Transcriptome profiling uncovers metabolic and regulatory processes occurring during the transition from desiccation-sensitive to desiccation-tolerant stages in Medicago truncatula seeds


  • Julia Buitink,

    Corresponding author
    1. Unité Mixte de Recherche 1191 Physiologie Moléculaire des Semences, Université d'Angers/INH/INRA, 16 Bd Lavoisier, 49045 Angers, France,
    Search for more papers by this author
  • Jean J. Leger,

    1. Ouest Genopole, Institut du Thorax, Institut National de la Santé et de la Recherche Médicale (UMR 533), Faculté de Médecine, 44035 Nantes, France,
    Search for more papers by this author
  • Isabelle Guisle,

    1. Ouest Genopole, Institut du Thorax, Institut National de la Santé et de la Recherche Médicale (UMR 533), Faculté de Médecine, 44035 Nantes, France,
    Search for more papers by this author
  • Benoit Ly Vu,

    1. Unité Mixte de Recherche 1191 Physiologie Moléculaire des Semences, Université d'Angers/INH/INRA, 16 Bd Lavoisier, 49045 Angers, France,
    Search for more papers by this author
  • Sylvie Wuillème,

    1. Unité de Biologie des Semences, INRA, RD 10, 78026 Versailles cedex, France, and
    Search for more papers by this author
  • Guillaume Lamirault,

    1. Ouest Genopole, Institut du Thorax, Institut National de la Santé et de la Recherche Médicale (UMR 533), Faculté de Médecine, 44035 Nantes, France,
    Search for more papers by this author
  • Alice Le Bars,

    1. Ouest Genopole, Institut du Thorax, Institut National de la Santé et de la Recherche Médicale (UMR 533), Faculté de Médecine, 44035 Nantes, France,
    Search for more papers by this author
  • Nolwenn Le Meur,

    1. Ouest Genopole, Institut du Thorax, Institut National de la Santé et de la Recherche Médicale (UMR 533), Faculté de Médecine, 44035 Nantes, France,
    Search for more papers by this author
  • Anke Becker,

    1. International NRW Graduate School in Bioinformatics and Genome Research, Institute of Genome Research, Center for Biotechnology, Bielefeld University, D-33594 Bielefeld, Germany
    Search for more papers by this author
  • Helge Küster,

    1. International NRW Graduate School in Bioinformatics and Genome Research, Institute of Genome Research, Center for Biotechnology, Bielefeld University, D-33594 Bielefeld, Germany
    Search for more papers by this author
  • Olivier Leprince

    1. Unité Mixte de Recherche 1191 Physiologie Moléculaire des Semences, Université d'Angers/INH/INRA, 16 Bd Lavoisier, 49045 Angers, France,
    Search for more papers by this author


To investigate regulatory processes and protective mechanisms leading to desiccation tolerance (DT) in seeds, 16086-element microarrays were used to monitor changes in the transcriptome of desiccation-sensitive 3-mm-long radicles of Medicago truncatula seeds at different time points during incubation in a polyethylene glycol (PEG) solution at −1.7 MPa, resulting in a gradual re-establishment of DT. Gene profiling was also performed on embryos before and after the acquisition of DT during maturation. More than 1300 genes were differentially expressed during the PEG incubation. A large number of genes involved in C metabolism are expressed during the re-establishment of DT. Quantification of C reserves confirms that lipids, starch and oligosaccharides were mobilised, coinciding with the production of sucrose during the early osmotic adjustment. Several clusters of gene profiles were identified with different time-scales. Genes expressed early during the PEG incubation belonged to classes involved in early stress and adaptation responses. Interestingly, several regulatory genes typically expressed during abiotic/drought stresses were also upregulated during maturation, arguing for the partial overlap of ABA-dependent and -independent regulatory pathways involved in both drought and DT. At later time points, in parallel to the re-establishment of DT, upregulated genes are comparable with those involved in late seed maturation. Concomitantly, a massive repression of genes belonging to numerous classes occurred, including cell cycle, biogenesis, primary and energy metabolism. The re-establishment of DT in the germinated radicles appears to concur with a partial return to the quiescent state prior to germination.


At the later stages of maturation, seeds of many species acquire the capacity to withstand the removal of the majority of their water, i.e. they become desiccation tolerant. This characteristic allows for prolonged survival in the dry state until conditions are favourable for germination. During seed imbibition, water is taken up and metabolic processes resume, leading to the emergence of the radicle through the seed coat, concomitant with the loss of desiccation tolerance (DT). In these radicles of germinated seeds, DT can be re-established by treatment of the germinated seeds in an osmoticum (Bruggink and van der Toorn, 1995; Buitink et al., 2003). This re-establishment is possible only during a small developmental window after germination: when radicles grow above a certain size, full DT can no longer be acquired. Possibly the acquisition of a greater tolerance to mild dehydration stress serves to enable seedlings to quickly adapt to adverse conditions during the early phases of establishment, and assures the survival of progeny.

The mechanisms of DT are thought to involve the synthesis of protective molecules, such as non-reducing sugars, late embryogenesis abundant (LEA) proteins, heat-shock proteins (HSP) and various stress proteins (for reviews see Bartels and Salamini, 2001; Hoekstra et al., 2001; Ingram and Bartels, 1996). In addition, it has been hypothesised that a controlled downregulation of metabolism might be essential to prevent the production of detrimental reactive oxygen species (Hoekstra et al., 2001; Leprince et al., 2000). Considering the multifactorial trait of DT, it is possible that several regulatory pathways act in parallel and interact with one another to induce DT. ABA and ABA-related signalling pathways play an important role in the acquisition of DT (Bartels and Salamini, 2001). Most of the research which has focused on elucidating signalling in DT has been performed on either non-seed organisms, such as the so-called resurrection plants (Bartels and Salamini, 2001; Collet et al., 2004; Neale et al., 2000; Ramanjulu and Bartels, 2002), or on rehydration of desiccation-tolerant bryophytes such as Tortula ruralis (Oliver et al., 2004). In seeds, much of the progress in the elucidation of signalling pathways and developmental programs controlling DT has come from the isolation and analysis of viviparous mutants and mutants impaired in embryogenesis and maturation, such as the fus3, lec1 and abi3 mutants of Arabidopsis, three master seed development regulator loci (Raz et al., 2001). In Arabidopsis and maize, some of the target genes of these activators are genes proposed to have a role in DT, e.g. small HSPs (Wehmeyer and Vierling, 2000) and group1 LEA proteins (Butler and Cuming, 1993). However, progress in unravelling the signal transduction pathways for the acquisition of DT has been impeded by the difficulty in discriminating them from those involved in other overlapping developmental programs, such as reserve accumulation and the acquisition of germinability. For instance, the desiccation-sensitive (DS) abi3/fus3 mutant of Arabidopsis exhibits an array of pleiotropic defects ranging from decreased accumulation of storage proteins to reduced dormancy and vivipary. In fus3/lec1 mutants, embryonic cell division, normally restricted to embryogenesis, continues during seed maturation (Raz et al., 2001). One could argue that the acquisition of DT is a transition that depends on developmental programs such as embryogenesis and reserve deposition. However, desiccation-tolerant embryos of Arabidopis have been isolated despite severe defects in embryogenesis (Wehmeyer and Vierling, 2000). Similarly, in wheat the competence to acquire DT is not dependent on the morphological development of the embryo (Golovina et al., 2001).

