By using imaging techniques, plant physiological parameters can be assessed in vivo without contact with the plant and in a non-destructive way, from a microscopic to a remote sensing scale. The major advantage of imaging consists in the instantaneous visualisation of heterogeneity in the studied parameter of a leaf or leaf area. To obtain such a detailed overview, classical techniques would need a multitude of repeated point measurements at each time point. Moreover, measurement on the same minute regions of a sample at each measuring session is nearly impossible. Especially destructive measurements are labour intensive and time consuming, and one has to assume that the different samples used for each determination are physiologically 'comparable'. Thus imaging techniques are an ideal method to visualise the evolution of patterns of plant physiological parameters.
Imaging has a vast potential for highlighting local changes in a parameter from surrounding 'background' levels. This discriminating power of imaging techniques allows an early detection of deviations from optimal physiological conditions. Imaging in the visual spectrum is a popular technique to record long-term physiological processes as growth and development. This type of monitoring is frequently designated time-lapse imaging. However, in addition to visual assessment, imaging in other spectral regions permits to reveal other fundamental plant physiological parameters as transpiration and chlorophyll fluorescence. Imaging in narrow spectral bands, coined hyperspectral imaging, makes it possible to extract information regarding concentrations of specific compounds in plants. Microscopic techniques, which will be mentioned briefly in this review, are generally applied for fundamental research on cellular and subcellular changes during development or upon stress. This still holds for the macroscopic scale, covering measurements on plant organs or single plants. At canopy and field scale, imaging techniques have the potential to identify emerging stresses and to guide sampling for identification of the stressor. These remote sensing applications are increasingly being applied in the rapidly evolving precision agriculture, which should improve crop production, while being more environmentally friendly than the 'classic' production methods
At ambient temperature, all objects emit far infrared light of approximately 10 μm wavelength (Nobel, 1991). Detectors sensitive in the 8-14 μm wavelength band convert this radiation into a temperature reading. Such detectors are the basis of non-imaging infrared thermometers, which yield an average temperature measurement of all objects within the field of view. Applications of these simple and affordable instruments include forest canopy studies and irrigation scheduling in field crops (Samson and Lemeur, 2000; Wanjura and Upchurch, 2000). Images can be generated from a thermal infrared detector by adding a scanning system. Each point measurement is assigned a pseudocolour value depending on the radiation captured. In this way patterns of radiation are converted to visual pseudocolour images representing temperature levels. Scanning thermal imaging systems have been used to monitor temperature distribution at the plant level since the mid-seventies (De Carolis et al., 1975). The last few years a new generation of thermal imaging systems has been commercialised. The use of arrays of detectors obviates the incorporation of expensive high-speed scanning systems. Moreover sensors that can operate at ambient temperature have been developed, whereas previously detectors needed to be cooled cryogenically to obtain a temperature resolution of 0.1 °C. These technical improvements make thermal cameras more affordable and user friendly, promoting an increase in industrial and expectedly in biological applications (Majumdar and Norton, 1999)
The energy budget of a plant, and thus its temperature is dependent on environmental factors. Sufficient light and water are the main factors in the energy budget and a prerequisite for satisfying yield. The most common plant stress is water shortage, which results in a higher leaf temperature due to decreased transpirational cooling. Most applications of thermography in plant physiology are based on monitoring changes in transpiration.
Monitoring of thermogenic flowering
As an exception in the plant world, flowers from a few plant families have the capa-city to generate heat, which is used to attract pollinators by volatilising attractants. Thermography has been used to study the timing of this thermogenesis phenomenon (Raskin et al., 1989). On the other hand, metabolism has a negligible influence on plant leaf temperature. Heat can be accumulated only in tissues that have a high heat capacity, such as the specialised flower structures mentioned above, (Breidenbach et al., 1997).
Thermographic visualisation of freezing
While evaporation uses latent heat, and thus has a cooling effect on leaves, freezing processes liberate latent heat. Thermal imaging was successfully applied to detect locally initiated freezing processes in plants (Pearce, 2001). It is particularly important to correlate the speed and spread of freezing with subsequent visible freezing damage (Pearce and Fuller, 2001). Such thermal studies of freezing processes in function of time can aid in developing frost-protection strategies (Wisniewski et al., 1997).
Thermography in mutant screening
Figure 1. Thermal imaging of drought stressed Arabidopsis.
