nm), are also used. A comprehensive description of lasers used in fluorescence confocal microscopy can be found in Gratton and vandeVen (2006). Currently, the price of standard gas lasers has decreased, and a wide selection of various single line lasers and lasers emitting a broad wavelength spectrum are available, including a ‘‘white light laser.’’ Durable diode lasers became available as well. Moreover, new low-cost sources of light, LEDs, have become available. LEDs are semiconductors, which exploit the phenomenon of electroluminescence. Originally used only as red laser pointers, they now include a range of devices emitting in ultraviolet, visible, or near-infrared spectral region. The critical part of an LED is the junction between two different semiconducting materials. One of them is dominated by negative charge (n-type) and the other by positive charge (p-type). When voltage is applied across the junction, a flow of negative and positive charges is induced. The charges combine in the junction region. This process leads to a release of photons. The energy (wavelength) of these photons depends on the types of semiconducting materials used in the LED. A set of properly selected LEDs can now be used as a source of stable exciting light in a fluorescence microscope. LEDs are extremely durable, stable, and are not damaged by being switched on and off quickly. These advantages will probably make them very popular not only in scanning confocal microscopes but in standard fluorescence microscopes as well. 110 3 Fluorescence Microscopy 3.2.3 Optical Filters in a Fluorescence Microscope In a fluorescence microscope, light beams of various wavelengths need to be selected and separated from others by means of optical glass filters. The filter that selects the desired wavelength from the spectrum of the source of exciting light is called an excitation filter. There is also a need to control the intensity of exciting light. This can be achieved by placing a neutral density filter in the light path, if it is not possible to regulate the emission intensity directly. Fluorescence is emitted in all directions; most of the exciting light passes straight through the specimen, but a large part is scattered and reflected by cells and subcellular structures. A majority of the reflected exciting light is directed by the dichroic mirror back to the light source, while a selected wavelengths range of the emitted fluorescence passes through toward the ocular or the light detector. In a standard widefield fluorescence microscope, the emission filter is mounted on a filter block which is placed in the light path to select the desired emission bandwidth. If only one set of spectral excitation and emission wavelengths were available in a microscope, the applicability of the instrument would be seriously limited, as only one group of spectrally similar fluorescent probes could be used. Therefore, a set of several filter blocks prepared for typical fluorescent dyes is usually mounted on a slider or a filter wheel. This allows for a rapid change of excitation and emission ranges. In fluorescence confocal microscopes, the dichroic filter is often designed so as to reflect two or three exciting wavelengths and transmit the corresponding emission bands. Additional dichroic mirrors split the emitted light into separate beams directed toward independent light detectors. This makes it possible to simultaneously excite and observe several fluorophores. Some widefield fluorescence microscopes incorporate electronically controlled excitation and emission filter wheels, which contain sets of several optical filters. The desired combination of filters can be quickly selected using a software package that drives the filter wheels and the shutters. Although microscope manufacturers offer standard filter blocks for most popular fluorescent probes, it is useful to discuss and order custom-made filter sets optimized for the source of exciting light and for the user’s specific applications. It is also prudent to buy an empty filter block (holder) when purchasing a new microscope in order to make it possible to build one’s own filter blocks when new applications are desired. ‘‘In-house’’ assembling one’s own filter block may require some skill and patience, because the position of the dichroic mirror in relation to the beam of exciting light is critical. Microscope manufacturers provide filter blocks with dichroic filters fixed in the optimal position. Aligning the dichroic mirror by the user should be possible but may be cumbersome in some microscopes. When buying individual filters, especially dichroics, it is important to choose high-quality filters, that is, low-wedge filters (flat, with both surfaces ideally parallel), with a low number of coating imperfections. Defects in coating will allow the exciting light to reach the fluorescence detector. The exciting light reaching the detector will be detected as a background, thus reducing the contrast and degrading image quality.  3.2 A Fluorescence Microscope 111 Also, when buying dichroics in order to build one’s own filter combinations, it is important to ensure that not only the shape and diameter but also the thickness of the filter is right for a given microscope design. Aligning a dichroic of a different thickness may turn out to be impossible. Interference filters will have an arrowhead on the edge pointing in a direction from which light should enter the filter. Optical filters have to be handled with care to avoid leaving fingerprints, grease, and dust on the surface. Newer hard-coated filters are quite robust and easier to clean. If cleaning is required, it can be done by blowing air or using a soft brush (to remove the dust) and subsequently using a piece of optical tissue (the tissue is not to be reused) with a small amount of pure methanol or isopropyl alcohol. The manufacturer of the filter will usually recommend the solvent that is appropriate for their products. It is important to ensure that the antireflection coating is not scratched or damaged. The state of current technology in the manufacturing glass optical filters is so advanced that filters of various optical characteristics can be custom-made by the optical companies. 