Combining next-generation pyrosequencing with microarray for large scale expression analysis in non-model species

Abstract

Background: The next generation sequencing technologies provide new options to characterize the transcriptome and to develop affordable tools for functional genomics. We describe here an innovative approach for this purpose and demonstrate its potential also for non-model species.

Results: The method we developed is based on 454 sequencing of 3' cDNA fragments from a normalized library constructed from pooled RNAs to generate, through de novo reads assembly, a large catalog of unique transcripts in organisms for which a comprehensive collection of transcripts or the complete genome sequence, is not available. This "virtual transcriptome" provides extensive coverage depth, and can be used for the setting up of a comprehensive microarray based expression analysis. We evaluated the potential of this approach by monitoring gene expression during berry maturation in Vitis vinifera as if no other sequence information was available for this species. The microarray designed on the berries' transcriptome derived from half of a 454 run detected the expression of 19,609 genes, and proved to be more informative than one of the most comprehensive grape microarrays available to date, the GrapeArray 1.2 developed by the Italian-French Public Consortium for Grapevine Genome Characterization, which could detect the expression of 15,556 genes in the same samples.

Conclusion: This approach provides a powerful method to rapidly build up an extensive catalog of unique transcripts that can be successfully used to develop a microarray for large scale analysis of gene expression in any species, without the need for prior sequence knowledge.

Figure 1

Pyrosequencing reads represent the 3' end of Vitis vinifera transcripts. The position of 9,366 reads from the non-normalized (NN) 454 library matching 1,948 gene models, and the position of 21,512 reads from the normalized (N) 454 library matching 3,749 gene models, expressed as a percentile of the length of the gene model to which the read mapped.

Figure 2

Number of reads per contig in the non-normalized (NN) and normalized (N) libraries. For each library, the number of contigs presenting the indicated amount of reads is plotted as a histogram and reported as a label for each histogram.