Lighting systems for Chl a fluorescence imaging
Most types of Chl a fluorescence imaging system require one or more sources of illumination for three different purposes; (i) to excite Chl a fluorescence during imaging, (ii) to provide constant actinic illumination; and (iii) to provide the multiple turnover pulses for measurement of Fm and F'm.
One of the earliest Chl a fluorescence imaging systems with the ability to image at Fo used a single source to provide all three illumination requirements (Oxborough and Baker 1997a, b), with images being taken at the prevailing photon irradiance. Fo was measured over a period of several seconds or minutes at a photon irradiance of less than 1 µmol m–2 s–1, F' under the prevailing actinic illumination level, and Fm and F'm at the end of a saturating pulse at a photon irradiance of several thousand µmol m–2 s–1. Calibration involved the taking of a series of reflected light images; one for each combination of photon irradiance and exposure time used during an experiment. Although this method worked well, it required the use of a very expensive Peltier-cooled CCD camera and long exposure times for measurement of Fo and F' at low photon irradiances. It is now more usual for imaging systems to use LEDs to provide ‘measuring pulses’, which maximize the signal-to-noise ratio of the camera system and allow for optimization of exposure times.
Provided the actinic effect of the measuring pulses is negligible, a certain amount of unevenness of illumination from this source is acceptable. This is because parameter images are always normalized, such that measuring pulse inhomogeneities are cancelled out. Conversely, the accuracy of parameter images is, to some extent, dependent upon the homogeneity of the actinic illumination that is incident on the sample. Consequently, it is essential that incident illumination from this source is as even as possible. The most important requirement of the illumination source providing the super-saturating pulses for imaging Fm and F'm is that the pulses are of sufficient intensity to close the majority of PSII centres within the entire imaged area.
Measuring light source
Measurement of Chl a fluorescence requires minimal spectral overlap between the measuring (excitation) light source and the detection system. The fluorescence emission spectrum of Chl a exhibits a peak at 682 nm plus a broad shoulder out to approximately 740 nm. Clearly, the largest Chl a fluorescence signal will be achieved if the detection system covers the entire emission spectrum (from approximately 670 nm to 750 nm), which would require that the excitation light does not emit photons of wavelengths longer than 670 nm. With conventional fluorometers, this requirement is easily satisfied; for example, by using a filtered xenon light source or blue LEDs to provide the excitation pulses. It is also possible to use a filtered xenon light source or blue LEDs to provide excitation pulses with imaging systems. However, the relatively high cost of these sources make them increasingly impractical as the area being imaged increases.
At the time of writing, the most cost-effective method of providing excitation pulses for the imaging of Chl a fluorescence over large areas (greater than a few square centimetres) is to employ orange-red LEDs. Although these LEDs have an emission peak at 620–630 nm, they also exhibit an emission ‘tail’ that extends to 660–670 nm. This effectively limits the measurement of Chl a fluorescence to wavelengths above approximately 710 nm. In addition to the obvious effect this has on signal size, limiting the measurement of fluorescence to this range of wavelengths also increases the fraction of the fluorescence signal that is emitted from PSI (see ‘Sources of error’ in the section on ‘Deriving information from chlorophyll a fluorescence’, above).
Constant actinic light and saturating pulses
With conventional fluorometers, the measuring pulses are usually bright enough that the additional contribution of fluorescence generated by actinic illumination can be ignored. With imaging applications, it is often not cost-effective to provide measuring pulses of sufficient intensity for this to be done. Consequently, the options are to switch off the actinic illumination while the measuring pulse is applied or to correct for the fluorescence that is generated by the actinic source. A system utilized by Zangerl et al. (2002) uses the former approach. With this system, an array of red LEDs, which provides the constant actinic illumination and super-saturating pulses for measurement of Fm and F'm, is switched off approximately 1 µs before the start of the measuring pulse, which is provided by an array of blue LEDs. A system described by Nedbal et al. (2000) uses an array of orange LEDs to provide measuring pulses, which are applied over and above the prevailing actinic illumination. The fluorescence generated by the actinic illumination is compensated for by taking two images, the first without the measuring pulse and the second with the measuring pulse. Subtracting the first image from the second generates an image that approximates the fluorescence generated by the measuring pulse alone. This approach can be very cost-effective, since the cost of non-LED based actinic illumination is significantly lower than LED-based illumination. A significant disadvantage is that a fraction of the dynamic range of the camera is sacrificed during image subtraction.
