• Open Access

Quantitative imaging of oil storage in developing crop seeds

Authors


* Correspondence (fax: +49-39482-5500, e-mail: borysyuk@ipk-gatersleben.de)

†These authors contributed equally to this article.

Summary

In this article, we present a tool which allows the rapid and non-invasive detection and quantitative visualization of lipid in living seeds at a variety of stages using frequency-selected magnetic resonance imaging. The method provides quantitative lipid maps with a resolution close to the cellular level (in-plane 31 µm × 31 µm). The reliability of the method was demonstrated using two contrasting subjects: the barley grain (monocot, 2% oil, highly compartmentalized) and the soybean grain (dicot, 20% oil, economically important oilseed). Steep gradients in local oil storage were defined at the organ- and tissue-specific scales. These gradients were closely coordinated with tissue differentiation and seed maturation, as revealed by electron microscopy and biochemical and gene expression analysis. The method can be used to elucidate similar oil accumulation processes in different tissues/organs, as well as to follow the fate of storage lipids during deposition and subsequent mobilization.

Introduction

The improvement of oil-producing crops has become a priority in the search for products for a range of industrial purposes and for the replacement of non-renewable energy sources. Recent technological advances have allowed the generation of plants characterized by an increased oil content and/or a novel fatty acid composition (Jaworski and Cahoon, 2003; Hildebrand et al., 2005). However, a number of significant technical challenges still remain to be solved.

Indeed, most of the fundamental research on oilseed metabolism has been performed in Arabidopsis. Although the majority of genes encoding lipid biosynthesis in plants have been identified and isolated (White et al., 2000; Ruuska et al., 2002; Beisson et al., 2003; Murphy, 2005), the specific pattern and fine regulation of gene expression in crop seeds are largely unknown. A further level of complexity is achieved by the post-transcriptional regulation of biosynthetic enzymes involved in lipid storage (Ohlrogge et al., 2000; Voelker and Kinney, 2001). This includes, inter alia, light, metabolite supply, oxygen delivery and the energy/redox state (Rawsthorne, 2002; Rolletschek et al., 2004; Ruuska et al., 2004; Weber et al., 2005). The levels of such factors can vary dramatically from tissue to tissue within the seed (Borisjuk et al., 2003), and therefore can have a significant influence on lipid accumulation in vivo (Rolletschek et al., 2005a,b). As seeds of oil-accumulating crops (such as soybean, canola and maize) differ in several aspects from the Arabidopsis seed model, a knowledge about lipid accumulation in vivo is necessary in order to obtain a rational choice of gene targets for crop improvement. Despite millenarian domestication history, the quantitative topology of lipid storage in crop plants could not be visually assessed until now.

The acquisition of quantitative data on the lipid distribution within seed organs largely relies on assays carried out after tissue dissection and extraction. Currently available non-destructive techniques allow either the quantification or visualization of lipids (Table 1). Recent major progress has been based on mass spectrometry (Adlof, 2003) and non-invasive magic-angle spinning 13C nuclear magnetic resonance (NMR) spectroscopy (Ratcliffe and Shachar-Hill, 2001). Both allow a detailed quantitative description of lipid composition, but are not appropriate for imaging. Moreover, neither chemical imaging nor other current lipid visualization methods are reliable for quantification in living seeds. Some techniques, such as fluorescence correlation spectroscopy imaging and multiphoton excitation, have the potential to allow accurate in vivo measurements of lipids and numerous other compounds. However, they suffer from interference from photodamage, light scattering and other characteristics of living plant tissues. Third-harmonic generation microscopy, for example, allows the visualization (Débarre et al., 2006) and measurement of lipids in biological liquids (Débarre and Beaurepaire, 2007). Unfortunately, the poor penetration of optical signals into and out of internal tissue layers is a particular problem in the context of developing crop seeds.

Table 1.  Overview of non-destructive methods for storage lipid analytical techniques
MethodSpecies/tissueRestrictionReference
Magic-angle spinning (MAS) 1H nuclear magnetic resonance (NMR)Barley seedNo visualizationRidenour et al. (1996)
MAS 13C NMRCanola seedNo visualizationHutton et al. (1999)
Multiphoton microscopyArbuscular mycorrhizal symbiosisNo multiple tissue layersBago et al. (2002)
Third-harmonic generation (THG) microscopyInsect, mammalian and plant tissuesLimited tissue penetrationDébarre et al. (2006)
Near-infrared spectroscopyMelaleuca cajuputiAir-dried samples onlySchimleck et al. (2003)
Raman spectrometry (RS)Eucalyptus treesLow tissue penetrationBaranska et al. (2005)
Fourier transform RSAuthentification of oilsLow tissue penetrationBaeten and Aparicio (2000)
Chemical shift imaging (CSI)Fennel mericarpNo quantificationRumpel and Pope (1992)
CSI combined with 13C NMR spectroscopyMaize kernelNo imaging and quantificationKoizumi et al. (1995)
CSIHerbarium (taxonomic distinctions)Dry tissue onlyMasson et al. (2001)
CSIBarley seedNo quantification, low resolutionGlidewell (2006)
CSICarum copticumessNo quantificationGersbach and Reddy (2002)
Nuclear magnetic resonance imaging (MRI)Barley and soybean seedsDry seeds onlyKano et al. (1990)
MRIPinus monticola seedsNo quantificationTerskikh et al. (2005)
MRICoconut fruitsVery low resolutionJagannathan et al. (1995)
MRI with correlation spectroscopy (COSY)Fennel mericarpDry tissue onlySarafis et al. (1990)
1H NMR spectroscopySoybean and barley seedsLiquid samples onlyKnothe and Kenar (2004)
1H NMR spectroscopy and MRIJuglans nigra seedsNo lipid imagingVozzo et al. (1996)
1H NMR spectroscopy (localized) and MRICherry fruitsNo quantificationIshida et al. (1998)
Conventional 1H NMR spin echo imagesGrape berry fruitsNo calibrationPope et al. (1993)
Combined spin echo imaging and CSISoybean seedDry tissue onlyBorisjuk et al. (2005)
Pulsed gradient spin-echo (PGSE) NMROil bodies in embryoNo visualizationBacic et al. (1992)
Spin-echo imaging and CSICarya illinoensis embryoNo quantificationHalloin et al. (1993)

