The tps1 mutant, which is disrupted in the TREHALOSE-6-PHOSPHATE SYNTHASE 1 gene, has been previously characterized as a recessive embryo lethal. tps1 embryos do not develop past late torpedo or early cotyledon stage. We report here that at the ultrastructural, biochemical, and transcriptional levels tps1 exhibits many features typically associated with the maturation phase. The appearance of storage reserve transcripts and organelles follows the same temporal pattern in tps1 and wild-type (WT) embryos in the same silique as does accumulation of storage lipid and protein. The mutant plastids accumulate large starch granules that persist until the end of seed development, in contrast with WT plastids where starch accumulation is transient. The transcriptome of tps1 embryos shows a coordinate downregulation of genes involved in starch and sucrose degradation. Interestingly, genes involved in lipid mobilization and gluconeogenesis are induced in tps1 embryos. The cell walls of tps1 embryos show a remarkable degree of thickening at the ultrastructural level and immunodetection of cell wall components shows that altered deposition of pectins accounts for this altered morphology. Consistent with this at the transcriptome level, genes involved in sugar nucleotide and pectin metabolism are altered in the mutant. The frequency of cell division in tps1 embryos is half that of the wild type at the heart and torpedo stages. These results suggest that TPS1 may play a major role in coordinating cell wall biosynthesis and cell division with cellular metabolism during embryo development.
Seed development in Arabidopsis can be divided into three phases. In the first, known as early embryo morphogenesis, the embryo develops through a series of cell divisions towards a torpedo-shaped morphology and acquires the basic cell patterning of the mature embryo. The second phase, embryo maturation, is dominated by the accumulation of storage compounds, differentiation of related cellular structures and cell expansion. During the final phase, the seed undergoes desiccation and the embryo enters a quiescent state (Harada, 1997).
The genetic programme directing morphogenesis and cell differentiation during embryo development requires the coordinated expression of a broad set of genes with major shifts in gene expression occurring as the embryo progresses through the different phases (Ruuska et al., 2002). Provision of nutrients from the mother plant is important in controlling progression through development, and sucrose in particular plays a central role in this process (Wobus and Weber, 1999). During seed development the embryo metabolizes sucrose imported from maternal tissues in order to sustain cell patterning, but, as the embryo progresses to the maturation phase, imported metabolites are directed to the accumulation of storage reserves. Distinguishing between metabolic and regulatory roles played by sugars can be extremely challenging, particularly as the metabolic roles are often essential and their disruption can also have secondary effects on development.
The discovery that a trehalose biosynthetic gene, TREHALOSE-6-PHOSPHATE SYNTHASE-1 (TPS1), is essential for embryo development in Arabidopsis has established a regulatory role for trehalose metabolism in this process (Eastmond et al., 2002). Unlike sucrose, trehalose does not constitute a major source of carbon or energy in plants, with its concentration in most cases (0.15 mg g−1 dry weight) being at least an order of magnitude lower than that of sucrose. Only in some desiccation-tolerant species has trehalose accumulation been reported, consistent with its role as osmoprotectant in microorganisms and animals (Goddijn and van Dun, 1999). Trehalose biosynthesis in plants occurs in two steps. The first, catalysed by TPS, involves the transfer of glucose from UDP-glucose to glucose-6P to form trehalose-6P (T6-P). This in turn is converted to trehalose by the second enzyme in the pathway, trehalose-6-phosphate phosphatase (TPP).
Overexpression of the Escherichia coli (otsA) or yeast trehalose-6P synthase homologues in plants produces alterations in growth and source–sink relations (Goddijn et al., 1997; Romero et al., 1997). Moreover, overexpression of AtTPS1 showed that this gene is a regulator of glucose signalling (Avonce et al., 2004). Such observations, along with the fact that trehalose is present at very low levels in most plant tissues, led to the proposal that, rather than contributing directly to metabolic processes, trehalose or associated metabolites instead play a regulatory role in plant development (Goddijn and Smeekens, 1998). Evidence that it is the product of the TPS reaction, T6-P, rather than the TPS protein itself that plays the regulatory role comes from the fact that the distantly related E. coli otsA gene can complement the Arabidopsis tps1 mutant phenotype, and wild-type Arabidopsis plants overexpressing the otsA gene are able to grow faster than untransformed controls on medium containing sugars (Schluepmann et al., 2003). A number of hypotheses have been proposed to explain the mechanism of action of the proposed signalling function of T6-P. Plants overexpressing TPP show accumulation of sugar phosphates, suggesting that glycolysis could be compromised by the depletion of T6-P (Schluepmann et al., 2003). In yeast, T6-P regulates glycolysis by inhibiting the activity of hexokinase (Thevelein and Hohmann, 1995) but, although several lines of evidence suggest that T6-P interacts with the hexokinase-dependent signalling pathway in plants (Avonce et al., 2004; Schluepmann et al., 2003), in vitro studies show that T6-P does not inhibit plant hexokinase enzyme activity (Eastmond et al., 2002; Wiese 1999).
