Control of rpoS transcription in Escherichia coli and Pseudomonas: why so different?


E-mail; Tel. (+39) 040 3757317; Fax (+39) 040 226555.


In Escherichia coli, the stationary phase alternative sigma factor σs controls the expression of genes involved cell survival in response to cessation of growth (stationary phase) and provides cross-protection to various stresses. Levels of σs increase dramatically at the onset of stationary phase and are regulated at the transcriptional, post-transcriptional and post-translational level, making this one of the most complex regulatory systems in bacteria. The basic mechanisms for the control of translation and σs proteolysis have been understood. However, studies on the transcriptional control in E. coli lag behind and are controversial. The cAMP-CRP complex and the two component BarA/UvrY system have been implicated and, ppGpp and polyphosphate appear to have a signalling role. σs has also been reported to be a general stress regulator in the fluorescent pseudomonads (Pseudomonas aeruginosa, P. fluorescens and P. putida) and recent studies on σs regulation highlight that transcriptional regulation in these bacteria apparently plays a major role. Global regulatory systems, the GacA/GacS two component system and quorum sensing all affect rpoS expression, as does the TetR family PsrA regulator that directly binds to- and activates the rpoS promoter in stationary phase. This striking difference in regulation between E. coli and Pseudomonas can be partly attributed to the differences in the functional role of σs in the two bacterial species. This report will review mainly recent studies on rpoS transcriptional regulation and will try to rationalize the current knowledge into a working model.


In their natural environment, bacteria are often challenged by constantly changing nutrient availability and by exposure to various forms of physical stress, including osmotic, oxidative and temperature shock. Exposure to starvation and stresses leads to a reduction or cessation of growth, resulting in a major switch in gene expression that allows the cells to cope with the new conditions. A very simple and effective mechanism employed by bacteria to bring about such a major switch in gene expression is the use of alternative sigma factors that alter RNA polymerase core specificity (Ishihama, 2000).

The stationary phase alternative sigma factor σs (also called σ38 and RpoS), first described in Escherichia coli, has now also been identified in many non-enteric bacteria belonging to the γ-subdivision of proteobacteria. The term stationary phase refers to a fixed physiological state regardless of what factors and/or environmental conditions led to cessation of growth. In E. coli, σs regulates more than 100 genes involved in cell survival, cross protection against various stresses and in virulence (Loewen et al., 1998; Ishihama, 2000). This set of genes is called the σs regulon and has been mainly studied in E. coli but σs has recently been described in the fluorescent pseudomonads as well, where it was shown to be a regulator in Pseudomonas putida, P. aeruginosa and P. fluorescens (Sarniguet et al., 1995;Ramos-Gonzalez and Molin, 1998; Suh et al., 1999).

The levels of σs are carefully controlled, increasing dramatically at the onset of stationary phase, when they reach 30% the level of the housekeeping σ70 subunit, thus increasing its ability to compete with other available σ subunits for core RNA polymerase (Jishage et al., 1996). The mechanisms governing σs levels are still the subject of extensive investigation in E. coli. These studies have revealed one of the most complex regulatory mechanisms in bacteria, regulation taking place at the level of transcription, translation and protein stability, all co-ordinated by the response to several stress signals (Hengge-Aronis, 2002).

Whereas the basic mechanisms for the control of rpoS translation and σs proteolysis are understood, the transcriptional regulation of rpoS in E. coli, remains the subject of some controversy. Several regulators and small molecules have been implicated in increasing rpoS transcription upon entry into stationary phase. Regulation of rpoS has also been recently addressed in the fluorescent pseudomonads, where transcriptional regulation apparently plays the major role. This review will focus on the transcriptional regulation of the rpoS gene, emphasizing the similarities and differences between E. coli and the pseudomonads as well as rationalizing the conflicting evidence in the literature.

