Starch turnover in developing oilseed embryos


Author for correspondence:
Alison M. Smith
Tel: +44 1603 450622


  • Starch accumulates early during embryo development in Arabidopsis and oilseed rape, then disappears during oil accumulation. Little is known about the nature and importance of starch metabolism in oilseed embryos.
  • Histochemical and quantitative measures of starch location and content were made on developing seeds and embryos from wild-type Arabidopsis plants, and from mutants lacking enzymes of starch synthesis and degradation with established roles in leaf starch turnover. Feeding experiments with [14C]sucrose were used to measure the rate of starch synthesis in oilseed rape embryos within intact siliques.
  • The patterns of starch turnover in the developing embryo are spatially and temporally complex. Accumulation is associated with zones of cell division. Study of mutant plants reveals a major role in starch turnover for glucan, water dikinase (absent from the sex1 mutant) and isoforms of beta-amylase (absent from various bam mutants). Starch is synthesized throughout the period of its accumulation and loss in embryos within intact siliques of oilseed rape.
  • We suggest that starch turnover is functionally linked to cell division and differentiation rather than to developmental or storage functions specific to embryos. The pathways of embryo starch metabolism are similar in several respects to those in Arabidopsis leaves.


The maturation phase of developing embryos of the model plant Arabidopsis thaliana and its close relative Brassica napus is characterized by rapid expansion growth and the synthesis and accumulation of storage reserves. Embryos of both species have an initial phase of starch accumulation, before the accumulation of storage oil (Eastmond & Rawsthorne, 2000; Baud et al., 2002; Hills, 2004). Starch levels decline later during development as rates of accumulation of storage oil and protein increase. Starch is almost undetectable at maturity, whereas storage oil and protein may each account for up to 40% of the dry weight (Baud et al., 2002; O’Neill et al., 2003).

Although starch is of no quantitative importance in the mature seed, at particular developmental stages it accounts for a significant fraction of embryo weight and is a major fate of carbon entering the embryo. In oilseed rape embryos the maximum starch content represents 8–10% of the dry weight (da Silva et al., 1997; Eastmond & Rawsthorne, 2000). The rate of accumulation in the 6 d before this point is c. 6 mg g−1 FW d−1 (estimated from da Silva et al., 1997). This rate is similar to that in embryos of plants in which the major seed storage product is starch. For example, starch accumulates at 7–10 mg g−1 FW d−1 through much of the development of pea embryos (estimated from Denyer et al., 1995). The absolute rate of starch synthesis in oilseed embryos may be higher than the rate of accumulation, because starch may be simultaneously synthesized and degraded. Starch turnover has been shown to occur in isolated developing embryos of oilseed rape, and in intact plastids prepared from developing embryos (da Silva et al., 1997; Eastmond & Rawsthorne, 2000). However, the extent to which these experimental systems reflect the metabolism of embryos in intact plants is not known.

Starch synthesis in plastids of oilseed embryos proceeds from glucose 6-phosphate via phosphoglucomutase, ADPglucose pyrophosphorylase (AGPase), starch synthases and starch branching enzymes (da Silva et al., 1997; Rawsthorne, 2002; Vigeolas et al., 2004). The pathway of starch degradation in embryos is not known. Recent research on starch degradation in Arabidopsis leaves at night has revealed a complex pathway (Smith et al., 2005; Niittyläet al., 2006; Fulton et al., 2008; Kötting et al., 2009). Starch polymers are initially phosphorylated by glucan, water dikinase (GWD1) and phosphoglucan, water dikinase (PWD) (Yu et al., 2001; Ritte et al., 2002; Baunsgaard et al., 2005; Kötting et al., 2005) rendering the granule surface more accessible to hydrolysis by β-amylases (BAM) and the debranching enzyme isoamylase 3 (ISA3) (Wattebled et al., 2005; Delatte et al., 2006; Fulton et al., 2008). Complete hydrolysis to maltose and glucose requires dephosphorylation via the glucan phosphate phosphatase SEX4 (Kötting et al., 2009) and malto-oligosaccharide disproportionation via DPE1 (disproportionating enzyme) (Critchley et al., 2001). Maltose is exported to the cytosol via the plastidial maltose transporter MEX1 (Niittyläet al., 2004), then further metabolized via the transglucosidase DPE2 (Chia et al., 2004). Although mutations affecting some of these enzymes are known to increase the starch content of the testa of the seed as well as the leaves (for example the sex1 mutation that affects GWD1; Caspar et al., 1991), the extent to which this pathway of starch degradation operates in the developing embryo is not known.

Little is known about the role of starch accumulation in developing oilseed embryos, but it has been the subject of considerable speculation. First, it was proposed that starch could serve as a carbon reserve for the synthesis of lipids and sugars later during development or to provide substrate for either the plastidial oxidative pentose phosphate pathway or the synthesis of sugars such as stachyose that accumulate during desiccation (Norton & Harris, 1975; Leprince et al., 1990). However, the maximum starch content during development is only a small fraction of the final oil content in these seeds, making it unlikely that starch supplies a significant proportion of the carbon and reductant needed for oil synthesis. Indeed, large reductions in starch content of oilseed rape embryos brought about by embryo-specific downregulation of AGPase had no significant impact on the final oil content of the seed (Vigeolas et al., 2004). Second, the early accumulation of starch may contribute to the establishment of the developing embryo as a sink organ. Conversion of imported sucrose to starch may serve to attract more carbon into the developing embryo and thus contribute to the sink strength of the seed before the major phase of oil synthesis (da Silva et al., 1997).