Alternatively, a non-destructive magnetic resonance (MR)-based approach would be attractive for both structural and quantitative investigations (Ratcliffe and Shachar-Hill, 2001; Köckenberger et al., 2004). However, many of the key attributes of plant tissue (in particular, their rigid cell walls and high water content) restrict the application of NMR-based lipid imaging (Köckenberger et al., 2004). To our knowledge, to date, no documented study has demonstrated the simultaneous visualization and quantification of lipids in developing seeds.

In this article, we describe a non-destructive MR-based method for the topographic and quantitative analysis of lipid storage in living seeds. Two plant models were chosen. The first was the barley grain (monocot), because its sophisticated lipid compartmentalization and extremely low lipid content (about 2% of dry weight) present a particular technical challenge. The second was the soybean seed (dicot), accumulating about 20% lipid and representing the most important industrial oil producer. Tissue-specific localization and gradients in lipid deposition were demonstrated and related to gene expression pattern and biochemical, structural and cytological parameters during seed development. The development of a quantitative in vivo lipid imaging technology represents a significant breakthrough in oilseed crop research.

Results

Molecular and biochemical analysis of oil storage in barley seeds

Over 250 genes involved in lipid metabolism were identified from expressed sequence tags (ESTs) generated from complementary DNA (cDNA) libraries of developing barley grains (Zhang et al., 2004). A set of 37 genes involved in fatty acid and triacyl glycerol (TAG) biosynthesis was defined by aligning barley ESTs with Arabidopsis thaliana and rice genomic sequences (see ‘Experimental procedures’). Their expression during seed development is shown in Figure 1[for a pathway scheme, see Figure 2; gene list given in Table S1 (see ‘Supplementary material’)]. Gene expression patterns in embryo, endosperm and pericarp were distinct from one another. The majority of the genes encoding the major enzymes involved in fatty acid biosynthesis were highly expressed in the embryo and endosperm, but poorly in the pericarp (Figure 1). Most were represented by two isoforms: one with preferential expression in the embryo, and the other with preferential expression in the endosperm. Overall, more genes were expressed at a higher level in the embryo than in the endosperm.

Figure 1.

Temporal expression pattern of genes encoding enzymes catalysing reactions of fatty acid and triacyl glycerol biosynthesis in developing barley seed. Numbers indicate days after fertilization (2-day intervals). Reaction steps and enzyme names are indicated by numbers (left panel) and in the abbreviated scheme (see Figure 2). Changes in gene expression levels from low (blue) to high (red) are also represented in Table S1 (see ‘Supplementary material’).

Figure 2.

Abbreviated scheme for the enzymes catalysing the reaction steps of fatty acid and triacyl glycerol (TAG) biosynthesis in developing barley seeds. The majority of the genes expressed in developing barley seeds encoding the major enzymes involved in fatty acid biosynthesis were represented by two isoforms (see Figure 1). Specific examples include the E1α and E1β subunits of pyruvate dehydrogenase, dihydrolipoamide-S-acetyltransferase and dihydrolipoamide dehydrogenase, which were all represented by two related sequences (steps 1–4, see also Figure 1). Sequences of genes encoding enzymes involved in the production of acetyl-CoA have plastid-targeted signatures. Similarly, two isoforms of acetyl-CoA carboxylase were present, one of which (BU984531) was strongly expressed in the embryo throughout development, whereas the other (BU971956) was expressed predominantly in the endosperm during later stages of development (step 5). The genes encoding enzymes catalysing the carboxylation of acetyl-CoA to malonyl-CoA, the transfer to acyl carrier protein (ACP) and others involved in the formation of acyl-ACPs (steps 6–11) were also present in two isoforms. Most of these, as well as ACP synthase II (step 12; BU986474), were more strongly expressed in the embryo than in the endosperm. Acyl-CoA synthetase (step 16) is encoded by distinct embryo- (BQ464463) and endosperm-specific (BU970052) genes. Together with the activation of fatty acid and TAG biosynthesis genes, oleosin (BU973985, step 25) and caleosin (BU989955, step 26) were also more highly expressed in the embryo than in the endosperm during the late developmental stage.

