Gradients of lipid storage, photosynthesis and plastid differentiation in developing soybean seeds


Author for correspondence: Hardy Rolletschek Tel: +49 39482 5686 Fax: +49 39482 5500 Email:


  • • This study establishes a topographical framework for functional investigations on the regulation of lipid biosynthesis and its interaction with embryo photosynthesis in developing soybean seed.
  • • Structural observations, combined with molecular and functional parameters, revealed the gradual transformation of chloroplasts into storage organelles, starting from inner regions going outwards. This is evidenced by electron microscopy, confocal laser scanning microscopy, in situ hybridization and histochemical/biochemical data.
  • • As a consequence of plastid differentiation, photosynthesis becomes distributed along a gradient within the developing embryo. Electron transport rate, effective quantum yield and O2 production rate are maximal in the embryo periphery, as documented by imaging pulse-amplitude-modulated fluorescence and O2 release via microsensors. The gradual loss of photosynthetic capacity was accompanied by a similarly gradual accumulation of starch and lipids. Noninvasive nuclear magnetic resonance spectroscopy of mature seeds revealed steep gradients in lipid deposition, with the highest concentrations in inner regions.
  • • The inverse relationship between photosynthesis and lipid biosynthesis argues against a direct metabolic involvement of photosynthesis in lipid biosynthesis during the late storage stage, but points to a role for photosynthetic oxygen release. This hypothesis is verified in a companion paper.


Oilseed crops such as soybean (Glycine max) accumulate large amounts of lipids, accounting for a significant portion of the global food and feed supply. Much effort has been devoted towards the identification of appropriate targets for biotechnological engineering of seed oils (Cernac & Benning, 2004; Hobbs et al., 2004). So far, however, these approaches have been of limited success (Thelen & Ohlrogge, 2002), underlining the importance of a more comprehensive view on lipid biosynthesis.

Storage lipids are mainly composed of triacylglycerols, synthesis of which takes place in two distinct cellular compartments. The production of acyl chains from acetyl CoA occurs within plastids, the subsequent lipid assembly being at the endoplasmic reticulum and deposition in oil bodies (Ohlrogge & Browse, 1995). Although the metabolic pathways and the genes involved are now well known (e.g. Beisson et al., 2003), the regulation of oil biosynthesis and concomitant accumulation of storage lipids are still poorly understood (Thelen & Ohlrogge, 2002), as are the factors determining the allocation of carbon to either oil or other storage products (Rawsthorne, 2002).

One important question is the contribution and/or regulatory function of seed photosynthesis. Light has a clear stimulatory effect on storage activity (Fader & Koller, 1985; Willms et al., 1999). Using inhibitors, Fuhrmann et al. (1994) showed that this effect is directly linked with photosynthetic activity within seeds. However, the physiological significance of the distinct photosynthesis-related processes remains unclear, and several hypotheses have been postulated. (1) Studies on 14CO2 uptake and carbon balances suggest that photosynthesis within seeds and fruiting structures predominantly refixes carbon derived from respiration and metabolism of the seed itself (Quebedeaux & Chollet, 1975; King et al., 1998; Aschan & Pfanz, 2003; Schwender et al., 2004), thus improving carbon efficiency. (2) As plastidic synthesis of fatty acids requires high amounts of ATP and reducing equivalents, photosynthesis was proposed as an energy supply (Asokanthan et al., 1997; Willms et al., 1999; Ruuska et al., 2004). However, due to the low light supply to the seeds, it has been questioned whether photosynthesis can play this role under in vivo conditions (Eastmond & Rawsthorne, 1998). (3) Redox signals mediated via the ferredoxin/thioredoxin cascade might modulate the activity of biosynthetic enzymes (Ruuska et al., 2004). (4) Finally, it has been proposed that photosynthetic oxygen release can promote storage activity of seeds under internal hypoxia (Rolletschek et al., 2003). This might be particularly important for oilseeds (Vigeolas et al., 2003).

Studies on the functional relationship between photosynthesis and lipid storage in seeds requires a detailed knowledge of the temporal as well as spatial patterns of these processes. Unfortunately, reports on the topography of photosynthesis, as well as lipid storage in oilseeds, are completely lacking. This gap considerably hampers the furtherance of our understanding on the regulation of storage lipid biosynthesis in developing seeds.

The overall motivation for this work was to investigate the topographical relationship of photosynthesis and lipid biosynthesis in developing soybean seeds as a framework for further functional studies. Toward this end, we combined a topographical analysis of photosynthetic activity across tissues [using pulse-amplitude-modulated (PAM) fluorescence and O2-sensitive microsensors] with ultrastructural studies, and spatial expression analysis (in situ hybridization) of plastidic metabolite transporters. In addition, a new nuclear magnetic resonance (NMR)-based technique was developed to visualize gradients in lipid deposition within seeds. Our results show a functional shift in plastid differentiation from chloroplasts to storage organelles, starting from the interior of the embryo towards its periphery. This shift is accompanied by the loss of photosynthetic capacity and the gain of storage activity. The inverse relationship between photosynthesis and lipid biosynthesis/deposition argues against a direct metabolic involvement of photosynthesis in lipid biosynthesis (via redox modulation or energy supply), but points to a role for photosynthetic oxygen release. This hypothesis was tested and verified in the second part of this study (Rolletschek et al., 2005).

Materials and Methods

Plant growth

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, snap-frozen in liquid N2 and stored at −80°C.

