SkateBase, an elasmobranch genome project and collection of molecular resources for chondrichthyan fishes

Chondrichthyan fishes are a diverse class of gnathostomes that provide a valuable perspective on fundamental characteristics shared by all jawed and limbed vertebrates. Studies of phylogeny, species diversity, population structure, conservation, and physiology are accelerated by genomic, transcriptomic and protein sequence data. These data are widely available for many sarcopterygii (coelacanth, lungfish and tetrapods) and actinoptergii (ray-finned fish including teleosts) taxa, but limited for chondrichthyan fishes. In this study, we summarize available data for chondrichthyes and describe resources for one of the largest projects to characterize one of these fish, Leucoraja erinacea, the little skate. SkateBase ( http://skatebase.org) serves as the skate genome project portal linking data, research tools, and teaching resources.


This article is included in the Elasmobranch
Biology & Conservation Collection Chondrichthyan fishes are composed of two subclasses, Holocephali and Elasmobranchii. Holocephalans are the more basal of the pair having first appeared more than 400 million years ago and include a single surviving order, Chimaeriformes, the chimaeras, with 39 extant species 1 . Elasmobranchs appeared approximately 350 million years ago and include more than 1000 species of sharks, skates, and rays 2 . Chondrichthyan fishes occupy a pivotal position at the base of the vertebrate phylogenetic tree. For research that includes an evolutionary component, representation of this diverse class affords a valuable perspective to evaluate all vertebrates.
Chondrichthyan fishes are circumglobal in distribution and occupy a wide range of ecological habitats. Their life history parameters are equally disparate but in general chondrichthyans are slow growing and late maturing fishes with an increased risk of extinction [3][4][5] .
Fecundity is as few as 1 or 2 for viviparous species such as the sand tiger shark, Carcharias taurus 6 and as high as 300 for the whale shark, Rhincodon typus 7 . They are of economic importance for fisheries as well as ecotourism. Management and assessment of stock is essential to ensure both ecotourism interests and food resources remain sustainable 8 . Management of fish populations has increasingly relied on molecular tools to investigate population structure, properly identify species, and compliance with fishing quotas 9-13 .
Elasmobranchs have been used as a model for biomedical research for more than 100 years. Elasmobranchs, like other cartilaginous fishes, exhibit many fundamental vertebrate characteristics, including a neural crest, jaws and teeth, an adaptive immune system, and a pressurized circulatory system. The skate is a powerful comparative model to study biological processes shared among jawed and limbed vertebrates such as development [14][15][16] , renal physiology 17-20 , immunology 21-26 , toxicology 27 , neurobiology 28 , and wound healing and regeneration 29 . They are the most ancient vertebrates to posses an adaptive immune system that generates antibodies using a V(D)J combinatorial mechanism 30 . Phylogenetically, cartilaginous fishes are the first vertebrates to possess a thymus, a central lymphoid organ that provides a microenvironment for the development of T cells 31 . The thymus shares a common organization with more derived vertebrates containing cortical and medullary regions 32,33 .
In addition to shared physiological characteristics, the diversity of specializations between species allows investigations of evolution within a single clade. For example, elasmobranchs use a plethora of reproductive strategies that span the full range of maternal investment from placental viviparity to strict lecitrophic oviparity. Besides sexual reproduction, captive elasmobranchs are capable of asexual parthenogenesis [34][35][36] . Of these reproductive mechanisms, the most tractable for research purposes is oviparity. Approximately 43% of chondrichthyans utilize oviparity including all Chimaeriformes, Heterodontiformes (bullhead sharks), Rajoidae (skates) and Scyliorhinidae (catsharks) 37 . Many species can be maintained in captivity and will breed and lay eggs throughout an annual season 38 . Artificial insemination has been reported for two oviparous species, the clearnose skate, Raja eglanteria 39 , and the cloudy catshark, Scyliorhinus torazame 40 . Additionally, sperm storage allows wild caught females to lay eggs for several years without requiring males or captive mating events 41 .
Leucoraja erinacea, the little skate, was chosen for a genome sequencing project to represent this clade of fishes because of their use as a biomedical model, experimental tractability, genome size, existing sequence data, and northeast regional distribution. The sequencing project is an ongoing effort of the North East Bioinformatics Collaborative (NEBC) of the North East Cyberinfrastructure Consortium (NECC), composed of the bioinformatics core facilities from Delaware, Maine, New Hampshire, Rhode Island, and Vermont funded by National Institutes of Health (NIH) Institutional Development Awards (IDeA) and/or National Science Foundation (NSF) Experimental Program to Stimulate Competitive Research (EPSCoR) programs.

