Surveying the plant’s world by magnetic resonance imaging


  • Ljudmilla Borisjuk,

    Corresponding author
    1. Leibniz-Institute of Plant Genetics and Crop Plant Research (IPK), Corrensstraße 3, Gatersleben, Germany
    Search for more papers by this author
  • Hardy Rolletschek,

    1. Leibniz-Institute of Plant Genetics and Crop Plant Research (IPK), Corrensstraße 3, Gatersleben, Germany
    Search for more papers by this author
  • Thomas Neuberger

    1. Department of Bioengineering, Pennsylvania State University, University Park, PA 16802, USA
    2. Huck Institutes of the Life Sciences, High Field MRI Facility, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author



Understanding the way in which plants develop, grow and interact with their environment requires tools capable of a high degree of both spatial and temporal resolution. Magnetic resonance imaging (MRI), a technique which is able to visualize internal structures and metabolites, has the great virtue that it is non-invasive and therefore has the potential to monitor physiological processes occurring in vivo. The major aim of this review is to attract plant biologists to MRI by explaining its advantages and wide range of possible applications for solving outstanding issues in plant science. We discuss the challenges and opportunities of MRI in the study of plant physiology and development, plant–environment interactions, biodiversity, gene functions and metabolism. Overall, it is our view that the potential benefit of harnessing MRI for plant research purposes is hard to overrate.


Magnetic resonance (MR) images derive from spatially encoded nuclear magnetic resonance (NMR) signals. The first MR images (at that time better known as zeugmatograms) were acquired less than 40 years ago by Lauterbur, (1973). The non-invasiveness of MRI has encouraged its widespread adoption and continuing development as a clinical tool (Simon and Mattson, 1996), and its value was recognized by the scientific community in the awarding of the Nobel prize for physiology or medicine in 2003 to its inventors, Lauterbur and Mansfield.

The earliest application of MRI in the plant sciences was based on the use of a clinical human NMR scanner (Hinshaw et al., 1979; Bottomley et al., 1986), but with the size of the capital expenditure needed to equip and maintain a dedicated MRI facility, along with the rather modest spatial resolution achieved at the time, optical microscopy maintained its role as the primary means of exploring the internal structures of plants. At the same time, due to a number of technical issues specific to plants – in particular the wide diversity with respect to organism size and their sessile nature – plants probably did not represent an attractive subject for NMR scientists. With the advances in hardware development in the last decades, the realization of ultra high magnetic fields, and the development of new imaging techniques, most of these problems have been solved, and the way has been paved for the application of MRI in plant research. Currently, however, the technique remains surprisingly underused, probably as a result of a widespread lack of awareness of its potential for solving outstanding issues in plant physiology. This review aims to highlight the current potential of MRI for applications in plant science, and also to provide a forward look at likely future developments in the technique.

Why Use MRI?

The principles by which MRI images are acquired differ fundamentally from those underlying conventional optical methods (Callaghan, 1993). The primary advantage of MRI is that both static and dynamic parameters can be spatially resolved, but importantly, the technique generates data in a non-destructive manner from the interior of the sample. In this way, the morphology/anatomy of opaque samples of whatever size, form or composition can be imaged, while at the same time allowing an assessment of a range of chemical parameters. Hence, this enables the visualization of the long-term dynamic behaviour of living plant tissue. It is possible to generate metabolic maps of the living plant body, and to use such a data to monitor various physiological processes. As an example, an MRI scan of a living plant stem can demonstrate the location of the xylem vessels, give information as to whether or not they are filled with liquid, derive the velocity and direction of this liquid’s movement and determine the identity of the metabolites dissolved in it (Metzler et al., 1995; Van As et al., 2009; Windt et al., 2009).

Few, if any, other analytical techniques addressing the physiology and development of living plants in their natural environment are as versatile as MRI (Ratcliffe, 2010). The other methods considered for this type of research are based on either visible range radiation (confocal laser scanning microscopy, optical coherence microscopy and optical projection tomography), on X-rays (high-resolution computed tomography) or on positrons (positron emission tomography, PET). A disadvantage of all optical techniques is that thick specimens usually need to be cleared with an organic solvent, precluding the possibility of live imaging (Lee et al., 2006). X-ray irradiation is incompatible with metabolite analysis, while the spatial resolution of PET is at best modest. There are a number of excellent techniques like Raman spectroscopy, matrix-assisted laser desorption/ionization (MALDI) and related mass spectrometry methods which provide very detailed chemical information on a spatial scale. However, their use is restricted to either surface imaging or tissue sections. In contrast, MRI can image in real time irrespective of sample thickness (Holbrook et al., 2001), and allows the measurement of both the distribution and dynamics of water and a range of plant metabolites.

What Prior Knowledge is needed to Understand MRI?


A brief explanation of the physics underlying MRI is given here, but in the interests of clarity for the non-specialist, more detailed information has been included in the Supporting Information (Data S1). Readers who are interested in more details are referred to a number of excellent textbooks. A comprehensive introduction to NMR has been published by Levitt (2008). A very detailed overview of the principles of MRI and the most common imaging techniques in medical imaging are described in Haacke et al. (1999). As in many plant MRI experiments a very high resolution is of essence; the book by Callaghan (1993) about MRI microscopy is of particular interest. An excellent and very detailed essay about radio frequency (RF) resonator design and construction can be found in Mispelter et al. (2006). And finally an open access internet book by Hornak (‘The basics of MRI’, is recommended.

An MRI system is designed to generate three different magnetic fields: the first (B0) is established by a large static magnet, the second (Gx, Gy, Gz) by a gradient coil set which generates three switchable spatially varying orthogonal magnetic fields, and the third (B1) by a RF resonator which provides a temporal varying magnetic field orthogonal to B0 (Figure 1a). To perform a MRI experiment the specimen to be examined is placed inside the strong magnet with the static magnetic field B0 always oriented in the z-direction. Besides permanent and electromagnets, superconducting magnets with field strengths up to 21 tesla (T) are used. Magnetic resonance imaging detects atoms having a non-zero nuclear magnetic moment (called ‘spin’). The most commonly used in vivo nuclei are 1H, 13C, 19F, 23Na, 31P and 39K (see Data S1). While nuclei with a spin of 3/2, for example, create additional possibilities for imaging, but play only a minor role, the following is concentrated on nuclei with a spin of 1/2. Furthermore, as living tissues have a high concentration of water and the majority of MRI images are images generated from protons within the water molecule, MRI of the protons within the water molecule will be discussed. This charged particle, the proton, has a spin 1/2 with an angular momentum. In the absence of a magnetic field the spins are oriented randomly, yielding an isotropic distribution. The net magnetization Mz a sum of all magnetic moments is zero. If a magnetic field B0 is applied the spins start to precess with frequency ω, called the Larmor frequency. The spin magnetic moment of each proton moves on a randomly oriented cone. The angle between the spin magnetic moment and B0 stays constant and the net magnetization Mz is still zero. Very small fluctuating magnetic fields resulting from the surrounding of the protons cause a slight change of the angle between the spin magnetic moment and B0. To reach the most energetically favourable state with the lowest magnetic energy the magnetic moments become slightly more oriented towards the external B0 field resulting in a non-zero net magnetization Mz. Once the lowest energetic state is reached the net magnetization Mz reaches its maximum value, called the equilibrium magnetization M0. The frequency of precession ω is proportional to the magnetic field B0, i.e. ω = γB0. The values of the nucleus-specific constant γ (also referred to as the gyromagnetic ratio) and the resulting ones of ω in a magnetic field of 17.6 T are given in Data S1.