3.2.4 Electronic Filters A new class of versatile optoelectronic elements that serve as filters has become available in recent years. These include the acousto-optic tunable filters (AOTFs). They provide flexibility and speed in choosing light wavelength and intensity that cannot be achieved with glass optical filters. This flexibility and speed is indispensable in modern, advanced fluorescence confocal microscopes. AOTFs work in a manner similar (though not identical) to diffraction grating. A specialized birefringent crystal is subjected to high-frequency (a few hun- dred megahertz) acoustic waves that induce a periodic pattern of compression (Figure 3.8). This, in turn, changes the local diffractive index of the crystal and imposes a periodic pattern of different refractive indices. Most AOTFs use tel- lurium dioxide (TeO2 ). This material is transparent in the range of 450–4000 nm. Only a selected band from a whole range of light wavelengths incident on the crystal is deflected. In this respect, the crystal resembles a band-pass filter rather than a diffraction grating where a whole range of wavelengths are diffracted (at different angles). The direction in which the light beam is deflected is fixed and does not depend on the wavelength. The wavelength of the diffracted band depends on the frequency of the acoustic wave. The intensity of the deflected light can be controlled by the amplitude of the acousto-mechanical wave incident on the crystal. The wave is impressed onto the crystal by a piezoelectric transducer, that is, a device that expands and shrinks according to the applied voltage. AOTFs can be used to rapidly change the wavelength as well as the intensity of the deflected light by changing the frequency and the amplitude of the incident acoustic waves that drive the crystal. The spectral width of the deflected light can be controlled by delivering multiple frequencies to the AOTF. Moreover, using several widely spaced frequencies will allow a set of wavelength bands to be deflected at the same 112 3 Fluorescence Microscopy Acoustic absorber First-order diffracted rays Input nonpolarized TeO2 crystal Undiffracted zeroth-order rays Piezoelectric Radio (acoustic) frequency First-order transducer source diffracted rays Figure 3.8 Architecture and principle of operation of an acousto-optic tunable filter (AOTF). (Source: Lichtmikroskopie online, Vienna University, modified.) time. These characteristics make AOTF a flexible component of a fluorescence microscope, which can serve simultaneously as a set of fast shutters, neutral density filters, and emission filters, allowing simultaneous detection of several emission colors. Changing the wavelengths and their intensity can be achieved at a high speed, namely, tenths of microseconds, while filter wheels need time on the order of a second. An AOTF can also be used as a dichroic filter, which separates the exciting light from fluorescence emission. These advantages make AOTFs particularly valuable for multicolor confocal microscopy and techniques such as fluorescence recovery after photobleaching (FRAP, Axelrod et al., 1976). 3.2.5 Photodetectors for Fluorescence Microscopy In most microscopy applications, there is a need to record fluorescence im- ages – often a large number of images within a short time – for subsequent processing, analysis, and archival storage. Originally, analog recording of fluo- rescence images on a light-sensitive film was extensively used, but in the past two decades, methods of electronic detection, analysis, and storage have become efficient and widely available. Currently, fluorescence microscopes are equipped with suitable systems for light detection, image digitization, and recording. The most common light detector in a standard widefield fluorescence microscope is a charge-coupled device (CCD) camera, while photomultipliers are used in laser scanning confocal microscopes. Other types of light detectors, including intensi- fied charge-coupled device (ICCD) and electron multiplied charge-coupled device (EMCCD) cameras, as well as avalanche photodiodes (APDs), are gaining increasing importance in modern fluorescence microscopy.  3.2 A Fluorescence Microscope 113 3.2.6 CCD – Charge-Coupled Device In a camera based on a CCD, an image is projected onto an array of semiconducting light-sensitive elements that generate an electric charge proportional to the intensity of the incident light (Figure 3.9). Two types of sensor architectures are currently produced: front-illuminated charge-coupled device (FI CCD) cameras, where light is incident on an electrode before reaching the photosensitive silicon layer, and back-illuminated charge-coupled device (BI CCD) cameras, where the incoming light falls directly on the silicone layer (Figure 3.10). Note that the structure of an FI CCD is similar to the anatomy of the human eye. In the retina, it is the nerve ‘‘wiring’’ that faces the incoming light. Only after light has passed the layer of nerve cells that are not sensitive to light, can the photons interact with rhodopsin in the photoreceptors. Light A CCD picture element Metal electrode Silicon dioxide Silicon substrate Depletion region Clocking the charge Figure 3.9 Converting photons into electric charges and their transfer in a CCD array. Front Illuminated Back Illuminated Incoming Incoming Polysilicon light light Thinned gate silicon Silicon dioxide Silicon Silicon Electrode Depletion region (a) (b) Figure 3.10 Architecture of a (a) front- and (b) back-illuminated CCD. 114 3 Fluorescence Microscopy 1 2 3 4 5 6 Image Amplifier Readout Figure 3.11 Charge transfer through a CCD array and into an amplifier. When the array of light-sensitive elements (capacitors, photodiodes; in digital microscopy called picture elements or pixels for brevity) of a camera is exposed to light, each of them accumulates an electric charge (Figure 3.9). These charges are later read one by one. An electronic control system shifts each charge to a neighboring element and the process is repeated until all charges are shifted to an amplifier (Figure 3.11). When the individual charges are dumped onto a charge amplifier, they are converted into corresponding voltage values. This process is repeated until all the light detected by a microscope in all pixels of the field of view is eventually translated, point by point, into voltage values. The values of these voltages are subsequently converted into discrete values (digitized) and translated into the levels of brightness (intensity of light) on the display screen (Figure 3.12). A CCD camera’s sensitivity and its ability to record strong signals depend on a number of factors, including the level of electronic noise and the ability to accumulate a large number of charges. Weak signals may be undetectable if they are comparable with the noise level. Strong signals may also be difficult to record faithfully. Exposing a CCD array to a large dose of light may result in filling the well with charges and reaching a maximum well capacity (saturation charge). This may cause charges spilling into the neighboring wells and result in deterioration of image quality (see below and Section 8.3.3). In an ideal situation, only the photons that strike the silicone surface of a light-sensitive element of the camera should actually cause accumulation of an electric charge, which is later shifted and read out (Figure 3.12). However, some electrons may occur and are recorded even in the absence of any incident light. This contributes to the noise of the displayed image (see below). As fluorescence intensities encountered in microscopy are usually low, the fluorescence signal may be difficult to detect in the presence of substantial camera noise (for a Accumulation Photon Electron of charges hv e in a well Voltage translated Transfering groups Amplification Conversion of into brightness of charges from of the electric electric signal of individual pixels well to well (clocking) signal into voltage on a computer monitor Figure 3.12 A schematic representation of the steps occurring from the absorption of a photon in a CCD array to the display of brightness on a computer screen. In some digital cameras, amplification of the signal is achieved already on the chip (see sections below).  3.2 A Fluorescence Microscope 115 comprehensive discussion of camera noise, see Section 8.3.3). In order to alleviate this problem, photons registered by the neighboring elements of an array can be combined and read as one entity. This procedure, called binning, is illustrated in Figure 3.13. A fluorescent structure in a specimen is symbolized by two black lines. It is overlaid with a fragment of a CCD array. The weak signals recorded in individual small pixels of a 6 × 6 array do not stand out well against the background noise generated by the electronics of the camera. When a 3 × 3 array of larger pixels is used, signals from the areas corresponding to four neighboring small pixels are summed up. Such signals become sufficiently strong to be detected well above the noise floor. The improvement in signal-to-noise ratio is achieved at the cost of spatial resolution. Note that Figure 3.13 is intended to explain the principle and benefit of binning, but it is a necessary oversimplification because it does not consider the dark noise generated in each pixel by thermal noise in the device; this noise will also be summed in the binning procedure. Longer recording times also help isolate weak signals from random noise. Weak fluorescence signals that are below the detection level of a standard CCD camera may be detected by ICCD and EMCCD cameras (see below). The light-sensitive element of a CCD camera cannot detect all incident photons because not all photons that arrive at a given pixel generate an electron. The ratio between the incident photons that generate an electric charge to all photons incident on an area of a detector is referred to as the quantum efficiency of a CCD camera. This efficiency is wavelength dependent. FI CCD cameras reach 60–70% quantum efficiency in the visible range and are essentially unable to detect UV Specimen and Pixels CCD array of the image 6x6 3x3 Figure 3.13 A schematic diagram describing the principle of binning in a CCD sensor. 116 3 Fluorescence Microscopy BI CCD 100 Quantum efficiency (%) FI CCD 80 60 40 20 0 200 300 400 500 600 700 800 900 1000 Wavelength (nm) Figure 3.14 QE and spectral response of front- and back-illuminated CCDs. (Source: Docu- mentation from Andor Company, simplified.) radiation (Figure 3.14). BI CCDs have a quantum efficiency (QE) up to 95% and a better spectral response, including an ability to detect in the UV range. The difference is primarily because the BI CCD exposes light-sensitive elements, rather than an electrode structure, as in FI CCD, directly to the incoming light. 