The information provided by ESTs of randomly isolated gene transcripts generated under specific abiotic stress conditions provides an opportunity for gene discovery in addition to identifying the biochemical pathways involved in plant physiological responses 10. Here, we describe ESTs obtained from salinity-induced cDNA library prepared from the leaves of the Glycine soja exposed to stress for a short period of time. Insert amplification of all random clones from cDNA library revealed inserts ranging between 500 bp and 2000 bp, with an average size of 1250 bp. A total of 2,219 clones were sequenced, and 2,003 cleaned EST sequences were generated for further analysis after trimming off vector sequences and removing of sequences shorter than 100 bp (GenBank_Accn: DT082443~DT084445). The average read-length of cleaned ESTs was 454 bp. The cleaned ESTs include 1936 5'end sequences and 67 3'end sequences (Table 1). The average G+C content of Glycine soja ESTs was 40%, which is similar to that of soybean 11. The 2003 ESTs were assembled into 375 contigs and 696 singlets (clusters) using the PHRAP program (Table 1). The frequency of EST distribution after clustering is shown in Fig. 1. Nine contigs had 10 or more ESTs, with the largest one containing 27 ESTs. Most contigs contained one to six ESTs. The redundancy level of EST collection was 65%, which means that continued sequencing of cDNAs selected at random from our libraries still has considerable potential to uncover novel sequences.

Blastn was used to compare the EST sequences of Glycine soja to Glycine max, Arabidopsis and rice. The E-value was set at 1e-30. Although the size of Glycine max Gene Index is smaller than the AGI and OGI, the sum of matching section between Glycine soja and Glycine max (3106) was far more than Glycine soja versus Arabidopsis or Glycine soja versus Oryza sativa (Table 2). Note that there is great difference in stress-tolerant characteristics between soybean and wild soybean, although they share a large amount of homologs in expressed sequences. This indicates that the discrepancy in stress responses may come from the subtle difference between the homologous sequences. It is therefore feasible to investigate the wild soybean's gene expression profile using the Affymetrix soybean chip.

In order to get more information about the expression pattern of Glycine soja ESTs, BLASTN was used to search against the Arabidopsis CDS from TAIR, and 244 ESTs were highly similar to genes from Arabidopsis. The corresponding Arabidopsis genes were searched for the expression data under salt stress since global expression profiling of the Arabidopsis was available from TAIR12. As a result, a total of 126 ESTs were predicted to be up-regulated in response to salt stress according to AtGenExpress, and may be induced by salt stress. This prediction will be confirmed by further analysis.

As shown in Tables 3 and Figure 2, all unigenes were classified according to terms of biological processes, molecular functions and cellular components, developed by the Gene Ontology Consortium 13 in Uniprot (EBI). These genes cover a broad range of the GO functional categories. However, due to the lack of gene products information, many transcripts cannnot be functionally categorized. These 'unknown' genes are likely the source of candidate salt-tolerant genes and further functional analysis will help elucidate their specific roles in salt tolerance 14.

We successfully classified 279 unigenes in terms of biological processes (Fig. 2A), 301 unigenes in terms of molecular function (Fig. 2B), and 262 unigenes in terms of cellular components. Since one gene product may be assigned to more than one GO terms, and one children term can fit into multiple parental categories, the total number of GO mappings in each of the three ontologies will exceed the number of genes.

A large proportion of genes were found to participate in the biological process of metabolism (69%), followed by cell growth and/or maintenance (13%). The accumulation of osmoprotectants by either altering metabolism or increasing transport is an important process of plants for the adaptation to environmental stress 15. It has been reported that in Arabidopsis, salinity induces programmed cell death in primary roots and the plants produce secondary roots which function better under abiotic stress 16. The increase in metabolism could be essential to nutrient redistribution and new tissue development, a strategy the plants adopted to cope with the changed environment.