Figure 1.

 Principles of magnetic resonance imaging (MRI).
(a) Experimental setup for MRI experiments. A carrot, as an example for the biological subject, is positioned in the centre of the MRI instrument. The lower part shows the orientations of the existing magnetic fields.
(b) Typical gradient echo pulse sequence applied during an MRI experiment.
(c, d) The MRI of a carrot generates first an image in Fourier (k-)space, which then has to be converted into image-space, showing the carrot segment analysed here.

The application of a linear polarized oscillating electromagnetic field (B1) with a duration of τB1 at the Larmor frequency perpendicular to B0 (referred to as an RF pulse), rotates the equilibrium magnetization M0 towards the xy-plane. The rotating xy-component (Mxy) of M0 induces a voltage in the RF resonator which is then recorded by the NMR spectrometer. Both the transmission of the B1 field and the detection of the NMR signal are achieved by either a single or by multiple RF resonators. To maximize the signal, these ‘antennae’ need to be fitted closely around the specimen. The combination of a large RF resonator with a small specimen would result in a reduced signal to noise ratio (SNR).

The decay of the signal after a single excitation pulse over time due to relaxation effects is referred to as free induction decay. The predominant relaxation effects relate to spin–spin relaxation (described by a relaxation time constant T2) and to variation in B0 caused by changes in magnetic susceptibility within the sample. The combined effects are described by the time constant T2* (T2* < T2). While there are ways to reverse the effect of variation in B0, spin–spin relaxation cannot be avoided. Spin-lattice or longitudinal relaxation is a further relaxation effect, which is characterized by the time constant T1. This process occurs either at the moment when the sample is first brought into the magnetic field, or after a RF pulse has rotated M0 towards the xy-plane. T1 describes the time required for the magnetization to re-establish its equilibrium value M0 along the z-axis.

Image generation and contrast

Thus far, the acquired signal has no spatial information. Unlike a light microscopy image which is acquired in image space, the MRI image is acquired in Fourier space (also referred to as k-space), and the actual image then has to be reconstructed via a multidimensional inverse Fourier transformation. The encoding of the MR signal is conducted by switching the spatially varying magnetic field gradients (Gx,y,z) in a certain manner as shown in the example of a gradient echo sequence (Haase et al., 1986) in Figure 1b. A more detailed description of the encoding procedure is given in Data S1. An illustration of real k-space data acquired from a slice through a carrot tap root is shown in Figure 1c, while Figure 1d shows the reconstructed image following a two-dimensional (2D) inverse fast Fourier transformation.

Unlike light microscopy images, MRI images are monochromatic, although colour coding can be added. Differences in grey-scale contrast reflect variation arising from three main sources. The one which the user cannot modify is the spin density within a given region of the specimen. Localized high water content, for example, results in an enhanced signal intensity. The other two sources are based on the relaxation times T1 and T2/T2*. Different tissue types within a plant can have different relaxation times. By adjusting the acquisition parameters of the pulse sequence a different contrast can be achieved (see Data S1).

Image resolution and imaging time

The voxel size of a 2D MRI image is defined by the slice thickness, and the ‘in-plane’ resolution Δx and Δy (see Data S1). Note that MRI is not limited to 2D imaging, as a genuine three-dimensional (3D) dataset can be acquired by applying an additional phase encoding (NPE2) in the slice selection direction. The imaging time Texp of a standard 3D MRI experiment is given by the product of NA (the number of times the experiment is repeated/averaged), NPE1 and NPE2 (the number of phase-encoding steps in two directions), and TR (the repetition time). Doubling NA does not double the SNR of the image; rather it increases it by a factor of √2 as the signal itself increases by a factor of two and the noise by √2. As the noise (called Johnson noise) is affected both by the temperature and the resistance of the RF resonator and the sample, the cooling of the RF resonator can be used to improve the SNR. Cryoprobes which cool the RF resonator but do not alter the temperature of the sensitive samples are commercially available, and have shown promising results in in vivo pre-clinical MRI ( As mentioned earlier, the signal from a voxel of a standard MRI experiment depends mainly on the spin density and the timing of the pulse sequence used. If the imaging parameters stay the same and the resolution is doubled in each dimension, the number of spins within a voxel will be reduced by a factor of eight. To achieve the same SNR as in the lower-resolution image the whole experiment has to be repeated 64 (82) times. Hence, experiments that took 1 h will take now almost 3 days, which is unacceptable for most experiments. To reduce scanning time, several rapid imaging techniques have been developed. A detailed description is given in Data S2.

The Trade-off between Physics and Plant Physiology

Magnetic resonance images which are informative in relation to structural and/or functional features can be created by various means (Köckenberger et al., 2004; Köckenberger and Granwehr, 2009; Van As et al., 2009). For most biologists, the complexities of the MRI methodology and NMR hardware may be hard to comprehend, while their biological background generally prevents them from making informed experimental decisions to fully exploit the potential of MRI technology. Standardized protocols, inasmuch as they exist at all, are geared to clinical practice. While the physicist aims to optimize the methodology by considering what combination of MR approaches is most likely to maximize the information content of the image and spectroscopic data, the physiologist’s requirements also need to be considered. The successful introduction of MRI into plant science therefore demands a close collaboration across two disciplines which are rather unfamiliar to one another.

The plant biologist used to seeing high-resolution optical-based images naturally expects the quality of MRI images to be at least as good. Such images are best generated when the water molecule is targeted, because its hydrogen nuclei provide a strong source of magnetization. Examples of fine-resolution 1H-MRI images obtained from plant material by this means include the spiral-shaped array of chloroplasts present inside of the cells of the alga Spirogyra (Ciobanu et al., 2002; Ciobanu and Pennington, 2004), for which a resolution of 3.7 × 3.3 × 3.3 μm3 was achieved. A second example relates to the geranium petiole, resolved to 2 × 2 × 50μm3 (Lee et al., 2001). Theoretically, using MRI (Glover and Mansfield, 2002) a resolution of approximately 1 μm could be reached, but that has not been, to the best of our knowledge, realized as of today.

Optimizing spatial resolution is a major objective, which implies the use of high magnetic and gradient field strengths (Lee et al., 2001; Ciobanu et al., 2002). The strongest currently commercially available MRI microscope has a magnetic field strength of 20 T (this corresponds to a 1H resonant frequency of 850 MHz, which is almost 14-fold stronger than a conventional clinical MRI scanner; see Figure S1). The trade-off is the orientation and the bore diameter of the instrument. Most plants are highly sensitive to the direction of gravity, so the orientation of the main magnetic field needs to be vertical (Chudek and Hunter, 1997). While a human system usually has a horizontal bore with a diameter of 60 cm, the diameter of a vertical bore of high field NMR systems (e.g. 20T) is 10 times smaller. Thus, the use of such a system is restricted to small specimens. The typical size of a ‘wide bore’ high-field system [the two leading manufacturers are Bruker Biospin ( and Agilent Technologies (] is bigger and reaches 8.9 cm in diameter. Plants of moderate size, such as Arabidopsis thaliana or part of plants (e.g. pods, leaves or seeds) are small enough to be subjected to high-field MRI. Despite imaging whole plants, specific regions can be targeted by the judicious placement of the RF resonators, a measure which simultaneously enhances sensitivity and resolution, and shortens the measurement time. The construction of these RF resonators can require technical innovation (Neuberger and Webb, 2009).

