The Automatic Microscope

Motorized microscope components and accessories enable the investigator to automate live-cell image acquisition and are particularly useful for time-lapse experiments that range in timescale intervals from milliseconds to tens or hundreds of minutes. A wide variety of aftermarket auxiliary components, such as electromechanical shutters, microprocessor-controlled filter changers (filter wheels), motorized stages, and axial focus control mechanisms can be retrofitted to a research grade microscope and interactively controlled by a companion workstation computer using commercially available image acquisition software packages. However, it should be noted that assembling a fully automated and optimized multi-dimension optical imaging system is an extremely complex task.

A variety of commercial systems are available (at very high cost), and these are often worth the investment for multi-user core facilities. Alternatively, costs can be reduced by at least 50 percent by assembling a system from scratch, but this effort should be restricted to laboratories that possess sufficient expertise and experience in optical microscopy. The primary problem in automatic microscope configuration is the integration of hardware and software components purchased from different sources into a well-coordinated, efficient system. Newer systems from the major microscope manufacturers are becoming increasingly better equipped with more advanced high speed shutters, axial focus control systems, filter changers, and illumination source and intensity controls, as well as other motorized components that are fully integrated to perform superbly with proprietary software.

A generalized live-cell imaging configuration utilizing an inverted tissue culture microscope is presented in Figure 1. All of the auxiliary components illustrated in the figure are necessary for maintenance of the culture while acquiring images under brightfield, phase contrast, differential interference contrast, and fluorescence illumination. The specimen chamber is positioned on the stage and is usually firmly attached (with the exception of simple microscope slides) over the circular or rectangular stage opening above the objectives. Maintaining the culture chamber environment is a Plexiglas enclosure that encases the microscope condenser system, stage, objectives, and nosepiece. A carbon dioxide sensor and feeder tank valve is controlled by the same unit that monitors the enclosure temperature, while additional units regulate the heating stage and objective heater. The excitation and emission filter wheels and CCD camera system are operated with a separate control unit that is interfaced to the host computer (not illustrated). All of the equipment is mounted on a breadboard-style vibration isolation platform.

Light Shutters

Electromechanical shutters are employed to block the light source(s) from illuminating the specimen between camera exposures and are particularly important when imaging delicate fluorescent specimens as a means of minimizing photobleaching and phototoxicity. These devices are absolutely essential for live-cell imaging over long periods of time in order to ensure and maintain cell viability. In addition, shutters can be utilized to select between multiple light sources and pathways as, for example, when rapidly interchanging between transmission and fluorescence modes to gather two images per time point in time-lapse sequences. Modern high-performance shutters are microprocessor-controlled and capable of operating in several modes, which allow user management over the speed and mechanism underlying shutter function. The fast mode provides the quickest open and close action sequence for which the shutter is capable. A slightly slower soft mode gradually opens and closes the shutter to reduce vibration and produce quieter operation. Control of light intensity without affecting wavelength is accommodated by a neutral density shutter mode that enables precise control of the effective aperture size in the shutter open state, enabling it to act as a neutral density filter. Many shutter controllers offer several operation modes and can interfaced with a workstation to coordinate the action between two or more shutters. The more robust shutters have lifetimes rated in the millions of cycles.

Presented in Figure 2 are several shutter designs that have found widespread use in live-cell imaging microscope configurations. The basic shutter system illustrated in Figure 2(a) is depicted in a cut-away view (Figure 2(b)) showing the internal operating mechanisms. The shutter blades are swung guillotine-fashion into place to close the aperture by the actuator and rocker-drive arm assembly that is dampened through a system composed of bumpers and a shaped spring wire. The shutter armature is electronically linked through the interface connector to an external control unit that can be operated alone or through a host computer. The cut-off section at the base of the shutter is designed to enable attachment in close proximity to the microscope frame without interfering with other parts. The filter wheel shutter illustrated in Figure 2(c) can act as a stand-alone shutter or attached to a motorized filter wheel (described below) for precise simultaneous control of exposure time and wavelength.

Shutters designed for microscopy applications have aperture diameters ranging from 25 to 35 millimeters. The smaller aperture models operate faster, but often limit the field of view and can produce vignetting on some microscopes. Shutters with larger apertures are slower, but reduce or eliminate edge effects on illumination. The housing encapsulating the shutter is usually machined or cast in aluminum and anodized, followed by treatment using a black dye to serve as a heat sink for the actuator coil. Many shutters can be uncased to increase flexibility in mounting, but this exposes the blades and other internal components to contamination from dust and debris. In order to minimize heating of the shutter assembly with hot arc-discharge lamps, shutters designed for microscopy applications should have reflective blades on the input side facing the illumination source. Typical high-end reflective coatings are made with an aluminum-magnesium fluoride alloy deposited on a beryllium-copper substrate, while more economical shutter blades are fabricated from polished stainless steel. The latter are not as highly reflective and do not dissipate heat as well as the alloy blades, but are much lower in cost and more durable.

