The small RNA SgrS: roles in metabolism and pathogenesis of enteric bacteria

Bacteria adapt to ever-changing habitats through specific responses to internal and external stimuli that result in changes in gene regulation and metabolism. One internal metabolic cue affecting such changes in Escherichia coli and related enteric species is cytoplasmic accumulation of phosphorylated sugars such as glucose-6-phosphate or the non-metabolizable analog α-methylglucoside-6-phosphate. This “glucose-phosphate stress” triggers a dedicated stress response in γ-proteobacteria including several enteric pathogens. The major effector of this stress response is a small RNA (sRNA), SgrS. In E. coli and Salmonella, SgrS regulates numerous mRNA targets via base pairing interactions that result in alterations in mRNA translation and stability. Regulation of target mRNAs allows cells to reduce import of additional sugars and increase sugar efflux. SgrS is an unusual sRNA in that it also encodes a small protein, SgrT, which inhibits activity of the major glucose transporter. The two functions of SgrS, base pairing and production of SgrT, reduce accumulation of phosphorylated sugars and thereby relieve stress and promote growth. Examination of SgrS homologs in many enteric species suggests that SgrS has evolved to regulate distinct targets in different organisms. For example, in Salmonella, SgrS base pairs with sopD mRNA and represses production of the encoded effector protein, suggesting that SgrS may have a specific role in the pathogenesis of some γ-proteobacteria. In this review, we outline molecular mechanisms involved in SgrS regulation of its target mRNAs. We also discuss the response to glucose-phosphate stress in terms of its impact on metabolism, growth physiology, and pathogenesis.

Over the last decade, small RNAs have emerged from relative obscurity to take their places as central regulators of gene expression in organisms from all three domains of life. While hundreds of small RNAs in dozens of bacterial genomes have been identified by computational or experimental methods, the functions of the vast majority of these remain a mystery. Nevertheless, we have learned a great deal about a small number of "model" bacterial sRNAs, including how their production is regulated, what targets they in turn regulate and the physiological outcomes of sRNA-mediated regulation. In this review, we first provide a brief overview of features of regulatory sRNAs that act on mRNAs through base pairing interactions. We then focus on one wellcharacterized sRNA, SgrS (sugar transport related sRNA) and describe its activities on target mRNAs and how these activities regulate bacterial metabolism and pathogenesis.

MECHANISMS OF REGULATION BY BACTERIAL SMALL RNAs BASIC CHARACTERISTICS OF sRNA-MEDIATED REGULATION
Several mechanistic classes of sRNAs have been identified in diverse bacterial species. Many characterized sRNAs act by base pairing with mRNA targets to control mRNA stability and translation. Such sRNAs are often between 50 and 300 nucleotides in length and require an RNA chaperone, Hfq, for their stability and regulatory effects on target mRNAs (Sledjeski et al., 2001;Moller et al., 2002;Zhang et al., 2002). Hfq is a hexameric ring protein with sRNA-and mRNA-binding faces (Mikulecky et al., 2004;Link et al., 2009;Zhang et al., 2013). On sRNAs, Hfq binds to stem-loop terminator structures preceded by A/U rich sequences (Valentin-Hansen et al., 2004;Otaka et al., 2011;Ishikawa et al., 2012). Hfq-binding sites are located in 5 untranslated regions (UTRs) of many mRNAs that are regulated by sRNAs (Soper and Woodson, 2008;Link et al., 2009;Salim and Feig, 2010;Salim et al., 2012). Binding of Hfq to both sRNAs and mRNAs increases their local concentrations, stimulates structural remodeling to facilitate pairing and increases annealing rates of cognate pairs (Fender et al., 2010;Maki et al., 2010;Soper et al., 2010;Hopkins et al., 2011). Other factors involved in sRNA-mRNA regulatory transactions include RNase E and components of the degradosome (Masse and Gottesman, 2002;Masse et al., 2003;Morita et al., 2005). Polynucleotide phosphorylase (PNPase, a 3 -5 exonuclease), RhlB helicase and enolase assemble on RNase E to form a degradosome complex required for bulk mRNA turnover in proteobacteria (Carpousis, 2007). RNase E binds A/U-rich single-stranded regions of RNAs and is responsible for sRNA-mediated mRNA degradation (Carpousis et al., 2009;Belasco, 2010;Prevost et al., 2011).

