Chapter 93

Regulation of Gene Expression during Entry into Stationary Phase

REGINE HENGGE-ARONIS

INTRODUCTION

The natural environments of bacteria are often characterized by limiting amounts of nutrients, and therefore, periods of negligible growth or apparent dormancy are the rule rather than the exception. Rapid growth under the usual laboratory conditions is thus not representative of the natural life style of bacteria, which has been commonly referred to as a "feast-and-famine" existence. Under conditions of nutrient starvation, some gram-positive bacteria such as Bacillus subtilis differentiate into spores, whereas Escherichia coli, Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), and related enteric bacteria enter into a physiological state termed stationary phase. Although not considered differentiated, stationary-phase cells of E. coli have many properties in common with spores, such as the ability to survive prolonged periods of starvation and a strong multiple-stress resistance.

In E. coli, the primary response to the limitation of a specific nutrient is an activation of a certain set of genes that allow higher-affinity uptake of the nutrient present in low concentration or the utilization of other substances that belong to the same class of nutrients. These nutrient-specific systems include the cyclic AMP (cAMP)-cAMP receptor protein (CRP) regulon for the use of alternative carbon sources, the NtrB/NtrC/σ ⁵⁴ regulon that is induced under conditions of nitrogen limitation, and the PhoB/PhoR regulon that is induced when sources of phosphorus are limiting, as well as systems for scavenging low concentrations of iron or other essential substances (see chapters 71, 86, and 87 in this volume). If alternative nutrient sources are present in the growth medium, the cells continue to grow and divide. The cAMP-CRP system also stimulates the expression of genes involved in the synthesis of flagella and in motility and chemotaxis (74) and thus increases the probability that the bacteria find themselves in a new environment more appropriate for growth. However, if the environment is totally exhausted for an essential nutrient, the cells enter into stationary phase. In contrast to the specific responses, the nature of the stationary-phase response does not appear to be dependent on the class of the missing nutrient.

Thus, the reaction to nutrient limitation can be seen as a two-stage process (with the two stages partially overlapping). If the induction of the nutrient-specific responses remains unsuccessful, i.e., growth cannot be resumed, the stationary-phase response is induced. This secondary response involves a transition from a metabolic state aimed at maximal growth and cell division to a maintenance metabolism and the induction of many genes whose function it is to provide maximal protection against a large variety of stress conditions. However, the stationary-phase response in E. coli does not involve an all-or-none switch or an irreversible commitment. Some stationary-phase-responsive genes actually exhibit inversely growth rate-related expression and thus are already partially induced under conditions of slow growth.

The stationary-phase response profoundly influences the entire cell physiology and therefore must be tightly and elaborately controlled. Several global regulators are involved in this control, and connections to other global control circuits are now becoming apparent, indicating that gene regulation during entry into stationary phase may provide a paradigm for a complex regulatory network in prokaryotes.

STRUCTURAL AND PHYSIOLOGICAL CHARACTERISTICS OF STATIONARY-PHASE CELLS

A variety of changes in cellular morphology and physiology were observed during entry into stationary phase (recently reviewed in references 106 and 174; see also chapter 106 in this volume). E. coli cells become smaller, develop a spherical rather than a rod-shaped morphology (88, 112), and have an increased tendency to form aggregates. Their cytoplasm is condensed, whereas the volume of the periplasm increases (160). Structural changes in many cellular components have been observed in starving cells. Apparent alterations in surface charge were attributed to an altered lipopolysaccharide (90). Membranes become more resistant to freeze-thawing (177). The concentration of cardiolipin increases relative to that of other phospholipids (83, 198), and unsaturated membrane fatty acids are converted into cyclopropyl derivatives (32). Changes in peptidoglycan composition and structure result in a higher resistance to autolysis induced by penicillin or chaotropic agents (109, 192). After several hours in stationary phase, alterations in the negative superhelical density of plasmids were observed (16), and the increased cellular content of the histone-like protein H-NS may result in a compactation of the nucleoide (195). Depending on medium composition, stationary-phase cells synthesize storage compounds such as glycogen and polyphosphate (156) and protective substances such as trehalose (78). In starved cells, stable RNA and protein turnover is stimulated (161). The degradation of ribosomes is particularly pronounced under conditions of phosphate starvation (37). On the other hand, 100S dimers of 70S ribosome monomers were observed (204), which might represent a storage form of ribosomes required for the rapid resumption of growth once nutrients become again available. In contrast to bulk RNA, mRNA half-life is more than twofold increased in glucose-starved cells, regardless of whether the transcripts encode proteins whose synthesis is stimulated, repressed, or unaffected after the onset of starvation (1).

The formation of several dozen proteins is stimulated during transition into stationary phase. A core set of proteins is induced regardless of the class of nutrient for which the cells are starved (67). It was proposed that many of these proteins are involved in stress protection (129), since stationary-phase cells are resistant to high-temperature heat shock and high concentrations of H₂O₂ (96), NaCl (95), alkylating agents, ethanol, acetone, toluene, deoxycholate (75), and acidic or basic pH (116, 175). The ability to survive under these very diverse stress conditions indicates that stationary-phase cells express systems for the protection of DNA, proteins, and membranes and for DNA repair that are not found in exponentially growing cells. Stationary-phase induction of these systems is not a response to specific stress conditions, since stationary-phase multiple-stress resistance is found in cells that have never encountered a particular stress agent. Although some of the corresponding structural genes may also belong to specific stress-inducible regulons, the stationary-phase regulation of gene expression must be different from the induction of specific stress responses.

STATIONARY-PHASE-RESPONSIVE GENES

While initially most stationary-phase-inducible proteins were known only as spots on two-dimensional O’Farrell gels (67), many stationary-phase-responsive genes have been identified recently (Table 1). Their gene products have a wide array of unrelated functions and can be found in all cellular compartments (see references 75, 76, 106, and 119 for recent reviews).