RpoS-dependent regulation in Pseudomonas versus Escherichia coli

The fluorescent pseudomonads can colonize multiple habitats and, consequently, are found in many different environments including soil, water, plants, animals and human infections. This adaptability is also reflected by the large number of ORFs and putative transcriptional regulators found in the annotated Pseudomonas genomes. For example, 24 putative sigma factors have been detected in P. aeruginosa and in P. putida (in comparison, E. coli has only seven sigma factor encoding genes), the highest number in all of the 50 annotated bacterial genomes (Martinez-Bueno et al., 2002). Pseudomonas putida and P. fluorescens are often beneficial root-colonizing bacteria whereas P. aeruginosa is frequently a human opportunistic pathogen and, consequently, many of the sigma factors are probably specifically associated with ability to colonize and survive a specific habitat. σs is a central regulator in the general stress response in E. coli, where it controls the expression of at least 100 genes in stationary phase (Ishihama, 2000 and references therein). In Pseudomonas on the other hand, σs has unique functions and has a less important general stress regulatory role (Jorgensen et al., 1999; Suh et al., 1999). In P. aeruginosa, for example, σs is involved in the production of extracellular alginate and exotoxin A (Suh et al., 1999), whereas in P. fluorescens, σs is implicated in the production of antibiotics and in suppressing soilborne plant pathogens (Sarniguet et al., 1995). The degree of requirement of σs for stress resistance following exposure to heat, low pH, high osmolarity, hydrogen peroxide and ethanol is less pronounced in P. aeruginosa than is in E. coli. Stationary phase σs-negative P. aeruginosa is more stress resistant than exponentially growing σs-positive cells, indicating that other factors associated with entry into stationary phase contribute to stress tolerance (Jorgensen et al., 1999).

It appears therefore σs is less important in stress survival in Pseudomonas and has more specific roles related to virulence and colonization that must be integrated into the overall cell-density dependent expression of virulence determinants. In fact studies thus far on the transcriptional regulation of rpoS in Pseudomonas have highlighted that it is influenced by global regulators involved in the regulation of virulence in response to cell-density and other yet unknown signals (see below).

Brief overview of regulation of rpoS translation and σs proteolysis in E. coli

The underlying mechanisms governing translational regulation and σs stability in E. coli have been reviewed recently and will be discussed only briefly here (Loewen et al., 1998; Hengge-Aronis, 2002; Repoila et al., 2003 The rpoS transcript is monocistronic with a long untranslated region of 567 bp. The regulation of translation was first demonstrated using rpoS-lacZ translational fusions, translation being induced by transition into stationary phase, by osmotic shock, by low temperature and by a shift to acidic pH. The secondary structure of rpoS mRNA together with trans-acting factors and small regulatory RNAs control translation under the different stress conditions. The RNA binding Hfq protein (also known as HF-I) stimulates translation by binding to rpoS mRNA secondary structure with the participation of additional factors like the DsrA RNA regulator (see below). The RNA-binding histone-like protein H-NS, on the other hand, has been implicated in negative translational regulation, as hns mutants display higher σs levels. The precise mode of action of H-NS is currently unclear, it might exert its regulatory role through counteracting positive factors like Hfq or through the nucleoid protein HU that is known to bind the rpoS leader region, thereby stimulating translation. These RNA binding proteins control rpoS translation in concert with the small regulatory RNAs DsrA, RprA and OxyS. DsrA is a 87-nucleotide RNA that is made at low temperatures, is stabilized by Hfq and functions as an antiantisense element in rpoS mRNA, disrupting intramolecular base pairing and promoting translational initiation at low temperatures. In addition, DsrA also interacts with hns leader RNA, exerting a negative effect on hns translation. RprA is another small (106 nucleotide) RNA that was identified by its ability, when present in multiple copies, to suppress dsrA mutations by stimulating rpoS translation. It also has a positive effect under osmotic shock and it is currently unknown if it, like DsrA, acts by base pairing to the rpoS leader. OxyS is a 109-nucleotide regulatory RNA that inhibits rpoS translation by binding Hfq in competition with the rpoS leader.

The major players in the complex rpoS translational control have therefore been identified, creating a network of positive and negative elements that interact with each other. Future aims will include understanding the fine-tuning of interactions and regulatory loops as well as the precise intracellular signals important for this control since, thus far, temperature, which acts through DsrA, is the only defined signal.

