The Metabolic Role of the Legume Endosperm: A Noninvasive Imaging Study^,,

Abstract

Although essential for normal seed development in the legumes, the metabolic role of the endosperm remains uncertain. We designed noninvasive nuclear magnetic resonance tools for the in vivo study of key metabolites in the transient liquid endosperm of intact pea (Pisum sativum) seeds. The steady-state levels of sucrose, glutamine, and alanine could be monitored and their distribution within the embryo sac visualized. Seed structure was digitalized as a three-dimensional model, providing volume information for distinct seed organs. The nuclear magnetic resonance method, combined with laser microdissection, isotope labeling, in situ hybridization, and electron microscopy, was used to contrast the wild-type endosperm with that of a mutant in which embryo growth is retarded. Expression of sequences encoding amino acid and sucrose transporters was up-regulated earlier in the endosperm than in the embryo, and this activity led to the accumulation of soluble metabolites in the endosperm vacuole. The endosperm provides a temporary source of nutrition, permits space for embryo growth, and acts as a buffer between the maternal organism and its offspring. The concentration of sucrose in the endosperm vacuole is developmentally controlled, while the total amount accumulated depends on the growth of the embryo. The endosperm concentration of glutamine is a limiting factor for protein storage. The properties of the endosperm ensure that the young embryo develops within a homeostatic environment, necessary to sustain embryogenesis. We argue for a degree of metabolite-mediated control exerted by the endosperm on the growth of, and assimilate storage by, the embryo.

The seed represents the link between the generations in higher plants. The seed coat, derived from maternal tissue, serves to both protect the filial generation and act as a conduit for the supply of nutrients (Murray, 1987). The diploid embryo and triploid endosperm both originate from the double fertilization event (Bewley and Black, 1994), but the developmental fates of these two organs differ greatly from one another. The embryo progresses through a fixed pattern of cellular divisions to form a series of specialized organs, whereas the endosperm, which initially grows coenocytically, either becomes a major storage organ (monocotyledonous species) or degenerates (dicotyledonous species) as the seed matures (Olsen, 2004; Sabelli and Larkins, 2009). The monocotyledonous endosperm serves as a crucial provider of nutrition during germination, while the role of the transient dicotyledonous endosperm has remained a puzzle despite more than a century of research.

The importance of the endosperm has become evident thanks to the elucidation of the molecular mechanisms underlying the control of seed size (for review, see Chaudhury et al., 2001; Berger et al., 2006). A series of genetic and epigenetic interactions in the endosperm is now known to play a crucial regulatory role over embryogenesis (Gehring et al., 2004; Nowack et al., 2007). The triploid endosperm appears to be a battleground over how resources are allocated to the embryo (Haig and Westoby, 2006). Given the higher dosage of maternal over paternal chromosomes, the mother may be able to exert greater control over nutrient allocation to the progeny and, in so doing, maximize her own fitness. The triploid state of endosperm appears to be advantageous for the transfer of nutrients from the maternal to the filial generation (Stewart-Cox et al., 2004; Sundaresan, 2005). Many genes/transcription factors active in the endosperm have been identified as having an effect on seed size (Ohto et al., 2005; Kondou et al., 2008), whereas the embryo contributes comparatively little to this trait (Garcia et al., 2005; Ingouff et al., 2006). The growth of the seed relies on a degree of coordination exercised by the endosperm, but how this is achieved at the metabolic level remains unclear.

The small seed size of the model plants Arabidopsis (Arabidopsis thaliana) and Medicago truncatula presents a technical challenge for the analysis of endosperm composition. The application of laser microdissection (Schneider and Hölscher, 2007) and/or noninvasive fluorescence resonance energy transfer nanosensors (Looger et al., 2005) can address some of these difficulties but does not help in the context of metabolic analysis, for which large-seeded species represent a more tractable subject (Zhang et al., 2007). Seed development in the agriculturally important species pea (Pisum sativum), soybean (Glycine max), broad bean (Vicia faba), and rapeseed (Brassica napus) has been extensively studied; thus, a preexisting body of knowledge surrounds both the spatial and temporal distribution patterns of metabolites in the embryo (for review, see Borisjuk et al., 2003) and nutrient loading into the seed (for review, see Patrick and Offler, 2001). However, the composition of the liquid endosperm and seed apoplast has only been touched upon (Gifford and Thorne, 1985; Borisjuk et al., 2002; Hill et al., 2003). This lack of experimental data largely reflects the difficulty of noninvasive sampling, without which the compartmentalization of metabolites and the biochemistry within the liquid endosperm cannot be determined. Current NMR technology now allows for the noninvasive study of metabolites in cell cultures, animals, and plants (Bourgeois et al., 1991; Metzler et al., 1995; Tse et al., 1996; Köckenberger, 2001; Neuberger et al., 2008).

Here, we sought to develop an analytical tool for the noninvasive imaging, quantification, and monitoring of metabolites in the living endosperm and to use this information to investigate the metabolic role of the endosperm in the legume seed. We have combined noninvasive NMR methods with more conventional histological and biochemical analyses to characterize the pea endosperm in both the wild type and a mutant in which the epidermal cell identity in the embryo has been lost (Borisjuk et al., 2002). The development of a giant endosperm in this mutant facilitates the elucidation of maternal-filial interactions and the mode of solute transport within the seed. The combination of approaches allowed a literally “new view” of the transient endosperm, its interrelation in vivo, and its function in legume seed.

RESULTS

Endosperm as an Immediate Environment of Embryo in Vivo

During its early development, the pea endosperm forms a motile, proliferating tissue lining the embryo sac cavity, later enveloping the embryo and suspensor (data not shown; Marinos, 1970). Endosperm becomes enlarged to occupy most of the embryo sac and forms a large vacuole in the chalazal region (Fig. 1A). Since the vacuole is completely enveloped by the endosperm cytoplasm, it belongs, spatially at least, to the endosperm and therefore can be considered as an “endosperm vacuole.” Cellularization of the endosperm cytoplasm occurs during the mid stage of development, at which time the endosperm vacuole becomes surrounded by a cell wall (Fig. 1, B and C), so that its status alters to become an apoplastic compartment within the endosperm. An NMR scan of a living endosperm vacuole (Fig. 1, D–I; Supplemental Movies S1–S3) shows a strongly labeled body (liquid-filled space) that embeds both the embryo and the suspensor throughout development. The layers of the endosperm cytoplasm could not be spatially resolved and appear to remain closely attached to the embryo sac wall and embryo surface. No air-filled spaces or other structures within the endosperm vacuole were identifiable. Early in development, the embryo floats within the endosperm vacuole (Fig. 1, D–I), thereby insulating it from mechanical pressure imposed by the seed coat. The suspensor binds the embryo to the embryo sac boundary wall, and its expansion pushes the embryo away from the micropyle (Fig. 1, D and G) and toward the mid (Fig. 1, E and H) or lower (Fig. 1, F and I) part of the embryo sac. The direct connection between the embryo and the maternal tissues is disturbed as the suspensor loses its integrity. The conventional assumption that the embryo is anchored to the boundary wall of the embryo sac (Marinos, 1970), therefore, needs to be reassessed.