Table 1Stationary-phase-activated genes in E. coli and S. typhimurium

The rpoS-encoded sigma factor σ ^S has been identified as a central regulator for many stationary-phase-responsive genes (113). The genes listed in Table 1 have therefore been grouped according to their dependency on σ ^S for expression. Three points should be noted with regard to Table 1. First, the list of stationary-phase-activated genes as given here is certainly incomplete. Especially σ ^S-dependent genes are being identified at an ever-increasing rate. Second, not all genes listed as σ ^S controlled are entirely dependent on σ ^S for expression. Often, these genes have more than one promoter and only one is controlled by σ ^S (examples are bolA [2], cfa [205], ecp-htrE [158], glgS [77], osmB [98], and proP [134]). Third, transcription of the genes listed as σ ^S regulated is not necessarily directly initiated by σ-containing RNA polymerase holoenzyme. Although few data are available, it seems likely that several genes are under indirect control of σ ^S. Cascade-like regulation within the σ ^S regulon is further described below.

IDENTIFICATION OF THE rpoS-ENCODED SIGMA FACTOR σS AS A CENTRAL REGULATOR OF STATIONARY-PHASE-RESPONSIVE GENES

Over the last 15 years, mutant alleles of rpoS have been discovered several times in different contexts. Only recently, however, it was recognized that these alleles (nur, katF, appR, and csi-2) represent a single regulatory gene locus.

The first rpoS allele identified was nur, for which a role in near-UV resistance was demonstrated (193, 194). katF was shown to be required for the synthesis of one of two catalases (hydroperoxidase II) (121) whose synthesis is growth phase controlled (120). nur and katF were found to be allelic (169). xthA, which encodes exonuclease III and is required for near-UV resistance (43, 168) and H₂O₂ resistance, was identified as a second gene under the control of katF (167). In parallel, mutations in appR that interfered with the synthesis of acid phosphatase (encoded by appA) were isolated (189). Expression of appA was known to be activated under conditions of phosphate starvation and anaerobiosis (190), and the appR mutations conferred additional growth phenotypes (189), suggesting a more global role for the appR gene. In a search for carbon starvation-responsive genes by using a gene fusion approach, the csi-2 allele was identified. A csi-2::lacZ fusion, as well as a csi-2::Tn10 subsequently isolated, interfered with starvation survival, stationary-phase thermotolerance, and H₂O₂ resistance, the expression of acid phosphatase, glycogen synthesis, and the synthesis of at least 16 proteins identified by two-dimensional gel electrophoresis (113). The csi-2::lacZ and csi-2::Tn10 insertions were localized within the katF open reading frame (113). Independently, it was also shown that a katF::Tn10 conferred pleiotropic defects (132) very similar to those described for insertions in csi-2 and that katF and appR are allelic (191).

Taken together, these studies indicated that this gene locus encodes a global regulatory protein of central importance for the control of many genes whose expression is stimulated by starvation or during entry into stationary phase. Its sequence turned out to have strong similarity to sequences of bacterial sigma factors (136), and therefore rpoS and σ ^S were introduced as appropriate designations for the gene and its gene product (113). In vitro transcription assays have recently confirmed the proposed sigma factor function of the rpoS gene product (141, 188).

CHARACTERIZATION OF σS AND ITS ROLE IN GENE REGULATION

Physical Characterization of the rpoS Gene

The rpoS gene is located at 58.9 min on the genetic linkage map of the E. coli chromosome (14), which corresponds to a location between coordinates 2885 and 2886 on the physical map (166). Its direction of transcription from the chromosome is counterclockwise. rpoS is the promoter-distal gene in an operon together with nlpD (115) which encodes a lipoprotein with a potential function in cell wall formation (87, 115). While the two nlpD promoters contribute to rpoS expression only during exponential growth, the major rpoS transcript starts at a position within the nlpD gene (111a, 186 ). A rho-independent terminator sequence is present immediately downstream of rpoS.

Sequence analysis has shown the presence of two potential start codons separated by 11 codons at the 5' end of rpoS (92, 136), and the second ATG was confirmed as the initiation codon (122). rpoS encodes a protein of 330 amino acids with a molecular mass of 37.9 kDa (92). Several variants of the rpoS gene have been found in natural populations and in various laboratory strains. Many commonly used laboratory strains (e.g., W3110 and most of its descendants) carry a stop codon within rpoS (92, 184, 185). Yet, these strains are not rpoS null mutants and, because of translational readthrough, express a certain amount of σ ^S. Long-term starvation survival of these strains is not affected, whereas trehalose synthesis is impaired (163, 184, 185) and glycogen accumulation is somewhat reduced (R. Hengge-Aronis and coworkers, unpublished results). Also, an rpoS allele in which an internal duplication of 46 bp results in a frameshift producing a protein in which the four C-terminal amino acids are replaced by 39 new residues has been found repeatedly (92, 220). This rpoS variant, which encodes a 41.5-kDa protein, arises with high frequency in aged cultures and confers to the mutant strain the ability to outgrow rpoS wild-type strains under conditions of prolonged starvation (220; see also chapter 106 in this volume). Again, this allele does not confer a null phenotype, but the mutant is somewhat attenuated with respect to several rpoS-dependent functions (92, 220).

σ S Is a Second Primary Sigma Factor

A comparison of the σ ^S amino acid sequence with sequences of other sigma subunits of RNA polymerase revealed that σ ^S is the closest relative of σ ⁷⁰ in E. coli (125). In contrast to alternative sigma factors, vegetative or primary sigma factors contain the RpoD box, a highly conserved motif in regions 2.3 and 2.4 (187), as well as characteristic motifs of 14 and 20 amino acids in regions 2.4 and 4.2, respectively (125). With a few exchanges, all of these stretches of conserved amino acids are also present in σ ^S (75, 125). σ ^S was thus classified as a nonessential sigma factor with strong homology to vegetative sigma factors (125). However, while σ ^S is nonessential for growth in the laboratory, it is essential for survival under conditions of nutrient starvation (113, 132, 137), i.e., conditions that prevail in nature.

In vitro, σ ^S binds to purified core RNA polymerase, and this binding can be competed for by the major sigma factor σ ⁷⁰ (141). In vitro transcription by σ ^S-containing RNA polymerase holoenzyme has been shown for various template genes. Surprisingly, bolA and xthA, which are σ ^S dependent in vivo (112, 167), were transcribed in vitro by polymerase containing either σ ^S or σ ⁷⁰. However, bolA transcription was stronger when initiated by σ ^S, and the transcriptional start sites were not identical in σ ^S- and σ ⁷⁰-containing assays (141). In an independent study, in vitro transcription with σ ^S-containing RNA polymerase was observed with typically σ ⁷⁰-recognized promoters such as trp, lacUV5, rnaI, and dnaQp2, and only one gene (fic) was found to be exclusively transcribed by σ ^S (188).