Wavelength Selection

One of the most important components for automatic multi-color fluorescence imaging is a device for rapidly switching between different wavelengths of light, either through the use of multiple filters, beamsplitting units that direct light through several pathways, monochromators, or acousto-optic tunable filters (AOTFs). In the standard configuration, the conventional fluorescence filter set is housed in an optical block and contains an excitation and barrier (or emission) filter, as well as a dichromatic mirror that directs excitation light to the specimen and transmits emission light to the detector. For live-cell imaging using more advanced filter technology, the dichromatic mirror is retained, but often substituted for a polychromatic derivative that contains multiple bandpass regions. Excitation and emission filters are removed from the optical block and placed in one or more of the external devices described below. Currently, there are several practical mechanisms for automatically interchanging fluorescence filters. The most common involves rotating filter wheels that are reliable, relatively inexpensive, and supported (in effect, driven) by a large number of aftermarket and proprietary image acquisition programs. The primary disadvantage of filter wheels is their limitation in switching speed. Among the major benefits of aftermarket filter wheels are their high light transmission efficiency and the flexibility to use a wide range of commercially available filters.

Filter wheels are designed to hold circular, flat optical interference filters, neutral density filters, heat filters, and ultraviolet filters in a wheel that can accommodate between 4 and 10 filters ranging in diameter from 25 to 50 millimeters (see Figures 3 and 4). The wheel is mounted in an aluminum housing with easy access to the filter slots for changing filters, and is driven by a precise stepper motor controlled by an external unit (illustrated in Figure 4(a)). The housing contains an optical port (into which each filter can successively rotate) for attachment to the microscope and illumination source through the use of specialized adapters, and many wheel housings also accommodate bolt-on shutter assemblies. High-end filter wheels are equipped with mounting ears that enable the housing to be securely fastened to an optical bench or vibration isolation table. Several filter wheels designs have slide-in or drop-in external filter holders for additional filters (such as neutral density) that are placed in series with interference filters mounted into the wheel. The filter wheel control module is coordinated through an internal microprocessor and functions as a standalone unit or can be coupled to a workstation computer for coordinated operation of several filter wheels, shutters, digital cameras, and other peripherals.

A standard economical filter wheel can require up to 100 milliseconds to rotate a new filter into position, while more expensive and faster wheels can do the job in approximately 25 milliseconds. Although these speeds might appear quite rapid at first glance, the switching timescales require at least 4 seconds to record two wavelengths in an axial series of 20 optical sections, thus strictly limiting the speed of image acquisition for rapid cellular processes. In addition, filter wheels are likely to create mechanical vibrations that can couple to the detector system through the microscope frame and lead to degradation of image quality. When attached directly to the microscope body, filter wheels and shutters are capable of producing vibrations during operation that can last for hundreds of milliseconds. This artifact can seriously decrease the resolution of images.

Vibration problems are also common in commercial research-level microscopes, which may contain a wide variety of motorized components. In an effort to restrict these vibrations, the manufacturers often limit the speeds of attached devices (such as axial drives, nosepieces, and condensers) or introduce time delays between cessation of motor activity and the onset of image acquisition. In many cases, however, these time delays decrease the overall performance of the system and often expose the specimen to unnecessary amounts of light. The manufacturers are also increasing their product inventories to include motorized shutters, filter wheels, fluorescent cube turrets, port controllers, and related accessories, all of which are coordinated through a single controller. These peripheral devices enable rapid, automated switching from one observation mode to another, and are ideal for live-cell imaging microscope configurations. Among the new options being offered are parfocal compensation to match the focus point when changing objectives, adjustable sensitivity for fine focus mechanisms, and coordination between motorized functions to ensure that timed sequences proceed without interruption.

A majority of the older microscopes that are already in the field are not able to take advantage of the newer automatic peripherals offered by the manufacturers and must rely on aftermarket equipment. When attaching these devices, it is prudent to mount the lamphouse, filter wheels, and shutter assemblies on a separate stand, external to the microscope, in order to eliminate vibration and to allow for synchronization between components without the necessity of introducing a delay time. In most cases, the complexity of attaching peripheral accessories requires precision alignment between the components and is best performed by mounting the entire assembly on a vibration isolation table or a breadboard equipped with shock absorbers. These platforms are available with pre-drilled and threaded mounting holes placed at 1-inch intervals. Illustrated in Figure 3 are recommended filter wheel and shutter configurations designed to reduce vibration artifacts. All components depicted in the figure are securely mounted to a breadboard or isolation table and feature a 1-millimeter air gap between the filter wheels and the microscope and digital camera system to reduce vibration coupling. In Figure 3(a), the illuminator collimator is mounted directly to a shutter housed on the filter wheel and supported by aluminum pillars. The gap between the shutter and the zoom adapter prevents rapid shutter movement from disturbing the microscope. Likewise, in Figure 3(b), there are air gaps between the filter wheel and the microscope frame, as well as between the camera and the filter wheel.

