We used the biotinylated CAP trapper method 29 to construct a full-length cDNA library of Thellungiella halophila (Shandong ecotype) from whole plants as well as from various tissues, including leaves, roots, flowers, siliques, and mature seeds, of plants treated with high salinity, chilling, freezing stress, and ABA (Table 1). The λFLCIII vector 33, which accommodates cDNAs in a broad range of sizes and is useful for the high-efficiency cloning of long cDNA fragments, was used for the construction of the cDNA library. To reduce the frequency of representation of highly expressed mRNAs in the library, normalization procedures 29 were employed in the construction process. The 20000 recombinant clones were randomly selected and sequenced from both ends. We determined 18636 and 16535 sequences from the forward and reverse directions, respectively, and from among the 20000 clones we obtained the forward or reverse sequences of a total of 19429 clones (Table 2). A total of 35171 sequences have been deposited in the DDBJ public sequence database (accession numbers, BY800476 to BY835646). We have named these "RTFL" (RIKEN Thellungiella Full-Length) cDNAs.

Figure 1 shows the size distribution of the Thellungiella cDNA inserts from 1161 randomly selected clones. The average size was approximately 1.54 kbp. Our group previously determined 20683 full-read cDNA sequences from the RAFL (RIKEN Arabidopsis Full-Length) cDNA collection, and these sequences are available in the RARGE database 34. The estimated average size of the RAFL cDNA inserts was 1.495 kbp (Motoaki Seki et al., RIKEN Plant Science Center, unpublished results). The average size of the Thellungiella cDNA inserts was thus slightly longer than the average cDNA inserts from Arabidopsis libraries and similar to those in other plants; for example, the average rice and wheat cDNA lengths are both about 1.5 kbp 1018.

The 35171 sequences were assembled by using the CAP3 program 35 to evaluate the level of sequence redundancy. Assembling these sequences generated 7402 contigs and 6556 singletons, and the sequenced 20000 cDNAs were clustered into 9569 nonredundant scaffolds that represented distinct genes (Table 2). Figure 2 shows the degree of redundancy in the sequences from the cDNA library. The majority of the cDNAs (6024 cDNAs, 63% of the total), consisted of a single cDNA in a cluster, and only 5% contained more than six cDNAs in a cluster, indicating that the normalization procedures were successful.

The cDNA sequence data were submitted for the BLASTN search to compare with the green plants (Viridiplantae) mRNA databases in GenBank. Of the 19429 clones that we obtained as clean sequences, 18 295 clones (94%) showed > 80% identity, whereas the remaining clones (6%) had no significant identity to any plant sequences in GenBank.

We examined the proportion of full-length cDNA clones in our library. We selected clones that (1) had sequences from both ends, and (2) showed an Expected Value (E) < 1.0e^-20 on the basis of a fastx search using forward sequences as queries against Arabidopsis proteins (TIGR v5, ATH1.pep, ftp://ftp.tigr.org/pub/data/a_thaliana/ath1/SEQUENCES/), with the correct direction of the reading frame. We considered clones to be full-length if they met the following criteria: (1) they contained the first methionine, and (2) the reverse sequence contained the polyA sequence. Consequently, we selected 12878 clones as calculation objects and classified 10880 (84.5%) of these clones as full-length clones. This frequency is nearly identical to the reported values from the Arabidopsis 14, rice 15, and wheat 18 libraries.

The 9569 nonredundant genes were submitted to InterPro 36 to obtain functional domain information. InterPro is an integrated resource for protein families, domains and functional sites that integrates the following protein signature databases: PROSITE, PRINTS, ProDom, Pfam, SMART, TIGRFAMs, PIRSF, SUPERFAMILY, Gene3D, and PANTHER. Protein matches in InterPro are pre-calculated by using InterProScan software, which combines the different protein signature recognition methods offered by the InterPro member databases into one resource and provides the corresponding InterPro accession numbers and Gene Ontology (GO) annotations 37. A total of 8289 sequences were assigned to InterPro IDs and GO terms [see Additional file 1]. According to the obtained GO terms, the 8289 genes were remapped and classified by using Plant GO Slim (http://www.geneontology.org/GO.slims.shtml; 38).