Taken together, the data show that σ ^S is not only structurally but also functionally closely related to σ ⁷⁰ and therefore may be regarded as a second primary sigma factor in E. coli.

σ S Promoter Consensus Sequence

There is good genetic evidence that regions 2.3/2.4 and 4.2 of σ ⁷⁰ are involved in binding the –10 and –35 promoter regions (summarized in reference 68). This was recently confirmed biochemically with σ ⁷⁰ constructs that lacked N-terminal regions of various lengths and consequently could directly bind to promoter DNA (49). The promoter-binding regions are highly homologous or even identical in sigma factors of different species that recognize the same promoter consensus sequence (e.g., σ ^{70 in} E. coli and σ ⁴³ in B. subtilis). σ ^S is quite similar to σ ⁷⁰ in these regions, which is consistent with the finding that in vitro, σ ^S and σ ⁷⁰ have overlapping promoter specificities (141, 188). However, there are σ ⁷⁰-dependent promoters that are not transcribed by σ ^S in vitro (mostly stringently controlled promoters), and conversely, the fic promoter is recognized by σ ^S but not by σ ⁷⁰ (188). This finding indicates that the promoter determinants recognized by these two sigma factors are probably similar but not entirely identical. An alignment of promoters that either in vivo require σ ^S for activity or are transcribed in vitro by σ ^S-containing RNA-polymerase suggests a –10 consensus (TATACT) similar to that of σ ⁷⁰ (TATAAT). However, the general similarity of σ ^S- and σ ⁷⁰-dependent promoters has hampered attempts to directly derive a precise σ ^S consensus by simply comparing promoter sequences and also suggests that promoters carrying a TATAAT-like sequence in the –10 region should not be routinely classified as σ ⁷⁰ controlled.

Among the known σ ^S-dependent promoters, no clear homology in the –35 region is apparent. This may reflect a requirement for additional cis-acting regulatory factors for the expression of many of these genes that can compensate for inefficient binding by RNA polymerase in the –35 region. Integration host factor (IHF) was recently found to be required for σ ^S-dependent transcription of the dps promoter in stationary-phase cells. Interestingly, transcription from the same promoter is also stimulated in response to H₂O₂ in growing cells but then requires σ ⁷⁰ and OxyR as an activator (8).

The identification of a σ ^S consensus sequence is further complicated by cascade regulation within the σ ^S regulon. With very few exceptions, it is unclear which of the promoters of the genes that in vivo require σ ^S for expression is under direct control of σ ^S. It seems that direct recognition by σ ^S would have to be demonstrated by the in vitro transcription with just the minimal set of purified components. Unfortunately, however, the in vitro transcription systems mentioned above do not produce the degree of specificity observed in vivo. Whether a promoter is predominantly recognized by a σ ^S- or σ ⁷⁰-containing holoenzyme in vivo is probably determined not only by the promoter recognition specificities of the two sigma factors but also by the relative amounts of the two holoenzymes, by the form of core enzyme involved, by the activities of cis-acting regulatory factors that may differentially influence binding and open-complex formation at a given promoter by the two holoenzymes, and by the topological state of the DNA in and around the promoter region. During transition into stationary phase, the cellular content of σ ^S increases severalfold relative to that of σ ⁷⁰ (63, 114, 188) and modified forms of σ ⁷⁰-containing RNA-polymerase with slightly different promoter recognition can be observed (154). In vitro transcription systems cannot reflect this complexity, and thus do not produce the same specificity as that observed in vivo.

As a way out of this dilemma, in vivo experiments have been performed with a mutant that produces temperature-sensitive σ ⁷⁰. Stationary-phase induction as well as osmotic induction of osmY (csi-5) is strongly reduced in rpoS mutants (although expression and even regulation can still be observed at a very low level, indicating that in the absence of σ ^S, σ ⁷⁰ can take over to some extent) (79, 111, 207, 215). It was shown that a shift to the nonpermissive temperature in the rpoD(Ts) mutant background did not affect osmotic activation of osmY, whereas osmoregulation of proU, known as a σ ⁷⁰-dependent gene (97), was strongly impaired, indicating that the osmY promoter is most likely under direct control of σ ^S (215).

A Role for σS in Growing Cells

While the cellular σ ^S concentration is highest in early-stationary-phase cells, exponentially growing cells also maintain a certain low level of σ ^S (63, 114, 188). Correspondingly, some σ ^S-dependent genes are also expressed during exponential growth. Apparently, expression of xthA is hardly stationary phase activated at all (167). Also, the uninduced level (in the absence of exogenous H₂O₂) of the katG product, hydroperoxidase I, is controlled by σ ^S (91).

The expression of many σ ^S-dependent genes is subject to hyperosmotic activation in exponential phase under conditions in which the cells continue to grow. Whereas in the cases of proU (127) and proP (134), the osmotically controlled promoters are not the σ ^S-dependent ones, there are many examples of σ ^S-controlled genes in which growth phase control and osmotic regulation occur at the same promoter. These genes include osmY (79, 216), osmB (98), otsBA (64, 78, 100), treA (21, 71, 162), bolA (79), and dps (pexB) (124). In addition, at least 16 unidentified proteins observed on two-dimensional O’Farrell gels exhibited σ ^S-dependent osmotic induction (79). Initially it was thought that σ ^S itself was unaffected by increased osmolarity, because an early translational rpoS::lacZ fusion was not osmoregulated (79). Recently, however, it has become clear that σ ^S expression is osmotically stimulated by a mechanism that operates at the levels of translation and σ ^S protein turnover (114) (see below). Thus, σ ^S should not be seen as a regulator exclusively involved in stationary-phase gene regulation.

REGULATION OF THE CELLULAR σS CONTENT

Experiments with rpoS::lacZ fusions indicated that rpoS expression is stimulated in stationary phase. This increase is most pronounced in LB medium (113, 137, 171). In minimal medium, starvation for glucose or ammonia results in weak induction only, whereas phosphate depletion strongly stimulates rpoS expression (113, 207). Because of the nature of the fusions used in these early studies, these results demonstrated transcriptional induction of rpoS.