Indirect evidence suggests that trehalose and T6-P may interact with the plant equivalent of the yeast sucrose non-fermenting-1 (SNF1) protein kinase complex. Sucrose non-fermenting related kinase 1 (SnRK1) genes have been identified in many plant species, and regulate the activity of key enzymes such as sucrose phosphate synthase (SPS) and ADP-gluc pyrophosphorylase (AGPase) in response to sucrose availability in the cell (Halford and Paul, 2003). Schluepmann et al. (2004) recently observed that expression of the Arabidopsis homologue of the yeast SNF1 kinase, AtAKIN11, correlates with T6-P levels, suggesting a link between trehalose metabolism and the SnRK1 signalling system. AKIN11 can complement the Δsnf1 mutation in yeast and interacts with the WD protein PRL1 modulating the sucrose response in Arabidopsis (Bhalerao et al., 1999; Ferrando et al., 2001). Interestingly, a recent report established that one of the mechanisms of action of T6-P in the regulation of starch synthesis is via redox activation of ADP-glucose pyrophosphorylase (Kolbe et al., 2005). This post-translational activation depends on the expression of SnRK1 and is a thioredoxin-mediated response to cytosolic sugar levels.
The tps1 mutant provides an excellent opportunity to investigate metabolite regulation of embryo development in plants. We have used a combination of biochemical, microscopic and transcriptomic approaches to show that tps1 embryos progress into maturation and cell differentiation, despite the fact that they suffer morphological arrest. We conclude that perturbation in trehalose signalling leads to decreased rates of cell division and deregulation of cell wall biosynthesis. These observations lead to a new conceptual framework to address the mechanisms by which trehalose metabolism and sugar signalling modulate growth and development in plants.
tps1 wrinkled seeds are able to germinate
Despite the fact that tps1 mutants arrest at the torpedo stage of embryo development, we found that, when wrinkled tps1 seeds were cold stratified and incubated on agar medium for an extended period of 1–3 weeks, 30–40% of the embryos expanded and broke out of the seed coat (Figure 1a). A limited expansion of underdeveloped cotyledons, hypocotyls and root tissue, as shown in Figure 1(a), occurred and lateral roots initiated. However, growth of tps1 plantlets was extremely slow and they did not progress beyond a small 5-mm-diameter rosette that remained in the vegetative stage until senescence (Figure 1b).
Development of tps1 embryos followed a normal patterning process, but the rate of growth was decreased. Consequently, at 18 days after flowering (DAF) when the wild-type embryos had reached the end of the maturation phase with fully expanded cotyledons, sibling tps1 embryos in the same silique had only just completed morphogenesis and reached the torpedo or early cotyledon stage (Figure 2a). Wild-type embryos reached the torpedo stage at 7 DAF (Figure 2b). The growth and expansion of cotyledons during maturation was driven by cell division and cell elongation, with storage deposition occurring at a high rate.
The cell morphology of WT and tps1 embryos was analysed at different stages during embryogenesis using electron microscopy (Figure 3). In WT late torpedo-stage embryos (7 DAF), cells in the base of the cotyledons contained large and prominent nuclei, numerous mitochondria and chloroplasts. Lipid bodies were also present along with vacuoles containing protein deposits (Figure 3a). WT bent cotyledon-stage cells at 15 DAF, however, contained more abundant oil and protein bodies with nuclei being less prominent (Figure 3c). Chloroplasts were also more abundant and the granal system was more differentiated (Figure 3c). In torpedo-stage tps1 embryos, the cellular structure was significantly different from that of the WT torpedo-stage cells and showed much greater similarity to that of the bent cotyledon-stage cells (Figure 3b). Oil and protein bodies were abundant and densely packed at the cell periphery, in an arrangement that was characteristic of WT embryos at the late maturation phase (Mansfield and Briarty, 1993).
Chloroplasts in WT bent cotyledon-stage embryos were more abundant than in WT torpedo-stage embryos (Figure 3a,c) and the granal system was more developed with infrequent small starch granules (Figure 3d,f). In marked contrast, torpedo-stage tps1 chloroplasts contained large starch granules that masked the internal membrane system (Figure 3e). Despite these large starch granules, plastid development in tps1 appeared to progress in a similar temporal pattern to that found in sibling WT embryos, with tps1 heart-stage embryos containing well-differentiated chloroplasts equivalent to those found in bent cotyledon-stage WT embryos (Figure 4c,d). WT heart-stage embryos contained pre-granal undifferentiated plastids (Figure 4a,b).
Another defining characteristic of tps1 at the ultrastructural level was the presence of much thicker cell walls compared with WT at both the torpedo and the cotyledon stages (Figure 3g–l). These thicker cell walls were more pronounced at the epidermal cell junctions (Figure 3g–i).
tps1 mutant embryos accumulated proteins, lipids and sugars
In order to further characterize the metabolic processes occurring in tps1 embryos, time-course analyses for key metabolites were performed in the mutant background and compared with the WT (Figure 5). Embryo volumes were first calculated by direct measurement of the length and width of embryos at different stages and assuming an approximately cyclindrical shape (Figure 5a). This allowed us to calculate metabolite concentrations and to compare metabolite content in embryos exhibiting different shapes and volumes. Consistent with the appearance of oil bodies, fatty acids accumulated in tps1 embryos, but the concentration was significantly less than that in sibling cotyledon-stage embryos (Figure 5b). Between 8 and 18 DAF, the amount of 16:0 and 18:0 decreased, while 18:3, 20:1 and 22:1 increased in WT embryos, in good correlation with previous reports in Arabidopsis ecotype wassilewskija (WS) (Baud et al., 2002). Similar trends were observed in tps1 embryos between days 12 and 18, showing that the evolution of fatty acid composition in the mutants did not differ from that in the WT (data not shown).