Figures 3 and 4 show the categorization of Arabidopsis genes and the 8289 Thellungiella genes assigned to the GO terms. In most categories, we observed no obvious differences between the numbers of sequences from Arabidopsis and Thellungiella, including genes involved in biological processes, cellular components, and molecular function. Notably, the number of sequences classified as transcription genes under biological processes in Arabidopsis was approximately twice that in Thellungiella (Fig. 4). Furthermore, the number of Arabidopsis genes associated with the nucleus under cellular components was also much higher than that in Thellungiella (Fig. 3A). Although the total number of transcription factors is not clear in Thellungiella genome, there seems to be no significant difference in the proportion of transcription factor genes between Arabidopsis and Thellungiella genome. The population of cDNAs may reflects the levels of gene expression. Thus, the expression level of Thellungiella transcription factors may be lower than that of Arabidopsis. Arabidopsis responds strongly to abiotic and biotic stresses at the transcriptional level. In contrast, Thellungiella does not initiate immediate changes in transcription in response to abiotic stresses, and instead constitutively expresses a large number of genes that correspond to stress-inducible genes in Arabidopsis 5. Thus, the difference between Arabidopsis and Thellungiella in their responses to various stimuli at the transcriptional level may reflect differences in the number of transcription-related genes between the organisms and may depend partly on the number of transcription factors.

Most Thellungiella genes have a high sequence identity (approximately 90% at the cDNA level) to Arabidopsis genes. Numerous studies of salt tolerance in Arabidopsis suggest that this plant contains most, if not all, the salt-tolerance related genes that might be found in halophytes39. The current hypothesis is that halophytes employ salt-tolerance mechanisms similar to those found in glycophytes, including Arabidopsis. However, subtle differences in this regulation result in large variations in salt tolerance between glycophytes and halophytes11. In addition, halophytes are hypothesized to exhibit specific salt-tolerance mechanisms resulting from the induction of halophyte-specific genes. We divided the 8298 genes into two groups on the basis of their sequence identities, using BLASTX searches against the Arabidopsis database. The group with high sequence identity to Arabidopsis genes (E value ≤ 1.0e^-50) included 7535 genes, and the group with low identity to Arabidopsis genes (E value > 1.0e^-50) included 763 genes. Previous studies revealed that a plasma membrane Na⁺/H⁺ antiporter (SOS1), a vacuolar Na⁺/H⁺ antiporter (NHX1), and a plasma membrane Na⁺ transporter (HKT1) are essential for the salt tolerance of Arabidopsis 404142, and these mutants exhibit a salt-hypersensitive phenotype. In contrast, plants that overexpress SOS1 and NHX1 show higher salt tolerance than wild-type plants 4344. The co-ortholog Thellungiella genes belong to the first gene group, exhibiting high identity to Arabidopsis genes. This suggests that some salt-tolerance mechanisms are common to both glycophytes and halophytes.

We compared the categorization of the whole Arabidopsis genome with the categories of the 763 Thellungiella genes that exhibited low identity to Arabidopsis genes (Fig. 4). Of the genes involved in biological processes, the number of genes in the categories for transport, DNA metabolic process, generation of precursor metabolites and energy, response to abiotic stimulus, multicellular organismal development, response to external stimulus, and cell differentiation in Thellungiella were more than 1.5 times the number in Arabidopsis (Fig. 4). Moreover, in regards to molecular function, the proportion of genes involved in transporter activity in Thellungiella was also higher than in Arabidopsis. Less NaCl accumulates in Thellungiella plants than in Arabidopsis under similar salinity conditions, suggesting that Thellungiella has a superior system for suppressing Na⁺ influx or for excreting Na⁺ 5. Electrophysiological analysis indicates that Thellungiella also exhibits high potassium/sodium selectivity, implying that Thellungiella has specific ion channel features that lead to superior homeostasis with respect to sodium and potassium 7. Arabidopsis that overexpresses a plasma membrane Na⁺/H⁺ antiporter gene, SOS1, shows salinity tolerance and represses its sodium uptake compared with that of wild-type plants 44. Likewise, the expression level of SOS1 in Thellungiella is higher than in Arabidopsis 545. Although SOS1 overexpression suggests a contribution of this gene to the salt tolerance of Thellungiella, the large proportion of transport genes may imply that Thellungiella has a distinct ion transportation system regulated by these specific genes.