Chlorophyll fluorescence imaging

Images of chlorophyll fluorescence parameters were obtained using an IMAGING-PAM chlorophyll fluorometer (Heinz Walz GmbH, Effeltrich, Germany). The device was equipped with the mini-measuring head (sample area 7 × 9 mm). Seeds were sliced and preilluminated for 8 min at an irradiance of 212 µmol quanta m−2 and s−1 blue light (peak 470 nm) so that steady state photosynthesis was achieved. Afterwards, actinic irradiances in the range 11–212 µmol quanta m−2 and s−1 were applied for 30 s, starting with the highest light intensity. Images of the steady state chlorophyll fluorescence (Ft) were acquired under actinic illumination, and maximum fluorescence yield (Fm′) was measured during 800 ms exposure to saturating light intensity. The data obtained were used to create rapid light-response curves which gave insight into the light-saturation properties of the object under investigation (White & Critchley, 1999). Images of the effective quantum yield of PSII (ΦPSII) were created from the digitized images of Ft and Fm′ by computing pixel-by-pixel and according to the equation: ΦPSII = (Fm′ − Ft) : Fm′. Such images of ΦPSII can be interpreted as a map of the relative photosynthetic electron transport rate (ETR), as ΦPSII is close to the overall quantum yield of photosynthesis (Genty et al., 1989). The ETR was calculated for selected sample areas as ΦPSII × PPFD × 0.5 × 0.84 (where PPFD = photosynthetic photon flux density). The size of the sample areas was 0.1 mm2.

Measurement of photosynthetic O2 release

Using oxygen-sensitive optical microsensors (Presens GmbH, Neuburg, Germany), gross photosynthetic rates were estimated as the rate of decrease in oxygen concentration within the seed tissue during the first few seconds following the extinction of light (for methodological details see Rolletschek et al., 2002a).

Cloning and sequencing of plastidic phosphate translocators

To clone three members of plastidic phosphate translocator (PT) gene families (glucose phosphate translocator, GPT; triose phosphate translocator, TPT; phosphoenolpyruvate/phosphate translocator, PPT), mRNA from young cotyledons of soybean was purified and cleaned using the Dynabead kit (Dynal Biotech, Oslo, Norway). First-strand cDNA was synthesized using the SuperScript II RNase H_RT kit (Invitrogen, Karlsruhe, Germany). Subsequently, fragment amplification was carried out using PCR with specific primers for GPT, TPT and PPT, respectively. The following primers were used: GmTPTf 5′-TGGGACTTCCTAAACGTGCT-3′; GmTPTr 5′-GTGCAAGACATCCTCCAGGT-3′; GmPPTf 5′-CTCTCCGCAACCCTTCCCCTAACT-3′; GmPPTr 5′-TAGCCACAGGTGCTAACAAGAA-3′; GmGPTf 5′-GCTTGAATTTTTCGCCTACCC-3′; GmGPTr 5′-GCCTGTGAATACAAGGAAGGTTCC-3′, for TPT, PPT and GPT, respectively). Amplification conditions were as follows: 3 min at 96°C; 30 cycles of 94°C for 30 s, 55°C for 60 s, and 72°C for 60 s, followed by incubation for 8 min at 72°C. PCR products were separated on a 1.0% (w/v) agarose gel and stained with ethidium bromide. The cDNA fragments were gel purified and subcloned into a TOPO vector kit (Invitrogen) for sequencing. DNA sequences were determined by the dideoxynucleotide chain-termination method. The sequencing data analysis was performed using lasergene software (DNA Star Inc., Madison, WI, USA). Homology search was performed using blastx (Altschul et al., 1997) against the National Center for Biotechnology Information (NCBI) public database for annotated functions of genes.

The deduced amino acid sequence of the GmPPT (GenBank accession number AY942815) showed high similarity to the PPT amino acid sequence from rice (78%, AAF86907); tobacco (75%, AAB40648); and Arabidopsis thalianaPPT2 (60%, AAF01540). The GmPPT also possessed conserved domains of the integral membrane protein DUF6 (Marchler-Bauer et al., 2005). The putative amino acid sequence of GmGPT (GenBank accession number AY942816) showed 88% identity to GPT of Pisum sativum (AF020814); 74.5% to ArabidopsisGPT1 (BAB08759); 71% to ArabidopsisGPT2 (AAC28500); and 71% to GPT of Zea mays (AF020813). The putative GmTPT (GenBank accession number AY942814) was aligned with deduced protein sequences of other known members of the TPT family and revealed similarity up to 73% with TPT of rice (AP003210); 89% with TPT of Pisum stativum (X68077); and ≈30% to other members of the TPT subfamily.

RNA isolation and hybridization procedure

Isolation of RNA, cDNA fragment labelling and Northern hybridization were performed as described by Heim et al. (1993). Radiochemicals were purchased from Amersham–Pharmacia (Freiburg, Germany). The cDNA fragments were used as probes after labelling with 32P-dCTP. RNA gel-blot hybridizations were carried out according to Church and Gilbert (1984): cDNA of GPT (1200 bp), TPT (798 bp) and PPT (841 bp) was cloned by RT–PCR based on EST sequences from soybean in the GenBank database as well as the Rubisco cDNA (full-length, small subunit). Signals on filters were quantified using a phosphor imager as well as quantification based on the quality of ribosomal RNA.

Lipid determination by NMR

After drying mature soybean seeds for 48 h at 80°C, the seed coat was taken off and the seed trimmed to 5 mm diameter. NMR spectroscopy and imaging were performed on a Bruker (Bruker Biospin, Rheinstetten, Germany) 750 MHz Advance system using a 5 mm birdcage resonator (Bruker). Global spectra were acquired with a spectral bandwidth of 21 kHz, a repetition time of 5 s and an averaging of 1. Spin–lattice (T1) and spin–spin (T2) relaxation times were measured with an inversion recovery sequence in the case of T1 and a Carr–Purcell–Meiboom–Gill sequence in the case of T2 to determine the relaxation times for the different lipid signals. After the spectroscopic experiments, 3D magnetic resonance imaging was conducted on the same seed. High-resolution proton distribution images were obtained using a standard 3D spin echo-pulse sequence, with an echo time of 3.7 ms, a repetition time of 0.5 s and four averages. The field of view was 5 × 6 × 6.5 mm3 and the matrix size was 128 × 154 × 168, resulting in an isotropic voxel resolution of (39 µm)3. The total experiment time was ≈14.5 h.

Metabolite analysis

Frozen plant material was rapidly weighed and immediately homogenized with an ice-cold mortar and pestle in liquid N2. Starch was determined after ethanolic extraction in the remaining pellet (Borisjuk et al., 2002). Total protein was determined from total nitrogen measurements (Rolletschek et al., 2002b). To analyse metabolites in abaxial and adaxial regions of cotyledons, embryos were freeze-dried and subsequently cut using a razor blade under a stereomicroscope (Stemi SV11, Zeiss, Germany). The two embryo fractions were weighed and extracted with trichloroacetic acid as described by Rolletschek et al. (2005). ADP-glucose was measured by liquid chromatography coupled to mass spectrometry (Rolletschek et al., 2005). Starch was analysed in the pellet remaining after extraction (Borisjuk et al., 2002).

Determination of fatty acid methyl esters

Seed material (2 mg) was added to 1 ml extraction medium [hexane : 2-propanol, 3 : 2 (v/v) with 0.0025% (w/v) butylated hydroxytoluene] containing triheptadecanoate as internal standard for the quantification of esterified fatty acids. Then seeds were immediately homogenized with a glass rod (potter) for 20 s on ice. The extract was shaken for 10 min and centrifuged at 13 000g at 4°C for 3 min. The clear upper phase was collected, and 600 µl of an aqueous solution containing 6.7% (w/v) potassium sulphate was added. After vigorous shaking and centrifugation at 13 000g at 4°C for 3 min, the upper hexane-rich layer was subsequently dried under a nitrogen stream. The remaining lipids were redissolved in 200 µl methanol. Then 20 µl of this solution was dried under a nitrogen stream, and 333 µl of a mixture of toluene and methanol (1 : 1, v/v) and 167 µl 0.5 mm sodium methoxide were added. After incubation of the samples for 20 min, 500 µl saturated NaCl and 50 µl HCl (32%, v/v) were added, and the fatty acid methyl esters (FAMEs) were extracted twice each with 0.75 ml hexane. The combined organic phases were evaporated to dryness under nitrogen, and the FAMEs were dissolved in 10 µl acetonitrile. Analysis of the corresponding FAMEs was performed with an Agilent (Waldbronn, Germany) 6890 gas chromatograph fitted with a capillary DB-23 column (30 m × 0.25 mm, 0.25 µm coating thickness, J&W Scientific; Agilent). Helium was used as carrier gas (1 ml min−1). The temperature gradient was 150°C for 1 min, 150–200°C at 8 K min−1, 200–250°C at 25 K min−1 and 250°C for 6 min. The amount of esterified fatty acids was calculated as the sum of all detected FAMEs.

Staining procedures

Lipid staining was conducted using fresh, freehand sections with Sudan B as described by (Brundrett et al., 1991). Briefly, Sudan red B (Sigma-Aldrich, St Louis, MO, USA) was dissolved in polyethylene glycol (400D, Sigma) by heating at 90°C for 1 h, and an equal volume of 90% (v/v) glycerol was added. Sections were stained for 24 h at room temperature, then rinsed with deionized water, fixed in 70% ethanol and photographed. Iodine staining was used to visualize starch granules (Borisjuk et al., 2002).

Confocal laser scanning microscopy

Confocal laser scanning microscopy (CLSM) was used to analyse chlorophyll distribution in developing seeds. For this purpose, 50 µm thick sections from early- to late storage embryos were cut on a vibratome. Illuminated with 488 nm laser light, chlorophyll autofluorescence was measured with a band-pass of 600–650 nm.

Transmission electron microscopy

Embryos of different developmental stages were chemically fixed for transmission electron microscopy (TEM) 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 postfixed with 1% OsO4 for 2 h. At the end of this procedure 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, samples were stained with 4% uranyl acetate and lead citrate. Digital recordings were made on a Zeiss 902 electron microscope (EM) at 80 kV.


Experiments were performed during the main storage phase in developing soybean seeds: early storage (50–120 mg f. wt); mid-storage (120–300 mg f. wt); and late storage (300–450 mg f. wt) stages (Fig. 1a). The storage phase was characterized by the near-linear accumulation of starch, lipid and protein. The net flux towards starch levelled off at around 300 mg seed f. wt (starch transition), after which the starch content declined. The onset of the storage phase was marked by strongly decreasing levels of hexoses and free amino acids (data not shown). Sucrose remained fairly constant at a level of ≈40 µmol g−1 f. wt, but increased at the time of starch transition to concentrations over 80 µmol g−1 f. wt. The ratio of whole-seed adenine nucleotides expressed as energy charge [AEC = (ATP + 0.5 × ADP)/(ATP + ADP + AMP)] was fairly high, but declined significantly before the starch transition occurred (Fig. 1b).

Figure 1.

Biochemical parameters of developing soybean seeds harvested at early, mid- and late storage stages. Dynamics of (a) starch (black circles), protein (grey circles) and lipid (white circles) accumulation; (b) adenylate energy charge. Error bars are means ± SD.

Photosynthetic abilities of the embryo form a gradient within cotyledons

We performed comparative investigations on photosynthetic parameters in abaxial and adaxial regions of embryos using PAM fluorescence. The maximal ETR of photosystem II reached the highest levels of 10.2 ± 0.8 µmol electrons m−2 s−1 in the young cotyledons, decreasing towards the later stages (Fig. 2, upper panels). This decrease was strongest in the adaxial regions where levels fell to zero in the late storage stage. In the early storage stage, both abaxial and adaxial regions were saturated at a photon flux density of ≈100 µmol quanta m−2 s−1. During the mid-storage stage, light saturation of the ETR in the outer regions of the embryo occurred at a twofold lower light intensity compared with the inner region. The slopes characterizing the light-saturation curve in the abaxial and adaxial regions were 0.130 ± 0.006 and 0.054 ± 0.004, respectively. During the late storage stage, the ETR of photosystem II decreased to undetectable levels within the adaxial region, but was still measurable within the surface area of the embryo (abaxial). This emphasizes the increasing difference between the adaxial and abaxial regions of cotyledons during development. Images of the effective quantum yield of photosystem II (measured at 63 µmol quanta m−2 s−1) showed a homogeneous pattern for small embryos of the early storage phase, but clear differences for the mid-/late storage stage with steep gradients declining towards the interior of the embryo (Fig. 2, lower panels). In summary, there is a gradual loss of photosynthetic ability starting from the embryo interior.

Figure 2.

Photosynthetic parameters of developing soybean embryo. Upper panels: electron transport rate (ETR, in µmol electrons m−2 s−1) measured in abaxial (a) and adaxial (b) regions in response to light intensity. Lower panels: images of the effective quantum yield of photosystem II measured at 63 µmol quanta m−2 s−1. Bar, 3 mm.

Photosynthetic oxygen release decreases towards the embryo interior

Due to restricted light supply towards the embryo interior and a gradual loss of photosynthetic ability, local differences in photosynthetic oxygen release might be expected. To test this assumption, we measured light-dependent oxygen production at 80 µmol photons m−2 s−1 within abaxial and adaxial regions using the microsensor technique (for methodological details see Rolletschek et al., 2002a). There were no differences for the early stage embryos. In contrast, during the mid-storage phase and later, the oxygen release rate was about fourfold higher in abaxial vs adaxial regions of the embryo (41.4 ± 2.3 vs 9.0 ± 0.1 µmol O2 g−1 f. wt and min). This indicates that embryo photosynthesis becomes distributed along a gradient in vivo.

Degreening starts from inner regions of the embryo and is coupled with structural changes in plastids

Photosynthesis in the embryo is associated with functional chloroplasts. During seed ripening, a colour transition occurs in the embryo from deep green to pale yellow. This degreening is indicative of a gradual loss of chlorophyll and disintegration of PSII (Saito et al., 1989). We have investigated spatial patterns of these processes within the growing embryo. To this end, comparative investigations of chloroplasts within adaxial and abaxial regions of the embryo were performed during early, mid- and late storage stages. The combined results of fluorescence analysis (CLSM) and ultrastructural investigations (TEM) are represented in Figs 3–6.

Figure 3.

Ultrastructural characteristics of embryo tissue at early storage stage. Overview (a) of specific chlorophyll fluorescence in cotyledons analysed by confocal laser scanning microscopy, and within abaxial (b) and adaxial (c) regions. Transmission electron microscopy of plastids in the abaxial region (d) shows three young chloroplasts, two containing small starch grains (white), surrounded by mitochondria, endoplasmic reticulum and storage deposits (black). (e) Representative picture of lamellae and grana in abaxial region. Structure of chloroplast in adaxial region (f); and fragment (g) showing typical grana. ab, abaxial; ad, adaxial; cw, cell wall; er, endoplasmic reticulum; g, grana; l, lamella; m, mitochondria; p, plastid; st, starch; vs, vascular bundle. Bars, (a) 200; (b,c) 20; (d,f) 5; (e,g) 1 µm.

Figure 4.

Ultrastructural characteristics of embryo tissue at mid-storage stage. Overview (a) of specific chlorophyll fluorescence in cotyledons analysed by confocal laser scanning microscopy, and within abaxial (b) and adaxial (c) regions. To stress differences in fluorescence, the contours of plastids that are approximately the same size are shown as dotted lines. Transmission electron microscopy of representative plastid structure in abaxial region (d): many differently sized starch grains are seen inside the plastid, as are well developed inner membrane structures of lamellae and grana. The close-up (e) demonstrates grana surrounding a starch grain. Plastid in adaxial region (f) is densely packed with large and small starch grains, but retains some grana. Fragment (g) shows grana and phytoferritin deposits. ab, abaxial; ad, adaxial; g, grana; li, lipid deposit; m, mitochondria; pf, phytoferritin; pg, plastoglobuli; pv, protein storing vacuole; st, starch; vs, vascular bundle. Bars, (a) 200; (b,c) 20; (d,f) 5; (e,g) 1 µm.

Figure 5.

Ultrastructure of embryonic plastids at late storage stage. Transmission electron micrograph of plastid in abaxial region (a) and fragment showing grana located between starch grains and accumulating lipid droplets or plastoglobuli (b). (c) Plastid in adaxial region possessing features of amyloplasts: fully packed with large starch grains and (d) partially distracted granal stack (arrow), large number of plastoglobuli, loss of membrane envelope (e), and massive deposits of phytoferritin (f). ab, abaxial; ad, adaxial; g, grana; li, lipid deposit; m, mitochondria; pf, phytoferritin; pg, plastoglobuli; st, starch; ?, absence of plastid envelope. Bars, (a,c) 5; (b,d,e,f) 1 µm.

