- Top of page
Salmonella species are enterobacterial pathogens that have been exceptionally well investigated with respect to virulence mechanisms, microbial pathogenesis, genome evolution and many fundamental pathways of gene expression and metabolism. While these studies have traditionally focused on protein functions, Salmonella has also become a model organism for RNA-mediated regulation. The present review is dedicated to the non-coding RNA world of Salmonella: it covers small RNAs (sRNAs) that act as post-transcriptional regulators of gene expression, novel Salmonella cis-regulatory RNA elements that sense metabolite and metal ion concentrations (or temperature), and globally acting RNA-binding proteins such as CsrA or Hfq (inactivation of which cause drastic phenotypes and virulence defects). Owing to mosaic genome structure, some of the Salmonella sRNAs are widely conserved in bacteria whereas others are very specific to Salmonella species. Intriguingly, sRNAs of either type (CsrB/C, InvR, SgrS) facilitate cross-talk between the Salmonella core genome and its laterally acquired virulence regions. Work in Salmonella also identified physiological functions (and mechanisms thereof) of RNA that had remained unknown in Escherichia coli, and pioneered the use of high-throughput sequencing technology to identify the sRNA and mRNA targets of bacterial RNA-binding proteins.
- Top of page
Proteins were long regarded as the only relevant players in the control of bacterial gene expression. However, the recent discoveries of unexpected numbers of small non-coding RNAs (sRNAs) and cis-encoded RNA control elements have challenged the above perception. Bacterial sRNAs are typically 50–250 nucleotides in length, generally untranslated and encoded in the ‘empty’ intergenic regions (IGRs) of bacterial chromosomes (Vogel and Sharma, 2005). Their synthesis is tightly regulated and often induced by a specific stress or virulence condition. Most sRNAs function as regulators by base-pairing with trans-encoded mRNAs, and thereby either repress or activate target genes at the post-transcriptional level. There are also some sRNAs that specifically modulate protein functions (Majdalani et al., 2005; Storz et al., 2005).
Whereas sRNAs exert control over trans-encoded targets, other non-coding RNAs regulate gene expression in cis. Besides the well-known class of antisense RNAs encoded opposite to mRNA genes (Wagner et al., 2002), cis-control includes the new classes of riboswitches and RNA thermometers. Riboswitches are highly structured RNA elements found in 5′ untranslated regions (5′ UTRs) of metabolic genes. A riboswitch commonly senses the metabolite that is synthesized or taken up by the downstream encoded protein(s), and thereby facilitates feedback regulation at the transcriptional or translational level; regulation is brought about by a programmed shift in RNA structure in response to binding of the metabolite. RNA thermometers are local structure elements that sequester the ribosome binding site (RBS) of an mRNA; structure formation is highly responsive to environmental temperature, and can thereby control mRNA translation in a temperature-dependent manner.
The systematic identification and functional analyses of non-coding RNAs were largely pioneered by work on non-pathogenic model bacteria (Majdalani et al., 2005; Storz et al., 2005). For example, a variety of approaches (Vogel and Sharma, 2005; Altuvia, 2007) have identified more than a hundred chromosomally encoded sRNAs in Escherichia coli, a number not yet matched in any other bacterium. Similarly, riboswitches have been mostly studied in Bacillus subtilis; ∼5% of all genes of this soil bacterium might be controlled by RNA-based metabolite sensing (Mandal et al., 2003).
Yet, there are well-characterized non-coding RNA regulators in pathogenic bacteria, e.g. the 514 nt RNAIII of Staphylococcus aureus, which controls many virulence factor mRNAs, or an RNA thermometer that controls prfA mRNA encoding a key transcription factor of virulence genes in Listeria monocytogenes (Romby et al., 2006; Toledo-Arana et al., 2007). Moreover, following the success of sRNA discovery in E. coli, bacterial pathogens are now being scrutinized to discover new sRNA regulators (Livny and Waldor, 2007).
Salmonella enterica serovar Typhimurium is one of the pathogens in which non-coding RNA functions have been studied extensively. This workhorse of research into enterobacterial pathogenesis is in many ways paradigmatic of the larger group of Salmonellaea; for simplicity, it will be referred to as Salmonella unless specified otherwise. Salmonella is a Gram-negative bacterium closely related to E. coli K12. Unlike the latter, non-pathogenic species, Salmonella invades and replicates in eukaryotic cells and causes disease in a variety of mammalian and non-mammalian hosts. For infection, Salmonella relies upon a range of laterally acquired virulence regions, the so-called Salmonella pathogenicity islands (SPIs). Of these, SPI-1 and SPI-2 encode type 3 secretion systems (T3SS), which translocate effector proteins to facilitate either invasion of non-phagocytic cells (SPI-1) or survival within macrophages (SPI-2). The secreted effectors are encoded by SPI-1 or SPI-2, by other minor SPIs, or by individual genes scattered throughout the Salmonella chromosome.
Both the evolutionarily close relationship with E. coli and the pathogen-specific aspects make Salmonella an attractive candidate for RNA research. First, many Salmonella sRNAs were originally identified in E. coli; whether these sRNAs serve identical functions in conserved general pathways or evolved new ones more relevant to the pathogenic lifestyle of Salmonella is a fascinating question. Second, > 25% of the total genetic material has been laterally acquired since Salmonella diverged from E. coli (Porwollik and McClelland, 2003). Do these Salmonella-specific regions contain new sRNAs or other RNA elements whose function would be missed in E. coli, and do RNA factors interconnect expression of the Salmonella core genome and virulence regions at the post-transcriptional level? Third, the auxiliary proteins (nucleases, RNA chaperones) typically involved in RNA-based circuits are highly conserved between the two species. Thus, RNA work in Salmonella can often take advantage of knowledge and mutants available for E. coli.
Research into Salmonella was never alien to aspects of non-coding RNA. For example, it revealed a novel mechanism of translational control by a Salmonella phage P22-encoded antisense RNA (Ranade and Poteete, 1993), or identified one of the first virulence phenotype of the ubiquitous tmRNA system (Julio et al., 2000). It also contributed much to understanding of cis-antisense control of extrachromosomal replication, e.g. of the pSLT plasmid essential for Salmonella virulence (Torreblanca et al., 1999), and identified a physiological role of Hfq (Brown and Elliott, 1996), a protein that since emerged as a key facilitator of the action of sRNAs.
In this review, we will focus on sRNAs and cis-encoded RNA control elements of Salmonella, and give an overview of their numbers and functions. Some of the molecules covered here have also been well investigated in E. coli or other bacteria; space constraints do not permit to extensively discuss this overlap.
Phenotypes of hfq mutants implicate sRNAs as important regulators in Salmonella
Perhaps the strongest evidence that sRNAs serve important functions in Salmonella stems from work on the RNA chaperone, Hfq. This protein preferentially binds A/U-rich single-stranded regions of RNA, and is required for both the intracellular stability of many regulatory sRNAs and their annealing with target mRNAs (Valentin-Hansen et al., 2004).
The numbers of phenotypes and deregulated genes observed in an hfq deletion mutant of Salmonella surpass those reported for any other pathogen. Deletion of hfq has for long been known to impair the expression of σS (Brown and Elliott, 1996), a general stress sigma factor essential for Salmonella virulence in mice (Fang et al., 1992). More recently, the hfq mutation was shown to attenuate the ability of Salmonella to invade epithelial cells, to secrete virulence factors, to infect mice and to survive inside cultured macrophages (Sittka et al., 2007). Loss of Hfq function also abrogates Salmonella motility, and deregulates > 70 abundant proteins. The latter includes the accumulation of outer membrane proteins (OMPs), which in turn cause chronic activation of the σE-mediated envelope stress response (Bang et al., 2005; Figueroa-Bossi et al., 2006; Sittka et al., 2007; 2008; Bossi et al., 2008). Moreover, Hfq was implicated in the control of Salmonella gene expression changes induced by the low-gravity condition experienced during spaceflight (Wilson et al., 2007).
Recent transcriptomic analysis revealed that Hfq controls, directly or indirectly, the expression of almost a fifth of all Salmonella genes, including genes in several horizontally acquired pathogenicity islands (SPI-1, -2, -4, -5), two sigma factor regulons and the flagellar gene cascade (Sittka et al., 2008). Given that Hfq primarily acts in concert with sRNAs, many of the above phenotypes in Salmonella may be attributable to loss of gene regulation by Hfq-associated sRNAs.
Salmonella expresses many sRNAs of mosaic origin
At least 70 sRNAs have been identified in Salmonella, examples of which are shown in Fig. 1. Almost half of these are encoded by genes that were originally described in E. coli (Hershberg et al., 2003); the genomic co-ordinates of the corresponding Salmonella genes were compiled by Papenfort et al. (2008). Expression in Salmonella was confirmed either in the course of functional analysis, e.g. of the CsrB/C, CyaR, GcvB, MicA, RybB, RyhB and SgrS sRNAs described further below, or by a global approach that used high-throughput sequencing of Hfq-associated RNA (Sittka et al., 2008).
Figure 1. Examples of Salmonella sRNAs and their genomic location. A. Secondary structures are based on in silico prediction of CyaR and IsrJ (Papenfort et al., 2008; S. Altuvia, pers. comm.) or in vitro chemical probing of GcvB, InvR and RybB sRNAs (Pfeiffer et al., 2007; Sharma et al., 2007; Bouvier et al., 2008). Shadowed nucleotides of the CyaR, GcvB or RybB sRNAs denote residues to interact with target mRNAs. B. Location of sRNA genes in the IGRs (drawn to scale) of the Salmonella chromosome (Salmonella typhimurium strain LT2). Neighbouring genes and the size of the IGRs are given for orientation.
Download figure to PowerPoint
Two parallel studies (Pfeiffer et al., 2007; Padalon-Brauch et al., 2008) took biocomputational approaches to identify Salmonella-specific sRNAs. Pfeiffer et al. (2007) searched for ‘orphan’ pairs of σ70-type promoters and ρ-independent transcription terminators in the IGRs of the Salmonella LT2 genome, and predicted 46 candidates (STnc10 through STnc460) of sRNA genes that were absent from E. coli K12, but mostly present in the early branching Salmonella species, S. bongori. This screen discovered the first sRNA from an enterobacterial pathogenicity island, i.e. the 80 nt InvR RNA (Fig. 1) that is expressed from the invasion gene locus, SPI-1 (Pfeiffer et al., 2007).
Padalon et al. (2008) explored the genetic islands of Salmonella, i.e. those IGRs that were > 100 bp and showed < 80% identity to their most similar sequence in E. coli K12. The predictions, which were largely based upon orphan ρ-independent terminators, resulted in 28 sRNA candidate genes. Expression of 19 sRNAs, now denoted Isr (A, B, etc.) for island-encoded sRNA, was verified by Northern blot analysis of a large panel of growth conditions reminiscent of the environments encountered by Salmonella upon host cell infection. In addition, several validated sRNAs were shown to be differentially expressed upon Salmonella infection of macrophages.
The function of these sRNAs is as yet to be elucidated. A significant number of them overlap with the 5′ or 3′ ends of open reading frames, and modulate the expression of these flanking open reading frames or are in turn affected by those same genes. IsrJ sRNA (74 nt) encoded between STM2614 and STM2616 is a very promising candidate for the study of sRNA-mediated control of virulence: IsrJ synthesis requires the major SPI-1 transcription factor, HilA and an isrJ mutant is impaired in host cell invasion and effector translocation (Padalon-Brauch et al., 2008).
A network of sRNAs controls the biogenesis of Salmonella OMPs
Regulation of OMP biogenesis is the best understood function of enterobacterial sRNAs (Guillier and Gottesman, 2006; Vogel and Papenfort, 2006), and is tightly linked to two prominent phenotypes of Salmonella hfq mutants, i.e. overproduction of OMPs and concomitant constitutive induction of the σE-controlled envelope stress response (Figueroa-Bossi et al., 2006; Sittka et al., 2007; Bossi et al., 2008). Figure 2A shows the currently known network of Hfq-dependent sRNAs that act to repress Salmonella omp genes at the post-transcriptional level. The expression of these sRNAs is controlled by diverse transcription factors, i.e. σE, HilD, Crp and OmpR, and activated by a specific stress or growth condition.
Figure 2. sRNA-mediated control of extracytoplasmic protein production in Salmonella. A. An emerging network of sRNAs that control OMP (blue circles) expression in Salmonella. Regulatory sRNAs are shown in orange circles within a schematic drawing of a Salmonella cell (the thick black or grey lines indicate outer or inner membrane respectively). Transcription factors of sRNA genes that are associated with membrane perturbation (OmpR, σE) are shown in black circles, whereas those of general (Crp) or virulence (HilD) regulons are shown in green circles. See text for more details on the OMP targets, and the input signals of the network. B. GcvB sRNA represses mRNAs of the periplasmic substrate proteins of many ABC transporters involved in amino acid uptake. The seven targets that have been validated by in vitro structure probing in Sharma et al. (2007) are shown as blue circles; grey circles denote additional proteins of transport systems or with roles in amino acid biosynthesis, and whose synthesis has been shown to be directly regulated by GcvB in Salmonella (C.M. Sharma et al., in preparation). The preferred amino acid, di- or oligopeptide substrates of relevant periplasmic binding proteins are shown in yellow circles. Control of the gcvB gene by two transcription factors (GcvR, GcvA) of the glycine cleavage system is indicated based on extensive work in E. coli by the Stauffer laboratory (see e.g. Urbanowski et al., 2000).
Download figure to PowerPoint
The 80 nt RybB sRNA (Fig. 1), which is transcriptionally induced by the envelope stress sigma factor, σE (Papenfort et al., 2006), is the most globally acting OMP regulator of Salmonella (Fig. 2A). It represses the synthesis of all major porins (OmpA/C/D/F) and many minor OMPs (Papenfort et al., 2006) by base-pairing with the 5′ UTRs or coding regions of omp target mRNAs; these RNA interactions are generally short (7–16 bp) and imperfect, and involve the conserved 5′ end of RybB sRNA (Bouvier et al., 2008; F. Mika et al., submitted). The 74 nt MicA sRNA is also activated by σE, and represses the mRNAs of two Salmonella porins, OmpA and LamB, by antisense pairing similarly to RybB (Udekwu et al., 2005; Figueroa-Bossi et al., 2006; Papenfort et al., 2006; Bossi and Figueroa-Bossi, 2007).
The σE response counteracts the accumulation of unfolded OMPs in the periplasm, and the role of MicA and RybB within this regulon is to halt OMP synthesis when porin production threatens outer membrane homeostasis (Papenfort et al., 2006). Hfq is mandatory for the stability and function of the two sRNAs (Papenfort et al., 2006; Viegas et al., 2007; Sittka et al., 2008) and, therefore, the loss of MicA/RybB-mediated omp mRNA repression might partly explain the chronic σE stress observed in Salmonella hfq mutants (Figueroa-Bossi et al., 2006; Sittka et al., 2007). In more general terms, the analyses of MicA and RybB introduced novel approaches to sRNA target identification in Salmonella, i.e. sRNA pulse-expression combined with global transcriptome profiling (Papenfort et al., 2006), and chromosomal mutagenesis of an sRNA gene (Bossi and Figueroa-Bossi, 2007), both of which will help unravel other sRNA-based circuits.
Similar to InvR sRNA, which will be covered further below, the ∼86 nt CyaR sRNA regulates a single OMP (Fig. 2A). CyaR targets the ompX mRNA encoding a small abundant porin that is highly overproduced in hfq mutants (Sittka et al., 2007). Until very recently, sRNA regulators were known for virtually every abundant porin except OmpX (Guillier and Gottesman, 2006; Vogel and Papenfort, 2006). Yet, an Hfq association of ompX mRNA in both Salmonella and E. coli (Zhang et al., 2003; Sittka et al., 2008) correctly predicted the existence of a cognate Hfq-dependent sRNA shared by these two species. CyaR was then identified in a ‘reverse target search’ approach by screening 35 Salmonella strains with mutations inactivating conserved sRNA genes for effects on OmpX synthesis (Papenfort et al., 2008).
Two features of CyaR are particularly interesting. First, cyaR expression is tightly controlled by the cyclic AMP receptor protein, CRP, which represents a link between porin repression and nutrient availability. Second, the target interaction regions of sRNAs are typically single-stranded. In CyaR, however, a conserved RNA hairpin featuring a C-rich apical loop (Fig. 1A) acts to sequester the Shine–Dalgarno sequence of ompX mRNA and to inhibit translational initiation.
Based on conservation of E. coli sRNA–target interactions, the MicC and MicF sRNAs (Mizuno et al., 1984; Chen et al., 2004) have been integrated into the OMP-regulatory network of Salmonella (Fig. 2A). Experimental screens similar to that for RybB targets (Papenfort et al., 2006) have not only confirmed MicC and MicF action in Salmonella, but also identified OmpD as an additional MicC target (V. Pfeiffer et al., in preparation). OmpD is the most abundant Salmonella porin, and its deregulation the primary cause of σE stress in hfq mutants (Bossi et al., 2008). It should now come as no surprise that a total of three Hfq-dependent sRNAs (InvR, MicC, RybB; Fig. 2A) control OmpD synthesis at the post-transcriptional level.
Control of ABC transporter expression by GcvB sRNA
Some of the above sRNAs regulate the synthesis of multiple OMPs, i.e. proteins with a shared cellular destination. Similarly, the Salmonella homologue of E. coli GcvB sRNA (Urbanowski et al., 2000) was shown to directly target seven mRNAs of periplasmic binding proteins of ABC uptake systems for amino acids and peptides (Sharma et al., 2007 and Fig. 2B). The ability of the 200 nt GcvB sRNA to interact with multiple mRNAs is based upon a ∼30 nt stretch of G and U residues (Fig. 1A); this G/U stretch recognizes extended C/A-rich elements in GcvB target mRNAs. The C/A elements overlap with the RBS of some targets, but are found far upstream of the start codon of others. In either case, the binding of GcvB prevents ribosome access to the mRNA, thus downregulating the target. Independent of GcvB action, the C/A elements have a stimulatory effect on ABC transporter mRNA translation (Sharma et al., 2007). Thus, multiple mRNA targeting by GcvB seems to have evolved through hijacking a translational enhancer element shared by numerous mRNAs of periplasmic transporters.
The work on Salmonella GcvB revealed novel aspects of sRNA action (Sharma et al., 2007). Not only did this sRNA commonly repress mRNAs at a ‘non-canonical’ position, i.e. much upstream of the RBS, but the same region in GcvB interacted with all identified target mRNAs. This region, i.e. the G/U element, is readily visible in alignments of GcvB homologues of a vast number of bacteria. Thus, an ultra-conserved region of the sRNA determines target mRNA regulation, and the recognition of this principle remedies some of the previous difficulty of sRNA target prediction (Vogel and Wagner, 2007).
GcvB is specifically expressed in fast-growing Salmonella (Sharma et al., 2007), i.e. when nutrients are plentiful, and the repression of ABC transporter synthesis by GcvB may help optimize amino acid uptake under this condition. Recent work has revealed an extended post-transcriptional regulon of GcvB, which also includes repression cycA mRNA encoding the major glycine transporter (Fig. 2B and C.M. Sharma et al., in preparation). As the gcvB gene is controlled by transcription factors of the glycine cleavage system (Urbanowski et al., 2000), the downregulation of CycA synthesis by GcvB might facilitate a feedback loop in the expression of ABC transporters.
Small RNAs mediate cross-talk of the Salmonella core genome and virulence regions
The Salmonella invasion gene island, SPI-1, has been one of the most intensely studied bacterial virulence regions. Following its original discovery almost 20 years ago, both its overall sequence and its ∼35 genes encoding the SPI-1 T3SS structural components, chaperones, effector proteins or transcription factors were subject to much scrutiny. Perhaps owing of the traditional focus on proteins, the SPI-1 gene encoding InvR RNA (Fig. 1 and Pfeiffer et al., 2007) remained unnoticed.
The invR gene, located at the right SPI-1 border, was presumably acquired along with the island very early at (or soon after) the divergence of E. coli and Salmonella. This sRNA gene is coexpressed with other SPI-1 genes under conditions that favour host cell invasion, and directly controlled by HilD, the protein that acts at the top of the SPI-1 transcription factor cascade (Fig. 3). The results of Hfq co-immunoprecipitation (co-IP) and deep sequencing analysis showed that InvR is the most abundant Hfq-associated RNA under conditions of host cell invasion (Pfeiffer et al., 2007; Sittka et al., 2008).
Figure 3. Example of sRNA-mediated cross-talk of Salmonella virulence region with core genome expression. Proposed model of porin repression by InvR sRNA (Pfeiffer et al., 2007). Two component systems encoded by the Salmonella core genome sense environmental signals that lead to activation of the SPI-1 transcription factor cascade (HilD, HilC, HilA, InvF), and subsequently, to the expression of the SPI-1 T3SS. HilD also activates expression of the SPI-1-borne invR gene. Together with the RNA chaperone Hfq, InvR acts post-transcriptionally to repress synthesis of the major OMP, OmpD, which is encoded by the Salmonella core genome. One might speculate that at times the repression of the most abundant porin by InvR might have aided the successful establishment of the membrane-spanning SPI-1 T3SS after lateral acquisition in the Salmonella lineage.
Download figure to PowerPoint
Despite what genomic origin and expression intuitively suggest, InvR is not directly involved in SPI-1 regulation or T3SS function. In contrast, InvR affects gene expression of the core Salmonella genome by acting as a post-transcriptional repressor of ompD mRNA (InvR targets the ompD coding region). The biological impact of this SPI-1-mediated repression of OmpD synthesis is not yet understood. However, one might speculate that, as invR is conserved in the early branching S. bongori species, the repression of the most abundant Salmonella porin by InvR might have aided the successful establishment of the membrane-spanning SPI-1 T3SS after horizontal acquisition in the Salmonella lineage (Pfeiffer et al., 2007). Note that OmpD overproduction destabilizes the envelope and abrogates SPI-1 effector secretion, and these phenotypes are successfully complemented by InvR expression (V. Pfeiffer and J. Vogel, unpublished).
IsrE sRNA (98 nt) represents another example of cross-talk function. It is an island-encoded paralogue of RyhB, a Fur-regulated sRNA of the Salmonella core genome (Vogel and Sharma, 2005; Ellermeier and Slauch, 2008; Padalon-Brauch et al., 2008). In E. coli, RyhB is an important regulator of iron homeostasis, and one of its function is the post-transcriptional repression of the core genome-encoded sodB mRNA under iron starvation conditions (Massé and Gottesman, 2002). In Salmonella, IsrE and RyhB act redundantly to downregulate sodB mRNA (Ellermeier and Slauch, 2008) and additively to slow down growth under iron-deplete conditions (Padalon-Brauch et al., 2008).
The Salmonella core genome extensively controls the virulence regions at the transcriptional level, e.g. by two-component systems that time the expression of invasion genes. The case of InvR shows that such cross-talk also takes place post-transcriptionally, through an Hfq-dependent sRNA. If true, then it might also work the other way around, and core genome-encoded sRNAs might control horizontally acquired virulence factors.
In line with this prediction, SgrS sRNA, which acts to combat phosphosugar stress in E. coli, regulates the production of a secreted effector protein, SopD, in Salmonella (K. Papenfort et al., in preparation). Intriguingly, the same region of SgrS involved in antisense targeting of mRNAs relevant to phosphosugar stress, e.g. the ptsG mRNA, is reused to modulate protein synthesis from sopD mRNA. This suggests that bacteria selectively employ their existing repertoire of sRNA regulators of general stress responses to co-ordinate the expression of laterally acquired virulence genes.
Small RNAs antagonize the global regulator, CsrA
Cross-talk of the Salmonella core genome with virulence regions is also facilitated by the CsrB (360 nt) and CsrC (240 nt) sRNAs. However, these two conserved sRNAs do not directly act upon mRNAs, yet commonly sequester (through GGA sequence motifs) the abundant RNA-binding protein, CsrA, which in turn directly modulates mRNA translation (Babitzke and Romeo, 2007). CsrA is a global regulator of gene expression, and its absence deregulates > 8% of all Salmonella genes (Lawhon et al., 2003). By a yet unknown mechanism, CsrA also controls the SPI-1-borne invasion genes, and is ultimately required for successful host cell invasion (Altier et al., 2000). In this circuit, the pool of free CrsA protein needs to be tightly controlled, as is evident from the attenuated host cell invasion phenotype caused by both deletion and overexpression of csrA (Altier et al., 2000); this is where the Csr sRNAs come into play.
A double csrB csrC deletion mutant is markedly impaired in host cell invasion in vitro (Fortune et al., 2006) and ∼100-fold attenuated in mouse infections (D. Becker and D. Bumann, pers. comm.), presumably due to hyperactivity of the CrsA protein. Intriguingly, however, the single csrB or csrC mutants do not show these phenotypes, suggesting redundancy of function, at least in the control of invasion gene expression.
The possible redundancy of csrB and csrC touches upon a key issue of the identification of sRNA functions in Salmonella. Thus far, none of the many global mutagenesis screens have discovered a sRNA gene as the cause of altered Salmonella virulence. However, these screens scoring the effects of individual gene disruption did not take into account that the function of individual sRNAs could be either backed up by another homologue, or masked by the action of a redundantly acting regulatory protein.
Riboswitches sense metabolite and magnesium concentrations
Many metabolic genes in bacteria are controlled at the RNA level by cis-acting riboswitches. Salmonella has been predicted to encode riboswitches for the sensing of cobalamin (vitamin B12), molybdenum cofactor, flavin mononucleotide and thiamine pyrophosphate (R. Breaker, pers. comm.). Of these, the cobalamin riboswitch located in the leader of cob operon mRNA has been exceptionally well investigated (Ravnum and Andersson, 2001 and references therein). The cob 5′ UTR is unusually long (462 nt) and highly structured. In the absence of cobalamin, the long-range interaction of an RNA element encoded approximately 100–250 bp upstream of AUG suppresses the formation of a short local RNA hairpin that would normally sequester the cob RBS to inhibit cob mRNA translation (Fig. 4A).
Figure 4. Post-transcriptional control by cis-encoded RNA sensors in Salmonella. A. A riboswitch mechanism for vitamin B12 (cobalamin) sensing by the first 5′ UTR of cob operon mRNA. Binding of cobalamin disrupts a long-range RNA interaction (red dotted line) that in the absence of cobalamin stabilizes a hairpin that sequesters the Shine–Dalgarno sequence of cob, and represses cob translation. The yellow structure denotes a ribosome. B. A sensor of Mg2+ concentration resides in the 5′ UTR of mgtA, encoding a membrane-standing Mg2+ transporter. In low-Mg2+ environment (extracytoplasmic Mg2+), the PhoPQ two-component systems directly activates mgtA transcription. However, if cytoplasmic Mg2+ are still high, Mg2+ binding to the mgtA 5′ UTR results leads to the formation of hairpin structure that attenuates mgtA transcription. Once the cytoplasmic Mg2+ concentration drops as well, the formation of an alternative stem-loop structure allows transcriptional readthrough into the mgtA coding region, and therefore MgtA protein synthesis. Internalization of Mg2+ by MgtA increase cytoplasmic Mg2+, which in turn attenuates transcription of full-length mgtA mRNA. Adapted from Cromie et al. (2006). C. Thermo-control of translation by an RNA thermometer contained in the 5′ UTR of Salmonella agsA mRNA that encodes a small heat shock protein. At regular growth temperature, a stable RNA hairpin sequesters the agsA Shine–Dalgarno to repress translation. Shift to non-permissive temperature activates both agsA transcription (by the heat shock factor, σ32) and translation by partial melting of the Shine–Dalgarno hairpin.
Download figure to PowerPoint
Cobalamin binding to the upstream RNA element disrupts the long-range interaction (Nahvi et al., 2004). As a result, the RBS hairpin forms and cob mRNA translation is switched off. The 25 cob operon genes, which are required for de novo synthesis of cobalamin, are translationally coupled. Thus, the RNA sensor of cobalamin located upstream of the first cob cistron can feedback-control the entire cob-encoded cobalamin synthesis pathway. The btuB mRNA encoding an OMP needed for import of extracellular cobalamin is feedback-regulated by a similar mechanism (Ravnum and Andersson, 1997; Nahvi et al., 2004).
RNA switches of Salmonella are not limited to sensing metabolites and performing translational control. Work on the mgtA mRNA, which encodes an Mg2+ transporter, revealed that its 5′ UTR served as a RNA sensor of Mg2+ (Cromie et al., 2006). Depending on the intracellular Mg2+ concentration, the 5′ UTR of mgtA adopts different stem-loop structures that allow or prohibit transcriptional readthrough into the mgtA coding region (Fig. 4B). Considering that Mg2+ is commonly required for proper RNA folding, its discovery as the specific determinant of a RNA switch mechanism provided a radically new example of RNA-mediated control in bacteria. In the context of mgtA, the RNA-mediated control of transcription elongation by cytoplasmic Mg2+ complements the tight control of transcription initiation by extracytoplasmic Mg2+ facilitated by the PhoP/Q two-component system (Cromie et al., 2006). Thus, the same ligand is sensed in different cellular environments to regulate distinct steps in gene transcription. Note that Mg2+ also seems to influence mgtA mRNA decay in an RNase E-dependent manner (Spinelli et al., 2008).
Mg2+ is a key signal in Salmonella virulence gene expression (Ohl and Miller, 2001), and it is an intriguing possibility that Mg2+-dependent RNA switches control additional Salmonella genes post the initiation of transcription. More generally, we must determine whether other specialized riboswitches sense substances emitted by host cells in the course of infection.
Sensing temperature with RNA
Thermosensing as an input function in the control of Salmonella gene expression is well documented at the level of proteins, e.g. TlpA, a DNA-binding protein encoded by the Salmonella virulence plasmid (Hurme et al., 1997). Yet, temperature is also sensed directly by Salmonella messengers, e.g. the ibpA and agsA mRNAs, which contain so-called RNA thermometers (Waldminghaus et al., 2005; 2007).
The 5′ UTR of agsA, which encodes a small heat shock protein, contains one such RNA thermometer (Fig. 4C). A stable hairpin blocks the Shine–Dalgarno sequence of agsA mRNA by base-pairing, and represses translation at normal growth temperature (37°C). A shift to a non-permissive temperature, i.e. heat shock (42°C), is predicted to open the hairpin, which is measurable by increased translation of an agsA reporter fusion. The results of in vitro RNA structure probing and 30S ribosome toeprinting experiments have fully supported this prediction, for example, showing that a stable translational initiation complex forms on agsA mRNA at 45°C but not at 30°C (Waldminghaus et al., 2007).
The agsA gene is also activated by heat shock at the transcriptional level. Thus, in contrast to the two-layer repression of mgtA by Mg2+ (see above), the agsA gene is activated by the same stimulus at both the DNA and the RNA level.
More roads to travel in the Salmonella RNA world
This brief review can only give a rough overview of the many roles that RNA may have in the control of Salmonella gene expression. For example, given that we have only glimpsed at Hfq functions, new pathways involving Hfq-dependent sRNAs will certainly continue to be discovered. High-throughput sequencing of Hfq-associated RNA species has identified > 700 mRNAs from diverse cellular pathways that interact with this protein in vivo, including many messengers from the Salmonella virulence regions (Sittka et al., 2008). This catalogue of putative sRNA targets provides departure points for improving in silico predictions and functional screens to identify new sRNA-based regulatory circuits. The method itself, which involves co-IP of RNA along with epitope-tagged Hfq protein and subsequent cDNA pyrosequencing, is general and should facilitate similar studies in many other pathogens. In Salmonella, the current Hfq co-IP data set is limited to one growth condition, and is definitely worth extending to more Salmonella-relevant stress conditions and infections of in vitro cultured cells or even animals.
A collection of Salmonella mutant strains for conserved sRNA genes (Papenfort et al., 2008) and for major factors relevant for sRNA processing (Viegas et al., 2007), combined with well-established methods to find and validate sRNA–target interactions in this organism (Papenfort et al., 2006; 2008; Bossi and Figueroa-Bossi, 2007), should also help test the extent to which sRNA functions in Salmonella are shared with other bacteria. There is evidence that even among Salmonella isolates, sRNA expression and thus function can radically differ. For example, the commonly used 14 028 s strain does not seem to produce the conserved DsrA and RprA sRNAs (Jones et al., 2006), two translational activators of rpoS mRNA in E. coli (Repoila et al., 2003). However, both sRNAs are produced and are functional as rpoS activators in the mouse-adapted SL1344 strain (Sittka et al., 2008 and K. Papenfort et al., in preparation).
The overall number of Salmonella sRNAs is likely to grow. First, most candidates identified in the screens by Pfeiffer et al. (2007) and Sittka et al. (2008) await experimental testing, and even the many growth conditions assayed by (Padalon-Brauch et al., 2008) might not have been sufficient to capture the expression of all currently predicted sRNAs. Second, several hundreds of Salmonella IGRs that correspond to sRNA candidate loci of E. coli (Hershberg et al., 2003; Livny et al., 2006) are yet to be validated in either of these two organisms. Essentially the same applies to riboswitches and RNA thermometers predicted in organisms closely related to Salmonella.
Last but not least, in a cellular environment, non-coding RNA is unlikely to be ‘naked’ and is more likely to be associated with RNA-binding proteins throughout its lifetime. Aside from the here-reviewed Hfq and CsrA proteins, work in Salmonella keeps discovering new involvements of RNA-binding proteins in post-transcriptional regulation. The latest examples of this include FljA, a protein long thought to be a transcription repressor in the flagellar gene expression cascade. It now turns out that FljA acts at the RNA level, by blocking the translation of fliC mRNA through specific binding to this messenger (Aldridge et al., 2006). Thus, the future analysis of Salmonella RNA-binding proteins has great potential to understand better RNA-based circuits in this model pathogen.