3.2.7 Intensified CCD (ICCD) Fluorescence microscopy invariably struggles with weak signals. Image acquisition is particularly difficult when a high speed of data recording is required. A standard CCD camera is often insufficient for the detection of faint fluorescence due to a high readout noise. Moreover, high sensitivity of a standard CCD sensor is achieved only at relatively low image acquisition rates. Much better sensitivity is offered by an ICCD camera. While a standard CCD-based camera produces only one electron in response to one incoming photon, an ICCD generates thousands of electrons. ICCDs use an amplifying device called a microchannel plate, which is placed in front of a standard CCD light sensor (Figure 3.15). A microchannel plate consists of a photocathode, a set of channels, and the layer of a phosphor. A photon incident on a photocathode generates one electron (by means of the photoelectric effect) that travels inside a channel, toward the phosphor layer and is accelerated by a strong electric field. Inside the channel, the electron hits the channel walls and generates more electrons that eventually reach the phosphor. There, each electron causes the emission of a photon that is subsequently registered by the sensor of a standard CCD camera. In this way, the microchannel plate amplifies the original weak light signal, and this amplified signal is eventually detected by a standard CCD sensor. In ICCDs, the signal-to-noise ratio can be improved by a factor of several thousand in comparison with a standard CCD camera. ICCD cameras feature high sensitivity that is required for the imaging of weak fluorescence signals (although the sensitivity of the primary photocathode is relatively low, not exceeding 50% quantum efficiency). However, the spatial  3.2 A Fluorescence Microscope 117 Photons Photon Photoelectric effect Secondary Photocathode electrons Electrons Electric field Phosphor Photons Emission of photons Photoelectric effect CCD array Readout Electrons Figure 3.15 Signal amplification in an intensified CCD sensor. resolution is low and background noise is usually high. Moreover, an ICCD chip can be damaged by excess light. When a high data registration speed is needed and the signal is weak, an EMCCD camera is a viable option. 3.2.8 Electron-Multiplying Charge-Coupled Device (EMCCD) Another option available for detection of weak fluorescence signals is an EMCCD camera. This device is characterized by a very low read noise and offers sufficient sensitivity to detect single photons. It features high speed, high quantum efficiency, and high digital resolution. In this type of camera, amplification of the charge signal occurs before the charge amplification is performed by the electron-multiplying structure built into the chip (hence called on-chip multiplication). An EMCCD camera is based on the so-called frame-transfer CCD (Figure 3.16; this technology is also used in some conventional CCDs) and includes a serial (readout) register and a multiplication register. The frame-transfer architecture is based on two sensor areas – the area, which captures the image, and the storage area, where the image is stored before it is read out. The storage area is covered with an opaque mask. The image captured by the sensor area is shifted to the storage area after a predefined image integration time, and subsequently read out. At the same time, the next image is acquired. The charge is shifted out through the readout register and through the multiplication register where amplification occurs prior to readout by the charge amplifier. The readout register is a type of a standard CCD serial register. Charges are subsequently shifted to the multiplication register, 118 3 Fluorescence Microscopy Image section Storage section On-chip charge to voltage conversion Output Readout Multiplication register register Voltage Secondary electrons Figure 3.16 Principle of operation and signal amplification in an EMCCD sensor. that is, an area where electrons are shifted from one to the next element by applying a voltage that is higher than typically used in a standard CCD serial register. At an electric field of such a high value, the so-called ‘‘secondary electrons’’ are generated. The physical process that is activated by this voltage is called impact ionization. More and more electrons enter subsequent elements of the multiplication register, and a ‘‘snowball effect’’ ensues: the higher the voltage (for an operator – the gain value) and the larger the number of pixels in the multiplication register, the higher the overall multiplication of the original faint fluorescence signal (Figure 3.16). The process of multiplication is very fast, so an EMCCD camera is not only very sensitive but also very fast while preserving the high spatial resolution of a standard CCD chip. The probability of generating a secondary electron depends on the voltage of the multiplication clock (Figure 3.17a) and the temperature of the sensor (Figure 3.17b). Although the efficiency of generating a secondary electron is quite low (approximately 0.01 per shift), this process can occur in all elements of a long multiplication register, bringing the final multiplication value to several hundred or more. But this amplification is not for free. As the generation of the electrons is  3.2 A Fluorescence Microscope 119 On-chip multiplication gain On-chip multiplication gain (normalized units) (normalized units) 100 100 0 0 0 1000 2000 3000 4000 −40 −20 0 20 (a) Voltage (au) (b) Temperature (°C) Figure 3.17 On-chip multiplication gain versus voltage (a) and temperature (b) in an EM- CCD camera. (Source: Roper Scientific Technical Note #14.) a stochastic process in each pixel, the extra gain increases the noise in the signal. This is accounted for by the so-called noise excess factor, F (see also Section 8.3.3). The probability of generating a secondary electron decreases with temperature, as shown in Figure 3.17. Thus, cooling a chip gives an additional advantage in terms of a higher on-chip multiplication gain. Cooling a chip from room temperature to −20 ◦ C increases the chance of generating a secondary electron by a factor of 2. Some EMCCD cameras are cooled to −100 ◦ C. The lower temperature also results in a lower dark current (see below). Thus, even a very weak signal becomes detectable because it stands out above the noise floor. Reducing the noise level at low temperatures is particularly important because all signals that occur in a well or a serial register, and do not represent the fluorescence signal, will be multiplied with the signal of interest. 