Table 3 [see Additional file 2] lists the Thellungiella genes with low identity (E value > 1.0e^-50) to the Arabidopsis genes classified under transporter genes. Several transporters, including chloride channels and P-type H⁺-ATPase, play important roles in the salt tolerance of plants. Homeostasis of Na⁺ and Cl^- is an important mechanism to reduce NaCl stress in higher plants. Chloride channels (CLCs) are a group of voltage-gated Cl^- channels originally identified in animals 46; they have diverse cellular functions such as stabilizing cell membrane potential and regulating cell volume and transcellular chloride transport 47. Recently, a chloride channel gene, GmCLC1, was cloned from soybeans 48. Transgenic tobacco BY-2 cells expressing GmCLC1 were able to drain Cl^- more efficiently from vacuoles than was the case in untransformed BY-2 cells, and the transgenics showed a higher NaCl tolerance 48. The plant cell membrane is energized by an electrochemical gradient produced by P-type H⁺-ATPase (proton pump). These pumps are encoded by at least 12 genes in Arabidopsis. One of the Arabidopsis P-type H⁺-ATPase genes, AHA4, was expressed most strongly in the root endodermis 49. The aha4 mutant plants exhibited a clear growth reduction under a mild stress of 75 mM NaCl compared with wild-type plants, and the ratio of Na⁺ to K⁺ in the aha4 mutants increased to between four and five times the values in wild-type plants. These results suggest that the aha4 mutants were compromised in their ability to exclude Na⁺ under salinity stress 49. P-type H⁺-ATPases were also found in a halotolerant cyanobacterium, Aphanothece halophytica, and a marine alga, Tetraselmis viridis 5051. Aphanothece halophytica grows under a wide range of salinity conditions (from 0.25 to 3.0 M NaCl), and Na⁺/H⁺ antiporters in A. halophytica play a crucial role in Na⁺ efflux to provide enhanced salt tolerance. Since the efflux of Na⁺ mediated by Na⁺/H⁺ antiporters utilizes protons as the motive force provided by a primary proton pump, H⁺-ATPase, the P-type H⁺-ATPase is thought to contribute to the salt tolerance of this species 52. On the other hand, vacuolar ATPase (V-ATPase) is the major proton pump that establishes and maintains an electrochemical proton gradient across the tonoplast. Expression of several V-ATPase subunits or an increase in V-ATPase activity induced by salt stress has been observed in a number of glycophytic species 53, suggesting that increased V-ATPase levels or activity are required to drive Na⁺ sequestration under salt stress. Recently, the V-ATPase-deficient det3 Arabidopsis mutant was shown to be extremely salt sensitive. Moreover, SOS2, a protein kinase that phosphorylates SOS1, interacted directly with the V-ATPase regulatory subunits B1 and B2 54. These studies indicate that V-ATPase activity plays a key role in salt tolerance. Although most Thellungiella genes show approximately 90% identity with Arabidopsis genes, the Thellungiella genes encoding transporters appear to be remarkably different from their Arabidopsis co-orthologs. Whether the sequence diversities among these genes are the source of the large differences in salt tolerance between Thellungiella and Arabidopsis is a topic of great interest.

Thellungiella halophila (Shandong ecotype) seeds were sown on Murashige and Skoog (MS) plates containing 0.8% (wt/vol) agar and 1% sucrose. The seeds were stratified at 4°C for two weeks and then transferred to 22°C under continuous light for germination and growth. Three weeks after germination, seedlings of Thellungiella were transferred to 250 mM NaCl (salt stress) or 50 μM ABA (ABA treatment) water, or were transferred onto separate 9-cm plastic pots filled with a 1:1 mixture of perlite/vermiculite and watered with 1000-fold diluted Hyponex™ (Hyponex, Osaka, Japan). One week after transfer onto the soil pots under 16 hours light – 8 hours darkness at 22°C, the seedlings were subjected to 4 °C (cold stress) or -6°C (freezing stress) in a growth chamber under 24 hours darkness.

