Figure 1 shows sections of developing guar seeds at 25 days after flowering (DAF) and of mature seeds at 40 DAF. The mature seeds have a large endosperm packed with reserves of carbohydrate (principally galactomannan), protein, lipid and minerals, which provide a reserve for the developing seedling for several days. In order to investigate developmentally regulated genes with a focus on galactomannan biosynthesis, two cDNA libraries were constructed. The "Early" cDNA library (library I) was made from seeds 15, 20 and 25 DAF, and the "Late" library (library II) from seeds at 30, 35 and 40 DAF. Developmental time points (DAF) were chosen for pooling based on maximal transcript levels of two key enzymes of galactomannan biosynthesis, galactosyl transferase and mannan synthase 41420. As described in our results below, the highest expression level of galactosyl transferase was detected by RT-PCR at 35 DAF and no mannan synthase expression was detected prior to 30 DAF. In total 16,476 ESTs from both cDNA libraries were sequenced, comprising 8,163 and 8,313 ESTs from libraries I and II, respectively. A total of 7,694 unique sequences, or unigenes (UG) were identified, of which 1,695 represented contigs and 5,999 represented singletons. Library I contained 4,804 unigenes, and library II contained 3,609. Surprisingly, only 719 unigenes were common to both libraries (Figure 2A). EST sequences of all clones are available at GenBank (Accessions EG974821 through EG991296).

ESTs were annotated with reference to gene function using the results of BLASTX comparisons with the GenBank non-redundant protein database (NR). EST sequences were grouped in three categories based on the "bit score" S' 21 of the aligned sequence segment with the top database hit after BLASTX comparison. The "secure" assignment group contains 1,662 unigenes (22% of the total) with the S' score value equal to or greater than 200; the "putative" assignment group contained 3,941 unigenes (51%) with the S' scores less than 200; the "no assignment" group contained 2,091 unigenes (27%) with no score. A BLASTX comparison of the 2,091 unigenes with no score was made against the Medicago truncatula genome v 1.0 22, which resulted in an additional 377 annotations. For sequences that did not have BLASTX scores, no protein similar to the translation product was present in the public databases at the time of analysis. We therefore assume that approximately 27% of the clones in the seed database encode previously undescribed proteins or may represent non-coding RNA.

The largest group of ESTs fell into the "putative" assignment group. This group could reduce dramatically with additional efforts to improve the length of the sequencing reads and quality of the sequence data. For most of the analyses described, only the "secure" assignment group was considered for distributing genes into functional categories in order to gain a preliminary understanding of metabolic processes during guar seed development (Figure 2B,C). However, both "secure" and "putative" assignment groups were used to identify candidate genes for specific biochemical pathways.

Seed development is genetically programmed and is associated with striking changes in metabolite levels. Differentiation occurs successively, starting with the maternal and followed by the filial organs, which later become highly specialized storage tissues. A complex regulatory network triggers initiation of seed maturation and corresponding accumulation of storage products. This includes transcriptional and physiological reprogramming mediated by sugar and hormone-responsive pathways 2324.

Galactomannan and seed storage proteins accumulate to high amounts in mature guar seeds, representing 26–32% and 23–31% of the seed dry weight, respectively 25. The biosynthesis of carbohydrate and storage proteins in guar seeds is probably preceded by increased transcriptional activity for these processes. Consistent with this hypothesis, the distribution of functional ontologies in the EST database (excluding unknown, hypothetical and non-classified genes) revealed major contributions from genes annotated as encoding proteins involved in signal transduction (10.9%), carbohydrate metabolism (10%), chaperone and proteolytic processes (9%), and translation and ribosomal structure (7.8%) (Figure 2B).

Mature seeds have very low metabolic activity, reflected by the lower representation of specific EST classes in library II. Genes annotated as involved in signal transduction were represented by four times as many ESTs, carbohydrate metabolism three times, chaperone and proteolytic activity 1.8 times, and translation and ribosomal structure 1.4 times, in library I compared to library II (Figure 2C, Additional file 1). However, three functional categories were represented by higher numbers of ESTs in library II. These include seed storage proteins (SSPs), and hormonal and stress/pathogen induced genes. SSPs accumulate to high levels during the late stages of seed development. Among the "stress/pathogen response" group of genes, one highly induced contig (UG00086) was represented by 46 ESTs in library II. This gene showed 81% amino acid similarity to a ripening-related protein from soybean (Glycine max) [GB# AAD50376] which is activated in soybean-soybean cyst nematode interactions and contains a conserved domain for the pathogenesis-related protein Bet v I family.