In some respects, plants are a more convenient experimental subject than are humans. A major advantage lies in the possibility of performing signal averaging over many hours, which is obviously inappropriate in the clinical setting where the length of time a patient can remain still is limited. This problem does not arise in plant specimens, which appear to be less sensitive to both the noise and strength of the magnetic field associated with MRI (Osuga and Tatsuoka, 1999; Paul et al., 2006), thus allowing both long-duration experiments and a wide range of short and intensive excitation pulses to be applied (Blümler et al., 2009); as a result, experiments can be focused on very high resolutions and/or quantifying the dynamic behaviour of tissues.

Further problems arise when long-term experiments are conducted as the plants need to be fully supplied with water, nutrients and light. To ensure this, customized controlled climate chambers have been devised (Van As, 2007). However, not much space is available for the insertion of a climate control device into the NMR system. A clever approach is the use of split coil magnets which provide opportunities to work with larger objects (Figure S1). While very high-resolution images are not available due to the low field strength, functional imaging with reasonable resolution delivers important results. Anyway, one has to consider that a high spatial resolution is not always needed, as is elaborated here and elsewhere (Köckenberger, 2001; Van As et al., 2009).

Magnetic Resonance Imaging for Outdoor Experiments

To apply MRI to plants growing in their natural environment, the NMR device needs to be transportable (Van As et al., 1994; Rokitta et al., 2000; Haishi et al., 2001; Wright et al., 2002; Goodson, 2006; Blümich et al., 2008). The NMR-MOUSE (mobile universal surface explorer) is an example of such a device. It is small and when placed on the surface of a specimen it allows the detection of the NMR signal from any relevant part of a plant within a few centimetres of its surface. Only the region covered by the magnet is investigated in more detail by the RF resonator. An application of such an instrument was demonstrated for monitoring leaf water dynamics spectroscopically (Capitani et al., 2009). The ‘C-shaped magnet’ generates a very homogeneous magnetic field (Rokitta et al., 2000; Wright et al., 2002; Utsuzawa et al., 2005) which is strong enough for imaging purposes. The suitability of this type of device for anatomical imaging has recently been demonstrated by in vivo monitoring of trees (Kimura et al., 2011; Umebayashi et al., 2011). Currently, however, such devices are relatively heavy to handle and/or their shape is insufficiently flexible (Halbach, 1980). A rather elegant solution to this problem is represented by cut-open force free NMR (NMR-CUFF; Figure S1), in which the magnet can be readily clamped (and later removed) around a tree trunk, a branch, a fruit or a plant stem (Raich and Blümler, 2004; Windt et al., 2011). The prototype device weighed just 3.1 kg, and was able to provide a flux density of 0.57 T over a 5-mm diameter sphere. The level of resolution obtained by NMR-CUFF remains limited, but this restriction is likely to be lifted by ongoing technical improvements in magnet design, which have seen the field strength and homogeneity generated by NMR-CUFF reach comparable levels typical for clinical fixed MRI device supplied 20 years ago.

Imaging of Plant Development

A major thrust of developmental biology is to understand how molecular and cellular processes produce 3D morphology. With its non-invasive character MRI is, unlike other imaging techniques, capable of gaining information with high spatial resolution, both structural and biochemical, as well as on temporal changes within the plant, and can therefore be used to monitor plant development processes.

Seed and bulb germination

One of the fundamental plant processes particularly amenable to MRI analysis is germination, which begins with the uptake of water into the seed (imbibition). As MRI does not require tissue transparency for image acquisition, it can be used to non-invasively trace the fate of imbibed water in seeds, and thereby identify which tissues are involved in water distribution. The germination process has been studied by MRI in most leading crop species (wheat: Rathjen et al., 2009; maize: Ruan and Litchfeld, 1992; legumes: Wojtyla et al., 2006; Garnczarska et al., 2007a; barley: Molina-Cano et al., 2002) and other species including trees (Köckenberger et al., 2004; Roh et al., 2004; Terskikh et al., 2005; Kikuchi et al., 2006). Tobacco seeds are as small as 1 mm in diameter, and could be imaged in vivo during their imbibition and germination by Manz et al. (2005) at a level of resolution sufficiently high to visualize the tissue-specific water penetration pathway and to characterize the dynamics of water uptake. In beans, an unexpected mechanical vibration of the seed was observed during the imbibition process (Kikuchi et al., 2006), and some of the regulatory mechanisms controlling the uptake of water were revealed (Koizumi et al., 2008). A future task will be the integration of in vivo MRI data with those on the complex gene regulatory and metabolic networks controlling seed germination (Bentsink et al., 2010). Germination is an important crop trait and application of non-invasive techniques, especially such as MRI, have the potential to facilitate crop improvement by contributing to both experimental and agricultural practice (e.g. evaluation of seed composition, quality, screening procedures and others). Monitoring water uptake has relevance for the food industry/biotechnology. As a result of the information gained from the NMR images, maltsters can improve the efficiency of the malting process (Horigane et al., 2006).

Some plant species have evolved the capacity to form storage bulbs as a vehicle for vegetative propagation, and their formation can be recognized quite early by the onset of certain changes in the internal structure of the relevant part of the plant. The non-invasiveness of MRI allows for the visualization of these changes in a way that is impossible to achieve non-destructively using conventional microscopy (Faust et al., 1997; Ishida et al., 2000; Ratcliffe et al., 2001). The higher free water content of actively developing organs within the bulb (inflorescence, florets, leaves) will result in a hyperintense signal in the MRI images (Van der Toorn et al., 2000). Structural, physiological and metabolic changes taking place inside the bulb can also be monitored (Robinson et al., 2000; Van der Toorn et al., 2000; Roh et al., 2004). Magnetic resonance imaging of Lachenalia aloides bulbs revealed the effect of elevated temperature on various internal storage processes (Roh, 2005). Bud development and dormancy induction in woody plants is also associated with water content and mobility. Magnetic resonance imaging can be used to examining the behaviour of water in buds and contribute to investigations on the mechanisms underlying the adaptation of plants to environmental and climate changes (Tanino et al., 2010; Kalcsits et al., 2009; Yooyongwech et al., 2008).