The increase in σs levels in stationary phase results in part from a substantial increase in its stability. The instability of σs in exponential growing phase is due to the activity of the ClpXP protease, which recognizes a 20 amino acid stretch between residues 170 and 190. The response regulator RssB is required for this degradation pathway; phosphorylated RssB directly interacts with σs and delivers it to the ClpXP protease complex. Again, the intracellular signal(s) involved in this control is(are) unknown and they could act at various steps in the degradation, such as phosphorylation or cellular levels of RssB.

rpoS promoters and evidence of transcriptional regulation

The rpoS gene has been identified in the genomes of several bacteria belonging to the γ-subdivision of proteobacteria, where it is immediately downstream nlpD, a lipoprotein structural gene whose expression is not stationary phase-induced. The rpoS promoter regions of E. coli, P. aeruginosa and P. putida all have one major promoter located within the nlpD gene. In E. coli, this promoter, called rpoSp or rpoSp1, initiates a monocistronic rpoS mRNA transcript comprising a 567 bp untranslated region and typical σ70 dependent − 10 and − 35 promoter sequences. Deletion and operon fusion studies demonstrated that transcription increases upon entry into stationary phase in rich LB medium (Lange et al., 1995). However, changes in transcriptional fusion expression in E. coli do not accurately reflect changes in rpoS mRNA levels during transition to stationary phase because the half-life of the mRNA increases (Zgurskaya et al., 1997).

Promoter studies performed in P. putida identified the transcription initiation site 373 bp upstream from the translational start codon, again with typical σ70 dependent − 10 and − 35 promoter sequences. Transcriptional fusions revealed significantly increased expression upon entry into stationary phase (Kojic and Venturi, 2001; Kojic et al., 2002). In P. aeruginosa, the promoter initiates transcription 366 bp from the translational start and again has a typical σ70 dependent promoter. RNA studies revealed a single mRNA band that was highly induced in stationary phase (Fujita et al., 1994). The untranslated nucleotide region of the rpoS transcripts of P. putida and P. aeruginosa are almost identical (> 90%) and the P. putida rpoS promoter is just as active and inducible in P. aeruginosa as in the original strain (Kojic and Venturi, 2001). However, the P. putida promoter displayed no activity in any phase of growth in E. coli (Kojic and Venturi, 2001), indicating that the E. coli transcription machinery cannot drive transcription from this promoter. Several transcriptional regulators and ‘alarmone’ signal molecules (listed in Table 1) are implicated in regulating these promoters as discussed below.

Table 1. . Genetic loci implicated in rpoS transcription in bacteria.
Gene(s)OrganismEffect on rpoS transcriptionReference
gacA/gacSP. fluorescensPositive effect, fusion studies, direct?Whistler et al. (1998)
rhlI/rhlRP. aeruginosaPositive effect, fusion studies, direct?Latifi et al. (1996)
psrAP. aeruginosaPositive effect, fusion and DNA binding studies, directKojic and Venturi (2001); Kojic et al. (2002)
P. putida  
relAP. aeruginosaOverexpression of relA has positive effect, fusion studies, indirectvan Delden et al. (2001)
gacA/gacSA. vinelandiiPositive effect, Northern studies, direct?Castaneda et al., 2001
barAE. coliPositive effect, fusion and primer extension studies, direct?Mukhopadhyay et al.(2000); Pernestiget al. (2001)
uvrYE. coliNegative effect, microarray and western studies, direct?Oshima et al. (2002)
crpE. coliNegative effect, fusion studies, direct?Lange and Hengge-Aronis (1994)
crrE. coliNegative effect, fusion studies, indirectUeguchi et al. (2001)
relA/spoTE. coliPositive effect, fusion studies, indirectHirsch and Elliott (2002)
ppk/ppxE. coliPositive effect, fusion studies, indirectShiba et al. (1997)

In E. coli, rpoS transcription is controlled by a two-component system and by the cAMP receptor protein (CRP) somehow through the signalling of ppGpp and polyphosphate whereas in Pseudomonas by a two component system, a TetR family regulator and by quorum sensing. In Pseudomonas these regulatory systems are also involved in the expression of genes in response to cell-density highlighting the integration of σs with other global systems governing the colonization and virulence of this species. The picture emerging is therefore that in Pseudomonas transcriptional level of regulation plays a major role because of the different functional role played by σs as well as the involvement of other global regulators. In E. coli, on the other hand, transcriptional control is apparently less important however, further investigations need to more precisely define the mode of action of the regulators thus far implicated and whether other trans-acting factors are involved.

Role of global two-component systems

The GacA/GacS system

The global two-component system GacA/GacS is well conserved in a variety of Gram-negative bacteria and has been implicated in the regulation of secondary metabolism and virulence factors and in quorum sensing (Heeb and Haas, 2001). GacA is the response regulator and GacS is the cognate histidine sensor kinase, though the environmental signal sensed by GacS is unknown. In P. fluorescens Pf-5, GacA/GacS influence rpoS transcription and σs accumulation (Whistler et al., 1998). Transcription, assessed with a chromosomal rpoS-lacZ transcriptional fusion, undergoes a threefold induction upon entry into stationary phase. In gacA and gacS mutants, this transcriptional induction occurs more gradually and to a slightly smaller level. Furthermore, σs accumulated in the mutants to only 20% of the levels detected in the wild type. It is not known at which level GacA/GacS act or whether it acts directly or indirectly. A similar effect of GacA/GacS has been observed in the nitrogen-fixing non-symbiotic soil bacterium Azotobacter vinelandii (Castaneda et al., 2001). In this case, Northern hybridization showed that rpoS mRNA was present in the stationary phase but not during exponential growth, demonstrating that it is under transcriptional growth phase regulation. No rpoS mRNA was detected in stationary phase in a gacA mutant, indicating an important role of GacA in transcriptional induction. Again, no further data are available on this effect and, as in P. fluorescens, it is not known whether this effect is direct. This is important to establish as GacA/GacS is a global regulatory system involved in several regulatory cascades (Heeb and Haas, 2001). In P. aeruginosa, for example, GacA/GacS positively control the production N-butyryl-homoserine lactone (also called C4-HSL), one of the two homoserine lactone quorum sensing autoinducers produced by P. aeruginosa (Reimmann et al., 1997). In P. aeruginosa, the C4-HSL quorum sensing molecule is involved in activation of rpoS transcription and vice versa (see below; Latifi et al., 1996; Whiteley et al., 2000).

The BarA/UvrY system

The BarA sensor kinase was first identified by its ability to suppress an envZ mutation through the control of OmpR in E. coli (Nagasawa et al., 1992). In uropathogenic E. coli it has also been implicated in the induction of a siderophore-mediated iron-acquisition system and as an important regulator in colonization of urinary tract infections upon attachment to eukaryotic host cells (Zhang and Normark, 1996). It has now also been demonstrated to induce rpoS transcription; consequently, rpoS mRNA levels are reduced in a barA mutant, as is the amount of σs protein (Mukhopadhyay et al., 2000). BarA is the homologue of Pseudomonas GacS, which is also involved in rpoS induction (see above). BarA is a member of the subclass of tripartite sensor kinases that catalyse the phosphorylation of their cognate response regulators via an ATP-His-Asp-His-Asp phosphorelay. It was recently shown to be a cognate kinase of UvrY, a response regulator of the FixJ family (Pernestig et al., 2001). Unexpectedly, UvrY does not appear to be involved in rpoS activation (Hengge-Aronis, 2002) and it is therefore probable that BarA activates rpoS through another, yet unidentified, response regulator. However, although a systematic microarray analysis of a uvrY deletion mutant displayed upregulation of the σs regulon through upregulation of σs synthesis it did not detect a major role for BarA in regulating the σs regulon, despite the fact that rpoS-lacZ expression is lower in the barA knock-out mutant (Oshima et al., 2002). Therefore, BarA and UvrY could both be involved in the regulation of σs levels but in different ways: BarA positively regulates rpoS whereas UvrY has a negative effect. It is not known whether they act interdependently to bring about these different controls and, just like the GacA/GacS system, the signals that modulate this two component system have yet to be determined. Interestingly, it has been reported that an autoregulatory loop exists between BarA and UvrY whereby UvrY stimulates barA expression (Suzuki et al., 2002).

It is important to note that the BarA/UvrY system is intimately related to another system called CsrA/CsrB. CsrA (carbon storage regulator) is a 61 amino-acid RNA binding protein and the central component of the system. It is capable of post-transcriptional repression or activation, depending upon the RNA target, of several stationary phase genes. For example, it inhibits glycogem metabolism, biofilm formation and gluconeogenesis whereas it activates glycolysis, acetate metabolism and flagellum biosynthesis in E. coli. The second component of the system is CsrB, a 366-nucleotide untranslated RNA which can bind, and consequently sequester, up to 18 CsrA subunits forming a large ribonucleoprotein complex thereby antagonizing the activity of CsrA. It has been observed that CsrA indirectly stimulates csrB transcription through UvrY indicative of an autoregulatory mechanism which determines the activity of CsrA without changing its level (Suzuki et al., 2002).

Role of quorum sensing

In Gram-negative bacteria, quorum sensing relies on autoinducer molecules [N-acyl homoserine lactones (HSLs)] that accumulate in the medium and allow individual cells to sense a population density. There are two known quorum sensing systems in P. aeruginosa, the Las and the Rhl systems. Each system has a transcriptional activator (LasR and RhlR) and an autoinducer synthase (LasI and RhlI), LasI synthesizes N-3-oxododecanoyl-homoserine lactone (3OC12-HSL) and RhlI, as mentioned above, synthesizes C4-HSL. The activator forms a complex with the cognate autoinducer at high threshold levels to induce transcriptional activation of other genes (Fuqua et al., 2001). The two quorum sensing systems in P. aeruginosa do not act independently, the las system is situated above the rhl system because it regulates rhlR (Latifi et al., 1996; Pesci et al., 1997).

Latifi et al. (1996) have observed that the expression of rpoS (as determined with plasmid rpoS-lacZ fusions) is abolished in las and rhl knock-out mutants. In addition a direct connection between the rhl system and activation of rpoS transcription was demonstrated in E. coli by the fact that an rpoS-lacZ fusion was activated in the presence of RhlR and C4-HSL (Latifi et al., 1996). However, another study observed that expression of a chromosomal transcriptional rpoS-lacZ fusion was similar in the quorum sensing mutants and in the wild type (Whiteley et al., 2000). The same study reports that σs negatively regulates rhlI expression, in accordance with the observation that rpoS mutants overproduce the excreted molecule pyocyanin, whose production is positively regulated by the RhlI/R quorum sensing system. Interestingly, σs is also involved in the regulation of quorum sensing in Ralstonia solanacearum, where it positively regulates the quorum sensing solR/I genes (Flavier et al., 1998). This is the only σs from β-proteobacteria that has been studied. Direct quantification of HSL was not performed and it is not known whether this regulation is direct.

In summary, two studies report a major link between quorum sensing and σs at the transcriptional level in P. aeruginosa, which probably results in co-ordinated regulation of virulence determinants and survival in stationary phase. The contradictory data can probably be explained by the use of different mutants [quorum sensing mutants are highly unstable and can accumulate secondary mutations (Beatson et al., 2002)], and of different experimental set-ups. In addition, it is important to study the role played by the GacA/GacS system in P. aeruginosa and to determine whether, as in P. fluorescens, it affects rpoS transcription. If so, one should determine whether this putative regulatory role is direct or via quorum sensing, as in P. aeruginosa GacA/GacS regulates the rhl system.

A complete and functional HSL-mediated quorum sensing system has not been detected in E. coli K-12. A LuxR family protein designated SdiA is present in E. coli, however, a HSL synthase gene has not been identified (Michael et al., 2001). SdiA is involved in the regulation of cell-division locus ftsQ and a genomic array study indicated its involvement in the regulation of several genes (Wei et al., 2001). It is not known if SdiA can act as an HSL receptor in E. coli but recent experiments provided evidence that SdiA of Salmonella enterica is an HSL receptor, implying a possible role of this protein in recognizing other Gram-negative bacterial species (Michael et al., 2001). The expression of rpoS in E. coli, however, appears not be affected by SdiA (Garcia-Lara et al., 1996).

Direct activation of rpoS transcription by the TetR family regulator PsrA in Pseudomonas

The TetR family regulator PsrA (pseudomonas sigma regulator) was identified in P. putida and P. aeruginosa through its involvement in rpoS transcriptional activation. psrA knock-out mutants in both species displayed a 90% decrease of promoter activity in stationary phase and a 50% decrease of σs levels (Kojic and Venturi, 2001). DNA-binding studies revealed that PsrA binds specifically to the rpoS promoter at a sequence − 35 to − 59 with respect to the + 1 transcription site consistent with a role as an activator of transcription. This sequence is conserved in P. putida and P. aeruginosa and contains the palindromic motif c/gaac N2-4 gttg/c. In addition, PsrA negatively autoregulates its own expression through binding to a similar sequence in its own promoter (Kojic et al., 2002). When produced in E. coli, PsrA activated a Pseudomonas rpoS-lacZ promoter fusion, but the activity was well below that obtained in the original strain (Kojic and Venturi, 2001). This is an indication that PsrA alone is not sufficient to stimulate fully the promoter; PsrA probably regulates other loci, because several putative PsrA DNA-binding sites are present in the P. aeruginosa genome and total protein SDS-PAGE analysis of P. aeruginosa psrA mutants indicates that genes are highly de-repressed compared to the original strain (M. Kojic and V. Venturi, unpubl. data). This implies that PsrA can act both as a repressor and activator probably depending on its position of binding within a promoter.

Involvement of the cAMP receptor protein (CRP) in E. coli

The rpoS promoter of E. coli contains two putative cAMP-CRP binding sites; one just upstream the − 35 σ70 consensus sequence and the other several nucleotides downstream from the + 1 transcription initiation position (Hengge-Aronis, 2002). Transcription of a rpoS-lacZ fusion was reported to increase in exponential phase of growth in cya (encoding adenylate cyclase, which synthesizes cAMP) and crp mutants, suggesting that the cAMP–CRP complex is involved in transcriptional repression rather than its classical role as an activator (Lange and Hengge-Aronis, 1994). The effect observed in the cya mutant could be decreased by the external addition of cAMP. Further experimentation is required to establish whether the CRP binding sites are functional and, if the observed cAMP-CRP effect is direct, whether the location of the putative cAMP-CRP binding sites is consistent with an inhibitory role. Consistent with a role for cAMP-CRP as a negative regulator is the observation that a crr knock out mutant displayed also a de-repressed expression of rpoS and accumulated 20 times more σs in exponential phase (Ueguchi et al., 2001). The Crr protein is a vital component of the phosphophenol pyruvate:carbohydrate phosphotranferase system of glucose uptake and is also a regulator of several cellular functions, including the modulation of cAMP activity. The observed effect of a crr mutation on a rpoS-lacZ transcriptional fusion can be restored to wild type levels in mid-exponential phase by adding cAMP, indicating that Crr acts indirectly, probably through modulating cAMP levels (Ueguchi et al., 2001).

Signal molecules implicated in triggering rpoS transcription initiation

The trans-acting factors involved in σs control most probably respond to an intracellular signal or multiple signals related to the cessation of growth caused by environmental stresses. The mechanim(s) by which bacteria sense stress and starvation is poorly understood since the multiple nature of these signals and the connection to specific stress conditions makes this subject difficult to investigate.

The involvement of the intracellular molecule guanosine 3′,5′-bispyrophosphate (ppGpp) as a positive signal of rpoS expression has been reported in E. coli and Pseudomonas (Gentry et al., 1993; Lange et al., 1995; van Delden et al., 2001; Hirsch and Elliott, 2002). ppGpp, the levels of which affect the stringent response, is synthesized by the ribosome associated RelA protein as a physiological response to nutritional stress. Under starvation conditions, the synthesis of ppGpp is also mediated by SpoT and only spoT relA double genomic knock-out mutants completely lack ppGpp. rpoS-lacZ activity and σs levels are reduced in E. coli spoT/relA double mutants. Promoter deletion studies have attributed this decrease not to lower transcription initiation but to reduced elongation, which could be due to premature transcriptional termination caused by the uncoupling of transcription and translation, thus reducing mRNA levels (Lange et al., 1995). Furthermore, artificial overproduction of ppGpp caused by the increased expression of relA increases σs levels (Gentry et al., 1993). More recent studies revealed that the key role of ppGpp is on basal rpoS expression, in both exponential and stationary phase, though ppGpp has a less important transcription inducing role, delaying full expression by 2–3 h (Hirsch and Elliott, 2002). In E. coli, ppGpp also increases σs levels at the post-transcriptional level as demonstrated by measuring the rate of σs synthesis by pulse labelling. ppGpp appears to induce translation indirectly by regulating non-ribosomal factor necessary for rpoS translation (Brown et al., 2002). Similarly, overexpression of relA in P. aeruginosa slows growth and prematurely activates rpoS transcription during early growth phase (van Delden et al., 2001). The quorum sensing system was also activated prematurely, signifying that the stringent response is also involved in regulating quorum sensing, even in an rpoS negative genetic background.

The ppGpp alarmone has a positive influence on the accumulation of inorganic polyphosphate (Gentry et al., 1993). Inorganic polyphosphate (poly-P) is a linear molecule of hundreds of orthophosphate residues that accumulates in many bacteria under stress conditions and in stationary phase and has an important role in regulatory responses (Kornberg et al., 1999). It is synthesized by polyphosphate kinase, encoded by the ppk gene, which polymerizes the terminal phosphate of ATP into a poly-P chain and is degraded by the exopolyphosphatase (poly[P]ase) encoded by the ppx gene. In E. coli, reducing the amount of poly-P to barely detectable levels by overexpressing a ppx gene results in a significant decrease in rpoS-lacZ transcription. Under these conditions, σs levels fail to increase upon entry into stationary phase (Shiba et al., 1997). Reducing the levels of poly-P has no effect on the concentration of ppGpp (Shiba et al., 1997), whereas poly(P)ase is inhibited by ppGpp. Consequently ppGpp leads to 100 to 1000-fold accumulation of poly-P (Kuroda et al., 1997). The stimulation of rpoS expression by ppGpp could therefore occur via poly-P. More evidence is required in order to establish the molecular mechanisms by which poly-P exerts its positive regulatory role and whether it is direct or via a poly-P binding protein or other factors. The fact that poly-P failed to activate rpoS transcription in vitro with reconstituted RNA polymerase points to an indirect involvement. The role of poly-P as an ‘alarmone’ has recently been addressed in P. aeruginosa, where it was shown to be important as a regulator of quorum sensing and several virulence factors (Rashid et al., 2000). The possible role in regulating rpoS expression was not investigated in Pseudomonas but, in the light of its function in virulence and stationary phase gene expression, it would appear appropriate to do so.

Homoserine lactone (non-acylated) was reported to play an important role in controlling rpoS transcription in E. coli (Huisman and Kolter, 1994). More recently, it appeared that homocysteine thiolactone and not homoserine lactone has a positive effect on σs levels (Goodrich-Blair and Kolter, 2000). Both compounds accumulate in nutritionally starved cells and are intermediates in the methionine biosynthesis pathway; but they are not precursors of the quorum sensing related acylated homoserine lactones. Future work is required to confirm the role of this amino acid biosynthetic pathway and the possible molecular mechanism and link to ppGpp in the control of rpoS/RpoS expression.