3.2.9 CMOS An alternative to a CCD is a complementary-metal-oxide semiconductor (CMOS) image sensor. Note that the term CMOS actually refers to the technology of manufacturing transistors on a silicone wafer, not the method of image capture. Like the CCD, CMOS exploits the photoelectric effect. Photons interacting with a silicon semiconductor move electrons from the valence band into the conduction band. The electrons are collected in a potential well and are converted into a voltage that is different from that in a CCD sensor, where the charge is first moved into a register and subsequently converted into voltage. The measured voltage is then passed through an analog-to-digital converter and translated into a brightness value for an individual pixel on the computer monitor. CMOS sensors are manufactured in a process where the digital logic circuits, clock drivers, counters, and analog- to-digital converters are placed on the same silicon foundation and at the same time as the photodiode array. In this respect, the architecture of a CMOS sensor is distinctly different from that of a CCD device, where the charge of each photodiode is transferred first to the chip, and then read out in sequence outside of the chip. 120 3 Fluorescence Microscopy Microlens Red color filter Transistor Silicon substrate Photodiode n+ Potential well Figure 3.18 Architecture of a single CMOS photodiode. (Source: A figure by M. Davidson, Microscopy Primer, University of Florida, modified.) The construction of a typical CMOS photodiode is presented in Figure 3.18. The actual light-sensitive element and the readout amplifier are combined into one entity. The charge accumulated by the photodiode is converted into an amplified voltage inside the pixel and subsequently transferred individually into the analog signal-processing portion of the chip. A large part of the array consists of electronic components, which do not collect light and are thus not involved in detecting the light incident on the sensor. The architecture of the array leaves only a part of the sensor available for light collection and imposes a limit on the light sensitivity of the device. This shortcoming is minimized by placing an array of microlenses over the sensor that focus the incident light onto each photodiode. The unique architecture of a CMOS image sensor makes it possible to read individual pixel data throughout the entire photodiode array. Thus, only a selected area of the sensor can be used to build an image (window-of-interest readout). This capability makes CMOS sensors attractive for many microscopy applications. CMOS technology has been refined in recent years so that today’s sensors compete successfully even with high-end EMCCD cameras in many low light microscopy applications. At the time of writing, 2048 × 2048 pixel CMOS sensors using 6.5 × 6.5 µm pixels with a quantum efficiency exceeding 70% and a 100 frames per second readout rate are available. 3.2.10 Scientific CMOS (sCMOS) CMOS-based digital cameras were originally inferior to high-end CCD cameras. Until recently, EMCCD cameras were the best choice in terms of sensitivity, speed of data acquisition, and digital resolution. Yet the CMOS technology has made  3.2 A Fluorescence Microscope 121 significant advances and the newest design, called sCMOS, is challenging even the EMCCD sensors in many demanding microscopy applications. The advantages of the sCMOS sensor include a large size array, small pixel size, low read noise, high frame rate, high dynamic range, and no multiplicative noise. 3.2.11 Features of CCD and CMOS Cameras CMOS and CCD cameras are inherently monochromatic devices, responding only to the total number of electrons accumulated in the photodiodes, and not to the color of light, which gives rise to their release from the silicon substrate. In fluorescence microscopy, detection of two or more colors is often required. Sets of emission optical filters are then used to collect images in selected spectral bands sequentially. When simultaneous measurements are necessary, two digital cameras are mounted on a microscope stand. Although the same type of CCD chip is used by many producers of digital cameras, the ultimate noise level, speed of data acquisition, and dynamic range of a given camera may be quite different. These differences arise from the differences in electronics and software driving the camera. The user should use the features and software options of the camera to identify the sources of noise, to calibrate the camera, and to optimize image collection for the specific type of experiment. 3.2.12 Choosing a Digital Camera for Fluorescence Microscopy It might seem obvious that a researcher who is planning to purchase a new fluorescence microscopy system should buy, funds permitting, the best digital camera on the market. However, there is no ‘‘best digital camera’’ for fluorescence microscopy as such. The camera should be selected for a given application. Manufacturers provide important information about the chip and the software, including the pixel size, quantum efficiency, full well capacity, the size of the various noise contributions, speed of image acquisition and the corresponding achievable resolution, tools for calibration and noise removal, etc. Careful analysis of technical parameters of various cameras available on the market is essential. However, nothing can be substituted for testing various cameras with a typical specimen, which is the researcher’s prime object of investigation. 