An increase in cellular σ ^S content during entry into stationary phase was also directly shown by immunoblot analysis. It was found that the intracellular concentration of σ ^S strongly increases under all starvation conditions tested, including transition into stationary phase in complex medium and starvation for glucose, phosphate, and amino acids (63, 114, 188). The rate of σ ^S synthesis as determined by pulse-labeling experiments increases fivefold during late exponential phase and reaches a maximum at the onset of starvation but decreases again in early stationary phase (114).

Transcriptional Control

All rpoS::lacZ fusions constructed in vitro in the various expression studies contained rather extended upstream segments (114, 122, 131, 137), and the analysis of upstream deletions indicated that more than 1 kb upstream of rpoS has to be present for full wild-type expression of the gene (115).

Evidence has been presented that rpoS is the second gene in an operon (115), with nlpD, the structural gene for a lipoprotein (87, 115), being promoter proximal. The two nlpD promoters are not growth phase controlled and contribute to low-level expression of rpoS in exponentially growing cells (115). The major rpoS-specific promoter, however, is located within the nlpD open reading frame. By subcloning smaller fragments of the region upstream of rpoS into a promoter probe vector, several segments that conferred promoter activity were found (with the most upstream segment carrying the nlpD promoters). Two transcriptional start sites located within the nlpD open reading frame were localized by primer extension analysis (186). Independently, one rpoS-specific transcriptional start site within nlpD was identified, and the corresponding promoter, which exhibits homology to the σ ⁷⁰ consensus, was designated rpoSp1 (111a). This start site corresponds to the second one in the formerly mentioned study (186), and a deletion analysis indicated that the region containing the first potential transcriptional start site as identified by Takayanagi et al. (186) does not contribute to rpoS expression (111a). In addition, primer extension experiments have shown the presence of a second, shorter transcript that appears to be a stable degradation product of the primary transcript originating at rpoSp1 rather than an independent transcript starting further downstream (111a). While the nlpD-rpoS operon promoters are not stationary phase activated (115), the rpoS-specific promoter located within nlpD exhibits growth phase-dependent activity (111a, 186).

From results of experiments with transcriptional fusions, it has been concluded that rpoS transcription is negatively controlled by the cAMP-CPR complex. In exponential phase, rpoS::lacZ fusions exhibit strongly increased expression in a Δ cya strain, which decreases immediately after the external addition of cAMP (113, 114). Putative CRP-binding sites are present in the region of the rpoS-specific transcriptional start site, but an involvement of these putative sites in the transcriptional control of rpoS has not been investigated.

Translational Control

Several studies have shown that translational rpoS::lacZ fusions exhibit higher factors of activation during transition into stationary phase than transcriptional fusions, even when the two types of fusions have identical fusion joints and upstream regions (114, 122, 131). Also, there is a clear discrepancy between the strongly increased cellular σ ^S content in glucose-starved cells (63, 114) and a fivefold-increased rate of σ ^S synthesis in late exponential phase in the one hand, and the lack of activation of transcriptional rpoS::lacZ fusions under the same conditions on the other hand (113, 114, 131, 207). All of this evidence indicates that rpoS expression is subject to posttranscriptional control. In addition, only translational rpoS::lacZ fusions encoding hybrid proteins containing substantial parts of σ ^S exhibit more than 10-fold osmotic induction (114), and so does σ ^S itself in immunoblot experiments (A. Muffler, D. Traulsen, R. Lange, and R. Hengge-Aronis, submitted for publication). No osmoregulation, however, can be observed for a transcriptional fusion with an identical fusion joint or for an early translational fusion (114). Recently, it was found that a stimulation of translation as well as an inibition of σ ^S turnover contribute to osmotic induction (Muffler et al., submitted).

rpoS mRNA is predicted to form a stable secondary structure in which the translational initiation region (comprising a Shine-Dalgarno sequence, the initiation codon, and a downstream box with complementarity to a 3' region in 16S rRNA) is involved in base pairing. Such a complex secondary structure has also been proposed for mRNA of rpoH (138, 219), which encodes the heat shock sigma factor σ ³² (70) and which was also shown to be under translational control (101, 138, 219; see also chapter 88 in this volume). In both cases, the translational initiation regions would be occluded by base pairing, and it is hypothesized that this secondary mRNA structure is resolved under inducing conditions. It seems likely that specific proteins that respond to growth phase, osmolarity, or temperature signals are involved either in the maintenance of these mRNA secondary structures under noninducing conditions or in their resolution once the respective signal is present.

While σ ^S does not control transcription of its own structural gene (114, 171), it seems to be involved in a negative feedback control of its own translation. This is suggested by the finding that late translational rpoS::lacZ fusions, which are subject to translational control, exhibit higher levels of expression in rpoS mutants than in isogenic rpoS ⁺ strains (114, 131). Most likely, this feedback mechanism is indirect and involves the activity of an unknown protein whose expression is dependent on σ ^S.

Control of σS Stability

Again like σ ³² (183), σ ^S is an unusually unstable protein in growing cells. Its half-life was determined as 1.4 or 2.5 min, as shown in pulse-labeling experiments followed by immunoprecipitation (114) or by inhibiting total protein synthesis with chloramphenicol and following the decay of σ ^S (186), respectively. At the onset of starvation in glucose-limited minimal medium, the σ ^S half-life changes to more than 10 min, and this change in stability substantially contributes to a continuing increase of the cellular σ ^S content since at the same time, the rate of σ ^S synthesis, which is maximal at the end of the exponential phase, is reduced again (114). In response to osmotic upshift, the σ ^S half-life changes to more than 45 min (Muffler et al., submitted).

For σ ^S, the molecular mechanism of differential control of protein stability has not been elucidated at the molecular level. σ ³² has been found to be stabilized by overproduction of phage λ cIII protein (15). cIII also interferes with the degradation of λ cII protein, which is the master regulator in the lysis-lysogeny decision in the life cycle of phage λ (17, 84). cII stability is governed by the HflB (FtsH) protein, which seems to be a component of a protease (17, 80). Recent studies have indicated that σ ^S turnover requires RssB, a novel type of two-component response regulator (A. Muffler, D. Fischer, S. Altuvia, G. Storz, and R. Hengge-Aronis, submitted for publication).

