RpoS and the bacterial general stress response

SUMMARY

The general stress response (GSR) is a widespread strategy developed by bacteria to adapt and respond to their changing environments. The GSR is induced by one or multiple simultaneous stresses, as well as during entry into stationary phase and leads to a global response that protects cells against multiple stresses. The alternative sigma factor RpoS is the central GSR regulator in E. coli and conserved in most γ-proteobacteria. In E. coli, RpoS is induced under conditions of nutrient deprivation and other stresses, primarily via the activation of RpoS translation and inhibition of RpoS proteolysis. This review includes recent advances in our understanding of how stresses lead to RpoS induction and a summary of the recent studies attempting to define RpoS-dependent genes and pathways.

INTRODUCTION

In Escherichia coli, the alternative sigma factor RpoS is the central regulator for a coordinated and extensive response of the cell upon entry into stationary phase, starvation for many nutrients, and for dealing with various other stressful conditions. Cells devoid of RpoS have significant changes in metabolism, are sensitive to oxidative stress, pH extremes, and DNA damage, among other stresses. This multipronged sensitivity of rpoS mutant cells has led to these stress-induced physiological and metabolic changes being defined as a “general stress response (GSR),” distinct from the condition-specific stress responses that cells also show. The RpoS regulon is extensive, with more than 300 promoters under its control. The broad range of inducing signals for RpoS accumulation is paralleled by a complex set of inducing mechanisms. Here, we review what is known about the regulation of RpoS in E. coli and how this reflects its roles. As much as we have learned about the E. coli response, significant gaps still exist in our understanding of this central metabolic switch.

THE GENERAL STRESS RESPONSE IN BACTERIA

The general stress response involves a cross-protection and a preventative response to many stresses (Fig. 1). The way this is regulated varies among species, suggesting independent evolution of unique systems in response to a common physiological requirement. Presumably, the existence of a GSR reflects the likelihood that a given species will encounter multiple challenges simultaneously due to changes in stress and nutrient availability.

Fig 1.

The bacterial general stress response. (A) Multiple stresses feed into synthesis and stabilization of RpoS, the primary regulator of the E. coli GSR, and (B) RpoS regulates multiple stress response pathways. Stress resistance pathways are represented in purple, within shaded blue region, metabolic changes in gold and in blue-green, other physiological changes upon RpoS induction. Induction by any of the stresses in A can lead to many of the outcomes in B.

GSR primary regulators: nature and conservation

Escherichia coli and many γ-proteobacteria utilize the group 2 specialized sigma factor RpoS as the central regulator for the GSR. Group 1 sigma factors, the essential RpoD-like sigma factors, are known as housekeeping regulators; group 2 sigma factors are closely related to the RpoD-like sigma factors but are non-essential (Fig. 2A). While all sigma factors contain domains called σ2 and σ4, responsible for binding to the −10 and −35 regions of promoter sequences, group 1 and group 2 sigma factors possess a σ3 domain linking these two, as well as N-terminal regions, longer for group 1 sigma factors than for group 2 (Fig. 2A). While they differ in their domain composition, RpoS and RpoD have high amino acid similarity, rendering differentiation based on amino acid sequence comparison between RpoS-like and RpoD-like sigma factors difficult [reviewed in reference (1)]. Consistent with these similarities, RpoS-like sigma factors recognize promoters that bear great similarity to the consensus RpoD-dependent promoters (Fig. 2B) (2, 3).

Fig 2.

Sigma factors and RpoS structure/function. (A) Domain organization of different sigma factor groups. Examples of sigma factors belonging to each group are in parenthesis, in black for those found in E. coli and in green examples of GSR sigma factors of Bacillus subtilis (σ^B) and of α-proteobacteria (σ^ECFG). (B) Key features in RpoS-recognized promoters and the RpoS sigma factor. Consensus sequence as defined in Peano et al. (4); see text for additional references (3, 5, 6). Length of spacer, shown in pink, includes extended −10 region.

Group 2 sigma factors are not present in all bacteria and appear to originate from the duplication of an ancestral version of rpoD (7). RpoS, the best studied of these, is present in most γ-, β-, and δ-proteobacteria. It appears to be absent in most other bacterial clades, including the α- and ε-proteobacteria, although group 2 sigma factors, in general, are widespread across bacterial phyla (8). Some γ-proteobacteria were shown to contain multiple group 2 sigma factors in addition to a canonical RpoS homolog. The RpoS-like RpoQ and RpoX in some Vibrio species seem to be involved in biofilm formation and virulence regulation (9, 10). It is not known whether these sigma factors originated from the same rpoD duplication or from an independent event. This suggests that many group 2 sigma factors remain to be discovered.

Even within the β-, γ-, and δ-proteobacteria, not all species contain an rpoS gene, including the β-proteobacteria Neisseria gonorrhoeae and the γ-proteobacteria Acinetobacter baumannii. Even though the GSR is not well documented for all of these species, A. baumannii appears to use other regulators for GSR, such as the BfmRS two-component system (11). While the spirochete Borrelia burgdorferi encodes an alternative sigma factor named RpoS, involved in the induction of genes necessary for its infectious cycle, this protein is thought to have evolved independently from RpoS, originating from another rpoD duplication (8).

Most α-proteobacteria regulate the GSR using a very different, group 4 sigma factor (σ^ECFG in many species) as a central regulator. Group 4 sigma factors are distinct from RpoD and RpoS, contain only the σ2 and σ4 domains, and are frequently associated with envelope stress and other specific stress responses (12, 13). Nonetheless, the genes that are part of the σ^ECFG regulon have significant overlap with genes found in the E. coli RpoS regulon, including genes involved in the oxidative stress and osmolarity response (e.g., dps, catalases, sodA, osmC) (14, 15). Outside of proteobacteria, σ^B, a group 3 sigma factor (containing domains σ2, σ3, and σ4) is involved in the GSR in Firmicutes and most gram-positive bacteria [reviewed in reference (16)].

Despite the lack of a conserved GSR regulator across the bacterial kingdom, the parallel evolution of the use of specialized sigma factors to regulate the GSR would suggest the importance of this type of response for bacterial survival. How the regulation of the sigma factors’ abundance and activity varies between species, and how this differential regulation allows these organisms to uniquely adapt to their own environments and stressors are only beginning to be understood.

Why, then, are alternative sigma factors the overwhelming choice for regulating GSRs? Certainly, they have properties needed to control a large regulon, and just as the vegetative sigma factor activities can be regulated by specific transcriptional activators and repressors, so too can these specialized sigma factors, allowing possible branching within the regulon. Having a sigma factor as the initiating and necessary part of such a global switch in transcription program may provide the most flexibility in combinatorial regulation of the range of downstream outputs.

Sigma factors generally act positively—to turn on genes. Negative regulation, seen in GSR regulons, can be through competition with other sigma factors for core polymerase, or via RpoS-positive regulation of negative regulators, including small regulatory RNAs (sRNAs) that pair with and repress mRNA targets. However, mechanisms for direct inhibition of gene expression by RpoS have been documented. Promoter occlusion, in which RpoS blocks initiation by other sigma factors or RpoS-dependent RNA polymerase (RNAP) pausing at the initiation of transcription, has been described, but their impact remains poorly understood (17–19). The characteristic of competition with the vegetative sigma is likely of primary importance, reflecting the function of the GSR regulator as a switch in lifestyle, as discussed in the next section.

GSR and RpoS as the switch between growth and survival

The earliest studies of strains carrying mutations in rpoS revealed phenotypes reflecting its importance for protecting cells against stress: sensitivity to oxidative stress, temperature and pH extremes, and high osmolarity, among others (20). While these phenotypes and the presence of a GSR in many bacteria attest to its usefulness, it is equally clear that under many conditions, cells are at a disadvantage when expressing the RpoS regulon. This reflects a trade-off between growth and survival that has been termed SPANC (self-preservation and nutritional competence) (21). Much of the disadvantage associated with high RpoS may reflect RpoS competing with RpoD and other sigma factors for core polymerase (22). Cells with high ratios of RpoS to RpoD, while stress resistant, were deficient in growing on multiple carbon sources, while those with reduced or defective RpoS had the opposite phenotypes.

Recent high-throughput fitness studies examining the competitive growth of transposon mutants of E. coli K-12 and other species in different experimental conditions corroborate these observations (23). While insertions inactivating rpoS were shown to be detrimental for growth in the presence of stresses, such as the DNA-damaging agent cisplatin, rpoS mutants were beneficial for growth using various poor carbon and nitrogen sources, such as arginine as the sole nitrogen source or acetate as the sole carbon source.

The negative effects of RpoS are also reflected in extensive studies documenting the frequent appearance of mutant alleles that decrease RpoS levels or activity in laboratory isolates and under long-term growth conditions (24–31). These observations vary depending on the specific growth and storage conditions, ranging from a 3-year-long evolution experiment to the isolation of infectious strains of E. coli from patients (32–34). With improvements in the capacity for sequencing evolved strains, additional understanding of when and how these mutations arise is available. Careful studies provide evidence that strains stored or transported on nutrient-rich agar slants or stabs frequently accumulate rpoS mutations (35, 36). These observations suggest a need to re-evaluate whether E. coli isolated from patients or in natural environments do indeed carry rpoS mutants, as initially suggested, or whether earlier storage conditions led to the mutations. One study compared two sets of human commensal and extraintestinal E. coli strains, one stored in stabs and with high levels of rpoS mutants and a second frozen without long lab manipulation and containing very few rpoS mutations, supporting the idea that rpoS mutants are arising in the lab and may not compete successfully in humans (37).

If cells are at a disadvantage under many growth conditions when RpoS levels are high, it may not be surprising if mutations arise that lower RpoS levels rather than inactivating the protein. In an examination of patient isolates, mutations were found in RpoS-regulated genes, and RpoS levels varied greatly in different isolates, but no mutations in the rpoS gene itself were found, suggesting that rpoS null alleles could lead to decreased virulence and, therefore, were not selected during evolution in the patients (34). Many laboratory stocks and evolved strains of E. coli K-12 carry an insertion sequence (IS1) that inactivates crl, a gene coding for a positive regulator of RpoS activity (38, 39). Another example of the complexity of the role of RpoS is demonstrated in studies by Moreau, evolving E. coli under phosphate starvation conditions (40). While mutations in rpoS as well as in metabolic genes arose quickly and were beneficial for short-term growth under Pi-starvation conditions, they were eventually detrimental for survival. The authors suggest that the RpoS regulon is needed to cope with high acetate levels produced by the evolving population (40).

MULTIPLE REGULATORY INPUTS AFFECT RpoS LEVELS AND ACTIVITY

As outlined above and shown in Fig. 1, the GSR is activated in response to a variety of different stresses or starvation conditions and provides broad cross-resistance to various stressors. The necessity to increase RpoS levels or activity under a variety of conditions implies the need for regulators responsive to each of these conditions. Studies of levels of regulation for RpoS in E. coli and Salmonella have highlighted the particular importance of post-transcriptional regulation. The signals that lead to upregulation are, as one might expect, frequently tied to the stresses that RpoS helps to mitigate. Once the stress or starvation has passed, it is equally important for the cell to restore RpoS levels and activity to the pre-stress state. While the return to a proper RpoS homeostasis upon recovery from stress remains less well studied than the induction of the RpoS response, it is becoming clear that this is likely at least as important as the conditions for inducing the GSR.

Transcriptional regulation of rpoS

In E. coli and many related bacteria, the gene encoding RpoS is found downstream of nlpD, encoding an outer membrane lipoprotein with roles in cell division. The primary promoter for rpoS, rpoSp, is found in the middle of the nlpD coding region, leading to a transcript with an unusually long (567 nt) 5′ untranslated region (UTR) (Fig. 3A) (41). Both the connection to nlpD and the presence of a long 5′ UTR are conserved in bacteria containing rpoS (8). The 5′ UTR plays a critical role in the regulation of rpoS translation, as discussed in the next section. The connection to nlpD has not been explained, although possibly characteristics of the transcription start site and elements of the 5′ UTR have led to the conserved synteny.

Fig 3.

Translational regulation of RpoS by sRNAs. (A) Genome context of rpoS. The major rpoS promoter is located within the upstream nlpD gene, forming a 567-nucleotide 5′ UTR. Another promoter is located upstream of nlpD and drives nlpD transcription but contributes very little to the expression of rpoS. Transcriptional activators and repressors shown are those discussed in the text and noted with an asterisk in Table 1. (B) (i) Translational regulation of rpoS. Closed hairpin structure of the 5′ UTR of rpoS mRNA inhibits translation initiation. This is overcome by base-pairing of the Hfq-dependent sRNAs ArcZ, DsrA, and RprA (in turquoise), with the upper strand of the hairpin, leading to the opening of hairpin and subsequent ribosome binding and activation of translation. Negative regulation of translation by the sRNAs OxyS through Hfq titration, and by CyaR by direct base-pairing to rpoS mRNA are shown. Other sRNAs implicated in rpoS regulation are discussed briefly in the text but not shown. (ii) Inducing signals leading to sRNA transcription. See text for references and further details.

Studies on transcriptional regulation of rpoS, as summarized in databases such as Ecocyc (42), identify multiple positive and negative regulators of the rpoS promoter. However, depending on how the regulation was measured, some of these may reflect changes in mRNA stability or readthrough, rather than transcriptional initiation. Transcriptional reporter fusions or measurements of rpoS mRNA levels will be perturbed by effects on RNA stability. Transcriptional termination, which has been shown to occur in the long 5′ UTR of rpoS, will also affect evaluation of initiation if measured with assays that measure transcription beyond the termination region (43). Classic experiments by Lange and Hengge-Aronis (44) using parallel transcriptional and translational reporter fusions demonstrated transcriptional control negatively correlated with growth rate, with differences in complex and minimal medium, but also showed major changes in translation and protein stability dependent on growth stage and conditions. It is also becoming increasingly clear that redundant levels of regulation exist for rpoS, further complicating attempts to define how a given regulator functions.