Protein accumulation in tps1 showed a similar trend to that of fatty acid accumulation, tripling between 12 and 18 DAF (Figure 5c). However, the concentration of proteins in the 18-DAF torpedo-stage tps1 was still only about half the corresponding values in WT.
In contrast to storage reserve compound concentrations, soluble sugar concentrations were significantly higher in tps1 embryos compared with the WT (Figure 5) In the WT, sucrose concentrations increased from the torpedo to the cotyledon stage and then declined slightly during the maturation and desiccation phases (Figure 5d). Hexose sugar concentrations in WT embryos decreased between 8 and 10 DAF and remained at low levels during the remainder of seed development (Figure 5e). These changes in sugar levels during embryo development were similar to those reported in previous studies in seeds of the WS ecotype (Baud et al., 2002). In tps1 embryos, sucrose and hexose concentrations were three to four times higher than those in WT embryos aged 12 DAF (Figure 5d,e). Soluble sugar concentrations in tps1 embryos decreased between 12 and 18 DAF to reach levels similar to those in the WT. Levels of starch were also 3 to 4 times higher in tps1 compared with the WT at 12 DAF and these also decreased up to 18 DAF (Figure 5f).
tps1 torpedo-stage and WT bent cotyledon-stage embryos exhibited similar gene expression profiles
To further characterize tps1, transcriptomic analyses were performed using Affymetrix ATH1 arrays (Affymetrix, High Wycombe, UK). The experimental design included torpedo-stage tps1 embryos (15 DAF) and two WT reference samples representing torpedo-stage (7 DAF) and bent cotyledon-stage embryos (15 DAF) These stages were chosen to allow comparison of tps1 mutant embryos with the morphologically similar torpedo stage and with the temporally equivalent upturned-U cotyledon stage (Figure 2b). The bent cotyledon stage represents the beginning of storage reserve accumulation, which proceeds until the embryo reaches maturity, when the dehydration programme is prevalent. The amount of material available for experimentation from the stalled torpedo-stage tps1 embryos is limiting, and to overcome this we performed a single round of RNA amplification on batches of 50 torpedo-stage tps1 and 50 torpedo-stage and bent cotyledon-stage WT embryos. RT-PCR analysis of TPS1 demonstrated the accuracy of the sampling method for embryos from hemizygous siliques, with transcript present in the WT samples but absent in the tps1 samples (data not shown).
Previous studies using cDNA microarray analysis of developing seeds showed that major changes in the expression of key genes occur between 8 and 13 DAF (Ruuska et al., 2002). Comparison of the transcriptome profiles of torpedo- and bent cotyledon-stage embryos confirmed the findings of these studies, showing a general induction of storage reserve-related genes, repression of starch metabolism-related genes, and upregulation of photosynthetic genes (Figure 6a,b; Tables 1 and 2).
Table 1. Expression of storage reserve- and lipid mobilization-related genes in wild-type torpedo- and cotyledon-stage embryos using the 22K ATH1 microarray (Affymetrix)
Average signal value
Fold change cotyledon/ torpedo
Signal values are the average of three independent replicates.
2S albumin 1 precursor
12S cruciferin seed storage protein
Fatty acid elongase 1
Biotin carboxyl carrier protein 2
Acyl carrier protein 1 precursor
3-ketoacyl-acyl carrier protein synthase I
3-ketoacyl-acyl carrier protein synthase III
Omega-6 fatty acid desaturase 2
Omega-3 fatty acid desaturase 3
Fatty acid elongase 1
Light-harvesting chlorophyll a/b binding protein
Ribulose-bisphosphate carboxylase small unit
Pyruvate orthophosphate dikinase
Table 2. Expression of putative cytoplasmic and plastidial glycolytic genes in wild-type torpedo- and cotyledon-stage embryos
Average signal value
Fold change cotyledon /torpedo
Signal values are the average of three independent replicates.
The gene expression profile of tps1 embryos was overall more similar to that found in bent cotyledon-stage embryos than to that found in torpedo-stage embryos (Figure 6c,d; Figure S1). Based on a 1% false discovery rate cut-off in the analysis of the microarray data, the number of genes differentially expressed between the torpedo-stage and tps1 samples was 2781, while the number of genes showing significant changes between cotyledon-stage and tps1 samples was 474 (Table 3). The expression of genes associated with storage reserve accumulation in tps1 was similar to that found at the bent cotyledon stage (Figure 6c; Figure S1), indicating that, in spite of the delayed growth, tps1 mutants induce maturation-phase storage reserve transcripts in a similar manner to normally developing embryos in the same silique. These data support our previous results showing a strong expression of reporter genes under the control of storage reserve promoters expressed in the tps1 genetic background (Gomez et al., 2005).