Signals Involved in the Control of σS Expression and Stability

The multilevel control of the cellular σ ^S content provides the potential for multiple signals being involved. While some intracellular signal molecules have recently been identified, the molecular details of their action in the control of σ ^S expression and/or stability have yet to be clarified.

cAMP as present in the cAMP-CRP complex plays a negative role in the expression of rpoS (113, 114). It is possible that the lack of transcriptional induction of rpoS in a culture starved for glucose (113, 114, 131) is due to the burst in cAMP production at the onset of glucose starvation (26). The physiological role of this interference of cAMP with rpoS transcription could be to avoid strong activation of rpoS and therefore the stationary-phase response in a carbon source downshift situation that can be managed by relieving catabolite repression.

Recently, evidence for a homoserine lactone-dependent signaling pathway that stimulates rpoS expression was presented (85; see also chapter 106 in this volume). A thrA mutant that does not synthesize homoserine and homoserine-phosphate was shown to have strongly reduced σ ^S levels. This defect could be suppressed by the external addition of homoserine lactone. In addition, an enzyme which, when overproduced, renders the pathway inoperative was identified (85). Homoserine lactones are produced and excreted into the external medium by marine Vibrio species, Agrobacterium tumefaciens, Pseudomonas aeruginosa, and other species and allow the bacteria to monitor their own population density. When the homoserine lactone (the "autoinducer") reaches a certain threshold concentration, a signaling pathway is activated that, for instance, stimulates the expression of the lux genes in Vibrio fischeri and Vibrio harveyi (60).

In a relA spoT double mutant, which is essentially free of guanosine 3',5'-bispyrophosphate (ppGpp) (211), the cellular σ ^S content is greatly decreased. Nevertheless, a weak induction following starvation for carbon, phosphate, or amino acids can be observed. The concentration of σ ^S seems to follow that of ppGpp, which can be manipulated by using a plasmid-encoded truncated relA allele under tac promoter control (63). ppGpp levels increase following the depletion of amino acids (triggering the stringent response), upon starvation for carbon, for nitrogen, and for phosphate (see chapter 92 in this volume), as well as in cells growing slowly in a chemostat (29). ppGpp thus seems to be a general starvation signal and may also be a signal that reflects the growth rate of a culture. Experiments with 5' deletions upstream of rpoS indicate that ppGpp does not influence transcriptional initiation at the nlpD or rpoS promoter, but that the lack of ppGpp results in a lower amount of transcript due to an uncoupling of transcription and translation followed by premature transcriptional termination (111a) in a way similar to that observed for lacZ (55, 200). In addition to its role in starvation, ppGpp may be involved in the osmoregulation of rpoS, since the cellular ppGpp level is transiently increased in response to osmotic upshift (73). In light of the finding that ppGpp is involved in the control of the σ ^S level in the cell and therefore influences the stationary-phase response, the physiological role of the stringent response may have to be reevaluated (see chapter 92 in this volume). A link between the stringent response and the general starvation response was also found in a marine Vibrio strain (56, 57, 143).

Yet another signal for σ ^S expression is the intracellular concentration of UDP-glucose. pgm and galU mutants, which are deficient in phosphoglucomutase and UDP-glucose pyrophosphorylase, respectively, and therefore do not produce UDP-glucose when grown on most carbon sources, show increased levels of σ ^S and higher expression of σ ^S-dependent genes. In addition, osmotic activation of the expression of these genes is reduced (20). UDP-glucose therefore appears to be a negatively acting signal molecule that down-modulates σ ^S levels under noninducing conditions. UDP-glucose influences the posttranscriptional regulation of rpoS (A. Muffler and R. Hengge-Aronis, unpublished results), but its molecular mechanism of action is not yet understood.

Experiments with gene fusions have also provided evidence that rpoS transcription in growing cells can be stimulated by a dialyzable, heat-stable factor present in spent medium of stationary-phase cultures (137). In addition, it was shown that weak acids such as acetic, propionic, and benzoic acids also stimulate the expression of transcriptional rpoS::lacZ fusions, leading to the suggestion that rpoS expression is modulated by the internal pH of the cell (137, 171). Alternatively, acetic acid itself, which is produced and excreted during growth on glucose, may be a signal for rpoS expression. However, the expression of a transcriptional rpoS::lacZ fusion is not affected in poxB and pta ackA mutants, i.e., mutants that are deficient in acetate production and metabolism (28).

THE STATIONARY-PHASE REGULATORY NETWORK

During transition into stationary phase, the expression of σ ^S-dependent genes is activated in a certain temporal order. In rich LB medium (for which most data are available), the expression of otsBA (78) and bolA (2, 19, 112) is activated early during the transition phase, whereas the expression of osmY (207), treA (78), and glgS (77) is stimulated somewhat later. osmB (78, 98) and the S. typhimurium spv genes, when cloned into E. coli (54, 142), are increasingly expressed approximately at the beginning of the stationary phase, and acid phosphatase (AppA) activity increases only after at least 2 h in stationary phase (66). Moreover, there are subsets of σ ^S-controlled genes that exhibit differential regulation in response to additional signals besides starvation or growth phase. For instance, expression of the cyxAB appA (also termed appCBA) operon (11, 12, 36), hyaABCDEF (23), and aidB (201, 202) is activated by anaerobiosis, and that of otsBA (64, 78), treA (21, 72), osmB (72), osmY (79, 216), bolA (79), dps (pexB) (124), and many others is stimulated in response to increased medium osmolarity. This corresponds to an increase in cellular σ ^S content in hyperosmotically stressed cells (114). However, csgA expression is highest at low osmolarity (150), and csiD and csiE are not osmoregulated (C. Marschall and R. Hengge-Aronis, Mol. Microbiol., in press). Also, regulatory patterns of certain σ ^S-dependent genes such as osmY (111) and mcc (135) are retained in rpoS null mutants although at greatly reduced absolute levels of expression.

All of these findings indicate that the expression of σ ^S-dependentgenes does not depend solely on the concentration of σ ^S in the cell but that additional regulatory factors differentially modulate the expression of these genes. Moreover, these factors may also control σ ^S-independent stationary-phase-responsive genes.