Another concern with attaching filter wheels to the emission port of an older microscope is the potential to affect parfocality, create bandpass wavelength variation, and promote aberrations by introducing flat optical plates into the focused light beam emerging from the microscope. Depending upon the configuration of the microscope optical system, an emission filter inserted into the exit beam may shift the focal plane. The severity of this artifact is determined by the thickness and refractive index of the filter. In addition, focused light passing at various angles through a bandpass interference filter can introduce variations in the center wavelength of the passband region for different parts of the beam. These problems are readily compensated by aftermarket adapters that create collimated (infinity) space at the microscope emission port, thus enabling the attachment of filter wheels and other optical devices into the parallel light path afforded by the adapter.

As an alternative to filter wheels, several multichannel and spectral imaging systems have been designed that enable simultaneous multi-color imaging or spectral separation and linear unmixing, without changing filters, using a single digital camera detector. These devices offer the advantage of eliminating motion-related artifacts that often occur with filter wheels during sequential imaging, and maintain perfect registration between images. Multichannel systems are quite useful when conducting quantitative, multicolor experiments (such as resonance energy transfer) where even single-pixel registration artifacts can lead to significant errors. Among the disadvantages of these units are limitations in filter capacity and a reduction in the effective imaging area of the sensor, which is reduced by one-half or one-quarter (depending upon the internal beamsplitter configuration and filter capacity). Newer multichannel systems partially offset the spatial limitations by splitting the light into two separate detectors (digital cameras), each equipped with an independent filter pathway and fine focus adjustment to ensure pixel registration.

The wavelength-switching devices illustrated in Figure 4 operate by either rotating new interference filters into the light path (Figure 4(a)) or by directing the emission light into separate pathways (Figures 4(b) and 4(c)). A typical filter wheel configuration is presented in Figure 4(a), where up to 10 individual filters units are accommodated in a single frame. The filter wheel is controlled by an external unit (not shown) and can rapidly switch between filters on a timescale of 25 to 40 milliseconds. Among the advantages of these wheels is the flexibility of rapidly changing filter sets and the numerous configuration possibilities for both excitation and emission light. The dual channel imaging system illustrated in Figure 4(b) features an optical train that is outlined in Figure 4(c). These multichannel imaging systems are described in detail in subsequent discussions, and are rapidly emerging as a useful alternative to filter wheels.

Multichannel imaging systems operate by collimating the light from the microscope exit port and passing it through a single dichromatic mirror to split the incident beam into two independent and spectrally distinct beams (see Figure 4(c)). One beam contains wavelengths below the cut-off point of the mirror, while the other beam contains wavelengths above the cut-off. The twin beams are folded through the optical system and may be adjusted with respect to spectral content, intensity, and polarization by passing each through the appropriate filter. Following the filtering process, the beams pass through a common imaging lens and form two spatially and spectrally distinct images on the faceplate of the detector. Note that each image is positioned so that it is projected over half of the detector array. In systems that split the light into four pathways, the single dichromatic mirror is replaced with a triple bandpass version that otherwise functions in essentially the same manner.

Variations in design enable the output of the multichannel system to be split between two camera systems (as previously discussed), enabling capture of the twin images using the entire detector array in applications where the full field of view is required. When combined with a slit system and a translating automatic stage, the multichannel system can be adapted to spectral imaging investigations. One of the primary advantages of multichannel imaging systems is the elimination of image acquisition delays and potential vibration problems due to filter wheel rotation events. Thus, any delay associated with capturing images falls on the limitations imposed by the detector readout speed. When coupled to electron-multiplying (EMCCD) cameras capable of fast readout at very low light levels, multichannel imaging systems are able to capture sub-cellular events that occur on much faster timescales than filter wheel configurations.

Live-cell imaging in confocal and multiphoton microscopy is limited to the available spectral lines of the laser systems attached to the instrument. The future of wavelength selection in widefield, spinning disk and swept field, as well as broadband total internal reflection fluorescence (white light TIRF) microscopy systems may ultimately shift, at least for many high-end applications, from bandpass interference filters to tunable monochromators, liquid crystal tunable filters (LCTFs) and AOTFs. Monochromators for optical microscopy combine an arc-discharge lamp (usually xenon) with a diffraction grating wavelength selection plate and an illuminator condenser system for output of light to the microscope (see Figure 8(d)). These devices, when controlled by fast galvanometers driving both the input slit and diffraction grating positions, are capable of delivering precise wavelength and bandwidth changes on the millisecond timescale. Liquid crystal tunable filters can be applied to investigations in spectral imaging, and feature high resolution, large aperture diameter, lack of image registration errors, and reasonable speed. On the downside, LCTFs have limited spectral range, variable bandwidths, low light transmission values (typically 30 percent of filters), and require an increase in the optical pathlength. The acousto-optic tunable filter is emerging as a workhorse for wavelength selection in confocal microscopy, but has yet to see widespread applications in other imaging modes. Problems such as dispersion, narrow bandwidth, low resolution, image smear, and other artifacts must be overcome before these devices become common in live-cell imaging applications using widefield fluorescence microscopes.