UG00177, in the hormone-inducible functional category, was represented by 26 ESTs in library II. The encoded protein showed 85% amino acid similarity to an auxin down-regulated gene from soybean 26, the function of which is yet to be determined. Five and seven ESTs" in libraries I and II, respectively, corresponded to genes involved in the biosynthesis of gibberellic acid (GA) (Additional file 1). Synthesis of GA in developing seeds is necessary to promote cell expansion 27.

Biosynthesis – Galactomannan is the major storage polysaccharide in guar seeds and accumulates in cell walls of the endosperm, accounting for up to 26–32% of the seed dry weight 25. Figure 3 shows an outline of galactomannan metabolism in guar, highlighting the importance of sucrose as a building block. In most plant species carbon is transported as sucrose. Cleavage of the O-glycosidic bond between the glucose and fructose units of sucrose is catalyzed by invertase (EC 3.2.1.26) and sucrose synthase (EC 2.4.1.13) 28. Invertase is a hydrolase, cleaving sucrose irreversibly into glucose and fructose, whereas sucrose synthase is a glycosyl transferase, converting sucrose in the presence of UDP into UDP-glucose and fructose. Two ESTs corresponding to different invertase unigenes were detected only in library I. Likewise, of the 11 unigenes corresponding to sucrose synthases, most were also represented by ESTs found in library I (Table 1).

During seed development, entry of carbon from the maternal coat cells into the seed apoplasm is mediated by membrane-localized sugar transporters 2930. Twelve unigenes annotated as sugar transporters were found in the guar seed cDNA libraries (Table 1). All ESTs, with the exception of UG05960, were detected in library I, suggesting that sugar transporters are actively transcribed, and presumably function, during early stages of guar seed development.

No hexokinase ESTs were detected in either of the cDNA libraries. Plant hexokinase (HXK) has been shown to be involved in sugar sensing and signalling, and is proposed to be a dual-function enzyme with both catalytic and regulatory functions 31323334. For example, transgenic Arabidopsis plants over-expressing AtHXK1 and AtHXK2 showed enhanced sensitivity to glucose containing medium 31. Overexpression of the Arabidopsis AtHXK1 in transgenic tomato plants led to reduced photosynthetic activity 32. HXK is presumably encoded by low abundance transcripts in developing guar seeds.

Phosphomannoisomerase (EC 5.3.1.8) converts fructose-6-phosphate (Fru-6-P) to mannose-6-phosphate (Man-6-P). This enzyme also functions in the reverse direction in the utilization of mannose released by hydrolysis of galactomannan on germination, after it is phosphorylated to Man-6-P 35. No ESTs annotated as phosphomannoisomerase were detected in either of the libraries. However, two unigenes corresponding to phosphomannomutase (EC 5.4.2.8), which reversibly converts D-mannose 6-phosphate to α-D-mannose 1-phosphate, were identified; four ESTs were found in library I and three ESTs in library II.

The direct precursors for galactomannan biosynthesis, GDP-D-mannose and UDP-D-galactose, are formed by the actions of GDP mannose phosphorylase (EC 2.7.7.22) and UDP-galactose 4-epimerase (EC 5.1.3.2). In vitro experiments have shown that the relative concentrations of these precursors can affect the M:G ratio of the galactomannan polymer 5. Of the three ESTs corresponding to GDP mannose phosphorylase, one was found in library I and two in library II. Two ESTs corresponding to UDP-galactose 4-epimerase were detected only in library I.

Two tightly membrane-bound glycosyltransferases together catalyze the formation of galactomannans. GDP-mannose-dependent mannosyltransferase transfers mannose residues to the end of the growing linear (1→4) β-linked mannose backbone of the galactomannan polymer 5620. Simultaneously, UDP-galactose-dependent galactosyltransferase transfers a galactose residue through a (1→6) α-linkage to a mannose at or near the nonreducing end of the growing mannan chain 56. Importantly, galactose can not be transferred to preformed mannose chains 4. The activities of the two transferases increase in parallel during the period of galactomannan synthesis, such that the M:G ratio in the polymer remains constant 456. UG07564, represented as a single EST in library I, was 100% identical to a recently described guar β-mannan synthase sequence 14. RT-PCR analysis with RNA from guar roots, leaves, stems, cotyledons and different development stages of seeds, revealed that this gene was only expressed in seeds, with maximum transcript accumulation at 35 DAF (Figure 4). In a previous study 14 10 ESTs corresponding to β-mannan synthase were found in a library derived from guar endosperm at 25 DAF. The low frequency of β-mannan synthase ESTs in our work may be due to the fact that our libraries were constructed from whole seed tissues.