Existing resources
There is a single order of holocephalans and 13 orders of elasmobranchs. The distribution of species in orders, families and genera is shown in Figure 1.

Chondrichthyan genomes
Currently there are multiple efforts to sequence an elasmobranch genome in various stages of completion (Table 2); however, only is illustrated in Figure 3. Despite the majority of species belonging to Elasmobranchii, the GenBank, UniProtKB/TrEMBL, and Gene databases are dominated by chimaera data derived from the genome sequence of the elephant shark, C. milii 43

RNA databases
Transcriptome sequencing seeks to characterize all genes expressed in a tissue or set of tissues in a sample. Technologies to identify the complete RNA transcript sequence have developed from studies of a small number of transcripts to comprehensive characterizations. The application of large-scale cDNA cloning of Expressed Sequence Tags (ESTs) gave initial characterizations of 5-prime and/or 3-prime ends of transcripts in several elasmobranchs including L. erinacea and S. acanthias (Table 3). EST sequence data are available in the EST divisions of the GenBank, EMBL and DDBJ databases that make up the INSDC. cDNA clones and their sequences from these EST projects have enabled the complete characterization of the full-length cDNA sequence of several genes. In the last five years, high-throughput RNA sequencing (RNA-Seq) has been applied to comprehensively examine the complete The distribution of data in NCBI databases, PDB, and UniProtKB for chondrichthyan orders is shown in Figure 4. When order Chimaeriformes is included ( Figure 4A) the distributions are disproportionate due to the large volume of annotated sequence data from the elephant shark genome. The distributions are repeated exclusively for elasmobranchs. To understand if the data distribution is representative of the number of species in each order, a species distribution is included in each chart. A cladogram ( Figure 4B) is linked to the chart legend and illustrates the phylogeny between chondrichthyan orders.

SkateBase
SkateBase (http://skatebase.org) is the public portal for the little skate genome project and is a valuable collection of data and learning resources. The NEBC little skate genome project team hosted three week-long workshops and a mitochondrial genome annotation jamboree with the goal of using the project data to develop a bioinformatics aware workforce and foster collaborative and distributed big data research. The lecture materials and worked annotation examples are included at SkateBase for educational use (http:// skatebase.org/workshops). The project vitae contains an overview and timeline of the genome project effort along with key personnel, project related publications and presentations, the curation team, and citation information for researchers utilizing the resource in their publication. A Gene Table currently represents manually curated genes derived from workshops and curriculum with extensive annotation evidence. The number of gene entries will continue to grow through usage and expansion of the SkateBase educational modules. Plans to update the annotation interface to enable community annotation by domain experts is planned for the future.
SkateBase provides links to web resources with chondrichthyan data including the Chondrichthyan Tree of Life, Elephant Shark Genome Project (http://esharkgenome.imcb.a-star.edu.sg), the first described genome for a chimaera 43 , and Vertebrate TimeCapsule, (http://transcriptome.cdb.riken.go.jp/vtcap), a project that aims to develop a gene database to represent evolution and development for vertebrates and currently includes transcriptome data for a hagfish (Eptatretus burger), shark (S. torazame) and birchir (Polypterus senegalus) 62 . SkateBase data is linked locally as well as from NCBI in the Gene Expression Omnibus (GSE26235), GenBank (AESE010000000) and Sequence Read Archive (SRA026856) to ensure convenient and easy access. A link to the American Elasmobranch Society (http://www.elasmo.org), a non-profit organization with the mission of advancing the scientific study of living and fossil sharks, skates, rays, and chimaeras and promoting education, conservation, and wise utilization of natural resources, connects domain scientists to the little skate genome project.
SkateBase data includes embryonic transcriptomes for three chondrichthyan species, a chimaera, C. milii, a shark, S. canicula and the little skate, L. erinacea as well as the first draft of the little skate genome. The assembled skate genome sequence gave a single highcoverage contiguous sequence that represented the entire length of the mitochondrial genome. The mitochondrial genome was subsequently annotated as part of a Jamboree in 2011 63 . The annotated sequence is represented by the NCBI Reference Sequence (RefSeq) project, accession NC_016429, and provides extensive information for each gene.
sequence of transcripts in tissues of cartilaginous fishes. Among the most valuable RNA-Seq datasets are those from whole late-stage embryos following organogenesis. Our project has generated these datasets for L. erinacea, S. canicula and C. milii 52 . Public RNA-Seq data sets can be found in the NCBI Gene Expression Omnibus and Short Read Archive (SRA) databases or the EBI ArrayExpress and European Nucleic Acid (ENA) archives (Table 3 and Table 4).