Involvement of Additional Global Regulators (CRP, Lrp, IHF, H-NS)

Many σ ^S-dependent genes are also under positive or negative control of other well-known global regulators whose role in stationary-phase gene regulation has only recently become evident. These regulatory dependencies are schematically summarized in Fig. 1. In most cases, the effects on the expression of gene fusions of mutations in the regulatory genes were studied in vivo, and the molecular details of the regulatory relationships have yet to be elucidated.

Fig. 1Regulatory dependencies within the stationary-phase regulatory network of E. coli. At the top are listed regulatory proteins that pleiotropically affect the stationary-phase regulation of the genes indicated below. Only stationary-phase-regulated genes for which an effect of at least two regulators has been found are included. Positive and negative influences are indicated by arrows and bars, respectively. These influences may be direct or indirect, and no information concerning the molecular mechanisms is included. Where available, these details and the corresponding references are given in the text and in Table 1.

Source:

cAMP-CRP has been implicated as an activator for approximately two-thirds of the carbon starvation-responsive genes (67, 172). Positive control by cAMP-CRP was shown for several cst genes (18, 172, 173), glgS (77), csiD and csiE (207), and poxB (28), as well as for the microcin C7 operon (mcc) (135). CRP boxes are present upstream of the transcriptional start sites of cstA (173) and glgS (77) at positions consistent with an activator function. Characteristically, the addition of cAMP to a Δ cya mutant does not immediately stimulate the expression of these genes, but additional factors that are present only during entry into stationary phase seem to be required (18, 207). At least for csiD and csiE, this additional factor may be a certain cellular level of σ ^S. On the other hand, some pex genes (129), rpoS itself (113, 114), and several σ ^S-dependent genes such as bolA (112), osmY (csi-5) (111, 207), csiF (207), and the Salmonella spv virulence genes (146) are under negative control of cAMP-CRP. For instance, late-exponential-phase expression of osmY (but not expression during early exponential phase and in stationary phase) is strongly increased in a cya crp mutant. This finding indicates that cAMP-CRP acts as a negative transition state regulator that determines the time of induction rather than the absolute level of expression of osmY in stationary phase (111). Also, in cells grown in minimal medium, cAMP-CRP inhibits expression of osmY (111) and bolA (112), and the sharp increase in the intracellular cAMP concentration (26) may be responsible for the weak activation of expression of these genes following glucose starvation. A recent study of global gene expression patterns suggests that repression by cAMP-CRP under conditions of glucose starvation is quite common in E. coli (30).

Leucine-responsive regulatory protein (Lrp) is a global regulator involved in the regulatory changes that follow a nutrient downshift from rich to minimal medium. Lrp represses genes required for the uptake and metabolism of nutrients present in rich medium and activates the expression of various biosynthetic genes (140; see also chapter 94 in this volume). In the case of osmY, Lrp plays a role similar to that of CRP. In rich medium, it acts as a negative transition state regulator, whereas in minimal medium, it interferes with osmY expression throughout the growth cycle (111). By contrast, Lrp seems to act as a positive regulator for stationary-phase expression of csiD (C. Marschall and R. Hengge-Aronis, unpublished results). Increased transcription of lrp was observed during slow growth in minimal medium, compared with rapid growth in rich LB medium (117).

IHF is a sequence-specific histone-like protein with a variety of functions that all involve the formation of higher-order nucleoprotein complexes, in which the DNA-bending activity of IHF may be crucial for proper arrangement of the other constituents (51, 59). IHF has a negative function in the regulation of osmY expression similar to that found for cAMP-CRP and Lrp (111). A positive role for IHF was found in the stationary-phase activation of expression of dps (8) and of the microcin B17 operon (mcb) (135), although the former is σ ^S dependent and the latter is not controlled by σ ^S (19, 112). The cellular level of IHF itself increases approximately 10-fold during transition into stationary phase (48), and increased transcription of hip (himD) and himA, the structural genes for the two IHF subunits, is partially σ ^S dependent under these conditions (13).

In contrast to CRP, Lrp, and IHF, which can be either positive or negative regulators of different stationary-phase-responsive genes, the histone-like protein H-NS inhibits expression in nearly all cases studied so far. Expression of mcc and csgA is reduced in rpoS mutants and is restored in rpoS hns double mutants, indicating that σ ^S can overcome H-NS-mediated repression (135, 150). Also, the transcription of the σ ^S-controlled spv virulence locus in S. typhimurium is negatively regulated by H-NS (147). Mutations in hns generally increase exponential-phase expression of many σ ^S-controlled genes and proteins (assayed as gene fusions and identified by two-dimensional gel electrophoresis, respectively). Also the expression of σ ^S itself is stimulated by a posttranscriptional mechanism in hns mutants (17a). These findings implicate H-NS as a pleiotropic exponential-phase inhibitor for the expression of many stationary-phase-responsive genes. A notable exception to this rule is provided by csiD, whose expression in exponential phase is only weakly affected by hns mutations but that seems to require H-NS for full expression during stationary phase (17a). While H-NS is already an abundant protein in growing cells, its concentration further increases during entry into stationary phase (44, 178, 195). Since exponential-phase inhibition by H-NS of the expression of many genes seems to be relieved in stationary phase, the main function of H-NS in starved cells could be in chromosome organization.

As apparent from Fig. 1, different stationary-phase-regulated genes are controlled by different combinations of regulatory factors. For instance, osmY expression is activated by σ ^S and inhibited by cAMP-CRP, Lrp, IHF, and H-NS (111), whereas the expression of mcc requires σ ^S and cAMP-CRP as positive factors but is inhibited by H-NS (135). The use of different combinations of regulators may explain the complex fine regulation of stationary-phase-regulated genes with respect to the time and extent of activation during entry into stationary phase in various growth media and in response to additional stress conditions. A large variety of regulatory patterns may result from combining a relatively small number of regulators, especially if these regulators can act either positively or negatively. Some of these factors act as transition state regulators that are responsible for an appropriate timing of expression of stationary-phase-activated genes. They may thus avoid stimulation of expression of these genes in a nutrient downshift situation that can be coped with by activating biosynthetic genes, by relieving catabolite repression, or by inducing alternative nutrient-scavenging systems (111). A similar concept of transition state regulators has been developed for gene regulation during the period that precedes sporulation in B. subtilis (182).