We currently have an incomplete state of understanding of the transcriptional regulation of rpoS. Table 1 lists all regulators of the rpoS promoter listed in Ecocyc (42) as of August 2023, summarizing our evaluation of their likely importance. Specific regulators with the most information available (highlighted with an asterisk) are also discussed in the text.

TABLE 1.

Regulators of rpoS transcription, as listed in Ecocyc (42)^a

Transcriptional regulator	Repressor or activator	Experimental approaches	Summary and comments	References
ArcA*	Repressor	Transcriptional fusion; EMSA and DNAase protection.	ArcA also regulates translation, via negative regulation of the sRNA ArcZ, a positive regulator of RpoS translation.	(45, 46)
Crp/cAMP*	Repressor/activator	Transcriptional and translational fusions; RpoS Westerns; effects of mutations in Crp sites in promoter.	Conflicting results in the literature, but mutations in Crp-binding sites suggest direct positive regulation.	(47)
MqsA*	Repressor	In vitro binding; transcriptional fusions.	Strength of repression is disputed in literature, but repressor activity is agreed upon.	(48, 49) (50)
Fur	Repressor	Undefined rpoS-lacZ fusion; mRNA levels.	Low rpoS mRNA levels in fur mutants could be indirect.	(51)
GadX	Activator	mRNA analysis of selected genes (microarray) upon GadX overexpression.	No evidence of direct regulation.	(52)
TorR	Activator	RpoS protein and rpoS mRNA levels; EMSA; site mutation; transcriptional fusion.	Deletion of torR leads to modest reductions in RpoS, particularly at pH2.0.	(53)
ppGpp	Activator	RpoS levels under stress and in strains lacking ppGpp.	Observed dependence of RpoS levels on ppGpp now known to be due to anti-adaptor upregulation, stabilizing RpoS and sRNA upregulation, increasing translation.	(54)

^a The strongest evidence is associated with those marked with asterisks, discussed in more detail in the text.

ArcA negative regulation

Using transcriptional fusions for rpoS that include the long 5′ UTR, Mika and Hengge (45) found higher expression in an arcA mutant and a corresponding increase in RpoS protein levels. In vitro, the ArcA response regulator protein bound and protected a region of the primary rpoS promoter, but the importance of the protected site for in vivo regulation was not explored.

Mandin and Gottesman identified ArcZ as an sRNA activator of RpoS translation (see section on Translational Regulation of RpoS) (46). ArcZ is negatively regulated by ArcA and, less dramatically, by its cognate sensor kinase ArcB, possibly explaining some of the effects of ArcA and ArcB on RpoS levels seen by Mika and Hengge. Expression of an RpoS-LacZ translational fusion, under the control of the arabinose-inducible pBAD promoter, was increased during early exponential growth in an arcA mutant, but that increase was fully abolished by an arcZ mutation (46). These results suggest that, while ArcA may bind and negatively regulate at the rpoS promoter, its effect on RpoS levels is at least in part due to its regulation of ArcZ.

Catabolite regulator protein and cAMP regulation of the rpoS promoter

Multiple groups have reported that cAMP and CRP regulate rpoS, and two sites for binding of CRP have been identified in the rpoS primary promoter. One of these is upstream of the core promoter elements, and one is within the initial portion of the 5′ UTR (47, 55). While some groups reported that CRP and cAMP were activating of rpoSp, others observed repression (56, 57). For example, in a recent report (55), deletion of crp significantly increased expression of an rpoS promoter reporter in minimal glycerol medium, suggesting negative regulation. However, the extent of the region in the reporter was not defined, and whether regulation was direct or indirect was not addressed. The likely answer is that both positive and negative regulation occur, depending on growth conditions, complicated by cAMP and CRP also acting at other levels (for instance, regulating mRNA stability and translation via the CyaR sRNA, see below). A recent report makes a strong case for direct positive regulation. Guo et al. (47) examined both a promoter fusion and RpoS protein levels during growth in rich medium. Early in exponential phase, RpoS and promoter fusion levels were increased in cya or Δcrp mutants, while at later time points when RpoS was detectable in WT cells, both protein and reporter activities were lower in cya and crp mutants. Mutating either of the putative cAMP-CRP sites in the fusion decreased expression of RpoS, and mutating both gave the lowest expression. This strongly suggests that there is direct positive regulation of rpoS transcription initiation by cAMP-CRP that depends on both the upstream and downstream sites. While it would be useful to test these mutants under other growth conditions, these results would suggest that negative regulation of rpoS may be indirect.

MqsA as a possible repressor of rpoS

Another reported negative regulator of rpoS is MqsA, the antitoxin and transcriptional regulator for the MqsR toxin (48, 50). Wood and coworkers found in vitro binding by MqsA to a site in the rpoS promoter and showed that mutating the site eliminated the ability of overproduced MqsA to repress an rpoS transcriptional reporter. Fraikin et al., in experiments using less dramatic overexpression of MqsA, did not see repression of an rpoS promoter reporter (49). Still to be determined is to what extent and under what physiologically relevant conditions MqsA regulates rpoS transcription.

GadX upregulation of rpoS and a potential positive feedback loop

GadX, a regulator of acid stress response genes, is listed in Ecocyc as an activator of rpoS, based on global RNA measurements when GadX was overexpressed (52). How direct or indirect this regulation is remains unexplored. Given that the gadX gene is itself positively regulated by RpoS (as well as a myriad of other proteins), this has the potential to provide a positive feedback loop for RpoS under the appropriate (GadX active) conditions. Both RpoS and GadX have important roles in the acid stress response, discussed below.

In summary, it seems likely that ArcA, Crp/cAMP, and possibly MqsA can directly interact with and regulate the major rpoS promoter within nlpD, but it is less clear when these transcriptional regulators are most important. In addition, ArcA and Crp/cAMP are known to regulate post-transcriptional regulators of rpoS, complicating a simple interpretation of the published studies. The discussion above focuses on transcription initiation. Evidence for regulation of termination within the rpoS 5′ UTR is discussed below, in the context of the activity of the small regulatory RNAs.

Translational regulation of RpoS

sRNAs as activators of RpoS translation

As discussed above, a long 5′ UTR precedes the ATG start codon for rpoS. We now know that secondary structure within this UTR interferes with rpoS translation, sequestering the ribosome-binding site in a base-paired structure. At least three small regulatory RNAs, with the help of the RNA chaperone Hfq, bind to and change the structure of the 5′ UTR, freeing the ribosome-binding site and allowing efficient translation (Fig. 3B).

The roles of the mRNA secondary structure, Hfq, and the sRNAs were defined in a series of studies in the mid- to late 1990s. Malcolm Winkler and collaborators first characterized mutations in the hfq gene in E. coli (58), finding that loss of hfq led to sensitivity to UV light, changes in cell shape, and osmosensitivity. This was followed by studies in both E. coli (59) and Salmonella (60) independently showing that a mutation in hfq blocked translation of rpoS. Subsequent genetic experiments by the Elliott lab identified cis-acting mutations in the rpoS leader that suppressed the effect of an hfq mutant, demonstrating the existence of an inhibitory hairpin (61, 62).

Separately, our lab had identified a small non-coding RNA, named DsrA, that positively regulated RpoS translation, particularly at low temperatures (63). Because only a few small RNAs had been previously identified, and their activity was only beginning to be studied, it was initially unclear how DsrA worked. However, the results of the Elliott lab allowed us to propose direct pairing of DsrA to the upper strand of the inhibitory hairpin, eliminating the hairpin structure and exposing the RBS; this model was confirmed using appropriate point mutations to disrupt base-pairing and compensatory mutations to restore base-pairing of DsrA to the rpoS 5′ UTR (64). Lease and co-workers suggested a similar model, as well as DsrA pairing with another target, hns (65). Subsequent work by our lab and others have further supported this model both in vivo and in vitro (66–68) and found other sRNAs, discussed below, that similarly positively regulate rpoS translation.

Our current understanding of the regulation of synthesis and degradation of the sRNAs that regulate RpoS and to what extent their induction can explain aspects of the ability of RpoS to respond to different stresses is reviewed here. Because these sRNAs not only pair with the rpoS 5′ UTR but also with other mRNAs, these sRNAs may help in “specializing” the GSR to specific stresses. Recent reviews of Hfq-dependent sRNAs and how they act can be found at references (69, 70).

DsrA

As indicated above, DsrA was the first sRNA described to activate RpoS translation. Given our finding that DsrA is particularly important at low temperatures (63), it was not surprising to find that the promoter of dsrA is activated at low temperature (71, 72), although the mechanism of this activation has not been determined. Recent work has also demonstrated that dsrA is positively regulated by ppGpp, an alarmone made at elevated levels in response to amino acid limitation and other stresses (73). It had previously been observed that ppGpp positively regulates RpoS production via increased transcription of the iraP anti-adaptor (see below) (73, 74), and this activation of DsrA synthesis—and thus RpoS translation—provides an additional explanation for the role of ppGpp in RpoS accumulation (54). It is less clear whether levels of ppGpp that affect iraP and dsrA transcription are the same. Also, not well defined are the range of DsrA targets, or whether that helps us understand why low temperature or ppGpp would lead to DsrA regulation. DsrA negatively regulates hns, as mentioned above, which encodes an abundant nucleoid-associated protein that acts as a transcriptional silencer for multiple RpoD-dependent promoters [see, for instance dps, (75)].

RprA

The second RpoS regulatory sRNA to be found was RprA, identified in a screen of multicopy plasmids containing fragments of the E. coli chromosome for the ability to turn on an RpoS-LacZ translational reporter (76). Like DsrA, RprA also pairs with the upper strand of the inhibitory hairpin in the rpoS 5′ UTR. rprA transcription is positively controlled by RcsB and the Rcs phosphorelay that regulates RcsB phosphorylation (77, 78). A variety of cell surface perturbations, including antimicrobial peptides and inhibitors of cell division, activate the Rcs phosphorelay and thus RprA. Other outputs from this cascade include synthesis of capsular polysaccharide. Thus, RprA links RpoS to a large regulon, generally associated with cell surface stress that is also likely important during biofilm formation.

ArcZ

The third characterized sRNA activator of RpoS is ArcZ. It was identified from a screen of a plasmid library expressing each of the then-known sRNAs (46). ArcZ is negatively regulated by the ArcB/ArcA two-component system. ArcB phosphorylates ArcA, generally under anaerobic conditions, and ArcZ is then repressed by ArcA-P. This contributes to RpoS levels, particularly in the initial recovery from the stationary phase (46). ArcZ regulates many other targets in addition to rpoS, making identification of physiological connections of those targets to RpoS a challenge. Given ArcA’s repression of rpoS transcription, the ArcAB two-component system appears to combine transcriptional and translational repression of RpoS, and the contribution of each has not been fully dissected. ArcZ negatively regulates motility, as does RcsB, suggesting that the cell decreases motility under at least two different conditions that also activate RpoS.

Are there more positive regulating sRNAs for rpoS translation? At least in stationary phase, deletions of arcZ and dsrA are sufficient to reduce the translation of rpoS almost as much as a deletion of the gene for the RNA chaperone Hfq (46). This suggests that any other positive regulators (including RprA) are not well expressed under the unstressed growth conditions tested in most published studies. However, as more stress conditions are examined, it is certainly feasible that more sRNA activators of rpoS will be found.

Negative sRNA regulators of RpoS translation

Several other sRNAs have been implicated in regulating rpoS translation (Fig. 3B). OxyS, an sRNA positively regulated by OxyR, was identified as a negative regulator of RpoS (79). Because there is no evidence of direct base-pairing of OxyS to the rpoS leader and Hfq overexpression partially overcomes the OxyS effect, it is likely that OxyS acts by titrating Hfq from the positively acting sRNAs (80). The CyaR sRNA, also identified as a negative regulator of RpoS, was subsequently demonstrated to base-pair with the rpoS leader, leading to degradation of the rpoS mRNA (81). Interestingly, ArcZ can pair with and negatively regulate CyaR (81); thus, when ArcZ is abundant, it may both positively regulate rpoS translation by base-pairing with the rpoS 5′ UTR and overcome the negative regulation by pairing with CyaR. As mentioned above, cAMP and CRP were found to negatively regulate rpoS transcriptional reporters in some experiments. It seems possible that this negative regulation is in part due to CyaR (82), which is positively regulated by cAMP and CRP, interacting with the rpoS 5′ UTR.

Since the screening done by us in 2010, many more Hfq-binding sRNAs have been identified that were not in the initial screening library (70). In particular, through approaches in which sRNAs and their targets are ligated together on Hfq and the resulting chimeras sequenced (RIL-Seq, GRIL-Seq), new potential sRNA regulators of rpoS have been identified (83, 84). In a recent comparative study by the Lory lab, sRNAs interacting with rpoS mRNA were identified in E. coli, Vibrio, and Pseudomonas. In E. coli, in addition to ArcZ and DsrA, MgrR (expressed at low magnesium ion concentration), AspX [encoded at the 3′ end of the aspA mRNA; a novel sRNA also found to be a decoy for RyhB (85)], GadY, and an antisense RNA to the ybiE gene (now named ZbiJ) were identified as sRNAs that target rpoS. When overexpressed, ZbiJ, MgrR, and AspX all negatively regulated an rpoS reporter fusion, with ZbiJ, the strongest regulator, predicted to pair near the RBS of rpoS mRNA (83, 86). A better understanding of when these novel sRNAs are expressed during growth or stress will provide further understanding of the physiological roles of these RNAs, both in their regulation of rpoS and other potential targets.