Table 3. Number of genes differentially expressed between wild-type (WT) torpedo- and cotyledon-stage embryos; WT torpedo-stage and tps1; and WT cotyledon-stage and tps1
Number of genes differentially expressed
FDR, false discovery rate.
Most of the genes that differed in expression between the WT torpedo-stage and tps1 torpedo-stage samples also differed between the WT torpedo-stage and WT bent cotyledon-stage samples, as would be expected, considering that the tps1 torpedo more closely resembles the transcriptome of the bent cotyledon stage (Figure 6b–d). Consistent with this, compared with the WT torpedo stage, the tps1 torpedo stage exhibited much higher level expression of genes typically induced during the transition from the torpedo to the bent cotyledon stage, including those involved in photosynthesis, the glyoxylate cycle, fatty acid synthesis, oil body synthesis and nitrate metabolism (Figure 6b,c). Interestingly, a number of late embryogenesis-associated genes and some stress-related genes were expressed in tps1 at even higher levels than in bent cotyledon-stage embryos (Figure S1), which is consistent with our observation that the tps1 torpedo-stage embryos survived desiccation (Figure 1).
Genes associated with sucrose and starch degradation were downregulated in tps1
In tps1, transcripts associated with sucrose synthesis, including two sucrose phosphate synthases (At1g04920 and At5g20280), were downregulated in comparison with torpedo- and cotyledon-stage WT embryos (Table 4). Transcripts associated with sucrose degradation, including two SUSs (At1g73370 and At5g20830) and vacuolar invertase (At1g12240), were also downregulated compared with both torpedo- and cotyledon-stage WT embryos (Table 4). Transcripts associated with starch synthesis were not significantly altered in tps1 embryos, but a number associated with starch degradation, including several amylases, glucan phosphorylase, and glucan water dikinases (At1g10760; SEX1, At5g26570), were repressed relative to the WT cotyledon stage (Table 4; Figure S1).
Table 4. Central carbon metabolism-associated transcripts repressed in tps1
Average signal value
Fold change cotyledon/tps1
*indicates absent calls in tps1, according to Micro Array Suite 5.0 software.
Sucrose synthase, putative
Citrate synthase-like protein
Sucrose–phosphate synthase, putative
Starch excess protein R1
Sucrose–phosphate synthase-like protein
1,4-alpha-glucan branching enzyme
Snf1-related protein kinase KIN11 (AKIN11)
AKIN11 expression correlates with T6-P levels in Arabidopsis plants overexpressing TPS and TPP (Schluepmann et al., 2004). In good agreement with these results, the expression of AKIN11 was twofold lower in the tps1 mutant compared with both the WT cotyledon and torpedo stages (Table 4), suggesting an interaction of TPS1 or T6-P with the SnRK1 signalling pathway.
Glycolysis-related transcript levels in tps1 embryos were broadly similar to those of WT cotyledon embryos, with the only exception being a cytosolic glyceraldehyde-3-phosphate dehydrogenase, which remained at the same level as in the torpedo stage instead of decreasing as in the cotyledon stage. Overall, plastidial glycolytic genes were induced during embryo maturation while their cytosolic counterparts were moderately repressed (Table 2). The transcription levels of the six hexokinase genes were not significantly altered in tps1 (Supplementary material). This lack of effect on glycolytic transcripts in tps1 is somewhat surprising given that a number of reports link alteration in trehalose metabolism and T6-P levels with altered glycolytic flux (Eastmond et al., 2002; Schluepmann et al., 2003; Thevelein and Hohmann, 1995) and hexokinase-mediated sugar sensing (Avonce et al., 2004).
Genes involved in lipid mobilization were induced in tps1 relative to bent cotyledon-stage embryos
In addition to the induction of storage lipid synthesis-related genes, WT bent cotyledon-stage embryos also showed a significant and coordinated induction of transcripts corresponding to the major pathways involved in the mobilization of storage lipids to sugars (Table 1; Figure S1). For instance, transcripts associated with β-oxidation, the glyoxylate cycle and gluconeogenesis were induced during maturation (Table 1; Figure S1). These results are to an extent anomalous, as they indicate an induction of both synthetic and degradative processes during maturation, but they are consistent with the findings of a number of previous reports and may be attributable to a significant amount of lipid re-modelling and fatty acid turnover occurring specifically during the late maturation process (Chia et al., 2005; Eastmond and Graham, 2001).
Transcripts associated with storage lipid mobilization pathways are induced in tps1 and, in the case of the glyoxylate cycle enzymes malate synthase (MS) and isocytrate lyase (ICL), transcript levels were more than twofold higher than in WT bent cotyledon-stage embryos (Table 5). When transgenic lines expressing the β-glucuronidase (GUS) reporter gene under the control of the ICL promoter were crossed into TPS1/tps1 plants, the expression of GUS in tps1 torpedo-stage embryos was higher than in WT bent cotyledon-stage embryos, in good correlation with the steady-state transcript levels obtained from the microarray analysis (Figure S2). In tps1 embryos the levels of transcripts associated with gluconeogenesis, such as phosphoenolpyruvate carboxykinase (PEPCK) and pyruvate phosphate dikinase, were also higher than in the WT cotyledon-stage embryos (Table 5). As with ICL, the expression of GUS under the PEPCK promotor was increased in tps1 embryos (Figure S2).