Figure 1.

Anatomy of seeds as visualized by conventional histology and noninvasive MRI. Images are representative for approximately 30 to 50 mg fresh weight. A, Histostaining of the cross section though the chalazal region of a seed (approximately 30 mg fresh weight) shows the embryo, endosperm, and endosperm vacuole surrounded by endosperm, embryo sac walls, and seed coat. Due to toluidine blue staining, seed tissues except the endosperm vacuole are blue stained. B, Fragment of a section though the micropylar region of a seed (approximately 50 mg fresh weight) demonstrates cellularized endosperm, which envelops the embryo and borders the embryo sac wall (arrows). Toluidine blue staining is shown. C, The same as in B but within the chalazal region of a seed. D to I, Fragments of three-dimensional imaging using MRI (Supplemental Movie S1) showing internal structures of a seed. The noninvasive views correspond to the longitudinal section (D–F) and cross section (G–I). e, Embryo; en, endosperm; ev, endosperm vacuole; s, suspensor; sc, seed coat.

The NMR data were used to reconstruct a digital model of individual seeds (Fig. 2, A–C; Supplemental Movie S4), which permitted a three-dimensional visualization of seed anatomy and, in particular, allowed for the measurement of the volume of various seed organs and the detection of any changes in their relative size over time (Fig. 2, D–F). During a period over which the seed volume of the wild-type pea expanded 4-fold, the suspensor disappeared while the size of the embryo increased. The size ratio between the endosperm and the seed coat did not vary. The growth of seed is coupled with the major increase in seed coat volume, in concert with the rapid growth of the endosperm. In the E2748 mutant (Supplemental Fig. S1, A and B), the early stages of embryogenesis are similar to those occurring in the wild type, but from the mid cotyledon stage onward, embryo growth stops and a callus-like body is formed (Borisjuk et al., 2002).

Figure 2.

Digital models of three-dimensional anatomy of a developing pea seed at different developmental stages based on the in vivo NMR data set. Seed coat, endosperm, suspensor, and embryo were identified and color coded. A to C, Images represent seeds of 30, 73, and 120 mg fresh weight. Different seed tissues are color coded. D to F, Volumes of different seed tissues calculated using AMIRA software: seed coat (green), endosperm (blue), embryo (red), suspensor (yellow). e, Embryo; en, endosperm; s, suspensor; sc, seed coat.

The Growth of the Seed Is Primarily Associated with the Initial Growth of the Endosperm

The embryo-endosperm ratio was assessed in seeds developing on heterozygotic mutant plants (Mm). The phenotype of the progeny of these plants was either wild type (Mm and MM) or mutant (mm; Supplemental Fig. S1C). Intact seeds appear identical, but magnetic resonance imaging (MRI) was able to differentiate between the mutant and the wild type, thus avoiding the need for dissection (Fig. 3, A, B, E, and F). The seed size and endosperm volume of the two types are similar up to a seed fresh weight of 30 to 50 mg (data not shown). Thereafter, morphological differences become apparent in the embryo sac, although not in the seed coat. By approximately 150 to 200 mg seed fresh weight, wild-type embryos have reached the same mass (volume) as their endosperm (“equilibration point”) and continue to grow at the expense of the endosperm (Fig. 3D). The wild-type seed grows to a fresh weight of 450 mg, at which point the embryo almost completely fills the seed (Fig. 3C). By this latter stage, all of the major developmental events have been achieved (stage 24, as identified by Marinos, 1970) and further increases in seed size are minimal. The mutant embryo's growth is retarded, although the volume of the liquid endosperm continues to increase (Fig. 3H). By the time the mutant seed fresh weight has reached approximately 150 to 200 mg, the endosperm has become 3- to 5-fold larger than the embryo; at a fresh weight of 300 mg, the endosperm ceases its growth and begins to dehydrate to form a shrunken seed (Fig. 3G). Thus, the growth of the seed is primarily associated with the early growth of the endosperm rather than with any later growth of the embryo.

Figure 3.

Developmental parameters of E2748 mutant (mm) and wild-type (Mm) phenotype seeds. A, Seed of wild-type phenotype by noninvasive NMR. B, The same seed but dissected by hand showing endosperm vacuole (ev) and embryo (e). C, Section through mature seed where embryo sac is occupied by embryo. D, Fresh weight (fw) accumulation of embryos (squares) and volumes of the endosperm vacuole (circles) in wild-type seed. Data points represent single measurements. E, Seed of mutant phenotype by noninvasive NMR. F, The same seed but dissected by hand showing liquid endosperm (ev) and embryo (e). G, Mature seed with characteristic shrunken phenotype. H, Fresh weight accumulation of embryos (squares) and endosperm vacuole volumes (circles) in mutant seed. Data points represent single measurements. The orange hatched lines in D and H indicate the stages of seeds shown in A/B and E/F, respectively.

Retardation in Embryo Growth in the Mutant Seed Does Not Induce Starch/Protein Storage in the Endosperm

In both the wild-type and mutant seed, histological and immunological analysis showed that both the deposition of starch granules and the accumulation of storage protein are restricted to the embryo (Supplemental Fig. S2, A–D). Transmission electron microscopy (Supplemental Fig. S2, E and F) confirmed that only a small number of lipid droplets are accumulated in the endosperm and also identified the formation of multiple endoplasmic reticulum and Golgi structures, each surrounded by mitochondria. Together, these organs represent the subcellular machinery utilized by a biosynthetically active cytoplasm. Even when embryo growth and storage activity are retarded, the endosperm retains its characteristics as a transient organ and does not coopt any of the storage functions of the embryo.

Experimental Design for the in Vivo Imaging of Metabolites

Noninvasive ¹H-NMR localized spectroscopy was used to analyze the in vivo composition of the endosperm vacuole in both wild-type and mutant seeds. ¹H-NMR metabolites were detected in the endosperm by voxel-selective point resolved spectroscopy (PRESS; Bottomley, 1984). Suc, Ala, Gln, lactate, and Val all appear as well-resolved peaks (Fig. 4A). Localized two-dimensional correlation spectroscopy (L-COSY; Fig. 4, B and C) was used to clarify the overlapping spectral range (3.4–4.4 ppm). The L-COSY pulse sequence is given in Supplemental Fig. S3. ¹H-NMR chemical shift imaging (CSI) with high spatial resolution (100 × 100 × 100 μm³) was applied on intact seeds at an early stage of development to image metabolite distribution. Structural FLASH (for fast low angle shot) multislice images were acquired at the end of the CSI protocol in order to topographically relate the spectroscopic data with the corresponding tissue structures at each pixel of the image.