The study by the Lory lab provides the opportunity to compare the well-studied sRNA regulation of RpoS in E. coli to the situation in Vibrio cholerae and Pseudomonas aeruginosa. A long 5′ UTR is found for rpoS in each of these species, but the RNAs found to interact with the rpoS mRNA were different. Both negative and positive regulators of the Vibrio reporter fusion were found; for Pseudomonas, the observed regulation was negative (83). Minimally, these results suggest that it is likely that sRNA regulation of RpoS extends well beyond E. coli and its close relatives but that the sRNAs used, when they are expressed and how they affect RpoS, are likely to differ with different species.

Other regulation within the 5’ UTR

While the evidence is strong that pairing with the sRNA opens up the secondary structure within the 5′ UTR, increasing translation and modestly stabilizing the mRNA (66), it has also been found that sRNA pairing can interfere with Rho-dependent termination within the 5′ UTR, positively regulating RpoS via transcriptional elongation (43). Rho requires a block of single-stranded RNA to bind and move along to a termination site; the sRNAs, by pairing to the 5′ UTR, may block that movement. Whether negatively regulating sRNAs such as CyaR may also act positively by blocking termination has not been evaluated.

RNase III, which cleaves double-stranded RNAs, may well be capable of cleaving the 5′ UTR hairpin in its closed (double-stranded) state and/or the double-stranded RNA formed by annealing of an sRNA to the upper strand of the hairpin. Consistent with this, our lab found a negative effect of RNase III on both mRNA and protein levels for RpoS (66). To what extent this is a regulatory effect or simply helps to keep RpoS levels low in the absence of stress is not known.

Given the length of the 5′ UTR for rpoS, it seems quite possible that other RNA binding or cleaving proteins may also interact with the leader. One such example is CspC/E, two of the nine small cold-shock family proteins. Overexpression of CspC or CspE stabilizes rpoS mRNA. Studies by the lab of Eliora Ron (87, 88) showed binding of CspC to the rpoS 5′ UTR, thereby stabilizing rpoS mRNA (89). As with some of the other regulators identified for rpoS, it is less clear if CspC/E are in fact regulatory (with their effects changing with conditions) or if they are necessary in the context of keeping the long 5′ UTR of rpoS protected sufficiently for regulation to occur.

Translational regulation: beyond sRNAs and the 5’ UTR

The factors described above act on the 5′ UTR, affecting mRNA folding and therefore ribosome accessibility, mRNA stability, and transcription termination. However, rpoS translation is also affected by factors independent of the 5′ UTR, some, not yet well understood, affecting translation initiation and others affecting translation within the RpoS ORF.

Initiation of RpoS translation increases dramatically in stationary phase (90) and in mutants of aceE, encoding a subunit of pyruvate dehydrogenase (91). In both cases, Hfq and the 5′ UTR are not required. What are the critical characteristics of this translation initiation region, and what signals or regulators affect translation initiation, remain to be determined. rpoS translation is decreased in ssrA mutants, encoding tmRNA (92), and in mutants of tRNA modification genes miaA and trmL (93, 94). The tRNA modification dependence can be bypassed by changing the leucine codons within the rpoS portion of an RpoS-LacZ translational fusion, suggesting that loss of modification has a direct effect on the translation of the rpoS ORF. Most recently, we found that in the absence of polyamines, or when cells carry mutations in rpsL that lead to hyperproofreading, RpoS levels decreased due to co-translational degradation of RpoS (95). Given that these regulatory effects are not seen for rpoD translation, these observations suggest that codon usage within rpoS has evolved to allow communication between the translation apparatus and RpoS levels in a way that would benefit from additional investigation.

Regulation of RpoS protein stability

Proteolytic control plays multiple roles for the regulation of RpoS in E. coli. First, degradation is important in keeping RpoS levels low during exponential growth; any RpoS synthesized, despite the transcriptional and translational levels of control outlined above, is rapidly degraded by the ClpXP protease (Fig. 4C). This degradation requires the atypical response regulator RssB, which binds to RpoS and escorts it to the ClpXP protease before being recycled to degrade more RpoS (96). This utilization of a recycled regulatory protein to facilitate degradation allows a minimal amount of RssB to facilitate the degradation of large quantities of RpoS.

Fig 4.

Regulation of RpoS proteolysis by RssB, ClpXP, and anti-adaptors. (A) Genome context for rssB, encoding an adaptor for RpoS degradation. One rssB promoter is located upstream of rssA; the other is downstream of rssA, immediately upstream of rssB. Both are controlled by RpoS and RpoD, with RpoD providing the basal level expression of RssB. (B) Activators, repressors, and inducing signals for the anti-adaptors IraM, IraP, and IraD are shown. See text for references. (C) Degradation of RpoS mediated by RssB and ClpXP and inhibition of RpoS degradation by the core RNAP and the anti-adaptors. Each of these anti-adaptors has a different interaction with RssB (not shown here), with only IraP preferring the phosphorylated form of RssB (as shown); binding of RssB to RpoS is also favored by phosphorylation.

Due to the rapid degradation of RpoS by the RssB-ClpXP system, any process that blocks or slows the degradation of RpoS provides an efficient “switch” that can be flipped, allowing the cell to rapidly accumulate RpoS and thus initiate the GSR. In E. coli, this inhibition of degradation is primarily accomplished by “Inhibitor of RssB Activity” or Ira anti-adaptor proteins. The Ira proteins are synthesized under specific stress conditions, bind to RssB, and prevent it from binding to RpoS and escorting it to ClpXP. Each of the three characterized Ira proteins of E. coli K12 is unrelated, suggesting independent evolution of multiple pathways to allow the rapid accumulation of RpoS and activation of the GSR. In addition to the anti-adaptors, RpoS can be stabilized by binding to core polymerase (96). Thus, factors that favor RpoS interaction with core will contribute to RpoS stabilization.

Finally, and perhaps most importantly, rapid RpoS degradation allows the bacterium an effective exit from the GSR when the response is no longer required. As noted above, RpoS and the GSR, while helpful under some growth/stress conditions, are disadvantageous under others. While new synthesis may cease, removal of the accumulated RpoS requires degradation. Embedded in the RpoS regulon are negative feedback loops ensuring that this acts effectively (97, 98).

Components for regulated degradation of RpoS in E. coli

RssB

RssB, also called SprE (stationary phase regulator) in E. coli or MviA (mouse virulence protein A) in Salmonella, is a 334-amino acid protein, with an N-terminal domain characteristic of the large response regulator family. It is encoded in the E. coli genome as the second gene in a two-gene operon (Fig. 4A). The function of the upstream rssA gene remains unknown. rssB is transcribed at a low level by RpoD, but transcription increases in an RpoS-dependent manner, providing a negative feedback loop likely important for restoring homeostasis in cells (97, 99, 100). RssB’s role in promoting RpoS degradation was first described in 1996; it was later shown to bind RpoS and deliver it to the ClpXP protease for degradation (101, 102).

RssB has proven challenging to resolve structurally, leaving many molecular details of how it functions still to be determined. The N-terminal REC (receiver) domain has the characteristics expected of this large family, including a conserved aspartate capable of being phosphorylated (103). Available structures suggest the C-terminal domain (C) of RssB resembles PP2C-type phosphatases, though it lacks many of the canonical residues required for phosphatase activity (104, 105). Unlike many response regulators, there is no evidence that E. coli RssB acts as a dimer or that a monomer-to-dimer transition is critical for its function (105). Also not well defined is how the two domains interact with RpoS or with ClpXP. There is little doubt that the flexible linker between these domains is critical for proper function, with the E. coli linker apparently very different from that of an RssB homolog from Pseudomonas aeruginosa (PDB 3EQ2).

Role of RssB phosphorylation

The conserved aspartate residue (D58 in E. coli RssB) in the N-terminal REC domain of RssB can be phosphorylated, leading to the expectation that regulation of RpoS degradation might depend on phosphorylation. In vitro studies suggested that the D58 residue is phosphorylatable using acetyl phosphate (AcP), as is true for many response regulators, and that this phosphorylation enhanced RpoS binding and functionality as an adaptor for RpoS degradation (96, 105, 106). However, unlike many response regulators, the rssB gene is not linked to a histidine kinase, and in vivo sources of phosphorylation are not clear. Mika and Hengge found a role for the sensor histidine kinase ArcB in phosphorylation of RssB in vivo and in vitro (45), but the conditions when this may be physiologically significant remain to be demonstrated.

AcP was reported by Bouche et al. to be a source of phosphorylation in vivo based on slower RpoS turnover in a strain unable to synthesize AcP due to deletion of ackA and pta (107). Under the growth conditions used by Bouche et al., AcP is synthesized by Pta and degraded by AckA. Our lab revisited the role of AcP in RpoS accumulation, comparing a mutation in ackA (higher levels of AcP) to a mutation in pta (low levels of AcP) to a deletion of both, and found that all three strains had increased levels of a degradable RpoS fusion protein (108). This suggested that disruption of metabolism, rather than absence of AcP, explains the Bouche et al. observations.

Most significantly, studies in the Silhavy lab clearly revealed that the phosphorylation of D58 was not necessary for the regulation of RpoS stability in response to GSR-inducing stresses. Mutating D58 to a phosphonegative residue, alanine, did not affect the ability of the cell to stabilize RpoS in response to carbon starvation or phosphate depletion or resume RpoS degradation for recovery when carbon or phosphate was restored (109). Thus, mechanisms other than phosphorylation must operate to regulate RpoS degradation. The discovery of the Ira anti-adaptors (Fig. 4C), discussed further below, provides at least part of the answer. Phosphorylation, which clearly happens in vivo, contributes to the basal level activity of RssB and may be important under specific, thus far untested conditions or possibly for other functions of RssB (see below) (109). It is also possible that in vivo phosphorylation, potentially from multiple sources, might be the default state for RssB and that it is dephosphorylation by unidentified phosphatases that will prove important as a regulatory signal under yet-to-be-defined conditions.

RpoS K173 is recognized by RssB

RpoS and RpoD share significant sequence and structural similarity, and recognize overlapping promoter motifs (Fig. 2B) (5). However, while RpoS is unstable, RpoD is not. The recognition sequence for RssB to bind RpoS for delivery to ClpXP centers around residue K173 of RpoS. A series of reporters carrying regions of RpoS fused to LacZ identified sigma region 2.4 as necessary for turnover of the fusion protein (110). K173, within region 2.4, is conserved in RpoS proteins, while glutamic acid is found at that position in RpoD. A K173E substitution completely blocks the ability of RssB to bind to RpoS and thus facilitate its degradation (110). While K173 defines one element of the interaction between RssB and RpoS, a full structure of the complex remains to be determined, as does a molecular understanding of how RssB presents RpoS to the next essential component of the degradation pathway, the ClpXP protease, without itself being degraded.

ClpXP

The ClpXP protease, a 1:1 complex of the ClpX ATPase hexamer and the ClpP multimeric proteolytic subunits, is one of five ATP-dependent proteases in E. coli. It is required for the degradation of RpoS, and mutations in other proteases have no effect on RpoS stability (111). The specificity of degradation is likely explained by the required role of RssB in delivering RpoS to ClpXP. Reconstruction of the degradation pathway with purified components demonstrated that RssB recognized RpoS in the absence of ClpXP and that this interaction was necessary for RpoS delivery to ClpXP (96). Both domains of RssB are necessary for interaction with RpoS (112). Because RpoS does not bind to ClpX in the absence of RssB, the interaction with RssB must remodel or reorient RpoS to enable recognition by ClpX (96). Consistent with this, the N-terminus of RpoS is required for ClpX recognition and unfolding to feed the polypeptide into the ClpP proteolytic chamber (113). RssB binds to the N-terminal zinc-binding domain (ZBD) of ClpX, and the ZBD is required for RpoS degradation (114). Further understanding of this process awaits, minimally, a structure of the entire degradation complex.

The observation that RssB interacts with RpoS in the absence of ClpX suggests the possibility of RssB acting as an anti-sigma factor, binding to and interfering with RpoS activity. This has been directly demonstrated in strains deleted for the ClpXP protease. In such mutants, where RpoS degradation cannot take place, high levels of RssB lead to a decrease in RpoS activity (115, 116). While this does not reflect its natural biological role in E. coli, it may give an indication as to an ancestral role of RssB as an anti-sigma, which later evolved the function of an adaptor for degradation.

Ira anti-adaptors: inhibitors of RssB activity stabilize RpoS

As summarized above, in vitro experiments demonstrate that RssB, ClpXP, and a source of ATP are necessary and sufficient to degrade RpoS. It is also clear that, under a wide variety of stress conditions, degradation of RpoS ceases, leading to its rapid accumulation. However, given that RssB derivatives that cannot be phosphorylated can still efficiently degrade RpoS in vivo and that stress conditions lead to stabilization, some other levels of regulation of RpoS degradation must exist (109). This regulation, it turns out, is performed by the Inhibitor of RssB Activity, or Ira, anti-adaptors, at least in response to many stresses.

The Ira anti-adaptors were first identified in genomic screens for the stabilization of RpoS-LacZ translational fusion proteins (117, 118). These anti-adaptors, IraM, IraD, and IraP, are each made and active under different conditions and are named for the condition under which they were first studied (Fig. 4B). IraM promotes RpoS stabilization in response to magnesium starvation and low pH, IraD in response to DNA damage, and IraP in response to phosphate starvation (117–120). Later studies have suggested other conditions under which they may also be expressed (see below). Despite unique mechanisms of induction and lack of sequence or structure similarity, each anti-adaptor binds to RssB and inhibits its ability to bind RpoS, thus inhibiting overall degradation.