As the aim was to relate the gene expression pattern with lipid content in various organs of the barley grain, gas chromatography was employed to quantify lipid accumulation. The endosperm, embryo and pericarp were dissected at a range of developmental stages (in parallel with gene expression analysis above), and the fresh weight and lipid content of these explants were measured (Figure 3).

Figure 3.

Lipid storage in barley seeds as revealed by a destructive mode. (a) Fresh weight accumulation in the developing embryo and endosperm. (b) Morphology of the barley grain at the main storage stage. Dotted lines show the level of the longitudinal section and cross-section through the fresh harvested grain. (c) Total lipid content in the developing embryo and endosperm. Sampling points correspond to the gene expression analysis given in Figure 1. (d) Lipid staining (red) demonstrated in a cross-section through the grain at the main storage stage. c, cross-section; DAF, days after flowering; e, endosperm; em, embryo; FW, fresh weight; l, longitudinal section; p, pericarp.

The embryo fresh weight increased rapidly from 13 days after flowering (DAF), but was negligible compared with that of the endosperm (Figure 3a). By contrast, the lipid content in the embryo was several fold higher than that in the endosperm (Figure 3c). The dynamics of lipid accumulation in the endosperm were distinct from those in the embryo. The level increased up to 25 DAF, but thereafter declined. The temporal pattern of lipid accumulation corresponded quite well to the temporal expression of genes related to oil storage. This is exemplified by oleosin in Figure S1 (see ‘Supplementary material’). The lipid content of the pericarp was negligible during the entire development (data not shown). There were no marked differences in fatty acid composition in the embryo and endosperm (data not shown).

Histochemical staining with Sudan B confirmed lipid accumulation in both the embryo and endosperm, but the characteristic pattern of staining suggested a tissue-specific nature of storage in both organs (Figure 3d). As lipid deposition was restricted to the suborgans of the tiny embryo and to a few peripheral cell layers of the endosperm, an accurate quantitative evaluation of lipid content based on tissue dissection was practically impossible.

Experimental design for simultaneous visualization and quantification of lipid in the developing barley grain

NMR-based chemical shift imaging (CSI) has been used for the detection of lipids in dry plant material and in oil-rich fruits and seeds (Table 1). In our experiments, CSI distinguished between the oil and water components in a dummy sample, but short-term measurements were not reliable for the analysis of living barley grains (data not shown). Given the small size, high water content and very low lipid content of the grain, a long measurement time (> 24 h) would have been required to achieve the level of spatial resolution necessary to distinguish between all the relevant tissue types. During this lengthy period, the living grain would have experienced changes in size, weight and shape. Thus, CSI was not appropriate for the assessment of the lipid distribution in the developing grain.

The distinct feature of CSI is that it provides a complete spectrum with chemical information for each pixel in the image. As we were only interested in the water and lipid signals, but no further chemical information beyond this, CSI was a very time-inefficient mode in which to acquire the data. Instead, a different approach was used. A single-slice image was acquired using either a short 90° sinc pulse (which excites protons in both water and lipid) or a longer 8-ms sinc pulse (which, as a result of its bandwidth of ~780 Hz, excites only one of the resonances). In both cases, the slice select gradient was present only during the 180° refocusing pulse. With offset frequencies taken from a global spectrum, both selective and non-selective images of the water resonance and the main lipid resonance could be acquired.

To validate the pulse sequences for the acquisition of frequency, selective images were applied to a dummy sample of vegetable oil and water (Figure 4a–c). Figure 4a shows a cross-section through an oil-containing glass tube placed inside a glass container filled with water. This non-selective image visualizes both water and oil. The selective images could specifically distinguish between water and oil (Figure 4b and c, respectively). The same protocol was then applied to living grains (Figure 4d–g). The structural representation of the tissue pattern (non-selective image; Figure 4e) and the distribution of water (Figure 4f) and lipid (Figure 4g) were visualized within the longitudinal section through the spike. When the lipid-specific pulse sequence was used, the lipid signal was detected only in the embryo and the surface of the endosperm (Figure 4g). By contrast, the highest water signal was found in the stem/rachis, the husk and the crease region of the grain (Figure 4f).

Figure 4.

Experimental design of the nuclear magnetic resonance (NMR) method for the in vivo investigation of developing seeds. Water and lipid selective images were measured using a dummy sample containing water and vegetable oil. (a) Non-selective image showing both water and oil in a cross-section through the tube. Water and oil are recognizable as bright signals, and glass is invisible/black. (b) Water selective image through the same tube with oil and glass invisible. (c) Lipid selective image with water and glass invisible. (d) Barley spike used for the NMR measurements. The red star indicates the individual grain visualized non-invasively in (e), (f) and (g). (e) NMR image of the virtual section though the barley spike. To obtain this image, spikes were placed inside the magnetic coil. Data acquisition/processing was obtained from intact spike/grains, and did not require prior sample preparation. (f) Water distribution in the grains. (g) Lipid distribution in the grains. The red star indicates the grain margin. 1, endosperm; 1*, surface of endosperm (aleurone); 2, embryo; 3, pericarp; 4, stem/rachis; 5, husk; bar, 3 mm.

To translate the signal intensity of the lipid-specific NMR image into actual lipid content, calibration was required. For this purpose, NMR measurements on living seeds were combined with subsequent gas chromatography applied on dissected samples. Based on the MR imaging results, distinct regions of seed were dissected and the total lipid content was measured using conventional gas chromatography (see ‘Experimental procedures’). The measured MR signal intensity of various tissue samples was plotted vs. the actual total lipid content (Figure 5). This calibration curve was used to quantify lipid content in various tissues of the grain, and the resulting quantitative image of lipid distribution represents a colour-coded ‘lipid map’ (shown in Figure 6).

Figure 5.

Calibration curve for nuclear magnetic resonance (NMR)-based lipid quantification. The NMR signal was plotted against a set of standard lipid concentrations. For details, see ‘Experimental procedures’. FW, fresh weight.

Figure 6.

Quantitative imaging of lipid in a living barley grain. Grain structures at early- (a), mid- (b) and late-storage (c) stages are presented on the left. The lipid content is colour coded. cl, coleoptile; co, coleorhiza; DAF, days after flowering; e, endosperm; em, embryo; g, grease region; n, nodule; p, pericarp; sc, scutellum.

Lipid mapping reveals steep gradients within the developing barley grain

The superimposition of lipid distribution with high-resolution MR images enabled a detailed analysis to be made of lipid gradients within any tissue and/or developmental stage of the seed. Figure 6 shows the lipid distribution in individual developing grains at the early-, mid- and late-storage stages (corresponding to 17, 25 and 35 DAF, respectively). At each of these stages, both a structural (non-selective) MR image, representing a longitudinal view of the grain, and a colour-coded lipid map within the same region are shown (also compare with Figure 3b, which shows a fresh harvested seed dissected by scalpel). The maps reach an in-plane pixel resolution of 31 µm × 31 µm, and thus identify lipid distribution close to the cellular level. At 17 DAF, the strongest signal (given in red) was localized in the scutellum and nodular region of the embryo. A clear decrease in lipid content occurred in the direction of the coleorhizae and coleoptile. Particularly steep gradients were observed in the endosperm, where the highest lipid levels were present within the aleurone layer, with very low levels in the basal endosperm. The pericarp did not show lipid deposition.

During the course of grain development, the general pattern of lipid distribution was rather stable, with the embryo and aleurone remaining as the major sites of lipid deposition. Within the embryo, the high lipid region expanded, occupying the scutellum and nodule (Figure 6c). By contrast, the proliferating regions (coleorhizae and coleoptile) avoided the accumulation of lipids during the complete maturation programme. Within the aleurone, the lipid level declined as development proceeded, although it always remained much higher than that in the starchy endosperm.

Overall, lipid mapping visualizes the tissue-specific deposition of lipids in a quantitative and non-invasive manner, with high spatial resolution.

High signal intensity in the barley lipid map is associated with the accumulation of lipid bodies

Lipid deposition in plant cells occurs mainly in lipid bodies, which possess a matrix of TAGs surrounded by a lipid monolayer that harbours oleosins as structural proteins (Huang, 1992). Transmission electron microscopy was used to relate the in vivo lipid map to the intracellular localization of lipids.

According to the lipid map (Figure 7a, top panel), distinct regions of tissues (Figure 7a, bottom panel) with high or low lipid level were dissected and analysed by transmission electron microscopy (Figure 7b–i). A very low lipid signal was characteristic of the pericarp, whose cells contain cuticular compounds, plastoglobuli and intracellular membrane systems, but only a few lipid bodies (data not shown). Similarly, the central part of the starchy endosperm, which generated no specific lipid signals in the MR map, contained neither lipid bodies nor oil deposits (except for the plastoglobuli within plastids). Tissues with a slightly elevated MR lipid signal, such as the meristem, had a small number of small lipid bodies, scattered randomly throughout the cytoplasm (e.g. the embryo radicle; Figure 7b,c). The moderate lipid level in the parenchymal nodule tissues (linking the coleorhizae with the coleoptile) was associated with numerous, large lipid bodies (Figure 7d,e). High lipid levels within the scutellar parenchyma and epidermal cells were associated with an increase in the size of the lipid bodies, and their spatial realignment along the plasma membrane (Figure 7f,g). In aleurone cells, a population of variously sized lipid bodies underlaid the plasma membrane and coated the surface of cellular organelles, such as the plastids and protein bodies (Figure 7h,i).

Figure 7.

Lipid distribution within the barley embryo combined with representative transmission electron microscopy images. (a) An in vivo lipid map at the mid-storage stage (top; lipid content is colour coded) and tissue structure within the same region (bottom). (b–i) Distribution of lipid bodies within the region of the coleorhizae (b, c), nodule (d, e), scutellum (f, g) and aleurone (h, i). cw, cell wall; m, mitochondria; n, nuclei; pb, protein bodies; pl, plastid; st, starch.

In embryonic tissues, the gradual increase in the NMR signal in the lipid map was associated with a shift from cells containing very small lipid bodies (coleorhizae and coleoptile) to cells accumulating numerous, variously sized, lipid bodies (nodule and scutellum). The highest lipid signal was observed in cells consisting predominantly of lipid bodies (the central region of the scutellum). Thus, the lipid map indicates the accumulation of lipid bodies, i.e. oil storage, within the tissues.

Lipid deposition in mature barley grains reflects the developmental arrangement of storage

The NMR method allows the lipid distribution in mature seeds to be visualized as a three-dimensional model (shown as a ‘movie’ in Figure S2, see ‘Supplementary material’). The model rotates around the longitudinal axis and allows the lipid distribution to be assessed within the entire grain, thereby giving a clear three-dimensional view of oil storage. Lipid is stored in the embryo and the aleurone, whereas the absence of signal in the central region of the (starchy) endosperm and the pericarp permits an unobstructed view of the interior of the grain. The transparent belt stretching along the longitudinal axis corresponds to the crease region in which the aleurone is absent.

The three-dimensional model facilitates the visualization of oil storage within the mature grain. It reflects the tissue-specific pattern of lipid accumulation during development.

Lipid map for developing soybean seeds

To demonstrate the applicability of the MR-based lipid mapping method for economically important oilseed crops, soybean was used. By contrast with barley, soybean seeds have a dicotyledon morphology, a chlorophyll-containing, large-sized embryo, and contain approximately 10-fold more oil (on a dry weight basis). Lipid storage is initiated at around 15 DAF, and maximum accumulation rates are reached during the main storage stage (24–42 DAF; Borisjuk et al. 2005). Frequency-selective MR imaging, as described in this paper, was used to visualize the lipid distribution in developing soybean seeds. The lipid map is shown in Figure 8b. Lipids were accumulated preferentially in the adaxial (basal) region of the embryo, decreasing towards the peripheral (abaxial) regions. This results in steep lipid gradients within single cotyledons, with about three-fold differences in local lipid levels. Lower levels were observed within vascular tissues. Within the seed coat, no MR signals were detected. The abaxial and adaxial regions of the seeds were dissected and the lipid content was analysed by gas chromatography. As shown in Figure 8c, the derived data were in good agreement with the MR-based lipid map. According to electron microscopy, lipid bodies accumulated preferentially in regions with a high MR lipid signal (similarly to barley, see above), i.e. were more obvious in adaxial vs. abaxial regions of the embryo (Figure 8d vs. 8e). It is concluded that the lipid mapping method described here is appropriate for developing oilseeds such as soybean.

Figure 8.

Quantitative imaging of lipid in an intact living soybean seed. (a) Non-selective nuclear magnetic resonance (NMR) image of seed structure: cross-section at the median part of the seed. (b) Lipid map of the seed shown in (a); lipid content is colour coded and forms a gradient with a maximal level in adaxial regions (red) and a minimum level in abaxial regions (blue–green). It should be noted that lipid deposits are absent in the seed coat and low in vascular tissues. (c) Lipid content in dissected tissues measured by gas chromatography. (d, e) Transmission electron microscopy of cells showing lipid bodies (dark spots) within adaxial (d) and abaxial (e) regions of the embryo. ad, adaxial region; ab, abaxial region; em, embryo; FW, fresh weight; sc, seed coat; vs, vascular tissues.

Discussion

Frequency-selective MR ‘lipid mapping’ opens up new perspectives for in vivo analytical techniques

MR imaging and spectroscopy are routinely used as powerful techniques in animal science and medicine, and have recently become popular in experimental botany (Ishida et al., 2000; Van der Weerd et al., 2001; Van As, 2007). However, the application of MR imaging for the visualization, quantification and analysis of the chemical content of plants remains a rather difficult task, mostly because of technical and methodological obstacles caused by specific characteristics of plant tissues (Ratcliffe and Shachar-Hill 2001; Köckenberger et al., 2004). Furthermore, NMR parameters by themselves do not possess any significant biological value. In order to make them useful for biologists, MR imaging data must be related to histological, biochemical and other characteristics of plant tissues.

In this article, we have demonstrated an original frequency-selective MR-based tool for the investigation of oil storage in developing crop seeds. Lipid deposition was visualized and quantified in living seeds and within different tissues at distinct stages of seed development. Such a comprehensive in vivo analysis outperforms any previous method. The major advantages are as follows.

  • 1The non-invasive mode of the method, i.e. avoidance of any tissue preparation.
  • 2The spatial resolution of lipid mapping is sufficiently high to distinguish local lipid content at the tissue level (e.g. aleurone layer of barley endosperm, vascular tissue of soybean). Moreover, mapping can be performed using moderate magnetic field strengths which are available in conventional scientific praxis. An even higher resolution can be reached by studying smaller objects, such as canola, linseed and others. In such cases, the application of small solenoid coils is recommended (Webb, 1997).
  • 3The short measurement time of about 1 h is a significant benefit of the new imaging method because the size and form of growing seeds change rapidly.
  • 4The simultaneous monitoring of anatomy and lipid deposition offers the possibility to relate lipid accumulation to the developmental changes of individual seeds. Following lipid mapping, seeds can be treated by any desired method (biochemical, histological, molecular analysis) or can just be allowed to grow.
  • 5Lipid mapping can be performed on different types of seed independent of anatomical differences or lipid content. The calibration procedure ensures the accuracy and quantification of mapping. We have demonstrated in this study how mapping can be applied to remarkably different types of seed [monocot (barley) and dicot (soybean)], which differ by more than five-fold in seed weight and 10-fold in lipid content.

Biotechnological relevance of the lipid mapping tool

Not all tissues/organs have become adapted to accumulate lipids. In some plants, the fruit is the major storage organ (e.g. oil palm, olive), whereas, in others, it is the embryo (e.g. soybean, rapeseed). Conventional breeding for high oil content has relied on improvements in the capacity of the organs appropriate (adapted in nature) for lipid storage. For example, maize varieties bred for oil production tend to form large embryos at the expense of the amount of starchy endosperm (Alexander, 1988). Lipid maps now show that, even within lipid-storing organs, only distinct tissues deposit oil to high levels. For example, in the barley embryo, only the scutellum and nodule are appropriate for oil storage, and thus are matter of selection. This observation could not have been made using the dissection of embryonic suborgans, which have been thought to contain highly variable but similar levels of lipid (reviewed by Morrison, 1993). Thus, lipid mapping allows the identification of more precise ‘topographical targets’ for breeding.

In the soybean embryo, the lipid content decreases towards the peripheral region of the embryo, defining only the inner (adaxial) region as the main site for oil storage. Although the underlying molecular mechanisms remain unknown, this demonstrates the ability of the lipid mapping tool to define regions with maximum and minimum oil accumulation capacity. Such knowledge identifies the distinct tissue areas for further detailed studies, including microdissection, microarrays, biochemical analysis, etc. This will provide the basis for new transgenic strategies to manipulate oil storage in soybean and other oilseed crops.

The non-invasive nature of the mapping method allows the dynamics of oil storage to be observed during development. This is particularly important as a result of the occurrence of transient lipid storage and/or lipid degradation during seed development (Chia et al., 2005). In the barley endosperm, the peak value within the aleurone layer is reached shortly after the onset of storage, but declines thereafter. This may be related to the fact that β-oxidation pathway genes in the aleurone are already activated at the peak of storage activity (N. Sreenivasulu, unpubl. data). As the utilization of lipid reserves before germination affects yield and seed quality, the identification of regions with transient lipid storage is of high biotechnological significance. For example, a transgenic approach could use tissue-specific promoters to locally block lipid degradation in regions previously defined by the mapping tool.

Gene manipulation/mutation enables the relocation and/or elevation of oil storage in alternative plant organs (Ogas et al., 1997; Dörte et al., 2004). Nevertheless, the performance of the transgenic strategy is complicated in vivo by multiple positional cues and the developmental regulation of lipid storage. The generation of transgenic lines is extremely time consuming and elaborate, and requires different types of analysis, multiple sampling and large amounts of seed material. The application of non-destructive lipid mapping enables scientific predictions to be checked during seed growth without killing the plant, the determination of time points during development which are most appropriate for certain analyses, and the observation in vivo of the effects of different factors (stress, light, herbicides). Lipid mapping not only enables the acquisition of data at the individual seed level, but also opens up the possibility of comparative non-destructive analysis for all grains within the spike or pod (or entire plant) without interrupting seed development. This feature is highly desirable for the selection of specific phenotypes in transgenic research and breeding programmes. Overall, this will reduce the number of samples and experiments with transgenic plants, as well as the related expenses and ecological risks.

In vivo topology of lipid storage is related to plastid and tissue differentiation

The accumulation of storage products in seeds is related to the differentiation and/or metabolic status of tissues, as demonstrated for starch- and protein-accumulating crops (Wobus et al., 2004; Weber et al., 2005). Similar studies on oil storage have been hampered by the lack of appropriate methods. The NMR tool revealed a close relation of lipid storage to tissue differentiation.

In soybean seed, lipid accumulation occurs in green photosynthetically active tissues of the embryo and temporally coincides with plastid differentiation: the chloroplasts within the interior of the embryos gradually differentiate into storage organelles (Saito et al., 1989). This functional shift is evident at the level of gene expression, loss of photosynthetic capacity and accumulation of oil bodies (Borisjuk et al., 2005; Rolletschek et al., 2005b). The in vivo lipid map of young soybean seeds demonstrates the spatial pattern of oil storage in developing seeds. It proves quantitatively that the accumulation of lipid occurs in inner regions of the embryo, and lipid gradients appear from the early developmental stages onwards. Differences in local lipid content are remarkable, and strongly support earlier hypotheses about the relationship of oil storage and plastid differentiation (Saito et al., 1989; Borisjuk et al., 2005). In this way, previous biochemical (Eastmond and Rawsthorne, 2000) and molecular biological (Ruuska et al., 2002) data on the central role of plastids in oil storage are demonstrated here in vivo at the topographical level.