Control plants (top row), an ABA-insensitive mutant (abi-2-1, middle row) and an abi-2-1 revertant showing a suppression of ABA-insensitivity (abi2-1R1, bottom row) were 17 days old when subjected to four days of drought stress. Leaf temperature of ABA-insensitive plants was 1 to 1.5 °C lower when compared to the control plants, whereas the temperature of abi2-1R1 plants was comparable to that of wild-type. The abi2-1R1 mutant was isolated during a thermographic screen of a drought-stressed mutagenised abi-2-1 population for plants with a higher leaf temperature. The scale is in °C. (reproduced with permission from Merlot et al., 2001.)
n case of water shortage, plants limit transpiration by closing their stomata. Stomata can be seen as microscopic valves that optimise the uptake (and thus assimilation) of CO2 while limiting the loss of water. Under water stress, stomata close via a process controlled by abscisic acid (ABA) (Hetherington, 1998). Since thermography readily visualises changes in transpirational cooling, this imaging technique is an efficient tool to screen plant populations for aberrant transpiration. Initially, thermography was used to isolate ABA-insensitive mutants in barley (Raskin and Ladyman, 1988). In recent years the technique has been applied to screening of the model plant Arabidopsis (Figure 1), since hormonal regulation of plant transpiration is a logical target for thermographic screening (Merlot et al., 2001). Thermal imaging is also well suited to check stomatal functionality in mutants isolated with other screening methods (Gray et al., 2000).
Estimation of transpiration
Although several physiological methods are available to measure transpiration, none of them can be easily applied at the field scale, neither do they permit long-term continuous measurements on a growing crop (Merta et al., 2001). By virtue of its non-contact methodology, thermography can be applied at any scale from the laboratory (Figure 2) to air-borne field applications. Moreover, automatic surveillance of the water status of a crop is an application within reach of current technological possibilities (Jones, 1999a). Since plant leaf temperature is dependent on weather conditions, the application of thermography to assess water stress on the field is proposed in conjunction with a dedicated calibration method, to compensate for the variable environmental conditions (Jones, 1999b). Importantly, stomatal closure also affects photosynthesis (Osmond et al, 1998; Cornic, 2000; Baker et al., 2001). Thus thermal imaging could complement chlorophyll fluorescence imaging measurements, making it possible to discern water stress from other stresses affecting chlorophyll fluorescence. This could be achieved with a combined portable thermal and fluorescence (multispectral) monitoring system (see 5.2).
Detection of pathogen infections
Figure 2. Thermographic visualisation of the ‘Iwanov’ effect in the first trifoliate leaf of French bean. The leaf was severed from the plant 10s after the first image was taken. The response results from cutting the leaf from the plant and consequently interrupting its water supply. The image series was captured with 1-minute intervals. First, a rapid cooling is apparent, characterised by initial stomatal opening. The onset of re-warming (see temperature indications in the images of the last row) is due to stomatal closure as the leaf water status declines. These changes in stomatal aperture, and thus transpiration and evaporative cooling, cause a decrease in leaf temperature followed by an increase. The temperature indications on each figure indicate the average temperature of the same 1cm2 region on the leaf surface (Reproduced with permission from Jones, 1999b).
During plant-pathogen infection, the physiological state of the invaded tissue is altered. This can be reflected in changes in photosynthesis, transpiration or both. Fluorescence imaging (see 5.2) and thermography are thus useful for rapid visualisation of emerging biotic stresses.
Some compounds produced by pathogens are known to induce stomata closure. Examples are hydrogen peroxide (McAinsh et al., 1996), elicitors generated during infection of tomato (Lee et al., 1999), toxins produced by the plant-pathogenic bacteria Pseudomonas syringae (Mott and Takemoto, 1989) and an unidentified product produced during the infection of soybean by the phytopathogenic fungus Phytophthora (McDonald and Cahill, 1999). Salicylic acid (SA), the central signalling component in the disease resistance response of many plant species (Delaney et al., 1994), also diminishes stomatal aperture (Larqué-Saavedra, 1979; Manthe et al., 1992). The stomatal movements that occur during plant-pathogen interactions can be imaged by thermography. A few examples are described below.
Figure 3. Two pairs of infrared and visual spectrum images at an early and a late stage of the incompatible interaction between tobacco and tobacco mosaic virus (TMV). The plant was infected and kept at 32°C for 29 hours to allow the spread of TMV. The plant was then shifted back to 21°C. Two and a half hours after the temperature shift a thermal effect emerged. The upper left panel shows the presymptomatic thermographic visualisation of the infection 8 hours after the temperature shift. The right panel shows the video image taken at the same time. No signs of cell death are yet visible. The lower left and right panels were captured 122 hours post temperature shift, and show extensive cell death at the places of infection (Chaerle and Van Der Straeten, unpublished).
