We constructed a full-length enriched cDNA library from male strobili of C. japonica at various stages of development. The mean length of the insert DNA derived from 91 randomly selected clones was approximately 1.6 kbp. We picked up 19,968 clones (clone IDs: CMFL001_A1 through CMFL052_N24) at random and sequenced them from their 5' and 3' ends. In total, 39,936 sequences were obtained and processed for removal of vector sequences, low-quality data and contamination by genomic DNA from E. coli. We retained 36,011 ESTs, which encompassed 18,843 sequences of 5' ends and 17,168 sequences of 3' ends derived from 19,437 clones. Each EST consisted of at least 100 contiguous nucleotides with a PHRED score greater than 20 (Table 1; Additional file 1). The average sequenced length of the ESTs was 517 bp. All sequences have been deposited in the dbEST division of DDBJ/EMBL/GenBank. The accession number of each EST is shown in Additional file 1.

We analyzed the full-length cDNA clones that included sites for the initiation of transcription of mRNA. An additional G at the 5' end is observed in many full-length cDNAs that are obtained by the biotinylated CAP trapper method, which we used in this study 20. Among the 5' sequence in our cDNA library derived from C. japonica male strobili, 16,977 (90%) had G at the 5' end (Table 2). The high percentage of cDNA clones with G at the 5' terminus suggests that the relative level of full-length cDNA clones in our library was high. The second nucleotide showed a strong bias towards purines (89% A or G; Table 2). This observation is consistent with the fact that transcription usually starts at a purine in Arabidopsis and rice 21. Subsequent positions did not show such an extreme bias towards purines. Our results suggest that most full-length cDNAs had one additional G at their 5' termini.

To confirm the addition of G at the 5' end of full-length cDNA clones, we compared the 5'-terminal sequences of cDNA clones that encoded Cry j 2 with genomic sequences, including the promoter and coding region of Cry j 2, that we determined in a previous study 22. We found seven clones that encoded Cry j 2 among the 5'-end sequences of 18,843 clones. All seven of them were 5'-extended cDNAs, as compared with three Cry j 2 cDNAs that we generated previously 22. Five of the seven had one G, one clone had two Gs and one clone had TG at its 5' terminus; these nucleotides were not present in the corresponding region of Cry j 2 genes (data not shown). The comparison between these seven cDNAs and Cry j 2 genes indicated that full-length cDNAs had one or two additional nucleotides at their 5' termini, with most clones having a single G.

We performed a BLASTX comparison of amino acids encoded by 5'-terminal sequences of 18,843 cDNA clones and proteins from Arabidopsis. Our analysis revealed that 9,208 ESTs (48.7%) exhibited strong homology to Arabidopsis proteins (E-value < 1e-20). Among these ESTs, the starting positions in the alignment of 7,480 clones (81.2%) were upstream of the initiation codon of the corresponding protein. We also performed a BLASTN comparison with nucleotide sequences of protein-coding regions in Arabidopsis. We found that 2,131 ESTs (11.5%) derived from 5'-terminal sequences of cDNAs exhibited strong homology (E-value < 1e-10) to protein-coding regions in the Arabidopsis genome. The starting positions in the alignment of 1,755 ESTs (82.4%) were upstream of the corresponding coding region of Arabidopsis. These results suggest that most of our cDNA clones are full-length cDNAs.

The assembly of ESTs can be expected to generate an overestimate of the actual number of genes represented since failure of ESTs to assemble can result from alternate splicing, differences in usage of polyadenylation sites, sequence polymorphism, and sequencing errors. Levels of redundancy after EST assembly have been estimated to range from 20% to 22% in previous studies of EST collections 2324. To reduce redundancy, we compared all consensus sequences (7,686 tentative contigs and 15,972 singletons) using BLASTN after assembly with PHRAP and then we grouped both 5'- and 3'-end sequences derived from the same respective clones together. Finally, our analysis indicated that ESTs derived from 19,437 clones could be grouped into 10,463 clusters as unique transcripts (Table 1; Additional file 1). The largest cluster contained 35 clones, but only 26 transcripts (0.25%) corresponded to more than 10 clones. We found that 6,320 of the 10,463 transcripts (60.4%) corresponded to only one clone and 2,078 clusters (19.9%) corresponded to two clones. The normalization step appeared to reduce the redundancy of our ESTs. Mitochondrial and chloroplast RNA sequences were not filtered, but they contributed one and three clones to the data set of 19,437 cDNA clones, respectively. No contamination by ribosomal RNA was detected in the data set. Figure 1 shows the functional classification of the putative proteins encoded by our ESTs, which was based on assignments in the COG database 25. Of the putative proteins derived from individual transcripts, 7,339 (70.1%) were assigned to 26 putative functional categories by BLASTX (E-value < 1e-5; Additional file 2).

We found that 8,059 (77.0%) of the 10,463 transcripts encoded peptides that were significantly similar to those in the UniProt database, 7,840 (74.9%) were similar to proteins in Arabidopsis, and 7,455 (71.3%) were similar to proteins in japonica rice at a BLASTX E-value of 1e-5 (Figure 2). When we compared our ESTs with sequences of ESTs from Pinus, Picea and Populus by TBLASTX, we found that 8,433 (80.6%), 8,400 (80.3%) and 7,763 (74.2%) transcripts included similar ESTs, respectively, at E-values of 1e-5. A larger overlap was found in the case of coniferous species (Figure 2). We also compared the 10,463 transcripts examined in the present study with ESTs that had been previously isolated from C. japonica, which included 10,328 ESTs derived from cambium and surrounding tissues, 3,677 ESTs from pollen, 3,222 ESTs from male strobili and 1,889 ESTs from other organs 1926. Slightly more than half (5,792; 55.4%) of our transcripts did not exhibit any similarity (BLASTN Evalue < 1e-10) to ESTs that had been previously isolated from C. japonica. Thus, the ESTs described in the present report contain novel information about protein-coding regions of the C. japonica genome.

We analyzed conserved domains using Pfam 27 as a database to predict the function of products of C. japonica transcripts. Overall, we found that 5,304 (50.7%) of the 10,463 transcripts encoded proteins similar to members of 1,664 Pfam protein families (E-value < 1e-10). Among these Pfam families, 147 families are denoted "DUF, Domain of Unknown Function" and 26 families are denoted "UPF, Uncharacterized Protein Family". In total, products of 4,965 (47.5%) of the transcripts from C. japonica male strobili were similar to members of 1,481 Pfam families when DUFs and UPFs were excluded. We found that 4,850 (46.3%) transcripts belonged to only one Pfam family, 433 (4.1%) transcripts belonged to two families, 17 transcripts (0.16%) belonged to three families, and two transcripts (0.02%) belonged to four families (Additional file 2). The most abundant protein families in male strobili of C. japonica corresponded to those most strongly represented in the genome of Arabidopsis (Table 3). Similar earlier analyses indicated that these domains occur frequently in sugarcane and white spruce also 2328.

We compared our deduced proteins with stamen and male gametophyte-specific proteins in Arabidopsis. Referring to previous studies of Arabidopsis, we identified amino acid sequences encoded by 1,145 stamen-specific transcripts and by 1,274 male gametophyte-specific transcripts in the Arabidopsis database 2930. We found that 754 (65.9%) stamen-specific transcripts and 617 (48.4%) male gametophyte-specific transcripts of Arabidopsis resembled 509 and 429 transcripts, respectively, from male strobili of C. japonica, at a TBLASTX E-value of 1e-10 (Additional files 3 and 4). Table 4 shows the Pfam domains that are found in stamen- and male gametophyte-specific genes of Arabidopsis and the numbers of transcripts that correspond to the families of genes in male strobili of C. japonica. Our results suggest that homologs of stamen- and pollen-specific transcripts of Arabidopsis are expressed in the male strobili of C. japonica.

We compared the peptides encoded by all the ESTs with known plant allergens in Allergome 31. Table 5 shows a list of pollen allergens that are similar to the deduced proteins encoded by ESTs derived from C. japonica male strobili. Proteins encoded by 180 transcripts (1.7%) exhibited partial sequence homology to plant allergens, which included known allergens of C. japonica. This result suggests the possibility that unidentified allergens in C. japonica might exist in our EST collection.