Figure 4.

Localized NMR spectra from the endosperm of wild-type seeds (approximately 260 mg fresh weight). A, PRESS spectrum. Resonances from the metabolites Suc, Ala, Gln, lactate (Lac), and Val are visible. B, L-COSY spectrum. The cross-peaks in the spectrum (indicated by the metabolite name labeling) confirm the assigned metabolites in A. The white square indicates the enlarged spectrum, which is presented in C. C, Enlarged spectrum from B. The cross-peaks detected in the range 3.2 to 4.2 ppm result from Suc. No other metabolites were detected in this spectral area.

The Spatial Distribution of Suc Differs from That of Amino Acids

Suc, Gln, and Ala are the major nutrients supplied to the zygote by the maternal plant. The distribution of Suc within the intact embryo sac at the early cotyledon stage showed that the highest concentrations are present in the central (adaxial) region of the cotyledon, decreasing toward its periphery (Fig. 5, A and B). Suc distribution within the endosperm vacuole is rather uniform. The lowest Suc levels coincide topographically with the suspensor. The distribution of Ala is quite distinct from that of Suc. The highest level is present in the endosperm and the suspensor, with very little accumulating in the cotyledon (Fig. 5, A and C). The distribution of Gln is identical to that of Ala (data not shown). Additionally, NMR spectra were obtained from three distinct regions of the liquid endosperm: one surrounding the embryo, one surrounding the micropyle, and the last surrounding the chalazal region (Supplemental Fig. S4). The acquired spectra did not differ significantly from one another, in accordance with the imaging data set. Thus, the liquid endosperm appears to be rather homogeneous, in contrast to the content of the cellularized embryo.

Figure 5.

Noninvasive mapping of metabolites in embryo sac of pea seed (approximately 50 mg fresh weight). A, Visualization of embryo sac by MRI shows positioning of seed coat, embryo, endosperm, and suspensor. Small boxes in pink show the region chosen for spectroscopic analysis (see right panels). B, Visualization of Suc distribution within the embryo sac. The bar represents the color code for specific signal intensity. C, Visualization of Ala distribution within the embryo sac. The bar represents the color code for signal intensity. Graphs at right show spectral characteristics of different regions of embryo sac: embryo, endosperm, and suspensor. e, Embryo; ev, endosperm vacuole; m, micropyle; s, suspensor; sc, seed coat.

NMR-Based Tools to Monitor Metabolites in the Intact Endosperm

The spectral parameters of the intact endosperm at the early and late stages of development differ markedly from one another (Supplemental Fig. S5), indicating that compositional changes do occur. As the purpose of these experiments was to quantify individual metabolite levels during development, the NMR signal amplitudes, acquired via the low-resolution CSI method, were converted into metabolite concentrations using a phantom replacement method (Soher et al., 1996). An external standard of known Suc, Ala, and Gln concentrations was measured using the in vivo protocol and used as a reference to correct the in vivo signal (for details, see “Materials and Methods”). For standardization purposes, the endosperm samples were isolated and analyzed biochemically immediately following the MRI experiment. The NMR in vivo Suc, Ala, and Glu signals were plotted against their concentrations obtained by the corresponding biochemical assay (Supplemental Fig. S6), and the resulting correlations were all high (Suc, r² = 0.99; Gln, r² = 0.93; Ala, r² = 0.63).

The Accumulation of Suc and Amino Acids in the Liquid Endosperm Follows Distinct Temporal Patterns during Development

The in vivo concentration of metabolites in the liquid endosperm of wild-type and mutant seeds was measured over the course of development (Fig. 6). The concentration of Suc increases steadily to approximately 250 mm until seed fresh weight reaches 150 to 170 mg and remains stable thereafter (Fig. 6A). The dynamics of Suc accumulation are similar in the wild-type and mutant seeds, but there are large differences in the accumulation of amino acids. During early development, both Gln and Ala levels increase (Fig. 6, B and C). However, from a seed fresh weight of 80 to 120 mg, both Gln and Ala levels decrease significantly in the wild type but continue to rise in the mutant. By the time seed fresh weight has reached 250 mg, the Gln and Ala concentrations in the wild-type endosperms are severalfold lower than in the mutant. Thus, Suc and amino acid levels in the endosperm appear to be under distinct control.

Figure 6.

Steady-state in vivo levels of metabolites in endosperm vacuoles of developing seeds measured by noninvasive NMR. A, Suc. B, Ala. C, Gln. WT, Wild type.

Sequences Encoding the Amino Acid and Suc Transporters Are Differentially Expressed in the Maternal and Filial Tissues of Individual Seeds

The accumulation of high concentrations of amino acids and Suc in the endosperm vacuole suggested that the uptake of assimilates in the endosperm is active. Amino acid permeases (such as AAP1 and AAP2; Tegeder et al., 2000) and Suc transporter/facilitator (SUF1 [Zhou et al., 2009] and SUT1 [Tegeder et al., 1999]), both of which are involved in assimilate transport in the legume seed, have been functionally characterized. However, these have all been localized in only the embryo and/or seed coat of pea, while their presence in the endosperm has not to date been investigated. To determine this, therefore, maternal seed coat and filial tissues (endosperm, embryo) were isolated using laser-assisted microdissection from seeds of approximately 35 mg fresh weight (Fig. 7, A–E), and the resulting RNA populations were analyzed by reverse transcription (RT)-PCR. The outcome of these experiments on individual deeds showed that PsAAP1, PsSUF1, and PsSUT1 transcripts are present in endosperm but not in embryo (Fig. 7F). PsAAP2 transcripts were observed at the limit of detection in the seed coat, endosperm, and embryo. Thus, the genes directing the transport of amino acids and Suc are up-regulated in endosperm tissue earlier than in the embryo.

Figure 7.

Tissue-specific analysis of transport-related genes using laser-assisted microdissection and RT-PCR of pea seed (approximately 35 mg fresh weight). A to C, Dissection procedure for embryo: before dissection (A), after dissection (B), and dissected embryo fragment (C). The red line in A indicates tissue chosen for dissection. D and E, Dissection of seed coat (D) and endosperm (E). The green lines indicate tissue chosen for dissection. F, Results of RT-PCR analysis of PsAAP1, PsAAP2, PsSUF1, and PsSUT1 genes expressed in endosperm, embryo proper, and seed coat. Amount of cDNA used per sample for PCR as a template was approximately 20 ng (see “Materials and Methods”). Actin mRNA abundance reflects approximate, but not exact, amount of cDNA template used for RT-PCR. e, Embryo; en, endosperm; sc, seed coat.