Table 5. Central carbon metabolism-associated transcripts induced in tps1
Average signal value
Fold change tps1/cotyledon
Short-chain alcohol dehydrogenase-like protein
Putative isocitrate lyase
Pyruvate kinase-like protein
Cell wall deposition was altered in tps1
As mentioned above, one of the most striking features observed in the electron microscopic analysis of tps1 torpedo-stage embryos was an increased deposition of cell wall material (Figure 3g–l). Transcript levels of genes related to cell wall metabolism were altered in tps1 embryos relative to the WT stages (Table 6). Several genes associated with cell wall synthesis were downregulated, including members of the UDP-D-xylose 4-epimerase and UDP-glc 4-epimerase families and several pectin esterases which were downregulated twofold with respect to WT (Table 6). In order to determine if these transcriptional and morphological changes are accompanied by changes in the composition of the cell wall, we performed immunolocalization using a set of monoclonal antibodies against specific cell wall components (Figure 7). Detection of partially methylated homogalacturonan (HG) compounds, using JIM5 (de-esterified) and JIM7 (methyl-esterified) antibodies (Clausen et al., 2003), showed that deposition of these polymers in tps1 torpedo embryos was similar to that in bent cotyledon-stage WT embryos. However, the pattern of HG deposition in the mutant differed from that in the wild type in the irregular distribution and increased amounts at cell junctions (Figure 7). JIM5 and JIM7 antibodies did not appear to detect epitopes in the cell walls of WT torpedo-stage embryos (Figure 7). The LM2 antibody, which is directed against β-linked arabinogalactan proteins (Smallwood et al., 1996), did not show any cross-reactivity with either torpedo- or bent cotyledon-stage WT embryos, but it did show a significant cross-reaction with tps1 cell walls (Figure 7).
Table 6. Cell wall-related transcripts downregulated in tps1 mutant embryos
Average signal value
Fold change cotyledon/tps1
Glycosyl hydrolase family 35 (beta-galactosidase)
Glycosyl hydrolase family 9 (endo-1,4-beta-glucanase)
Putative cellulose synthase catalytic subunit
Laccase (diphenol oxidase)
UDP-glucose 4-epimerase-like protein
Glycosyl hydrolase family 35 (beta-galactosidase)
Identical to beta-fructosidase
Glycosyl hydrolase family 38 (alpha-mannosidase)
Uridine diphosphate glucose epimerase
Similar to endo-beta-1,4-d-glucanase
Wax synthase-like protein
Similar to callose synthase
Frequency of cell division was decreased in tps1 embryos
To establish if the delayed morphogenesis and growth in tps1 embryos is attributable to decreased cell elongation, we measured cell length, but found no difference between tps1 and WT embryos (data not shown). We also monitored the rate of cell division by crossing hemizygous TPS1/tps1 plants into the cell cycle reporter line expressing the cyclin B1 promoter–GFP fusion, and following expression of the reporter in WT and tps1/tps1 embryos. WT embryos showed a consistently higher frequency of cell division at both heart and torpedo stages compared with tps1, with most of the divisions in the wild type occurring in the cotyledon primordia (Figure 8a,b). The number of cell divisions decreased during maturation and there were no cell divisions at the beginning of desiccation (data not shown). The cell division frequency of both heart- and cotyledon-stage tps1 embryos was approximately half that found in the equivalent WT stage (Figure 8e).
TPS1 regulates embryo growth but not cell differentiation
For a normal developmental programme to be executed, processes such as cell patterning and differentiation must be tightly coordinated with changes in metabolism in order to fulfil the nutritional requirements of each developmental stage. The changes in the transcriptomic profile between WT torpedo and bent cotyledon embryo stages clearly reflect a shift of metabolism towards the synthesis of reserve compounds. The transition from invertase- to sucrose synthase-mediated sucrose degradation, the induction of storage protein transcripts, and the induction of transcripts associated with the synthesis of fatty acids and TAGs (triacylglycerol) reflect a coordinated reprogramming of gene expression in the transition from cell patterning to maturation. Our results show these changes on a genome-wide scale and agree well with previous results obtained from whole developing seeds (Ruuska et al., 2002), validating the sampling and RNA amplification procedures used in the present work.
The data presented here demonstrate that the tps1 mutant embryos are able to use the sucrose provided by the mother plant for the synthesis of storage reserves in a temporal manner similar to that in WT embryos. Nevertheless, the profile of metabolites in tps1 during seed development shows that, although there is accumulation of both lipids and proteins during seed development, the concentration of these is significantly lower on a per volume basis than in maturing WT embryos (Figure 5b,c). Together with the increased amount of soluble sugars and starch observed in tps1 (Figure 5d–f), this observation suggests that there is a perturbation of metabolism that reduces rather than blocks the flux of carbon into storage reserves.