IraM

IraM, one of the major Ira proteins found in E. coli, plays a principal role in inhibiting the degradation of RpoS under magnesium starvation conditions. Its transcription is dependent on the PhoQ/PhoP two-component system and was first characterized as leading to RpoS stabilization under magnesium starvation, when PhoP is active (117). As discussed below, acid stress also leads to PhoP activation and thus IraM-dependent RpoS stabilization. The iraM gene sits at the edge of the e14 prophage and shows characteristics of recent horizontal transfer (e.g., low GC content) and, as for other horizontally transferred genes, is repressed by H-NS (121). There are homologs of iraM in some species of Enterobacterales, named rssC and iraL, with different transcriptional regulation and thus stabilizing RpoS under different conditions. IraL is produced and stabilizes RpoS during exponential phase in uropathogenic E. coli and Shigella, aiding in the adaptation of these pathogens to survival in their hosts (122). Induction conditions for RssC, found in Salmonella, have yet to be described. IraM is included in a list of potential antimicrobial drug targets based on its presence in pathogenic Enterobacterales but not in mammals (123).

IraD

IraD was identified in a screen for genes able to stabilize an RpoS-Lac fusion when overexpressed. The previously uncharacterized gene (formerly yjiD) was named IraD for DNA damage, reflecting work showing it was induced upon hydrogen peroxide and contributed to cell survival after treatment with hydrogen peroxide or long-term treatment with AZT; resistance was also RpoS dependent (117, 119). iraD is induced in response to DNA damage via a mechanism independent of the canonical SOS response, recently shown to be regulated by DnaA (120). Independently of the SOS response, the promoters of iraD were also shown to be induced by ppGpp on the entry into the stationary phase, although ppGpp may act indirectly, via the SspA stringent response protein (120, 124) (Fig. 4B).

In addition to transcriptional regulation, IraD translation is also negatively regulated by the RNA-binding protein CsrA. CsrA negatively controls iraD translation by blocking ribosome entry for an upstream leader peptide, indirectly regulating iraD translation (119). Because CsrA is usually associated with promoting rapid growth and inhibiting stationary phase metabolism, CsrA’s function here may be to contribute to keeping RpoS activity low when not needed and/or promoting recovery of low RpoS levels after stress conditions have passed (125).

The structure of IraD in complex with RssB demonstrates binding along the length of RssB (104). IraD bridges both domains of RssB, apparently locking RssB’s domains in an inactive conformation and preventing RpoS recognition (104, 105). The non-phosphorylated form of RssB favors interaction with IraD (106), and in the crystal structure, the site of RssB phosphorylation was mutant. A recent study based on interactions of the N-terminal domain of RssB with IraD or RpoS concluded that IraD outcompetes RpoS for binding to RssB along the same interface, leading to the stabilization of RpoS (126), but is not consistent with studies of the full-length RssB.

IraP

IraP, the first identified anti-adaptor protein, was found through a genomic library screen looking for inhibitors of RssB-dependent degradation of RpoS and was shown to inhibit RpoS degradation during phosphate starvation. The purified protein acts by binding to RssB and preventing it from acting as an adaptor (118). iraP transcription is induced during phosphate starvation. While a two-component system activated in response to phosphate starvation (PhoR/PhoB) exists, it is not involved in iraP expression. Rather, iraP transcription is dependent on SpoT and ppGpp; the iraP promoter is directly positively regulated by ppGpp (73, 74) (Fig. 4B). iraP transcription has also been reported to be activated by the biofilm transcriptional activator CsgD directly or indirectly, resulting in RpoS stabilization, possibly providing coordination between biofilm formation and the rest of the RpoS response (127, 128). In Salmonella, while the ppGpp-dependent promoter of iraP is conserved, a second promoter, activated by PhoP, is found (129). Thus, in Salmonella, where PhoQ/PhoP is an essential virulence regulator, IraP has been co-opted into the PhoP regulon. This reinforces the conclusion noted for iraM above and likely to be found in future studies—bacteria with different lifestyles have evolved variations in regulation of the anti-adaptors, allowing them to induce the RpoS response as needed.

The studies examining regulation of RpoS degradation in E. coli and Salmonella have provided an understanding of how RpoS proteins can be stably maintained at a low level and quickly rise in response to stress. However, RssB and anti-adaptors are not found exclusively in E. coli and Salmonella, and the characterization of these proteins and their homologs in other species has helped inform our understanding in the E. coli model system in turn.

RssB and proteolytic regulation outside of E. coli

Evolutionary analyses have found rssB homologs only in γ-proteobacteria, although functional studies have been primarily in Enterobacterales. The lack of clear ira gene homologs outside of Enterobacterales suggests that if there are anti-adaptors, they are distinct enough in sequence not to have been detected or that other regulatory mechanisms may have evolved for RssB and/or RpoS in these organisms.

Enterobacterales

In part due to its critical role in maintaining widespread transcriptional activity within E. coli, RssB and its regulatory role has been characterized in several species of Enterobacterales, where homology of both RssB and the RpoS regulon is best conserved. Closely related to E. coli, the Salmonella RssB homolog (MviA) has been particularly well characterized. MviA appears to perform a similar function to RssB in its mediation of RpoS turnover, with its deletion resulting in stabilization and thus overproduction of RpoS resulting in general stress tolerance (130). Deletion or downregulation of RssB led to hypervirulence due to overproduction of RpoS. The importance of proper regulation of RpoS is seen not only in mammalian pathogens such as Salmonella and E. coli but also in plant pathogens. Mutations in the Pectobacterium carotovorum RssB homolog, called ExpM, have reduced virulence in tobacco (131). The RpoS regulon has been linked to pathogenicity and virulence gene induction (see below), and thus the RssBs of these pathogenic species serve as a potential avenue for perturbing disease development in a variety of bacterial pathogens.

Vibrionales

The Reidl group has recently demonstrated that the RssB homolog in Vibrio cholerae O1 also functions as an adaptor for the degradation of RpoS via the ClpXP protease (132, 133). Synthesis of this RssB homolog is fully dependent upon RpoS, consistent with a role for RssB as a feedback regulator, leading to RpoS degradation after the stress has passed and RpoS is no longer needed. While RssB is stable in E. coli, in V. cholerae RssB was shown to be itself degraded dependent on ClpP after RpoS levels return to normal. Like E. coli RssB, there is a decrease—though not complete abrogation—of RssB activity when the phosphorylation site is mutated. This parallel system provides an interesting variation on E. coli, suggesting that different species of γ-proteobacteria have adapted to differentially regulate RpoS proteolysis.

Other γ-proteobacteria

Less has been reported for RssB and proteolytic control of RpoS protein outside of Vibrionales and Enterobacterales. While a resolved structure exists of a full-length RssB protein from Pseudomonas aeruginosa (PDB 3EQ2), this structure differs from more recent structures of E. coli RssB (104, 105), in that the Pseudomonas protein has an extended and potentially inflexible linker and was isolated as a dimer, not seen for E. coli RssB. In addition, the P. aeruginosa protein, unlike E. coli, carries the sequences characteristic of a fully functional PP2C phosphatase domain. A recent study reported that the P. aeruginosa RssB and the closely related RssB of Azotobacter vinelandii promote the degradation of RpoS in vivo. In these experiments, RpoS recognition and degradation required an additional STAS domain-containing protein, encoded downstream of rssB, named rssC (134). Such STAS proteins are frequently found to act as anti-anti-sigma factors (e.g., GigB in Acinetobacter baumannii and RsbV in Bacillus subtilis), promoting sigma factor activity, contrary to what was found in this study (135–138). Future work will allow a better understanding of how these more distant RssB homologs function and how their activity is regulated.

Other roles for RssB

One question which remains open is whether RssB has roles beyond the regulation of RpoS. For instance, might it also be an adaptor for delivery of other proteins to ClpXP? Thus far, the phenotypes associated with loss of RssB can, for the most part, be attributed to high levels of RpoS and can be reversed by deletion of rpoS. One reported RssB activity, independent of RpoS, was in modulating the level of polyadenylation of specific mRNAs (139, 140), but the molecular basis of this activity remains unknown.

Regulation of RpoS activity

The sections above describe the complex regulation of RpoS levels, keeping it low in the absence of stress and allowing rapid increases upon entrance into the GSR. RpoS protein is also regulated at the level of activity, primarily by mechanisms that affect its ability to compete with the vegetative RpoD sigma factor for core RNA polymerase.

Both the core RNAP and the housekeeping sigma factor RpoD are constitutively produced and stable during growth. Each alternative sigma factor, however, is produced and active under specific conditions and must compete with the highly abundant RpoD for RNAP binding (Fig. 5A). The RNAP core enzyme concentration is limited in the cell, while the concentration of RpoD is at least threefold that of RNAP. Additionally, RpoD has a strong affinity for the core, and RpoS has the lowest affinity constant of all sigma factors for the core (KD of 0.25 vs 4.26 nM) (141–144). The competition between the sigma factors for binding to the limiting amount of free core RNAP is a key part of the regulation of sigma factor activation. A variety of strategies are used by E. coli to promote the access of RpoS to the core when cells need it. Some of these are specific regulators of RpoS activity, but global strategies to inhibit RpoD binding to the core polymerase are beneficial for the binding of all alternative sigma factors to the RNAP core enzyme, including RpoS.

Fig 5.

Regulators of RpoS activity and roles of ppGpp for RpoS levels. (A) Regulators of RpoS activity. Direct or indirect activators of RpoS appear in turquoise and repressors in red. Crl promotes RpoS binding to the core RNAP, shown in brown, while FliZ can associate with some RpoS-dependent promoters, inhibiting their recognition by RpoS. Rsd, DksA, ppGpp, and 6S RNA inhibit RpoD activity through different mechanisms highlighted in the text, indirectly favoring RpoS. (B) Positive effects of ppGpp on RpoS levels. DksA and ppGpp induce the expression of four positive regulators of RpoS, including the anti-adaptors IraD and IraP, the sRNA DsrA, and the RNA chaperone Hfq. Increased levels of anti-adaptors protect RpoS from degradation (Fig. 4); DsrA together with its chaperone Hfq promotes the translation of rpoS mRNA (Fig. 3). RssB phosphorylation favors RpoS binding; it aids but is not required for IraP activity.

Direct regulation of RpoS activity by Crl

The Crl protein was first identified as a central biofilm regulator (crl stands for its role in the synthesis of biofilm-specific curli appendages) and was later shown to be an activator of RpoS-dependent transcription. Crl functions as a facilitator and stabilizer of the RNAP and RpoS association to form holoenzyme and increases the concentration of the RNAP-RpoS pool in vivo. Crl accumulates during exponential phase and continues to be synthesized until entry into stationary phase, suggesting the importance of Crl when RpoS levels are limited. Initially, the Crl regulon was defined as comprising a portion of the RpoS regulon, although it can likely facilitate expression of all RpoS-dependent promoters under appropriate conditions (145, 146). Additionally, crl expression is stimulated by indole produced by TnaA under alkaline conditions, while under nitrogen starvation conditions, crl expression is negatively regulated, reducing RpoS activity (17, 145).

Structural and biochemical studies have demonstrated that Crl binds to σ2 of RpoS as well as the RNAP subunit β′ and show that Crl stabilizes the holoenzyme by helping tether RpoS to core RNAP (147, 148). While RpoS and RpoD are close in terms of sequence and promoter recognition characteristics, RpoS lacks the long non-conserved region (NCR) between domains σ1.2 and σ2.1 (Fig. 2A). Crl is unable to bind to RpoD due to a clash with this NCR, while RpoS contains the residue R82, which is critical for Crl binding (Fig. 2B). Studies of RpoS from P. aeruginosa, which lacks a Crl homolog, showed that RpoS-R82 is not conserved in this species and, therefore, cannot bind to E. coli Crl, but mutating the residue 82 to arginine in RpoS from P. aeruginosa enables Crl binding (149). In addition, the so-called DPE motif within the σ2 domain of RpoS was shown to be required for Crl binding (147) (Fig. 2B). Because core binding protects RpoS from degradation, Crl also helps to stabilize RpoS; this further increases the expression of RpoS-dependent promoters, including genes that contribute to shutting off the RpoS response, such as rssB (97).

There is growing evidence that Crl not only stabilizes the interaction of RNAP core with RpoS but also acts to facilitate promoter binding and open complex formation during the initiation step of RpoS-dependent transcription. Crl was shown to increase the in vitro transcription of an RpoS-dependent promoter by a pre-assembled RNAP-RpoS complex, suggesting that Crl may stabilize the σ2 of RpoS, necessary for promoter melting (147, 148, 150, 151).

Crl is restricted mainly to γ-proteobacteria but is less widespread than RpoS. Among γ-proteobacteria, Crl is restricted mostly to Enterobacterales, Aeromonadales, and Vibrionales. It cannot be excluded that species containing RpoS but lacking Crl use different strategies to favor RpoS-RNAP binding.

Inhibition of RpoD

In addition to factors like Crl that directly favor RpoS activity/association with core, other components of the RpoS system interfere with RpoD, thus indirectly promoting the ability of RpoS to act (Fig. 5A). For each regulator, we would like to understand when and how it participates in the activity of the RpoS regulon, and how that activity is reversed upon exit from the GSR. Some but not all of this is known for the factors discussed here.

6S RNA

6S RNA is a 180-nucleotide ncRNA synthesized in response to slow growth and stationary phase [reviewed in references (152–154)]. The double-stranded structure of the 6S RNA mimics an open DNA promoter region, interacting with σ4.2 of RpoD when bound to RNAP and preventing its interaction with the −35 box at promoters (155). This freezes the RpoD-core polymerase complex. When nutrient levels increase upon exit from the stationary phase, RpoD can transcribe a short RNA from the 6S RNA, releasing RpoD (156–159). The gene coding for 6S RNA, ssrS, is under both RpoD and RpoS control. Deleting ssrS leads to a decrease in RpoS-dependent gene transcription and an increase in the expression of RpoD-dependent genes. Studies have shown that about half of E. coli genes change expression in the absence of ssrS (155).