It is not known how many isoforms of β-mannan synthase and galactosyl transferase are involved in galactomannan biosynthesis in guar. To highlight additional candidate β-mannan synthase genes, we considered all ESTs which show similarity to glycosyl transferase family 2 members, which are able to transfer GDP-mannose to a range of substrates. By this criterion, four additional ESTs representing putative β-mannan synthase were found, three from library I and one from library II (Table 1).

UDP-galactose-dependent galactosyltransferase belongs to glycosyl transferase family 34 36. Two ESTs corresponding to galactosyltransferase were detected in our EST database; UG05797, from library II, showed 100% identity to a guar galactosyltransferase sequence available in GenBank, whereas UG03477, also from library II, showed 62% similarity to a galactosyltransferase from rice (Oryza sativa) (Table 1). RT-PCR analysis of different guar tissues showed the presence of UG03477 transcripts only in seeds, with maximal accumulation at 35 DAF (Figure 4), consistent with an involvement of this gene in galactomannan biosynthesis.

Hydrolysis – Three enzymes are involved in the hydrolysis of galactomannans during seed germination: β-mannosidase, which hydrolyses the oligomannans released by prior endo β-mannanase activity; β-mannanase, which cleaves the mannan backbone; and α-galactosidase which concomitantly removes the galactose side-chain units 37. Galactomannan hydrolases were the most abundant class of ESTs involved in galactomannan metabolism in the seed EST libraries. Of the five genes annotated as β-mannanase, UG00260 and UG00259 were highly represented in library I, by 10 and 12 ESTs respectively. RT-PCR analysis showed the highest expression level for UG00260 to be at 20–25 DAF (Figure 4). Thus, β-mannanases are actively transcribed during early seed development in guar. Schroder et al (2006) recently demonstrated that a tomato endo-β-mannanase can carry out a transglycosylation in the presence of mannan-derived oligosacchrides 31. This observation may support our findings of high steady-state levels of β-mannanase transcripts in developing seeds.

Of the three β-mannosidase genes detected only in library II, UG00294 was the most highly expressed, being represented by 14 ESTs. RT-PCR confirmed elevated transcript levels for this gene at 30–35 DAF (Figure 4). α-Galactosidase appeared to be less highly expressed; from four identified unigenes, only three ESTs were present in library I and one in library II (Table 1). Early transcriptional activation of galactomannan hydrolyzing enzymes is consistent with seed biology. Upon imbibition, pre-formed enzymes present in the aleurone layer are secreted to mobilize the stored reserves during seed germination 38. Nevertheless, it does raise the question of whether degradative enzymes are ever in proximity with galactomannan during its biosynthesis, such that overall chain length or composition is modified prior to storage.

Seed storage proteins (SSPs) are a set of proteins that accumulate to high levels in seeds during the late stages of development. During germination, SSPs are degraded and the resulting amino acids are utilized by the developing seedlings as a nutritional source 3940. In mature guar seeds, protein accounts for 23–31% of seed dry weight 25.

Five classes of unigenes representing seed-specific proteins were identified in both guar libraries and showed similarities to oleosin, glycinin, conglutin, "seed specific protein," and legumin. All except glycinin did not pass the "secure" assignment threshold of S ≥ 200 (Figure 5A, Table 2). Usually, SSP sequences predominate in cDNA libraries derived from seeds 16. The SSPs were not subtracted from the libraries described here. A single SSP, UG00199, represented the largest class of clones, with 602 ESTs in library II and comprising 3.7% of the total ESTs from both libraries. The predicted translation product of this gene contained 146 amino acids and showed 51% amino acid identity to the delta-conglutin seed storage protein precursor from Lupinus albus. Conglutin delta is related to the 2S super-family of storage proteins 41. 2S storage proteins are widely distributed in dicot seeds, including the economically important genera Brassica 42 and Pisum 43, as well as the model plant Arabidopsis 44. The family is characterized by low molecular weight proteins that contain relatively high levels of cysteine and glutamine.

RT-PCR analysis of guar conglutin transcripts showed maximal expression level in seeds at 35 DAF, and a low but detectable level of expression in cotyledons (Figure 5B). Amplification of conglutin from genomic DNA showed the PCR product to be the same size as the cDNA, indicating that the gene lacks introns (Figure 6C). DNA gel blot analysis of the conglutin, which contains a SacI restriction site in its open reading frame, revealed a low copy number in guar genomic DNA (Figure 6A–B).

Guar (Cyamopsis tetragonoloba) plants, cultivar HES 1401 (now known as Monument, Plant Variety Protection Number: 200400301), were used in this study. This cultivar grows up to 11 dm tall and has the greatest amount of soluble dietary fiber in the seeds 25. Individual plants were grown in 2 gallon pots containing 75% soil (Metro Mix 350, Sun Gro Horticulture, Bellevue, WA) and 25% sand at a temperature of 26°C/22°C (day/night). Plants were fertilized at time of watering using a commercial fertilizer mix (Peters Professional 20-10-20 (N-P-K) General Purpose, The Scotts Company, Marysville, Ohio).

Seeds from guar cultivar HES 1401 were harvested 15, 20, 25, 30, 35, and 40 days after flowering (DAF). Total RNA was extracted from 200–500 mg of ground tissue from the six different seed stages collected from 10 plants using TRI Reagent (Molecular Research Center, Inc. Cincinnati, OH) following the manufacturer's recommendations. Poly A+ RNA was isolated using an Oligotex mRNA Mini Kit (Qiagen, Los Angeles, CA). cDNA was prepared from polyA+ enriched, pooled samples of equivalent amounts of total RNA from each time point. Two cDNA libraries were generated: an "early" seed library (15, 20, and 25 DAF, library I), and a "late" seed library (30, 35, and 40 DAF, library II). The cDNA was directionally ligated into the Uni-Zap XR vector (Stratagene, Los Angeles, CA) and packaged using Gigapack III Gold packaging extracts. Phagemids containing cDNA inserts were in vivo excised from the recombinant Uni-ZAP XR vector using ExAssist helper phage and the E. coli strain XL1-Blue MRF' (Stratagene, Los Angeles, CA). Excised plasmids were plated using SOLR cells (Stratagene, Los Angeles, CA).

Plasmid preparations were made using a Beckman Biomek 2000 robot following standard protocols. Average insert size (1–1.5 kb) was evaluated by agarose gel electrophoresis. cDNA clones were sequenced (single pass, 5'-end sequencing) using an Applied Biosystems 3730 sequencer. Base calling and conversion of binary trace files (.ab1) to human readable text files (.phd.1 and .seq) was completed using Applied Biosystems Sequence Analysis 5.1 program, which essentially is based on Phred 45. Raw sequences were screened and cleaned with NCGR's X Genome Initiative (XGI) program 46, which removed the low quality (N-rich) reads, poly-A and low-complexity regions, vector and primer oligonucleotide sequences. 16,476 quality EST sequences with a minimal length of 50 bp were saved for downstream analysis. These include 8,163 from library I and 8,313 from library II. EST sequences were further clustered and assembled into consensus (unigenes) with TIGR Assembler 47 using its default parameter settings (at least 40 bp overlap with 94% identity). The assembly process generated 7,694 consensus sequences, including 1,695 contigs and 5,999 singletons. BLAST search against the most current version (January 24, 2006) of NCBI non-redundant protein database (NR) was performed with the Personal BLAST Navigator (PLAN) system 48. Annotations, including gene ontology (GO) annotation 39, on each query with the top hit that passed filters e-value ≤ 0.1 and score S' ≥ 40 were further analyzed. The BLASTX search adopted the commonly-used BLOSUM62 scoring matrix. The use of both e-value and score S' 21 filters ensures that only satisfactorily precise (low e-value) and relatively long (high score) alignments are studied 49.

Guar seeds from 25 and 40 DAF were frozen in liquid nitrogen and sectioned to 15 micron by a microtome in a Leica CM1850 cryostat. Sections were stained with toluidine blue (0.05% w/v) to reveal non-neutral cell wall polysaccharides.

One μg of total RNA was used in a first strand synthesis using SuperScript III Reverse Transcriptase (Invitrogen Life Technologies, Chicago, IL) in a 20 μl reaction with oligo-dT primers according to the manufacturer's protocol. Two μl of the first strand reaction was used for PCR with Takara Ex Taq (Fisher Scientific Company, Palatine, IL) according to the manufacturer's protocol. PCR products were analyzed on an agarose gel. The sequences of primers used in RT-PCR experiments are listed in Table 3.

Young leaves from guar cultivar HES 1401 were frozen and ground in liquid nitrogen. Genomic DNA was extracted from 0.5 g ground tissue using Plant DNAZOL Reagent (Invitrogen Life Technologies, Chicago, IL) according to the manufacturer's protocol.

Ten μg of genomic DNA was digested with SacI, SacI/EcoRI or Hind III and loaded on a 0.8% agarose gel. The gel was capillary blotted to nylon Hybond-N+ membrane (Amersham Pharmacia Biotech, Pittsburgh, PA). The blot was hybridized and signal detected using ECL direct nucleic acid labelling and detection systems (Amersham Pharmacia) according to the manufacturer's protocol. Probe was synthesized by PCR using primers complementary to the conglutin gene listed in Table 3.