REPRESSION BY sRNAs
The canonical mechanism of negative regulation by sRNAs involves base pairing interactions that directly inhibit translation initiation because the sRNA sequesters mRNA sequences required for stable ribosome binding, in the region from ∼20 nt upstream to 20 nt downstream of the start codon (Beyer et al., 1994;Huttenhofer and Noller, 1994). As more detailed studies of sRNA regulatory mechanisms have been performed, variations on this theme have been discovered. Binding of sRNAs to mRNA sequences outside the region recognized by the ribosome can still inhibit translation initiation Sharma et al., 2007;Bouvier et al., 2008;Holmqvist et al., 2010), for example, by recruiting Hfq to bind at a site overlapping the ribosome binding site (RBS) (Desnoyers and Masse, 2012) or by sequestering mRNA sequences that apparently act as translational enhancer elements (Bandyra et al., 2012;Desnoyers et al., 2013).
Regardless of the mechanism, sRNA-mediated translational repression is often coupled to mRNA degradation by an RNase E degradosome-dependent pathway. Translating ribosomes protect mRNA from RNase E degradation (Dreyfus, 2009), thus sRNA inhibition of translation unmasks RNase E recognition sites and makes target mRNAs susceptible to degradation (Prevost et al., 2011). Though translational repression and mRNA degradation are typically coupled, this is usually not obligatory for gene silencing. In other words, mutations that abrogate RNase E-dependent turnover of sRNA targets have no impact on translational repression Rice and Vanderpool, 2011;Rice et al., 2012). While less commonly described, in some cases sRNA-mediated translational repression has no significant effect on mRNA turnover (Moller et al., 2002). Conversely, there are a handful of known cases where sRNAs do not directly affect translation but rather specifically target mRNAs for degradation (Desnoyers et al., 2009;Pfeiffer et al., 2009).

ACTIVATION BY sRNAs
sRNAs can also activate gene expression post-transcriptionally. Positive regulation of target mRNAs is achieved via activation of translation or stabilization of the target transcript. Some mRNAs have intrinsic secondary structures in their 5 UTRs that hinder translation, for example, because the RBS is sequestered. Binding of sRNAs to these 5 UTRs can prevent formation of translation-inhibitory secondary structures and therefore activate translation (Morfeldt et al., 1995;Lease et al., 1998;Prevost et al., 2007). Another mechanism of activation by sRNAs involves sRNA-mRNA base pairing that alters accessibility of RNase E recognition sites on mRNA targets. sRNA-mRNA binding can induce cleavage of a target transcript, resulting in a processed mRNA with an accessible RBS (Obana et al., 2010) or alternatively sRNA base pairing with an intrinsically unstable mRNA can occlude an RNase E recognition site and prevent cleavage .
SgrS also represses manXYZ, which encodes a PTS transporter of relatively broad substrate specificity. SgrS repression of manXYZ is carried out through a more complex mechanism compared to ptsG. Two distinct SgrS binding sites on manXYZ mRNA are located within early manX coding sequences and within the manX-manY intergenic region (Figures 1, 2C). SgrS binding at the manX site is responsible for translational repression of manX, but has no effect on translation of manY or manZ (Rice et al., 2012). The intergenic SgrS binding inhibits translation of manY and manZ (translation of manY and manZ is coupled) and this is independent of manX regulation (Rice et al., 2012). SgrS binding at each site individually does not affect manXYZ mRNA stability; pairing at both sites is required for RNase E-dependent degradation of manXYZ mRNA (Rice et al., 2012).
A third SgrS target, yigL mRNA, which encodes a haloacid dehalogenase-like phosphatase (Koonin and Tatusov, 1994), is positively regulated by SgrS . Synthesis of YigL is induced by SgrS in response to glucose-phosphate stress. The yigL gene is in an operon with the upstream pldB, however SgrS activates only yigL. In the absence of SgrS, RNase E processing of the pldB-yigL transcript yields an mRNA that is susceptible to further degradation. When SgrS is produced, it base pairs with a sequence on the processed pldB-yigL mRNA (Figures 1, 2C) and prevents further degradation. This mechanism of positive FIGURE 1 | Current model for the role of SgrS in the glucose-phosphate stress response. The top panel illustrates the main features of the phosphoenolpyruvate phosphotransferase system (PTS), which transports numerous carbohydrates as well as glucose analogs (αMG, 2DG: α-methyl glucoside and 2-deoxy glucose, respectively). Glucose-phosphate stress is associated with accumulation of sugar-phosphates (hexagons with attached green circles). The stress response is initiated by the activated transcription factor, SgrR, which induces SgrS synthesis. SgrS has two functions; the first is base pairing-dependent regulation of target mRNAs (illustrated in lower panel), the second is production of the ∼40 amino acid protein SgrT. SgrT acts to repress activity of the EIICB glc (PtsG) transporter (top panel). The base pairing activity results in repression of two mRNA targets encoding PTS sugar transporters, ptsG and manXYZ, and activation of a third mRNA target encoding a phosphatase, yigL (described in detail in the text). Altogether, the base pairing activity of SgrS on these various targets inhibits further uptake of sugar-phosphates by inhibiting production of sugar transporters and promotes sugar efflux by providing neutral sugar substrates that are pumped out by an unknown efflux pump (indicated by a "?"). regulation is unusual in that an initial processing event is required to allow SgrS access to its binding site because cleavage within pldB frees the pldB region from translating ribosomes . Moreover, activation of yigL by SgrS is translationindependent. SgrS stabilizes yigL mRNA by occluding a specific RNase E cleavage site upstream of the yigL coding region, not by enhancing yigL translation . A similar translation-independent mechanism of mRNA stabilization was recently described for the RydC sRNA-cfa mRNA regulatory pair (Frohlich et al., 2013).

GLUCOSE-PHOSPHATE STRESS PHYSIOLOGY
Targets of SgrS include sugar transporters and a sugar phosphatase. SgrS-mediated repression of sugar transporters diminishes cells' capacity to take up sugars and therefore reduces further phosphosugar accumulation (Figure 1). However, this effect is not immediate: PtsG protein has a half-life of ∼80 min , so merely stopping new synthesis of PtsG would not provide a fast remedy for the problem of phosphosugar accumulation. The activation of YigL synthesis by SgrS addresses this problem since dephosphorylation of sugars allows their efflux (Figure 1) (Winkler, 1971;Haguenauer and Kepes, 1972;Papenfort et al., 2013). Growth competition experiments between wild-type and sgrS mutants provided insight into how regulation of different SgrS targets contributes to stress resistance and growth during glucosephosphate stress . When cells are stressed while growing in rich medium, SgrS-mediated regulation of ptsG mRNA alone is sufficient to confer wild-type levels of growth. In contrast, cells stressed in minimal media are far more growth inhibited, and repression of ptsG alone is not sufficient to rescue growth. In minimal media stress conditions, repression of ptsG and activation of yigL are necessary, but not sufficient for full growth rescue . These findings illustrate the poorly understood influence of nutrient availability on the severity of glucosephosphate stress. Moreover, these results highlight the fact that additional unknown SgrS targets are involved in the stress response. Phosphosugar intermediates of central metabolism provide precursors for biomass and energy, yet, as illustrated by glucosephosphate stress, excessive accumulation of phosphosugars is detrimental to cell growth. Other types of phosphosugar stress also cause growth inhibition or cell lysis (Yarmolinsky et al., 1959;Englesberg et al., 1962;Irani and Maitra, 1977;Lee et al., 2009). In most cases, the mechanisms responsible for phosphosugarassociated inhibition or lysis have not been defined. However, recent work suggests that in some cases phosphosugars themselves are not directly inhibitory. Rather, accumulation of phosphosugars is accompanied by depletion of other metabolites, and stress is ameliorated by supplementation with the limiting metabolites (Lee et al., 2009(Lee et al., , 2013Richards et al., 2013). Glucose-phosphate stress is so far associated with accumulation of a few sugar-phosphate intermediates of upper glycolysis (Morita et al., 2003;Vanderpool and Gottesman, 2004;Sun and Vanderpool, 2013). A recent study implicates depletion of intermediates of lower glycolysis, particularly phosphoenolpyruvate (PEP) as an important cause of glucose-phosphate stress. When αMG is taken up and phosphorylated, it cannot be metabolized to replenish glycolytic intermediates. Thus, PEP utilized to drive αMG uptake is not replaced via glycolytic metabolism. Under these conditions, SgrS regulation of target mRNAs and production of SgrT limits PEP consumption by reducing levels and activity of PtsG (Figure 1). In sgrS mutants, exposure to αMG results in strong growth inhibition (Vanderpool and Gottesman, 2004;Richards et al., 2013) that is largely reversed by supplementing stressed cultures with glycolytic intermediates (Richards et al., 2013). The ratios of PEP to pyruvate seem to be particularly relevant for growth during glucose-phosphate stress. Increasing pyruvate levels during stress results in lysis of sgrS mutant cells, whereas increasing PEP levels rescues cell growth (Richards et al., 2013). The observation that stress (and growth inhibition) is more severe when cells are growing in minimal compared to rich media is also consistent with metabolite depletion as an underlying cause of glucose-phosphate stress. In rich media, cells do not have to synthesize many biosynthetic intermediates. In contrast, growth in minimal media requires de novo biosynthesis of amino acids. Thus, depletion of glycolytic intermediates during glucose-phosphate stress would have more severe effects on growth under conditions where these same intermediates are needed as precursors for biosynthesis. Consistent with this idea, supplementation of minimal media with amino acids improves stress recovery in minimal medium .
The transcription factor SgrR also plays an important, but not fully characterized role in glucose-phosphate stress physiology. SgrR activates expression of sgrS and at least two other genes during glucose-phosphate stress: setA, encoding an efflux pump (Liu et al., 1999;Sun and Vanderpool, 2011), and alaC (formerly yfdZ), a glutamate-pyruvate aminotransferase (Vanderpool and Gottesman, 2007;Kim et al., 2010). The role of alaC in helping cells recover from glucose-phosphate stress is unknown. In contrast, setA, which is encoded just downstream of sgrS, is important for growth recovery under certain stress conditions (Sun and Vanderpool, 2011). Given its function as an efflux pump, the hypothesis that SetA was responsible for export of αMG was tested, but was not supported (Sun and Vanderpool, 2011). Thus, the role of SetA in glucose-phosphate stress also remains elusive.