Mitochondrial genomes
Individual mitochondrial genes such as cytochrome c oxidase subunit I (CO1 or COX1) and NADH-ubiquinone oxidoreductase chain 2 (NADH2 or MT-ND2) have been used extensively to construct molecular phylogenies [55][56][57] . The Fish barcode of life (FISH-BOL) a working group of the International Barcode of Life Project (iBOL), has CO1 barcodes for 54% of elasmobranchs and 62% of holocephalans (http://www.fishbol.org, accessed July 24, 2014). Recently, whole mitochondrial sequences are increasingly popular for their increased granularity when resolving branches of phylogenetic trees 1 . Whole mitochondrial genome sequences currently are available for 72 species of sharks, skates, rays and chimaeras. These sequences are accessible in the GenBank, EMBL and DDBJ databases summarized in Table 5 58 .

Chondrichthyan Tree of Life
Currently, molecular data for cartilaginous fishes is being collected as part of the Chondrichthyan Tree of Life project (http:// sharksrays.org). The project website currently includes 5 elements: 1) an interactive phylogenetic tree 55 ; 2) scientific illustrations of specimens; 3) range information for all extant species; 4) interactive comparative anatomy through segmented CT scan data; and 5) DNA sequence for 1265 single copy orthologous genes 59 . Project data will be available in public databases as well as through the project website once collection and analysis is complete (personal communication, Gavin Naylor, Medical University of South Carolina).

Protein databases
Given the improved technologies to characterize full-length transcripts using RNA-Seq, there are increasingly more protein sequence data for chondrichthyans.  61 . Of these, 76% are derived from 2 species from a single family, Torpediniformes, the electric rays, and in total only 10 species are represented in PDB.    Whole embryos were used to build the transcriptome libraries available at SkateBase 35 . Two C. milii embryos, stage 32, were combined and used to build a chimaera library. The transcriptome library for S. canicula was assembled from six pooled embryos, stages 24-30. The embryonic skate transcriptome library was assembled using six pooled embryos ranging in stage from 20-29. This combination of stages encompasses a large portion of the developmental period for these fishes and represents a catalog of genes important for organogenesis of all or part of every physiological system. Early developmental events are similar for nearly all elasmobranchs regardless of reproductive mode or adult body form enabling the data to be useful for more than just the specific species from where it was derived 64 . Since all three embryonic transcriptomes contain a similar stage embryo direct comparison for temporal expression patterns is possible. Skatebase includes tools for data investigation, SkateBLAST, a sequence retrieval tool, Skate Contig Lookup, and genome browsers for three skate whole mitochondrial sequences, L. erinacea, the thorny skate, Amblyraja radiata, and, the ocellate spot skate, Okamejei kenojei. Skatebase contains resources that can be used for teaching and research purposes. As an example, two use cases follow, one for sequence or homology based research and the other for education.