n the well-characterised plant-pathogen model system tobacco - tobacco mosaic virus (TMV), resistance is accompanied by a necrosis response termed hypersensitive response (HR). In this incompatible plant-pathogen interaction, SA starts to accumulate before the appearance of visible necrosis (Enyedi et al., 1992; Malamy et al., 1992). As was hypothesised from the above facts, thermography monitored an early increase in leaf temperature at the points of TMV-infection (Chaerle et al., 1999, Figure 3). The visualisation of the emerging 'hot-spots' was presymptomatic, because the first signs of necrosis only became visible 8 hours later. Moreover, the full extent of the thermal effect corresponded with the area of cell death 4 days later. The concentration of water vapour in the air circulated through a cuvette clamped onto an infected leaf area, as measured by continuous infrared gas analysis (IRGA), revealed a local decrease in transpiration. This decrease correlated with the increase in leaf temperature measured thermographically. A robotised system coupled to an 'on-line' 'real-time' continuous visualisation method (see www.plantgenetics.rug.ac.be/~lacha) was used for thermography-aided selective sampling of visually asymptomatic tissue. The increase in SA content of these samples as a function of time corresponded well with the measured decrease in transpiration as infection progressed. Such site-specific sampling could also be applied to the molecular characterisation of the early stages of plant-pathogen interactions. Panoramic views of gene expression during biotic stress were recently obtained using microarrays of a selected set of genes (Maleck et al., 2000; Reymond, 2001; Schenk et al., 2000). This approach could thus be extended to early symptomless time points by taking advantage of imaging-aided selective tissue sampling.
The agriculturally more important Brassicanapus infectionwith Phomalingam was also visualised presymptomatically by thermal imaging. The incompatible interaction between Brassica and Phoma results in stem canker, a disease characterised by stem necrosis (Lamkadmi et al., 2000). In addition to the thermal data, changes in protein patterns were detected before the appearance of disease symptoms, and amongst the differentially induced polypeptides, a phosphatase was identified (Lamkadmi et al., 1996).
Cell death is a common symptom during the HR in incompatible plant-pathogen interactions. In mutants and in some transgenic lines from several plant species, spontaneous cell death occurs (Dietrich et al., 1994; Johal et al., 1995; Mittler and Rizhsky, 2000). Comparable to the visualisation in the tobacco-TMV system, the onset of cell death could be revealed before the appearance of visual symptoms in lsd2 Arabidopsis mutants and in tobacco bacterio-opsin transgenics (Chaerle et al., 2001). Thermography visualised the dynamic evolution and spreading of these cell death phenomena in function of time. Since cell death at the surface of plant leaves results in evaporation of the contents of disrupted epidermal cells, high-contrast images can be thermally recorded.
Non-destructive testing in biology
Together with process monitoring and preventive maintenance control, non-destructive testing is a frequent application of thermography in many industrial sectors. In the sector of processing of biological materials, thermography can for instance be applied for non-destructive testing in the wood industry (Wyckhuyse and Maldague, 2001).
Active thermography, which uses pulses from high-power infrared lamps to heat transiently the monitored samples, has also found widespread application in industrial manufacturing. Transient heating permits thermographic detection of local differences in heat capacity, linked to structure or composition, by monitoring the speed of either warming or subsequent cooling. This technique has also been used in medical applications (Fujimasa, 1998) and in post-harvest processing of fruit (Offermann et al., 1998). An agricultural robotics application was developed based on active thermal imaging to detect diseased potato tubers (Lefebvre et al., 1993). Furthermore, the active thermography methodology was used to visualise water distribution in plant leaves (Kümmerlen et al., 1999). This setup monitors the temporal evolution of water content in intact plants, which is impossible with the commonly used destructive methods.
Thermal remote sensing
Applications at the remote sensing scale were attempted as soon as portable thermal imagers became available, and are presently refined for integration into precision farming (Liu et al., 2000). These measurements however depend on environmental conditions, which influence the thermal properties of the visualised crop. Calibration of images according to weather conditions is necessary for comparison between image data obtained during different measuring periods and growth seasons (Nilsson, 1995). A field trial with winter wheat proved that root nematode infection increases canopy temperature in comparison with a control field (Nicolas et al., 1991). In the case of an economically important rice disease, rice blast, leaf temperature was shown to correlate with the severity of disease in the field, whereas remote sensing in the visual spectrum could not discern infested from control fields (Yamamoto et al., 1995).
Robotised thermography is exploited to screen animal cell cultures for their thermal metabolic reaction to pharmacological compounds (Paulik et al., 1998). Such a high-throughput non-destructive testing setup could be used for plant cell cultures as well. Recently, thermal imaging was applied to monitor in-vitro reactions of enzymes with stereospecific substrates (Reetz et al., 2001). Exothermic reactions show up as hot-spots in microtiter plates used for screening. This application is proposed for identification of enantioselective enzymes, which have a broad application range in microbiology and biotechnology.