Although the tps1 embryos are morphologically in the late torpedo stage, cell differentiation does not seem to be directly affected by the mutation. Chloroplast development occurs independently of embryo morphology, with heart-stage tps1 embryos exhibiting chloroplasts that are similar in structure to WT sibling embryos in the more advanced early cotyledon stage. As tps1 chloroplast development proceeds, starch granules accumulate and persist, whereas in WT the accumulation of starch is transient at the onset of the maturation phase. This persistence of elevated starch levels in tps1 embryos is somewhat at odds with a recent report of the post-transcriptional activation of ADP-glucose pyrophosphorylase and induction of starch synthesis by T6-P in isolated plastids (Kolbe et al., 2005). It is possible that in tps1 starch accumulation is primarily a result of downregulation of starch degradation, as indicated by altered expression of related genes, but the accumulation will also demand an active ADP-glucose pyrophosphorylase under what must be depleted T6-P conditions.
This study of tps1 embryos shows very clearly that the major biochemical processes associated with the maturation phase of embryo development can proceed independently of the morphological stage. Previous reports showed that storage reserve gene induction occurs in a number of embryo-lethal mutants that are arrested as early as the globular stage of embryo development (Devic et al., 1996; Yadegari et al., 1994). The comprehensive description of the tps1 embryo phenotype provides a demonstration that not only storage reserve gene expression, but also various related processes such as chloroplast development and entire transcriptome profile advance independently of morphological stage.
TPS1 is essential for seedling establishment
Torpedo-stage tps1 embryos survive seed desiccation and germinate, but fail to develop further. Various transcripts including those of several late embryogenesis-associated (LEA) genes increase in tps1 embryos (Figure S1) and some of these are likely to be associated with conferring desiccation tolerance as in WT embryos. The failure of germinated seeds to establish demonstrates that the TPS1 gene is absolutely essential for vegetative growth of the seedling. Using a dexamethasone-inducible transcription system to drive the expression of a TPS1 transgene in the tps1 mutant background, it was possible to recover mature plants, but these remained in the vegetative phase and did not produce inflorescence unless the TPS1 transgene was again induced (van Dijken et al., 2004). The phenotype we report here is much more severe and demonstrates that TPS1 is absolutely essential for vegetative growth. It is possible that the inducible system used in the previous study (van Dijken et al., 2004) gave a very low level of TPS1 expression that was sufficient to produce a less severe phenotype.
Perturbation of metabolism in tps1
Both starch and soluble sugar levels were significantly elevated per unit volume in tps1 compared with the wild type, whereas storage lipid and protein levels were decreased (Figure 5). This suggests that, although the gene expression programme proceeds into maturation, there is a bottleneck in the flux of carbon into storage reserves that may lead to accumulation of sugar nucleotides and consequently starch and sucrose. There were no major alterations in transcripts associated with glycolysis in the transcriptome analysis of tps1, but we did find significant changes in a number of transcripts associated with sucrose and starch metabolism and in particular a decrease in transcripts associated with starch breakdown (Table 4; Figure S1). Interestingly, we also found a significant decrease in AKIN11 transcript levels in tps1, which is consistent with previous reports showing a correlation between expression of this gene and T6-P levels in Arabidopsis plants overexpressing TPS and TPP (Schluepmann et al., 2004). AKIN11 encodes a plant homologue of the yeast SNF kinase complex that regulates a number of enzymes, including sucrose phosphate synthase and ADP-glucose pyrophosphorylase, which play key roles in sucrose and starch synthesis, respectively. Both sucrose and starch are elevated significantly in tps1 embryos, which is counterintuitive to the idea that AKIN11 regulates activity of the enzymes involved in their synthesis, given that AKIN11 transcript levels are decreased in tps1. The elevated levels of starch in tps1 are also contradictory to the recent finding that T6-P causes the oxidative activation of ADP-glucose pyrophosphorylase via SnRK1 (Kolbe et al., 2005). One explanation for this apparent paradox is that sucrose and starch accumulate as a consequence of the severe decrease in embryo growth and the associated decrease in the flux of carbon into storage protein and lipids.
Cell wall thickening in tps1
The progressive delay in morphogenesis in tps1 suggests that the primary defect altering embryo morphology could be associated with processes such as cell division and/or expansion rather than central carbon metabolism. tps1 embryos have much thicker cell walls compared with WT at both the torpedo and the cotyledon stages. Interestingly, transcript levels of several genes involved in the interconversion of sugar nucleotides required for the synthesis of the cell wall are significantly altered in the tps1 torpedo-stage embryos relative to the two WT stages (Table 6). Among these are genes encoding sugar nucleotide interconversion enzymes that have been proposed as an important link between the dynamics of the cell wall and primary metabolism (Seifert, 2004).