SkateBLAST
A common task for researchers is searching for genes of interest in a genome or transcriptome. Knowledge of the gene sequence at the DNA or RNA level is needed for many different studies, including phylogenetic analysis or designing primers for quantitative PCR gene expression studies. Here we describe the major steps necessary to identify relevant sequences for a gene of interest using the BLAST sequence similarity tool at SkateBase. SkateBase features a web interface to BLAST, named SkateBLAST, that builds upon the ViroBLAST package version 2.2 65 , with custom modifications allowing parallel cluster-based execution of queries and enhanced display of results. The overall workflow consists of a) entering a query sequence and selecting the database to search; b) evaluating the alignments returned; c) retrieving the sequence from one of the SkateBLAST databases; and d) checking to make sure that the retrieved sequence aligns best to the query sequence. The following description provides a brief tutorial on the overall workflow while describing tools at SkateBase. Figure 5 demonstrates the use of SkateBLAST to find expressed sequences for the gene, suppressor of cytokine signaling 6 (SOCS6  Figure 5. Example of using SkateBase and NCBI resources to find transcriptome data for SOCS6. A. SkateBLAST query form showing the four steps to align the UniProt sequence for human SOCS6 (O14544) against the skate embryonic transcriptome using tblastn.
Step 1 is to enter the sequence in FASTA format. The second step is to choose the tblastn program that will align the query protein sequence against translated sequences in all six possible reading frames. The third step is to select the embryonic transcriptome as the sequence database to search. The fourth step is to launch the search. B. The complete BLAST output can be accessed by clicking the "Inspect BLAST output" link at the top of the summary report page. This is necessary to examine the sequence alignments. C. Four important fields in the output should be examined carefully to interpret the alignments and determine which returned alignment best represents the skate ortholog to SOCS6. First, the alignment score, E-value, alignment length and percent identity can be used to interpret the overall alignment significance. Alignment coverage with respect to the query protein sequence and the subject transcriptome sequence can be interpreted by comparing the alignment coordinates to the length of the query protein sequence and length of the transcriptome sequence. In this example, the entire query protein sequence is covered by this transcriptome sequence. D. The SkateBase Contig Lookup tool can be used to retrieve the transcriptome sequence found in the SOCS6 tblastn search in FASTA format. Sequences from the skate genome assembly or the skate, S. canicula or C. milii transcriptome assemblies can be retrieved using this tool. E. Output from the NCBI ORF Finder tool showing a 536aa ORF in the skate transcriptome contig that best represents SOCS6 (left). Alignment from blastx search of the skate transcriptome sequence (contig 15542) against human UniProt using NCBI BLAST to validate that the contig aligned best to human SOSC6 rather than another human gene. cellular proliferation through its SH2 domain 66 . The first step to identify SOCS6 in the skate transcriptome begins with entering the protein sequence for human SOCS6 that was obtained from UniProt and searching this sequence against the skate transcriptome using the tblastn program. The next step is to evaluate the alignments to determine which transcriptome sequences best represent SOCS6. When interpreting the pairwise alignments from SkateBlast as in any BLAST tool, it is important to examine: a) alignment statistics; b) alignment coverage; and c) presence of protein domains that you may expect to be conserved. The alignment statistics are reported to ascertain whether you would expect the given alignment by chance or not. There are three key alignment statistics, the expectation (E)-value, percent identity and alignment length. The E-value represents the probability that you would expect an alignment with that alignment score or better by random chance, thus the lower the E-value, the better the alignment. Conversely, the greater the percent identity (percent identical sequence) and alignment length, the more similar the two sequences are assumed to be. Alignment coverage with respect to the query or subject sequence (alignment length divided by the length of the query or subject sequence) can also be an important consideration, as low coverage suggests that important regions of one or both sequences may not be represented in the alignment. Finally, there may be particular sequence features, such as protein domains, that you would expect to find in the alignment. If those domains are missing, then it suggests that you have a partial or misleading alignment.
Once a transcriptome sequence of interest, such as contig15542, is identified in the SkateBLAST results, you must do a reciprocal search of that sequence against a database of protein sequences to confirm that the sequence aligns best to your gene of interest. You can retrieve the full sequence directly from the BLAST tool or using the Skate Contig Lookup tool ( Figure 5D): a) specify the transcriptome that you had originally searched using SkateBLAST; b) enter the sequence identification or contig number is entered into the query box; and c) select the 'GO' button. The user can copy the returned sequence and use it for further exploration of sequence homology at NCBI or similar databases.

SkateBase classroom use case: teach concepts of gene and protein annotation
SkateBase includes valuable teaching resources derived from the project workshops on gene and protein annotation. Infrastructure for sequence annotation was developed and modules for use in teaching are available. Access to the teaching modules is through the Curator Access link from the homepage and permission is granted by request using the email link at the bottom of each page, info@SkateBase.org. Once successfully logged into the site, access to pre-computed blast results, guides and examples, annotation forms, and links to external tools helpful for sequence analysis are available. Gene annotation begins with a transcriptome contig identified through a SkateBlast search as illustrated above. The portion of the transcript that codes for protein is identified using an open reading frame or ORF finder tool. Annotation follows a workflow where complimentary sequences from the transcriptome and genome are aligned allowing annotation of both sequences using Sequence Ontology vocabulary 67 . The evidence is recorded in an annotation form that records information about the annotator and sequences and includes a comment box for questions and comments between students and teachers or curators and annotators. The annotation form records the pairwise alignment of the transcriptome and genome contigs, notes concerning mismatches or gaps, as well as output from the ORF tool. The untranslated regions (UTR) at the beginning and end of each sequence, 5'UTR and 3' UTR regions, as well as the intron/exon structure for the genomic contig and CDS for the transcriptomic contig are recorded in the Gene Annotation Form. When completing the Gene Annotation Form, the appropriate activity must be selected and can be customized to specify the user's course ID, institution or workshop title to track annotation history. Protein annotation uses the rapid annotation interface for proteins, RACE-P, developed by the PIR. A UniProt accession number is required to initiate a new annotation form. The form is composed of 6 blocks of information, protein information, gene information, a bibliography, Gene Ontology (GO), computational analysis using tools such as Pfam 68 , TMHMM 69 , SignalP 70 , COILS 71 , NetPhos 72 and EMBOSS 73 , and protein family evidence.