The Spatial Expression Pattern of Sut1 in the Endosperm Is Consistent with the Delivery of Suc from the Seed Coat

A Sut1-specific sequence was used as a probe for the in situ localization of Sut1 expression in the endosperm during the early stages of seed development. This showed that transcripts are more abundant in the endosperm than in the embryo (Fig. 8, A–D), in accordance with the RT-PCR outcome (Fig. 7F). The major zone of expression lies at the peripheral surface of the endosperm cytoplasm, with only a modest level of expression occurring in the endosperm cytoplasm attached to the embryo. Thus, the external and internal endosperm membranes appear to differ from one another functionally with respect to Suc transport (arrows in Fig. 8A). The localization of Sut1 expression to the embryo sac-endosperm interface implicates this region as being important for the transport of Suc from the maternal apoplast to the endosperm vacuole. To relate Sut1 expression to Suc levels in the endosperm, total Suc amounts were calculated from the product of its concentration and the vacuole volume (Fig. 8E). The up-regulation of Sut1 in the endosperm coincides temporally with the onset of Suc accumulation in the vacuole. Thus, the endosperm appears capable of acquiring Suc transport activity and begins to accumulate Suc before the embryo does. From the mid cotyledon stage onward, the expression shifts to the embryo (Tegeder et al., 1999; Borisjuk et al., 2002).

Figure 8.

Spatial expression pattern of Sut1 in endosperm of pea seed and accumulation of Suc in endosperm of wild-type phenotype and mutant. A, In vivo image corresponds to cross section through the seed. Expected direction of Suc flow is arrowed. Image is representative for approximately 30 to 50 mg fresh weight. B to D, In situ hybridization visualizes local up-regulation of Sut1 within the peripheral regions of endosperm (³³P labeling appears in black by bright-field illumination). Nonlabeled tissues are blue colored due to staining with toluidine blue. Arrows indicate nonlabeled region of endosperm. E, Amount of Suc accumulated in endosperm vacuole during seed development in mutant (mm) and wild-type (Mm) seed. Sucrose content was calculated by volume of endosperm liquid × Suc concentration. Arrows indicate temporal up-regulation of SUT-1 in either endosperm or embryo. ch, Chalaza; e, embryo; en, endosperm.

Endosperm Metabolite Levels Respond to the Onset of Storage Activity in the Embryo

An analysis of the temporal pattern of starch and protein accumulation in the embryo enabled the clarification of the relationship between assimilate level and storage activity. In the wild type, starch deposition begins at a low rate once the seed fresh weight had reached 120 mg but rises as seed fresh weight increases from 150 to 200 mg (Fig. 9A). The Suc concentration in the endosperm vacuole does not respond to the initiation of starch storage in the embryo, but the amount of Suc in the wild-type endosperm declines markedly from this time onward (Fig. 8E). In the mutant endosperm, Suc content remains high (Fig. 8E), corresponding to a much lesser amount of starch being stored in embryo (Borisjuk et al., 2002). However, the Suc concentration was similar to that in wild-type endosperm (Fig. 9A). Thus, the Suc reserves in the endosperm appear to be accessed by the growing embryo, but homeostatic mechanisms then come into play to control/balance its concentration.

Figure 9.

Steady-state level of metabolites in endosperm vacuole and accumulation of storage products in embryo during seed development. A, Levels of Suc measured by photometry in endosperm vacuole of wild-type (wt) and mutant seed and starch accumulation rate in wild-type embryo. B, Levels of Gln measured by HPLC in endosperm vacuole of wild-type and mutant seed and protein accumulation rate in wild-type embryo. C, ¹⁵N abundance in proteins after in vitro cultivation of wild-type embryos at different levels of Gln in the nutrition medium. Data points represent means ± sd of three to five measurements. FW, Fresh weight.

Protein storage was initiated in the wild-type seed once its fresh weight reached 100 to 150 mg (Fig. 9B), achieving a peak rate at 200 to 250 mg. The increase in protein storage activity coincides with a fall in the level of Gln (and Ala) in the endosperm vacuole. In mutant seeds, the accumulation of protein per embryo is much reduced (Borisjuk et al., 2002). Correspondingly, the endosperm vacuole level of Gln in the mutant seed remains high over a sustained period during development (Fig. 9B).

The developmental decline in the endosperm Gln (Ala) concentration coincides temporally with rising demands of the growing embryo and therefore may serve to limit the uptake of amino acids by the embryo, at the same time depressing the rate of storage protein deposition. In an in vitro experiment designed to determine the limiting endosperm level of Gln, intact wild-type embryos were incubated in the presence of various concentrations of ¹⁵N-labeled Gln (while the concentrations of Ala, Suc, and inorganic compounds were maintained at their in vivo levels). Increasing the available level of Gln from 10 to 60 mm enhanced the ¹⁵N signal in the protein fraction (Fig. 9C), but further increases had no effect. Because the in vivo endosperm Gln concentration declines to below 60 mm during development, the indication is that protein storage in the embryo becomes limiting from a seed fresh weight of 100 to 150 mg.

Metabolite Levels in the Endosperm in Response to the Environment

A series of experiments was designed to explore whether the level of nutrients present in the endosperm is sensitive to the external environment. The effect of nutrient starvation was tested by removing intact seeds (150 mg fresh weight) from the pod and using NMR to monitor sugar and free amino acid contents in the endosperm vacuole. Over a 4-h period, metabolite levels in the endosperm remained unaffected by the disconnection of nutrient supply (Supplemental Fig. S7). Next, the nutrient level present in a 100-mg seed fresh weight endosperm in the middle of the day was compared with that present in the middle of the night. The Suc concentration varied between 170 and 194 mm, and that of Gln between 64 and 75 mm, but critically, the level appears to be independent of time of day. Finally, detached seeds were exposed for 2 h to either 100% oxygen or 100% nitrogen to examine the effect of exogenous oxygen concentration on the endosperm nutrient level. This treatment has been shown to have a large effect on nutrient uptake and respiratory and storage metabolism of seed (Rolletschek et al., 2005b). However, neither hyperoxia nor anoxia materially altered the levels of Suc, Gln, or Ala (data not shown). Thus, the liquid endosperm that surrounds the developing embryo represents a medium where the nutrient level remains balanced even in the face of major changes in the external environment.

DISCUSSION

Recent research has identified the endosperm as an integrator of genetic programs controlling seed development (Berger et al., 2006) and has suggested that its metabolic role needs to be reinterpreted. Here, we have applied NMR to investigate the metabolism of the liquid endosperm in vivo within the developing pea seed. Our study argues for a metabolic role of the endosperm in legume seeds.