3.2.13 Photomultiplier Tube (PMT) While the CCD and CMOS cameras briefly introduced above are used for recording the whole image of a field of view essentially at the same time in a parallel process, a photomultiplier is used as a point detector. In other words, it records the intensity of light only in one selected point of the image at a time. Thus, photomultiplier 122 3 Fluorescence Microscopy Photocathode Secondary high voltage (−) electrons Anode 500–2000 V Incident To current- to-voltage photon amplifier Focusing Dynode electrode Figure 3.19 Schematics of a photomultiplier. tubes (PMTs) are not used in standard widefield fluorescence microscopes, but serve as light detectors in laser scanning confocal microscopes. A PMT is a signal-amplifying device that exploits the photoelectric effect and a secondary emission phenomenon, that is, the ability of electrons to cause emission of other (secondary) electrons from an electrode in a vacuum tube. Light enters a PMT through a quartz (or glass) window and strikes a photosensitive surface (a photocathode) made of alkali metals (Figure 3.19). The photocathode releases electrons that subsequently strike the electrode (dynode), which releases a still larger number of electrons. These electrons hit the next dynode. A high voltage (1–2 kV) is applied between subsequent dynodes. The electrons are accelerated and the process is repeated leading to an ampli- fication of the first electric current generated on the photocathode. The current measured at the last dynode is proportional to the intensity of light that was incident on the photocathode. The gain obtained by amplifying the electric current through subsequent dynodes can be as high as 107 –108. However, the voltage, which is applied to the dynodes, causes a low-level electron flow through the PMT even in the absence of light. This is translated into a non-zero-level reading on a fluorescence image. The quantum efficiency of a PMT does not exceed 30%. PMTs are very fast detectors of UV and visible light. They can be used to detect single photons and follow extremely fast processes, as the response time can be as low as several nanoseconds. Typically, in laser scanning confocal microscopes, sets of two to five PMTs are used as fluorescence detectors in selected wavelength bands. 3.2.14 Avalanche Photodiode (APD) An APD is also a signal-amplifying device that exploits the inner photoelectric effect. A photodiode is essentially a semiconductor p–n (or p–i–n) junction (Figure 3.20). When a photon of sufficient energy strikes the diode, it excites an electron, thereby  3.3 Types of Noise in a Digital Microscopy Image 123 n-Contact (cathode) Incident p-Contact Holes (anode) photons Electrons SiO2 p layer n n-Layer p-Layer Figure 3.20 Schematics of an avalanche photodiode. (Source: A figure by M. Davidson, Microscopy Primer, University of Florida, modified.) creating a free electron (and a positively charged electron hole). This, in turn, creates a flow of electrons (‘‘an avalanche’’) between the anode and the cathode, and electron holes between the cathode and the anode, because a high voltage is applied between the anode and the cathode. Electrons accelerated in the electric field collide with atoms in the crystalline silicone and induce more electron–hole pairs. This phenomenon amounts to a multiplication effect. In this respect, an APD bears similarity to a PMT. The quantum yield of an APD can reach 90%, that is, it is substantially higher than that of a PMT and the response time is several times shorter. However, the gain is lower, in the range of 500–1000. Single-photon avalanche photodiodes (SPADs) are currently used as detectors in fluorescence lifetime imaging microscopy (FLIM). 3.3 Types of Noise in a Digital Microscopy Image If a light detector were ideal, an image collected in the absence of any specimen should be completely black. However, even in the absence of any fluorescence in the sample, CCD sensors still generate certain readout values that are greater than zero. In laboratory vocabulary, these weak unwanted signals that do not represent fluorescence are generally called background or noise. The adverse influence of noise on the quality of the recorded image is under- standable. Let us assume that the noise signals have a value in the range between 1 and 10 on a scale of 1–100. If a signal representing fluorescence has an intensity of 80, it will be readily detected, but if a weak signal has an intensity of 10, it will not be distinguishable from the noise (Figure 3.21). Averaging a large number of image frames should make the signal detectable over the noise level, but an experimenter rarely has the luxury of collecting many images of the same field of view because photobleaching will inevitably diminish the fluorescence signal, while the noise level will remain the same. The range of intensities of fluorescence signals that 124 3 Fluorescence Microscopy Incident photon CCD Array Signal Electronic noise Output signal Fluorescence lost in noise (a) Signal Noise averaged averaged Output signal n - times n - times Fluorescence detectable above noise (b) Figure 3.21 Signal averaging and detection stable fluorescence signals are averaged, the of weak fluorescence signals. (a) A weak noise is averaged out and becomes relatively signal cannot be detected if it is compara- low in comparison with the signal. This sim- ble with the level of noise generated by the plified scheme does not take dark noise into electronics of a camera. (b) When images of account. can eventually be recorded above the level of noise is called the dynamic range of the detector. More precisely, dynamic range is the ratio between the maximum and the minimum level of signal that can be detected. Thus, the dynamic range of a CCD camera is equal to the saturation charge (full well capacity) divided by the readout noise (i.e., the noise generated in the absence of light), when both are expressed as the number of electrons. A higher dynamic range of a camera means a broader