The anti-RpoD factor Rsd and its antagonist, HPr

RpoD is subjected to direct inhibition in E. coli and other closely related bacteria by the anti-sigma factor Rsd (also called AlgQ in P. aeruginosa, where it is part of the alginate operon). Rsd is widespread in γ-proteobacteria, although it is poorly studied outside of E. coli and P. aeruginosa (160, 161). A structure of the complex of the E. coli RpoD σ4 and Rsd, coupled with biochemical characterization, has shown that Rsd binds to σ2-σ4 of RpoD, thus preventing its binding to both the core RNAP and the target promoters (160, 162). rsd transcription is under both RpoD and RpoS control, leading to its accumulation during stationary phase, promoting the binding of sigma factors other than RpoD to the core RNAP (163, 164). The overproduction of Rsd causes an increase in the expression of the genes under RpoS control, while its deletion leads to the increase in the activity of RpoD promoters during stationary phase (143).

It was shown recently that Rsd also participates in the regulation of the stringent response in E. coli by binding to SpoT and stimulating its ppGpp hydrolase activity. The absence of Rsd was shown to lead to an increase in the stability of ppGpp during carbon downshift (165). Given ppGpp’s effects on RpoS production and stability discussed here, this secondary effect of Rsd may act to prevent RpoS oversaturation in the cell, maximizing the efficacy of the RpoS already present without necessitating unnecessary RpoS production.

In E. coli, the activities of Rsd are antagonized by the histidine-containing phosphocarrier protein HPr, which is a component of the sugar phosphotransferase system (PTS). The phosphorylated state of HPr influences its binding to Rsd, as seen in partner-switching systems regulating numerous sigma factors (136). Cells grown in the presence of glucose as a carbon source result in unphosphorylated HPr, leading to a strong inhibition of Rsd by HPr and thus favoring RpoD activity (Fig. 5A); in the presence of non-PTS sugars such as glycerol, HPr is phosphorylated, inactive for Rsd binding and inhibition (165–167), favoring alternative sigma factors such as RpoS. The protein HPr, therefore, connects carbon metabolism to the regulation of the activity of sigma factors, including RpoS, and highlights the presence of multifunctional proteins involved in the coordination of metabolism and stress responses.

ppGpp and the GSR

The small molecule ppGpp is a major effector of the stringent response, which allows the rewiring of gene expression of cells in transition from rapid growth to starvation conditions. First, ppGpp, together with its partner protein DksA, directly interacts with RNAP and inhibits the transcription of rRNAs and tRNAs (which constitute the major transcriptional activity in growing cells), releasing the RNAP-RpoD holoenzyme and increasing the availability of RNAP core for binding to alternative sigma factors. Second, ppGpp can directly bind to and enhance the transcription of certain promoters, including the iraP promoter (discussed above). In total, ppGpp impacts the expression of about 700 genes in E. coli (168).

RpoS is not made in strains lacking both ppGpp synthetases, RelA and SpoT, and ppGpp plays pleiotropic roles for RpoS production (Fig. 5B) (54). First, rpoS mRNA is more abundant with higher ppGpp concentration, meaning ppGpp likely increases either rpoS transcription or mRNA stability. Second, DksA and ppGpp have been shown to promote hfq transcription, which is in turn necessary for RpoS translation in combination with sRNAs (73). The transcription of three genes that directly promote higher RpoS levels is positively regulated by ppGpp. The promoter for the sRNA DsrA is stimulated directly by DksA and ppGpp (73). The promoters of two anti-adaptors, iraD and iraP, are triggered by DksA and ppGpp under somewhat different stress conditions (73, 74, 124). While the ppGpp-dependent induction of iraD happens during stationary phase, ppGpp-dependent induction of iraP is triggered during phosphate starvation.

These direct effects of ppGpp on RpoS production and stability are additive with the indirect effects on RNAP availability from ppGpp transcriptional effects, which, when combined, create an intricate and very complex pathway leading to the activation of RpoS in response to stress and starvation.

Direct inhibition of RpoS activity by FliZ

The protein FliZ, which is coded with flagella synthesis genes, was shown to specifically inhibit RpoS promoter activity, by directly binding to RpoS-dependent promoters and thus preventing RpoS binding (Fig. 5A) (169). FliZ is a component of the flagella synthesis pathway and, by modulating RpoS activity, may promote the activity of the flagella synthesis sigma factor RpoF (also called σ²⁸). It is unknown whether FliZ contacts specific RpoS promoters or has a global effect on the RpoS regulon. Similarly, its role and importance in E. coli physiology remain unclear (169).

Additional factors favoring RpoS activity

The global transcriptional pattern is also impacted by the accessibility of promoters by the RNAP holoenzyme. DNA harbors different global or local topologies across the genome, varying with growth phase and conditions, especially during stationary phase where the nucleoid is relatively decondensed (170). Multiple DNA-binding proteins involved in the regulation of nucleoid structure (e.g., Fis, IHF, H-NS, HU, or Dps) are induced at different growth phases and have a strong positive impact on DNA relaxation, freeing access to promoters for binding and, therefore, overall gene transcription (144, 171, 172). Additionally, potassium glutamate, which accumulates during osmotic stress, appears to play a role in the rewiring of global transcription, as salts appear to negatively impact the transcription of the strong ribosomal promoters (173). At the same time, it may activate RpoS-dependent transcription. Potassium glutamate was also shown to shift the RNA polymerase-RpoS complex from a poised to an active state for transcription, thus inducing RpoS-dependent gene expression, particularly of osmotically induced genes such as osmY (174–176).

Heterogeneity in RpoS activity

Growing evidence shows that RpoS activity is heterogeneous at the single-cell level within an E. coli population. Patange et al. showed that the noisy gene expression of RpoS-dependent genes (bolA in this study) is necessary for cell survival after stress (177). A large variation in the activity of the bolA reporter was observed during exponential growth, in which RpoS is expressed, produced, and activated at low levels. The variation observed was abrogated in a strain deleted of rpoS and was not observed on RpoD-regulated promoters, suggesting a specific noisy expression of RpoS-dependent genes at limiting RpoS levels. Using time-lapse microscopy, this study found that the fraction of cells with high RpoS activity survived subsequent oxidative stress better compared to cells with low RpoS activity. The regulators of this noisy RpoS activity have not been defined.

FROM STRESSES TO RESPONSES: RpoS REGULATORY INPUTS AND OUTPUTS

The GSR has the property of being induced in response to multiple stresses through different modes of RpoS activation (Fig. 6). In addition, the GSR has a downstream readout that is significantly broader than what is needed for response to the initial stress. Nonetheless, as with other stress responses, one would minimally expect that the genes of the regulon would provide protection or repair mechanisms for the inducing stresses. Here, we provide an overview of known members of the E. coli RpoS regulon, their likely roles with respect to known stress conditions, and discuss the connections between inducing signals and regulon output for a selected set of stresses.

Fig 6.

Input signals at different levels of RpoS regulation. High cell density or starvation for some nutrients leads to an increase in ppGpp levels, affecting multiple levels of RpoS control (Fig. 5B), by promoting transcription and translation and/or inhibiting degradation. High pH also impacts multiple levels of RpoS control. Upstream regulators for transcription of the sRNAs DsrA, ArcZ, and RprA sense multiple signals leading to RpoS translation. Anti-adaptor induction upon various stresses leads to RpoS protein stabilization. This figure combines the detailed analysis shown in parts of Fig. 3 to 5.

Defining RpoS-dependent promoters

rpoS originated from the duplication of rpoD, and RpoS has co-evolved together with its recognized promoter sequences. As a result of this shared origin, RpoS and RpoD have overlapping recognition elements in promoter sequences. However, a consensus RpoS recognition sequence has been identified, including defining a few features of the promoter that would favor RpoS (summarized in Fig. 2B) (2–6, 178). The strongest specificity in RpoS-dependent promoters is the C nucleotide at position −13, which appears to be mostly conserved in RpoS-dependent promoters. C−13 interacts specifically with the positively charged K173 residue of RpoS σ3 (179). Nonetheless, many promoters that are transcribed by RpoS, and thus are defined as part of the RpoS regulon, are also read by RpoD, likely under somewhat different conditions.

The pattern of induction of RpoS-dependent genes can differ significantly, with some induced at low levels of RpoS and others only when RpoS is higher, while the timing of induction of RpoS-regulated genes varies with the nature of the stressor (180, 181). Additionally, as with RpoD-dependent promoters, RpoS-dependent promoters can also require additional transcriptional activators or be subject to transcriptional repressors. Hengge and coworkers used microarrays to compare the sets of RpoS-dependent genes in three distinct conditions, stationary phase, osmotic shock, and acid stress. Out of the 481 genes upregulated by RpoS in at least one condition, 140 were found in all three conditions, suggesting that while RpoS induces a core set of genes, additional stress-specific regulators are recruited under specific conditions (182). These regulators may promote the reinforcement of the transcription of RpoS-dependent genes needed in a specific condition, such as the acid stress response. For instance, the hdeAB operon is induced by RpoS in an “insensitive” manner, meaning only when RpoS levels are high (180). However, its transcription is also controlled by the acid resistance regulators GadX, GadW, and GadE and by the regulators RcsB and AppY, further activating hdeAB transcription during acid stress but not under other stresses which may induce RpoS. Peano et al. found that not all promoters bound by RpoS show transcription from the gene, further suggesting that additional negative or positive regulators may be acting at RpoS-dependent promoters (4).

Analysis of RpoS regulon data in E. coli

In E. coli, multiple studies have been done to define the RpoS regulon, using techniques ranging from microarrays, RNA-seq combined with ChIP-seq, or in vitro transcription assays. RpoS targets found in these studies vary greatly depending on the method used, but RpoS has been found to affect the transcription of up to ~1,000 genes (~ 23% of all genes) in E. coli.

Various issues render it challenging to precisely define the RpoS regulon. RpoS-recognized promoters are difficult to detect bioinformatically, as they closely resemble RpoD-dependent promoters, and RpoS readily recognizes non-optimal sequences (Fig. 2B). Much of the work on defining the regulon has been done using RNA levels and/or reporter expression, comparing wild-type cells to a deletion of rpoS and thus including both direct and indirect effects. In vitro and ChIP-seq approaches should define direct interactions of RpoS but may not address whether a bound promoter is actually transcribed by RpoS. Consistent with the idea that there is not a simple on/off switch for RpoS-dependent promoters, many studies suggest that the degree of dependence on RpoS changes with growth phase and levels of RpoS. Wong et al. measured the expression of different genes at varying RpoS concentrations, defining three classes of response to RpoS by genes of its regulon; Dong and co-workers found genes that are RpoS induced in exponential phase but not in stationary phase (180, 183, 184). Finally, within a given species such as Escherichia coli, pathogenic strains, prophages, and plasmids introduce further variability, including, in numerous cases, targets for RpoS.

We compiled results from the many global studies to define RpoS regulon members in E. coli K-12 (Tables S1 and S2). Table S1 (“Complete regulon”) lists all genes identified as within the RpoS regulon based on any of six different global analyses (see “References summary” tab for further information). This includes all genes positively regulated by RpoS and identified in various global and associated follow-up experiments, annotated to show in which study they were found and the nature of the evidence (4, 180, 182, 183, 185–187). An estimate of the strength of evidence of RpoS direct control was developed by summing the number of studies in which a gene was found to be RpoS dependent (Table S2; excel document, “Ordered RpoS regulon,” grouped either by total count or by likely functions).

If a direct RpoS target is itself a regulator, turning on its synthesis will lead to indirect downstream effects of RpoS. If such regulators are subject to other environmental inputs, this provides the possibility of downstream branching of the RpoS pathway, allowing specialization within the global stress response. A subset of these is discussed below, tracing the inducing signals related to this function, where known, as well as the RpoS-dependent genes and their roles. The full supplemental tables provide evidence of the many roles of the RpoS regulon still to be explored. Not discussed here, in part because they were generally not annotated in the global approaches used to assemble our list of RpoS regulon genes, are the RpoS-regulated small RNAs. Among these are SdsR (188) and ArrS (189). Genes repressed by these sRNAs will be indirectly negatively regulated by RpoS, at the level of mRNA stability and/or translation. sRNA-activated genes would appear as positively regulated by RpoS, possibly without evidence of direct RpoS interaction with the promoters, although it is certainly possible for RpoS to act at multiple levels on a given transcript (directly, promoting transcription, and indirectly, via action of one of these sRNAs), as it appears to do for regulation of cfa (190).

Defining roles for RpoS-dependent genes

Another approach to define key components of the RpoS regulon is to identify genes necessary for resistance to various stresses associated with the RpoS response, though this presents its own difficulties. If there is redundancy in a resistance pathway, mutating a single gene may not uncover sensitivity. In other cases, a complex resistance pathway may show sensitivity when one of a collection of genes is deleted but will not necessarily reveal all members of that resistance cascade. One of the clearest instances of defining what is sufficient for RpoS-dependent resistance to high osmolarity resulted from an evolution experiment showing that a mutation leading to the overexpression of the otsAB operon bypassed the RpoS dependence for adaptation to high salt (191). In this instance, two linked genes were sufficient to bypass the need for RpoS. If multiple genes downstream of RpoS are needed for stress resistance, it may be difficult to identify them by genetic approaches. For instance, Chen and Goulian found that while RpoS was important for the dehydration tolerance in E. coli, no effectors promoting the tolerance were identified, suggesting multiple targets may be important (192). Another study has defined the RpoS-dependent effectors that are necessary for SDS resistance during carbon starvation as the membrane integrated proteins SanA, DacA, and YhdP (193). However, these genes were not identified in the analyzed studies of the RpoS regulon, highlighting the difficulty of specifically and completely identifying RpoS-regulated genes. As a further complication, cells frequently have more than one pathway for responding to a stress. The response to oxidative stress, discussed below, provides a clear example of this redundancy. Presumably, RpoS-dependent genes will be most important under slow growth/starvation conditions, while the specific stress responses may operate most efficiently when a specific stress is encountered during exponential growth. Our current understanding of the role of RpoS in a few of these pathways is outlined here. Some of these have long been recognized as part of the RpoS GSR; others are suggested by the appearance of multiple RpoS-dependent genes within a metabolic pathway. Additionally, according to the RpoS regulon analysis presented here, roughly 28% (83 out of 299 putative targets) of E. coli putative and/or confirmed RpoS-activated genes appear to be uncharacterized with no predicted function.