Seed development

Developing seeds are valuable targets for MRI, and various approaches have been used to characterize this phase of the plant life cycle. Glidewell (2006) used MRI to study developing barley grains from anthesis to maturity, generating 3D images of caryopses as well as quantitative T2 maps (see Data S1). Chemical shift imaging (CSI; see Data S1) was applied to detect changes in the tissue distribution of water, soluble carbohydrates and lipids. Further developments of the method elucidated quantitative lipid maps (Neuberger et al., 2008). Gruwel et al. (2008) applied diffusion-weighted MRI on wheat grain and endosperm pore size. Furthermore, the embryo cell dimensions could be obtained. Ishimaru et al. (2009) used MRI to explain the formation of distinct phenotypes of rice grains grown under different temperature conditions. MRI was combined with physiological measurements, laser microdissection and expression analysis. Garnczarska et al. (2007b) used MRI to study water content/distribution during maturation of lupin seeds, elucidating the spatial/temporal relationship to dehydrin proteins. Melkus et al. (2009) modelled the 3D structure of developing pea seeds, quantifying the volume ratio of different seed organs including the tiny suspensor. Magnetic resonance imaging was linked to NMR spectroscopy and allowed quantification of local concentrations of metabolites in different regions of the seed (Figure 2, Video clip S1). Hayden et al. (2011) applied an MRI approach for the integrative study of seed development in two oat cultivars, combining lipid mapping, metabolite and transcript profiling.

Figure 2.

 Non-invasive study of seed structure and metabolite distribution in pea during early developmental stages.
(a) Hand section through the pod and seeds showing seeds filled with liquid endosperm.
(b) Three-dimensional NMR-based model of pea seed.
(c) Selected longitudinal section from NMR three-dimensional dataset used for modelling.
(d, e) Distribution of sucrose (in d) within the embryo sac showing elevated levels in endosperm versus suspensor and gradient distribution of sucrose in the embryo; sucrose concentration is colour-coded. Image in (e) represents the reference image (cross-section) showing the seed coat, embryo, endosepermal vacuole and suspensor. For more information, see Melkus et al. (2009).
Abbreviations: e, embryo; ev, endospermal vacuole; sc, seed coat; s, suspensor.

Fruit growth

Glidewell et al. (1999) monitored the whole developmental process of blackcurrant (Ribes nigrum) fruits attached to the plant. Their use of various gradient echo imaging sequences, chemical shift effects, etc. and both 2D and 3D reconstructions allowed for a correlation between NMR signal intensities and specific tissue features, such as cell size, air inclusions and lipid content. The quantification of fruit composition in oil palm carried out by Shaarani et al. (2010) identified a tissue-specific pattern of oil and water distribution. Windt et al. (2009) were able to demonstrate that most of the water translocated into the tomato fruit travels through the xylem and not the phloem, thereby resolving a long-standing difficulty in modelling fruit growth. Magnetic resonance imaging has also found applications in the study of certain parameters of fruit quality (Chudek and Hunter, 1997; Musse et al., 2009; Haishi et al., 2011).

Root growth

Most recent improvements in MRI technology have enabled the investigation of root development. Magnetic resonance imaging can visualize the 3D geometry of roots not only in liquid or clear media but also inside soil or sand (Kaufmann et al., 2009; Blossfeld et al., 2011; Hillnhütter et al., 2011). This allows access to the ‘hidden’ root architecture and how it relates to local soil composition, environmental and biotic factors.

Imaging of Water Dynamics in Living Plants

The distribution of water, nutrients and regulatory compounds in plants relies on the functions of the vascular system. This system, consisting of phloem and xylem, is deeply embedded in plant tissues, thus any functional investigation becomes a technical challenge. To explain the driving forces of solute movement within the vascular system two theories have been proposed: the cohesion–tension (CT) theory (Dixon and Joly, 1894) to explain xylem transport and the pressure-flow hypothesis (Münch, 1930) to explain phloem transport. Their validity has remained a matter of debate over decades.

The development of non-invasive NMR based technologies created a basis for in vivo study of xylem and phloem transport in living plants (Köckenberger, 2001). Pioneering flow MRI by Köckenberger et al. (1997) was followed by the development of fast imaging techniques such as fast low-angle shot (FLASH; Rokitta et al., 1999; Peuke et al., 2001) and q-space imaging (Scheenen et al., 2007). Dedicated hardware was developed and strategies to visualize and quantify dynamics of the plant vascular system in wide ranges of plant species were proposed (Windt et al., 2006; Van As, 2007; Van As et al., 2009). Currently, xylem and phloem flow and their mutual interactions is one of the most popular subjects for MRI (Hölttäet al., 2006; Van As, 2007; Windt et al., 2009). A particular contribution of MRI has been the in vivo characterization of fluxes in xylem/phloem (Figure 3), the quantification of the diurnal pattern of solute flow, monitoring of embolism repair and the defining of certain structure–function relationships, such as between sieve tube geometry and phloem flow (Peuke et al., 2001; Salleo et al., 2004; Kaufmann et al., 2009; Mullendore et al., 2010). Examples of important in vivo observations include the facts that water flow through the plant as a whole responds both to the nitrogen source and root/stem cooling (Scheenen et al., 2001; Peuke et al., 2006, Schulze-Till et al., 2009). Takase et al. (2011) exploited MRI to relate the solute flow to the expression of aquaporin genes in A. thaliana. The growing consensus is that the plant vascular network, far from being a passive plumbing net, is in fact a finely regulated transport system.

Figure 3.

 High resolution magnetic resonance imaging (MRI) demonstrating water flow dynamics in tomato.
(a) Tomato truss; arrow shows the peduncle connecting tomato fruits to the stem.
(b) Microscopy image (light microscopy) of the perimedullary tissue showing localization of phloem (ph) and xylem (x).
(c) Volume flow map of influx and efflux in the peduncle before truss pruning. Influx in the outer ring is shown in blue and efflux in red. Influx in the inner ring is shown in green. The influx in the inner ring corresponds with the position of the perimedullary tissue.
(d) High-resolution colour-coded quantitative volume flow map (see colour bars on the top panel). Image courtesy of Dr C. Windt, Forschungszentrum Jülich GmbH, Germany. For more information, see Windt et al. (2009).

Functional Imaging of the Abiotic Stress Response

The geographical dispersion of a plant species is known to be greatly affected by the frequency with which abiotic stresses, in particular drought, salinity, cold or heat, are experienced (Boyer, 1982; Araus et al., 2002). The common link between all these stresses is that at least some of their detrimental effect is caused by a disruption to the plant’s moisture status (Verslues et al., 2006).

Drought stress

Magnetic resonance imaging was attempted in 1986 on intact drought-stressed Vicia faba plants (Bottomley et al., 1986) and Pelargonium spp. (Brown et al., 1986), but it has only recently become possible to revisit these pioneering experiments using current MRI equipment and methodology (Van der Weerd, 2002a; Scheenen et al., 2007). While conventional physiological experiments have tended to focus on the response of particular organs, 1H-MRI is well suited to the study of stress responses more holistically. In a MRI-based comparison between maize and pearl millet (Pennisetum glaucum), Van der Weerd et al. (2001) demonstrated differences in the drought response of the two species.

The stem is the logical location for an MRI probe, because it connects the root with the leaf, and its tissue architecture is highly suited to the acquisition of high-resolution images (Van der Weerd et al., 2002b). Xylem embolism (where solute flow is blocked by an air inclusion) is one of the early effects of drought stress. Various hypotheses have been proposed to explain how such embolisms can be corrected (e.g. Salleo et al., 2004), but so far none has been fully validated. Magnetic resonance imaging provided the first direct observations of xylem cavitation and embolism repair in an intact plant (Holbrook et al., 2001), experiments which were later extended to a wide range of species (Clearwater and Clark, 2003; Scheenen et al., 2007; Kaufmann et al., 2009).