From Fig. 1, also many questions become apparent. Do the regulatory factors directly or indirectly control the expression of the genes indicated? Do CRP, Lrp, IHF, and other regulators act together with σ ^S, or are these regulatory relationships independent of each other? Does the structural and functional organization of the promoter region of each of the target genes reflect the multicomponent regulation? The answers to these questions are just beginning to emerge and are summarized in the following section.

Cascade Regulation within the σS Regulon and Connections to Other Regulatory Circuits

There is evidence that several σ ^S-controlled genes encode regulatory proteins (Fig. 2). AppY contains a helix-turn-helix motif characteristic of many DNA-binding proteins and seems to be a member of the AraC family of regulators (12). It controls the expression of the cyxAB appA (11, 12, 36) and hyaABCDEF (23) operons, whose expression is stimulated under anaerobic conditions. Besides being σ ^S controlled, appY expression is also regulated by the ArcB/ArcA two-component system and is thus under oxygen control (T. Atlung, personal communication).

Fig. 2Cascade regulation within the ?S regulon and connections to other environmentally controlled regulons in E. coli. For details and the corresponding references, see the text.

Source:

BolA is a helix-turn-helix protein (2) involved in cell shape determination (4). The only gene identified so far that seems to be under BolA control is dacC, which encodes penicillin-binding protein 6 (2), a protein involved in septum synthesis (139) and stabilization of peptidoglycan in nongrowing cells (199). Penicillin-binding protein is induced during transition into stationary phase in a σ ^S-dependent way (25, 50).

Dps is a protein present in high concentration in late-stationary-phase cells. In vitro, Dps forms highly structured complexes with DNA, although it does not exhibit a typical DNA-binding motif. As shown by two-dimensional gel electrophoresis, the expression of as many as 23 unidentified starvation-induced proteins is affected in a dps mutant (7). Like H-NS (152) and possibly Lrp (35, 206), Dps may be one of several abundant DNA-binding proteins that have regulatory as well as chromosome-organizing functions. In addition, the dps (pexB) mutant is highly susceptible to killing by hydrogen peroxide (7), and recently it was found that Dps expression is also stimulated by the external addition of hydrogen peroxide. For this induction the H₂O₂-sensitive regulator OxyR and σ ⁷⁰ are required, although the same promoter is used as during σ ^S-dependent stationary-phase induction (8, 124).

Also, a csiD mutant exhibits pleiotropic alterations of the total protein synthesis pattern as determined by two-dimensional gel electrophoresis, indicating that CsiD has a regulatory function (D. Weichart and R. Hengge-Aronis, unpublished data). The only known phenotype of the csiD mutant is a partial defect in stationary-phase thermotolerance. Besides requiring σ ^S, csiD expression is positively controlled by cAMP-CRP (207) and by Lrp. In contrast to that of many other stationary-phase-regulated genes, csiD expression in growing cells is hardly affected by mutation in hns but seems to be positively controlled by H-NS in stationary phase (17a).

The csgDEFG operon encodes several proteins involved in the control of expression of the curli subunits encoded by csgBA. The C terminus of the csgD gene product is homologous to the DNA binding domains of members of the LuxR family of transcriptional activators. The csgD gene is transcribed in a σ ^S- and OmpR-dependent manner during entry into stationary phase and is essential for the expression of the csgBA operon. CsgD in conjunction with CsgG also stimulates the transcription of the csgDEFG operon. In addition to being growth phase controlled, the two csg operons are transcribed only at low osmolarity, probably as a result of activation by the nonphosphorylated form of OmpR (M. Hammar and S. Normark, personal communication).

These examples illustrate that within a branched regulatory cascade, secondary regulators can be the point of integration of additional signals that differentially influence the expression of certain subsets of genes within the larger regulon (Fig. 2). Stationary-phase induction of certain gene products is thus linked to oxygen control via AppY, to the OxyR-mediated oxidative stress response via Dps, and to catabolite repression via CsiD. In addition, differential control at the level of the secondary regulators can neutralize or even revert regulation exerted at a higher level of the hierarchy, as exemplified by osmotic regulation within the σ ^S regulon. Whereas the expression of most σ ^S-dependent genes is hyperosmotically activated because the level of σ ^S itself increases under conditions of high osmolarity, the requirement for an additional activator that is present only at low osmolarity (nonphosphorylated OmpR) results in the opposite regulatory pattern for the csgDEFG and csgBA operons.

σ S-Independent Stationary-Phase Gene Regulation

While a majority of the known stationary-phase-responsive genes require σ ^S for expression, there are genes whose expression is not affected by mutations in rpoS (Table 1). On two-dimensional O’Farrell gels, some 20 proteins have been observed to exhibit σ ^S-independent induction in carbon-starved cells (132). Among these are DnaK and GroEL, whose formation and therefore also stationary-phase induction require σ ³² (94). The expression of transcriptional lacZ fusions to rpoH, the structural gene for σ ³², is stimulated during entry into stationary phase in complex medium, and this activation is not affected by a mutation in rpoS (R. Hengge-Aronis, unpublished results).

Some σ ^S-independent stationary-phase-regulated genes seem to require ppGpp for expression. The cellular level of ppGpp increases in response to starvation for amino acids and sources of carbon, nitrogen, and phosphate or under conditions of slow growth, and two synthases (encoded by relA and spoT) are involved in this increase (see chapter 92 for details). ppGpp thus appears to be a general starvation signal. The expression of rpoS itself is strongly reduced in ppGpp-free relA spoT double mutants (63). Recently, it was shown that the formation of the stringent starvation protein (SspA) is stimulated under all starvation conditions mentioned above and is reduced in a relA mutant (210). As demonstrated by two-dimensional gel electrophoresis of total cellular protein, SspA also seems to have a global regulatory function (210). Moreover, SspA has the ability to bind to RNA polymerase (89, 210). A relA dependence of starvation activation was also described for the genes stiA, stiB, stiC, and stiE in S. typhimurium (179). However, stiA and stiC are positively σ ^S controlled, and it could be that ppGpp acts indirectly via σ ^S (if σ ^S expression in S. typhimurium is ppGpp dependent as it is in E. coli). Interestingly, stiB is under negative control of σ ^S, and it has been suggested that σ ^S may be required for the formation of an unknown repressor for stiB. The relA gene product may be involved in overcoming the action of this repressor (149). Stationary-phase induction of hip (himD) and himA (encoding the two subunits of IHF) was greatly diminished in a strain unable to produce ppGpp because of a relA spoT double mutation. Since this effect was more pronounced than that of an rpoS mutation, ppGpp stimulates the expression of these genes by a mechanism that is at least partially independent of σ ^S (13).