Transcription factors play important roles in the formation of floral tissues throughout organogenesis. To identify transcription factors, we first identified 37 Pfam domains that encode transcription factors by comparing data in the AtTFDB (Arabidopsis thaliana transcription factor database) 32 with the domain annotations of Arabidopsis proteins. As a result of a search for sequence similarity between ESTs from C. japonica male strobili and the 37 identified Pfam domains, we identified a total of 207 (2.0%) unique transcripts that encoded putative transcription factors, which could be assigned to 29 protein families (Table 6). Transcripts whose products resembled the C3HC4 zinc finger domain, MYB-like transcription factor, and AP2 were abundant in the male strobili of C. japonica.

Among the putative transcription factors, ten transcripts encoded MADS-box proteins (Table 6). In addition, we found another two transcripts whose products were similar to MADS-box proteins with E-values < 1e-8. We determined the entire sequence of representative cDNA derived from each transcript (accession numbers AB359027 through AB359038). Figure 3 shows the phylogenetic tree constructed from an alignment of the sequences of MADS, I-, and K-domains. One cDNA clone (CMFL009_N22) did not encode a K-domain and it was classified as the cDNA of a type I MADS-box gene. Six of twelve C. japonica MADS-box genes were classified as DEF/GLO/GGM13-like genes and two genes were categorized as TM8-like genes. The other three genes were categorized as AG-, AGL6-, and TM3-like genes, respectively (Figure 3).

Male strobili of C. japonica were collected at seven-days intervals from mid August to mid November of 2003 and 2004. All the samples were immediately frozen in liquid nitrogen and stored at -80°C until use. RNA was extracted from each set of strobili, and a total RNA mixture derived from all the samples was used for construction of a full-length enriched cDNA library.

Total RNA was isolated by a previously described method 52 with slight modifications, using the SV Total RNA Isolating System (Promega, Madison WI). Frozen male strobili were powdered with Multi-beads shocker^® (Yasui Kikai, Osaka, Japan). Then we added 7.5 ml of a solution of 100 mM Tris-HCl (pH 9.5), 20 mM EDTA, 1.4 M NaCl, 2% (w/v) polyvinylpyrrolidinone, 5% (w/v) β-mercaptoethanol and 0.5 mg/ml spermidine per gram of powdered male strobili with through mixing. After incubation of the mixture at 65°C for 5 min, RNA was extracted twice with an equal volume of a mixture of chloroform and isoamyl alcohol (24:1, v/v) and centrifugation at 15,000 × g for 20 min. One-fourth volume of 10 M LiCl was added to the final aqueous phase. Total RNA was allowed to precipitate for 2 h and collected by centrifugation at 15,000 × g for 30 min. The pelleted RNA was dissolved in SV RNA Lysis Buffer (Promega) and purified with the SV Total RNA Isolating System according to the protocol provided by the manufacturer. Poly(A)⁺ RNA was prepared with a μMACS™ mRNA Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the protocol provided by the manufacturer.

A full-length enriched cDNA library was generated from poly(A)⁺ RNA by the biotinylated CAP trapper method using trehalose-thermoactivated reverse transcriptase as described previously 3653. The resultant cDNAs were normalized, inserted into λ-FLC-III vectors and packaged 5455. The packaged DNA produced approximately 1.1 × 10⁶ plaques and the original phage library was amplified once. Part of the one-round-amplified phage library was excised in vivo and resultant plasmids were used to transform E. coli DH10B. Colonies of bacteria were picked with picking machines (Q-bot; Genetix, New Milton, UK) and transferred to 384-microwell plates. DNA templates corresponding to cDNA inserts were prepared from glycerol stocks by the rolling circle amplification (RCA) method with a TempliPhi™ DNA Amplification Kit (GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA). Products of RCA were sequenced with BigDye^® Terminator v3.1 Cycle Sequencing Kits (Applied Biosystems, Foster City, CA) and automatic sequencers (ABI 3700; Applied Biosystems). The M13-21 primer (5'-TGT AAA ACG ACG GCC AGT-3') was used for forward sequencing and the 1233 primer (5'-AGC GGA TAA CAA TTT CAC ACA GGA-3') was used for reverse sequencing. Sequences were quality-trimmed by reference to the high-quality (PHRED 20 or better) contiguous region, as determined with PHRED software 56 and then vector and poly(A) regions were removed. Sequences of less than 100 nucleotides after trimming were discarded. Contaminating regions of E. coli genome were identified with BLASTN 57, and ESTs with an E-value < 1e-10 were excluded. The remaining sequences were used for further analysis.