The root is increasingly recognized as a key player in the adaptation of plants to drought stress (Pennisi, 2008; Lopes et al., 2011). Magnetic resonance imaging can generate images of roots in soil, modelling their structure, monitoring moisture changes in the rhizosphere and carrying out functional studies of plant nutrition (Pohlmeier et al., 2008; Blossfeld et al., 2011; Hillnhütter et al., 2011). Spin density MRI analyses of drought-stressed maize roots have successfully localized cavitation events and allowed the visualization of the refilling process, shedding light on the identity of certain in vivo processes underlying drought tolerance (Kaufmann et al., 2009).

The response of plants to field drought is more complex then that induced in controlled experiments, in which care is taken to ensure that drought is the sole stress being imposed. Magnetic resonance imaging experiments carried out in the field allow us to monitor ‘integrated’ plant responses instantly at the time and place they arise (Capitani et al., 2009; Windt et al., 2011). Quercus ilex leaves have been used to monitor what changes occur in vivo over the course of progressive drought (Sardans et al., 2010). Here, measurement of the water content in the plate and reap of the leaf indicated a non-homogeneous response to stress. This information suggests how gene expression studies could be based on topographical information. Progress in this direction may deliver the use of MRI as a diagnostic tool to help the scheduling of irrigation. It could also be developed into a crop breeders’ selection tool for identifying genetically superior individuals with respect to water use efficiency.

Some plant species have evolved a very high level of drought resistance (Schneider et al., 2003; Liu et al., 2007), and MRI might help to unravel the underlying mechanisms. The African resurrection plant, Myrothamnus flabellifolia, is able to switch between a highly desiccated state and a fully hydrated green plant within 24 h of watering (Figure 4). Lipid composition, water movement within its shoots and leaves during drying and rehydration episodes have been visualized by Schneider et al. (2003) using MRI. This analysis provided evidence that the key transport tissues are equipped with lipids, and that the spatial arrangement of the xylem enables repeated cycles of hydration and dehydration in an organized manner.

Figure 4.

 Comparison of light and magnetic resonance (MR) microscopic imaging for the study of rehydration in the African resurrection plant Myrothamnus flabellifolia.
(a) The plant before and after watering, demonstrating the rehydration potential.
(b) Three-dimensional reconstruction measured on a dry intact Myrothamnus branch at a height of 5 cm.
(c) Light microscopy image of a cross-section of an air-dry Myrothamnus branch stained with the lipophilic dye Nile red. Yellow fluorescence indicates lipids (p, pith; lta, lipid-rich tracheid assemblies; lt, leaf trace).
(d) High-resolution 1H-NMR lipid distribution images of an air-dry Myrothamnus branch (pp, pith periphery; lp, lipid pieces).
(e) 1H-NMR imaging visualizes non-invasively the spreading of water (bright areas) within a virtual cross-section of an air-dry branch during rehydration. For further details see Schneider et al. (2003).

Cold stress

Episodes of low-temperature stress challenge plants in a multitude of ways, and the responses should be considered as a syndrome rather than as a single reaction (Beck et al., 2007).

Prolonged exposure to near freezing temperatures can cause functional changes and tissue damage. Continuous MRI-based monitoring over a 240-h period allowed a detailed study to be made of the in vivo response to cold stress of woody lianas (Clearwater and Clark, 2003). Scheenen et al. (2002) used a combination of the imaging of flow and T2 to analyse the effect of cooling the roots on the water status of intact cucumber plants. Their major findings were that cooling induced a substantial decrease in water uptake, due to a root response and embolisms in the xylem (Scheenen et al., 2007). These authors also reported the restoration of functioning xylem, as observed by flow, not only filling of vessels. Cooling the stem of Ricinus communis caused both the leaching of sucrose from the stem phloem vessels and the short-term inhibition of mass flow at the beginning of cold treatment (Peuke et al., 2006).

An early consequence of cold stress caused by exposure to sub-zero temperatures is the dehydration of tissues due to the freezing of water, and the subsequent damage to membranes upon thawing (Yamazaki et al., 2008; Yadav, 2009). 1H-MRI is well suited to detect how the plant cell copes with the nucleation and expansion of ice (Ishikawa et al., 1997; Ide et al., 1998), because it can monitor the development of hardening and de-hardening, which is difficult to achieve using a destructive assay. It also allows for monitoring over prolonged periods of exposure in the natural environment. As long ago as 1995, MRI was used to investigate freezing tolerance in wheat (Millard et al., 1995). The MRI analysis was able to identify clear behavioural differences between acclimated and non-acclimated plants. It was even possible to identify the most cold-sensitive part of the plant. Non-invasive approaches are also advantageous for forest species. The flower buds of hibernating trees differ in their susceptibility to freezing damage. Methods have been developed based on MRI to determine the response to sub-zero temperatures of various tree species (Ishikawa et al., 1997; Price et al., 1997; Ide et al., 1998). Comparisons of MRI images obtained from trees exposed to freezing conditions have revealed that Acer japonicum flower buds, leaves and stem bark tissue had become frozen when the temperature fell to −7°C, but the lateral primordia retained their viability down to −40°C. This strategy (harmonized freezing) might help to ensure the survival of trees. In Arabidopsis, freezing tolerance is associated with the avoidance of damage to the plasma membrane and/or membrane repair (Yamazaki et al., 2008). A better understanding of the mechanisms underlying freezing tolerance at the whole plant level will require a determination of the relationships between ice management, cell wall properties and membrane resealing, and this may be more easily achieved by incorporating MRI analysis with more conventional biological approaches.

The Host–pathogen Interaction

Plants are challenged by a range of viral, bacterial and fungal infections, and an intimate and dynamic means of monitoring the host–pathogen interaction would greatly enhance our understanding of the infection process. Although the damage caused by infection is often readily visible, it can sometimes remain hidden within the plant. A good example of the use of MRI in a plant pathology context has been given by an analysis of diseased sycamore (Acer pseudoplatanus) trees, in which both the various infection pathways exploited by diverse pathogens could be well defined and the effect of disease on the water status of the wood monitored (Pearce et al., 1994). MacFall et al. (1994) imaged gall formation in pine seedlings following their infection by the bacterium Chronartium quercuum. A further example is described by Goodman et al. (1996) regarding the fungus Botrytis cinerea, where 3D MRI datasets were used to successfully define the location of infected regions inside diseased strawberry fruit. It appeared that the pathogen breaks down parenchymal cell walls, with the result that cell contents leaked into the intercellular spaces (Goodman et al., 1996; Chudek and Hunter, 1997). The application of MRI contributed to the understanding of Pierce’s disease as well, which was a major problem in some Californian vineyards during the early 1990s. It had been assumed that the sole pathogen involved was the bacterium Xylella fastidiosa which colonized the xylem, and thus compromised water translocation throughout the plant. However, MRI analysis showed that the blockage of the xylem resulted from the host’s active responses to infection rather than from the proliferation of the pathogen itself (Alonso et al., 2007).