Focus and Stage Control

Focus motors are attached to the fine focus transmission gearset of a microscope to enable automated axial focus control through the image acquisition software. These devices can be employed in conjunction with autofocus functions in the software or to gather a stack of optical sections for subsequent deconvolution analysis and/or three-dimensional image reconstruction. Motorized stages can be utilized with image acquisition software to automate translation of the stage between two or more viewfields or from one well to another in a multi-well culture dish. In order to collect axial image stacks (referred to as a z-series), the microscope objective fine focus mechanism must be stepped through the specimen at precise intervals under the control of image acquisition software. A variety of motorized stepper focusing devices that drive either the entire nosepiece or the microscope stage are available from the microscope manufacturers and aftermarket distributors. Alternatively, a piezoelectric device can be attached to the nosepiece to translate a single objective up and down along the microscope optical axis. The primary advantages of stepper motors are their lower cost and virtually unlimited travel distance, which offers more control over gathering image stacks. In addition, because stepper motors are designed to work through the microscope transmission gearset, they enable the use of all objectives in a revolving nosepiece. The primary disadvantage of stepper motors is that they are slower than piezoelectric devices and exhibit significantly more hysteresis. The fastest mode of capturing z-sequences is to use piezoelectric devices, which feature much greater precision and speed than stepper motors. However, piezo drives have limited travel distance (between 100 and 200 micrometers) and are more expensive.

Presented in Figure 5 is a high-performance motorized stage assembly containing a live-cell imaging chamber. These stages are very useful for eliminating the possibility of lateral stage drift and gathering sequential images of two or more viewfields during time-lapse sequence acquisition. Aftermarket motorized stages are designed with either direct current (DC) stepper or servo motors, the latter of which characteristically produce higher translation speeds (often double that of a stepper). The ability to precisely control the position of the stage (and its repeatability) depends on the encoder feedback system, which also dictates the resolution of the stage motion. The primary advantage of stepper over servo motors is their durability and excellent slow speed control. The travel range of a typical motorized stage is 2 to 4 inches in both the x and y directions, and many have encoders capable of 40 to 100 nanometer resolution and 300 to 750 nanometer repeatability (the ability to return to the same field). Most aftermarket motorized stages are capable of mounting a full range of specimen holders, such as multi-well plates, flasks, Petri dishes, as well as standard glass microscope slides. In addition, many stages can accommodate environmental control stage adapters for live-cell imaging experiments. Motorized stages require a separate controller system that is usually equipped with a joystick device for manipulating lateral motion. The controller can either be employed as a stand-alone device or interfaced to a host computer (through an RS-232 serial port or universal serial bus) for integration with other accessories.

Fluctuations in the axial position of the microscope focal plane during the collection of sequential images from living cells is one of the most serious and frequently encountered problems in time-lapse microscopy. Often termed focus drift, changes to the microscope focal plane usually occur due to temperature variations in the imaging chamber or within the room in which the instrument is housed. A variety of commercially available software and hardware solutions have been introduced by the microscope and aftermarket manufacturers to contend with focus drift. Several of the hardware devices are autofocus systems that measure the distance between the objective front lens and the specimen by sensing light or sound reflected from the lower surface of coverslip (closest surface to the objective). This approach can be hampered, however, when high resolution oil immersion objectives are used, due to loss of contrast and reflectivity as the sensing light passes through the oil. The most advanced autofocus systems use low intensity near-infrared laser or light emitting diode (LED) sources to reflect a beam of illumination through the objective and onto the upper surface of the coverslip (supporting the cells and bathed by the culture medium), subsequently recovering the reflected light with the objective and passing it onto a detector that controls a feedback circuit to adjust the position of the objective in relation to the coverslip-culture medium interface.

Proprietary and aftermarket microscope autofocus systems can compensate for focus drift produced by temperature fluctuations (as discussed above), gravity, viscosity changes in immersion media, and vibration. The most advanced systems operate by continuously comparing the image acquired at the current axial position with a reference image that was gathered during setup. The reference is usually the reflection pattern obtained from the coverslip, whose intensity is easily measured with a specialized sensor circuit. The microscope axial control is then able to compensate for thermal and mechanical drift to reposition the coverslip in the original focal plane, thus providing sharp focus. The operator can define a specific value (termed offset) for the desired axial focus, which often deviates from the exact position of the coverslip interface with aqueous tissue culture media or buffers. Autofocus systems can correct focus drift in a timeframe of milliseconds, a clear advantage over software-based algorithms, which often take several seconds (and can produce significant phototoxicity and photobleaching).

Among the variables that should be investigated when evaluating autofocus systems is their ability to handle complex imaging of multiple features at several axial and lateral positions on a single specimen. Another important feature is the frequency and accuracy of focus correction and whether compensation is continuous or applied immediately before each image is captured. There are advantages to each approach. For experiments utilizing real-time observation, continuous feedback systems are more efficient than those that correct independently of a software prompt. However, when capturing images is the primary consideration, enabling autofocus immediately prior to acquisition is preferred because of the strict requirement to synchronize image capture with the measurement of a reference point (conducted each time focus is evaluated to determine if drift has occurred). In time-lapse experiments where a significant time interval occurs between image captures, the relatively short period required to measure focus position is inconsequential. On the other hand, when collecting image streams at high speed (2 to 30 frames per second), autofocus systems may produce an artifact known as jitter when they attempt to correct focus by inadvertently translating the stage during image capture.

Illustrated in Figure 6 is the primary advantage of autofocus systems in gathering image sequences during lengthy time-lapse experiments. Each frame in the sequence represents a time interval of 4 hours. The specimen is a culture of gray fox lung fibroblast (FoLu line) cells stably transfected with human beta-actin fused to mCherry fluorescent protein. Fluorescently labeled actin concentrates in the cytoskeletal stress fibers. Note that the cell in Figures 6(a) through 6(d) slowly drifts out of focus due to thermal fluctuations over the 16-hour time-lapse sequence. The bright and clearly defined stress fibers (Figure 6(a)) become less defined as the focus begins to drift (Figure 6(b)) until only labeled actin in ruffles and the cytosolic pool can be visualized (Figures 6(c) and 6(d)). In contrast, with autofocus, the sharp actin stress fibers (Figure 6(e)) disperse as the central cell in the viewfield enters metaphase (Figure 6(f)), but re-emerge as the daughter cells re-attach to the coverslip and spread (Figures 6(g) and 6(h)). Maintaining focus during the gathering of extended time-lapse image sequences is often fundamental to the success or failure of an experiment.

One of the final considerations in evaluating autofocus systems is the total length of time required for image acquisition, which can range from seconds to hours to days. In the longer experiments, continuous feedback systems may be hampered by mechanical drift due to hysteresis or creep arising from artifacts in piezo actuators. In the short term, these problems are usually inconsequential, but can be of significant concern for long-term acquisitions. Both of these problems can be compounded with the use of high numerical aperture (and magnification) objectives where the narrow focal depth requires stringent control. Regardless of the potential artifacts introduced by autofocus systems, their current performance and continued development potential is proving to be one of the greatest assets to high-resolution fluorescence microscopy of living cells.

Illumination Sources for Live-Cell Imaging

The traditional illumination systems employed in widefield microscopy rely on a tungsten-halogen source for transmitted light and a short-arc gas lamp for fluorescence excitation. A relatively limited number of laser configurations have been introduced as light sources for widefield investigations, but the widespread emergence of the confocal and multiphoton microscopes has vastly increased the use of lasers in optical microscopy. Because signal-to-noise is perhaps the single most important variable for gathering data in live-cell imaging, the specimen chamber must be illuminated with very high light intensities in order to maximize signal and achieve the full resolution of the optical system. However, when focused at full power through a 100x objective of numerical aperture 1.4, a 100-watt mercury arc-discharge lamp will seriously affect the viability of mammalian cells in just a few seconds. Similar damage is produced at a slower rate when cells are illuminated at full intensity with a tungsten-halogen or xenon lamp, but high-powered lasers can destroy cells even faster than a mercury lamp. As a result, the intensity of light impacting the specimen must be carefully regulated with regards to both intensity and the wavelength spectrum. Only the limited portion of the spectrum that is necessary for the investigation (in effect, for fluorophore illumination, photoactivation, photobleaching, etc.) should be used in live-cell imaging experiments. In virtually all scenarios, illumination levels and wavelength restrictions represent a compromise between image quality and maintaining cell viability.

One of the best criteria for determining if the illumination intensity is too high is to observe whether the cells enter and complete mitosis under the experimental conditions. Some cells, such as embryos, are relatively resistant to visible light, probably because they lack pathways to arrest the division cycle in response to DNA damage. However, many of the more popular mammalian cell lines, including rat kangaroo, pig kidney, and Indian Muntjac, as well as most primary cultures, are extremely sensitive to light. Regardless of the light source employed for the experiment, filters should be installed in the optical pathway to prevent contamination of the illuminating light from even trace amounts of ultraviolet and infrared radiation.

Even though the modern bandpass filters commonly utilized for fluorescence imaging have high blocking levels in the visible portion of the spectrum, they still tend to pass some radiation at extremely low and high wavelengths. Therefore, it is wise to install inexpensive glass filters in the light path to block wavelengths at the ends of the visible light spectrum. The application of these additional filters is particularly important with mercury arc lamps since they produce very high levels of ultraviolet light. Tungsten-halogen lamps emit excessive levels of infrared light and require infrared filters for all live-cell investigations. In contrast, xenon lamps generate much less ultraviolet light, although they have emission peaks in the near-infrared, while the increasingly popular metal-halide lamps produce spectral lines similar to mercury with higher levels of emission in the longer wavelengths.

The spectral profiles of several common illumination sources for live-cell imaging are presented in Figure 7. The distinct peaks present in the mercury and metal-halide lamps are discussed below. The tungsten-halogen lamp has a profile that produces relatively little output in the ultraviolet spectral region, but gradually increases before leveling in the near-infrared wavelengths. The relative power output of the filament lamp is approximately 25-percent of the mercury arc-discharge lamp at center of the visible wavelength region (550 nanometers). In contrast to the mercury lamp, the xenon arc-discharge lamp has low power output in the visible region with most of the energy being concentrated at wavelengths above 800 nanometers. However, the output profile of the xenon lamp is far more continuous in the visible wavelengths than those of mercury and metal-halide lamps. All of the light sources illustrated in Figure 7 have found significant utility in imaging living cells.

For imaging cells with contrast-enhancing techniques in brightfield mode (principally DIC, HMC, and phase contrast), the most common light source is the 100-watt tungsten-halogen lamp. In long-term experiments, this lamp is particularly stable and is subject to only minor degrees of output fluctuation (temporal and spatial) under normal operating conditions. To ensure increased temporal stability for time-lapse experiments requiring hundreds or even thousands of images over an extended period, a regulated power supply can be installed on the system. Usually, this effort is only necessary for laboratory environments where the line power is subject to frequent drops in voltage. The overall light intensity in brightfield imaging modes is reduced when green or red filters are used to block ultraviolet and blue wavelengths for purposes of preserving cell viability. In many cases, a 546-nanometer green interference is used for DIC and phase contrast, but a longer wavelength filter will usually work as well. Note that the classical use of a green interference filter to correct for chromatic aberration in brightfield imaging modes is no longer necessary with modern advanced fluorite and apochromatic objectives.

In general terms, the wavelength range of the illumination source that is least deleterious to the specimen should be employed for brightfield imaging. Extensive investigations have revealed that most cells have little tolerance for ultraviolet and infrared light, and are the least sensitive to red wavelengths, followed by green and blue. Thus, from the biological standpoint, it is reasonable to use red (600 to 650 nanometers) light for live-cell investigations whenever possible even though the longer wavelengths force a compromise on resolution, and some CCD cameras are less sensitive in this region. The resolution issue is less important than cell viability and is usually limited to a greater extent by internal cellular motions, temperature drifts, and imperfections in the optical and illumination systems. The limitations in camera sensitivity are rapidly being addressed by the manufacturers, who are striving to produce designs that are more homogeneously responsive across the entire visible spectrum.

Short arc plasma lamps have the highest luminance and radiance output of any continuously operating light source and approach, very closely, the ideal model for a point source of light. However, the arc-discharge lamps exhibit significantly greater fluctuation in intensity than do filament (tungsten-halogen) lamps because the gas plasma is inherently unstable and affected both by magnetic fields and the erosion of the electrodes. Short term stability is primarily affected by three artifacts of the arc created between the tungsten electrodes. Arc wander occurs when the attachment point of the arc on the conical cathode tip surface traverses the electrode in a circular pattern, usually requiring several seconds to move in a full circle. Flare refers to the momentary change in brightness when the arc relocates to a new area on the cathode with a higher emissive quality than the previous attachment point. Convection currents in the xenon gas or mercury vapor due to a temperature differential between the arc and the envelope generate arc flutter, which is manifested by rapid lateral displacement of the arc column. Metal-halide arc-discharge lamps contain halogens, such as iodine and bromine, in addition to mercury, and operate in a process known as the tungsten/halogen cycle where the halogens prevent vaporized tungsten emitted by the electrodes from being deposited on the walls of the envelope, thus extending the useful lifetime and stability of the lamp. These light sources are currently among the most preferred illumination sources in fluorescence microscopy.

Only 45 percent of the radiant output from a standard mercury lamp (HBO; 100 watts) lies between the useful fluorescence microscopy wavelengths of 350 to 700 nanometers. Furthermore, a majority of the energy is concentrated in prominent spectral lines at 366 nanometers (10.7 percent), 436 nanometers (12.6 percent), 546 nanometers (7.1 percent), and 579 nanometers (7.9 percent). The useful power output from a xenon lamp (XBO; 75 watts), although relatively uniform in the 350 to 700 nanometer range, constitutes only 24.5 percent of the total with most of the energy (approximately 74 percent) residing in longer near-infrared wavelengths. The xenon arc-discharge lamp is often employed when a wide range of excitation wavelengths is required. The spectral output of a xenon lamp simulates sunlight because it provides featureless intense broadband illumination without prominent lines in either the ultraviolet or visible wavelength regions.

Several of the most advanced illumination sources currently available for live-cell imaging are presented in Figure 8. The metal-halide lamp (Figure 8(a)) is rapidly becoming one of the most versatile sources for widefield microscopy, often being marketed as a replacement for the mercury arc-discharge lamp. External xenon lamphouses (Figures 8(b) and 8(c)) are usually coupled to a collimating lens (mounted on the microscope input port) via a single-mode fiber or a liquid light guide. The most versatile light source is the monochromator (Figure 8(d)), which can be employed to select specific wavelengths for excitation, but is also the most expensive option. Usually powered by a xenon or similar arc lamp, monochromators can replace the lamphouse and excitation filter wheel to enable extremely fast wavelength tuning in fluorescence microscopy. Together, the illumination options illustrated in Figure 8 provide the microscopist with a wide range of choices that can fit most budgets.

