Compendium of zebrafish lncRNAs

The lncRNA annotations were independently derived from manual curation of data from published literature and supplementary resources [16–18]. The lncRNA annotations and their genomic loci were collated. The bulk of annotations came from the three recent RNA-Seq datasets, which characterized lncRNAs in developmental stages as well as adult tissues in zebrafish. The similarities and differences within the three RNA-seq datasets with respect to the sample used, analysis protocols and lncRNAs identified have been discussed in a recent review on the field [19]. A merged annotation of these lncRNAs was made and this served as the template for the analysis of their expression levels in various datasets.

Analysis of publicly available RNA-Seq data

Raw RNA-Seq data for each study was downloaded from Sequence Read Archive (SRA) and the samples were analysed using the standard pipeline as detailed. The list of datasets used and descriptions are available as Table A in S1 File, Fig 2. TopHat was used for the alignment of reads to the reference genome (Zv9 genome assembly), which performs ultra fast short read mapping using bowtie based on exon–exon splice junctions [20]. Transcript assembly for different runs of each sample was done with cufflinks and the different assemblies were then merged for each sample using cuffmerge. Further downstream analysis for differential expression (DE) was carried out using cuffdiff. The lncRNAs were further classified and named on the basis of transcript type and their corresponding expression pattern.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Matrix reporting sample count of various datasets used across different developmental time-points.

https://doi.org/10.1371/journal.pone.0129997.g002

Mapping ChIP-Seq data on the lncRNome

Genome-wide ChIP-Seq datasets from five studies were retrieved from SRA and aligned to the reference genome of zebrafish using—Mapping and Assembly with Quality (MAQ) [21]. A complete list of the datasets used and descriptions is available in Supplementary Data I. The peaks were called using—Model-based Analysis of ChIP-Seq (MACS) [22] as described previously [23]. Histone modification marks included developmental time-points dome, shield, epiboly, 24hpf, 48hpf and adult stages of the zebrafish [18, 24–27].

Integrating Ribo-Seq information

Zebrafish ribosomal profiles, as predicted by Chew et. al. [28], were also integrated in the database. Sequencing reads for ribosome-protected fragments for different developmental time-points spanning 2–4 cell, 256 cell, 1k cell, dome, shield, bud, 28 hpf and 5 dpf stages were retrieved from SRA. The dataset was preprocessed by removing the adapter sequences and discarding the reads that mapped to rRNA using Bowtie2 [29]. The remaining reads were then aligned to the zebrafish transcriptome and Zv9 genome assembly with TopHat2 [30] as described by Chew et. al. Coverage of Ribo-Seq reads across the zebrafish developmental time course can be analysed over the genome browser to check if the transcript can have any possible coding potential. In addition, the Open Reading Frames (ORFs) were predicted for each of the transcripts using getORF, which is available as part of the EMBOSS suite[31].

Genomic variations and mutation information

A number of insertional mutagenesis approaches have been employed in zebrafish towards understanding gene functions by closely following up phenotypes using molecular methods. Retroviral genomic insertions from publically available datasets have also been included in the resource for each of the lncRNA [13, 32]. Apart from this, presence of important variations reported in dbSNP within the exonic regions of the lncRNAs was also checked and catalogued [33].

Database design and architecture

The resource has been built in MySQL and the web interfaces have been coded in Perl-CGI. For each putative lncRNA, information related to the corresponding stage specific expression, open reading frames, the retroviral insertion maps and variant data has been compiled in different annotation tables and linked to provide a user-friendly interface. To explore the lncRNAs across entire genome, taking into consideration various available annotation marks, a genome browser has been embedded within the interface. Alignment maps of histone modification marks, ribosome profiling, expression levels, transcription factors and variations have been loaded into the browser. Tracks for RefSeq genes, ENSEMBL genes and the genes nearest to lncRNA are also added to enable accurate annotation and functional analysis of the transcripts taking into consideration the reports from all the associated studies.