Advantages and Limitations of the Noninvasive Measurement of Endosperm Metabolite Levels

Physical access to the liquid endosperm is much more problematical than to the embryo (Borisjuk et al., 2003; Hill et al., 2003; Morley-Smith et al., 2008). A recognized disadvantage of invasive sampling is its induction of a wounding response, quite apart from the inevitable disruption of the spatial distribution of metabolites. The resulting artifactual concentrations of metabolites can be disturbed by 1 order of magnitude (for review, see Gifford and Thorne, 1985), and as a result, little effort has been made to systematically study metabolite levels in the liquid endosperm. MRI technology, however, provides a noninvasive means of measuring metabolite concentrations (Bourgeois et al., 1991; Soher et al., 1996; Vanhamme et al.,1997; Tkác et al., 1999; Pohmann and von Kienlin 2001; De Graaf, 2007).

The use of NMR to track some of the major seed metabolites has a number of advantages. First, the seed remains undamaged by the process of sample preparation, so wound responses are avoided; second, the use of a CSI method with a direct free induction decay acquisition and a long repetition time allows for metabolite quantification with a minimal requirement for postprocessing correction; and finally, the acquisition and processing procedure is simple and robust, because only one pulse and a phase-encoding gradient are needed for signal encoding. Some plant tissues, however, are refractory to MRI (Ratcliffe and Shachar-Hill, 2001; Köckenberger et al., 2004), and the method appears particularly unsuited for the analysis of low-concentration metabolites. Achieving good levels of sensitivity in small seeds will need modification of currently available coils (Neuberger and Webb, 2009).

In Vivo Imaging Reveals an Absence of Metabolite Gradients in the Liquid Endosperm But Their Presence in the Embryo

The NMR analysis captured the in vivo state of the seed at a distinct phase of its development. In particular, it was possible to identify the presence of a Suc gradient within the embryo, falling from a high level in the central (adaxial) part of the cotyledon to a low level at the periphery. This gradient corresponded spatially with the differentiation pattern of the embryo (Hauxwell et al., 1990; Borisjuk et al., 2002, 2003). Differentiation of the endosperm has been described at both the cytological (Boisnard-Lorig et al., 2001) and gene expression (Fig. 8; Tanaka et al., 2001; Bate et al., 2004) levels. However, the endosperm vacuole appears to be rather homogeneous with respect to Suc, Ala, and Gln (Fig. 5; Supplemental Fig. S4), suggesting that it represents a large homogenous compartment. In contrast to what pertains in the embryo tissue, equilibration within the endosperm vacuole can take place rapidly, thus allowing for an efficient transfer of nutrients from the source (seed coat) to the sink (embryo). This scenario may well be advantageous in the context of the seed's energy economy (Borisjuk and Rolletschek, 2009) and may favor metabolite exchange between two generations (Lersten, 2004).

The suspensor, which connects the embryo to the mother plant (Fig. 2; Supplemental Movies S1 and S2), contains similar Ala/Gln levels as the endosperm, but its level of Suc is substantially lower (Fig. 5B). Thus, the embryo is exposed to a particular Suc level at its attachment region to the suspensor. Since Suc can act as a signal molecule (Chiou and Bush, 1998), this may locally affect cell differentiation (Rolland et al., 2006) in concert with the suspensor's function in embryo patterning, as postulated from in vitro studies (Supena et al., 2008).

The Role of Suc in Determining the Homeostatic Properties of the Endosperm

Suc uptake is responsible for approximately 70% of the dry matter in the mature pea seed (Patrick and Offler, 2001). Its rate is determined by the activity of Suc transporters (Tegeder et al., 2000; Rosche et al., 2002), which in turn is largely controlled by the embryo's carbohydrate demand (Zhou et al., 2009). The up-regulation of Sut1 in the embryo, which is primarily localized within the outer regions of the cotyledons, is coupled with an increase in assimilate sink strength (Borisjuk et al., 2002). Our data have demonstrated that Sut1 is also expressed in the endosperm, where its up-regulation anticipates its expression in the embryo, and that its expression is largely confined to the region adjacent to the seed coat. The endosperm's surface area appears rather larger than that of its partner embryo (Fig. 2A). Sut1 expression in the endosperm is followed by a rapid accumulation of Suc in the endosperm vacuole. Even though similar events take place in the embryo, the endosperm grows much more rapidly than the embryo does. Thus, it is reasonable to suppose that the endosperm is predetermined as a primary sink for Suc and follows a similar strategy in terms of Suc uptake as does the embryo. A defective mutation in the Arabidopsis endosperm-specific Suc transporter AtSUC5, which promotes the rate of Suc transport into the endosperm, has been associated with a delay in embryo development (Baud et al., 2005). We have shown here that the amount of Suc in the endosperm is inversely proportional to the rate of embryo growth and storage: starting from approximately 150 to 200 mg seed fresh weight (“equilibrium point”), the total amount of Suc in endosperm decreases (Fig. 8E), whereas starch accumulation rate in embryo increases (Fig. 9A). This indicates the role of the endosperm as a carbohydrate source for the embryo.

In contrast to the total amount of Suc present in the endosperm vacuole, its steady-state level is not affected by the growth of the embryo (Fig. 9A). The temporal gradient of Suc suggests that its steady state in the endosperm vacuole is developmentally fixed and is relatively unaffected by short-term starvation (Supplemental Fig. S7), diurnal switches, or external atmospheric conditions, unlike the phloem and seed coat apoplast, which deliver Suc from maternal tissues (Geiger and Servaites, 1994; Tuan et al., 1997; Kallarackal and Komor, 1998; Munier-Jolian and Salon, 2003; Kang et al., 2007). Thus, it is the endosperm itself that exerts control over the Suc concentration in the endosperm vacuole. The amount of Suc stored in the vacuole is higher than might be predicted on the basis of the demand of the developing embryo (Zhou et al., 2009) and can support embryo growth for over 20 h if Suc delivery is cut off. Thus, transient variation in the external environment can be buffered to provide the homeostasis necessary for embryo growth. Suc can be sensed by embryo cells (Rolland et al., 2006) and regulates both gene expression and embryo development (Iraqi and Tremblay, 2001; Boyer and McLaughlin, 2004). With Suc acting simultaneously as both a growth substrate and a signal molecule, the endosperm is able to sustain the growth of the embryo even in the face of transient variation in the external environment.

The in Vivo Steady-State Level of Free Amino Acids in the Endosperm Vacuole

Free amino acids represent the major source of nitrogen for the growing embryo (Tegeder et al., 2000). Amino acid uptake by the embryo is in part passive, especially during its early development (DeJong et al., 1997), so that a favorable concentration gradient needs to be established between the embryo and the endosperm. Our in vivo data have demonstrated the existence of such a gradient. The endosperm up-regulates the expression of amino acid transporters earlier in development and grows more rapidly than does the early embryo, after which it accumulates Ala and Gln in its vacuole (Fig. 6). Thus, a transient internal source of nitrogen was generated in the embryo sac, in the form of a liquid medium enriched with amino acids (Fig. 5C). In this situation, the passive uptake of amino acids by the embryo is encouraged and may indeed represent an essential component of embryo nutrition during its early development.

