Sampling the Arabidopsis Transcriptome with Massively Parallel Pyrosequencing

Abstract

Abstract Massively parallel sequencing of DNA by pyrosequencing technology offers much higher throughput and lower cost than conventional Sanger sequencing. Although extensively used already for sequencing of genomes, relatively few applications of massively parallel pyrosequencing to transcriptome analysis have been reported. To test the ability of this technology to provide unbiased representation of transcripts, we analyzed mRNA from Arabidopsis (Arabidopsis thaliana) seedlings. Two sequencing runs yielded 541,852 expressed sequence tags (ESTs) after quality control. Mapping of the ESTs to the Arabidopsis genome and to The Arabidopsis Information Resource 7.0 cDNA models indicated: (1) massively parallel pyrosequencing detected transcription of 17,449 gene loci providing very deep coverage of the transcriptome. Performing a second sequencing run only increased the number of genes identified by 10%, but increased the overall sequence coverage by 50%. (2) Mapping of the ESTs to their predicted full-length transcripts indicated that all regions of the transcript were well represented regardless of transcript length or expression level. Furthermore, short, medium, and long transcripts were equally represented. (3) Over 16,000 of the ESTs that mapped to the genome were not represented in the existing dbEST database. In some cases, the ESTs provide the first experimental evidence for transcripts derived from predicted genes, and, for at least 60 locations in the genome, pyrosequencing identified likely protein-coding sequences that are not now annotated as genes. Together, the results indicate massively parallel pyrosequencing provides novel information helpful to improve the annotation of the Arabidopsis genome. Furthermore, the unbiased representation of transcripts will be particularly useful for gene discovery and gene expression analysis of nonmodel plants with less complete genomic information.