Studying bacterial transcriptomes using RNA-seq

Abstract

Genome-wide studies of bacterial gene expression are shifting from microarray technology to second generation sequencing platforms. RNA-seq has a number of advantages over hybridization-based techniques, such as annotation-independent detection of transcription, improved sensitivity and increased dynamic range. Early studies have uncovered a wealth of novel coding sequences and non-coding RNA, and are revealing a transcriptional landscape that increasingly mirrors that of eukaryotes. Already basic RNA-seq protocols have been improved and adapted to looking at particular aspects of RNA biology, often with an emphasis on non-coding RNAs, and further refinements to current techniques will improve our understanding of gene expression, and genome content, in the future.

Figure 1

Methods for preparation of cDNA. All methods require the extraction of nucleic acids from a sample of cells, followed by the enzymatic removal of DNA. Ribosomal RNA may then be depleted to increase the sequence coverage of other transcripts. To identify putative transcriptional start sites, samples are first treated with tobacco acid pyrophosphatase (TAP), which converts the triphosphate group at the 5′ end of intact transcripts to a monophosphate. This is required for the ligation reaction to attach an adapter to the 5′ end; polyadenylation or oxidation of the 3′ end of the RNA is used to ensure the specificity in the orientation of this reaction. This allows the 3′ part of the cDNA, corresponding to the extreme 5′ end of the original transcript, to be targeted for sequencing. In order to obtain sequence data covering the entire transcriptome, small cDNA molecules must be randomly generated from throughout the RNA sample. This has frequently been achieved through random hexamer-primed reverse transcription; using only the first strand for sequencing library construction allows information on the direction of transcription to be maintained. Alternatively, the RNA may be fragmented, and information on the template strand for transcription retained through orientation-specific, stepwise attachment of adapters. One method involves dephosphorylating the 5′ end so the first adapter can only be ligated to the 3′ end of the transcript; the complementary approach is to polyadenylate the 3′ end such that the first adapter is only found attached the 5′ end of the RNA. One technique not shown is the use of fragmented RNA as a template for random hexamer-primed reverse transcription, as performed by Oliver et al. A wider range of methods has been applied in obtaining similar information from eukaryotic transcriptomes (see text).

Figure 2

Display of RNA-seq data. Data from a Salmonella bongori transcriptome, prepared as described in Ref. [9], displayed using Artemis. Using BamView, the total coverage is shown displayed as a plot (a), as raw reads aligned against the reference sequence (b) and as reads assigned separately to the two strands of the genome (c). A strand-specific coverage plot is also shown (d) and the genome annotation is displayed underneath.