Jatropha curcas, a potential source of biodiesel, is generally regarded as being a hardy, drought- and disease-resistant plant, but it can be heavily damaged by the whitefly-borne Jatropha mosaic virus. Sidhu et al. (2010) have recently demonstrated the value of MRI and high-resolution magic angle spinning NMR spectroscopy for studying this viral infection. The contrast of T1 and T2 weighted images detected differences in the spatial distribution of water, lipids and macromolecules in infected versus healthy stems. Alterations in certain anatomical structures and in the rate of sap translocation could then be correlated with metabolic changes in infected plants.

Pine wilt disease is characterized by the formation of embolized tracheids following the invasion into the resin canal of the pine wood nematode Bursaphelenchus xylophilus. Infected trees eventually die as a result of compromised xylem conductivity. Umebayashi et al. (2011) devised a compact MRI system featuring a C-shaped magnet and a movable U-shaped RF coil, which allowed the trunk’s internal structure to be imaged at a high level of resolution. The dynamics of disease spread and the resulting damage could then be precisely documented. These experiments suggest a future place for MRI-based devices in the early diagnosis of some tree diseases.

Current developments in MRI have allowed the non-invasive detection of below-ground symptoms in sugar beet caused by the beet cyst nematode and/or soil-borne root rot. Magnetic resonance imaging monitored a synergistic relationship between the two pathogens, providing new insight into plant–pathogens interactions (Hillnhütter et al., 2011).

Sustaining Biodiversity

Biodiversity is threatened by a combination of over-exploitation, pollution and climate change, raising the priority of conserving plant genetic resources. Seed storage under low temperatures represents an efficient means of preserving many flowering plant species, while some non-seed tissues (e.g. tubers, bulbs, meristems) can be cryopreserved. The longevity of seeds is an issue in all germplasm banks (Nagel and Börner, 2010), and the lack of non-destructive methods to assess seed viability means that seed numbers inevitably become depleted over time, forcing stocks to require regeneration on a regular basis. The ability of MRI to provide a non-invasive assessment of the integrity of a seed’s internal structure, to detect the presence of internal pathogens (Köckenberger et al., 2004), to visualize the distribution of lipids or water within a seed (Ishida et al., 2004; Neuberger et al., 2008) and to monitor the physical state of moisture within a seed as a result of storage at low temperatures (Borompichaichartkul et al., 2005) are all highly relevant for developing an efficient germplasm conservation strategy. Systematic comparisons of the structure and composition of freshly harvested versus stored seeds (possibly augmented by artificial seed ageing measures) could succeed in defining what parameters are associated with seed viability (Gruwel et al., 2002; Borisjuk et al., 2011). Scaling MRI techniques appropriately and developing a cost-effective hardware platform will be needed to promote the application of MRI in this area. It should be noted that electron paramagnetic resonance (EPR) in combination with the use of spin probes offers an alternative mean for non-invasive observation of seed viability and longevity (Golovina et al., 2010).

Most parts of the plant cannot be maintained intact over the long term, and are substantially altered during fixation or cryopreservation procedures. The creation of a virtual library providing 3D models of these materials based on MRI data of living plants could enable a indispensable digital collection of a mass of biodiversity information and make it accessible for future generations of scientists. Efforts are under way to develop appropriate hardware, software and methodology.

Gene Expression and Function

Prospects for employing MRI reporter genes

Currently exploited reporter genes, such as those encoding β-glucuronidase, luciferase or GFP, are based on histochemical staining or fluorescence. Optical projection tomography has extended the resolution of these reporters in plant material to three dimensions within a single cell, or in some cases within tissue sections with a thickness up to 15 mm (Lee et al., 2006; Truernit et al., 2008). The current peak resolution achieved in animal material is represented by the transgenic ‘brainbow’ mouse, in which the simultaneous expression of multiple fluorescent proteins has resulted in the recognition of some 90 distinguishable colours (Livet et al., 2007). Two new promising classes of reporter genes are now emerging, one of which relies on affinity for specific radioisotopes (Serganova et al., 2007) and the other on MRI (Gilad et al., 2008). A particular feature of MRI reporter genes is that in principle they can combine gene expression data with anatomical and functional information. In the most advanced of these, the reporter gene product interacts with a reagent containing the element gadolinium (Gd) (Gilad et al., 2008). The Gd enters the root cell symplast, moves in conjunction with the flow of solutes and can be well traced in plants (Gussoni et al., 2001; Zhang et al., 2009). Gadolinium is non-toxic for plants, both in its chelated and unchelated forms (Quiquampoix et al., 1990), but its membrane permeability needs to be considered. Another opportunity is provided by the Escherichia coli gene encoding polyphosphate kinase (PPK) (Ki et al., 2007). Polyphosphate kinase does not require an exogenously supplied substrate and can be visualized by 31P-MRI. The enzyme catalyses the synthesis of inorganic (largely immobile) polyphosphate from ATP, and has been expressed constitutively in plants (Van Voorthuysen et al., 2000; Nagata et al., 2006). A disadvantage of this system is the low sensitivity of 31P-NMR. A further option is the use of iron-based reporter genes (Hill et al., 2011), which are associated with good contrast in 1H-MRI. The heterologous expression of ferritin genes has been achieved in a number of plant species, but aspects related to the accumulation of iron and its complex regulation complicate the picture (Van Wuytswinkel et al., 1999; Drakakaki et al., 2000; Jiang et al., 2006). Finally, the switchable chemical exchange saturation transfer (CEST)-based reporter genes (Liu et al., 2011) have the feature that they are able to simultaneously visualize more than one target. As yet MRI reporter genes have not been developed for plant material, but it is likely that they will be in the future.

Bridging the gap between gene expression and function

The non-invasive monitoring of plant processes in vivo offers the potential to establish relationships between gene expression and physiological events, which can help in the elucidation of gene function. The role of aquaporins in maintaining plant moisture status, water hydraulics and stress tolerance has been controversial for some time (Katsuhara et al., 2008), but their visualization using 1H-MRI has resolved much of the argument. When Takase et al. (2011) monitored the behaviour of water in the A. thaliana root, a diurnal pattern of water content was observed in the basal zone of the root, and this rhythm was maintained even when the plants were kept under continuous light or darkness. Imaging data were compared with the expression profiles of two aquaporin-encoding genes, known to control water uptake (Chaumon et al., 2005) and whose expression followed a circadian rhythm under continuous light. The circadian oscillation in water dynamics was abolished in a mutant compromised for the detection of the circadian signal (Liu et al., 2001). Thus the inclusion of MRI data allowed a linkage between function (water dynamics) and gene expression. The wider use of MRI for this sort of research can be expected to yield many novel insights into gene function (Yooyongwech et al., 2008).

Another example is the Jekyll-gene in barley which has been shown to have a role in sexual reproduction (Radchuk et al., 2006). The localized up-regulation of Jekyll appears to be coupled with cell autolysis in the developing grain, while its down-regulation slows the growth of the endosperm. On this basis, it was suggested that the function of JEKYLL is associated with the allocation of nutrients between maternal (pericarp) and filial (endosperm) tissue. Later, 13C/1H-MRI was applied to visualize allocation of 13C sucrose in plants engineered to repress Jekyll expression to various extents (Melkus et al., 2011). These experiments showed that the quantity and distribution of sucrose were dependent on the degree of Jekyll repression, approving the role of JEKYLL in nutrient allocation during the process of grain filling.

The analysis of mutant plants

Mutants have proven invaluable for defining gene function, but not uncommonly their primary effect is concealed by pleiotropy. In such cases, non-destructive methods may be required to identify the primary effect of the mutation, presenting an opportunity for MRI, based on its ability to simultaneously monitor a range of structural, metabolic and physiological parameters. This type of analysis is rare in the plant world as yet, and MRI technology is still challenging when applied to small targets such as seeds of A. thaliana. Fortunately, the novel model plant species (rapeseed, rice, maize, etc.) should be more amenable to MRI.

Fast Seefeldt et al. (2007) were able to use 1H-NMR imaging to both identify and characterize β-glucan (BG) mutants in barley. The presence in food of BG lowers both its cholesterol content and glycaemic index. Magnetic resonance imaging was proved to be effective for delineating the internal structure of the grain, and for identifying varietal differences in the grain’s water-holding capacity. Another use of MRI was to characterize a pea mutant (Borisjuk et al., 2002), as part of a wider attempt to understand the role of the liquid endosperm. Applied to the seed of a mutant which develops a giant endosperm, MRI was able to determine non-destructively 3D structures and the volume of each of the seed’s component organs (Melkus et al., 2009). Both the concentration and the distribution inside the liquid endosperm of some major metabolites were obtained in vivo. The endosperm is the major seed storage organ in monocot crop species, and NMR spectroscopy has been widely applied to help understand the metabolism of the endosperm and its regulation (Alonso et al., 2011). Linking such efforts with MRI should accelerate progress in this field.

Imaging of Plant Metabolism

The study of plant metabolism and its compartmentalization provides a number of potential MRI applications, since the technology offers the non-invasive measurement of various metabolite concentrations (Bourgeois et al., 1991; Soher et al., 1996; Vanhamme et al., 1997; Tkác et al., 1999; De Graaf, 2007). In order to be informative for the biologist, MRI data have to be related to known histological, biochemical and other characteristics of the tissue, and this represents an area where substantial progress has been achieved in recent years – in particular, in the imaging of the commonest assimilates exported into and distributed within the developing seed, and in the quantification of seed storage compounds.

Visualization of lipid storage and degradation

Regulation of oil storage activity in vivo is complex and requires non-invasive approaches. Various NMR-based methods have been used for lipid detection both in dry plant material and in oil-rich fruits/seeds (reviewed in Neuberger et al., 2008). When CSI was employed as a non-invasive means of visualizing lipid distribution in the mature soybean seed, clear lipid gradients were observable, in accordance with the differentiation pattern of the plastids, which are the site of fatty acid synthesis (Borisjuk et al., 2005). A disadvantage of CSI is its relatively long experiment time. Hence, it is only of limited use for delivering a reliable picture of events within a developing seed. In a more recently developed approach, reliance was placed on the slightly different resonance frequencies of water and lipids, which could be exploited using a frequency-selective MRI technique (Neuberger et al., 2008). This method shortened the measurement time up to 10-fold, and delivered a spatial resolution close to the cellular level.

The simultaneous imaging of anatomy and lipid deposition offers the opportunity to relate lipid accumulation with seed development. Using this approach in the developing barley grain revealed concentrated lipid deposition in particular regions of the embryo (scutellum and nodule), as well as in the aleurone layer of endosperm, a structure which is only a few cell layers thick (Figure 5, Video clip S2). At the same time, the regions where lipid degradation occurs later in the maturation process were identifiable. In high-oil cultivars of oat, lipid occupies the entire endosperm as demonstrated by the MRI-based analysis (Figure 6, Video clip S2). To date, this mode of lipid mapping has been applied to seeds of oat (Hayden et al., 2011), oilseed rape and barley (Neuberger et al., 2009) as well as tobacco, maize, wheat, Jatropha, pine, cotton, linseed and sacred lotus (our own unpublished data). Combining oil topology with the analysis of gene expression and metabolites has the potential to identify key factors in the regulation of lipid metabolism in vivo (Hayden et al., 2011) and is expected to provide novel insights into the control of storage in crops. In the future one can anticipate an equivalent approach being taken to study the fate of storage lipids during germination.

Figure 5.

 Quantitative imaging of lipid in a living barley grain. (a) Fragment of a barley spike used for the magnetic resonance imaging (MRI) analysis. (b) Longitudinal tissue section showing the internal structure of the grain. (c) Lipid staining in a longitudinal tissue section using Sudan/ethanol procedure (lipids stained in red). (d) An MRI based three-dimensional model of the spike shown in (A) (see also Video clips S2 and S3 (e) Non invasive visualization of the spike demonstrating the internal structure of grains/spike with resolution of 35 μm. (f) Quantitative map representing lipid deposition within the grain in vivo; lipids are mainly found in the embryo and the aleurone layer; lipid content is colour coded. Abbreviations: al, aleurone layer; em, embryo; en, endosperm; np, nucellar projection; p, pericarp. For further details see Neuberger et al. (2008).

Figure 6.

 Quantitative imaging of lipid in a living oat grain.
(a) Oat grain pictured using a light microscope.
(b) Cross-tissue section showing the endosperm and pericarp.
(c, d) Quantitative map representing lipid deposition within the grain in vivo (corresponding to the cross-section shown in (b)) in the low-oil cultivar Freja (c) and the high-oil cultivar Matilda (d). Lipid content is color-coded.
Abbreviations: al, aleurone layer; en, endosperm; em, embryo; p, pericarp. For further details see Hayden et al. (2011).

Visualization of metabolite distribution

A further focus of MRI relates to the imaging of individual metabolites, giving information on their distribution, transport and conversion within the cell (Ratcliffe et al., 2001; Köckenberger et al., 2004). As the particular composition and architecture of plant tissues reduces the sensitivity of MRI, only abundant metabolites such as sucrose (Verscht et al., 1998; Szimtenings et al., 2003) and free amino acids have been successfully targeted to date. Nevertheless, the non-invasiveness of MRI has provided a number of analytical opportunities, which are unobtainable by destructive sampling which induces the wounding response. Chemical shift imaging has only a minimal requirement for post-processing correction, and the acquisition and processing procedure tends to be relatively simple and robust, because only a single pulse and phase-encoding gradient are needed for signal encoding. An example is provided by the use of 1H-NMR CSI to image metabolite distribution in intact pea seeds at various stages of their development (Melkus et al., 2009). Structural FLASH (Haase et al., 1986) multi-slice images were acquired at the end of the CSI protocol in order to topographically relate the spectroscopic data with the corresponding tissue structures. As a result, it was apparent that the spatial distribution of sucrose (as well as of glutamine and alanine) within the endosperm vacuole tends to be rather uniform, but at the same time is notably different from that in either the suspensor or the cellularized embryo (Figure 2c,d). The sucrose concentration gradient was somewhat different from that of the free amino acids, and was in accordance with the expression pattern of genes encoding metabolite transporters. At the same time it was possible to demonstrate how endosperm metabolite levels respond both to the onset of storage activity in the embryo and to specific environmental cues, and to identify the endosperm glutamine concentration as representing a limiting factor for protein storage in the legume embryo. Improving the level of sensitivity obtainable from small seeds will need some modification of currently available RF resonators (Neuberger and Webb, 2009). Another MRI application for metabolite imaging is the study by Wenzler et al. (2008) of carbohydrate metabolism involved in forming floral nectar (Anigozanthos flavidus). These authors combined cyclic J cross-polarization and 1H spin-echo imaging (a technique implemented by Heidenreich et al., 1998) to show the localization of 13C-labelled glucopyranose and the glucose moiety of sucrose inside the peduncle during a 13C-feeding experiment.