RpoS and the oxidative stress response

Before rpoS was recognized to encode a sigma factor, multiple groups studying different processes identified mutations (and named the mutated gene), all of which turned out to be allelic with rpoS. These names thus reflect some of the major phenotypes associated with the loss of RpoS. One such name, katF, resulted from the observation that mutations in this locus led to loss of one of two E. coli catalases, hydroperoxidase II (HPII). HPI, encoded by katG, is induced by hydrogen peroxide, dependent upon OxyR (194). HPII was not induced by hydrogen peroxide but instead by entry into stationary phase, and transcription of katE, encoding HPII, was found to be abolished in the absence of katF. We now know that this RpoS-dependent regulation of catalase is one of the clear hallmarks of the RpoS regulon.

Reactive oxygen species (ROS) are formed through aerobic carbon metabolism in exponentially growing cells as well as in cells starved for nutrients such as phosphate or nitrogen [reviewed in references (195–197)]. However, a failure to keep ROS levels in the cell low creates a highly toxic intracellular environment. Additionally, ROS can arise from environmental stresses or host-secreted toxins and can accentuate the effects of antibiotics like aminoglycosides. The three major ROS forms are superoxides, hydrogen peroxide, and hydroxyl radicals. Superoxides (${\mathrm{O}}_2^{-}$) are acted upon by superoxide dismutases to create hydrogen peroxide (H₂O₂) and oxygen (O₂); the catalases can then transmute hydrogen peroxide into water and oxygen. Directly or indirectly, superoxide and hydrogen peroxide can lead to the promotion of the Fenton reaction, which uses Fe (II) to produce hydroxyl radicals, leading to cell death. To study the enzymes that cells use to combat sensitivity to ROS in the lab, H₂O₂ and superoxide-generating molecules (e.g., paraquat) are frequently added to growing cells to provide a source of ROS. However, given the creation of ROS during respiration and carbon metabolism, cells must perpetually get rid of ROS. Thus, it is not always clear whether endogenous or exogenous sources (or both) drive the need for the multiple layers of ROS defense in bacteria like E. coli.

Bacteria undergoing oxidative stress induce several enzymes to inactivate superoxide and H₂O₂. Some of these are fully RpoS dependent; others are expressed independently of RpoS but under the regulation of specific regulators of RpoD such as SoxRS and OxyR, and in some cases, both RpoD-dependent and RpoS-dependent promoters are found for the same genes [reviewed in references (198, 199)]. Three superoxide dismutases exist in E. coli to convert superoxide into hydrogen peroxide. SodC is the sole periplasmic superoxide dismutase and is dependent upon RpoS for expression. SodC would deal with the ${\mathrm{O}}_2^{-}$ made in the periplasm, most likely by dihydromenaquinone under stationary phase conditions in which glucose is present, as for instance during Pi-starvation (200). Of the two cytoplasmic superoxide dismutases, SodB, an Fe-S protein, is RpoS independent, but SodA, a Mn-dependent enzyme, is positively regulated by SoxRS and RpoS. E. coli contains two catalases, KatG (OxyR dependent) and KatE (RpoS dependent) that convert high levels of H₂O₂ into O₂ and H₂O (197). In addition, the AhpCF peroxidase converts low levels of H₂O₂ into H₂O, consuming NADH, in response to OxyR activation (197). The gene encoding ahpF has also been identified in some studies as part of the RpoS regulon (Tables S1 and S2). Thus, just among the superoxide dismutases and catalases, there are redundant resistance pathways subject to separate, RpoS-dependent and RpoS-independent regulation.

In addition to enzymes that detoxify ${\mathrm{O}}_2^{-}$ and H₂O₂, other aspects of the cell’s metabolism can be regulated to reduce the creation of ROS molecules. The Fenton reaction depends on free iron, and thus keeping iron levels low will reduce this reaction. The primary defense against the Fenton reaction is Dps, abundant in stationary phase dependent upon RpoS. Dps stores iron, protecting DNA from hydroxyl radicals (197). Other mechanisms also contribute by reducing iron in the cell. For instance, the manganese and iron importer MntH is positively regulated by OxyR but was also found to be activated by RpoS, independently of OxyR (201). MntH was shown to play a role in diminishing oxidative stress by importing manganese, which can be used instead of iron to remetalate inactivated iron proteins (202).

A major endogenous source of superoxide and hydrogen peroxide appears to be the accidental auto-oxidation of flavoproteins (196). Some RpoS-dependent functions may participate in protection from this process. RpoS-regulated wrbA encodes an NAD(P)H:quinone oxidoreductase that likely could play a role in reducing quinones, helping to block interaction of the semiquinone with oxygen and production of superoxide (203).

Limiting oxidation within the cell is crucial for cell survival. The use of thiol redox pathways promotes the maintenance of a reduced state, contributing to preventing the formation of ROS. osmC, encoding an osmotically inducible peroxidase, is transcribed by RpoS during stationary phase. OsmC metabolizes organic hydroperoxides via oxidation of cysteine residues within the protein (204, 205). The antioxidant glutathione (GSH, Glu-Cys-Gly) efficiently reduces ROS species primarily by reducing oxidized cysteine residue to CysSH + GSSG and is also produced through RpoS-regulated enzymes. These enzymes include the glutamate-cysteine ligase GshA (involved in glutathione biosynthesis via glutamate utilization) and the putative glutathione S-transferase YncG, whose expression increases in response to hydrogen peroxide stress.

Nitric oxide is another toxic oxidative metabolite produced by the host immune system to neutralize pathogens. The products of its oxidation can then react with DNA, thiols, and iron-containing enzymes. RpoS activates the nitric oxide dioxygenase Hmp, which protects the cytochromes bo and bd under aerobic conditions by eliminating nitric oxide compounds (206, 207).

RpoS and acid resistance pathways

Acid stress is encountered by bacteria under a variety of circumstances. Oxidative or other stresses and the corresponding response can include metabolic changes that lower intracellular pH. Fermentation leads to the acidification of the cytoplasm during prolonged phosphate starvation, eventually triggering cell death (208). Similarly, high osmotic shock also leads to a decrease in intracellular pH; the response to this shock, rapid K⁺ uptake coupled to H⁺ export, leads to the alkalinization of the cytoplasm. As a result, other acids accumulate to restore a proper intracellular pH. In addition, the environment can also impose acid stress. Extreme low pH is found in some areas of the host (e.g., the stomach) and is a challenge for neutralophilic bacteria such as E. coli (209). Therefore, bacteria not only need to adapt to the stress but also need to rewire their metabolic pathways to adapt to the toxicity generated in response to stress. The RpoS regulon is intimately connected to this rewiring.

Acid-specific RpoS-dependent response

E. coli possesses multiple pathways to re-equilibrate intracellular pH [reviewed in references (210, 211)], each under somewhat different and very complex regulation. This reflects the different conditions under which the various pathways are needed. For instance, extremely low pH requires a different response than moderately low pH. These regulatory pathways are also affected by the availability of substrates needed for various proton export pathways. As with oxidative stress, there are both RpoS-independent and RpoS-dependent pathways, but under many growth conditions, rpoS mutants are highly sensitive to low pH (212, 213). In addition to the dedicated pathways for counteracting low pH, changes to cell metabolism can also play an important role in avoiding intracellular generation of low pH. A full discussion of these interwoven pathways is beyond the scope of this review (211).

Induction of RpoS at low pH via IraM stabilization

Levels of RpoS increase when cells are exposed to even moderately low pH, consistent with a necessary role for RpoS in the response to low pH. Much of this induction can be attributed to stabilization of RpoS via the anti-adaptor IraM.

iraM is positively regulated by the PhoQ/PhoP two-component system (117) (Fig. 4B). The response regulator PhoP is phosphorylated and thus active when Mg²⁺ levels are low in the cell, in response to acid stress or in response to antimicrobials such as polymyxin [reviewed in reference (214)]. While the regulation of PhoP can be complex, when PhoP is active, IraM levels rise, and RpoS is stabilized (117). For instance, acid stress leads to activation of the EvgS/A two-component system, which in turn activates PhoP, resulting in the upregulation of iraM and increased levels of RpoS. Deletion of MgrB, an inhibitor of PhoP activity, also leads to increased RpoS levels in the cell, primarily (if not fully) dependent upon PhoP (215). Cells carrying a constitutively active EvgS or lacking MgrB are resistant to low pH, and RpoS and IraM are needed for this acid resistance (AR) (216). Interestingly, in Salmonella, where the PhoQ/PhoP system plays a central role in virulence, iraP rather than iraM is transcribed from a PhoP-activated promoter, also leading to RpoS stabilization when PhoP is active (129). Thus, in both Salmonella and E. coli, acid resistance in stationary phase is associated with the activation of PhoP under acid stress, leading, among other PhoP-regulated effects, to RpoS stabilization via an anti-adaptor.

The acid resistance pathways

Acid response involves multiple strategies, including the passive action of molecules that can buffer the cytoplasm and active responses that correct the pH homeostasis, generally termed the AR pathways. These pathways are activated in different acidic conditions as well as during stationary phase. They also allow a reversal in membrane potential (increase of positive charges within the cell), equilibrated with excess proton export. At least five acid resistance pathways are known in E. coli, each using different molecules. Most commonly, this involves amino acids coupled with the appropriate antiporter and activated depending on the severity of the acid stress. The AR2 pathway is RpoS dependent and provides protection for highly acidic conditions (defined as pH 2–3) in the presence of glutamate (Fig. 7). In this pathway, glutamate decarboxylases GadA and GadB use up intracellular protons to convert glutamate to GABA. The antiporter GadC exports GABA from the cell, importing glutamate and glutamine to be used as a substrate for continuing cycles of this reaction (Fig. 7) (217). The transcriptional activator GadE is required for expression of the AR2 pathway, and the complexity of regulation in this system is reflected in the multiple transcription factors implicated in regulating its expression (see, for instance, Ecocyc entry). A role for the AR2 system in responding to metabolic changes, rather than low external pH, is suggested by the observation that a gadABC triple mutant had decreased viability when grown under anaerobiosis during phosphate starvation (208).

Fig 7.

RpoS-dependent acid stress response. RpoS induces the expression of genes involved in glutamate decarboxylation (AR2 pathway), chaperones, acid and proton export, and in membrane structure. Shown are RpoS-dependent proteins with known roles and involvement in acid resistance. Regulators and other proteins whose roles are unclear are not mentioned here.

Strategies to increase the glutamate pool during acid stress include the conversion of glutamine to glutamate. Glutamine is more available in the human digestive tract than glutamate, and because it is a major nitrogen source for digestive bacteria, its transport and utilization are necessary for bacterial survival in the gut. The RpoS-dependent and acidic condition-responsive glutaminase GlsA converts glutamine into glutamate and ammonia and thus acts as part of AR2 when glutamine is available, but glutamate is not (Fig. 7) (209, 217, 218). AR1 can also be considered part of the RpoS regulon since it uses the RpoS-dependent AR2 enzymes and the intracellular pool of glutamate (219).

In E. coli, gadE and most of the AR2 components, with the exception of gadB and gadC, are encoded in a so-called “acid fitness island,” and increases in RpoS lead to increased expression of most of these genes. However, additional environmental inputs, including low pH, are likely needed as well, and different genes within this cluster show different degrees of dependence on RpoS. This island includes a gene coding for the starvation-induced lipoprotein Slp, which has been shown to promote resistance to low pH through an unknown mechanism. slp forms an operon with the DNA-binding regulator dctR, whose targets are unknown, although the double mutant Δslp ΔdctR leads to a decrease in cell viability compared to the WT after exposure to highly acid conditions (220). The island also contains genes coding for the acid-responsive periplasmic chaperones HdeA and HdeB, as well as a membrane protein of unknown function HdeD, which is known to be involved in acid stress resistance. mdtE and mdtF, coding for the multi-drug TolC-dependent efflux pump MdtEF, are involved in the efflux of bile acids under extreme acid conditions (Fig. 7) (221). The DNA-binding regulator AppY was shown to regulate the expression of both the acid fitness island and other genes involved in acid resistance (222). In addition to its role as a regulator, the overproduction of AppY leads to the stabilization of RpoS (117), suggesting another direct connection between acid and mechanisms to increase RpoS availability.

Other RpoS-dependent acid response strategies

Mild acid stress responses include the induction of the electron transport chain, upregulating cytochromes and dehydrogenases that expel protons (Fig. 7). This strategy is sufficient to re-establish pH homeostasis under mildly acidic conditions. While most of these enzymes are RpoD induced, the hydrogenase hya is RpoS induced at low pH during anaerobiosis and may play a role during acidic stress (223). The induction of the RpoS-dependent Cfa protein, which produces cyclopropane fatty acids, decreases membrane fluidity and the permeability to protons (224).

Additional stress-induced chaperones are transcribed by RpoS and active at low pH. The protein Dps protects DNA from damage in case of oxidative stress (225), and the protein chaperone Hsp31, encoded by the hchA gene, is involved in the periplasmic stress response during anaerobiosis under acidic conditions (226).

Acid stress also causes positive charge to build within the cell in the form of hydrogen ions, leading to the inversion of the membrane potential (219). A full cellular response to acid stress must, therefore, include the production of proton transporters that help restore a proper charge balance across the membrane. RpoS positively controls the chloride-proton antiporter ClcA and the fluoride exporter CrcB, which mediates the export of excess fluoride and thus counteracts the accumulation of weak acids in the cell (Fig. 7). RpoS controls the activation of many different proton transporters, such as the putative cation transporter ChaB that is hypothesized to pump out antibiotics and toxic compounds in addition to hydrogen ions, providing a role for its increased expression in anaerobic conditions at low pH (223, 227). At high pH, NhaA, a sodium-proton antiporter, is necessary for resistance to Na⁺ and high pH, dependent in stationary phase on an RpoS-dependent promoter (228).