Concluding remarks

In recent years, a large number of trans-acting and cis-acting factors have been implicated in post-transcriptional and post-translational regulation of σs levels in E. coli. Studies of these factors have led us to a basic model of this rather complex regulation system. Transcriptional regulation, on the other hand, has been studied sporadically and, consequently, the picture needs to be clarified. Many operon fusion and mRNA studies have been performed in E. coli, where the cAMP/CRP complex and the two component system BarA/UvrY act as transcriptional regulators that ensure higher transcription rates under starving conditions (summarized in Fig. 1). The inhibitory role of the cAMP/CRP complex and the stimulatory role of BarA/UvrY need to be further addressed at the molecular level to complement the genetic studies that have already been performed (see Fig. 1).

Figure 1.

A working model for regulation of rpoS transcription in E. coli. The cAMP–CRP complex could negatively regulate rpoS via binding to two putative CRP boxes. Consistent with the negative role of cAMP–CRP is the requirement for the Crr protein, which stabilizes cAMP. The two component sensor BarA positively regulates rpoS expression. The identity of the two-component regulator that pairs with BarA (A?) to mediate this regulation is unknown. UvrY is specifically phosphorylated by BarA and negatively regulates rpoS. UvrY stimulates barA expression. The ppGpp induces the expression of rpoS, probably via the control of poly-P levels by inhibiting exopolyphosphatase. It remains unclear whether poly-P does so directly or via another molecule(s) or factor(s). See text for details.

The role played by σs in Pseudomonas is similar to, but distinct from, that of σs in E. coli probably a result of the difference in physiology and metabolism. Thus σs-mediated stress resistance is likely also to proceed via different sets of genes. In addition, σs in Pseudomonas is known to be involved in the expression of a subset of extracellular toxins that are also regulated by other global regulators in response to cell-density and other environmental factors (Jorgensen et al., 1999; Suh et al., 1999). These differences might help explain the observed differences in the regulation of rpoS in the two bacteria. Transcriptional regulation appears to play a major role in Pseudomonas but studies at the post-transcriptional and post-translational level have not been performed. The TetR family regulator PsrA has been directly implicated, whereas GacA/GacS and quorum sensing have been shown to influence positive transcriptional regulation of rpoS (summarized in Fig. 2). The regulatory mechanisms through which GacA/GacS influence quorum sensing and σs levels highlight the complexity of interactions involving several global regulators. A regulatory cascade probably exists between GacA/GacS, quorum sensing, PsrA and σs in response to stationary growth phase and these factors need to be organized hierarchically through standardization of experiments and the use of well-characterized mutant strains (see Fig. 2).

Figure 2.

A working model for regulation of rpoS transcription in fluorescent pseudomonads. The TetR family PsrA regulator directly binds and positively regulates rpoS transcription at the onset of stationary phase. The GacA/GacS two component system has a positive effect on rpoS transcription but whether it is directly or through another regulator is unknown. In P. aeruginosa, GacA/GacS positively regulates the RhlR/I quorum sensing system and it remains unclear whether and how the Rhl system regulates rpoS or whether RpoS negatively regulates the rhl system. It is still unknown if and how PsrA fits into this cascade, which already involves several global regulators. ppGpp probably induces the expression of rpoS but it remains unclear whether it does so directly or via another molecule(s) or factor(s). See text for details.

Identifying signalling molecules that are involved in rpoS expression can be particularly difficult, as a mutation that causes a change in the growth behaviour can have an indirect consequence on σs accumulation. In E. coli, the mechanism by which the nutrient stress alarmone ppGpp stimulates rpoS expression, probably via inorganic polyphosphate, is far from clear. The signalling mechanism(s) occurring in post-transcriptional and post-translational levels of σs control is(are) also poorly understood.


I thank RpoS co-researchers C. Aguilar, I. Bertani, G. Degrassi, G. Devescovi, M. Kojic and M. Sevo for their contribution in the results of this review. This work was supported by I.C.G.E.B. (International Centre for Genetic Engineering and Biotechnology, Trieste, Italy).