The aim of our work was to assess the location, nature and importance of starch synthesis and degradation during oilseed development. First, we examined the spatial and temporal pattern of starch accumulation in developing Arabidopsis seeds. Second, we investigated the pathway of starch degradation during Arabidopsis embryo development by examining the impact on starch turnover of mutations affecting enzymes with known roles in starch turnover in leaves. Third, we examined developmental and diurnal changes in the rate of starch synthesis in embryos of B. napus by feeding intact siliques with [14C]sucrose. Finally we examined the effect upon the final composition of Arabidopsis embryos of mutations that alter the rate of starch turnover during development.

Materials and Methods

Plant material and growth conditions

Unless otherwise specified, Arabidopsis plants were grown in a controlled environment room with 16 h light : 8 h dark, 250 μmol photosynthetically-active radiation m−2 s−1, at 22°C. For microscopy and measurements of starch in seeds, reproductive growth was restricted to the main stem and all secondary and side shoots were removed upon emergence. Flowers were tagged on the day of anthesis. For the determination of the lipid content of mature seeds, plants were grown as described by Andriotis et al. (2010) and Hobbs et al. (2004).

All of the mutant lines used in this study have been described previously: see references in the Results section. For all except sex1, the same mutant line was used throughout and is stated in the text where appropriate. For sex1, the use of either sex1-1 or sex1-3 is specified in the legends. phs1-1 is in the Ws background; all others are in the Col-0 background.

For 14C feeding experiments with detached siliques, Brassica napus cv Topas was grown in a glasshouse with supplementary lighting to give a 16-h day (day minimum 18°C, night minimum 15°C).

Microscopy and starch staining

For all experiments, developing siliques were rapidly harvested and opened under a dissecting microscope.

For determination of embryo developmental stages, seeds were cleared at room temperature on microscope slides using Hoyer’s solution (Liu & Meinke, 1998). Clearing was complete within 2 h for young seeds, or overnight for older material.

For starch staining, seeds were placed in 70% (v : v) ethanol and boiled for 5 min, then incubated in an iodine solution (50% (v : v) Lugol’s solution) for 20 min. Excess staining solution was removed by washing with distilled water.

Whole-mount preparations of all seed material were viewed under differential interference contrast (DIC) optics with a Nikon Eclipse 800 microscope.

For light microscopy of sections, seeds were fixed in 2.5% (w : v) glutaraldehyde in 50 mM sodium cacodylate (pH 7.2) at room temperature under vacuum for 1 h and then overnight in fresh fixative at room temperature. Fixed samples were treated with 1% (w : v) osmium tetroxide in 50 mM sodium cacodylate buffer (pH 7.2) for 1 h, then dehydrated in an ethanol series and gradually infiltrated with LR White resin. Following three changes of pure resin, samples were polymerized for 16 h at 60°C. Sections were cut on a Leica UC6 ultramicrotome (Leica Microsystems, Milton Keynes, UK) and stained with toluidine blue O, counterstained with iodine solution (50% (v : v) Lugol’s solution) and viewed with a Nikon Eclipse 800 microscope.

[14C]sucrose feeding experiments

Siliques were detached from B. napus plants and the cut pedicels incubated in Murashige and Skoog medium containing 20 mM [U-14C]sucrose at 30 MBq mol−1 (GE Healthcare, Little Chalfont, Bucks, UK.), as described by Morley-Smith et al. (2008). Incubations were for 12 h at 25°C, in either 300 μmol quanta photosynthetically active radiation m−2 s−1 or darkness. Incorporation of radiolabel into seed storage products was linear for the experimental period. At the end of the incubation seeds were removed from the siliques and embryos rapidly dissected from them and frozen at −80°C until further analysis.

Measurement of radioactivity in starch was as described by Runquist & Kruger (1999). Four embryos from each silique were extracted in hot ethanol and the insoluble residue subjected to autoclaving and incubation with α-amylase and α-amyloglucosidase. The radioactivity rendered soluble by the incubation was measured by liquid scintillation counting.

To permit calculation of the absolute rate of starch synthesis, the specific radioactivity of the glucose 6-phosphate (Glc 6-P) pool within the embryo was measured according to Morley-Smith et al. (2008; see the Supporting Information, Table S1). Briefly, Glc 6-P was assayed by an enzymatic cycling assay in neutralized trichloroacetic acid extracts of up to 10 pooled embryos from a single silique. The 14C content of the Glc 6-P was determined by treatment of extracts with Glc 6-P dehydrogenase to convert Glc 6-P to 6-phosphogluconate, passage through a column of Dowex 50 resin, then elution with 80 mM succinate from a column of Dowex 1 resin (Cl form) to resolve 6-phosphogluconate from other phosphoesters. Calculations of the rate of starch synthesis assumed that starch was derived from the Glc 6-P pool within the embryo. The 14C recovered in starch (dpm) was divided by the specific activity of Glc 6-P (dpm nmol−1) to give the amount of starch synthesized.

Analytical methods

For measurement of embryo starch content, embryos were rapidly dissected from seeds from freshly-harvested siliques. Samples of 30 embryos, obtained from at least five plants per genotype that were allowed to develop only the main shoot, were either frozen in liquid nitrogen and stored at −80°C or immediately homogenized. Samples were homogenized in ice-cold 0.7 M perchloric acid and centrifuged at 13 000 g, for 45 min at 4°C. Starch was assayed on the pellets according to Smith & Zeeman (2006). Briefly, the pellets were washed once in water and twice in 80% (v : v) ethanol, then resuspended in 200 mM sodium acetate (pH 4.8), boiled for 10 min and digested overnight with α-amyloglucosidase and α-amylase. The amount of glucose released in the incubation was determined enzymatically. Measurements of starch in whole seeds and testas were made in the same way.

Measurement of the lipid content and fatty acid content of seeds and embryos were made by NMR spectroscopy (Hobbs et al., 2004; Andriotis et al., 2010) and fatty acid methyl ester analysis (Chia et al., 2005) respectively.