Discussion
The volume of data in GenBank continues to grow exponentially, doubling nearly every 18 months. The first sequences for chondrichthyes appeared in 1983 and the overall data trend for chondrichthyans is similar to all of GenBank with three exceptions. First, the rate of increase is less than GenBank. Second, the number of sequences deposited during the first decade of the 21 st century was nearly stagnant in comparison. Third, a large spike is observed in late 2012 attributed to the Elephant Shark Genome Project data ( Figure 6). Molecular data is increasingly important for all aspects of research utilizing chondrichthyan fishes 74 . It can be a forensic tool to understand species when fins are landed without carcass and ensure protected species and quotas are respected [75][76][77][78][79] . For migrating species molecular data serves as a surrogate to classical tagging data to understand population structure and range [80][81][82][83] . In studies of evolution, molecular data provides estimates of divergence time and supplements morphological and ecological traits as the basis for a phylogeny. The benefits and uses of molecular data for these fishes are limited

C. milii genome
only by the amount of data available. SkateBase provides the only genomic data publically available for an elasmobranch in addition to embryonic transcriptomes, data tools, and educational resources.
Sequencing projects require significant funding and personnel commitments but generate a large amount of information that can be translated to knowledge by domain experts. The efficiency of this process is affected most by allowing the scientific community to access the data. The value of data sharing can be measured by the number of publications that result from its distribution. To date, 19 publications in peer-reviewed journals have used data derived from SkateBase (http://skatebase.org/vitae). Molecular data are the means to investigate genes and develop reagents for gene expression studies by PCR or in situ hybridization. Small scale sequencing efforts that generate limited or fragmented data often get deposited to hard disks and remain 'buried' and out of reach. Efforts to deposit this data at public sequence repositories are encouraged to build the foundation of data required to describe this dynamic and ancient clade of fishes. We invite investigators to contact the authors in an effort to survey the volume of private data available for potential distribution through SkateBase.
The transcriptome data at SkateBase serves as a platform to teach molecular techniques, technologies, and bioinformatics in the context of studying elasmobranchs. As next generation sequencing (NGS) technologies evolve it is important for scientists and students to understand how the sequence was generated and caveats of workflow for each data type in order to recognize errors and customize analysis algorithms. The educational materials and infrastructure at SkateBase have been used by University of Delaware, Georgetown University, MDI Biological Laboratory, University of Maine at Machias, University of Rhode Island, and most recently the Virginia Institute of Marine Science to teach gene and protein annotation concepts. We invite and look forward to continued expansion of the SkateBase educational platform as we refine the infrastructure and expand the data available for investigation through continued sequencing efforts.

Author contributions
The little skate genome project is a collaborative effort with participation from all authors. JTW and BLK prepared the first draft of the manuscript. SWP oversaw the design and implementation of the SkateBase web portal and SkateBLAST tool. All authors were involved with the revision of the draft manuscript and have agreed to the final content.

8.
Introduction -In the first sentence replace "Chondrichthyan fishes" with "Living chondrichthyan fishes." Paragraph 2 -The text implies that the only benefits of chondrichthyans are ecotourism and fisheries. Perhaps a mention of ecological services (e.g. food-web dynamics) or other benefits of chondrichthyan conservation would be appropriate here. Paragraph 4 -"asexual" parthenogenesis seems redundant.
Paragraph 5 -perhaps some more details on the distribution, life history, reproductive mode, genome size etc. of little skate is in order. Why is the geographic location of the species relevant? Figure 1 is hard for me to decipher. I can't figure out what the two separate outer rings indicate and why the spacing between them is different for the orders vs. the families, genera, and species. I really think this figure is too complex and tries to cram too much information that is peripheral to the goals of the paper into a single figure. I may be old fashioned, but I think a table that listed the numbers of families, genera, and species would have been sufficient. There also needs to be a citation on the source of the information as the taxonomy and number of recognized species of chondrichthyans is changing. Figure 4B -A citation is needed for the cladogram.
I have read this submission. I believe that I have an appropriate level of expertise to confirm that the total number of species, genera and families for the clade." I also suggest explaining the figure in a little more detail. What do the numbers mean? Do the colors represent anything? Figure 2 legend: "is" should not be italicized. Table 1: What do the numbers in the columns mean? Are these numbers of entries?  This can be rectified by using black, white and yellow for three of the above orders in the figure. Figure 5 legend: I suggest changing "Four important fields in the output should be examined carefully to interpret the alignments and determine which returned alignment best represents the skate ortholog to SOCS6. First, the alignment score, E-value, alignment length and percent identity can be used to interpret the overall alignment significance." to "Four important fields in the output should be examined carefully to interpret the alignments: the alignment score, E-value, alignment length and percent identity." I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
No competing interests were disclosed. Competing Interests: