We have determined the TSSs for 78% of the S. Typhimurium ORFs during growth conditions in which model the extracellular virulence gene expression programme (STEX). To date this is the most extensive and accurate map of the TSSs for this bacterium. Our analysis also identified secondary TSSs for many genes and operon structures. Our MEME based promoter analysis of the first genes of operons identified conserved regions in the promoters which were found to closely resemble consensus binding sites for σ24, σ28, σ32, σ38 & σ54 factors; many of the predicted sigma factor-dependent genes had previously been experimentally verified in either E. coli or Salmonella. We verified the expression of 38 out of 87 predicted sRNAs and 45 out of 62 known sRNAs [28, 68](and J. Vogel; pers. comm.) and also extended the repertoire of sRNAs encoded within the S. Typhimurium genome by 55. Of the predicted sRNAs we were unable to verify, it is possible that they were not expressed under the growth condition studied here. We also observed that the location of the TSSs of a subset of the predicted sRNAs did not correspond to the predicted start sites; from this we infer the bioinformatic approach used to identify the TSSs may require experimental-based refinements to enhance accuracy. We identified 302 candidate antisense transcripts for the S. Typhimurium genome for which we defined a conserved -10 hexamer upstream of the TSS. Although from this study, we cannot rule out the possibility that the expression of asRNAs are a result of promiscuous transcription, other work suggests this is not the case, at least for H. pylori .
Our dRNA-seq approach to identifying ppGpp-dependent transcription was validated by comparison with a microarray-based determination of the ppGpp-dependent transcriptome performed under identical growth conditions. The GC rich discriminator region located between the TSS and -10 region of ppGpp-repressed genes has been shown to play a role in destabilising the RNAP-ppGpp complex of rRNA promoters . We were able to correlate decreased transcript levels of ppGpp-repressed genes with the abundance of GC residues within the discriminator region . Our data showed no correlation between AT content of the discriminator region and the level of ppGpp-activation. However in agreement with Da Costa et al , we did find that in general, ppGpp-activated genes contained a higher overall discriminator AT content. Interestingly we note that SPI1 and SPI2 encoded genes contain AT-rich discriminator regions and the only sigma factor known to contribute to SPI1 expression is σ70; this suggests the possibility of a direct activation of SPI1 regulatory genes by ppGpp, rather than via sigma factor competition, as has already been suggested [9, 71].
Many regulons controlled by alternative sigma factors, including σ38 and σ32 are poorly induced in cells lacking ppGpp . In order to determine whether this was also the case for S. Typhimurium, we analysed our alternative sigma factor promoter database for ppGpp-dependency. We found that almost all the genes belonging to the σ28 and σ32 regulons and more than half of the σ24 -dependent genes were ppGpp-repressed. In contrast, the σ38 regulon showed no tendency towards ppGpp-activation or repression (additional file 2: Table S1). Previous work has also shown that, in contrast to E. coli, ppGpp does not control RpoS levels in S. Typhimurium during late-log and stationary phase growth . We conjecture that the ppGpp-repression of some of the alternative sigma factor regulons may represent an adaptation to favour σ70 dependent virulence gene expression under the STEX growth conditions studied here.
It is generally accepted that elevated levels of ppGpp during amino acid starvation (stringent response) result in repression of stable RNAs (rRNA and tRNA). Consistent with this we observed repression of the rRNA operons in the wild-type compared to the ΔrelAΔspoT strain (additional file 2: Table S6). However, all but one of the tRNA mono- and polycistronic operons showed elevated transcript levels in the wild-type compared to the ΔrelAΔspoT strain; a similar ppGpp-dependent elevation of tRNA levels was found in stationary phase Rhizobium etli relative to early exponential phase . It is possible that elevation of tRNA levels could be a consequence of ppGpp-dependent differential processing or stability rather than direct ppGpp-dependent regulation. Indeed tRNA has been reported to remain stable under starvation conditions that induce rRNA degradation in E. coli . In support of the possibility of ppGpp-dependent differential processing or stability of tRNAs we observed that expression of RNaseP, a ribozyme responsible for 5' end processing of tRNAs, was ppGpp-activated in S. Typhimurium (additional file 2: Table S1). Similarly, we note that the R. etli RNaseP was also ppGpp-activated (29). We hypothesise that the ppGpp-dependent activation of RNaseP may result in reduced tRNA processing in the ΔrelAΔspoT strain and subsequent removal of incorrectly processed tRNAs via RNA quality control mechanisms .
For the known and predicted sRNAs described in this study, a MEME analysis was able to identify conserved -10 (TATTNT) and -35 (TTGaCA) regions upstream of the predicted TSSs (Figure 3D). A manual inspection of the smaller new candidate sRNA dataset identified AT rich -10 hexamers in 69% of the promoters (data not shown). A manual inspection of all of the sRNA promoters described in this study failed to find any of the well-defined alternative sigma factor binding motifs and in fact only four sRNAs, (MicA, RybB, GlmZ and GlmY) have so far been shown to be positively controlled by σE and σ54 in E. coli . This suggests that, at least for the sRNAs transcribed under the growth conditions studied here, their expression is mostly σ70 dependent and perhaps reflects the major role sRNAs play in maintaining house-keeping functions and regulating virulence determinants. In contrast to the discriminator regions of ppGpp-repressed genes, we were unable to identify a conserved GC rich region in the set of ppGpp-repressed sRNAs. Several of the ppGpp-repressed sRNAs (OmrA, OmrB, MicA, MicF and CyaR) have been shown to act as repressors of genes encoding porins and outer membrane proteins (OMPs) suggesting that ppGpp may indirectly activate these target genes. OMPs are important virulence factors and play a significant role in the bacterial adaptation to environmental conditions. Other highly ppGpp-repressed sRNAs shown to play a role in Salmonella virulence include IsrI, IsrP and CsrB. In addition, the sRNA chaperone, Hfq was ppGpp-repressed by 5.6-fold thus expanding the role of ppGpp in the regulation of Salmonella virulence gene expression (Figure 5) . The IsrI and IsrP sRNAs are expressed during infection of J774 macrophages . IsrI is also expressed during stationary phase, and under low oxygen or magnesium levels; IsrP is expressed under low magnesium and extreme acid conditions of pH2.5 . CsrB is part of the csr system shown to play a role in the regulation of invasion gene expression in S. Typhimurium . The SPI1 encoded sRNA, InvR, has previously been reported to be ppGpp-activated . Our data confirms InvR as the most highly ppGpp-activated sRNA we detected under these growth conditions (12.5-fold; additional file 2: Table S1). Although Hfq has been shown to reduce the stability of InvR , we note that despite a 5.6-fold ppGpp-dependent repression of Hfq trasncript levels, InvR remains highly ppGpp-activated. This suggests that ppGpp is able to modulate InvR transcript levels via a Hfq independent mechanism. InvR represses synthesis of the major outer membrane protein, OmpD . It is suggested sRNAs such as InvR have evolved to modulate OMP levels, which can be deleterious to the cell . OmpD has also been shown to facilitate Salmonella adherence to human macrophages and intestinal epithelial cell lines [78, 79]. Potential targets for the new ppGpp-dependent sRNAs include fabH, involved in the initiation of fatty acid biosynthesis, 6 genes involved in transport of sugars, nitrite, peptides and branched chain amino acids, and 3 transcriptional regulators, nadR, rob, and STM2275 (see additional file 2: Table S4).
We discovered extensive antisense transcription within the S. Typhimurium genome under the growth conditions studied here. Similarly, a considerable abundance of asRNA transcription was also discovered in E. coli and H. pylori [24, 54]. Interestingly, we observed candidate asRNAs to several virulence genes from SPI1, 2 and 6 and identified 4 putative ncRNAs which were classified as opposite intergenic between genes encoding several major SPI1 regulators including hilA and hilD (additional file 2: Table S5). One of the two candidate ncRNAs between hilA and hilD (Sla0508) was highly ppGpp-activated by a factor of at least 34-fold. HilD has been implicated in cross-talk between SPI1 and SPI2 expression . Under the growth conditions used in this study, SPI2 is not highly expressed compared to SPI1. It is therefore tantalising to suggest that ncRNAs may play a role in modulating expression of the STEX and STIN virulence gene expression programmes.