Transcriptome analysis can be a suitable tool to investigate the regulatory and protective mechanisms involved in the acquisition of DT, by shedding light on the shifts of metabolic and cellular processes that are necessary to be prepared for the dry state. The physiological model of re-establishment of DT in germinated radicles is particularly useful for this purpose, because it uncouples acquisition of DT from other developmental programs during maturation and germination. The legume Medicago truncatula, closely related to economically relevant legumes such as Medicago sativa, Phaseolus vulgaris and Pisum sativum, is a major target species for the understanding of storage reserve accumulation in proteineous seeds (Bell et al., 2001). Considering the overlapping time frame of the acquisition of DT and the accumulation of storage reserves in M. truncatula, it is possible that several regulatory mechanisms between both phenomena are either similar or might interact. Indeed, in the abi3 mutants, both storage accumulation and Em gene regulation are affected. Seeds of M. truncatula are relatively large and can be easily selected after germination, according to their radicle length, to obtain a synchronised and reproducible curve for the re-establishment of DT (Buitink et al., 2003).

Gene expression profiling has been studied during seed development (Girke et al., 2000) in order to investigate either early reproductive stages (Dong et al., 2004; Hennig et al., 2004) or seed filling (Ruuska et al., 2002), as well as seed germination (Nakabayashi et al., 2005; Soeda et al., 2005), but has not been studied in relation to the acquisition of DT. Transcript profiling has been carried out on drought stress in vegetative tissues of Arabidopsis thaliana (Bray, 2002; Seki et al., 2002). Even if DT can be considered as one type of drought tolerance (Alpert and Oliver, 2002), it is unknown whether the regulatory pathways are similar between desiccation and drought tolerance. Because tissues that have the ability to survive complete drying also have to pass through the hydration level comparable to drought stress, it could be expected that at least part of the mechanisms involved in DT act at hydration levels corresponding to drought stress. Examples are the expression of LEA proteins both during maturation and upon drought, and the finding that overexpression of a Myb transcription factor (TF), cpm10, of the desiccation-tolerant resurrection plant Craterostigma plantagineum can induce drought tolerance in Arabidopsis (Villalobos et al., 2004). Thus, the understanding of the mechanisms necessary for the tolerance of desiccation and their regulatory networks might be beneficial in the search of candidate genes to improve drought tolerance (Ramanjulu and Bartels, 2002). In particular, the understanding of the re-establishment of stress tolerance in the germinated radicles could be transferred to older seedlings in order to improve seedling establishment.

In this study, 16086-element (16K) microarrays of M. truncatula were used to perform gene expression profiling during the re-establishment of DT in sensitive radicles of germinated seeds. The time-dependent changes in gene expression of different functional classes revealed the main shifts that these DS radicles undergo to become resistant to complete drying. In addition, gene expression was compared at two stages of seed development, before and after the acquisition of DT, to identify overlapping pathways involved in the acquisition and re-establishment of DT.


Acquisition, loss and re-establishment of DT in seeds of M. truncatula

To investigate gene expression related to DT, precise developmental stages were characterised (Figure 1). DT is acquired midway through the maturation (Figure 1a), between 16 and 20 days after pollination (DAP). The capacity to germinate (12–14 DAP) as well as seed filling processes (10–30 DAP) take place in a similar time frame. Mature, non-dormant seeds start to germinate at around 13 h when imbibed at 20°C (Figure 1b). The DT of the radicle is lost immediately after the protrusion of the radicle (Figure 1b). It is possible to re-establish DT in these radicles just after germination by submitting the germinated seeds to an osmotic shock for several days (Buitink et al., 2003). When seeds with DS radicles of 2.8 mm are incubated in an osmotic polyethylene glycol (PEG) solution of −1.7 MPa at 10°C, DT is gradually re-established (Figure 1c). After 6 h, 10% of the seed population have desiccation-tolerant radicles, whereas after 16 h, the percentage has increased to 62% and reaches a final value of 90% after 72 h of incubation. This physiological system was used to investigate gene expression at different time points during the re-establishment of DT and upon the acquisition of DT during maturation.

Figure 1.

 Characterisation of acquisition, loss and re-establishment of desiccation tolerance (DT) in seeds of Medicago truncatula.
(a) Acquisition of germination and DT in seeds of M. truncatula during development. Percentage of germination (closed circles) was determined at the indicated days after pollination (DAP) on 100 seeds by incubation in water at 20°C and scored as the protrusion of the radicle. DT (open circles) was determined by dehydration of 100 seeds for 3 days in an airflow of 42% RH and counting the percentage of germinated seeds subsequent to rehydration. DW accumulation (closed triangles) was determined gravimetrically on triplicate of five seeds (average ± SE). (b) Germination and loss of DT during imbibition at 20°C. Germination (closed symbols) and DT (open symbols) percentages were determined as described above on triplicates of populations of 50 seeds (average ± SE). (c) The re-establishment of DT in radicles of germinated seeds. Germinated seeds with a protruded radicle length of 2.8 mm were incubated at 10°C in a polyethylene solution (−1.7 MPa) in the dark prior to drying (closed symbols). After drying the seeds for 3 days at 42% RH, DT was scored as the resumption of radicle growth after rehydration. Each data point represents the average (±SE) of between three and five independent experiments of 30–50 seeds. Water content was determined gravimetrically on a triplicate of 25 radicles (closed squares; average ± SE).

Clustering and classification of the kinetics of gene expression during the re-establishment of DT

During the re-establishment of DT, gene expression in the protruded radicles at different time points during the PEG incubation was compared with that of the 3-mm-long DS radicles (time point zero of PEG incubation). Based on a three-level analysis of concordance of the replicates (between the replicates on the same slide, between the two slides that serve as technical replicates and between the two biological replicates of the PEG time course each containing 300 individuals), concordant genes were selected for each time point. Subsequently, the minimum and maximum values for both time courses were calculated by permutation (26 = 64 assemblies) and genes were considered variant when the 90th percentile of the range >0.7. In addition, only those genes for which at least four out of six time points gave concordant values were considered. Finally, the biological reproducibility of the two PEG time courses was assessed using the Pearson's correlation coefficient, and only genes with r > 0.7 were retained. In this manner, a total of 1310 genes were differentially and reproducibly expressed. A list of these genes is presented in Supplementary. To verify the accuracy of the microarray data, the expression of several genes belonging to different clusters were also analysed by Northern blot. Both microarray data and Northern blot analysis showed comparable trends, irrespective of whether genes were either lowly or highly expressed.