Database features and navigation

Information on a specific lncRNA is organized as a simple and browse-able interface, which can be searched using either their gene names, aliases, genomic loci or by the nearest protein-coding gene. To detail the salient features of the database, we describe the specific annotation for a well characterized lncRNA in zebrafish- megamind. The screenshot of the resource (Fig. A in S1 File) shows the annotation of the lncRNA. The expression profile of the lncRNA across developmental stages and tissues support the earlier observation that the lncRNA is highly expressed in brain among adult tissues, while provides additional information that it is developmentally regulated. The genome browser provides an option where the user can visualize the transcript in the context of various other integrated experimental datasets, including ribosome profiling data and histone modifications across developmental time points. Analysis suggests histone modification—H3K4me3 is closely associated with the lncRNA gene body, while activator mark H3K27ac shows association with the lncRNA promoter. The resource also provides a ready reference to genomic variations in the lncRNA loci and ready links to relevant citations describing the lncRNA and the sources of relevant datasets integrated.

An overview of the nearest gene and distance between the TSS of the corresponding lncRNA is provided in the list displayed on querying the input. Selecting a transcript provides detailed information on each transcript organized in the sections as detailed below:

Transcript information

The database provides basic annotation of the lncRNA transcripts along with the nomenclature, aliases and genomic coordinates. The database content is largely derived from the three major and recent publications which include 691 lncRNAs predicted in early embryogenesis, 1,133 in late developmental stages and 442 from adult tissues. Each of the lncRNAs was further categorized into sense intronic, overlapping, intergenic and promoter associated depending on their genomic context in relation to protein-coding genes. Analysis revealed a total of 86 intronic, 485 overlapping, 309 promoter associated and 1,386 linc-RNAs. All the three studies showed a preponderance of intergenic lncRNAs. This observation could arise because the transcript overlaps with Ref-Seq protein coding genes were filtered for the annotation of lncRNAs.

Genome Browser

The entire genome can be explored with an unparalleled speed through the genome browser featuring localized annotations for each of the transcript [34]. It displays various feature tracks that include exonic regions, nearest gene, variations, expression levels in different stages and five epigenetic marks, transcription factors and ribosome profiles across different stages from various samples simultaneously in a single panel.

Expression and regulation

The availability of a catalog of lncRNAs encompassing all the annotations and the expression levels of the transcripts from RNA-Seq data offers a unique opportunity towards creating a spatiotemporal map of gene expression in lncRNAs. In addition to the track displayed in genome browser, expression levels across ten developmental time points and five adult tissues are represented graphically with log₁₀ FPKM values plotted across different stages. This section details the conditions in which the lncRNA is highly expressed.

A number of recent reports have characterized the promoter epigenetic marks of lncRNAs and have suggested that the promoter epigenetic marks in lncRNAs are similar to that of protein coding genes [35]. Drawing parallels, it would be imperative to understand the epigenetic marks associated with lncRNAs. Histone modification marks encompassing H3K27ac, H3K36me3, H3K4me1, H3K4me3 and H3K27me3 reported in zebrafish across different developmental time points have been integrated and provided in the genome browser. For ease of interpretation, activator marks are shown in green color while the repressor marks are depicted in red color. In addition, ChIP-Seq datasets encompassing a number of critical transcription factors have also been integrated in the genome browser. This includes transcription factors such as Nanog-like, Mxtx-2, gata-1, Sox-2, Pou5f1, Cdx-4 and Sal-4 [36–39].

Mutant information

The resource provides an easy access to information regarding mutants thereby aiding researchers to study them in detail towards understanding the biological mechanisms and phenotypes associated with the particular lncRNA. A systematic mapping of a total of 15,223 publicly available retroviral insertions from ZFIN [13] showed a total of 111 insertions mapped to 126 lncRNA transcripts. A set of 156 insertions reported in ZETRAP have also been included in the resource as a track in the browser [32].

Apart from this, in context to the queried transcript the predicted open reading frames and important variations falling within the exons and references pointing to relevant literature information are also provided. In addition, the experimental datasets for ribosome profiling during developmental time-points are also provided as a brows-able track on the genome browser.

In summary, the resource thereby allows the study of zebrafish as a model organism with a broad perspective taking into consideration the genomic, transcriptomic and epigenetic context. A comparative analysis of the features of zflncRNApedia vis-à-vis other two major resources–ZFIN and lncRNAdb, is reported in Fig. B in S1 File suggesting unique salient features in the resource. The list of predicted transcripts, corresponding expression levels, variations and retroviral insertion maps have been provided for download at the home page as tab-delimited text files. In addition, the compendium of lncRNA annotations could be visualized on UCSC genome browser as a track hub.