Figure 6.

Storage product accumulation as revealed by electron microscopy and histological staining. (a–e) Early storage stage: representative transmission electron micrograph of embryo tissue showing deposit of starch, protein and lipid (a); starch distribution visualized by iodine staining (b); lipid distribution by Sudan III staining (c); smooth (d) and rough (e) endoplasmic reticulum within biosynthetically active regions. (f–i) Mid-storage stage: fragment of cell showing lipid, starch and protein deposition within abaxial (f) and adaxial region (g), extensive development of endoplasmic reticulum around plastid (h) and protein vacuoles (i). (j–l) Late storage stage: gradient in lipid distribution within cotyledons visualized by Sudan III staining (j); storage product deposition within adaxial (k) and abaxial (l) region: lipid bodies within the cytoplasm are larger in (k) than in (l). ab, abaxial; ad, adaxial; cw, cell wall; er, endoplasmic reticulum; li, lipid deposit; pf, phytoferritin; pg, plastoglobuli; pr, electron-dense protein deposit; st, starch; sv, storage vacuole containing protein; vs, vascular bundle. Bars, (a,f,g,i,k,l) 5 µm; (b,c) 300 µm; (d,e,h) 1 µm; (j) 1 mm.

According to CLSM, plastids of young embryos displayed a strong chlorophyll autofluorescence when illuminated with 488 nm laser light (Fig. 3). Chlorophyll fluorescence was nearly evenly distributed throughout the tissues during the early stage (Fig. 3a–c). Ultrastructural observations showed that most chloroplasts of young embryos (≈70 mg f. wt) contained grana with well differentiated thylakoids, as well as starch grains and plastoglobuli. Although chloroplast morphology did not differ significantly in the different parts of the embryo (Fig. 3d,e), chloroplasts from the adaxial region were, on average, slightly larger than those from the abaxial region. This may be due to the starch granules, which were more numerous and larger in adaxial compared with abaxial chloroplasts (Fig. 3f,g).

During the mid-stage there was a pronounced chlorophyll gradient with a maximum in the abaxial region (Fig. 4). Chlorophyll fluorescence in individual plastids was stronger in the abaxial compared with the adaxial region (Fig. 4b,c). Adaxial and abaxial plastid populations could easily be distinguished by ultrastructural investigations. Chloroplasts within abaxial regions still contained well developed grana with stacks of four to 10 thylakoid membranes, which were loosely distributed between starch grains of moderate size (Fig. 4d,e). Chloroplasts from adaxial regions, however, were partially distorted by extensive accumulation of starch (Fig. 4f,g). The grana of these chloroplasts not only had lost their structural regularity, they also contained significantly fewer thylakoid membranes (Fig. 4g). Occasionally the accumulation of phytoferritin and plastoglobuli was observed.

The late storage stage was characterized by a near absence of chlorophyll autofluorescence, with the exception of the most peripheral tissue layers (data not shown). The integrity of plastids was often distorted by large, complex starch grains, especially in adaxial regions (compare Fig. 5a,b and c–f). Clusters of plastoglobuli (Fig. 5d) and accumulation of phytoferritin (Fig. 5f) were commonly observed in the plastid stroma. Grana were found only in abaxial plastids, but were significantly smaller and had fewer thylakoid membranes compared with those from previous stages (Fig. 5d).

In summary, our data show that degreening is accompanied by characteristic changes in the ultrastructure of plastids, and occurs along a gradient.

Differentiation of storage tissue occurs in coordination with plastid development

Plastid development is an integral part of tissue differentiation. We combined EM, histological staining procedures and conventional biochemical analyses to determine how plastid differentiation is reflected in the deposition pattern of storage products within the embryo.

During the early storage stage, clear EM evidence was found for starch, lipid and protein accumulation all over the embryo parenchyma simultaneously (Fig. 6a). For starch, the distinction between adaxial and abaxial regions was already notable by investigation of the chloroplast population (see above). Starch deposition was maximal in the adaxial region, as also evident by the positive iodine–potassium iodine reaction (Fig. 6b). At the same time, lipid bodies appeared more numerous in adaxial regions. This was confirmed by the intense Sudan III reaction of fresh material (Fig. 6c). Topographically, lipid accumulation occurred within tissues with high starch density. Protein deposition occurred in the opposite direction, and was more evident within cells of the abaxial region (confirming data of Perez-Grau & Goldberg, 1989; Saito et al., 1989). Extensive development of endoplasmic reticulum (ER) was observed: smooth ER (Fig. 6d) consisting of smooth membranes enclosing electron-translucent spaces and forming a network through the cytoplasm; and rough ER (Fig. 6e) associated with ribosomes and the Golgi system.

During the next (mid-storage) stage (Fig. 6f–i), differences in plastid structure became obvious. Biochemical analysis revealed about twofold lower levels of starch in abaxial vs adaxial regions (30.8 ± 5.3 and 57.3 ± 7.7 mg g−1 f. wt, respectively). In addition, the level of ADP-glucose (the direct precursor of starch biosynthesis) was likewise twofold lower (data not shown). Storage protein deposits in the abaxial tissues were larger than in the adaxial region. The protein storage vacuoles within the abaxial region were different in shape and richly filled with material which was dispersed, and/or built aggregates with different electron density (Fig. 6f). In adaxial regions, storage vacuoles were smaller and filled with more homogeneous material (Fig. 6g). Cytoplasm was changed via significant proliferation of rough ER, surrounding protein vacuoles and lining plastids in multiple layers (Fig. 6h,i). Deposition of lipids occurred predominantly in adaxial regions, as revealed by histostaining (Fig. 6j). According to EM, lipid bodies of different sizes were prevalent in cytoplasm of cells within the adaxial region. Many were attached to the inner side of the cell membrane underlying the cell wall or surrounding protein vacuoles. Lipids were more freely distributed within the cytoplasm of abaxial cells, tending towards the outer cell membrane, and were smaller (Fig. 6k,l).