IDENTIFICATION OF SgrS IN ENTERIC BACTERIA
SgrS homologs were identified in many γ-Proteobacteria, including Escherichia sp., Salmonella sp., Shigella sp., Yersinia sp, Serratia sp., Klebsiella pneumoniae and Erwinia sp. (Horler and Vanderpool, 2009). The sgrR-sgrS intergenic region (containing the sgrS promoter) is highly conserved, suggesting that SgrR regulates sgrS expression in all these organisms. All identified SgrS homologs contain a Rho-independent terminator and most possess an additional stem-loop structure upstream of the terminator; these two structures are important for Hfq binding to SgrS (Figure 2A) (Horler and Vanderpool, 2009;Otaka et al., 2011;Ishikawa et al., 2012). While the overall conservation of SgrS is low, a short stretch of ∼13 nts near the SgrS 3 end is nearly invariant (Figure 2B). This SgrS sequence is complementary to the translation initiation regions of ptsG mRNAs in all species where an SgrS homolog was found (Horler and Vanderpool, 2009). Mutation of residues G176 and G178 within the conserved region of E. coli SgrS abrogates SgrS-mediated repression of ptsG mRNA and prevents recovery from glucosephosphate stress (Maki et al., 2008). Introduction of analogous mutations in the conserved regions of SgrS homologs from Salmonella, E. carotovora, Y. pestis and K. pneumoniae similarly prevented regulation of ptsG (Wadler and Vanderpool, 2009). Regulation of other targets is less well conserved among SgrS homologs. The SgrS sequences required for base pairing with manX are upstream of the conserved region ( Figure 2C) and are poorly conserved among SgrS homologs. SgrS homologs from Salmonella and K. pneumoniae have the same predicted SgrS-manX base pairing interaction and manX translation is regulated as expected. In contrast, E. carotovora and Y. pestis SgrS homologs have changes in the manX pairing site resulting in loss of complementarity to their cognate manX and were accordingly shown not to regulate manX translation (Rice and Vanderpool, 2011).