Our results showed that 4% of the unigene set responds to external stimulus, while 2% responds to stress (Fig. 2A). These two catgories form the basis for mining the stress-regulated genes. Genes encoding dehydration-induced ERD15 protein (DT083772), late embryogenesis abundant (LEA) protein (DT084384) and other stress-induced proteins were found in these categories. Submergence induced gene, induced by anaerobic stress, was also found in the ESTs sequenced (DT082680). There were also other genes function as scavengers of reactive oxygen species, such as catalase, glutathione S-transferase, and superoxide dismutase. These gene products are needed to maintain the redox homeostasis under abiotic stress. It was reported that overexpression of H₂O₂-scavenging enzymes increased the tolerance of plants to abiotic stress17. Metallothioneins (MT) are a group of low-molecular-weight (LMW) metal-binding proteins with a high cysteine content that are thought to be involved in metal ion metabolism and detoxification 18. MT-like transcripts have been reported to be highly up-regulated in response to salt stress in barley 1920. Type 2 metallothionein (DT083320, DT083023) was present in our database.

In addition, proteins involved in the regulation of signal transduction pathway (Fig 2B) have been categorized separately. In plant cells, calcium functions as a second messenger coupling a wide range of extracellular stimuli to intracellular responses 21. Calmodulin, one major class of Ca²⁺ sensor characterized in plants, which was present in the Glycine soja ESTs (DT083725), is involved in stress signal transduction suggested by several lines of evidence 212223.

Genes for transcription factors that contain typical DNA binding motifs, such as MYB, bZIP, have been demonstrated to be stress inducible 24. Transcription factors containing similar domains are present in the Glycine soja ESTs and may be important in regulating the response to salt stress.

Seeds of Glycine soja (50109) were inoculated in half-strength solid MS medium (pH5.8) in the dark until germination. Plants were grown at 25°C in a greenhouse with a photoperiod of 15 h light/9 h dark. One-month-old seedlings were transferred into 150 mM NaCl solutions. Equal leaves were sampled at 0.5 h, 1 h, 3 h and 6 h and immediately frozen in liquid nitrogen. Frozen tissues were stored at -80°C until use.

Total RNA was isolated from plant materials with Trizol (Invitrogen) according to the manufacturer's instructions. The RNA concentration was determined by spectrophotometry, and its integrity was assessed by electrophoresis in 1% (w/v) formaldehyde-agarose gels 25.

For the full-length cDNA library, 2 μg of mRNA were used for cDNA synthesis using the SMART cDNA synthesis kit (Clontech, Palo Alto, CA, USA) according to the manufacturer's protocol. The resulting double-stranded cDNAs were digested with SfiI and ligated into the SfiI site of λ TriplEx2. The phagemids were packaged according to the instruction of Gigapack III Plus-7 packaging extract kit (Stratagene company). The average titer of the libraries was ~2 × 10⁵ pfu/ml.

Homologous recombination with E. coli BM25.8 was conducted to convert the phage libraries to the plasmid form. 8300 colonies were randomly selected and activated as templates of PCR reactions. The primers are as follows: P5':5'-GGCCATTACGGCCGGG-3'; P3':5'-CCGAGGCGGCCGACATG-3'. PCR was performed for 30 cycles of 30 s at 94°C, 30 s at 69°C and 2 min at 72°C. The PCR products were electrophoresed next to DNA size markers to estimate the molecular sizes of the insert DNAs. The clones with inserted fragments' size ≥ 500 bp were sequenced by Shanghai Sangon Company.

The trimming process, which included the removal of low-quality sequences, poly(A) tails, ribosomal RNA, and vector regions, was conducted as described by Telles and da Silva 26 with minor modifications. In addition, sequences shorter than 100 bases were not included in the analysis.

The resulting sets of cleaned sequences were assembled into contigs by PHRAP program27 using the following parameters: minmatch 100, minscore 94.

To assign annotation to contigs, BLASTX was used to search the Uniprot (EBI) with terms from the Gene Ontology Consortium28 controlled vocabularies. The expectation value (e-value) cutoff for BLASTX was set at 1e-5.

In order to survey the similarity between soybean and wild soybean expressed sequences, our set of ESTs was blasted against local installations of GMGI (Glycine max Gene Index, release 12), AGI (Arabidopsis Gene Index, release 12) and OGI (Oryza sativa Gene Index, release 16) from TIGR. The Glycine soja ESTs were also blasted against Arabidopsis CDS from TAIR (release 6) at 1e-15. The raw data (cel file) of microarray experiment of Arabidopsis from TAIR (AtGenExpress) were used to identify up-regulated CDS of Arabidopsis response to salt stress. The software RMAExpress (Ben Bolstad) was used to scale/normalize the raw data.