During the late storage stage, cells within abaxial regions were mainly made up of densely packed protein storage vacuoles, in contrast to cells in the adaxial part which contained significantly fewer proteinaceous structures (not shown).

Taken together, histological and EM studies demonstrated the heterogeneity of storage tissue: both starch and lipid started to accumulate from adaxial regions towards the periphery of the embryo. This spatial pattern follows the gradual loss of photosynthetic ability and differentiation of plastids, as described above.

Lipid deposition within mature seeds as revealed by noninvasive NMR

To obtain an overview of lipid distribution within the whole mature seed we used the noninvasive NMR technique (Köckenberger et al., 2004). MR micro-imaging as well as NMR spectroscopy (Schneider et al., 2003) was adapted for measurements on seeds. An example of the NMR spectra from mature seeds is shown in Fig. 7a. There are three main peaks over a range of 7 ppm in the proton spectrum, tentatively assigned to triacylglycerols, glycolipids and phospholipids. The acquired 3D spin–echo experiment finally allowed visualization of lipid gradients in the trimmed seed in a noninvasive manner. Two tangential ‘sections’ of the seed (through the mid-part and its basal part with axis) are shown in Fig. 7b. The colour scale represents signal intensity in relative units and corresponds to lipid content in appropriate tissue areas. The highest signal intensity was found in the inner part of cotyledons, and lowest levels in the embryo axis (signals in red and blue, respectively). Remarkably, there was a clearly distinguishable, steep gradient in lipid deposition. The inner (adaxial) region of cotyledons showed about fivefold higher lipid contents compared with peripheral regions. This pattern in mature seeds clearly confirmed the lipid gradient already formed during embryo development.

Figure 7.

Lipid deposition within mature seeds as revealed by noninvasive nuclear magnetic resonance. (a) Proton spectrum showing three main peaks tentatively assigned to glycolipids, phospholipids and triacylglycerols. (b) Image of signal intensity (given in colour scale) within two tangential sections through mid (α) and basal (β) part of embryo, as shown in (c). ab, abaxial; ad, adaxial; ra, radicle. Bar, 3 mm.

Expression of plastid-related genes is temporally and spatially regulated in the developing embryo

The shift from photosynthetic to storage activity within developing embryos goes along with alterations in carbohydrate exchange between plastids and cytoplasm. Hence genes encoding key players in this process can be used as molecular markers for plastid differentiation (Kubis et al., 2004; Weber, 2004). We isolated putative plastid metabolite transporters of soybean by RT–PCR (see Materials and Methods) and, together with ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco, small subunit), analysed their expression in vegetative tissues as well as developing embryos.

Rubisco showed highest expression during the early storage stage, and decreased later to undetectable levels (late storage stage) (Fig. 8a,c). Tissue separation followed by Northern analysis further revealed a shift in expression from adaxial to abaxial regions of cotyledons (mid-storage stage, data not shown). The triose phosphate–phosphate translocator (TPT) is characteristic for photosynthetic tissues (Weber, 2004). Expression of TPT_RT was found in all green vegetative tissues (flower, pod, leaf, stem) but not in roots (Fig. 8b). In embryos, TPT expression was upregulated in embryo from ≈50–100 mg seed f. wt (Fig. 8a,c), and thereby restricted to early developmental stages as shown for Rubisco.

Figure 8.

Expression of plastid-related genes in developing embryo and vegetative tissue of soybean. (a–c) Relative transcript level and expression of Rubisco (RBC) and putative plastidic metabolite transporters TPT, PPT and GPT by Northern hybridization in different tissues and stages of embryo development. Cross-section of early stage cotyledon (d, toluidine blue staining) and spatial expression pattern of GPT by in situ hybridization [signals seen as white grains in dark field (e) and enlargement (f)]. Localization of starch by iodine staining in parallel sections [abaxial (g) and adaxial (h)]. ab, abaxial; ad, adaxial; vr, vascular region. Bars, (d,e) 1; (f–h) 0.2 mm.

The glucose-6-phosphate–phosphate translocator (GPT) imports glucose-6-phosphate into the plastid (Weber, 2004). GPT_RT was not expressed in leaves but in roots and other sink organs (Fig. 8b). In embryo, the transporter was upregulated during the early storage stage (100–150 mg seed f. wt), decreased during further development, but showed a second maximum at the late storage stage (350–400 mg f. wt) (Fig. 8a,c). In situ localization of GPT_RT in tissue sections of embryo (early stage) revealed strong specific labelling of storage parenchyma within adaxial regions, gradually decreasing towards the periphery (Fig. 8d–f). Iodine staining in parallel sections points to the colocalization of elevated GPT_RT expression and starch accumulation (Fig. 8g,h). During mid- and late storage stages, GPT_RT labelling was more or less equal in cotyledon tissues (data not shown).

The phosphoenolpyruvate/phosphate translocator (PPT) is required for the importing of phosphoenolpyruvate into plastids (Weber, 2004). Its expression was found in both green and nongreen plant tissues, with highest levels in sink organs (flowers, pod, sink leaf; Fig. 8b). In embryo, PPT_RT showed the highest expression between 100 and 150 mg seed f. wt (Fig. 8a,c). Moderately high expression was also observed during the late storage stage. Spatial expression patterns with a maximum within the adaxial region of the embryo were similar to those seen for the GPT_RT (data not shown).