Dynamic imaging of metabolites

Dynamic NMR protocols (or ‘functional imaging’) can be used for applications beyond the reach of current MRI, such as attempts to monitor the transport and conversion of major metabolites. Flow-encoded NMR measurements are effective where velocities are measured in mm h−1 (Szimtenings et al., 2003; Van As, 2007). However, they are difficult to perform when velocities lie in the μm h−1 range. The detection, imaging and quantification of sucrose can be achieved by using 1H-NMR to target protons associated with carbon nuclei (Tse et al., 1996; Melkus et al., 2009). The advantage of using the 1H signal (instead of 13C) is its high MRI sensitivity. It is impossible, though, to follow certain sucrose molecules through the plant, and only steady-state levels are observed. As the natural abundance of 13C is very low and the dominant 12C isotope is not visible by NMR, a combination of 13C-NMR and the feeding of 13C-labelled substrates to the plant can be used to track the 13C-labelled metabolites on their way through the plant. By combining NMR spectroscopy and imaging, it is possible to obtain both metabolic and spatial information regarding 13C-enriched molecules and their metabolic derivatives from the same experiment. Various inverse detection schemes have been developed to further improve the detection sensitivity of the 13C nucleus (Bax et al., 1983; Rothman et al., 1992). These pulse sequences provide a range of flexible strategies for the detection of 13C nuclei and are in general more sensitive than direct detection of 13C (Heidenreich et al., 1998; De Graaf et al., 2003).

A recent example of dynamic NMR is given by Melkus et al. (2011), tracking the allocation of assimilates in barley seeds. A tool has been developed to not only detect specific metabolites, but also to produce an adequate level of spatial and temporal resolution over the course of a prolonged period of monitoring. In this approach the gradient enhanced heteronuclear multiple quantum coherence (geHMQC; Hurd and John, 1991) sequence was applied using a double-tuned RF resonator and a high magnetic field strength. The metabolic images were captured either via direct or inverse 13C detection schemes following 13C feeding. These results demonstrated for the first time how sucrose diffuses in vivo inside a developing cereal grain (Figure 7, Video clip S3). The cellular pathways were identified at a sub-millimetre level and the tissue-specific velocity of sucrose allocation was determined. 13C/1H-NMR delivered a five fold higher in-plane resolution than PET (Jahnke et al., 2009), and facilitated dynamic observations. Furthermore, in contrast to MR, PET lacks the information concerning from which specific molecule a decaying 11C nucleus has originated. The 13C/1H-NMR method allowed for the straightforward co-registration of the structural and the metabolite images, and therefore the exact identification and localization of metabolites within a tissue. In the barley caryopsis, the nucellar projection has been shown to represent the exclusive gateway for sucrose inflow, and possesses several structural, metabolic and gene expression features enabling this function (Melkus et al., 2011). Further applications in other major crops can be expected to identify bottlenecks in the supply of photo-assimilate to sink organs such as the seed, thereby providing novel targets for the molecular (biotechnological) modification of crop species.

Figure 7.

 Monitoring of 13C sucrose allocation during onset of seed filling in barley.
(a) The uptake of 13C in the barley caryopsis occurs by feeding 100 mm13C sucrose to the stem (left panel). The red cage shows the position of the double-resonant 13C/1H-NMR coil.
(b) Visualization of 13C sucrose allocation within the caryopsis (see also Video clip S3). The time post the start of incubation is indicated. For details see Melkus et al. (2011).

Dynamic MRI, metabolic modelling and systems biology

Systems biology is a holistic approach, describing the complex interactions in biological systems. Magnetic resonance imaging can substantially contribute to such an approach because it considers the plant’s complexity: each organ comprises distinct cell and tissue types, each of which may be governed by a distinct metabolic network which all interact with each other. This compartmentalization needs to be considered when analyzing the regulation and control of plant metabolism in vivo (Sweetlove and Ratcliffe, 2011). An example of the use of MRI for the analysis of metabolic compartmentalization is represented by an analysis of the barley endosperm, which was assumed a priori to be metabolically homogeneous. The use of a geHMQC sequence enabled the detection of 13C alanine, derived by supplying 13C sucrose to the plant (Rolletschek et al., 2011). Dynamic imaging was able to demonstrate that 13C alanine synthesis is restricted to the innermost most hypoxic region of the endosperm (Figure 8). In combination with biochemical and flux balance analysis, a spatially resolved metabolic model of the starchy endosperm has since been derived, with the aim of obtaining an improved interpretation of localized metabolic activity. The metabolic compartmentalization occurring in the starchy endosperm provides a measure of physiological flexibility, and contributes to the high carbon conversion efficiency shown by the starchy endosperm of the cereal (Alonso et al., 2011). Apart from such seed-targeted experiments, a number of related applications are conceivable. Experimental plants could be fed with various 13C substrates in order to define the routes by which the corresponding compounds are taken up, distributed and/or metabolized. When applying other isotopically labelled nuclei (e.g. 15N, 19F, 31P), one needs to consider the shift in sensitivity of MRI.

Figure 8.

 Functional imaging of metabolite dynamics in the living endosperm of barley grains by use of magnetic resonance imaging (MRI). (a, b) A MR image showing the steady-state distribution of 13C sucrose fed to intact caryopses; the 1H reference image is shown in b. (c) A MR image showing the localized synthesis of 13C-alanine and its preferential accumulation in the central endosperm region. (d) Diagram indicating the direction of sucrose flow (red arrows) within the caryopsis (see also Video clip S3). (e, f) Flux maps depicting mitochondrial metabolism in peripheral (e) versus central (f) endosperm regions as derived from flux balance analysis.
Abbreviations: en, endosperm; np, nucellar projection; p, pericarp. For details see Rolletschek et al. (2011).

Taken together, we argue that dynamic MRI opens up new perspectives for the non-invasive analysis of metabolic compartmentalization, metabolic modelling and the identification of metabolic markers in plants.


Axel Haase (University of Munich) and Ulrich Wobus (IPK Gatersleben) are gratefully acknowledged for their support in commencing our NMR research on plants. The authors thank Gerd Melkus (University of California) and Johannes Fuchs (University of Würzburg) for dedicated contributions. We also thank Peter M. Jakob (University of Würzburg), Andrew Webb (University of Leiden) and Thomas Altmann (IPK Gatersleben) for continuous support. We acknowledge funding by the German Federal Ministry of Education and Research, the Deutsche Forschungsgemeinschaft and BayerCropScience.