cDNA libraries and EST sequencing
Twelve cDNA libraries were constructed from various tissues, organs, and cell lines, including four blue catfish libraries and eight channel catfish libraries (Table ). More than 600,000 sequencing reactions were attempted to sequence a total of 307,296 cDNA clones from both ends. A total of 438,321 ESTs were generated from this project, of which 128,711 sequences were from blue catfish and 309,610 were from channel catfish (Table ). Of these EST sequences, 219,831 were sequenced from the 5' end of the transcripts, and 218,490 were sequenced from the 3' end. A total of 194,136 clones had paired reads from both 5' and 3' ends of the same transcripts. The lengths of the ESTs range from 100 to 877 bp, with an average length of 576 bp and a median length of 655 bp (Figure ). Addition of these sequences to the 10,764 blue catfish ESTs and 44,767 channel catfish ESTs in GenBank before the start of this project increased the number of catfish ESTs to almost a half million sequences (139,475 blue catfish ESTs and 354,377 channel catfish ESTs; Table ). The ESTs from blue catfish and channel catfish have been deposited in GenBank under accession numbers [GenBank:FC996013-FC999999, FD000001-FD380635 and GH640296-GH693994].
cDNA library information and sequencing summary
Length distribution of Joint Genome Institute EST sequences.
All existing catfish EST sequences were used to produce three assemblies: blue catfish ESTs; channel catfish ESTs; and blue catfish and channel catfish ESTs for inter-specific analysis. Assembly of 139,475 blue catfish ESTs resulted in 54,815 unique sequences (22,009 contigs and 32,806 singletons) whereas assembly of 354,377 channel catfish ESTs resulted in 70,717 unique EST sequences (28,941 contigs and 41,776 singletons) (Table ). Details of the catfish EST assembly are available online [36
In order to identify inter-specific SNPs, we also conducted the assembly of all available 493,852 ESTs from blue catfish and channel catfish. This assembly allowed the formation of 45,306 contigs (66,272 singletons), from which potential inter-specific SNPs could be identified. The number of inter-specific contigs was significantly larger than that from either species, potentially due to the formation of new contigs of related transcripts that were singletons in either species (see Discussion). A majority of contigs contained only two (43%) or three (13%) sequences (Figure ), and average contig depth was nine sequences. With the ESTs being sequenced mostly from normalized libraries, the vast majority of contigs had 50 or fewer sequences. However, some extremely large contigs were found. The largest contig, containing 7,208 sequences, putatively identified as apolipoprotein, was repeatedly sequenced from all libraries, and was prevalent in the pre-existing non-normalized libraries in GenBank. As previously reported [37
], contig depth is one of the two most important factors affecting EST-derived SNP qualities. Therefore, the information on contig depth is highly useful.
To assess the similarity between blue catfish and channel catfish sequences, the consensus sequences from each species were compared to each other using BLASTN at a stringency of 1E-10. Of blue catfish and channel catfish sequences with a minimal 200-bp matching region, sequence similarity ranged from 77% to 100%, with an average similarity of 95%. Over 50% of blue catfish and channel catfish homologous sequences have similarity levels over 95% (Figure ).
Distribution of sequence similarity between blue catfish and channel catfish sequences.
Gene identification and annotation
Putative gene identification was conducted either by ab initio identification of open reading frames (ORFs) or by BLASTX similarity search of public protein databases. Of the 111,578 total unique catfish sequences (total catfish EST assembly), ORFs were detected from 83,198 (75%) unique sequences, with an average ORF length of 450 bp (minimum = 51 bp, maximum = 14,674 bp; Figure ), and the remaining 28,380 sequences (25%) contained no ORFs (Figure ). These ORF-less ESTs were likely ESTs sequenced within the untranslated regions (UTRs) or within intron-retaining cDNAs.
Open reading frame (ORF) length distribution from unique sequences of the all catfish assembly.
Figure 5 Analysis of open reading frames (ORFs). (a) Percentage of ORFs among unique sequences from the all catfish EST assembly; (b) Percentage of ORF greater than 100 bp among unique sequences from the all catfish EST assembly; (c) Percentage of ORFs equal to (more ...)
There was a positive correlation between the length of ORF and BLASTX match. Of the identified ORFs, 91% had a length of more than 100 bp. Within these ORFs, 53% had significant (1E-10) BLASTX matches (Figures ). However, only 6% of the ORFs with less than 100 bp had significant BLASTX matches (Figure ).
A total of 41,311 (37%) unique sequences had significant BLASTX matches within the nr database, and 34,860 (31%) had significant BLASTX matches within the Uniprot database (Table ). Over 98% of unique sequences with significant matches were identified with ORFs, which indicated the reliability of ORF searching. After examination of putative protein identities from the BLASTX searches, homologous sequences were identified from the catfish ESTs. Of the 41,311 sequences with BLASTX hits, 22,642 (approximately 55%) and 17,948 (approximately 43%) unique proteins were identified through searches against the nr and the Uniprot protein databases after removing the redundant protein hits, respectively.
Summary of BLASTX search analysis of catfish ESTs
Assessment of the sequenced catfish transcriptome
In order to assess the level to which the catfish transcriptome has been captured, the unique catfish sequences (111,578) were also searched against the NCBI Refseq and Ensembl databases. A number of significant hits were identified within zebrafish, medaka, Tetraodon, human, mouse, and chicken reference protein databases (Table ). After removal of the redundant protein hits, 14,988 - 11,059 unique reference proteins were identified within zebrafish, medaka, Tetraodon, human, mouse, and chicken databases respectively (Table ). The unique catfish sequences had hits to 54% to 57% of the unique proteins of zebrafish, medaka, and green-spotted pufferfish. To allow comparison of catfish unique protein coverage with that expected between species with complete genome sequences, all Tetraodon Ensembl proteins were searched against the medaka Ensembl protein database. A total of 22,150 Tetraodon proteins have significant hits to 15,054 (61% of total unique) medaka Ensembl proteins with a cutoff E-value of 1E-10. Similarly, zebrafish Refseq proteins were searched against the human Refseq protein database. In this case, 24,971 zebrafish proteins have significant hits to 13,789 (36% of total unique) human proteins with a cutoff E-value of 1E-10. Taken together, these numbers provide strong evidence that this project has captured a large majority of the catfish transcriptome.
A total of 14,776 cumulative unique genes were identified from catfish based on BLASTX searches against the Refseq/Ensembl database (Table ), including 8,075 genes identified from both blue catfish and channel catfish, 1,881 unique genes from blue catfish and 4,820 unique genes from channel catfish (Figure ). As expected based on sequencing coverage depths, significantly more unique genes were identified from channel catfish than blue catfish.
Comparison of shared and unique gene identities of channel catfish and blue catfish from a total of 14,776 unique genes.
To assess the evolutionary conservation of the identified unique genes, the number of hits to unique genes in each species of zebrafish, medaka, Tetraodon, human, mouse, and chicken were compared. A total of 8,592 (58% of total number of unique catfish genes) putative known unique genes were found in all six species: 11,303 (76%) were found in all three fish species; and 14,515 (98%) were found in at least one of the three fish species (Figure ), indicating a high level of conservation of gene content among catfish and other teleost fish species.
Conservation of catfish gene identities with other species. Number of catfish homologous genes identified from other species using BLASTX searches.
Prediction of full-length cDNAs
The catfish EST sequences provide a platform for the identification and characterization of full-length cDNA clones without having to use expensive and labor-intensive primer walking sequencing. In the context of this work, full-length cDNA inserts were defined as a cDNA from a single clone with the start codon and poly (A) tail contained within the clone. A total of 10,037 channel catfish and 7,382 blue catfish putative full-length cDNAs were identified from the assembly with a cutoff E-value of 1E-5. A well characterized full-length cDNA set from catfish will be crucial in ongoing studies of teleost gene duplication and gene family structure, as well as aiding in annotation of the catfish whole genome sequence. Current efforts are focused, therefore, on characterization and re-sequencing of these full-length cDNAs.
Microsatellite and SNP marker identification
A total of 20,757 microsatellites were initially identified from 15,082 unique sequences, including di-, tri-, tetra-, penta- and hexa-nucleotide repeats (Table ). After removing the microsatellites without enough flanking sequence for primer design, 13,375 unique sequences with microsatellites had sufficient flanking sequences (50 bp) on both sides of the microsatellites to design primers for genotyping. Our previous research indicated that over 72% of EST-derived microsatellites were polymorphic in one resource family [12
], suggesting the potential utility of these microsatellites.
Summary of microsatellite marker identification from catfish ESTs
A total of 48,702 putative SNPs and 14,803 putative insertions/deletions (indels) were identified from the blue catfish EST dataset assembly; 102,252 putative SNPs and 41,660 putative indels were identified from channel catfish EST dataset assembly (Table ). These putative SNPs indicated an SNP rate of 3.2 SNPs per kilobase of transcribed sequences in blue catfish, and 4.1 SNPs per kilobase of transcribed sequences in channel catfish. Obviously, such SNP rates were calculated from the total consensus sequence length and, therefore, the deeper the EST sequencing was, the greater the possibility for the identification of a SNP within the consensus sequences.
Summary of SNP identification from the catfish ESTs
Over 303,000 putative SNPs and 100,000 putative indels were identified from the all catfish EST assembly results (Table ). EST-derived SNPs are often prone to sequencing errors. Therefore, the putative SNPs were subjected to filtering using only those with contig sizes of at least four sequences and the minor allele presence of at least two sequences in the contigs, and indels were not used for further analysis [37
]. These parameters were previously shown to select markers with 70% success rate for genotyping. After filtering, 3,891 and 15,827 SNPs were identified from the blue catfish and channel catfish EST dataset assemblies, respectively. A subset of 48,594 filtered SNPs were obtained from the all catfish EST assembly; these SNPs included 32,235 transitions and 16,359 transversions (Table ). The filtered SNP frequency in the transcribed sequences was 0.25 SNP in blue catfish, 0.64 SNP in channel catfish, and 1.6 SNP in the all catfish assembly per kilobase. A total of 19,398 filtered insertions and deletions (Indels) were discovered, that is, 0.64 indels per kilobase of the transcribed sequences. Of the 48,594 SNPs, over 90% were identified from the contigs containing 5 or more sequences (Table ).
Quality assessment of the filtered putative SNPs identified from the catfish ESTs based on the number of sequences per contig and the sequence frequencies of the minor alleles
The assessment of the rates of inter-specific SNPs and intra-specific SNPs may have practical applications. We therefore assessed these SNP rates using the EST data. First, SNPs were identified from contigs containing at least four sequences with at least two sequences from either channel catfish or blue catfish in the all catfish EST assembly. Inter-specific SNPs were defined as those that have sequence variations between blue catfish and channel catfish, but no sequence variations within blue catfish or within channel catfish; similarly, SNPs were identified within blue catfish but not within channel catfish or vice versa; and SNPs were identified within both channel catfish and blue catfish at the same SNP positions (Figure ). Of the 48,594 filtered SNPs, 42,080 were identified from contigs comprising both channel catfish and blue catfish ESTs, and 6,514 were identified from contigs composed of ESTs from either channel catfish or blue catfish, including 5,396 from channel catfish contigs and 1,118 were identified from blue catfish contigs. Of the contigs containing ESTs from both blue catfish and channel catfish, the estimation of percentage of inter- and intra-specific SNPs was conducted based on the identification of SNPs from 1,000 randomly selected contigs. Of the 48,594 filtered SNPs identified from the all catfish assembly, over 18,000 (39%) were inter-specific SNPs; with 523 (1%) intra-specific SNPs for blue catfish, 2,352(5%) intra-specific SNPs for channel catfish, and 3,790 (8%) intra-specific SNPs for both channel catfish and blue catfish. However, approximately 17,000 SNPs could not be determined because overall the SNPs qualified as SNPs with at least two minor allele sequences, but only one of the minor allele sequences was from one of the two species of blue catfish or channel catfish (Table ). Additionally, the number of inter-specific SNPs may be overestimated, due to failure to capture minor allele sequences from one or both species in the current EST data. However, the sequence differences between species should be greater than those within species. Although a large number of filtered inter-specific SNPs were identified (18,000 out of 48,000 total filtered SNPs), they were identified from a relatively small number of contigs. The 18,000 filtered inter-specific SNPs were identified from approximately 2,800 contigs, with an average of 6.6 SNPs per contig.
Figure 8 Categorization of four different types of SNPs from the all catfish EST assembly and examples of SNPs whose categories could not be determined. (a-d) Types of SNPs from the all catfish EST assembly that can be identified from the all catfish EST assembly. (more ...)
Estimation of proportions of inter-specific and intra-specific SNPs from the set of filtered SNPs identified from the interspecific all catfish EST assembly