To obtain an overview of the time-dependent changes of gene expression during re-establishment, k-means clustering was employed (for a review, see Quackenbush, 2001) to attempt to identify possible biologically relevant clusters of genes in the expression dataset from the DS 2.8-mm protruded radicles during the PEG incubation. The 1310 genes that were retained could be divided into five clusters of upregulated genes (I–V up) and four clusters of downregulated genes (I–IV down; Figure 2). The clusters are ranked based on their time-dependent expression profiles, first an early, mainly transient expression for the first two clusters, thereafter a cluster with a sharp and early increase in expression (III), and finally the clusters the expression profiles of which continuously increase with the re-establishment of DT (IV and V). Two clusters containing genes with a transient upregulated expression profile were identified with a maximum around 3 h (cluster I up; 30 genes) and 6 h (cluster II up; 72 genes) of PEG incubation, respectively (Figure 2). Thereafter, a group of two clusters (III up) showed a sharp continuing increase in expression from 1 to 16 h of PEG incubation, one with an average increase of four-fold (IIIa up; 109 genes) and the second of 16–20-fold (IIIb up; 62 genes). For clarity, cluster IIIa is not shown in Figure 2, but the expression profiles are comparable with those of cluster IIIb, except that the magnitude of the expression is lower. Finally, two more clusters consist of genes the expression profiles of which either just precede or coincide with the increase in the percentage of DT (cluster IV, 186 genes; cluster V, 122 genes). The downregulated genes were grouped in four expression profiles (Figure 2). The first two clusters contain genes with a transient downregulation profile, peaking at 1–3 h (I down; 30 genes) and 9 h (cluster II down; 115 genes). The third cluster contained genes the expression of which gradually decreased after 1–16 h (cluster III down; 248 genes). Finally, a fourth cluster showed an expression profile that gradually decreased during the re-establishment of DT (IV down; 329 genes).

Figure 2.

 Clustering of the expression patterns of M. truncatula genes during the re-establishment of DT in radicles during polyethylene glycol (PEG) incubation.
Time course of changes in gene expression in radicles during PEG treatment of seeds leading to the re-establishment of DT. Clusters were constructed using k-means clustering (Genesis). Genes were divided in five clusters of upregulated genes (a–e) and four clusters of downregulated genes (f–i). The average expression ratio of each cluster is indicated by a thick solid line. Curves depicting the increase in the percentage DT and decrease in the water content are indicated in j (see Figure 1). Data are expressed as the log2 values of the ratio treated/untreated radicles.

Based on comparison with current releases of the PIR and TrEMBL databases, as well as Interpro searches, genes in the different clusters were annotated. Subsequently, the genes were grouped in functional categories according to the functional catalogue of the Munich information center for protein sequences (MIPS) classification. Table 1 gives a summary of the classification, arranged for the clusters with upregulated (from I up to V up) and downregulated genes (from I down to IV down). The clusters showing similar expression profiles but differing in the amplitude of the expression ratio (i.e. IIIa and IIIb up) were joined together for the classification analysis.

Table 1.  Classification of genes either up- or downregulated in the different expression clusters during the re-establishment of desiccation-tolerant (DT)
  1. Clusters were designed by k-means clustering and are presented in Figure 2.

Metabolism 7 29 22 26 16 3 6 40 64
 Amino acid metabolism003310165
 Nitrogen and sulfur metabolism000000001
 Nucleotide metabolism003411033
 Lipid, fatty acid (FA), isoprenoid met24182121014
 Metabolism vitamins, cofactors003221010
 Secondary metabolism41953502418
Energy 0 2 4 10 5 0 0 12 17
 Glycolysis and gluconeogenesis011300011
 Glyoxylate cycle010100000
 Pentose-phosphate pathway000000000
 Pyruvate dehydrogenase complex000100003
 Tricarboxylic-acid pathway000100015
 Electron transport001220085
 Metabolism energy reserves000000011
Storage reserves 0 0 0 1 5 0 1 1 0
Cell cycle and DNA processing 0 0 2 6 3 1 2 13 17
Transcription 5 2 11 7 6 3 9 20 19
Protein synthesis 0 0 3 3 2 0 2 1 8
Protein fate 0 7 9 15 8 2 9 9 19
Transport 0 4 10 9 4 1 12 25 34
Cellular comm., signal transduction 4 4 10 13 10 6 8 16 22
Cell rescue, defence and virulence 1 4 36 13 20 3 5 11 10
 Stress response0128680110
 Disease, virulence, defence122171224
Cell fate 0 0 3 4 2 0 2 5 6
Biogenesis/growth 0 3 3 8 3 0 1 22 8
Unclassified 2 1 9 10 4 1 8 11 12
Unknown 11 16 48 63 40 10 50 62 93

Genes involved in metabolic processes that are related to carbon and lipid metabolism were the most abundant in all clusters (Table 1). Genes related to secondary metabolism are transiently upregulated early during the PEG incubation (see cluster II up). For example, many genes involved in secondary metabolite pathways synthesising the phytoalexin medicarpin are expressed. At the later stages, when DT is installed, genes in this pathway are downregulated (cluster IV down). Cluster III–IV up contain many genes that are involved in detoxification (Table 1, for instance encoding gluthatione peroxidase, catalase, cationic peroxidase, l-ascorbate peroxidase and alcohol dehydrogenase (ADH). In fact, other genes involved in anaerobic respiration (for instance pyruvate decarboxylase,) are also represented in these clusters. At the end of the re-establishment of DT, many genes belonging to various classes are downregulated (cluster III–IV down). The expression of genes related to primary metabolism, such as amino-acid metabolism, carbon metabolism, lipid metabolism and secondary metabolism, as well that of genes involved in transport and energy metabolism, such as TCA cycle and electron transport, decrease at the final stages of the re-establishment of DT, just like those involved in transcription and cell cycle and DNA processing, suggesting a general downregulation of metabolic processes and cellular activity in the desiccation-tolerant radicles.

A class that contains mainly upregulated genes is related to cell rescue, defence and virulence, with genes involved in stress response and detoxification. This class contains almost all the LEA protein genes (cluster III up), apart from two LEA proteins, Em6, a member of group 1 and D-34, a member of group 5 that can both be found in cluster V (Figure 3a). This cluster V also contains genes that are typically upregulated during later seed development, such as those involved in seed storage reserves. Genes involved in transcription (mainly TF), protein fate and cellular communication are upregulated as well as downregulated and present in all clusters. Some examples of different profiles of upregulated TFs during the PEG incubation are homologues to DREB2 (cluster I), AP2-like TF (EREBP) (cluster II), regulator protein ROM2, AREB-like protein, Aintegumenta and DREB (cluster III), and a homologue of MYBST1 (cluster IV) (Figure 3).

Figure 3.

 Expression patterns of genes encoding late embryogenesis abundant (LEA) proteins (a) and transcription factors (TF) (b) during the re-establishment of DT in radicles of Medicago trunculata.
DT was re-established by an incubation of seeds with 3-mm-long protruded, sensitive radicles (3 mm) in a PEG solution (3 mmP). Data are expressed as the log2 values of the ratio of treated/untreated radicles. Data points represent the average (±SE) of the technical and biological replicates (between four and eight values). (a) A selection of genes encoding LEA proteins that are present in either cluster IIIb up (top) or cluster V up (bottom). DHN, dehydrin; Em6, an LEA group 1 and D34, a group 5 LEA protein. (b) Genes encoding TFs exhibiting various expression profiles: DREB2 TF, cluster II up; AREB, aintegumenta and DREB TF, cluster III up; ROM and MYBST TF, cluster IV up. Expression data can be found in.