In barley seed, the lipid map of the embryo was compared with its differentiation pattern (Duffus and Cochrane, 1993). A low lipid content was characteristic of developmentally younger suborgans, and vice versa. The highest lipid level was found in the scutellum and nodule. These organs are differentiated earlier in development and provide transient storage and nutrient delivery to the dividing cells of the coleorhizae and coleoptile (root and shoot of the next generation). A high lipid content is associated with a specific intracellular alignment of lipid bodies along the plasma membrane, as shown by transmission electron microscopy (Figure 7). The migration of lipid bodies towards the membrane is a common phenomenon in plant tissues during cell dehydration and maturation, but does not occur on short premature drying of the embryos (Dasgupta et al., 1982; Leprince and Hoekstra, 1998; Golovina et al., 2001). The intracellular alignment of lipid bodies might serve to decrease cell solute leakage from this region of the embryo, which would otherwise be promoted by the change in its cellular water status (Perdomo and Burris, 1998; Cordova-Telleza and Burris, 2002). This would lead to a more organized dehydration and acquisition of desiccation tolerance. It is noteworthy that the intracellular arrangement of lipid bodies is essential for correct germination, and is also a significant determinant of seed end use quality (Cordova-Tellez and Burris, 2002).

Clearly, the subcellular role of lipid bodies is much more complex than their simple action as relatively inert carbon stores. Lipid ‘droplets’, long considered as inert storage vessels for energy-rich fats, have recently been shown to play an important role in eukaryotic cells (for a review, see Martin and Parton, 2006). They have even earned recognition as possible organelles (Beckman, 2006). In his review, Murphy (2001) compared the widespread distribution and dynamic metabolic role of intracellular lipid bodies in animals, plants and microorganisms. Recent progress in the characterization of these organelles has relied on the availability of powerful new analytical methods. For example, modern microscopic techniques have allowed the real-time monitoring of the formation and dynamics of lipid droplets (Martin and Parton, 2006). Our lipid mapping tool enables both the accumulation and degradation of lipid to be followed during entire development on the organ/tissue scale. It is necessary to determine precise targets for a further move to the (sub)cellular level of investigation. The non-invasive nature of lipid mapping makes this easy to accomplish with modern microscopic tools (Supatto et al., 2005; Débarre et al., 2006; Ogilvie et al., 2006), molecular cell biology and biochemistry. Combining the NMR tool with other methods will allow different levels of complexity (organ/tissue/cell/organelle scale) to be studied, and promises to be a fascinating and productive field for future scientific research.

Conclusions

Visualization has become a very popular approach in plant science: ‘Seeing is understanding’ (Breithaupt, 2006). The MR-based lipid mapping method described in this article allows the simultaneous visualization and quantification of lipid in a non-destructive manner in living plant tissues, a capability that has not been available to date. The method is applicable at the level of the individual tissue/seed/entire plant during ontogeny. A knowledge of lipid gradients is fundamental for the understanding of local regulatory networks covering storage metabolism, and for the development of new approaches for plant breeding and transgenic research.

Experimental procedures

Plant material

Barley plants (Hordeum vulgare cv. Barke) were cultivated in growth chambers under a light/dark regime of 16/8 h at 20/14 °C. Flowering ears were tagged for the determination of DAF. Soybean plants [Glycine max (L.) Merr.] were cultivated in the glasshouse under a light/dark regime of 16/8 h. Intact seeds at distinct developmental stages were harvested during the mid-light phase and, if used for metabolite analysis, were snap-frozen in liquid N2 and stored at –80 °C. The dissection of barley seed tissues (pericarp, embryo and endosperm) was performed as described in Sreenivasulu et al. (2006).

Determination of total esterified fatty acids

Total fatty acids were analysed as their corresponding fatty acid methyl esters (FAMEs) by gas chromatography (Hornung et al., 2005). FAMEs were prepared by direct transmethylation with methanol containing 2% (v/v) dimethoxypropane and 2.75% (v/v) sulphuric acid. After 1 h at 80 °C, 0.2 mL of 5 m NaCl was added and FAMEs were extracted with 2 mL of hexane. The organic phase was dried with Na2SO4 and evaporated to dryness under nitrogen. FAMEs were dissolved in 10 µL of acetonitrile. The analysis of FAMEs was performed using an Agilent 6890 gas chromatograph (Waldbronn, Germany) fitted with a DB-23 capillary column (30 m × 0.25 mm; coating thickness, 0.25 µm; J&W Scientific/Agilent, Palo Alto, CA, USA). Helium was used as carrier gas (1 mL/min). The temperature gradient was 150 °C for 1 min, 150–200 °C at 8 °C/min, 200–250 °C at 25 °C/min and 250 °C for 6 min. The amount of total lipids was calculated as the sum of all detected FAMEs.

Histological methods

Lipid staining was conducted using fresh, free-hand sections with Sudan B, as described previously (Borisjuk et al., 2005).

Lipid determination by MR imaging

MR imaging experiments were performed on a Bruker 500-MHz Avance system (Bruker Biospin, Rheinstetten, Germany). A custom-built solenoid coil (diameter, 6 mm) was used as the radiofrequency (RF) coil. The proton spectra of intact developing seeds allowed the resolution of the individual water and multiple lipid peaks. A spin echo imaging sequence was applied with the 90° excitation pulse being set to be highly frequency selective (sinc shape; length, 8 ms; bandwidth, 777 Hz). With offset frequencies taken from a global spectrum, selective images of the water resonance and the main lipid resonance were obtained. In order to obtain an image containing both water and lipid signals, the frequency-selective RF pulse was replaced by a short, 1-ms sinc pulse. The typical experimental time needed to create one high-resolution, frequency-selective in vivo lipid image was 1 h 8 min, using a field of view of 8 mm, data matrix of 256 × 256 complex points, in-plane resolution of 31 µm, slice thickness of 0.75 mm and TR/TE = 1000/12.1 ms. Image processing and analysis were performed with programs written in IDL (ITT Visual Information Solutions, Boulder, CO, USA).

Calibration procedure for MR imaging

To translate the signal intensity of the lipid-specific MR image into actual lipid content, a calibration procedure was applied. Individual barley seeds of different developmental stages (early-, mid- and late-storage stages) were analysed by MR imaging under identical experimental conditions. The signal intensities for pixels in distinct areas of the seed were averaged (given as regions of interest, ROIs). Once defined by MR imaging, those areas of the particular seeds were subsequently dissected rapidly using a stereomicroscope. The samples (dissected parts of the embryo, pericarp, branch endosperm, chalazal region) were freeze-dried and the total lipid content was measured using conventional gas chromatography (described above). The measured MR imaging signals of the various areas (normalized ROI values) were plotted vs. the known total lipid content (gas chromatography data) of the respective dissected tissue samples. The data are illustrated as calibration curves, where the ordinate shows the normalized MR imaging signal intensity and the abscissa shows the total lipid values in the corresponding tissues (Figure 5). Each data point on the graph represents the average and standard deviation of two to three MR imaging measurements and two to eight gas chromatography readings. The fitted calibration curve was used to quantify the lipid content in various parts of the grain, providing quantitative images of lipid distribution (‘lipid map’).

In order to provide absolute quantification of the lipid content and distribution at each spatial location, ideally, the T1 and T2 values of the lipid should be measured at each pixel in the image. Such a procedure would lengthen the measurement time by at least an order of magnitude. In order to obviate the need for this, the data were collected with a relatively long value of TR, which minimizes the effects of any variations in lipid T1 (as a function of spatial location or developmental stage). In addition, the minimum value of TE was used to minimize the effects of any variations in the lipid T2 value. The calibration curve indicates that the choice of these TR and TE times enables a good correlation to be established between the lipid content measured by MR imaging and that by gas chromatography. Indeed, the standard deviations of the gas chromatography measurements are much higher than those from MR imaging.

Transmission electron microscopy

Embryos of different developmental stages were chemically fixed with 2% glutaraldehyde and 2% formaldehyde in cacodylate buffer (50 mm, pH 7.0) for 16 h. After three 20-min washes with the same buffer, the embryos were post-fixed with 1% OsO4 for 2 h. At the end of this procedure, the embryos were washed again with buffer and aqua dest, followed by dehydration in a graded ethanol series and subsequent embedding in Spurr's low-viscosity resin. After thin sectioning, the samples were stained with 4% uranyl acetate and lead citrate. Digital recordings were made on a Zeiss 902 electron microscope (Zeiss, Jena, Germany) at 80 kV.

Macroarray analysis

To identify genes encoding enzymes of fatty acid and TAG biosynthesis, sequence similarity searches were performed by comparing barley EST sequences obtained from developing seed libraries (Zhang et al., 2004) with GenBank sequences using blastx or blastn search algorithms. To define gene family members, barley EST sequences were aligned with available Arabidopsis and rice genomic sequences. A custom-made, high-density, 12 000 cDNA array was fabricated using sequences expressed in developing grains (Sreenivasulu et al., 2006), and used to study the temporal gene expression patterns of members of gene families of fatty acid and TAG biosynthesis during pericarp, endosperm and embryo development. Furthermore, the transcriptome data of embryo development were subjected to J-express software, which uses a reference transcript as ‘query gene’ and ranks other functionally related genes according to strong expression correlation match. This approach was used to find other functionally related genes co-expressed with oleosin transcripts.

Acknowledgements

We are grateful to U. Tiemann and K. Lipfert for artwork and A. Stegmann for excellent technical assistance. Special thanks are due to G. Melkus and V. C. Behr for help and discussion with MR imaging. We acknowledge funding by the Deutsche Forschungsgemeinschaft (Project number RO 2411/2-1/2-2 and BO 1917/2-1) and the Federal Ministry of Education and Research (BMBF; GABI SEED II grant). Andrew G. Webb and Thomas Neuberger acknowledge support of the Alexander von Humboldt Stiftung, Wolfgang Paul Preis.

Ancillary