range of fluorescence intensities that can be faithfully recorded (Figure 3.22). It should be noted, however, that the dynamic range of a light detector defined in that way does not provide any information about its absolute sensitivity. A weak fluorescence signal can be ‘‘fished out’’ of the noise by increasing the integration time (Figure 3.23). This simple procedure will only be useful if the rate of photobleaching does not offset the benefit of integration. Integration takes considerable time. Therefore, although weak signals can eventually be recorded by a CCD camera, the process may be relatively slow. Another way to detect a weak signal is to use an ICCD or an EMCCD. These devices use two different ways of amplification of the signal before it is actually read out, that is, before an unavoidable addition of the read noise takes place. In order to fully appreciate different strategies that were used by digital camera developers aiming at enabling detection of weak signals, a brief discussion of various types of noise is required.  3.3 Types of Noise in a Digital Microscopy Image 125 Fluorescence signal Recorded signal intensity (number of electrons) Fluorescence signal Noise Fluorescence signal Fluorescence Noise signal Noise Noise Signal Noise A B C D Figure 3.22 A schematic representation of Sensors A, B, and C have similar dynamic dynamic ranges and sensitivities of four hy- ranges, that is, the ratio between the max- pothetical digital cameras. The input signals imum recordable fluorescence signal and received by cameras A, B, C, and D are such the level of noise is similar, but the ability as to fill the charge wells; thus they are dif- to detect the strong signals differs – it is the ferent in each case. The recorded signals best for camera C. Camera A is more sen- consist of a noise contribution and fluo- sitive than camera B or C. Camera D has a rescence photons, as shown schematically very low dynamic range but it has a very low by the bars. The electrons resulting from noise level; thus, it is the most sensitive of noise and fluorescence that fill the wells to the four. Camera D is not suitable for de- a maximum capacity are shown symboli- tecting strong signals. Different levels of the cally below the bars. Cameras A, B, C, and D maximum recordable signal of these cameras generate different levels of noise; therefore, are a consequence of different well depths of their ability to detect weak signals differs. their sensors. Some sources of noise were mentioned when speaking about the principles behind and construction of various camera types. Generally, one can identify three major sources of noise in a digital microscopy image registered by a camera: (i) dark current noise, (ii) photon noise, and (iii) read noise. The dark current (dark noise) arises from electrons that are generated in a well of a semiconductor sensor in the absence of any external light due to electron emission by thermal motion. When the integration time on the CCD chip is in- creased, the accumulating thermal charge also increases. This leads to a detectable background in the image. As the electron emission is dependent on temperature, cooling the camera chip efficiently reduces the dark current. The dark current can be decreased from the final image by subtraction. Dark current noise should not be 126 3 Fluorescence Microscopy Incident photon CCD Array Signal Fluorescence lost in noise Signal generated Noise Output signal by fluorescence added by incident on a CCD electronics (a) array Incident photons Integration ×n Signal Fluorescence detectable above noise Integrated Noise Output signal added by signal (b) electronics Figure 3.23 Integrating elevates weak sig- after signal integration remains constant. nals above the noise level. Integration, which This simplified diagram does not take dark is the summing up of incident fluorescence noise into account. (a) Signal collected as photons, improves signal-to-noise ratio be- one frame, that is, without integration and cause the level of electronic noise added (b) integration of signals. confused with background signal arising from low-level autofluorescence or fluores- cence arising from nonspecific binding of an antibody in an immunofluorescence preparation. The photon noise or shot noise results from the quantum nature of light. It is a term that refers to the temporal distribution of photons arriving at the surface of a sensor. Even if the fluorescence-emitting object is flat and uniformly stained, the frequency of photons arriving at the light-sensitive element of the sensor is governed by chance. The frequency of photon arrival follows the so-called Poisson statistics. This implies that the number of photons originating in a given small volume of a continuously illuminated specimen and subsequently reaching the detector varies in different time intervals (we ignore photobleaching for simplicity). This also implies that when the same (nonbleaching) voxel in the specimen is imaged repeatedly, and in a given measurement the number of detected photons  3.4 Quantitative Fluorescence Microscopy 127 is n, the subsequent determinations of the number of photons vary within a range √ of n. Shot noise can be quite misleading. Inexperienced microscopists often take the grainy structure of an area in the image for a real variation of the fluorescence signal, or interpret differences between local signal intensities as evidence for a difference in the concentration of a fluorescence label. However, such features of the image may merely be a consequence of a very low number of photons that are typically collected in fluorescence microscopy studies. The read noise arises in the process of converting a charge generated in the sensor well into voltage and digitization. Camera manufacturers provide information about the noise generated ‘‘on-chip’’ by specifying a root-mean-square number of electrons per pixel (RMS per pixel). For instance, 10 e− RMS means that a read noise level of 10 electrons per pixel is expected. Thus, the signal obtained after readout of the charges would show a standard deviation of 10 electrons, even if all pixels contained identical numbers of electrons. At low signal levels, photon noise is the most significant noise contribution. As vendors may interface the same type of a light-sensitive chip with different electronics, the levels of electronic noise in cameras from different sources may also be different. In addition to these major sources of noise mentioned above, other factors may result in unpredictable signal variability or nonzero levels. These factors include the nonuniformity of photoresponse, that is, noise values dependent on the location in the image sensor, and the nonuniformity of dark current. Nonuniformity of photoresponse is a consequence of the fact that individual pixels in a CCD chip do not convert photons to electrons with identical efficiency. Pixel-to-pixel variation is usually low in scientific grade cameras and generally does not exceed 2%. The correction for differences between light sensitivity of individual pixels can be achieved by recording an image of an ideally uniform fluorescent object, for instance, a dye solution, and creating a correction mask using standard image processing tools. Dark current nonuniformity results from the fact that each pixel generates a dark current at a slightly different rate. The dark current may also drift over a longer period, for instance, owing to a change in the sensor’s temperature. In summary, different types of noise contribute differently to the final image. On-chip multiplication represented by ICCD, CMOS, and EMCCD digital cameras has opened new avenues of research by making it possible to detect very weak signals quickly and with a high spatial resolution. 3.4 Quantitative Fluorescence Microscopy 3.4.1 Measurements of Fluorescence Intensity and Concentration of the Labeled Target A fluorescence microscopy image is usually treated as a source of information about structure, that is, about the spatial distribution of a molecule of interest. Let us use an immunofluorescently stained image of microtubules in a fibroblast 128 3 Fluorescence Microscopy as an example. On the basis of such an image, the observer can establish the presence or absence of tubulin in a given location of a cell. The local intensity of the fluorescence signal is not used here as a source of information about the local concentration of tubulin. Such information is not available because the strength of the fluorescence signal depends on the location of the microtubule in relation to the plane of focus. Using the fluorescence signal as a source of information about the local concentration of tubulin is also impossible because immunofluorescence staining may be nonuniform, as the antibody may be sterically hindered by other proteins from accessing the microtubule. Thus, in this example, the only information required and expected is the architecture of a network of microtubules. The interpretation of the image is based on the assumption that all microtubules stain with an antibody to a degree, which makes them detectable. Often, not just the presence or absence, but also the local concentration of the labeled protein or other molecular target is of interest. Extracting this in- formation requires using a fluorescence microscope as an analytical device. In all honesty, however, one has to admit that a standard widefield fluorescence microscope is not made to be an analytical device capable of straightforward measurements of the quantities of fluorescently labeled molecules within a spec- imen. This task can be performed more adequately by confocal microscopes. A rough estimate of local concentrations can be made in a widefield fluorescence microscope; however, a microscopist should keep the following considerations in mind. Any attempt to estimate relative (within the same field of view) or absolute concentrations (see below) of fluorescent molecules is based on the tacit assumption that the amount of the bound fluorescent probe is proportional to the amount of the molecule of interest. This assumption is rarely true. Let us take DNA- binding fluorescent probes as an example. Among some 50 fluorescent probes that bind DNA, only a few bind DNA in a stoichiometric manner (4 ,6-diamidino- 2-phenylindole (DAPI), Hoechst, propidium). Most DNA dyes also bind RNA, thus RNA has to be hydrolyzed before measuring the local concentration of DNA. Hoechst and DAPI have low affinity for RNA, but propidium is an example of a probe with high affinity for RNA. Nevertheless, propidium is a popular DNA probe used for measuring DNA content in cells by flow and laser scanning cytometry. Measurement of the absolute DNA amounts is also complicated by the fact that propidium, as well as other DNA-affine dyes, compete with proteins for binding to DNA. It has been demonstrated that the amount of propidium bound to DNA is higher in fixed cells, following removal of some of the DNA-associated proteins. This means that assessment of DNA content requires careful calibration within the investigated system, including using the same type of cells the same procedure of RNA removal, etc. When these measures are taken, the assessment of the local concentration of DNA can be quite precise. Simple and reliable determinations of relative amounts of DNA in individual cells can be achieved by staining DNA with DAPI or Hoechst and recording images with a laser scanning cytometer (LSC, Figure 3.24) (Zhao et al., 2011).  3.4 Quantitative Fluorescence Microscopy 129 EdU incorporation (a) (b) DNA content Figure 3.24 Laser scanning cytometer (LSC) (200 µM) for 60 min, showing a reduced determination of DNA content per cell, a