Carbon metabolism and accumulation of sucrose

Major protective compounds produced during the acquisition of DT are non-reducing soluble sugars, mainly sucrose and oligosaccharides (Hoekstra et al., 1994). A large number of genes involved in various catabolic and anabolic pathways related to carbon metabolism were present in different clusters (Table 1), suggesting a sequential regulation of the carbon resources in relation to the acquisition of DT. This observation prompted us to analyse the change in sugars during the re-establishment of DT in the sensitive radicles. Prior to PEG-treatment, sucrose levels in the radicles are low, around 15 mg g−1 dried weight (DW) (Figure 4, open symbols). During the PEG incubation, sucrose accumulates in two phases: first, within the first six hours, there is a very rapid increase to 65 mg g−1; second, there is a further slower accumulation to around 85 mg g−1 (Figure 4, open symbols). An assessment of the profiles of differentially expressed genes encoding enzymes involved in starch, sucrose and lipid metabolism suggested different origins for the sucrose accumulation at different time points during the PEG incubation. To verify this hypothesis, changes in starch, soluble sugars and lipids were measured to see if the break-down of these compounds could account for the sucrose accumulation (Table 2). One of the first upregulated genes involved in carbon metabolism is encoding β-amylase, the expression of which is already increased by three-fold after 1 h of PEG incubation (Figure 4a). Its expression coincided with the degradation of starch (Table 2). At later stages, between 6–16 h, there was activation of the gene expression of an enzyme involved in a second pathway of starch degradation, starch phosphorylase (Figure 4a), in parallel to a second peak of starch degradation (Table 2). At the end of the PEG incubation there was a net production of starch, concomitant with the expression of a gene encoding starch synthase cluster V up. The expression of genes related to sucrose synthesis, encoding sucrose phosphate synthases (SPS), were increased at the later stages of the re-establishment (cluster IV up) after 6–9 h of PEG incubation (closed symbols, Figure 4b). At the same time, the expression of two sucrose synthase (SuSy) genes increased, the most highly expressed isoform being SuSy2. It is a general observation that the expression of SPS and SuSy genes is upregulated during dehydration/osmotic stress conditions (Dejardin et al., 1999; Pelah et al., 1997), and antisense expression of SPS in potato plants led to decreased SPS activity and completely suppressed the water-induced stimulation of sucrose synthesis (Geigenberger et al., 1999). After 6–16 h there is a substantial decrease in the hexose pool that could contribute to the sucrose accumulation. In addition, oligosaccharides (mainly stachyose with smaller quantities of raffinose, verbascose and an unidentified oligosaccharide) that were still present in 2.8-mm long radicles were gradually degraded during the first 16 h of PEG incubation (Table 2).

Figure 4.

 Sucrose accumulation and expression patterns of genes encoding enzymes related to carbon metabolism during the re-establishment of DT in radicles of M. truncatula.
DT was re-established by an incubation of seeds with 3-mm-long protruded radicles (3 mm) in a PEG solution (3 mmP). Sucrose accumulation (open diamonds) was determined in triplicate (average ± SE). Expression data are expressed as the log2 values of the ratio of treated/untreated radicles. Data points represent the average (±SE) of the technical and biological replicates (between four and eight values). (a) Expression profiles of genes involved in starch degradation; β-amylase (two isoforms, solid circles), starch phosphorylase (solid triangles). (b) Expression profiles of genes involved in sucrose metabolism: SPS, sucrose phosphate synthase (two isoforms, closed circles); SuSy1, sucrose synthase 1 (upward triangles); SuSy2, sucrose synthase 2 (downward triangles). (c) Expression profiles of genes involved in lipid mobilisation; acyl-coA synthase (solid triangles), acyl-coA dehydrogenase (two isoforms, solid squares). Expression data can be found in.

Table 2.  Changes in composition of starch, soluble sugars, starch and lipids during thepolyethylene glycol (PEG) incubation in 2.8-mm-long protruded radicles
PEG incubation0 h1 h3 h6 h9 h16 h72 h
  1. Net either production or consumption of lipids, starch and soluble sugars (μg mg−1) in 3 mm-long protruded radicles of Medicago truncatula at different time intervals during PEG incubation. Average of three replicates ±SE.

Sucrose18.5 (3.8)30.8 (8.0)47.3 (6.8)63.1 (5.0)70.7 (2.6)74.2 (3.0)84.9 (4.8)
Hexoses26.9 (4.3)30.5 (3.1)22.5 (3.2)35.4 (8.7)17.6 (1.1)16.1 (3.4)13.1 (2.1)
Oligosaccharides46.4 (1.0)42.9 (2.1)40.6 (1.5)34.3 (1.9)30.7 (0.3)29.9 (0.5)30.1 (1.4)
Starch22.8 (2.2)15.5 (1.3)12.6 (3.1)14.5 (4.1)12.1 (0.6)5.5 (0.2)17.1 (2.1)
Lipids155.7 (0.9)146.5 (2.7)143.8 (2.6)138.8 (4.4)135.7 (2.1)133.6 (2.7)111.0 (4.6)

Breakdown of triacylglycerol via the glyoxylate cycle could be another source of sucrose production though gluconeogenesis. The expression of lipase genes did not give any indication of activation of reserve breakdown, their expression either increased or decreased. However, the expression of genes encoding enzymes involved in β-oxidation increased sharply at the later stages of PEG incubation (Figure 4c). The decrease in the total lipid content within the first 3 h seemed to be post-transcriptionally regulated, whereas at the later stages, lipid breakdown coincided with acyl-coA synthase and dehydrogenase expression (Table 2).

Overall, it appears that mainly starch and lipid reserves, and to a lesser extent oligosaccharides, are mobilised at different time points during the re-establishment of DT in order to provide sucrose for osmotic adjustment and protection, and that this degradation can be partially deduced from the expression profiles of genes involved in carbon metabolism.

Comparison of DT-associated gene expression that is either developmentally regulated or induced by incipient water loss upon PEG incubation

To investigate whether the genes that are either up- or downregulated in desiccation-tolerant radicles of germinated seeds after the re-establishment can also be identified in embryos in which DT was acquired during seed maturation, an additional screen was performed on seeds harvested at two time points during maturation: at 14 DAP, representing seeds that are desiccation sensitive (Figure 1a) and at 20 DAP, representing seeds that have acquired DT. After selecting those genes with concordant values between the different levels of replication, genes were retained when M > 0.8, M indicating the mean of the log2 expression ratio. In total, upon the acquisition of DT during seed development, 538 and 395 genes were detected that were either up- or downregulated, respectively. A Venn diagram was constructed between both DT datasets (acquisition of DT during maturation, 20 DAP versus 14 DAP and re-establishment of DT after germination, 3 mm 72 h PEG versus 3 mm). We found 187 upregulated and 108 downregulated genes in common between the two datasets (Figure 5). A list of these genes is provided in. A classification of the genes is shown in Table 3. Main classes of genes that are specifically upregulated in relation to DT are those related to storage reserves, such as lipid bodies, carbon metabolism, transcription and cellular communication, and cellular defence. Classes that contain mainly downregulated genes are those related to lipid and energy metabolism, cell cycle and DNA processing (Table 3). Table 4 shows a list of genes involved in regulation (signalling and TFs) and protection found to be in common for both DT datasets, indicating the identification of the genes based on homology as well as their expression ratio. To obtain an idea of either the tissue or condition specificity, the number of expressed sequence tags (ESTs) found in each M. truncatula EST library are provided in (electronic northern).

Figure 5.

