Towards the ictalurid catfish transcriptome: generation and analysis of 31,215 catfish ESTs

cDNA library construction and sequencing of catfish ESTs

EST sequencing was conducted in two phases. In phase I, 200–300 clones were sequenced from each library to provide a list of the most abundantly expressed genes. In phase II, the most abundantly expressed genes (Supplemental Table 1) were subtracted from the clones to be sequenced by screening with overgo probes, to provide a higher gene discovery rate under a restricted budget. Overgo probes were designed for 200 genes, and the probes were used for colony lifting hybridization. Subsequently, only negative clones were picked for phase II sequencing. The number of ESTs generated from each library is given in Table 1. A total of 20,451 ESTs were successfully sequenced from channel catfish, and 10,764 ESTs were sequenced from blue catfish. These ESTs have been submitted to NCBI dbEST [GenBank: BM438128–BM439194, BQ096608–BQ097456, CF261473–CF266494, CF970744–CF972299, CK401558–CK426402, and EE993123–EE993655]. Furthermore, a database, ESTIMA: Catfish, was established for free public access [29]. These ESTs represent a significant fraction of the EST resources from channel catfish and the sole publicly available transcripts from blue catfish.

Sequence assembly

Sequence annotation

The putative identities of the sequenced ESTs were determined using BLASTX searches against the non-redundant (nr) database in GenBank. Of the 20,451 channel catfish ESTs, 10,859 (53%) had significant hits (cutoff E-value of e^-5), while the remaining 9,592 ESTs (47%) had no significant similarity to any sequences contained in GenBank (Table 2). Similarly, of the 10,764 blue catfish ESTs, 5,456 (50.7%) had significant hits (cutoff E-value of e^-5), while the remaining 5,308 ESTs (49.3%) had no significant similarity to any sequences contained in the database. While a significant fraction of ESTs could not be identified by similarity searches, our results are comparable to other EST work in fishes. The unidentified transcripts are still valuable sources of microsatellite markers, and can be furthered sequenced if determined to be important in QTL analysis or expression profiling with microarrays. Additionally, many of these currently unknown transcripts will likely be identified when they cluster with additional transcripts produced in the future.

Assessment of the sequenced catfish transcriptome

To link these catfish EST resources to a comparative genome analysis framework, we conducted systematic TBLASTN searches on all existing catfish ESTs using Tetraodon chromosome-linked proteins as queries. The TBLASTN search parameters were set to select the top catfish hit and used a relatively more stringent cutoff E-value of e^-10. Approximately 50% of annotated Tetraodon genes had a significant hit against catfish ESTs (Table 3), providing a rough assessment of the percentage of the catfish transcriptome now sequenced. However, BLAST-based comparisons between sequences of the two species have several shortcomings. First, rapid intraspecific diversification of gene families within catfish and Tetraodon has obscured gene homologies between the species. Second, short and/or divergent protein sequences would be excluded with the stringent parameters used. Altogether, 6,720 unique catfish ESTs were returned as the top hit of one or more Tetraodon proteins (Fig. 1). The factors mentioned above, especially gene family diversification, could also be responsible for the modest number of catfish hits. The majority of these catfish ESTs (3,929) were hit by a single Tetraodon query. However, a sizeable proportion (22%) were hit by three or more Tetraodon queries (Table 3, Fig. 1). A survey of the catfish ESTs hit by 20 or more Tetraodon queries revealed that these represented large gene families often functioning in developmental processes. Examples of the families hitting single catfish ESTs included the protocadherin clusters, notch proteins, zinc fingers, netrin family, and Hox proteins. Analysis of the chromosomal origins of these repetitive Tetraodon queries indicates that many are clustered tightly together and have likely resulted from rapid tandem gene duplication in their local environ [31, 32]. High sequence conservation between members of these gene families may obscure their relationships with homologous families in catfish. Alternatively, transcripts representing gene family members in catfish may not have yet been sequenced.

Total number of Tetraodon protein queries	27,918
Tetraodon genes that had hit(s) to catfish ESTs	14,512
Number of catfish ESTs hit by a single Tetraodon gene	3,929
Number of catfish ESTs hit by a two Tetraodon genes	1,329
Number of catfish ESTs hit by a multiple Tetraodon genes	1,462

Potential for comparative genome analysis and directed gene mapping

Two species system for identification of ancient and recent gene duplications

Gene duplication is a widespread phenomenon in vertebrate species and a particularly important trait of teleost fish. It has been proposed that a whole-genome duplication event occurred in the teleost lineage after its split from the tetrapod lineage, but that only a subset of the duplicated genes has been retained [35]. ESTs are a valuable tool for the identification of duplicated genes in species with and without a sequenced genome. For species with completed genome sequences, a large EST collection is invaluable for gene annotation in duplicated regions and facilitates the study of sub-functionalization among duplicated gene copies [36]. In species like catfish, where whole-genome sequencing is yet to be initiated, ESTs are important early indicators of gene copy numbers. However, comparisons of highly similar transcripts often do not allow researchers to differentiate between allelic variants and gene duplications. Using a two species system of EST sequencing and analysis, such as in channel catfish and blue catfish, can help to distinguish between these two possibilities. We surveyed the catfish EST resources for gene duplication events, applying the rationale that allelic variation within the same species should be smaller than the variation present between orthologues from different species [37]. Using this rationale, two highly similar channel catfish sequences would be considered paralogues if one of them is more closely related to a transcript from blue catfish than related to the other transcript from channel catfish. Likely instances of catfish gene duplication that included ESTs from both species were identified by BLASTN searches and then reciprocal BLASTX searches carried out. Most identified "duplications" were members of large, previously identified gene families that were the result of ancient gene duplications, i.e. similar members are present in other species such as Danio rerio. More informative were cases where all selected catfish ESTs were highly similar to the same BLASTX hit and/or gene copy number could not be predicted based on BLAST results, indicators of more recent gene duplication within catfish. Examples of these cases, where allelic variants could not be distinguished from gene paralogues based on the data from a single species alone, were subjected to phylogenetic analysis (Fig. 2). The putative blue catfish orthologues provided the context necessary to differentiate between the highly similar channel catfish transcripts. The ability to utilize the ESTs from the two closely related catfish species for analysis of local gene and genome-based duplications was one of the reasons for continuing EST sequencing efforts in catfish conducted by the Joint Genome Institute (see below).

Subtraction probes for normalization of cDNA libraries

Sequencing a large number of cDNA libraries widened the range of the catfish transcriptome sequenced while providing information concerning the most abundantly expressed genes in a variety of tissue types. This information is critical to ensure high numbers of unique transcripts can be obtained when sequencing a library to greater depths. To produce a list of most abundantly expressed genes for further subtraction, we conducted cluster analysis of all catfish ESTs in the dbEST database of NCBI. Clusters were sorted by size and those containing 2 or more transcripts per 10,000 sequences were selected as subtraction drivers for use during the construction of normalized/subtracted cDNA libraries to be used for a large-scale EST project (Supplemental Table 2). Through the Community Sequencing Program, a project for sequencing 300,000 clones of catfish ESTs was recently approved by the Joint Genome Institute (JGI) of the Department of Energy (DOE). Subtraction of highly abundant genes based on information gained through the current project should markedly increase the number of unique transcripts obtained by JGI sequencing, and initial sequencing and quality control determination by JGI of the subtracted cDNA libraries we produced using this strategy confirmed this assessment.

Tissue samples and RNA isolation

All procedures involving the handling and treatment of fish used during this study were approved by the Auburn University Institutional Animal Care and Use Committee (AU-IACUC) prior to initiation. Channel and blue catfish were raised in troughs in the hatchery of the Auburn University Fish Genetics Hatchery for four weeks before harvesting of tissues. To create resource cDNA libraries containing a full complement of gene transcripts, including those expressed after infection, both healthy and infected catfish were used. Channel catfish and blue catfish were challenged with Edwardsiella ictaluri using procedures adapted from Dunham et al. [42]. Fish were divided into 2 groups, the non-challenged controls and the fish for challenge (N = 240). The fingerlings used for disease challenge were placed into a 150 L tank containing 1.1 × 10⁶ E. ictaluri cells/ml for 1 h. The challenged fish were then removed and stocked into a 1000 L tank. At time of sampling, fish were euthanized with MS-222 at 300 mg/L before dissection. Tissue samples were collected from 15 control and 5 infected fish each at 24 h, 3 d, and 7 d during the challenge, pooled, quick-frozen in liquid nitrogen, and stored at -80°C until RNA extraction. The following tissues were collected: channel catfish gill, head kidney, trunk kidney, intestine, liver, skeletal muscle myomere, olfactory organ, ovary, pituitary, spleen, stomach, and testes; blue catfish head kidney, heart, intestine, liver and spleen. Equal tissue weights of all the control and infected pools for each tissue within a species were combined, ground to a fine powder with mortar and pestle in the presence of liquid nitrogen and thoroughly mixed. A fraction of the tissue samples was used for RNA isolation. Total RNA was isolated following the guanidium thiocyanate method [43] using the Trizol reagent (Invitrogen, Carlsbad, CA) following manufacturer's instructions. Poly(A)⁺ RNA was purified from total cellular RNA using the Poly(A)⁺ Pure kit (Ambion, Austin, TX) according to the manufacturer's instructions.

Library construction

Initial sequencing of four catfish cDNA libraries, channel catfish brain, head kidney, skin, and spleen, was previously reported [27, 28, 44, 45]. Fifteen additional libraries from the tissues listed above were constructed here closely following protocols used previously. Briefly, the cDNA libraries were constructed using the pSPORT-1 Superscript Plasmid Cloning System from Invitrogen. This cloning system provides a vector with capacity for uni-directional cloning of cDNAs that support choices of EST sequencing from either the 5'-, or 3'-end of the transcript. In this work, all ESTs were sequenced from upstream of the transcripts (5' sequencing) to provide a longer length of ESTs. Two micrograms of Poly(A)+ RNA were used in each initial reaction. Procedures followed instructions provided by the manufacturer with the exception that ElectroMax DH12S cells (Invitrogen) were used for electroporation of the cDNA library. The quality of the cDNA libraries was determined by number of primary recombinants and average insert size. Before sequencing analysis, the primary cDNA libraries were amplified once [46]. The pooled libraries were frozen in liquid nitrogen and stored at -80°C.

Colony lifting hybridization and sequencing

Colony-lifting hybridization [46] was conducted using overgos as probes to reduce the sequencing redundancy of a set of 200 genes determined to be highly expressed by preliminary sequencing of the libraries. Oligonucleotides were custom made by Sigma Genosys (St. Louis, MO). Overgos were designed to overlap for 8 bases where the sense and antisense oligos pair, leaving the remaining 5' overhang for filling in using labeled nucleotides, P³²-dATP and P³²-dCTP [47, 48]. All the overgos were labeled in a single reaction. Overgo hybridization was conducted at 45°C overnight using conditions as previously reported [49]. The filters were washed using 2× SSC at room temperature four times for 15 minutes each. After exposure of X-ray films, bacterial plates were aligned to match the patterns of the exposed colonies on the X-ray film. Negative colonies were picked for sequencing and manually arrayed into 384 well plates containing LB with antibiotics and 10% glycerin and stored at -80°C until sequencing. Sequencing was conducted using ABI PRISM 3730 automated sequencers located in the Core Facility of Purdue University.

Sequence analysis, EST clustering, and sequence annotation

ESTs were trimmed for vector and adaptor sequences. Base calling was performed using the Phred program with quality cut-off set at 20. Sequences were assembled in CAP3 using a criteria of a minimum overlap of 70 bp sharing 90% sequence identities for clustering. Cleaned ESTs were used as queries for BLASTX searches against the nr database at NCBI and annotated based on the top, informative BLAST hit. A cutoff E-value of e^-5 was used for annotation. The channel catfish and blue catfish ESTs were submitted to dbEST. A database was developed to facilitate information dissemination. ESTs were annotated using the Gene ontology (GO) terms and the results built into the database.

Catfish ESTs and comparative analysis

Chromosome-assigned proteins of Tetraodon nigroviridis as well as those from undetermined chromosome locations were downloaded from the protein database of NCBI. All proteins linked to a given Tetraodon chromosome were uploaded separately as query files onto the University of Illinois Keck Center's Gridblast server. All catfish ESTs from NCBI's dbEST were uploaded as a database on the same server. The TBLASTN search parameters were set to select the top catfish hit, using a cutoff E-value of e^-10. Resulting text files were parsed to obtain Tetraodon query IDs, catfish hit IDs, and e-values and these were imported into Excel spreadsheets. Results were further sorted to separate those catfish ESTs hit by a single Tetraodon query and those hit by multiple Tetraodon queries. Catfish ESTs hit by a single Tetraodon query were uploaded to Msatfinder [50] to search for microsatellites contained in the sequences. BLASTX searches were carried out on those catfish ESTs hit by 20 or more Tetraodon queries.

Assessment of gene duplication

Channel catfish TIGR consensus (TC) sequences, composed in part by the ESTs reported here [51] were used as queries for BLASTN searches against the est_others database of NCBI, limiting the entrez query to Ictalurus. Top hits with perfect matches (E-value = 0.0) were the channel catfish sequences from the TC. If additional highly similar hits (E-value <e^-25) from both channel catfish and blue catfish ESTs were present, these sequences were noted for further analysis as potential gene duplicates. Reciprocal BLASTX searches were carried out using at least three ESTs from the initial searches, with at least one of these ESTs from blue catfish. When all ESTs shared the same top BLASTX hit, they were translated, and areas of amino acid overlap identified. Phylogenetic trees were drawn by the neighbor-joining method [52] within the Molecular Evolutionary Genetics Analysis (MEGA 3.0) package [53]. Data were analyzed using Poisson correction and gaps were removed by complete deletion. The topological stability of the trees was evaluated by 1,000 bootstrapping replications.