There are some tantalizing connections between the changes observed at the transcriptional level in tps1 and the altered cell wall structure and composition in tps1 embryos shown in Figures 3 and 7. In addition to the transcripts corresponding to sugar nucleotide interconversion enzymes there are also striking decreases in transcript abundance for several pectin-modifying enzymes (Table 6). It is possible that the observed increase in cell wall pectins could be a result of the downregulation of several pectin esterase genes in tps1 (Table 6). Work in tomato (Lycopersicon esculentum) has shown that antisense inhibition of pectin esterase gene expression results in an accumulation of esterified pectins in cell walls, possibly as a result of limiting polygalacturonase activity on the pectins (Hall et al., 1993). Increased pectin levels in plant cell walls can also arise as a consequence of decreased cellulose biosynthesis, as has been observed in mutants with impaired cellulose biosynthesis (His et al., 2001) and in cells treated with chemical inhibitors of cellulose biosynthesis such as isoxaben (Manfield et al., 2004). In this context it is interesting to note that there is also a marked decrease in transcript abundance of a cellulose synthase (At2g25540, AtCesA-10) in tps1. Very little is known about the regulation of cell wall biosynthesis in plants (Scheible and Pauly, 2004). The altered cell wall structure and related gene transcripts in tps1 demonstrate that trehalose metabolism plays a role in regulating this process. The challenge now is to establish whether this is a direct regulatory role in cell wall metabolism, or a more indirect role arising as a consequence of altered nutrient status or sugar nucleotide metabolism in the mutant plants.
Cell division compromises embryo growth in tps1 mutants
Although essential for organogenesis, there is controversy about the role that cell division plays during plant development. Modulation of the expression of certain cell cycle genes has been shown to modify the rate of cell division in Arabidopsis. Using these approaches it has been shown that neither inhibition nor stimulation of cell division severely affects cell differentiation, suggesting that cell division proceeds almost independently of this process (Inze, 2005). Our data show that cell division is markedly reduced in tps1, and this may be sufficient to account for the smaller size of the mutant embryos. This reduction in cell division could be directly attributable to perturbation of cell wall biosynthesis – thickening of cell walls could delay progression through the various phases of cell division. Alternatively, trehalose metabolism could play a more direct role in regulating cell division as an integrator of nutritional status and growth. In cell cultures, sugar availability has been proposed as a regulator of the G1 phase of the cell cycle via CycD3 cyclins (Riou-Khamlichi et al., 2000). The control of embryo growth during seed development is tightly regulated at the level of cell division, as demonstrated by the fus3/lec mutants (Raz et al., 2001).
We conclude that, in addition to playing a key role in central metabolism, the Arabidopsis TPS1 gene product modulates growth by determining cell wall deposition and cell division. The developmental arrest imposed by the tps1 mutation does not delay the progression of the transcriptome programme or cellular differentiation into the maturation state, but it does impact on primary carbon metabolism resulting in a dramatic accumulation of sucrose and starch.
Arabidopsis ecotype Columbia (Col-0) plants and the tps1-1 mutant (Eastmond et al., 2002) were stratified on soil at 4°C for 4 days and then grown at 20°C in a 16 h light/8 h dark photoperiod at 200 μmol m−2 sec−1 irradiance. After flowering, individual seeds were tagged in order to establish the developmental stage of the siliques at the moment of harvesting. Germination of tps1 wrinkled seeds on plates was carried out after the seeds were surface-sterilized and placed on half-strength Murashige–Skoog medium with 0.8% agar and vernalized at 4°C for 4 days.
WT and tps1 embryos were fixed in 2.5% glutaraldehyde/4% formaldehyde in 100 mM phosphate buffer, postfixed in 1% osmium tetraoxide and dehydrated through an acetone series, and this was followed by infiltration and embedding in Spurrs resin. Sections (80 nm) were stained with saturated uranyl acetate in 50% ethanol and Reynolds lead citrate. The sections were visualized using a FEI Tecnai G2 electron microscope (FEI, Hillsboro, OR, USA). For confocal microscopy, embryos were stained for 10 min in propidium ionide, and observed using a Zeiss LSM 510 confocal microscope (Carl Zeiss, Oberkochen, Germany).
Inmunofluorescence studies of cell wall components were performed in embryos fixed overnight in 4% paraformaldehyde and embedded in LR White resin (London Resin, Reading, UK). Sections (0.5 μm) were blocked in 1% Marvel milk powder (Premier International Foods, Spalding, UK) in phosphate-buffered saline (PBS) for 1 h and incubated with LM2, LM5, LM6, LM7, JIM5, and JIM7, at 1/20 dilution (PlantProbes, Leeds, UK). Control sections were incubated with blocking solution only. After washing with 1% Marvel milk powder in PBS, the sections were incubated with FITC-conjugated anti-rat secondary antibody (Sigma-Aldrich, Poole, UK) for 2 h. After washing three times, sections were mounted in CitiFluor (Agar Scientific, Stansted, UK) and observed by fluorescence microscopy. Cell division was monitored in segregating siliques of TPS1/tps1 plants crossed into transgenic plants expressing GFP under the control of the CycB1;1 promoter (Peter Doerner, University of Edinburgh, UK, unpublished results). Embryos at 4 and 7 DAF for WT, and 8 and 15 DAF for tps1/tps1 were observed with a Zeiss LSM 510 confocal microscope to detect cell division. Z-stack images were taken from at least 15 embryos from each stage in order to determine the number of cell divisions.