We searched for the nucleotide and amino acid sequences in TAIR (The Arabidopsis Information Resources; genome release version 6.0) 58 homologous to nucleotide and predicted amino acid sequences encoded by the 5'-end sequences of ESTs using the BLASTN and BLASTX program. We also searched for sequence homologies between the 5'-end sequences of ESTs and the cDNA and promoter region of Cry j 2 using the BLASTN program. Assessment of possible contamination by chloroplast and mitochondrial genomes and by ribosomal RNAs was performed with the BLASTN program (E-value < 1e-10). We used the chloroplast and mitochondrial genomes of Arabidopsis thaliana and genes for ribosomal RNAs from A. thaliana, Populus species, Platanus occidentalis and Zea mays as query sequences.

Related cDNA sequences from both 5' and 3' ends were grouped as contig sequences with the PHRAP program according to the following criteria: -penalty -5; -minmatch 300; -minscore 300; and -trim_qual 20 59. After grouping ESTs as contigs, we used the BLASTN program to compare contig sequences. When contig sequences overlapped by more than 200 nucleotides and were more than 99% homologous in the overlapping regions and, in addition, when the overlap began within 5 bp from the ends of the contig sequences, we grouped the contig sequences together. We performed BLASTN searches separately for the EST sequences of between 100 and 200 bp and the resultant contig sequences. When an EST sequence and a contig were more than 98% homologous and the length of non-overlapping sequence in the EST sequence was less than 10 bp and, in addition, the overlap began within 5 bp from the end of the EST sequence or the contig, we grouped the EST sequence and the contig together. Finally both 5'- and 3'-end sequences derived from the same clone were grouped together.

We categorized ESTs on the basis of the putative functions of encoded products using a database of clusters of orthologous groups from seven eukaryotic genomes (COG) 25. We used databases of proteins from more than three species (KOGs), proteins from two species (TWOGs) and lineage-specific expansion groups (LSEs). We compared the results of BLASTX analysis of each sequence in each cluster and adopted the functional category with the highest score.

We performed similarity searches with the BLASTX and TBLASTX programs. We used RefSeq of Arabidopsis from the National Center for Biotechnology Information 60, the database of the Knowledge-based Oryza Molecular Biological Encyclopedia 61 and UniProtKB/TrEMBL Release 33.3 62 as protein databases, and we used Pinus Gene Index Release 6.0, Spruce Gene Index Release 2.0, and Poplar Gene Index Release 3.0 from the TIGR Gene Indices 63 as EST databases. Sequences of stamen- and pollen-specific transcripts of Arabidopsis were obtained from previous studies 2930. Similarities to ESTs that had previously been derived from C. japonica were determined with the BLASTN program. Similarities to known plant allergens were determined by BLASTX comparison with proteins in Allergome database http://www.allergome.org31. We annotated protein families using a list of Pfam domain sequence (Pfam-A.fasta; release 21.0) 27 using the BLASTX program and an E-value < 1e-10. We obtained domain annotations of Arabidopsis thaliana proteins from the TAIR website 58 and selected Pfam families with an E-value < 1e-10.

To reconstruct a phylogenetic tree of MADS-box genes, we obtained amino acid sequences from EMBL/DDBJ/GenBank DNA databases and aligned them using the Clustal W program 64. The alignments of sequences of MADS, intervening (I-) and keratin-like (K-) domains were edited manually using the MacClade program 65. Maximum likelihood distances were calculated with the NJdist and ProtML programs in MOLPHY 66. A neighbor-joining (NJ) tree was obtained with NJdist, and was based on ML distances in the JTT model 67. This tree was used as the starting tree for a local rearrangement search with the ProtML program 66. The local bootstrap probability of each branch was estimated by the resampling-of-estimated-log-likelihood (RELL) method 68.