Prevalence of alternative splicing choices in Arabidopsis thaliana

Abstract

Background: Around 14% of protein-coding genes of Arabidopsis thaliana genes from the TAIR9 genome release are annotated as producing multiple transcript variants through alternative splicing. However, for most alternatively spliced genes in Arabidopsis, the relative expression level of individual splicing variants is unknown.

Results: We investigated prevalence of alternative splicing (AS) events in Arabidopsis thaliana using ESTs. We found that for most AS events with ample EST coverage, the majority of overlapping ESTs strongly supported one major splicing choice, with less than 10% of ESTs supporting the minor form. Analysis of ESTs also revealed a small but noteworthy subset of genes for which alternative choices appeared with about equal prevalence, suggesting that for these genes the variant splicing forms co-occur in the same cell types. Of the AS events in which both forms were about equally prevalent, more than 80% affected untranslated regions or involved small changes to the encoded protein sequence.

Conclusions: Currently available evidence from ESTs indicates that alternative splicing in Arabidopsis occurs and affects many genes, but for most genes with documented alternative splicing, one AS choice predominates. To aid investigation of the role AS may play in modulating function of Arabidopsis genes, we provide an on-line resource (ArabiTag) that supports searching AS events by gene, by EST library keyword search, and by relative prevalence of minor and major forms.

Figure 1

Recognizing alternative splicing events and EST support. The diagram presents the genomic alignments for two alternative gene models (shaded purple) alongside alignments for six ESTs. The models propose two AS events, one in which two introns with different boundaries overlap and another in which one model retains an intron that is absent in the other. For each AS event, an exon in one gene model (called e_overlap) overlaps an intron (labeled i_overlap ) in another gene model; this overlap forms the basis of the AS event detection algorithm described in the Methods section. For the five-prime AS event, which involves an alternative acceptor site, the differentially-spliced intron in Model 1 also overlaps an alternative intron (labeled i_alt ) in Model 2. The orange regions represent Difference Regions (D_R), segments of genomic sequence that are included in one gene model and not the other. Also shown are ESTs supporting different splicing decisions. For individual AS Events, the gene model that contains the D_R is called G_P (Gene Present) and the model that lacks the D_R is called G_A (for Gene Absent); ESTs that overlap a D_R can support either the G_A or G_P choice, but not both. In the case of AS events involving retained introns, overlapping ESTs provide support for removal of the intron when their boundaries coincide with the intron's boundaries. Alternatively, they provide support for retention of the intron when they contain a block of alignment that begins 20 bases on either side of the intron boundary and extends at least 20 bases within it.

Figure 2

Intron support and intron position. The diagram shows the percentage of supported introns among introns removed from sequential 50 base pair regions with respect to the spliced transcript, for spliced transcripts 2,000 bases or larger.

Figure 3

Size distribution of Difference Regions in the TAIR9 gene models. The size distribution for Difference Regions in which each AS choice is supported by one or more ESTs is shown for AS events involving exon skipping (top left), retained introns (top right), alternative donors (bottom left) and alternative acceptor sites (bottom right.) Not shown are five retained intron Difference Regions larger than 600 bases.

Figure 4

Histograms showing the distribution of percent minor form among alternative splicing events in Arabidopsis. Percent minor form is calculated as the number of ESTs supporting the less frequently observed choice divided by the total number of ESTs supporting either choice, multiplied by one hundred. Only AS events were counted for which there was at least one EST supporting each alternative and where at least (a) 5 (b) 10 (c) 15 (d) 20 (e) 25, or (f) 30 ESTs overall support both forms together.

Figure 5

ArabiTag on-line tool for assessing frequency of splicing choices from Arabidopsis gene models. Screen captures showing query and results pages from ArabiTag http://www.transvar.org/arabitag are shown. Searching for locus id AT2G39730, encoding rubisco activase, a well-known example of conserved AS in plants [14,34], retrieves three AS Event ids and their corresponding Difference Regions. The alternative splicing choices for the third AS event on the list are supported by multiple ESTs. Clicking the AS Event id link opens an AS Report page in ArabiTag that shows the breakdown of support for each choice, by library, in three different tables according to which choice the ESTs support. Libraries that contain a mix of different ESTs that support both choices appear in the first table, libraries with ESTs that support just one alternative (G_A) appear in the next table, and libraries with ESTs that support the other alternative (G_P) appear in the final table. In all cases, the number of ESTs from each library that support G_A or G_P are also listed, thus providing the user with information about possible tissue or sample-type specific alternative splicing. However, in this case, both rubisco activase forms are widely expressed, suggesting they may co-occur in the same cells or cell types.

Figure 6

ArabiTag linking to Integrated Genome Browser. Screen captures showing query and results pages from ArabiTag http://www.transvar.org/arabitag are shown. A page describing alternative splicing of AT2G39730 (encoding rubisco activase) offers an "IGB link" icon that, when clicked, directs the currently running instance of the Integrated Genome Browser to display the region specified in the IGB link. For convenience, the page also includes a link to the on-line launch page for the IGB software http://www.bioviz.org/igb. To use the IGB link, users download and launch a copy of the IGB software, load the spliced EST and TAIR9 mRNA data sets from the BioViz DAS2 and BioViz Quickload data sources, setting the load mode for the TAIR9 mRNA data set to "Whole Genome." Clicking the IGB link within the ArabiTag Web page directs IGB to zoom and scroll to the Difference Region for this AS choice. Clicking the "Refresh Data" button instructs IGB to retrieve all the spliced ESTs that overlap the region, allowing the user to compare their boundaries with the TAIR9 gene models. Clicking an EST (as shown in the image) activates an edge-matching function, where all items with boundaries identical to the selected item (an EST in this case) acquire an edge-matching icon - a white bar drawn on top of the matching boundary. Right- (or control-clicking) the EST offers the user the option to visit the NCBI Web site and view the dbEST record for the selected EST, revealing it comes from a library prepared from salt-stressed seedlings.

Figure 7

Integrated Genome Browser showing the entire spliced EST data set for chromosome 2. Here, the user has set the spliced ESTs track to allow maximum expansion, thus allowing the ESTs to occupy as much vertical space as necessary to show all ESTs mapping to chromosome two, even for regions with many hundreds overlapping ESTs arising from highly-expressed genes. Two of the largest EST stacks are from AT2G21660 and AT2G30570, encoding a small, glycine-rich RNA binding protein (ATGRP7) and a protein similar to photosystem II reaction center W, respectively.