A considerable effort has been expended on synthesizing fluorophores that have absorption maxima located near the prominent mercury spectral lines. For example, the classical probes rhodamine and MitoTracker Red absorb the 546 and 579 mercury lines, respectively, with high efficiency, while the Alexa Fluor series of dyes have maxima corresponding to most of the mercury peaks (350, 405, 440, 546, and 568). For many years, mercury arc-discharge lamps were the most prevalent light sources employed in fluorescence microscopy, but as discussed above, they do not provide a uniform field of illumination and exhibit large fluctuations in output over short periods of time. Both of these artifacts can present significant problems in the quantitative analysis of fluorescence. Metal-halide lamps contain spectral lines similar to HBO mercury lamps, but also exhibit off-peak intensities that are about 50-percent more powerful. Therefore, fluorophores that are not strongly excited by peaks in the mercury lamp emission spectrum, such as fluorescein and Alexa Fluor 488, produce brighter images with metal-halide illumination sources. In addition, the metal-halide lamps produce a more uniform irradiance than mercury lamps, a quality that is easily detected with digital cameras. Xenon lamps, although not as bright, are also more stable than mercury lamps and provide almost constant excitation illumination for fluorophores across the entire visible spectrum. In order to overcome the problem of non-uniform illumination with arc-discharge lamps, a liquid light guide (discussed below) can be utilized to couple the lamphouse to the microscope input port. The light guide produces a more uniform field by scattering the light passing through. Temporal fluctuations in output are more difficult to control, but the most success has been with feedback mechanisms that control input power to the lamp.

Both transmitted and fluorescence microscopes employed for live-cell imaging are usually configured to operate under Köhler illumination. This universal lighting scheme serves to uniformly illuminate the image field using a spatially complex source (arc-discharge or tungsten-halogen lamps) by imaging only a portion of the source at the focal plane of the condenser in transmitted mode or the objective rear focal plane in epi-illumination fluorescence mode. The source light striking the specimen is even, although this light may not arrive from all possible azimuths with equal distribution. The field aperture (which effectively lies in an intermediate image plane) is imaged onto the specimen to limit the area illuminated without affecting the angle of the illuminating light. In the case of highly non-uniform sources, a diffusion filter may be used to further improve uniformity at the focal plane. Köhler illumination is not the most efficient system because it does not make use of the full surface of the source or the full angular distribution of the emitted light. However, in microscopes equipped with only a lamphouse that feeds illumination directly into a condenser system, Köhler illumination is still the best choice for microscope alignment.

Although the function of Köhler illumination systems is to ensure uniform illumination and to control its coherence, it is also possible to achieve homogeneous illumination using a light guide. Scrambling of the light, effectively decreasing its spatial or temporal coherence, is also accomplished by the application of light guides. The most widely used and practical method of coupling a light source to the microscope, while also reducing coherence, is to focus the light into a flexible length of single-mode optical fiber or a liquid light guide (see Figure 9). Thermal motion in the liquid light guide constantly alters the optical path and scatters light so that both spatial and temporal coherence are effectively eliminated. In the case of a coiled single-mode optical fiber, the cladding reflections constantly change because the fiber flexes slightly, producing an exit beam that is effectively uniform in intensity over time and space. The technique of vibrating the fiber (at a frequency of up to 100 kilohertz) is also effective in scrambling the light. The light phase is scrambled due to the varying path lengths of light waves passing through the fiber, although the high radiance and monochromaticity are preserved. The exit beam is described by a top-hat intensity profile rather than the Gaussian profile that is characteristic of laser light. To avoid possible heat damage to the light scrambler, infrared radiation as well as other unwanted emission wavelengths should be removed before they enter the fiber or light guide. Ideally, only wavelengths critical to image formation should leave the light source (optical fiber and liquid light guide transmission profiles are presented in Figure 9). Rather than relying on broadband mirrors, cold mirrors and bandpass interference filters should be employed to select the light wavelengths transmitted to the aperture disk and to allow unwanted heat to escape.

A critical issue in using a liquid light guide or optical fiber in microscopy is the efficiency of coupling the external source lamp output into the optical fiber. Most fibers have a numerical aperture between 0.2 and 0.55, and this value should be matched to the collection optics for the source. Several manufacturers provide lamphouses designed for implementation with liquid light guides in which this condition is met. The combination of a 75-watt xenon arc in an elliptical reflector, a cold mirror, and an optically matched 3 to 5 millimeter-diameter liquid light guide can deliver a light output in excess of 2 milliwatts per nanometer. The fiber end becomes the effective light source for the microscope regardless of the size of the lamp arc, resulting in a decrease in radiance compared to the arc itself. However, when the goal is uniform illumination of a large-diameter aperture, as is the case in the spinning disk confocal instrument, an extended source is not as detrimental to performance. The only requirement is a collimating lens of sufficient diameter to efficiently collect the output from the light guide and project it onto the disk scanner.

In spinning disk confocal microscopy, the scanning mechanisms employ an illumination method that fills the rear aperture of the objective. Coupling the available light from a light source to the specimen is somewhat more involved in the non-laser case, because sources such as arc lamps radiate into a sphere rather than producing a parallel beam. Consequently, reflectors are required to direct light from the backside of the arc toward the specimen. Unneeded wavelengths, such as infrared and ultraviolet, can be allowed to pass through dichromatic mirrors (an infrared transmitter is denoted a cold mirror and an ultraviolet transmitter a hot mirror) and be absorbed in the lamp housing. The dissipation of heat by a cold mirror reduces source movement caused by thermal expansion of the mechanical and optical components. In all forms of microscopy, lenses near hot sources must be mounted appropriately to allow for thermal expansion. A variety of aftermarket light sources are available that couple the microscope to an external light source (usually xenon or metal-halide) via an optical fiber or liquid light guide (see Figure 8). In addition, specialized monochromators are available for precise wavelength selection. Although expensive, these devices offer an excellent alternative for quantitative microscopy. The final exercise in choosing the appropriate illumination source and wavelength region for live-cell imaging should be approached with the understanding that high-intensity light of any color is inherently deleterious to living cells. Regardless of whether or not the wavelength range is optimized, the light source must be shuttered at all times except when images are being acquired.

Software for Live-Cell Imaging

Digital cameras designed for optical microscopy are generally interfaced to a host computer workstation, which allows the user to adjust camera settings, preview specimens, gather images, and store the data collected during an investigation. Among the popular camera interface buses are FireWire (IEEE-1394), the universal serial bus (USB), traditional serial ports (RS-422), small computer system interface (SCSI), and a wide spectrum of proprietary interfaces that are supported by camera-specific integrated circuit boards inserted into the computer. A computer dedicated to live-cell image acquisition will require a fast microprocessor and a significant amount of random access memory, as well as sufficient hard drive space to store images and data. The recommended starting point is a 3 gigahertz (GHz) processor and at least 1 gigabyte (GB) of memory. Hard drives are relatively cheap, so the workstation should have one or more 250+ GB drives and will significantly benefit from a high-performance graphics card. A disk drive for writing data to compact disk (CD) or digital versatile disk (DVD) is also recommended. The monitor should be a high-resolution flat panel to conserve desk space and provide bright, sharp, flicker-free images.

The image acquisition software provided with most digital camera systems is used to adjust camera settings such as gain, offset, exposure time, and binning. Several of the more sophisticated packages are also able to perform varying degrees of image processing tasks, including control of brightness, contrast, and gamma, adding pseudocolor, image merging, creating overlays, and producing digital video sequences. The most advanced software packages will provide extensive tools for post-acquisition image analysis, deconvolution, and complex measurements, as well as total control of the additional hardware (such as filter wheels, shutters, z-axis drives, stages, etc.) necessary for automated live-cell imaging. Drivers are available from both camera and software manufacturers to interface almost every model of digital camera to the image acquisition and microscope control software programs. In addition, many camera companies also provide software development kits for researchers interested in programming specialized features into camera driver software. Modern confocal microscopes are equipped with software than enables the researcher to immediate analyze results from experiments that include colocalization, photobleaching, resonance energy transfer, photoactivation, as well as a variety of other techniques. When choosing software for a customized live-cell imaging configuration, the investigator should consider a package that allows the most flexibility in designing experiments.

Conclusions

During the configuration stage of an optical microscope system for live-cell imaging, the most important consideration to bear in mind is that a compromise must often be made between keeping the cells viable and achieving the best temporal and spatial resolution. Most commercial microscopes are designed to attain the highest possible image quality, but are not necessarily configured to maintain healthy cell cultures. However, with careful planning, a standard off-the-shelf microscope can readily be modified into a true live-cell imaging system. The essential considerations are specimen chambers, focus drift, digital camera systems, illumination sources, shutters, filter combinations, vibration compensation, and software to interface the entire system into a cohesive unit. Although there are many alternatives to the systems discussed in previous sections, the concepts described in this review have proven their utility in numerous investigations.


BACK TO LIVE-CELL IMAGING

Contributing Authors

Michael E. Dailey - Department of Biological Sciences and Neuroscience Program, 369 Biology Building, University of Iowa, Iowa City, Iowa, 52242.

Alexey Khodjakov and Conly L. Rieder - Wadsworth Center, New York State Department of Health, Albany, New York, 12201, and Marine Biological Laboratory, Woods Hole, Massachusetts, 02543.

Jason R. Swedlow, and Paul D. Andrews - Division of Gene Regulation and Expression, MSI/WTB Complex, University of Dundee, Dundee DD1 5EH, Scotland.

Jennifer C. Waters - Nikon Imaging Center, LHRRB Room 113C, Department of Cell Biology, Harvard Medical School, 240 Longwood Avenue, Boston, Massachusetts, 02115.

Nathan S. Claxton, Scott G. Olenych, John D. Griffin, and Michael W. Davidson - National High Magnetic Field Laboratory, 1800 East Paul Dirac Dr., The Florida State University, Tallahassee, Florida, 32310.