Genome-wide detection of novel regulatory RNAs in E. coli

Abstract

The intergenic regions in bacterial genomes can contain regulatory leader sequences and small RNAs (sRNAs), which both serve to modulate gene expression. Computational analyses have predicted the presence of hundreds of these noncoding regulatory RNAs in Escherichia coli; however, only about 80 have been experimentally validated. By applying a deep-sequencing approach, we detected and quantified the vast majority of the previously validated regulatory elements and identified 10 new sRNAs and nine new regulatory leader sequences in the intergenic regions of E. coli. Half of the newly discovered sRNAs displayed enhanced stability in the presence of the RNA-binding protein Hfq, which is vital to the function of many of the known E. coli sRNAs. Whereas previous methods have often relied on phylogenetic conservation to identify regulatory leader sequences, only five of the newly discovered E. coli leader sequences were present in the genomes of other enteric species. For those newly identified regulatory elements having orthologs in Salmonella, evolutionary analyses showed that these regions encoded new noncoding elements rather than small, unannotated protein-coding transcripts. In addition to discovering new noncoding regulatory elements, we validated 53 sRNAs that were previously predicted but never detected and showed that the presence, within intergenic regions, of σ(70) promoters and sequences with compensatory mutations that maintain stable RNA secondary structures across related species is a good predictor of novel sRNAs.

Figure 1.

Transcription termination at known riboswitches. (A) Ratios of mean expression values (MEV) of upstream leader regions to those of their corresponding ORFs. Positive values show that the MEVs of riboswitches are markedly higher than the MEVs of their downstream coding regions. (B) Expression profiles of leader sequences and downstream coding regions. The y-axes denote coverage at each nucleotide position, limited to a maximum coverage of 1000 (dashed lines). Numerical positions of transcription start sites (black arrows) follow the coordinates for the E. coli K12 genome (NC_000913.2). Wide gray arrows on x-axes, depicting ORFs, are not drawn to scale.

Figure 2.

Transcription termination at putative regulatory leader sequences. (A–I) Expression profiles of leader sequences and downstream coding regions. The y-axes denote coverage at each nucleotide position, limited to a maximum coverage of 1000 (dashed lines). Putative transcription start sites (black arrows) and their locations in the E. coli genome (NC_000913.2) are shown on x-axes. Lengths of the corresponding ORFs (wide gray arrows) are not drawn to scale. (J) In vitro transcription of thiI coding region by T7 RNA polymerase in the presence of 100 μM S-adenosylmethionine (SAM) or S-adenosylhomocysteine (SAH) relative to its expression in the absence of SAM and SAH. Data represent means of three experiments ± standard deviations.

Figure 3.

Detection of novel sRNAs in intergenic regions. (A–J) Expression profiles of intergenic regions. The y-axes denote coverage at each nucleotide, limited to a maximum of 1000 (dashed lines). Positions of transcription start sites (TSS) and terminators (stem–loop structures) found within each intergenic region are depicted. Nucleotide positions follow the numbering of the E. coli genome (NC_000913.2). Wide arrows on x-axes contain named ORFs (gray), sRNAs (pink), and an unnamed small ORF (green). Lengths of flanking genes are not drawn to scale.

Figure 4.

Hfq stabilization of sRNAs. Abundances of sRNAs in wild-type E. coli relative to the amounts in an isogenic hfq-deleted strain (normalized to 1, black line). The intergenic location of each sRNA is indicated on the x-axis. For each sRNA tested, data represent the means of three experiments ± standard deviations. Statistically significant differences from expression in hfq-deleted strain are indicated by (***) p ≤ 0.001 and (**) p ≤ 0.01 (unpaired t-test).