The switch to an active uptake system occurs rather late in development and is based on enhanced levels of AAP1 (an amino acid permease) and PTR1 (a peptide transporter) in the embryo proper (Miranda et al., 2001). This transport system appears to be controlled by assimilate availability (Bennett and Spanswick, 1983). Protein synthesis appears to be nitrogen limited, with the embryo's nitrogen uptake activity via AAP1 being rate limiting for storage protein synthesis (Rolletschek et al., 2005a; Sanders et al., 2009). Here, we have explored the time point at which protein storage becomes nitrogen limited and how this relates to the level of the majority amino acids present in the endosperm vacuole. We have shown that the levels of Ala/Gln gradually increase in the wild-type endosperm until the developmental stage reached by a seed of fresh weight 100 to 150 mg. The subsequent decline coincides temporally with an increase in the rate of protein accumulation in the embryo. In contrast, in the mutant seeds, the levels of Ala and Gln in the endosperm vacuole remain high (Fig. 6, B and C), presumably reflecting a much lower demand for amino acids by the embryo. Our current model suggests that during early development, when the protein accumulation rate and sink strength of embryo are negligible, amino acid demand is exceeded by its supply. However, as soon as protein storage is initiated in the embryo, the balance between demand and supply of amino acids is altered and an exhaustion of reserves leads to a decline in the amino acid concentration in the endosperm vacuole. Later, the control of the steady-state amino acid level in the endosperm becomes demand driven.

A concentration of 60 mm Gln is associated with a peak Gln incorporation rate (Fig. 9C), but the endosperm Gln concentration remains below this threshold during most of the seed maturation process (Fig. 9B). Thus, protein storage appears to be nitrogen limited during the entire main storage phase. The low availability of Gln in the endosperm vacuole is likely insufficient to satisfy the requirements of the embryo during the mid to late developmental stages. Biochemical approaches attempting to increase levels of seed protein, therefore, should focus on increasing the amino acid content of the endosperm.

The Transient Metabolic Role of the Endosperm

The coordination of seed development clearly requires feedback between the filial and the maternal tissue (Garcia et al., 2005). The signaling of seed size in a range of species emanates not from the embryo but rather from the endosperm (Felker et al., 1985; Maitz et al., 2000; Garcia et al., 2003). The fact that a retardation in embryo growth does not affect seed size in the pea mutant confirms this notion. The question that arises is how this control can be metabolically mediated. The filial Suc and amino acid pool sizes are clear candidates as metabolic signals, because these respond to altered rates of solute delivery and storage activity and are known to regulate symporter gene expression via derepression (Patrick and Offler, 2001). The endosperm is more advanced than the embryo with respect to the accumulation of soluble metabolites and the up-regulation of corresponding transporters. In the pea mutant, the filial pool is largely confined to the endosperm, which accumulates severalfold more soluble metabolites than does the wild-type one. Thus, the endosperm is able to unload nutrients from the seed apoplasm independently of the size of the embryo sink. Solute withdrawal from the seed apoplasm can be sensed by a turgor-homeostat mechanism located in maternal seed tissue, which acts to balance effluxer activity as required (Patrick and Offler, 2001). The growth of the endosperm, therefore, may trigger a wild-type-like feedback signal to the maternal tissue, which could be transmitted via a calcium signaling cascade (Zhang et al., 2007) to drive cell elongation in the seed coat of the mutant. This mechanism would allow the endosperm to retain the coordination of seed development despite retarded embryo growth. This is attributed to the endosperm-mediated metabolic control of seed growth and reflects the evolutional predetermination of endosperm: to favor nutrient transfer from the maternal to the filial generation (Sundaresan, 2005).

The frequent observations that the endosperm can drive seed growth and that a retarded endosperm is associated with reduced embryo growth (Chaudhury et al., 2001; Choi et al., 2002; Garcia et al., 2003; Ingouff et al., 2006) have been taken to suggest that nutrients are delivered from the endosperm to the embryo. However, a body of evidence suggests that embryo is autonomous (Weijers et al., 2003; Ungru et al., 2008; Pignocchi et al., 2009) and that metabolite delivery to the embryo bypasses the bulk endosperm (Yeung and Meinke, 1993; Stadler et al., 2005; Morley-Smith et al., 2008). According to our in vivo data, the extent of the trophic subordination of the embryo to the endosperm may alter during the course of seed development. During its early stages, the embryo is physically connected to the maternal seed coat via the suspensor (Supplemental Movies S1 and S2), so it is not possible to exclude the possibility of this alternative means of metabolite delivery to the embryo. Later, following the destruction of the suspensor, the endosperm blocks any direct contact between the maternal tissue and the embryo, so it must be involved in any metabolite exchange. The endosperm accumulates Suc and amino acids in its vacuole (Fig. 6), and these represent the embryo's sole source of nutrition. The growth of the embryo is affected when the metabolism of the endosperm is compromised (Baud et al., 2005; Ohto et al., 2005; Kondou et al., 2008). This behavior demonstrates the trophic subordination of the embryo to the endosperm. Here, we have shown that the endosperm affords the metabolic environment of the embryo and mediates seed coat growth, at a stage before the embryo's nutrition can be ensured by its own sink strength, and until cell division, tissue differentiation, endopolyploidization, and storage have all been completed. Reciprocal crosses have been shown to express the same extent of maternal dominance over the determination of seed size (Lemontey et al., 2000).

Proteomic and transcriptomic analyses in M. truncatula have indicated that, as the embryo acquires its own sink, so the endosperm's metabolism shifts from a highly active to a quiescent state (Gallardo et al., 2007). At this stage, the transient nature of storage in the dicotyledonous endosperm is crucial: the endosperm deposits soluble metabolites, which are neither preserved nor returned to the maternal organism. Instead, the embryo utilizes these reserves, and, as it grows, it takes over the physical space previously occupied by the endosperm. A transient endosperm, therefore, allows the embryo to avoid competition for nutrition and space within the maternal organism, which could otherwise restrict embryo growth under stressful conditions or affect gene expression as a result of mechanical pressure (Weijers et al., 2003; Desprat et al., 2008). Therefore, the endosperm largely accomplishes its function before seed maturation. Its job complete, the liquid endosperm then disappears. As Schiller puts it in his play The Genoese Conspiracy, “the Moor has done his duty, the Moor can go” (Schiller, 1799).

MATERIALS AND METHODS

Plant Material

Pea (Pisum sativum ‘Erbi’) and E2748 mutant (Johnson et al., 1994) plants were grown under a 16-h-light, 19°C/8-h-dark, 16°C regime. Because homozygous E2748 mutant seeds are lethal, plants were produced by embryo rescue. For this purpose, the developing embryo of a homozygous mutant seed (identified by phenotype, following Johnson et al. [1994]) was hand dissected, germinated, and grown under sterile conditions until a shoot was formed, at which point the seedling was transferred to soil. The wild-type phenotype was restored by transferring pollen from a wild-type plant.

Sampling and Biochemical Analysis

Endosperm liquid was isolated by microsyringe (two endosperm samples per individual seed). Set volumes (3–10 μL) were added to 200 μL of 80% ethanol and processed as described elsewhere (Borisjuk et al., 2002). Free amino acids were measured by HPLC, and soluble sugars were measured by a coupled photometric assay (Rolletschek et al., 2005a, 2005b). Starch was measured enzymatically (Rolletschek et al., 2002). Total nitrogen was assessed from powdered seed samples (dried overnight at 70°C) by elemental analysis (Rolletschek et al., 2002). Crude protein was calculated from total nitrogen × 6.25. Accumulation rates were calculated from measured starch/protein contents and the embryo weight. Fresh and dry weight accumulation rates were derived by weighing the seed before drying and removing the embryos before reweighing.

Immunoassay

Seeds were fixed in either 2.5% (v/v) glutaraldehyde, 50 mm sodium cacodylate buffer (pH 7) or 4% (w/v) paraformaldehyde, 50 mm potassium phosphate buffer (pH 7) under a mild vacuum for 4 h at room temperature, rinsed in cacodylate buffer, dehydrated, and embedded in butyl-methyl methacrylate by polymerization at 20°C for 48 h under UV light. Sections of 3 to 5 μm thickness were cut on a microtome. The immunolocalization procedure was performed using an affinity-purified anti-legumin polyclonal antibody and the corresponding VASTASTAIN ABC-AP kit (Alkaline Phosphatase Substrate Kit III) and evaluated microscopically.

Electron Microscopy

Seed sections were immersed in 2% (w/v) paraformaldehyde, 0.5% (v/v) glutaraldehyde in 50 mm potassium phosphate buffer, 5 mm EGTA (pH 7.2), 5 mm CaCl₂, and 3% (w/v) Suc. After 2 h at room temperature, fixation was continued overnight on ice using freshly prepared solution. All subsequent steps were performed as described elsewhere (Borisjuk et al., 2002).

Laser Microdissection, RT-PCR Analysis, and in Situ Hybridization

Seeds were frozen at −80°C, cut into 10-μm sections at −20°C, mounted onto PET membrane-covering glass slides (Carl Zeiss MicroImaging), and lyophilized at −20°C. The sections were subjected to laser-assisted microdissection using a PALM Laser-Microbeam (Bernried) device. Small tissue samples were collected using laser pressure catapulting and larger ones by a hand-operated microneedle. Poly(A) mRNA was extracted directly using the Dynabeads mRNA DIRECT kit (Dynal Biotech) using the manufacturer's protocol. Linear amplification and cDNA synthesis were carried out from approximately 1 ng of mRNA using the ExpressArt mRNA Amplification Nano kit (AmpTec) according to the manufacturer's protocol. PCR was based on a template of 20 ng of cDNA in a 25-μL volume containing 0.2 μm gene-specific primers, 200 μm desoxyribonucleotide triphosphate (Pharmacia), 0.5 units of Taq polymerase (Roche), and 1× polymerase buffer supplied with the enzyme. The primer sequences used and predicted amplicon sizes are given in Supplemental Table S2. The PCR regime consisted of a 94°C incubation for 2 min followed by 30 cycles of 94°C for 30 s, 57°C for 30 s, and 72°C for 60 s. PCR products were separated by 0.8% agarose gel electrophoresis and visualized by ethidium bromide staining.

cDNA labeling and in situ hybridization were carried out as described by Borisjuk et al. (2002). The complete cDNA of Vicia faba SUT1 (Weber et al., 1997) was used as probe after labeling with [³³P]dCTP.

In Vitro Incubation of Pea Embryos

Freshly isolated, intact embryos were transferred into 25-mL vials containing an incubation buffer consisting of 250 mm Suc, 80 mm Ala, 20 mm KCl, 4 mm CaCl₂, 2 mm K₂SO₄, 2 mm KH₂PO₄, 3 mm MgSO₄ and 10 mm MES/KOH (pH 5.6). Subsequently, ¹⁵N-labeled Gln (Campro Scientific) was added to reach final concentrations of 10, 30, 60, 90, 120, and 150 mm. Osmolality was held constant by the addition of mannitol. The vials were incubated in light (25 μE) with gentle agitation. After 20 h, the embryos were rinsed twice in distilled water and immediately frozen. The protein faction was extracted, and the ¹⁴N/¹⁵N isotope pair was analyzed using elemental analysis coupled to isotope ratio mass spectrometry (Borisjuk et al., 2007).

NMR Protocol

Instrument

The experiments were performed on a 17.6-T wide-bore (89 mm) superconducting magnet (Bruker BioSpin) equipped with actively shielded imaging gradients. A 200 mT m⁻¹ gradient system and a 15-mm birdcage resonator were used for the identification and quantification of metabolites. High spatial resolution spectroscopic imaging experiments were performed with a 1 T m⁻¹ gradient system and a 20-mm birdcage coil.

Plant Material

For high-resolution metabolite imaging, seeds were detached from the pod and immediately immersed in buffer (as used for in vitro incubation of embryos but lacking Suc, Ala, and Gln). For all other experiments, the seed was detached from the pod and presented to the instrument within minutes. During the period of measurement, the seed was kept in the dark at 20°C and ambient atmosphere.

Metabolite Identification

Localized one-dimensional PRESS (repetition time [TR] = 2 s, echo time [TE] = 20 ms, voxel = 1 mm³, no. of averages [NA] = 32) and L-COSY (Thomas et al., 2001; TR = 2 s, TE = 10 ms, voxel = 1.5 mm³, acquired points in the indirect spectroscopic direction [t1] increments = 1,024, NA = 1) were used for metabolite identification.