A Thellungiella full-length cDNA library was constructed from a mixture of mRNA extracted from stress-treated plants and various tissues of Thellungiella. Thellungiella plants were subjected to various stress treatments: high-salinity (250 mM NaCl for 1, 2, 3, 7, and 14 days), cold temperatures (4°C for 2, 4, 8, and 24 hours), freezing temperatures (-6°C for 1, 2, 4, or 8 hours), or ABA (50 μM ABA for 1, 2, 4, and 8 hours). Control plants were grown under unstressed conditions under16 hours light – 8 hours darkness at 22°C. After the stress treatments, mRNA was extracted from whole plants (salt stress and ABA treatment) or rosette leaves (cold and freezing stresses) collected at each point in time. Rosette leaves and cauline leaves, roots, flowers, and siliques were collected from 7- to 10-week-old plants, and mature seeds were collected from 12- to 20-week-old plants.

Total RNA was prepared by using TRIZOL Reagent (Life Technologies, Rockville, MD, USA) from the treated samples. A full-length cDNA library was constructed as previously reported 142728 by means of the biotinylated CAP trapper method using trehalose-thermoactivated reverse transcriptase 28. We used the λFLCIII 33 vector, which accommodates cDNAs in a broad range of sizes and is thus useful for the high-efficiency cloning of long cDNA fragments, for construction of the cDNA libraries 33. The λFLCIII vectors can also be bulk-excised by a Cre-lox-based system free of size bias to generate the plasmid libraries. Normalization 29 was also introduced in the construction of the full-length cDNA library to reduce the frequency of highly expressed mRNAs in the library. The method is based on hybridization of first-strand, full-length cDNA as the tester and cellular biotinylated RNA extracted from stress-treated plants and various tissues of Thellungiella as the normalizing driver.

The DNA of each clone was directly amplified from 384 bacterial cultures in a glycerol stock plate by the RCA method 55 using a TempliPhi HT DNA amplification kit (GE Healthcare, United Kingdom). End sequencing of 20000 clones was performed using ABI 3700 capillary sequencers (Applied Biosystems, Foster City, CA, U.S.). The M13 (-21) primer (5'-TGTAAAACGACGGCCAGT-3') and the 1233 primer (5'-AGCGGATAACAATTTCACACAGGA-3') were used for forward and reverse sequencing, respectively.

We used sim4 software for the detection of vector sequences 56. Raw sequence data were base-called using the Phred software 5758. Regions of low quality found at both edges of each raw sequence were discarded, and we extracted only the high-quality region (Phred quality score > 14, and more than 20 bases repeated). After this initial evaluation, sequence data shorter than 100 bases in length or with many low-quality regions (Phred quality score ≤ 14, and more than 50% of its total length) were omitted. The ESTs were assembled by using CAP3 software 35 with its default parameters. All sequences were submitted to the DNA Databank of Japan (DDBJ) with accession numbers BY800476 to BY835646.

The sequence lengths of the Thellungiella cDNA inserts were determined from a total of 1,161 clones by digestion with SfiI (in the cDNA cloning site) or PCR amplification using T3 (5'-TGTAAAACGACGGCCAGT-3') and T7 primers (5'-AATACGACTCACTATAGGG-3').

To examine the proportion of full-length cDNA clones in this library, we selected the following clones as calculation objects: (1) clones with sequences from both ends, and (2) clones showing an Expected Value of (E) < 1.0e^-20 in a fastx search using forward sequences as queries against Arabidopsis proteins (TIGR v5, ATH1.pep), with the correct direction of the reading frame. We used the following criteria to classify clones as full-length: (1) clones must include the first methionine, and (2) the reverse reading sequence must include the polyA sequence.

In order to obtain a non-redundant set of transcripts, the clones were clustered according to clone names. To accomplish this, we parsed the .ace file from the CAP3 program output to build scaffolds, which are groups of sequences that represent a unique transcript for which the relative position and orientation of the fragments can be inferred. Using clone names, the contigs or singletons corresponding to the two ends of a given clone were joined by adding 20 N's in the middle of both sequences. Since 20 is more than the default window size in BLAST searches, these N's did not interfere with the BLAST analyses.

Once these scaffolds were created, the sequences were submitted to InterPro 36 to obtain functional domain information. Protein matches in InterPro were pre-calculated with InterProScan software, available from http://www.ebi.ac.uk/Tools/InterProScan/3759. InterProScan provided the corresponding InterPro accession numbers and GO annotation in the results 37. The genes assigned to InterPro ID were classified according to the GO terms developed by InterPro using Plant GO Slim (http://www.geneontology.org/GO.slims.shtml; 38).