Location of starch in developing Arabidopsis seeds

We used iodine staining to follow the pattern of starch accumulation during development in Arabidopsis seeds. In our growth conditions seeds contained globular-stage embryos at 4 d after flowering (DAF), and by 6 DAF most embryos had reached the heart stage and entered the maturation phase of development. They expanded to the walking stick – early cotyledon – stage by 12 DAF. Development was complete by 16 DAF (seeds contained fully-expanded embryos that had started to desiccate; Fig. 1).

Figure 1.

 Arabidopsis seed development and starch accumulation. (a) Developmental progression of wild-type (Col-0) embryos under the growth conditions used in our study. Whole-mount preparations of seeds from developing siliques were viewed under differential interference contrast (DIC) optics. Developmental stages shown are (left to right) globular embryo at 4 d after anthesis (DAF), heart stage embryo at 6 DAF, torpedo stage embryo at 8 DAF, walking stick stage embryo at 10 DAF, upturned U (expanded cotyledon) embryo at 12 DAF and embryo at 16 DAF. Bars, 50 μm for the first four stages; 100 μm for the last two stages. (b) Starch in the testa of developing Arabidopsis seeds. Seeds were dissected from siliques at the same DAF as in (a), and stained with iodine solution. (c–f) Embryos dissected from developing seeds, stained with iodine solution to visualize starch and observed under DIC optics. (c) Embryo at the torpedo stage, 8 DAF. Iodine staining shows starch granule accumulation (blue coloration) mainly in two distinct regions, one immediately above the tip of the radicle and a second at the hypocotyl region (marked with asterisks). (d) Embryo at the walking stick stage, 10 DAF. Starch is present in the radicle and hypocotyl. (e) Embryo at the early cotyledon stage, 12 DAF. Starch is present in cotyledons as well as the radicle and hypocotyl. (f) Embryo at the expanded cotyledon stage, 14 DAF. Starch has disappeared apart from a zone above the tip of the radicle (asterisk). (g–i) Thin sections of embryo stained with toluidine blue and counterstained with iodine solution. (g) Embryo at torpedo stage, as in (c). Iodine staining shows starch granule accumulation (brown dots) mainly immediately above the tip of the radicle and at the hypocotyl region (marked with asterisks). r, radicle; h, hypocotyl; c, cotyledon. (h) Section through the radicle of an early cotyledon-stage embryo, 12 DAF. The inset shows a magnification of a single cell from this section. Starch granules are found in all embryonic tissues apart from the vasculature (v). (i) Section though the root apical meristem of a cotyledon-stage embryo as in (f). Starch granules are present throughout the meristem. Bars, (c) 25 μm, (d) 20 μm, (e–g) 50 μm, (h) 20 μm, (i) 10 μm.

In testas, starch was present at 4 DAF. Staining intensity of intact testas increased up to cotyledon stage (Fig. 1). In thin sections, starch granules were visible in the outer integument layers of seeds containing heart-stage embryos. The number and the size of starch granules increased up to cotyledon stage (Fig. S1). After cotyledon stage, iodine staining in the testa decreased, and no starch was present in the testas of mature seeds (Fig. 1).

The distribution of starch in the embryo was examined by iodine staining, and by microscopy of thin sections. The two approaches gave identical results. There was very little starch in heart-stage embryos, although small starch granules were occasionally detected in the basal region of the embryo (Fig. S1). In torpedo-stage and walking-stick stage embryos (8–10 DAF) starch was present mainly in two distinct zones, above the tip of the radicle and at the hypocotyl region below the apical meristem (Fig. 1c,d,g). Staining intensity in both locations peaked at the early cotyledon stage (12 DAF; Fig. 1e). At this stage starch was also present in the cotyledons. In thin sections, starch granules were absent from the vasculature but visible in all other hypocotyl and radicle cells, including the radicle tip and meristem (Fig. 1h,i; Fig. S1). Beyond 12 DAF, the starch content of the embryo decreased rapidly. By 16 DAF starch was visible only above the radicle tip (Fig. 1g) and it had disappeared in mature, desiccated embryos (see Fig. 3).

Starch was also present in the endosperm of developing Arabidopsis seeds. After endosperm cellularization (c. 8 DAF), starch granules of various sizes were visible in thin sections stained with iodine. Their size and number increased by 12 DAF (Fig. S1). Endosperm starch content then declined and starch was absent from the endosperm of mature seeds (not shown).

Starch content of developing Arabidopsis embryos

Previously, starch content in Arabidopsis seeds has been assayed only on a whole-seed basis (Baud et al., 2002). We found that starch in the testa accounted for most of the starch content of the whole seed at 12 DAF (Fig. 2a). To determine the pattern of starch accumulation specifically in the embryo, measurements were made on embryos dissected free of testa and surrounding endosperm tissue. Embryo starch content peaked at 12 DAF. At this point, lipid had started to accumulate (Fig. 2b). Starch content declined rapidly over the next 4 d and the storage lipid content increased dramatically (reflected in the kinetics of accumulation of eicosenoic acid: this fatty acid is found exclusively in storage lipids (Lemieux et al., 1990)). It was not possible to dissect embryos free from contamination with surrounding tissues beyond 16 DAF.

Figure 2.

 Accumulation of storage reserves in developing Arabidopsis seeds. (a) Starch contents of whole seeds and testas of wild-type (closed bars) and sex1-3 (open bars) plants at 12 d after anthesis (DAF). Note that the testa accounts for most of the starch in the seed, and that testa starch contents are strongly elevated in sex1-3. Values are mean ± SE of measurements made on three biological replicates, each of 30 seeds per testas from five plants. (b) Pattern of accumulation of starch and fatty acids during embryo development. For starch, values are mean ± SE of measurements on three biological replicates, each of 30 embryos pooled from five plants. Similar results were obtained with two further, separately grown batches of plants. For fatty acids, values are mean ± SE of measurements on four biological replicates, each of 50 seeds pooled from five plants. Closed circles, total fatty acid methyl ester (FAME) content; open circles, eicosenoic acid (C20:1) content. Eicosenoic acid is found exclusively in storage lipids: the kinetics of its accumulation specifically reflect those of storage lipid accumulation.