CONSERVATION OF SgrT
While the exact molecular function of SgrT has not been reported, available data strongly suggest that this small protein interacts directly with PtsG protein to inhibit its activity (Wadler and Vanderpool, 2007). Most SgrS homologs contain open reading frames similar in size to E. coli SgrT (∼40 amino acids) (Horler and Vanderpool, 2009). While the primary amino acid sequence of putative SgrT homologs was not well conserved, homologs from Salmonella, Klebsiella, and Erwinia were functional when expressed in an E. coli sgrST mutant (Wadler and Vanderpool, 2009). Interestingly, some species with SgrS homologs appear to lack a functional SgrT. In Yersinia sp., SgrS appears to be truncated at the 5 end, and SgrS from Yersinia species ranges in size from ∼85 to 140 nt and lacks the sgrT open reading frame. In pathogenic E. coli O157:H7 strains, a point mutation in the SgrS 5 region alters the sgrT start codon, presumably abrogating SgrT production in these strains.
Differential presence and absence of SgrT in organisms that possess SgrS led to a closer comparison of E. coli K12 and Salmonella SgrS sRNAs. In E. coli K12, sgrT alone (without the region of SgrS involved in base pairing with mRNAs) was not sufficient to allow growth rescue during glucose-phosphate stress conditions. This is in part due to very low levels of SgrT produced from the native sgrS allele in E. coli (Wadler and Vanderpool, 2009). E. coli SgrS has a sequence in the 5 region that forms a structure that inhibits sgrT translation. On the other hand, Salmonella SgrS does not have the same inhibitory structure and therefore produces more SgrT than E. coli SgrS (Wadler and Vanderpool, 2009;Balasubramanian and Vanderpool, 2013). While native levels of SgrT production have not been investigated in Erwinia or Klebsiella species, it was observed that ectopic production of SgrT homologs from these organisms in an E. coli sgrST mutant rescued growth during glucose-phosphate stress (Wadler and Vanderpool, 2009). Thus, SgrT is functionally conserved when it is present, but levels of SgrT production vary among bacteria.

SgrS REGULATION OF sopD mRNA
Although SgrS is conserved among enteric bacteria, divergence in primary sequence has resulted in species-specific target regulons, exemplified by the finding that Erwinia and Yersinia SgrS homologs do not regulate their cognate manXYZ homologs (Rice and Vanderpool, 2011). Another instance of species-specific regulation by SgrS is regulation of the Salmonella-specific gene sopD (Papenfort et al., 2012). SopD is an effector delivered to host cells through the Type 3 Secretion Systems (T3SSs) encoded on Salmonella pathogenicity island (SPI)-1 and SPI-2 (Brumell et al., 2003) and it functions as a general virulence factor in mice (Jiang et al., 2004;Bakowski et al., 2007). Regulation of sopD by SgrS involves base pairing interactions between the conserved region of SgrS and the early coding sequence of sopD mRNA ( Figure 2C); the interaction inhibits translation initiation and stimulates sopD mRNA degradation (Papenfort et al., 2012). Interestingly, Salmonella encodes a second SopD protein, SopD2, which shares 42% identity with SopD and likely arose from a duplication (Jiang et al., 2004). The predicted SgrS-sopD2 base pairing interaction differs from SgrS-sopD at only a single position, a wobble G:U base pair instead of the G:C base pair. Remarkably, this interaction that differs by only a single hydrogen bond prevents regulation of sopD2 by SgrS (Papenfort et al., 2012).
While the biological significance of sopD regulation by SgrS is not yet clear, the inclusion of sopD in the Salmonella SgrS regulon illustrates plasticity in the evolution of sRNA regulons. The presence of sgrR-sgrS-sgrT in the same genomic context in pathogenic and non-pathogenic γ-proteobacteria (Horler and Vanderpool, 2009) suggests that this is an ancestral, or "core" RNA among these organisms. Yet, this core sRNA has acquired the ability to regulate a gene that was horizontally acquired by Salmonella. Studies of other SgrS homologs in pathogenic and non-pathogenic enteric bacteria will surely shed light on the breadth of regulatory activities of this fascinating dual-function sRNA.