RNA preparation, amplification, array hybridization, data analysis and MapMan display
Fifty WT embryos at the torpedo stage (6 to 7 DAF), 50 WT embryos at the bent cotyledon stage (15 DAF), and 50 tps1 embryos (15 DAF) were dissected from the seed and used for RNA preparation. Once dissected the embryos were immediately placed in RNAlater solution (Qiagen, Valencia, CA, USA). RNA from embryos was extracted using the Absolutely RNATM Nanoprep kit (Stratagene, Amsterdam, the Netherlands) as follows. Embryos were centrifuged at 2500 g on a bench top microcentrifuge and the RNAlater solution was replaced by lysis buffer from the Absolutely RNATM Nanoprep kit. Embryos were homogenized using a plastic pestle and subsequently vortexed in the presence of glass beads for 30 sec. The resulting homogenate was filtered through a QiaShredder column (Qiagen). After the addition of 1 volume of 70% ethanol the RNA extraction was performed following the manufacturer's protocol.
RNA was vacuum-concentrated to a volume of 11 μl. cDNA synthesis and RNA amplification were performed using the MessageAmpTM aRNA kit (Ambion, Austin, TX, USA) according to the manufacturer's instructions. Briefly, first-round cDNA synthesis was primed with an T7 oligo(dT) primer. A volume of 1 μl of T7 oligo(dT) primer was added to 11 μl of RNA, heated to 70°C for 10 min, cooled down to 42°C and added to a mix containing 2 μl of 10X first-strand buffer, 1 μl of ribonuclease inhibitor, 4 μl of dNTP mix and 1 μl of reverse transcriptase. After a 2-h incubation at 42°C, second-strand synthesis was performed by the addition of 63 μl of nuclease-free water, 10 μl of 10X second-strand buffer, 4 μl of dNTP mix, 2 μl of DNA polymerase and 1 μl of RNase H. Reactions were incubated for 2 h at 16°C. cDNA was purified according to the manufacturer's instructions and vacuum-concentrated to 16 μl. For in vitro transcription, 8 μl of the cDNA was mixed with 8 μl of NTP mix, 2 μl of 10X reaction buffer and 2 μl of T7 enzyme mix. Reactions were incubated at 37°C for 16–24 h and then subjected to DNase I digestion. Amplified RNA was purified and vacuum-concentrated to 10 μl. One round of amplification was enough to produce amplified RNA in the μg scale.
Prior to microarray analysis, 0.5 μg was used for reverse transcription using Superscript II RT (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions, and the resulting cDNA was amplified using actin and TPS1 specific primers (Eastmond et al., 2002) in order to confirm the presence of transcripts for TPS1 in WT samples and their absence in tps1 embryos (data not shown).
Biotin labelling of aRNA was carried out according to the protocol Gene Chip Eukaryotic Small Sample Target Labelling Assay Version II, supplied by Affymetrix Ltd (High Wycombe, UK), using random primers (Invitrogen) for the first-strand synthesis reaction and proceeding with the biotin labelling after the synthesis of the second-strand cDNA. Three biological replicates of 50 embryos per sample were hybridized, washed, stained and scanned in Affimetrix ATH1 chips, following the procedures described in the Affimetrix technical manual. The expression levels of genes were measured by detection calls and signal intensities using the Micro Array Suite 5.0 software (Affymetrix Ltd) with a target signal (TGT) of 100. Sixty-four Affymetrix controls and 6860 Arabidopsis genes that were detected as ‘Absent’ in all nine chips were removed from the total of 22 810 probe sets. Pair-wise differentially expressed genes were identified using SAM software (Tusher et al., 2001), using the data for the remaining 15 886 Arabidopsis genes. The parameters for SAM were adjusted so that the false discovery rate for every pair of comparison was approximately 1%. In some cases, the SAM q-values above 1% were retrieved for genes of interest and considered as significantly differentially expressed between the cotyledon and the tps1 pair.
Samples for metabolite analysis were taken from WT plants at 8, 9, 10, 12, 15, and 18 DAF. tps1 material was harvested at 12, 15, and 18 DAF. All samples were composed of 50 embryos excised from the seed. The volume of the embryos was estimated by measuring the length and average width of 50 embryos for each developmental stage, assuming that embryos have a cylindrical shape.
Starch, sugars and proteins were measured on batches of 50 embryos following the procedure described by Baud et al. (2002). Fatty acids were quantified by gas chromatography, as previously described (Larson and Graham, 2001). Batches of 50 embryos excised from the seed were used for these analyses.
We thank Meg Stark for technical assistance with transmission electron microscopy, Stuart Casson and Keith Lindsey for assistance with RNA amplification from Arabidopsis embryos, Oliver Thimm for help with mapman, Peter Eastmond for critical reading of the manuscript, Simon McQueen-Mason for his help with the analysis of cell wall, and Peter Doerner for the CycB1;1::GFP line. This work was funded by the Biotechnology and Biological Sciences Research Council through a grant to IAG (87/917231).