A wider role for glutamate in the GSR?

Multiple studies have shown that glutamate accumulation has an important role in stress survival under different conditions, being necessary, together with K⁺, for both acid resistance and the osmotic response (229–231). Glutamate was also shown to stimulate RpoS activity (173, 175, 232), and its levels are lower in the absence of RpoS (233). Furthermore, RpoS induces numerous proteins involved in the biosynthesis of glutamate. The RpoS-dependent glutaminase GlsA was shown to participate in acid resistance (Fig. 8—pathway 1). Both RpoS and RpoD activate the expression of gltB and gltD coding for the glutamate synthase that catalyzes glutamine-to-glutamate conversion using oxoglutarate (Fig. 8—pathway 2). These genes were shown to be overexpressed under mildly acidic conditions (234), further suggesting the importance of glutamate accumulation under acidic conditions. Another RpoS-dependent strategy increases the intracellular glutamate pool via GabT and GabD, using GABA and oxoglutarate to produce glutamate and succinate (Fig. 8—pathway 3). These enzymes, induced by pH downshift and osmotic shock, are also involved in the degradation of lysine following its decarboxylation into cadaverine, a pathway that includes additional RpoS-regulated enzymes and produces succinate (230).

Fig 8.

RpoS-dependent glutamate synthesis pathways. Glu stands for glutamate, Gln for glutamine, and Arg for arginine. GABA is γ-aminobutyric acid. In blue are proteins under RpoS control (strong evidence), in darker blue are proteins potentially under RpoS control (GltB, GltD, and SpeA; less compelling evidence). (1) The glutaminase GlsA converts glutamine into glutamate, releasing ammonia. (2) Glutamate biosynthesis pathway through GltB and GltD; these convert glutamine to glutamate and also convert the ammonia and oxoglutarate into glutamate. (3) GABA degradation pathway. GabT converts GABA into glutamate; GabD converts the products of the reaction into succinate. (4) Putrescine import and degradation pathways. The transporters Pot and PuuP import putrescine, which is further degraded by PatA and PuuABCD, leading to succinate, glutamate, and GABA production. (5) Arginine import and degradation pathways. Arginine is imported through the Art transporter and subsequently degraded into glutamate and succinate by the Ast pathway. Arginine can also be converted into putrescine by the SpeA and SpeB enzymes.

Many additional genes involved in the metabolism of glutamate were found as potential RpoS targets in global studies but have not been examined for involvement in the acid stress response (235). Many of these genes fall into two pathways consistently found as high confidence RpoS targets (Table S2), suggesting that they may be appropriate targets for future research. First, putrescine and oxoglutarate are catabolized to GABA and glutamate using the putrescine symporter PuuP and the enzymes PatA, PuuB, PuuC, and PuuD (Fig. 8—pathway 4). The PuuBCD proteins are encoded in an operon along with the GABA aminotransferase PuuE which promotes GABA degradation, forming succinate and glutamate; PuuA and PuuP are encoded from the divergent operon. Also, part of the puuD operon is the PuuR DNA-binding repressor for both operons, probably promoting a feedback loop control on the production of the enzymes and transporter when needed (236). Second, the major arginine degradation pathway AstABCDE, which produces glutamate and succinate, and the arginine ABC transporter ArtIQMP are all under the control of both RpoS and RpoD (Fig. 8—pathway 5) (237, 238).

Acid stress response and metabolic rewiring

The ways in which bacteria such as E. coli deal with acid stress are clearly intricately connected to broader cell physiology and metabolism, and possibly, this degree of interaction is why the GSR is needed—responding to one stress is rarely without other stressful consequences for the cell.

Anaerobiosis and phosphate starvation conditions induce metabolic pathways resulting in the production of weak organic acids such as acetate, thus leading to a decrease in intracellular pH (208, 239). For example, pyruvate is metabolized by the constitutively produced pyruvate dehydrogenase during optimal growth conditions but is replaced by the pyruvate oxidase PoxB at the transition to stationary phase, during phosphate starvation, or in response to osmotic shock. PoxB synthesis is activated by RpoS and converts pyruvate produced during glycolysis into acetate and CO₂ (182, 239, 240). While the rerouting of pyruvate metabolism via PoxB has disadvantages, such as the accumulation of acetic acid and therefore a decrease in cytoplasmic pH, it is necessary for bacterial survival, as a poxB mutant leads to earlier death in prolonged Pi-starved cultures (208). Indeed, the NAD-independent PoxB may lead to a decrease in oxidative stress by reducing the production of NADH, reducing its utilization by the NADH dehydrogenases and the resulting production of H₂O₂. The mild acid stress during long-term Pi-starvation can be overcome by the addition of glutamate, inducing the Gad acid resistance system (AR2) (208). This example highlights the interconnectivity of the different stress response pathways, with the cell trading mild acid stress for decreased oxidative stress, using a network of genes regulated by RpoS.

During aerobiosis and mild acid stress, components of the electron transport chain, including dehydrogenases and the cytochrome bo oxidase, appear upregulated and participate in the export of protons, increasing the intracellular pH (210, 234). In the same conditions, the periplasmic and oxidative stress responses are also upregulated, likely to respond to the membrane burden and to the increase in oxygen radical production during acid stress (234).

Other stress response pathways

Osmotic stress response

High osmolarity leads to severe damage to the cell, as water content decreases and impacts cell shape and growth. E. coli adapts to an increase in osmolarity by accumulating K⁺ ions and other compounds called osmoprotectants (231). E. coli produces glycine betaine and trehalose as osmoprotectants during osmotic stress, as well as during cold stress and stationary phase. Glycine betaine is the most efficient osmoprotectant to protect cells from dehydration, but this relies on the import of proline or betaine. RpoS controls the expression of many genes coding for glycine betaine and proline import, such as the ABC transporters OsmF-YehYXW, as well as the symporter ProP and the porin OmpF that mediate the diffusion of other small solutes. RpoS also activates the expression of otsA and otsB, coding for the enzymes for trehalose biosynthesis. A mutant called otsX, affecting expression of otsA and otsB in different E. coli strains, turned out to be allelic with rpoS (241). As mentioned earlier, constitutive expression of otsAB is sufficient to provide osmoresistance to a strain deleted for rpoS (191). Trehalose can also be degraded and used as a sole carbon source by E. coli during low- or high-osmolarity conditions, a pathway that includes the enzymes TreA, TreF, and Glk, also under RpoS-positive control.

Additionally, both the mechanosensitive channel MscL, which allows excess intracellular solutes to exit, and the osmotic shock-induced YbiO, shown to protect cells against low osmolarity, are induced by RpoS (Table S2) and were both shown to participate in the osmotic stress response (242, 243). A set of osmotically inducible genes encoding proteins localized to the periplasmic compartment, many of which were found to also be dependent upon RpoS, has been identified and includes the outer membrane lipoproteins OsmB and OsmE, and the periplasmic chaperone OsmY, shown to inhibit the aggregation of membrane proteins (244, 245).

Import and export

Many types of channels, pumps, and transporters are under RpoS control. Notably, the efflux channel protein TolC, which shows an RpoS-dependent increase in stationary phase, pumps out many toxic exogenous compounds (e.g., antibiotics, detergents, and bile compounds) as well as some metabolites as siderophores for iron uptake (246). TolC is required for the function of many efflux pumps, including MacAB (for antibiotics and toxins), MdtEF (for extreme acid and nitrosative stress resistance), and MdtABC (for resistance to antibiotics, bile salts, and SDS stress), all three upregulated by RpoS (221, 247). MdtM, a multidrug efflux pump that acts in synergy with the AcrAB/TolC efflux system to enhance efflux of toxic compounds and bile salts is also activated by RpoS. RpoS also contributes to the expression of multiple porins, likely playing roles in the movement of small molecules into and out of the cell (Tables S1 and S2).

Stress responses and metabolism

Bacteria must trade off using energy to grow with survival and stress resistance. RpoS plays a central role in this balance, regulating metabolic processes directly and indirectly. Maharjan and co-workers created a set of E. coli strains expressing various ratios of RpoS to RpoD and showed that stress resistance increased with a high RpoS/RpoD ratio, while metabolic efficacy decreased (248). While the genes involved in stress resistance have been well studied, the basis for the decrease in metabolic efficiency is only beginning to be understood. An examination of the RpoS regulon (15% of genes in Table S2) identifies several metabolic genes, highlighting the role of RpoS in the rewiring of metabolism during the GSR. RpoS appears to take over RpoD housekeeping functions when cells encounter non-optimal growth conditions, in some cases substituting enzymes that may be less energetically efficient but avoid accentuating an existing stress or creating a stress not already present. One such example, the induction of poxB under Pi-starvation conditions, which triggers acid stress, was discussed above, and further examination of the RpoS-RpoD regulatory trade-off will likely yield further examples of the cell rewiring its metabolism to minimize stress.

Enzymes of central metabolism

RpoS positively controls genes involved in glycolysis, anaerobic respiration, fermentation, and amino acid degradation. Indeed, RpoS positively regulates the expression of many metabolic genes, including enzymes involved in the TCA cycle (e.g., FumC and AcnA), glycolytic enzymes such as the aldolase FbaB, and proteins involved in the non-oxidative pentose phosphate pathway such as TktB and TalA (249, 250). These proteins are isozymes of the RpoD-regulated enzymes FumA/FumB, AcnB, FbaA, TktA, and TalB, respectively. It was shown that RpoS-regulated isozymes take over for RpoD-regulated ones during stationary phase and limited carbon conditions, although the physiological reason behind a switch in these isoenzymes remains unclear. In the case of the FumABC isozymes, FumA and FumB both contain iron-sulfur clusters that are susceptible to oxidative stress, and thus one potential reason the iron-independent FumC takes over in RpoS-inducing conditions is to overcome FumA/B inactivation (251, 252).

RpoS positively regulates the expression of the carbon storage regulator CsrA (called RsmA in some other γ-proteobacteria), an RNA-binding protein that is a central regulator of carbohydrate metabolism and is usually associated with rapid growth. As noted above, CsrA negatively regulates translation of the anti-adaptor iraD (253). CsrA activity is negatively regulated by the non-coding RNAs CsrB and CsrC, and these are also upregulated dependent upon RpoS. This has led to the suggestion that increased synthesis of CsrA is part of a homeostatic loop, poising cells to return to rapid growth (125). Minimally, the presence of csrA in the RpoS regulon highlights the likelihood that some of the regulon actually acts negatively on RpoS.

Growth on non-optimal carbon sources and balancing acetate/acetyl-CoA levels

Multiple studies have found several genes involved in glycerol transport and metabolism are upregulated by RpoS, including the G3P importer encoded by ugpBAEC operon and metabolic enzymes such as GlpD and DhaKLM (Tables S1 and S2). The crucial role of RpoS in glycerol metabolism suggests that RpoS may play a central role in adaptation to non-optimal carbon sources, enabling stress adaptation under nutrient-limiting conditions (254).

Acetate, a common alternative carbon source, accumulates upon glucose consumption at the entry of stationary phase. It was shown that a strain deleted for rpoS harbors decreased acetate levels and biomass in batch aerobic fermentation. This effect is mainly due to the loss of the RpoS-dependent induction of the pyruvate oxidase PoxB, which promotes the accumulation of acetate in stationary phase. The expression of the PoxB pathway would also decrease acetyl-CoA levels upon entry into stationary phase (249). This perturbation of cellular metabolism also impacts RpoS synthesis, as a decrease in acetyl-CoA concentration leads to an increase in RpoS translation and stability, suggesting the crucial role of this regulator in rewiring metabolism (91). Furthermore, RpoS activity is stimulated by acetate (107, 255). Thus, acetate/AcCoA levels appear to be one of the central signals for RpoS within the cell, as high acetate upregulates RpoS presence at multiple levels.

Fermentation and respiration pathways

RpoS plays a central role in the redirection toward an anaerobiosis and fermentative energy metabolism during the transition to stationary phase, by positively regulating enzymes involved in anaerobic respiration and fermentation while indirectly downregulating aerobic respiration. RpoS activates the expression of the fumarate reductase coded by the frdABCD operon, the cytochrome bd-II coded by the appCBXA operon and the operon ynfEFGH, coding for the putative selenate reductases, oxidoreductase, and menaquinol dehydrogenase. RpoS also activates the expression of genes coding for a number of oxidoreductases, including YdbK, involved in the response to oxidative stress, possibly by reducing flavodoxin, thus preventing the formation of ROS (256). RpoS controls the expression of multiple enzymes including the hydrogenases coded by the hya operon which may also play a role in acid stress resistance (223).

Cell and membrane integrity response

Cells undergo major RpoS-dependent changes in shape and membrane architecture under stress and in stationary phase. Cells become significantly shorter. As mentioned above, RpoS and RpoD have significant overlap in their regulon, including genes essential for cell division, such as ftsQ and ftsB, required for proper septum formation (257). The changes in relative concentration of these or other cell division proteins upon RpoS induction may contribute to the shorter cell length phenotype observed under stress conditions.

RpoS regulates the expression of DNA-binding regulator BolA, involved in the regulation of bacterial shape via cell permeability, motility, and biofilm formation (258). BolA controls the expression of the mreBCD operon, which is essential for cell elongation and maintenance of cell shape (259). BolA also represses the expression of the penicillin-binding proteins DacA and DacC, which are involved in peptidoglycan maturation. In addition, BolA activates expression of the central biofilm regulator CsgD, as does RpoS itself (260, 261).

RpoS positively controls the expression of multiple membrane proteins involved in peptidoglycan maintenance, including its tethering, maturation, and recycling. Murein (Braun’s) lipoprotein, encoded by lpp, is the most abundant lipoprotein in E. coli, and promotes the tethering of peptidoglycan to the outer membrane; it is found as a putative RpoS target in multiple studies (Table S2). The outer membrane lipoprotein Blc, which is involved in lipid transport and storage under stress conditions, is also under RpoS control. Multiple proteins involved in the maintenance and remodeling of the peptidoglycan structure are also under RpoS control, including the peptidoglycan maturation proteins LdtDEFA which promote survival during stationary phase, play a major role in the remodeling, and repair the peptidoglycan during cell envelope stress and are essential in cells with outer membrane defects (262). Kbp, a potassium-binding protein that contains a LysM domain associated with cell wall degradation, is necessary for proper peptidoglycan structure. Multiple peptidoglycan recycling enzymes are under RpoS control, as well as transporters involved in the import of peptidoglycan components (see Tables S1 and S2 for details). The PqiABC and its homologs YebS (renamed LetA) and YebT, intermembrane transporters, were shown to contribute to membrane integrity during stresses and are under RpoS control (263).

High-throughput RpoS regulon studies have also identified genes involved in the synthesis of lipopolysaccharides and other components of the membrane biosynthesis pathways (see Table S2).

This broad array of membrane structural genes under RpoS control highlights the central role of RpoS to promote both the adaptation of cell shape and the repair of membranes in response to a wide set of stresses.

DNA structure, function, and repair

Stationary phase cells have RpoS-dependent changes in several properties associated with DNA structure and function: nucleoid condensation, resistance to DNA-damaging agents, and mutagenesis rates (264, 265). Many proteins that are used during exponential phase have been identified as having, in addition to an RpoD-dependent promoter, RpoS-activated promoters. For instance, transcription of the ihfA and ihfB genes, encoding the subunits of integration host factor, a nucleoid-associated protein with roles in maintaining nucleoid structure and function, increases sixfold in stationary phase, at least partially dependent upon RpoS (266). gyrB, encoding a subunit of DNA gyrase, responsible for maintaining the negative supercoiling of DNA is similarly under both RpoD and RpoS control. The role of the RpoS-dependent increase in transcription during stress/starvation phase for these genes remains to be explored (266, 267).

A major contributor to the protection of DNA during stationary phase is the very abundant Dps protein, mentioned earlier. Dps binds to DNA, promoting the compaction of the nucleoid during stationary phase and oxidative stress, preventing damaging agents from reaching the DNA. Dps also functions as a ferroxidase, promoting iron capture and detoxification, thus preventing the mutagenic Fenton reaction from occurring close to the DNA (268).

While proteins such as Dps can protect DNA from damage, RpoS also supports the expression of proteins to repair a variety of types of DNA damage, including damage caused by metabolites accumulating in stationary phase cells. Methylglyoxal is a regular product of glycolysis and degradation of some amino acids but is highly toxic, resulting in glycation of DNA and proteins. The methylglyoxal reductases DkgA and DkgB, which detoxify methylglyoxal, are upregulated by RpoS. Deglycase enzymes that repair glycated proteins as well as glycated GTP, DNA, and RNA, include YhbO and Hsp31 and are also RpoS induced (269). Alkylating agents, present in the environment as well as byproducts of metabolism, are highly toxic, as they introduce lesions into DNA and RNA [reviewed in reference (270)]. In E. coli, the pathway responsible for the repair of the alkylated DNA is called the Ada response. This response is spearheaded by the DNA-binding activator Ada, which is responsible for the expression of AlkA, AlkB, AidB, and itself; all of these are similarly regulated by RpoS (Table S1). The exonuclease III coded by xthA, induced during oxidative stress by OxyR and RpoS, has multiple functions including the repair of hydrogen peroxide-damaged DNA as well as 3′ - 5′ exonuclease and ribonuclease activity. This stress response role likely explains why the deletion of xthA is lethal in the presence of hydrogen peroxide (271).

While the pathways above can repair DNA damage, RpoS induction simultaneously leads to expression of error-prone polymerases DinB (PolIV) and PolII. A global identification of mutants that increased stress-induced mutagenesis identified rpoS and upstream direct and indirect regulators of RpoS (272). These error-prone polymerases appear to be expressed in only a subpopulation of cells experiencing a high RpoS response. This population has been called “gamblers,” suggesting that a subset of the bacterial population, under stress, “bets” on the occasional helpful mutation providing relief or resistance to intense stress among the many harmful ones. A drug that inhibits the RpoS response, and therefore blocks this “gambler” mutagenesis, was found to prevent the emergence of ROS-resistant mutants (264, 265). While the precise mechanisms that determine which cells adopt this “gambler” phenotype are thus far undetermined, such observations reinforce the heterogeneity of the RpoS response.

RpoS role in E. coli virulence

A common bacterial strategy for survival in harsh and/or competitive environments includes the development of biofilm, which is used to protect cells from biotic and abiotic stresses. RpoS is known to be the central regulator for the formation of macrocolony biofilms in E. coli (273). RpoS is essential for the production of curli and cellulose, and these genes represent about 5% of overall RpoS targets (Table S2). While the analysis of RpoS-dependent functions described above was primarily done in E. coli K-12, it is clear that RpoS is a central regulator of virulence and host colonization in many species, with the dependence on RpoS playing different roles depending on the species, its colonization method, and the experimental conditions. Although RpoS is not required for host colonization by some species, it appears to be essential for some species to invade the host (212, 274–276). This suggests an important role of RpoS as a major regulator of virulence, with different roles in different bacteria. These complicated pathways vary depending on strain and species and require specific attention, beyond the scope of this review [see more detailed reviews in references (276–279)].

CONCLUSIONS AND FUTURE DIRECTIONS

Regulation of RpoS in E. coli as a model case: converging inputs for GSRs

Here, we have reviewed our current understanding of the general stress response in E. coli, with an emphasis on the stress sigma factor RpoS and its roles. The multiple levels of RpoS control, summarized in Fig. 5 and 6, provide some understanding of how many different starvation and stress signals (Fig. 1A) converge to increase RpoS availability and to allow RpoS to compete successfully for core polymerase.

As much as we know about regulation of RpoS, it is clear that there is still more to learn, even in E. coli K-12. How the known regulators collaborate under different growth conditions is not well explored. Additional sRNA regulators and additional anti-adaptors are still to be investigated, the networks that regulate the sRNAs and anti-adaptors are not fully defined, and some levels of regulation remain poorly understood.

Comparisons between the regulatory inputs in E. coli K-12 and those in other organisms, including other E. coli, are already informative and should continue to provide further understanding of ways in which the GSR is induced. While a detailed comparison to other species is beyond the scope of this review, a few examples provide a glimpse of what might be expected. For instance, it is very clear that the regulatory cascades and mechanisms studied in this model organism can, via fairly simple changes, be reconfigured to allow bacteria to adjust to other environments and requirements. For instance, expression of the anti-adaptor IraL is high during exponential phase in uropathogenic E. coli and Shigella, where it is needed for pathogenesis (122). In Vibrio cholerae, the RssB adaptor regulates RpoS degradation, as it does in E. coli, but RssB itself is unstable, helping to drive recovery from the stress response (132). While a long 5′ UTR with secondary structure that can inhibit translation appears to be found in many organisms, the sRNAs that may interact with this clearly vary, as seen in a comparison of E. coli to V. cholerae and P. aeruginosa (83). Even if the sRNAs are the same in close relatives to E. coli K-12, it would not be surprising if promoters, for instance, varied, as they do for the anti-adaptors.

Further afield from E. coli, the general stress response in gram-positive organisms such as B. subtilis, mediated by the alternative sigma factor σB, is also regulated by multiple inputs, in this case primarily parallel cascades of anti-sigma and anti-anti-sigma proteins [reviewed in references (136, 280)]. This comparison sets one strong expectation for a central regulator of a GSR: multiple regulatory inputs, each capable of leading to availability of the central regulator.

Also clearly important for RpoS and other GSR systems are mechanisms that keep the regulator low when cells are growing well and reverse induction at the end of the stress. Implicit in this are two concepts: one, that the GSR acts as a global switch in the transcriptional program and thus behavior of the cell, useful in some conditions and harmful under others, and two, that the transition into this program likely needs to be reversed. In E. coli, both the inhibition of translation by the hairpin in the 5′ UTR of rpoS and rapid degradation of any translated protein keep the GSR off when not needed; their role in returning the cell to homeostasis is less explored, but evidence, not discussed here, suggests that the sRNAs themselves are unstable, limiting when they will stimulate RpoS synthesis (281) and that RpoS degradation resumes quickly after many stresses (98). Examination of the recovery process remains a subject that certainly deserves more attention.

Defining the effects of inducing the GSR: the RpoS regulon

If a GSR sigma factor is available and active, understanding what genes it regulates and when becomes the next challenge. Many genes, originally studied individually, have been shown to be important for stress resistance and to be dependent upon RpoS for expression. Our compilation of the results of multiple high-throughput approaches to defining the RpoS regulon (Tables S1 and S2) demonstrates how much of the RpoS regulon still needs to be understood. Of the 299 genes that we judged as reasonably “high confidence” (Table S2), 83 we list as in “unknown” pathways, even if some of the proteins have predictable activities (Table S2, Ordered per pathway then count). For many of the others, while their general pathway can be assigned, that assignment does not yet tell us why these are part of the RpoS regulon. It is certainly possible that these pathways and genes become critical when E. coli is in natural environments outside of the lab, a vast field for future study.

Particularly in need of further studies is how the RpoS GSR remodels metabolism, and why this is important. Tables S1 and S2 identify positively regulated genes; many aspects of metabolism are also negatively regulated during the GSR. As discussed here, in some cases, rerouting aspects of central metabolism is likely critical for avoiding toxic metabolite accumulation (195), but in many other cases, much more remains to be learned.

General principles for GSR regulators

Studies on a wide range of bacterial species suggest that a GSR, while not universal, is frequently found, particularly in bacteria that are not confined to a single environment. Can we use the E. coli example to help identify the most productive approaches to studying these? Some thoughts are highlighted here:

1. Is there a central regulator identified for a GSR? Sigma factors are certainly high likelihood candidates, but other regulators have been found as well. The expectation would be that a mutant that disables this regulator will lead to multiple, apparently unrelated defects in stress responses, possibly only when cells enter stationary phase. Thus, querying phenotypes under more than one growth condition will be important. Overproduction of a candidate regulator that leads to multiple resistances may also be diagnostic.

2. If such a regulator of the GSR is identified, it may be expected that overexpression/inappropriate expression, while improving stress resistance, may also be detrimental under non-stressed growth conditions. Defining the conditions when the GSR is beneficial and when it is harmful can be critical in uncovering how it is regulated.

3. Multiple inputs and thus multiple types of regulation may be expected to converge on the levels and/or activity of the GSR regulator. Those may include variations on regulated proteolysis and sRNA-dependent control of translation, as seen in E. coli, cascades of anti-sigmas and anti-anti-sigmas, as seen in many gram-positive bacteria and some proteobacteria, but are also likely to include yet unexplored regulatory mechanisms. If we have learned nothing else from the E. coli GSR, it is that everything that can be regulated will be, sometimes in ways we did not know we should be considering.

Biographies

Sophie Bouillet is a post-doctoral visiting fellow in the laboratory of Susan Gottesman, Laboratory of Molecular Biology in the National Cancer Institute at NIH. She received her PhD in microbiology from Aix-Marseille Université, France under the supervision of Dr Chantal Iobbi-Nivol, in the lab of Dr Vincent Méjean (CNRS) in 2017, where she studied a new RpoS regulation pathway in Shewanella oneidensis. She held a post-doctoral position in the lab of Dr Ann Stock, Rutgers University from 2018 to 2020, studying molecular mechanisms of histidine kinases. She then joined the Gottesman lab in 2020 to pursue her interest on the regulation of stress responses in Escherichia coli.

Taran S. Bauer received his B.S. in Biological Sciences from the Cornell University College of Agriculture and Life Sciences. At Cornell, Taran studied the virulence mechanism and disease progression of an emergent strain of Xanthomonas citri pv. malvacearum in the laboratory of Dr. Adam Bogdanove. He further studied applied agricultural research during his undergraduate career, spending a summer working with Christy Hoepting and the Cornell Cooperative Extension on pesticide efficacy in agricultural fields throughout upstate New York. He is currently a postbaccalaureate fellow in the Laboratory of Molecular Biology at the National Institutes of Health. Under the advisement of Susan Gottesman, he studies the functional conservation of RssB homologs throughout gammaproteobacteria with the goal of characterizing sigma factor regulatory mechanisms in other species. He plans to take these experiences into a Ph.D. program in the future, with plans to study molecular microbiology and host-microbe interactions.

Susan Gottesman is Chief of the Laboratory of Molecular Biology in the Center for Cancer Research, part of the intramural program of the National Cancer Institute in Bethesda, Maryland. She also holds the title of NIH Distinguished Investigator. She received her undergraduate education at Radcliffe College, working with Boris Magasanik, her PhD with Jonathan Beckwith at Harvard Medical School, and did postdoctoral fellowships at the NIH, with Max Gottesman, and at MIT, with David Botstein, before establishing her research group at the NCI in 1976. Her lab has been interested in novel signalling pathways and post-transcriptional mechanisms of regulation, primarily studied in E. coli. Studies in her lab on the roles of energy-dependent proteolysis and small regulatory RNAs in gene regulation in response to stress have both been relevant to understanding how and when the RpoS general stress sigma factor accumulates in cells.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mmbr.00151-22.

Supplemental table legends. mmbr.00151-22-s0001.docx.

Legends of Tables S1 and S2.

Table S1. mmbr.00151-22-s0002.xlsx.

Complete RpoS regulon.

Table S2. mmbr.00151-22-s0003.xlsx.

Ordered RpoS regulon, high confidence targets.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.