A wide variety of life forms, including insects, jellyfish and bacteria, can emit visible light through enzymatic reactions on high-energy substrates. Substrate availability is the major limitation for the long-term application of these reporter systems in plants. To detect the relatively weak light emission, measurements have to be carried out in darkness with cooled image-intensified CCD cameras. This technique is essentially different from fluorescence imaging (see 5) where chromophores, as for example GFP, absorb light from an intense light source and emit light at a lower wavelength.
The photoprotein aequorin, naturally present in the jellyfish Aequorea victoria, emits blue light upon exposure to free calcium. This Ca2+ reporter protein consists of a polypeptide apoaequorin and its luminophore coelenterazine. To avoid the invasive method of loading of the calcium reporter molecule aequorin into plant cells, transgenic plants expressing apo-aequorin were generated. Thus, measurements of free calcium concentration dynamics at the whole plant level were made possible (Knight et al., 1991). Aequorin needs to be reconstituted in vivo in apoaequorin- producing transgenic plants by floating seedlings on an aqueous solution of coelenterazine. To localise precisely the emission of aequorin bioluminescence, luminescence images are superimposed with reflection images taken with the same microscope, under dark-field visual spectrum illumination. Since Ca2+ is a ubiquitous second messenger involved in multiple stress responses, aequorin imaging has been adopted to assess calcium signalling in diverse stress responses, including the oxidative burst after pathogen attack and guard cell signalling (Grant et al., 2000; Wood et al., 2000).
The luciferase enzyme (LUC), encoded by the firefly luc gene, is a popular in vivo reporter of gene expression. As is also the case for the above-described aequorin imaging, the substrate (luciferin for luciferase imaging) has to be kept at a sufficient concentration either by regular spraying or overnight imbibition. Since luciferase has a half-life of 3 hours, continuous assessment of changes in its expression is possible. The prime application has been the study of biological rhythms in plants (McWatters et al., 2000; Somers et al., 2000). In addition, plant-pathogen interactions, oxidative stress and wounding have been studied using this reporter system (reviewed in Chaerle and Van Der Straeten, 2001). Recently, luciferase imaging has been used to isolate mutants in gibberellic acid response (Meier et al., 2001). To spatially correlate the effects of photooxidative stress - as observed with high resolution fluorescence imaging (see 5.2) - with gene expression, plants were transformed with the LUC reporter under the control of a light-stress induced peroxidase promoter (Baker et al., 2001). Parallel imaging of chlorophyll fluorescence and luminescence will help to elucidate the effects of light-inhibition of photosynthesis.
A different type of application makes use of the luciferase operon, which encodes both a luciferase and its substrate, present in certain marine bacteria. Plant-pathogenic bacteria transformed with this operon can be visualised in planta. This possibility was exploited to selectively sample symptomless tissue already invaded by the pathogen (Gay and Tuzun, 2000). Comparison between the responses of different cultivars to bacteria can help to characterise resistance factors.
The human vision system is based on detection of reflected radiation in the 400-800nm spectral range. Consumer photographic emulsions mimic this spectral sensitivity. The spectral sensitivities of custom emulsions and of CCD detectors in video cameras extend into the near infrared. Consumer video cameras have a filter that leaves out the near infrared radiation (NIR). Imaging in the visible spectrum is used to obtain reference images for fluorescence, thermal and bioluminescence imaging. This approach makes it straightforward to correlate patterns observed in other spectral regions with concurrent or later emerging visual symptoms.
On the microscopic scale, custom-designed measuring cuvettes allow to keep live samples in controlled conditions for long periods of time. Cytological and morphological changes in symbionts and host cells were visualised by time-lapse imaging (Giovannetti et al., 2000). Confocal fluorescence microscopy (see 5.1) of the interaction between Nicotiana roots and arbuscular mycorrhizae (Vierheilig et al., 2001) could be extended in time by using dedicated observation chambers. The same holds for plant-pathogen interactions (Howard, 2001). At the same measuring scale, stomatal movements in response to stress can be imaged in function of time, in a custom-built gas-exchange chamber (Kaiser and Kappen, 2000). Individual stomata within the enclosed 25x50mm leaf area can be imaged repeatedly through electronic remote control. Simultaneous information on stomatal aperture, CO2 assimilation and transpiration, the latter both being determined on-line using IRGA (see 2.5), provides all necessary information to study the limitation of decreased stomatal aperture on photosynthesis in function of light and/or humidity (see 2.4).