All-in-one LED-based illumination systems
An all-in-one LED-based illumination system must obviously employ LEDs that satisfy the wavelength criteria outlined in the ‘Measuring light source’ section, above. Beyond this requirement, the most difficult criterion to satisfy is the provision of super-saturating pulses for measurement of Fm and F'm. As noted earlier, the highest output is currently provided by orange LEDs, although blue LEDs have also successfully been used in this type of system (Barbagallo et al., 2003; Oxborough, 2004).
Cameras and frame grabbers
Existing Chl a fluorescence imaging systems are based around cameras that utilize a charge-coupled device (CCD) sensor for image capture. These silicon-based devices are divided into a two-dimensional array of wells, which accumulate electrical charge through the absorption of incident photons. At some point, the charge that accumulates within each well must be converted to a number; a process that involves an analogue to digital converter (adc). With digital camera systems, the adc is located within the camera itself. With analogue camera systems, the adc is part of a frame grabber, which is normally located inside a computer. In terms of both cost and performance, there is little to choose between analogue and digital systems.
The integrated signal size (S) from any Chl a fluorometer system, can be defined in terms of equation 5. With imaging systems, S represents the charge accumulated by the wells within the sensor array, which is proportional to the number of photons absorbed.
S Fx I x A x t + kdt + R(5)
where: S, integrated signal size; F, chlorophyll fluorescence yield; I, incident PPFD; A, the fraction of incident photons absorbed; t, integration time; kd, rate constant for dark-noise; R, read-noise.
Two important sources of noise within imaging systems are included in equation 5. The first is so-called dark-noise (expressed as kdt), which is the accumulation of charge due to thermal events within the CCD and is proportional to the integration (exposure) time, t. The second is noise associated with reading and digitization of the image, which is independent of t. It is generally the case that read-noise degrades image quality far more than does dark-noise.
Camera dark-noise can be decreased to insignificant levels by cooling the CCD (which is usually achieved using an integrated Peltier device). This allows for the very long integration times that are often required when imaging Fo (see below), particularly when working at the microscopic level. Unfortunately, the cost of Peltier-cooled CCD cameras is relatively high.
For a usable fluorescence image to be generated, there is an obvious requirement for a certain number of photons to be absorbed by the CCD. Signal size can be improved by increasing the integration (exposure) time and/or increasing the incident photon irradiance. However, either increase has the potential to impact on the de-excitation processes at PSII. The effect is most acute when imaging Fo, where there is a requirement that the measuring light has minimal actinic effect, to minimize the closure of PSII centres while the image is being accumulated.
The shortest measuring pulse that can provide enough photons to generate a usable image is likely to be in the region of 10–100 µs. This is orders of magnitude longer than the time-constant for charge-stabilization at PSII, of approximately 300 ps (Roelofs et al., 1992; Dau and Sauer, 1992), and comparable to the time-constants for the opening of a closed PSII centre through the transfer of an electron from Q–A to plastoquinone or semi-plastoquinone at the QB-site, which are a minimum of 200 and 400 µs, respectively (Crofts et al., 1993; Robinson and Crofts, 1983). Consequently, the ratio of the yield of PSII centre closure to the yield of PSII centre reopening during the integration period is very high with a measuring pulse of this duration. This maximizes the fraction of PSII centres that are closed during the integration period, which increases the probability of Fo being overestimated.
While there is essentially nothing that can be done to decrease the yield of PSII centre closure significantly (since this would require integration times of considerably less than 1 µs), the yield of PSII centre reopening during the integration period can be increased by simply increasing the integration time and decreasing the incident photon irradiance, such that the product of the two is unchanged.
There are two ways of increasing the integration time: (i) increase the length of a single exposure; or (ii) average multiple integration periods that have a relatively long dark interval between them. For example, the generation of a usable Fo image might require a single integration period of 1 s at a photon irradiance of 1 µmol m–2 s–1. As an alternative, the same number of photons could be delivered to the sample during 10 widely spaced integration periods of 100 µs each, at a photon irradiance of 1000 µmol m–2 s–1. While the first method will accumulate 1000 times as much dark-noise as the second, the second method will accumulate 10 times as much read-noise as the first. Consequently, the best method will depend on the characteristics of the imaging hardware being used. For example, the very low rate of dark-noise accumulation by cameras that have a Peltier-cooled CCD sensor makes the single exposure method the best option in this instance. With other types of camera, the best option will depend upon the relative levels of dark-noise and read-noise. As noted in the ‘Cameras and frame grabbers’ section, read-noise is generally a more significant problem than dark-noise.