Impact on embryo starch content of mutations affecting enzymes of starch degradation in the leaf

To provide information about the pathway of starch degradation in oilseed embryos, we first examined the starch contents of mature seeds of Arabidopsis mutants lacking enzymes of leaf starch metabolism (Caspar et al., 1985; Lin et al., 1988; Smith et al., 2005; Fulton et al., 2008; Kötting et al., 2009). As expected, no staining was observed in wild-type seeds or seeds of pgm1-1 and adg1-1 mutants lacking plastidial phosphoglucomutase and the small subunit of AGPase, respectively (Fig. 3). By contrast, staining was observed in mature seeds of several mutants that accumulate excess starch in leaves. These were the sex1 and sex4-5 mutants, lacking the starch phosphorylating enzyme glucan, water dikinase 1 (GWD1: Yu et al., 2001; Ritte et al., 2002) and the glucan phosphate phosphatase SEX4, respectively (Niittyläet al., 2006; Kötting et al., 2009); the bam4-1 mutant lacking one of the chloroplastic β-amylases (BAM4), and the quadruple mutant bam1bam2bam3bam4 lacking four chloroplastic β-amylases (hereafter referred to as bam1234: Fulton et al., 2008) (Fig. 3). Seeds of sex1, sex4 and bam4 contained three to nine times more starch than wild-type seed (Fig. 3) and seeds of bam1234 plants contained nearly 20 times more. Most of the starch in the mature seeds of these mutants was in the testa, but iodine staining revealed that starch was also present in the embryo, mainly in a region above the tip of the radicle (Fig. 3c). At 12 DAF, starch content of sex1-3 seeds was elevated in both embryo and testa (Fig. 2b).

Figure 3.

 Accumulation of starch in seeds of Arabidopsis mutants lacking enzymes of starch metabolism. (a) Mature seed were stained for starch with iodine solution and observed under differential interference optics. Lines were: Col-0, Ws, sex1-1, sex1-3, sex4-5, bam1-1, bam3-1, bam4-1, bam1234, amy3-2, isa3-1, phs1-1, mex1-1, dpe2-5, adg1, pgm1-1. Bar, 200 μm. (b) Starch content of whole mature seeds of mutant and wild-type plants. Values are mean ± SE of measurements made on three biological replicates, each of 30 seeds from five plants. (c) Mature embryos of Col-0, sex1-3, bam4-1 and bam1234 plants, stained with iodine solution to visualize starch and observed under differential interference optics. Bar, 200 μm.

Seeds of three other mutants defective in starch degradation in leaves did not contain starch at maturity. These were the mex1-1 (Niittyläet al., 2004) and dpe2-5 (Chia et al., 2004) mutants, defective in maltose transport from the chloroplast to the cytosol and maltose metabolism in the cytosol respectively, and the bam3-1 mutant, lacking the BAM3 plastidial isoform of β-amylase (Fulton et al., 2008). We also examined mutants lacking enzymes of starch degradation present in leaves but not required for normal rates of starch degradation at night. Seeds of the amy3-2 mutant (lacking the plastidial isoform of α-amylase: Yu et al., 2005), the phs1-1 mutant (lacking the plastidial isoform of glucan phosphorylase: Zeeman et al., 2004) and the bam1-1 mutant (lacking the BAM1 plastidial isoform of β-amylase: Fulton et al., 2008) did not contain starch at maturity (Fig. 3).

We measured the amount of starch in embryos isolated from seeds between 8 DAF and 16 DAF. In wild-type embryos, starch content rose from < 0.01 μg per embryo at 8 DAF to peak at 0.14 μg per embryo at 12 DAF, then fell to 0.01 μg per embryo by 16 DAF. As expected, the starch content of pgm1 embryos was close to the detection limit of the assay throughout development (Fig. 4). Embryos of a strong and a weaker sex1 mutant (sex1-3 and sex1-1 respectively; Yu et al., 2001), bam4 and bam1234 mutants contained 1.5–2.5 times as much starch as wild-type embryos at 12 DAF. In bam1234, starch content was much higher than in wild-type embryos from torpedo stage onwards (Fig. 4). The starch content in these mutants remained higher than in wild-type embryos beyond 12 DAF, and starch was still present in the embryo at maturity. In sex1 mutants containing the null allele sex1-3, the starch content of embryos 16 DAF was 50% higher than at 12 DAF and c. 35 times higher than in wild-type embryos. Iodine staining of sex1-3, bam4 and bam1234 mutant embryos revealed that starch persisted longer in the cotyledons, and especially in the zone above the tip of the radicle, than in wild-type embryos (Fig. 4). The starch content of embryos of sex4, amy3, isa3 and phs1 mutants was not increased relative to wild-type embryos between 8 and 16 DAF (Fig. 4).

Figure 4.

 Accumulation of starch in developing embryos of mutant and wild-type Arabidopsis plants. (a–f) Values are mean ± SE of measurements made on three biological replicates, each of 30 embryos obtained from five plants. Each graph shows values for wild-type plants (squares), plus one or two mutant lines as follows: (a) sex1-1 (diamonds) and sex1-3 (circles); (b) pgm1-1 (circles); (c) sex4-5 (triangles); (d) bam4-1 (triangles); (e) isa1-3 (triangles) and amy3 (circles); (f) phs1-1 (triangles). (g) Developing embryos of Col-0, sex1-3 and bam4-1 were stained with iodine solution and viewed under a dissecting microscope. Numbers beneath the panel correspond to the days after flowering (DAF) at which the samples were harvested. (h) Starch contents of developing embryos of wild-type and bam1234. Upper: plants at 12 d and 14 d DAF. Closed bars, Col-0; open bars, bam1234. Values are mean ± SE of measurements made on three biological replicates, each of 30 embryos obtained from five plants. Lower: developing embryos of bam1234 were stained with iodine solution and viewed under a dissecting microscope. Numbers beneath the panel correspond to DAF at which the samples were harvested. Bar, 200 μm.