T1 Measurement, Quantification, and Metabolite Mapping

T1 measurements were performed by extending the CSI sequence with a nonselective adiabatic inversion pulse (hyperbolic secant pulse; 3 ms), a variable inversion time (TI) before signal excitation, and a 10-s recovery delay after signal acquisition. The experiments were performed for five different TIs (nonlinearly distributed between 98 and 5,000 ms). To avoid side-lobe effects from the spatial response function and contamination from surrounding areas, an acquisition-weighted k-space sampling (hanning) was applied (Pohmann and von Kienlin, 2001). The parameters for the T1 CSI protocol were as follows: TR = 10 s, number of total scans (NS) = 100, field of view (FOV) = 8 × 8 mm², spatial resolution = 1 × 1 mm² in plane, 0.2-mm slice thickness, spectral = 1,024 points, bandwidth = 8 kHz. The VAPOR (for variable pulse power and optimized relaxation delays) suppression scheme using frequency-selective hermite pulses of 300 Hz bandwidth (Tkác et al., 1999) was used to ensure water suppression. The T1 values for the metabolite resonances at various developmental stages are given in Supplemental Table S1. A low-spatial-resolution CSI method with a long repetition time (TR = 10 s), NS = 200 (hanning weighted), FOV = 8 × 8 mm², spatial resolution = 0.8 × 0.8 mm² in plane, 1-mm slice thickness, spectral = 2,048 points, bandwidth = 8 kHz, acquisition delay = 1 ms, VAPOR water suppression (hermite pulses; bandwith = 200–600 Hz) was employed to quantify metabolites in the endosperm. The duration of the spectroscopic imaging experiment was 33 min, 20 s. The high-spatial-resolution CSI experiment was performed with a 1 T m⁻¹ gradient system, using the following experimental parameters: TR = 1.5 s, NS = 27,500 (hanning weighted), FOV = 5 × 5 mm², spatial resolution = 100 × 100 μm² in plane, 100-μm slice thickness, spectral = 2,048 points, bandwidth = 8 kHz, acquisition delay = 2 ms, VAPOR water suppression (hermite pulses; bandwith = 500 Hz). The total duration for the spectroscopic imaging experiment was 11 h, 27 min.

Single Voxel Spectroscopy

Changes in metabolite concentration in the pea endosperm were observed over 6 h using voxel-selective PRESS (TR = 2 s, TE = 20 ms, voxel size = 1 mm³, NA = 128). The experiments were performed under nitrogen or oxygen (premixed gas bottles; purity of 99.999%) by aerating the seeds via a gas tube mounted within the NMR instrument. This treatment induced almost immediate changes in the oxygen level inside the coil, as checked by oxygen microsensors (Presens; for details, see Rolletschek et al., 2005b). An oxygen level of less than 2 μm is regarded as anoxic conditions.

Morphological Imaging

The structural and morphological imaging of the seeds was achieved using gradient-echo multislice images (FLASH) with the following parameters: TR = 500 ms, TE = 30 ms, FOV = 10 × 10 mm², matrix = 128², slice thickness = 0.5 mm, NA = 4, total acquisition time = 1 min, 42 s. Three-dimensional data sets were acquired using a three-dimensional spin-echo sequence with TR = 1 s, TE = 7.7 ms, FOV = 20 × 7.5 × 7.5 mm³, matrix = 256 × 96 × 96, isotropic resolution = 78 μm³, total measurement time = 2 h, 33 min for a 120-mg seed and TR = 1 s, TE = 6.9 ms, FOV = 20 × 7 × 7 mm³, matrix = 336 × 116 × 116, isotropic resolution = 60 μm³, total measurement time = 3 h, 44 min for 30- and 73-mg seeds.

Postprocessing and Quantification

To estimate the metabolite concentration in the endosperm, an external 5-mm-diameter reference tube containing 50 mm aqueous Suc, Glc, Gln, or Ala was measured by the same spectroscopic imaging protocol used for the in vivo measurements. Signal correction between reference and in vivo samples was based on Soher et al. (1996) and DeGraaf (2007). The long repetition time (TR = 10 s) of the CSI experiment enabled sufficient longitudinal spin relaxation, so correction factors for partial saturation were not required. The repetition time was more than 5-fold the longest metabolite T1 (Supplemental Table S1). No B1 field correction was applied, because the 15-mm birdcage resonator had a homogenous B1 distribution in phantom measurements. The in vivo metabolite concentration (KI) was calculated from the known external reference metabolite concentration (KR) using:

where S_I and S_R are the in vivo (I) and reference (R) signal amplitudes acquired with the chemical shift imaging method. The correction factor for the different coil loading between the in vivo seed and reference (C_LOAD) was involved in the quantification process (Soher et al., 1996). C_LOAD can be estimated from the difference in the radio frequency (RF) power:

where ΔA is the difference in power attenuation for the 90° pulse between the two experiments, measured in decibels. The accuracy of C_LOAD was checked using eight water phantoms with different solute fractions of sodium chloride (0.1%–20%). The attenuation required to achieve a 90° pulse and the water signal amplitude were measured for each phantom.

Data Processing

Raw spectroscopic imaging data were preprocessed using an in-house computer program written in MATLAB 7.0.1 (MathWorks). The NMR signal was analyzed with the interactive time domain fit algorithm AMARES (Vanhamme et al., 1997) within the jMRUI software package (Naressi et al., 2001).

MRI Calibration

To ensure the reliability of the method, an external calibration procedure was applied. Individual seeds at selected developmental stages were analyzed under identical experimental conditions. Immediately after the MRI, endosperm samples were isolated by microsyringe, their free amino acid content was derived by HPLC, and their soluble sugar content was derived by photometry. The MRI signals were plotted against the measured metabolite concentrations.

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers AY956395, AY956396, DQ221698, AF109922, and X90378.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure S1. Phenotype of mutant E2748.

• Supplemental Figure S2. Distribution of storage product accumulation in embryo versus endosperm in mutant seed.

• Supplemental Figure S3. Schematic representation of the L-COSY pulse sequence (Thomas et al., 2001) for the application on pea seed.

• Supplemental Figure S4. Localized PRESS spectra from different regions of the endosperm of wild-type seed (approximately 120 mg fresh weight)

• Supplemental Figure S5. Localized CSI spectra from the endosperm of wild-type seeds at two different developmental stages.

• Supplemental Figure S6. Correlation of noninvasive measurement with data derived from conventional assays in dissected samples of endosperm.

• Supplemental Figure S7. Compositional changes of liquid endosperm during the starvation experiment measured by NMR.

• Supplemental Table S1. T1 relaxation times of metabolites in the pea endosperm at 17.6 T.

• Supplemental Table S2. Primers used in RT-PCR analyses.

• Supplemental Movie S1. Internal structure of seed by three-dimensional NMR imaging (corresponds to Fig. 1, D and G).

• Supplemental Movie S2. Internal structure of seed by three-dimensional NMR imaging (corresponds to Fig. 1, E and H).

• Supplemental Movie S3. Internal structure of seed by three-dimensional NMR imaging (corresponds to Fig. 1, F and I).

• Supplemental Movie S4. Digital model of three-dimensional anatomy of a pea seed (corresponds to Fig. 2B).