 Venn diagram comparing upregulated (a) and downregulated (b) genes expressed upon the acquisition of DT and after the re-establishment of DT in germinated seeds.
Expression ratios were obtained from comparison of desiccation-sensitive (DS) 14-day-old embryos with desiccation-tolerant, 20-day-old embryos harvested during seed development (mat; 20 DAP/14 DAP) and from the comparison of DS, 3-mm-long radicles of germinated seeds with desiccation-tolerant, germinated radicles in which DT was re-established after 72 h of PEG treatment (−1.7 MPa, 10°C) (3 mmP; 3 mm 72 hPEG/3 mm). Genes were selected that were concordant at the three levels of experimental design and for which the log2 mean expression ratio >0.8. Only genes that showed differential expression during the PEG kinetics were considered. The statistical association between the differentially expressed genes (upregulated and downregulated) in both datasets was verified using a χ2 test (χ2 = 575; d.f. = 4, n = 5724; P < 0.001).

Table 3.  Classification of genes either up- or downregulated both upon the acquisition of DT during maturation (20 DAP/14 DAP) and after the re-establishment of DT in 3-mm-long radicles of germinated seeds by 72 h of PEG treatment (3 mm 72 h PEG/3 mm)
  1. Genes were selected that were concordant at all levels of the experimental design, differentially expressed during the PEG kinetics and for which the log2 mean expression ratio >0.8.

Metabolism 31 28
 Amino acid metabolism57
 Nitrogen and sulfur metabolism22
 Nucleotide metabolism01
 Lipid, FA, isoprenoid. met111
 Metabolism vitamins, cofactors30
 Secondary metabolism31
Energy 8 17
Storage reserves 2 0
Cell cycle and DNA processing 1 10
Transcription 12 2
Protein synthesis 0 3
Protein fate 10 8
Transport 6 8
Cellular comm, signal transduction 13 3
Cell rescue, defence and virulence 31 2
 Stress response220
 Disease, virulence, defence10
Cell fate 3 5
Biogenesis 5 1
Unclassified 8 2
Unknown 56 16
Table 4.  Expression ratio of Medicago truncatula genes related to either regulation or protection that are upregulated both after the acquisition of DT during maturation (20 DAP/14 DAP) and after the re-establishment of DT in 3-mm-long radicles of germinated seeds by PEG treatment (3 mm 72 h PEG/3 mm)
TIGR IDIdentification (homology)PEGMAT
Transcription factors (TF)
 TC79073Q7Y0Y9 dehydration responsive element binding protein2.84.3
 TC88745Q8LGQ4 aintegumenta-like protein1.94.2
 TC87039Q8W0K0 P0007F06.2 protein NAM2.62.9
 TC81623T01257 probable GT-1-like TF2.82.9
 TC77054Q84UB0 TF Myb12.12.8
 TC77331AAS58510 MYB TF1.92.4
 TC89342Q9FLG5 similarity to NAM2.22.2
 TC92282O13337 putative transcriptional regulator5.42.0
 TC76842AAR37423 putative ethylene response factor 52.01.9
 TC93578BAD19065 auxin response factor 53.31.9
 TC77157E86148 T1N6.12 protein NAM4.81.8
 TC80971Q9NPA8 E(Y)2 homolog (DC6)1.91.7
Cellular communication
 TC93145Q7XYX9 SNF4b20.632.7
 TC90076T48379 AtRab GTP-binding protein1.99.5
 TC77426AAS87371 zinc finger protein2.66.9
 TC88465Q7XAB6 cyclin D3–12.94.1
 TC89205O81223 calcineurin B-like protein 43.93.7
 TC78763Q8GRK2 somatic embryogenesis receptor kinase1.93.2
 TC85793T31428 fiber annexin3.73.1
 TC83546P54637 protein-tyrosine phosphatase 32.42.6
 TC77188Q84XC0 calcineurin B-like-interacting protein kinase1.92.4
 TC87699BAD00043 MAP kinase phosphatase2.72.3
 TC87242S56716 protein kinase SPK-33.62.0
 TC79837Q9AWQ0 receptor protein kinase PERK1-like2.21.8
 TC91913Q93WK5 two-component response regulator-like APRR7 (pseudo-response regulator 7)3.52.0
Heat shock proteins
 TC78900P51819 heat shock protein 833.22.2
 TC85567S16247 heat shock protein
LEA proteins
 TC79093Q9ZTY1 35kDa seed maturation protein54.0104.0
 TC76866O49817 CapLEA-231.749.2
 TC85220S61428 PM1020.437.5
 TC87025S27757 embryonic abundant protein. 59K47.836.9
 TC79790Q40848 late embryogenesis abundant protein15.226.6
 TC87163T07661 maturation protein PM311.321.6
 TC88632P93510 Em protein16.920.5
 TC84892Q9XES7 Seed maturation protein PM278.018.0
 TC85295T06255 dormancy associated protein6.417.1
 TC78559Q9SPJ6 maturation protein pPM3213.316.9
 TC84691O49816 CapLEA-13.012.9
 TC76868CAA12026.1 LEA protein13.310.0
 TC80189S04046 embryonic abundant protein gD-341.87.8
 TC85950Q9FNW8 seed maturation protein LEA 45.17.1
 TC86014Q9SW70 stress-related protein4.66.5
 TC86281T07089 dehydrin3.15.3
 TC77847S12095 embryonic abundant protein2.04.7
 TC87755S52657 seed biotin-containing protein SPB658.72.9
Stress proteins
 TC77899Q9SWB3 seed maturation protein PM392.84.0
 TC81698Q9SWB3 seed maturation protein PM392.93.7
 TC92949Q9SF06 F26K24.22 protein USP domain8.92.8
 TC82648Q9SF06 F26K24.22 protein USP domain32.32.5
Pathogenesis-related proteins
 TC82272Q7XA41 disease resistance protein RGA26.02.2
 TC88591Q6E2Z6 1-Cys peroxiredoxin4.217.1
 TC87513Q84V96 aldehyde dehydrogenase 1 precursor8.83.2
 TC77042Q94BV3) probable phospholipid hydroperoxide glutathione peroxidase.mitochondrial precursor5.13.1
 TC80834Q9SML3 cytochrome P450 monooxygenase2.52.4
 TC82902Q9LHT0) putative short chain alcohol dehydrogenase3.22.0
 TC76569Q9FT05 cationic peroxidase4.11.9
 TC80356T07120 probable cytochrome P450 CP72.01.8

Several genes encoding TFs were expressed in relation to DT, such as three TFs belonging to the family of NAM proteins and three TFs of the MYB family (Table 3). Interestingly, a gene encoding a dehydration responsive element binding protein (DREB) TF was highly abundant in the seed library derived from seeds harvested at late maturation, thereby confirming its upregulation at the later stages of maturation, as found in this study. Only a small number of signalling genes that were abundantly expressed in seeds (although some are also expressed in symbiosis) were common between both datasets. One example is SFN4b, a putative activating subunit of the sucrose non-fermenting related kinase (SnRK1) complex, the expression of which was already correlated to DT (Buitink et al., 2004). Whether SNF4b is indeed involved in activating the SnRK1 complex in seeds is unknown, but the SnRK1 complex seems to play an important role in regulating carbon and energy metabolism (Thelander et al., 2004). A large number of genes related to stress and defence were found to be highly upregulated in both datasets (Table 4). In contrast to regulatory genes, those involved in protection were more seed specific, especially several of the LEA proteins, protecting macromolecules against the detrimental effects of dehydration (Hoekstra et al., 2001). In addition, two HSPs were identified, one of them being a small HSP (sHSP). Quantitation of sHSP protein in desiccation-intolerant seeds of abi3-6, fus3-3 and lec1-2 mutants showed that all had <2% of wild-type HSP17.4 levels, correlating to a reduction in sHSPs with desiccation intolerance (Wehmeyer and Vierling, 2000). Genes that were classified in detoxification are mostly involved in detoxification of reactive oxygen species, such as 1-cys peroxiredoxin (Prx) and peroxidases, confirming the importance of free-radical processing systems in DT. 1-Cys Prx is an antioxidant that is also present in the resurrection plant Xerophyta viscosa (Mowla et al., 2002). Recently, it was suggested that 1-Cys Prx is employed to sense and/or react to seed environmental conditions, thus preventing germination taking place under unfavourable conditions (Haslekås et al., 2003).


By unravelling changes in gene expression in the radicles of germinated M. truncatula seeds in response to the PEG treatment, this study gives insight into the metabolic and regulatory changes necessary to induce DT. Even so, caution is required, because gene expression does not necessarily lead to the translation of either the gene product or enzyme activity, either mRNA stability could have been altered or mRNA can be stored in mRNPs to function in the rehydration phase (Wood and Oliver, 1999). Nonetheless, an overview of the main transcriptional shifts during the re-establishment of DT may provide an initial framework from which new areas can be explored to better understand the regulatory and protective mechanisms necessary to become desiccation tolerant. It is unlikely that all the regulatory mechanisms leading to the re-establishment of DT in germinated radicles are similar to those involved in the acquisition of DT during maturation, considering the different developmental stages of the organs before DT was acquired. For instance, upon the re-establishment of DT in the germinated radicles, downregulated genes could be related to either the arrest of radicle growth or germination processes that have not yet been initiated in developing seeds. Indeed, only few downregulated genes that are involved in transcription and cellular communication were found to be common between the two datasets. In addition, the re-establishment of DT was achieved though a combination of drought and cold stress, applied by the PEG treatment and incubation at 10°C, respectively. Given the responsiveness of certain transcripts to both cold and drought, a number of genes are likely to change as a result of the combination of the stresses (Seki et al., 2002). Even so, similar functional classes are either up- or downregulated at the later stages of the re-establishment of DT and after the acquisition of DT during maturation (compare Table 1 with Table 3).

An overview of the expression of genes belonging to different classes regulated during the re-establishment of DT is given in a schematic diagram (Figure 6). In general, genes can be classified in relation to their response in terms of time-scale, as was found for genes involved in drought response (Ramanjulu and Bartels, 2002). It has been speculated that the early responsive genes may provide initial protection and amplification of signals, whereas genes that are responding later may be involved in adaptation to stress conditions. Consistent with this idea, during the re-establishment of DT, different time frames exist, both on the transcriptional level and at the metabolite level. At first there is an adaptive stress response, probably caused by the decrease in water potential, and a second phase is linked to the acquisition of tolerance (Figure 6), as discussed hereafter.

Figure 6.

 Schematic diagram representing the time frames of expression of the major gene classes during the transition from the DS to DT stage in radicles of M. truncatula embryos. Classes contain either upregulated (blue) or downregulated (red) genes. Radicles are rendered DT by an osmotic treatment in a poly-ethylene glycol solution (−1.7 MPa, 10°C). Gene clustering suggests that the radicles go through several adaptive phases: first, a stress response involving osmotic adaptation and initial protection, and second, a phase involving the acquisition of DT.

Early stress response and adaptation

Within the first few hours of PEG incubation, when water content decreases as a result of the osmotic adjustment, a set of early responsive genes are transiently expressed (cluster I and II). The main classes represented in these clusters are those of transcription, secondary metabolism and protein fate (Table 1, Figure 6). Within cluster III of the upregulated genes, which have an expression profile that sharply increases within the first 6–9 h of PEG incubation, many genes belong to the class of protection (Tables 1 and 4). Almost all of the LEA protein genes follow this expression profile (Figure 3a), as well as several genes encoding stress proteins and defence proteins. Another functional class with a large number of genes that can be found in this cluster is protein fate, encompassing endopeptidases, ubiquitin ligase and cystatin. Proteolytic activity is high during stress conditions, and thus expression of cystatin, an endogenous cysteine protease inhibitor gene, might be required to block the protease activity. Our results are consistent with previous observations showing that cystatin and other dehydration-inducible genes encoding proteinase inhibitors are expressed in desiccation-tolerant plants (Ramanjulu and Bartels, 2002), some of which have been described as also being induced by water deficit. Also upregulated in these clusters are several genes involved in transport mechanisms from the endoplasmic reticulum to the Golgi apparatus, aquaporins, and in particular outer membrane lipoproteins, small extracellular proteins displaying high specificity for small hydrophobic molecules (Flower, 1996).

Several of the transcriptional regulators in the clusters with genes that are early expressed are homologues of stress-responsive TFs, such as homologues of TINY and DREB2. Intriguingly several of these TFs were also expressed during maturation, implying that this early stress response is also developmentally regulated and is possibly part of the cellular program that will participate in the acquisition of DT (Table 4). It has been shown that a reduction in water content comparable with that measured in this study promotes the acquisition of DT in developing embryos of wheat (Black et al., 1999), also suggesting that comparable signalling pathways operate both during acquisition and the re-establishment of DT.

Expression profiles in parallel with acquisition of DT: return to quiescence

The expression of the genes in the last two clusters of the upregulated genes (IV and V up) and the last one of the downregulated genes (IV down) either increases or decreases at the later stages of the PEG incubation (Figure 2). Cluster IV of downregulated genes shows a massive repression of genes involved in primary and energy metabolism, such as in amino-acid metabolism, carbon metabolism, lipid metabolism, secondary metabolism, TCA cycle and electron transport (Table 1). Depression of metabolism is a key survival strategy for organisms that undergo long periods of either anoxia or freezing (MacDonald and Storey, 1999a), and is also hypothesised to play a role in the DT of seeds (Leprince et al., 2000). Indirect evidence for this comes from the reduced rate of respiration after the re-establishment of DT in radicles of cucumber seeds (Leprince et al., 2000). Several studies with different systems have shown that regulated suppression of activities of membrane ion channels and ion pumps is a key factor in achieving the energy conservation that results in metabolic rate depression (MacDonald and Storey, 1999b; Staples and Hochachka, 1997). Interestingly, genes encoding H+ATPases are downregulated at the end of the re-establishment of DT as was found for drought stress (Seki et al., 2002). In addition to the genes involved in metabolism, those belonging to the classes of communication, protein fate, transport biogenesis and cell cycle are equally downregulated at the end of the re-establishment (Table 1).

Several classes of upregulated genes found in cluster V are typical for the later stages of seed development, such as seed-storage proteins, oleosins and caleosins, and proteinase inhibitors, consistent with the idea of the existence of common regulatory pathways in germinated and developing seeds. Two LEA proteins, Em6 and a member of the D-34 family can also be found in this cluster, with their expression profile differing from that of most of the other LEA proteins (Figure 3a). Interestingly, a recent proteomic study identified this EM6 and another member of the same D-34 family as being specifically linked to DT (Boudet et al., 2006). A second group of detoxification genes are also present in this cluster, such as catalase, peroxidase, 1-cys Prx and lipoxygenase genes that are upregulated during maturation as well. Overall, based on gene expression, it appears that at the end of the re-establishment of DT in the germinated radicles there is a partial return to the developmental stage that the seeds were in prior to germination.

Upregulated TFs during the re-establishment of DT are involved in both abiotic stress and seed development

A large number of TF and signalling molecules were identified that share homology to known genes previously studied in other species, proving some clues about putative regulatory and signalling pathways that might be involved in DT. In the future, confirmation of their action in M. truncatula, as well as further analysis of the function of either kinases or TFs with unknown functions that are specifically linked to DT will be required to ascertain their role in the regulation of seed DT. A review of the functions of the TFs based on data mining suggests that they fall into two main categories: (i) those related to developmental and maturation processes occurring in seeds or (ii) those involved in regulating responses to abiotic stresses.

A bZIP DNA binding protein identified in cluster III up shares homology with ROM2, a repressor of maturation (MAT)- and LEA-specific gene expression in dicots, acting during seed desiccation (Chern et al., 1996). Another Myb TF that was found in cluster IV encoded a tuber-specific and sucrose-responsive element binding factor. The expression of a homologue of this TF from A. thaliana (MYBR1) is affected by the fus3, lec1 and abi3 mutations (Kirik et al., 1998). Here, ABI3 was present but not differentially expressed, and a homologue of FUS3 could not be identified. Yet, considering that several genes upregulated in relation to the re-establishment of DT are also targets of ABI3 (Nambara et al., 2000), and the finding that abi3/fus3 double mutants are DS, might indicate that these TFs are participating in the re-establishment of DT. It has been argued that late embryo development is programmed as an insertion into the sequential growth between early embryo development and post-germinative growth (Raz et al., 2001). The above observations suggest indeed that germinated radicles that undergo the transition from a DS to tolerant stage are returning to a preceding developmental program.

In addition to these developmentally regulated TFs, several homologues of known, drought/stress-induced TFs were also upregulated during the re-establishment of DT. In this study, one AREB-related TF was sharply upregulated during the PEG incubation. In Arabidopsis, the transcription of two ABRE-binding proteins (AREB1 and AREB2) is upregulated by drought, as well as NaCl and ABA (Uno et al., 2000). Furthermore, three genes are expressed in radicles of M. truncatula seeds during the osmotic treatment that show homology to DREB proteins involved in an ABA-independent pathway. Two of them show a sharp and early increase (Figure 3b, cluster II and III), whereas the third one increases at the later stages of the PEG incubation (cluster V). In Arabidopsis, DREB1/CBFs function in cold-responsive gene expression, whereas DREB2 is involved in drought-responsive gene expression (Liu et al., 1998). One of these DREB TFs was also upregulated upon the acquisition of DT during seed maturation Table 4), suggesting that these TFs are also developmentally regulated and arguing for partial overlap of the regulatory pathways involved in drought and DT (Alpert and Oliver, 2002).

Origin of sucrose accumulation

One of the striking observations is the large number of differentially regulated genes that are involved in carbon metabolism (Table 1. Apparently, a considerable effort is put into regulating the production of sucrose (Figure 4, Table 2). Non-reducing sugars are thought to have multiple protective functions in DT (Hoekstra et al., 2001). Early during the drying process they are likely to act as compatible solutes and to participate in preferential exclusion. Upon further drying, they are thought to replace the hydrogen-bonding of water molecules, to interact with membranes and other macromolecules and to participate, most likely with proteins, in the formation of a highly viscous, stabilising, glassy state. The observation that the increase in sucrose content upon PEG incubation coincides with the loss of water suggests that the response to osmotic stress causing this increase is particularly rapid (Figure 1c, Figure 4). So, where does this sucrose originate from? It does not appear to be transported from the cotyledons to the radicles during the PEG incubation, because even in the absence of cotyledons, excised radicles become desiccation tolerant and accumulate a similar quantity of sucrose upon osmotic treatment (Buitink et al., 2003). Most likely, the breakdown of storage reserves such as starch and lipids would be responsible for the sucrose production (Table 2). Indeed, the first upsurge of sucrose can be explained, at least partly, by the lipid breakdown as well as starch breakdown. Cold stress leads to the upregulation of β-amylase gene expression (Seki et al., 2002), and it has been shown that starch degradation is important for enhanced freezing tolerance during an early phase of cold acclimation (Yano et al., 2005). However, between 3 and 6 h, further sucrose accumulation does not seem to be explained completely by the breakdown of starch, oligosaccharides and lipids (Table 2). Considering the simultaneous mobilisation of all the different storage reserves to produce sucrose, and the fact that the main storage reserve of M. truncatula seeds are storage proteins, one wonders whether M. truncatula radicles also mobilise their storage proteins to produce the required sucrose. An observation that points to a role of breakdown of seed storage proteins comes from a proteomic study on the same physiological model system (O. Leprince and J. Boudet, UMR Physiologie Moléculaire des Semeuces, Univ Angers/INH/INRA, personal communication). Fragments of legumins are present in the 3-mm-long germinated radicles, derived from the breakdown of the storage proteins during the germination process. These fragments have disappeared after the PEG treatment, indicating that the amino acids served either as building blocks for protein production or were shuttled through the TCA cycle back into gluconeogenesis to produce sucrose. Further studies will be needed to either confirm or discard this hypothesis. Nonetheless, the accumulation of sucrose appears to be very important for the radicles, considering the activation of a multitude of metabolic pathways related to its production during the PEG incubation.

Experimental procedures

Biological material

Seeds of M. truncatula Gaertn. (cv Paraggio, Seedco, Australia) were imbibed in distilled water at 20°C in the dark. The percentage of germination was determined by counting the number of seeds that showed visible radicle protrusion. To determine the percentage of desiccation-tolerant seeds during imbibition, seeds were removed at different intervals of imbibition and dried for 2–3 days at 20°C under a flow of air at 42% relative humidity (RH) (Buitink et al., 2003). Seeds were considered desiccation-tolerant when the radicle either protruded from the seed coat or resumed growth upon re-imbibition. Data represent the average of three independent replicates of 50 seeds each. To re-induce DT in germinated, DS radicles of seeds of M. truncatula, 20-h-imbibed germinated seeds with a protruded radicle length of between 2.7 and 2.9 mm (referred to as 3 mm in the text) were selected and submitted to an osmotic treatment by incubation in a PEG 8000 solution (−1.7 MPa) at 10°C in the dark, according to the method described by Buitink et al. (2003). At different time intervals, seeds were removed from the PEG solution, rinsed thoroughly and dried as described above. Each time point represents 150–300 seeds originating from 3–5 independent experiments. Water content was determined gravimetrically on a triplicate of 25 radicles (average ± SE). To determine the acquisition of DT during maturation, plants of M. trunculata were grown under a 16-h light regime (250 μmol m−2 s−2 photosynthetic photon flux density) at 23/21°C, and flowers were labelled when pollination had occurred. At different time points, 100 seeds were removed from the pods and either imbibed directly to determine the percentage of germination or dried as described above to determine the percentage of DT. Water contents were determined gravimetrically on a triplicate of five seeds.

Isolation of Total RNA, mRNA and Northern blot analysis

Total RNA was isolated according to Verwoerd et al. (1989) from 50 embryos at two maturation stages (14 and 20 DAP) and from 300 radicles of germinated seeds harvested at different time points during PEG incubation. Seeds were removed from the PEG solution, rinsed and radicles were directly frozen into liquid nitrogen to prevent a wounding response. Total RNA aliquots (400–800 μg) were used to purify mRNA using the PolyATract mRNA isolation system III kit according to the manufacturer's protocol (Promega, Madison, WI, USA). The quality of mRNA was assessed using an Agilent 2100 bioanalyser (Agilent, Palo Alto, CA, USA). For the reference of the kinetics of the PEG-treatment, 20 independent mRNA extractions of 300 untreated 3-mm-long radicles were pooled together. A biological replicate consisted of two separate mRNA extractions of 300 radicles from either two independent PEG time courses or two times 50 embryos (maturation).

To test the accuracy of the microarray results, Northern blot analysis was performed as described in Buitink et al. (2004) using RNA aliquots from separate extractions on seven genes with different expression profiles. The constitutively expressed rRNA 18S was used to check for equal loading.

Microarrays, Cys-labeling of hybridisation targets, hybridisation, and image acquisition

Mt16kOLI1 microarrays contain 16 086 70-mer oligonucleotide probes (Qiagen, Hilden, Germany) representing all tentative consensus (TCs) of The Institute for Genomic Research (TIGR) M. truncatula Gene Index 5 (, as well as different GAPDH controls (Hohnjec et al., 2005). For each M. truncatula probe, duplicate spots are present in the same grids throughout the Mt16kOLI1 microarrays. Cys3- and Cys5-labeled cDNA was prepared using the CyScribe cDNA Post-Labelling Kit (Amersham Pharmacia Biotech, Piscataway, NJ, USA). Two replicate mRNA extractions from the time points of two independent time courses of PEG-treated radicles and embryos aged for 14 DAP were each labeled individually with Cys3. The reference pool (either untreated radicles or 20 DAP) was labeled with Cys5. Immediately before use, microarrays were rinsed once in 0.2% SDS (20°C, 2 min), three times in milliQ (10 sec), 20 min in milliQ at 50°C and dried by centrifugation (40 g, 3 min, 20°C). Each Cys3-labeled sample was mixed with an equal quantity of Cys5-labeled sample, pre-incubated with yeast tRNA and poly A+ RNA, denatured (2 min 100°C, 30 min 37°C) and hybridised to the microarrays. Two independent hybridisations were performed for each RNA sample, leading to four hybridisations per sample. After 16 h of hybridisation at 42°C, microarrays were washed once in 2× SSC, 0.2% (w/v) SDS (1 min 20°C), once in 1× SSC and twice in 0.2× SSC (2 min, 20°C). Slides were dried by centrifugation (40 g, 3 min, 20 C) and treated with Dyesaver (Génisphère, Hatfield, PA, USA). Hybridised arrays were scanned by fluorescence confocal microscopy on a ScanArray Express Scanner (PerkinElmer, Boston, MA, USA) at a laser power ranging from 75% to 100% and photomultiplier tube gain settings ranging from 65% to 100%. Measurements were obtained separately for each fluorochrome at 10 μm/pixel resolution. Data are submitted to ArrayExpress under the accession number E-MEXP-664.

Data analysis, k-means clustering and electronic Northerns

Signal intensities were extracted with the genepix pro 5.0 image analysis software (Axon Instruments, Union City, CA, USA). Image data processing was performed using the madscan software (Le Meur et al., 2004). This dynamic procedure physically validates the quality of the raw data points and the quality of the microarray, and it corrects systematic and random biases by normalising the filtered data (for details see For each time point a total of four slides containing two measurements were analysed, except for 72-h PEG, where each biological replication was represented by one slide. Data were retained based on concordance between the measurements that was calculated at the three levels of the experimentation. First, only data where the two values originating from the same slide were considered concordant were retained. Measurements were considered concordant when the absolute difference between the different replicates was <0.4 for M<1 and 1 for M > 1, with M specifying the log2 expression ratio. Subsequently, only data where the average values originating from the two slides of the same biological replicate were concordant were retained and the final data set was limited to those data for which the average values originating from the two biological replicates were concordant. This way, a minimum of two values from each biological replication had to show concordance between each other and between the biological replicates in order for the gene to be considered for further analysis. For the 72-h PEG time point, all the four measurement of the two slides had to be concordant in order to pass further analysis. On average, about 10 000 genes per time point were considered concordant and thus retained. In order to select differentially expressed genes, the range (absolute difference between the minimum and maximum values) for each of the 26 = 64 assemblies, representing the two biological replications of the PEG time courses (containing six time points) was calculated by permutation. Genes were considered variant when the 90th percentile of the 64 ranges >0.7. Possible differences as a result of gene-specific dye bias were thus eliminated. In addition, the reproducibility between the two PEG time courses was confirmed by retaining the genes for which the Pearson's product moment correlation coefficient r > 0.75. Using this strict data treatment, a final number of 1310 genes were considered as being reproducibly and differentially expressed during the PEG incubation. In order to compare desiccation-tolerant and -sensitive tissues, either by comparing the radicles of germinated seeds incubated for 72 h in PEG solution with those without treatment or by comparing seeds aged 20 DAP with those aged 14 DAP, genes were selected based on concordance of the three levels of experimental design as discussed above. Subsequently, genes were considered differentially expressed when M > 0.8. To account for possible artefacts caused by gene-specific dye bias, only genes that were shown to be differentially expressed during the PEG kinetics (determined when either M > 0.7 or when the difference in log2 expression ratio compared with the value at 1 h PEG was >0.7) were considered. k-Means clustering was performed using genesis software (Sturn et al., 2002). The number of ESTs from the M. truncatula seed libraries were obtained from the expression summary of the TIGR MtGI8 ( The description of the EST libraries was retrieved from MtDB2.0, the M. truncatula Consortium database (

Metabolite determination

Analyses of starch, soluble sugars (hexoses) and lipids were carried out on excised radicles at different time points during PEG incubation. Analyses were performed in triplicate on three different biological samples and expressed on a DW basis. DW was determined gravimetrically. Starch and soluble sugars were determined according to Baud et al. (2002). Total fatty acid (FA) concentration was determined by measuring methyl ester derivatives following the protocol of Baud et al. (2002). Oligosaccharide contents were determined according to Hoekstra et al. (1994) using DIONEX-HPLC on a triplicate extraction of 25 radicles, using melisitose as an internal standard.


We thank B. Jettner (Seed-Co Australia Co-Operative Ltd, Hilton, SA, Australia) for the generous gift of M. truncatula cv Paraggio seeds. Drs C. Rochat and M. Miquel are gratefully acknowledged for helpful comments on the manuscript and the anonymous reviewers for suggestions on the statistical approaches. This work was supported in part by grants from the Ministère de l'Agriculture, de l'Alimentation, de la Pêche et des Affaires Rurales, Contrat Etat-Région 2000–2006, Conseil Général de Maine-et-Loire, Conseil Régional Pays de la Loire, CNRS, INSERM and INRA.