Developmental changes in the transcription of genes encoding enzymes of starch synthesis and degradation

We analysed publicly-available microarray data in order to determine changes in transcript levels during embryo development for genes encoding proteins putatively involved in starch metabolism ( (Zimmermann et al., 2004). We considered transcripts for chloroplast-located enzymes putatively involved in the synthesis of starch from hexose phosphate in embryos (Kang & Rawsthorne, 1994), and in the degradation of starch to maltose and glucose in leaves at night (Table S1; Smith et al., 2004, 2005; Fulton et al., 2008). The analysis omitted those genes that were not reliably detected on microarrays (P > 0.06).

Abundances of transcripts encoding enzymes of starch synthesis were generally high in early development, before the point of peak starch content (stage 1–4, Fig. 5a). Transcripts encoding enzymes of starch degradation showed a similar pattern of abundance: amounts generally rose before the onset of starch accumulation (stage 1–3), then fell throughout the period of accumulation and decline (Fig. 5b). The only marked exception to this pattern was the transcript encoding BAM1, which rose throughout development.

Figure 5.

 Expression pattern during embryo development of genes encoding proteins involved in starch metabolism. (a) Enzymes of starch synthesis, (b) enzymes of starch degradation, (c) transporters on the plastid envelope. Publicly available microarray data were obtained from the Genevestigator website (; Zimmermann et al., 2004). Each point is the average from three biological replicates. The list of genes included in the analysis is presented in the Supporting Information, Table S1. The developmental stages shown are: globular, 1; heart, 2; early torpedo, 3; torpedo, 4; walking stick, 5; cotyledon, 6. Gene names are: ISA, isoamylase; PGM, phosphoglucomutase; ADG, ADPglucose pyrophosphorylase; PGI, phosphoglucose isomerase; SS, starch synthase, SBE, starch branching enzyme; GBS, granule-bound starch synthase; DPE2, maltose transglucosidase; PHS, glucan phosphorylase; LDA, limit dextrinase; DPE1, disproportionating enzyme; GWD, glucan, water dikinase; BAM, β-amylase; AMY, α-amylase; SEX4, glucan phosphate phosphatase; GLT, plastidial glucose transporter; GPT, plastidial glucose 6-phosphate transporter; TPT, plastidial triose phosphate transporter; MEX, plastidial maltose transporter.

Transcripts for plastid envelope transporters putatively involved in starch metabolism had distinct patterns of change through embryo development (Fig. 5c). Transcript for GPT1, involved in the import into the plastid of glucose 6-phosphate for starch synthesis (Kang & Rawsthorne, 1994), peaked just before the point of peak starch content. Transcript abundance for MEX1, necessary for export of maltose from chloroplasts at night (Niittyläet al., 2004), was low throughout development, whereas transcript abundance of GLT1, the transporter implicated in export of glucose from chloroplasts at night (Smith et al., 2005), followed a pattern similar to that of GPT1. Transcript abundance for the triose-phosphate transporter TPT fell throughout development from an initially high level.

Rates of starch synthesis in developing oilseed rape embryos

To investigate developmental and diurnal changes in the flux of carbon from sucrose into starch during embryo development, we measured the incorporation into starch of 14C from [14C]sucrose supplied to the petioles of intact, developing siliques of B. napus (as in Morley-Smith et al., 2008). The rate of incorporation of carbon from sucrose into storage products in embryos in this experimental system is comparable with that in intact plants, and linear rates of radiolabel incorporation are maintained for many hours. To estimate the absolute flux, we measured the specific activity of glucose 6-phosphate in embryos and assumed that this was identical to the specific activity of glucosyl residues incorporated into starch. We performed these experiments with siliques of B. napus rather than Arabidopsis because steady-state concentrations of metabolites are unlikely to be retained through the protracted period required for the isolation of the very small embryos of Arabidopsis. The pattern of starch accumulation and loss during embryo development is very similar in B. napus and Arabidopsis (see da Silva et al., 1997; Eastmond & Rawsthorne, 2000). Experiments were performed at three different stages of seed development: early-, mid- and late-oil, as defined by Eastmond & Rawsthorne (2000). At the early-oil stage the starch content of embryos is increasing rapidly. At mid-oil, embryos have maximum amounts of starch and are accumulating oil rapidly. In late-oil stage embryos, oil is reaching its maximum value and starch content is declining steeply.

We found that 14C from sucrose supplied to silique petioles was incorporated into starch in embryos at similar rates in all three developmental stages when the incubation was in the light (Fig. 6). At the first two stages, incorporation was much lower when incubations were in the dark. In late-oil embryos dark and light incubations resulted in approximately the same incorporation of 14C into starch (Fig. 6).

Figure 6.

 Rates of starch synthesis in developing embryos of oilseed rape. The rate of starch synthesis was measured by supplying [14C]sucrose to whole, detached siliques of oilseed rape for 12 h in the light (open bars) or the dark (closed bars). Embryos were dissected from the siliques and used for determination of 14C in starch and in glucose 6-phosphate, and for measurement of glucose 6-phosphate content. The specific activity of glucose 6-phosphate and the 14C content of starch were used to calculate the rate of starch synthesis. Siliques at three developmental stages were used: early oil corresponds to an embryo weight of c. 2.5 mg and is before the maximum starch content; mid-oil corresponds to an embryo weight of c. 3.5 mg and is at or shortly after the peak of starch content; and late oil corresponds to an embryo weight of c. 4.5 mg when oil content is at its maximum and starch content has declined to < 20% of its maximum (Eastmond & Rawsthorne, 2000).

Effect of embryo starch turnover on oil synthesis

We compared the lipid contents of mature wild-type, sex1, bam1234 and sex4 seeds. Although seed weights of wild-type and mutant plants were the same, seeds of sex1 had almost 30% less lipid on both a dry weight and a per seed basis than seeds of wild-type plants (Table 1). Very similar results were obtained for total fatty acid content. The fatty acid composition of sex1 seed was not distinguishable from that of wild-type seed (not shown). Measurement of total lipid content in seeds of the bam1234 mutant also revealed a strong reduction relative to wild-type seeds (Table 1). Similar results have been reported for the starchless pgm1 mutant: the lipid content of mature seeds is reduced by c. 40% relative to that of wild-type plants while seed weight is near normal (Perriappuram et al., 2000). However, the sex4 mutation has very little impact on seed lipid content. The sex4 mutant accumulates less starch in leaves and seeds than either the sex1 or the bam1234 mutants, so the lack of effect on lipid content may simply reflect this less severe starch phenotype.

Table 1.   Impact of mutations that increase embryo starch content on the oil content of mature Arabidopsis seeds
LineaSeed weightbOil content
(μg per seed)(% weight)(μg per seed)
  1. aPlants were grown in a controlled environment in a randomized block design.

  2. bFor seed weights and measurement of oil content by NMR, values are mean ± SE of values from nine biological replicates. Each replicate was a pool of seeds from a pot containing five plants. Seed weights were determined from batches of 100 seeds.

  3. cPlants were grown in an unheated glasshouse in a randomized block design, between October and December. Natural daylight was supplemented with lighting to give a 16-h day.

  4. Values for mutant lines marked with asterisks are statistically significantly different (Student’s t-test) from wild-type values: *, = 0.00054; **, = 0.00040; ***, = 0.000000014.

Wild type15.7 ± 0.2632.1 ± 1.65.0 ± 0.2
sex1-315.5 ± 0.3123.7 ± 0.5*3.7 ± 0.1**
sex415.9 ± 0.3430.5 ± 0.74.9 ± 0.7
Wild typec 45.0 ± 0.1 
bam1234c 30.1 ± 0.4*** 


Spatial and temporal patterns of starch accumulation in the seed

Our detailed examination of the location of starch in developing Arabidopsis seed provides an unexpectedly complex and dynamic picture. First, as previously reported, starch accumulates in the embryo, the endosperm and the testa. Starch appears in the testa at the early-globular stage of embryo development before embryo starch starts to accumulate, and is present mainly in the outer but also in the inner integuments (Beeckman et al., 2000; Western et al., 2000; Windsor et al., 2000). The starch content of the testa is four times that of the embryo when embryo starch content is at its peak. In our experiments starch persists in the testa well into the cotyledon stage of embryo development when embryo starch has largely disappeared. The starch content of the cellularized endosperm follows a temporal pattern more similar to that of the embryo than the testa, rising from c. 8 DAF, peaking at c. 12 DAF, then declining rapidly.

Second, starch accumulates in specific temporal and spatial patterns within the developing embryo. Although we have not quantified amounts of starch in different tissues of the embryo, the patterns we observe by iodine staining and by microscopy of thin sections are consistent with each other and with changes over time in the starch content of the embryo as a whole. Accumulation starts at c. 8 DAF in two zones: a hypocotyl zone subtending the apical meristem and a zone subtending the radical meristem. Starch then appears throughout the hypocotyl and radicle and in the cotyledons. Cotyledon and hypocotyl starch disappears by 14 DAF; starch then disappears from the radicle. Thus cotyledon starch persists for only 3–4 d, while starch is present in the zone subtending the radicle meristem for at least 6 d.

Previous consideration of carbon partitioning between storage products in developing oilseeds has assumed that starch metabolism occurs in the same cells as all other storage functions of the seed or embryo. In particular, the progression of events during the maturation phase of the seed has been portrayed as a shift from starch to lipid storage (e.g. Eastmond & Rawsthorne, 2000; Schwender et al., 2003; Vigeolas et al., 2003, 2004; Lin et al., 2006; Musgrave et al., 2008). Our results show that these assumptions are oversimplifications. Within the Arabidopsis seed, most of the flux of carbon from imported sucrose to starch occurs in the testa at relatively early stages of development, whereas most of the flux to storage oil and protein occurs in the embryo at later stages of development. Within the Arabidopsis embryo, our data indicate that flux to starch may be larger and occur over a longer time in zones within the axis (radicle and hypocotyl) than in the cotyledons. By contrast, most of the lipid accumulation in Arabidopsis embryos is in the cotyledons rather than the axis. In mature seeds the cotyledons contain 60% and the axis 27% of the total seed oil (Li et al., 2006). Thus the quantitative and temporal relationship between the flux of carbon to starch and the flux to oil is very different in different parts of the embryo.

The temporal and spatial patterns of starch accumulation in the developing embryo are strongly reminiscent of patterns of activity of several enzymes of primary metabolism, visualized by histochemical stains (Baud & Graham, 2006). For four enzymes implicated in the synthesis of starch from sucrose, AGPase, phosphoglucose isomerase, UDPglucose pyrophosphorylase and phosphoglucomutase, activity was much higher in torpedo-stage embryos in zones close to the radicle meristem and subtending the apical meristem than in other embryo tissues. Rates of cell division are high at this developmental stage (Mansfield & Briarty, 1991). Activity then became more broadly distributed through embryo tissues, declining towards embryo maturity. Other enzymes for which activity was highest in one or both of these zones at torpedo stage included glucose 6-phosphate dehydrogenase, 6-phosphogluconate dehydrogenase, pyrophosphate-dependent phosphofructokinase, glyceraldehyde 3-phosphate dehydrogenase and alcohol dehydrogenase (Baud & Graham, 2006). These observations imply that zones subtending the radicle and apical meristems are sites of high general metabolic activity as well as starch turnover at torpedo stage.

Starch turnover in the embryo

Starch within the developing oilseed embryo is subject to substantial turnover throughout the maturation phase of development. Evidence for this comes from two main observations. First, mutations that eliminate enzymes exclusively involved in starch degradation can affect not only the rate of decline in starch content but also the rate at which starch content increases and the peak starch content in the embryo. This is true for the sex1, bam4 and bam1234 mutants. Thus starch degradation occurs at the same time as synthesis from the onset of accumulation. Second, [14C]sucrose supplied to intact siliques of oilseed rape is metabolized to starch within the embryo at comparable rates during the periods of both net accumulation and loss of starch. This shows that synthesis is occurring throughout the period of loss of starch from the embryo.

Our data show directly that starch turnover occurs in vivo, confirming indications from previous indirect, in vitro and invasive experiments. These include demonstrations that activities of enzymes of both starch synthesis and degradation are present throughout the periods of starch accumulation and loss in oilseed rape seeds (da Silva et al., 1997), that substrates supplied to isolated oilseed rape embryos are used for starch synthesis at developmental stages at which the total starch content was declining (Eastmond & Rawsthorne, 2000), and that 14C from [14C]sucrose injected directly into seeds is incorporated into starch during the decline in seed starch content (Vigeolas et al., 2003).

We observed that the rate of starch synthesis in oilseed rape embryos is higher in the light than in the dark at early- and mid-oil stages. This raises the possibility that at least part of the starch is produced via photosynthetic carbon assimilation in the light. Isolated embryos are capable of light-driven CO2 assimilation, and sufficient light reaches the embryo on the plant to permit assimilation (King et al., 1998; Ruuska et al., 2004). However, stable isotope analysis of metabolism in isolated embryos is not consistent with the operation of a complete Calvin cycle, and provides no evidence for flux of carbon from CO2 via Rubisco to fructose 6-phosphate, the Calvin cycle intermediate required for starch synthesis (Schwender et al., 2004).

An alternative explanation for the light stimulation of starch synthesis is that the supply of ATP for the generation of ADPglucose (the substrate for starch synthases) via AGPase is greater in the light than in the dark. There is good reason to think that the supply of ATP in general is higher in the light than in the dark. First, the light reactions of photosynthesis operate in the embryo. The provision of ATP and reductant via these reactions is held to contribute to the light stimulation of fatty acid synthesis in Brassica embryos observed in our silique-incubation system (Morley-Smith et al., 2008; Fig. 6) and in isolated embryos (Bao et al., 1998; Ruuska et al., 2004). Second, the oxygen tension is higher in siliques in the light than in the dark, because of oxygen generation via photosynthesis in the light. At ambient external oxygen (21 kPa) in the light, the oxygen tension inside the silique is 16 kPa (Porterfield et al., 1999; Vigeolas et al., 2003) whereas in the dark it drops to c. 12.5 kPa in the silique (Porterfield et al., 1999). When reductions of this magnitude are imposed by placing siliques in external oxygen tensions below ambient in the light, the ATP : ADP ratio of the seeds is reduced by c. one-third and the flux of carbon to starch by up to half (estimated from Vigeolas et al., 2003).

Pathways of starch metabolism in the seed

Our data are consistent with the idea that starch degradation in the Arabidopsis seed is primarily via a pathway similar to that described for leaves at night. Starch turnover is affected by several mutations that eliminate enzymes necessary for normal rates of starch degradation in leaves. The loss of GWD1 (in the sex1 mutant) has a strong effect, as it does in leaves. The rate of loss of starch is reduced in the testa, and in all three starch-accumulating zones in the embryo. Thus phosphorylation of the granule surface seems to be important in starch degradation in nonphotosynthetic tissues as well as in leaves. The importance in seeds of subsequent dephosphorylation via the glucan phosphate phosphatase SEX4 is less clear. In leaves, loss of this protein reduces the rate of starch degradation and results in build-up of phosphorylated malto-oligosaccharides in the chloroplast stroma (Kötting et al., 2009). In seeds, the rate of starch loss from the testa is reduced but there is no marked effect on the pattern of starch accumulation and loss in the embryo. This does not necessarily mean that SEX4 is unimportant in starch degradation in the embryo. A detailed analysis of starch-related metabolites in wild-type and sex4 mutant embryos would be required to establish whether SEX4 plays a role in embryo starch degradation.

β-Amylases are essential for normal starch degradation in both the testa and the embryo, whereas loss of either the only plastidial α-amylase, AMY3, or the plastidial glucan phosphorylase, PHS1, has no effect. This is also observed in leaves: β-amylases are important for starch degradation at night (Fulton et al., 2008) whereas AMY3 and PHS1 are not (Zeeman et al., 2004; Yu et al., 2005). However, the embryo phenotypes of the bam mutants indicate that the relative importance of BAM isoforms may differ between embryos and leaves. Loss of either BAM3 or BAM4 reduces the rate of starch degradation in leaves. In seeds, loss of BAM4 reduces the rate of decline of the starch content but loss of BAM3 has little effect. This result is particularly interesting because BAM4 is reported to lack β-amylase activity. Genetic analysis suggests that it facilitates or regulates starch degradation, probably through interaction with other starch-degrading enzymes (Fulton et al., 2008). The strong embryo phenotype of the quadruple bam mutant bam1234 indicates that, as in leaves, more than one BAM isoform can contribute to starch degradation in the plastid.

Loss of the debranching enzyme ISA3 has no effect on the starch content of seeds even though it is essential for normal starch degradation in leaves. This result suggests that one or more of the other three debranching enzymes present in Arabidopsis plastids – ISA1, ISA2 and limit dextrinase (LDA) (Delatte et al., 2005; Wattebled et al., 2005; Streb et al., 2008) – can also participate in the cleavage of α-1,6 glucosidic linkages.

Our results do not establish unambiguously the fate of glucose, maltose and malto-oligosaccharides released by starch degradation in embryo plastids. At least some of this carbon may be metabolized inside the plastids via glycolysis or the oxidative pentose phosphate, after conversion to hexose phosphate via glucan phosphorylase (in the case of malto-oligosaccharides) or the plastidial hexokinase (At1g47840, Claeyssen & Rivoal, 2007). Maltose and glucose may be exported from the plastid via the maltose transporter MEX1 and the glucose transporter GLT respectively. In the cytosol, maltose may be converted to hexose phosphate via the transglucosidase DPE2 and the glucan phosphorylase PHS2, as appears to happen in leaves (Smith et al., 2005).

Our results support the view that measurements of transcript abundance cannot provide direct information about the nature and control of fluxes in primary metabolism (Smith et al., 2004; Smith & Stitt, 2007). There is essentially no relationship in the Arabidopsis embryo between the relative importance of the enzymes of starch metabolism – deduced from mutant analysis – and the relative abundances of transcripts for these enzymes. For example, amounts of transcript for BAM4 are very low throughout embryo development and those for BAM3 are much higher, yet loss of BAM4 has a larger overall effect on starch degradation than loss of BAM3. Similarly, transcripts for DPE2, PHS1, PHS2 and BAM1 are very abundant at torpedo stage yet none of these enzymes is necessary for normal starch turnover in the embryo.

The role of starch turnover during seed development

Previous suggestions for a role for starch accumulation and turnover during embryo development attempted to relate this phenomenon to specific features of embryo metabolism and development. Suggested roles included a carbon reserve for the synthesis of lipids and sugars later during development, a source of substrate for the plastidial oxidative pentose phosphate pathway or for the synthesis of sugars that accumulate during desiccation and the establishment of the developing embryo as a sink organ (see the Introduction). Taken as a whole, our results lead us to a different conclusion. Starch accumulates and turns over in the embryo in cells that are dividing or have recently divided – initially adjacent to the radicle and apical meristems and then in the cotyledons. We suggest that starch turnover is a normal feature of cells passing from division to differentiation phases, in embryos and in many other plant organs.

Our suggestion is supported by numerous examples. In the shoot apex of Pinus pinaster starch accumulates in some meristematic tissues and in the pith parenchyma that lies immediately beneath the apical meristem. Starch accumulation correlates with the organogenic activity of the apical meristem and the development and growth of new organs from the apex (Jordy, 2004). During the early stages of leaf development in tomato, starch granules are present specifically in the submeristematic region underlying the leaf primordium and in the primordium itself (Pien et al., 2001). Starch also accumulates in cells adjacent to meristems in regenerative callus (Karim et al., 2006) at either side of the intercalary meristem in the base of flower scapes of Narcissus (Chen, 1966) during growth of the primary thickening meristem of Allium cepa stems (Ernst & Bufler, 1994), in the apical meristem of Antirrhinum (Mérida et al., 1999), and in apical meristems of Sinapis alba following evocation of flowering by long-day treatment (Havelange et al., 1974).

In general, these observations are consistent with the idea that starch acts as a temporary reserve of carbon in cells that are dividing or in the early stages of differentiation. The existence of a carbon reserve may be important in such cells because they are expected to undergo rapid metabolic and developmental changes, and hence large fluctuations in demand for carbon. Mutations that affect starch turnover do not cause obvious defects in development of embryos or other plant parts (e.g. pgm mutants of Arabidopsis (Caspar et al., 1985), tobacco (Hanson & McHale, 1988) and pea (Harrison et al., 2000)), but the possibility of subtle effects on meristematic activity in starch mutants cannot be ruled out. A starch reserve may be of particular significance for meristem function in fluctuating environments or under environmental stress.

Mutant plants with strongly increased or reduced embryo starch content also have reduced seed lipid content. This could reflect a direct relationship within the embryo between flux of carbon into lipid and flux into and out of starch. However, the effect of the starch mutations on seed lipid content could also be indirect. All of the mutations also profoundly affect starch turnover in other parts of the plant, thus their effects on seed lipid content could result from an altered supply of carbon to the embryo during its development. For example, the mutations profoundly alter carbohydrate metabolism in the leaves and thereby affect short- and long-term carbon partitioning in the plant as a whole (Caspar et al., 1985; Schulze et al., 1991; Zeeman & ap Rees, 1999; Smith & Stitt, 2007). The mutations also strongly affect starch turnover in the testa, which may have consequences for carbon partitioning within the seed. An argument for an indirect effect of the starch mutations on seed lipid content is provided by experiments in which starch accumulation was reduced specifically within the embryo of oilseed rape plants. When AGPase was downregulated using an embryo-specific promoter, the transgenic embryos had only half the starch content of normal embryos but had near-normal final lipid contents (Vigeolas et al., 2004).


We thank Sameer Sengupta for technical assistance, Sheila Mitchell for expert horticultural support, Alan Jones (John Innes Centre) for GC-MS analysis of fatty acid methyl esters, and Sue Bunnewell (John Innes Centre) for help with microscopy. Seeds of bam mutants were the kind gift of Sam Zeeman (ETH, Zurich). This work was supported by a Core Strategic Grant from the Biotechnology and Biological Sciences Research Council (UK) to the John Innes Centre, and by a Biotechnology and Biological Sciences Research Council Grant to AMS (208/P19448).