In summary, expression of Rubisco and TPT_RT associated with photosynthetically active plastids was restricted to the early storage stages where photosynthetic activity of embryos was highest. In contrast, expression of both GPT_RT and PPT_RT was upregulated at the onset of starch and lipid accumulation, and coincided spatially with deposition of these storage products.


We present here for the first time gradients of lipid deposition in oilseeds, and describe how these gradients are related to structural and functional changes in plastids and tissue differentiation. Within the interior of embryos, photosynthetically active chloroplasts differentiate into storage organelles, expressing metabolite transporters characteristic for heterotrophic metabolism. This is followed by local accumulation of both starch and lipid. The gain of storage functions is accompanied by the apparent loss of photosynthetic capacity. Hence photosynthetic activity and storage overlap during early seed growth, but become separated spatially during the main storage phase. This indicates a functional shift in the interrelationship of photosynthesis and storage during soybean seed development.

A comparative view of storage product deposition in soybean seeds

Accumulation of storage products occurs in a well known temporal manner in oil-storing seeds, including soybean (Yazdi-Samadi et al., 1976). However, their spatial deposition patterns have only been partially investigated (Meinke et al., 1981; Perez-Grau & Goldberg, 1989), and lipid accumulation patterns are unknown. Recent investigations emphasize the complexity of the interrelation between starch and lipid storage (Da Silva et al., 1997; White et al., 2000; Ruuska et al., 2002; Hills, 2004; Vigeolas et al., 2004). In this context we analysed the spatial pattern of lipid biosynthesis/deposition.

Starch and lipid accumulate simultaneously (Fig. 1). Starch deposition shows a clear spatial gradient with a maximum in the adaxial region (mid-storage stage). Further on, starch is utilized while lipid accumulation continues up to maturity. Visualization by NMR of lipid distribution in mature soybean embryo as a 3D model revealed gradients similar to that of starch (Fig. 7). Replacement of starch by lipid provokes speculation that the final gradient of lipid deposition is simply a product of starch-to-lipid conversions during development. Consideration of starch as a possible carbon provider for oil synthesis, and the negative correlation between starch and lipid accumulation in oilseed plants (Da Silva et al., 1997; Periappuram et al., 2000), might support this assumption. However, our ultrastructural and histochemical data indicate clearly that the gradients in both lipid and starch accumulation already occur simultaneously during the early storage stage. Thus the lipid distribution pattern observed in mature seeds does not simply reflect starch-to-lipid conversion during late storage stages, but results likewise from lipid biosynthesis distributed along a gradient. It was suggested recently that starch biosynthesis is involved in establishing the (local) sink activity of seeds required for the onset of oil accumulation (Da Silva et al., 1997). Decreasing starch biosynthesis via transgenic downregulation of AGPase causes a developmental delay in lipid accumulation in canola, but does not affect the final oil accumulation (Vigeolas et al., 2004). In this context, the colocalization of starch and lipid deposition observed here appears to be quite reasonable.

The coordination between oil and protein synthesis varies between oil-accumulating species as well as between different seed types (rapeseed, Norton & Harris, 1975; castor bean, Greenwood et al., 1984; Arabidopsis, Baud et al., 2002; tobacco, Tomlinson et al., 2004). In soybean embryo, oil accumulation coincides temporally with the accumulation of protein (Fig. 1). However, protein deposition starts in the abaxial region of seeds (as evidenced by mRNA pattern; Perez-Grau & Goldberg, 1989). Proteins and lipids therefore show an inverse deposition pattern. This unexpected finding may, in part, explain the independence of protein and oil biosynthesis (Schwender & Ohlrogge, 2002 and citations therein). Notably, it points to the importance of spatial patterns of temporally coinciding storage processes, and raises the question as to which factors determine these gradients and confer an evolutionary advantage.

Differentiation of plastid and storage tissues occurs along a gradient within developing embryos

Previous work on legume seeds (soybean, Saito et al., 1989; pea, Smith et al., 1990) suggested that storage plastids are derived directly from chloroplasts via plastid differentiation. Our study demonstrates the spatial pattern of plastid differentiation in the soybean embryo. We observed processing of the membrane system related to photosynthetic functions (grana stacks, where PSII is assembled, Hermann et al., 1985; nonapressed membranes, associated with PSI, Staehelin & Arntzen, 1983), and demonstrated their degradation (reducing of grana, accumulation of ferritin and plastoglobuli) as well as starch deposition in plastids (Figs 4–6). These processes were coupled with the gradual loss of specific chlorophyll fluorescence (associated with granal stacks, Apuya et al., 2001), and emphasized gradients from adaxial to abaxial parts of the embryo. Such structural changes in chloroplasts are known to be coupled with their photosynthetic abilities (Rolletschek & Borisjuk, 2005), and are evident by spatially resolving measurements of the effective quantum yield of PSII, photosynthetic electron transport rate and photosynthetic O2 production.

Saito et al. (1989) suggested that stages in plastid differentiation may be well coordinated with cell functions. This was later confirmed by genetic manipulation of biosynthetic functions of plastids (schlepperless, Apuya et al., 2001; EDD1, Uwer et al., 1998; GPT in Vicia, Rolletschek and coworkers, unpublished data). We conclude that the functional shift from photosynthesis to storage is coupled with changing histological/biochemical parameters of the embryo tissues, revealing the developmental gradient within the embryo.

The shift from photosynthesis to storage is coupled with alterations in plastid-related gene expression

The acquisition of photosynthetic competence requires the activation of both nuclear and chloroplast genes (Mullet, 1988). We demonstrate here that expression of Rubisco (small subunit) temporally coincides with the highest photosynthetic activity of the embryo (early to mid-storage stage, Figs 2, 8). A similar expression pattern was described for chlorophyll a/b binding protein (Saito et al., 1989). While it is not associated with the photosynthetic apparatus, TPT is highly abundant in photosynthetic tissues. This transporter delivers photosynthetically derived triose phosphates from chloroplasts to the cytoplasm in source tissues (leaves). The TPT_RT transcript described here is restricted to green tissues, and can be attributed to photosynthetically active plastids. In embryo, its expression is much lower than in leaves, corresponding to the presumably low input of embryonic photosynthesis to the overall carbon supply (Sriram et al., 2004).

Storage activity requires high import capacities via metabolite translocators. The GPT catalyses the plastidic glucose-6-phosphate import, feeding mainly starch biosynthesis and the oxidative pentose phosphate pathway (Rawsthorne, 2002; Weber, 2004). The PPT imports phosphoenolpyruvate and can contribute substantially to fatty acid biosynthesis (Kubis et al., 2004; Weber, 2004). Although transcript abundance alone cannot be used to infer the in vivo activity of translocators (Kubis et al., 2004), they are considered as molecular markers for plastids involved in lipid and starch synthesis (Arabidopsis: Ruuska et al., 2002; White et al., 2000; Brassica: Kubis et al., 2004). GPT_RT is highly abundant in storage tissues. Its expression coincides with the onset of starch/lipid accumulation in the young embryo, confirming findings for other lipid-storing seeds (Arabidopsis: White et al., 2000; Brassica: Kubis et al., 2004). GPT_RT remained highly expressed when starch degradation occurred. This points to a direct role for GPT in lipid synthesis.

Because GPT is upregulated by sugars (Quick et al., 1995), its expression may be included in the reprogramming of chloroplasts into starch- and fatty acid-synthesizing plastids under elevated carbohydrate supply. By transition of autotrophic leaves to heterotrophic growth conditions, their plastids show elevated GPT activity and store starch using glucose-6-phosphate as a substrate (Quick et al., 1995), like heterotrophic plastids from seeds. This links the sugar status and its proposed role as a metabolic signal (Weber et al., 2005) with plastid differentiation.

The expression of GPT_RT and PPT_RT occurs concomitantly in developing soybean seeds. Because biosynthesis of starch and fatty acids shares the same compartment and utilizes the same carbon and energy sources (Möhlmann et al., 1994; Periappuram et al., 2000), such an observation seems quite reasonable. Moreover, the high ATP level required for maintenance of high biosynthetic activity (Neuhaus & Emes, 2000) is locally elevated at sites of synthesis (Rolletschek et al., 2005). Supposedly, energy supply is a decisive factor for channelling carbon towards lipid.

Taken together, the upregulation of putative carbohydrate-importing translocators precedes morphological and functional changes of plastids and biochemical alteration of tissues.

Spatial regulation of photosynthesis and lipid biosynthesis affects their cross-talk in developing embryos

Lipid biosynthesis is known to be strongly energy-dependent, and is stimulated by light (Neuhaus & Emes, 2000; Rawsthorne, 2002). It has been suggested that seed photosynthesis provides significant amounts of energy in the form of ATP and redox equivalents, needed for fatty acid biosynthesis (Asokanthan et al., 1997; Willms et al., 1999; Ruuska et al., 2004). The ‘photosynthetic energy supply’ hypothesis relies essentially on the colocalization of photosynthesis and lipid biosynthesis. This is because long-distance transport of ATP/NADPH from sites of synthesis to those of utilization is unlikely. During the early storage stage of soybean embryos, there is a spatial codistribution of photosynthesis and lipid biosynthesis. However, starting from the mid-storage stage onwards, photosynthesis becomes restricted to peripheral (abaxial) regions. This results in a marked separation of local photosynthetic and lipid biosynthetic activities within the soybean embryo tissue, and makes a key role of photosynthetic energy supply for lipid synthesis unlikely. Rubisco may act without the Calvin cycle, significantly increasing the carbon efficiency in oilseed rape embryos, as demonstrated recently by Schwender et al. (2004). This mechanism might also be important for soybean embryos, and is under investigation (J. Ohlrogge, personal communication). However, gene expression of Rubisco decreases dramatically from mid-storage stage onwards (start of degreening), and it is not known if high Rubisco activity is maintained in degreening tissues. Even if this is the case, lipid biosynthesis requires a high local energy supply. This raises the question: how can photosynthesis, being spatially separated from lipid biosynthetic regions, mediate energy supply? We propose that oxygen is photosynthetically produced at high rates at the embryo periphery throughout development. It can easily diffuse towards the embryo interior where it promotes respiration, followed by local increases in the energy state and enhanced biosynthetic fluxes. The physiological significance of photosynthetic O2 supply is based on internal O2 limitation of seed metabolism in vivo. Our ‘photosynthetic oxygen supply hypothesis’ has been verified using microsensors, ATP bioluminescence imaging, metabolite profiling and isotope flux analysis in the second part of our study (Rolletschek et al., 2005).


We are grateful to U. Tiemann and K. Lipfert for artwork, and K. Blaschek, A. Schwarz, A. Stegmann, U. Siebert and P. Meier for excellent technical assistance. We thank M. Melzer (IPK Gatersleben) for providing technical equipment. We acknowledge funding by the Deutsche Forschungsgemeinschaft (Project numbers RO 2411/2-1/2-2, WE 1641/6-1/6-2). We would like to thank Sebastian Aussenhofer for technical support and work with AMIRA. Thomas Neuberger and Andrew G. Webb acknowledge financial support from the Alexander von Humboldt Foundation, Wolfgang Paul Programme.