In vitro expression of glgC and glgA (which are part of an operon encoding the glycogen synthetic enzymes) can be stimulated by ppGpp (165). In vivo, the expression of a glgC::lacZ fusion correlates with the intracellular ppGpp content (157). The glgCAP operon is not σ ^S controlled, although rpoS mutants do not produce glycogen (77). Whether glgCAP is subject to cAMP-CRP control is still a matter of debate. The finding that the expression of a lacZ fusion in the chromosomal copy of glgA is affected neither by a Δ cya mutation nor by the external addition of cAMP indicates that cAMP-CRP is not involved in the transcriptional regulation of the glgCAP operon (207). On the other hand, a 10-fold stimulation by cAMP of the synthesis of glcogen synthase (GlgA) was found in an in vitro coupled transcription-translation system (165), and the expression of a plasmid-encoded glgC::lacZ fusion is 5-fold reduced in a cya mutant (157). In addition, the glgCAP and glgBX operons are under the control of the product of the csrA gene, a pleiotropic negative regulator that also affects the expression of gluconeogenic genes (164).

Some other σ ^S-independent genes, e.g., the cst genes, require cAMP-CRP for expression (18, 173). cAMP-CRP-dependent genes previously known to be involved in nutrient uptake in growing cells also exhibit some degree of stationary-phase activation, especially when exposed to carbon starvation (5, 38, 40, 113). The cstA gene may belong to this category, since it apparently encodes a peptide transport system (173). Glucose starvation activation was also shown for phoA fusions to mglB and rbsB, encoding the periplasmic binding proteins for galactose/glucose and ribose, respectively, and to lamB, the structural gene for an outer membrane sugar-specific porin (5). Under glucose-limiting conditions in a chemostat, the expression of the mgl system and of lamB is even higher than in exponentially growing cells in the presence of specific exogenous inducers and confers a growth advantage at very low substrate concentrations (38, 40). As shown for the galactose-transporting systems, the inducer galactose is synthesized internally under limiting nutrient conditions (39).

The stationary-phase-induced microcin B17 operon (mcb) is under positive control of OmpR (81, 82) and is negatively regulated by MprA (41, 42). While OmpR is required for a wild-type level of expression of mcb, stationary-phase activation, albeit at a reduced level, was also observed in an ompR mutant background (31).

It should also be mentioned here that one can isolate from stationary-phase cells three forms of σ ⁷⁰-containing RNA polymerase (S1, S2, and S3) which differ from the single form (L1) in growing cells. It is believed that these forms carry covalent modifications, and it was shown that they have somewhat different promoter preferences in in vitro transcription assays (153, 154). It may be that the S forms of RNA polymerase stimulate expression of some stationary-phase-responsive genes.

MOLECULAR ORGANIZATION OF STATIONARY-PHASE-ACTIVATED PROMOTER REGIONS

For several stationary-phase-regulated genes, the transcriptional start sites have been mapped, and in a few cases, deletion analyses have provided some evidence as to the location of upstream regulatory sequences. So far, no features exclusively characteristic of these growth phase-controlled promoter regions have been found. Also, no clear differences between σ ^S-dependent and σ ^S-independent promoters are apparent. In this respect, the mcb promoter seems of particular interest, since it is nearly identical with the promoter of bolA, a gene under σ ^S control (2, 31). Yet, the expression of mcbA::lacZ fusion is not affected (112) or is even somewhat stimulated (19) in an rpoS mutant, and in vitro transcription of mcb is initiated by σ ⁷⁰ (σ ^S was not tested) (19). This finding suggests that also sequences outside the core promoter region are involved in discriminating between σ ^S- and σ ⁷⁰-dependent promoters. Segments of curved DNA were found in the promoter regions of bolA (p1), katE, and xthA (53), and sequences with the characteristics of bent DNA are also present upstream of the osmY and glgS promoters (77, 111, 215). Since the mcb promoter region does not exhibit bending, it was suggested that DNA curvature may contribute to recognition by σ ^S (53). However, since bent DNA regions are also found in the upstream regions of many σ ⁷⁰-dependent promoters (155), and the rate of transcriptional initiation can be stimulated by inserting segments of bent DNA upstream of a promoter (62), DNA curvature may contribute to overall promoter strength rather than to discrimination between different sigma factors for transcriptional initiation.

Several stationary-phase-regulated genes are transcribed from single transcriptional start sites. These include the σ ^S-controlled genes fic (197), katE (203), osmY (111, 215), poxB (28), treA (162), and csiE (Marschall and Hengge-Aronis, in press) and the σ ^S-independent mcb operon (31). Sometimes these promoters exhibit dual control. The activities of the promoters of osmY (79, 207, 216) and treA (71, 78, 162) are stimulated by growth phase or starvation signals as well as by increased osmolarity. Transcription of these genes might be directly initiated by σ ^S, and this regulatory pattern probably reflects an increase in cellular σ ^S content under both conditions (114).

Other growth-phase-controlled genes have more than one promoter. Either all of them contribute to stationary-phase activation, as in the case of glgS (77), or only one of the promoters is stationary phase activated, and this promoter may be σ ^S-controlled as was found for bolA (2, 19, 112), cfa (205), osmB (78, 98), proP (134), and wrbA (214). The other promoters may be subject to regulation by different signals and thus establish a connection to other regulatory circuits. Three and four transcriptional start sites have been found for cstA (173) and glgC (165), respectively. For glgS, four start sites have been mapped, with the second being σ ^S regulated and the others being cAMP-CRP dependent (77). Both for cstA and glgS, these multiple start sites are approximately one helix turn apart from each other, suggesting some kind of "fuzzy" multiple binding site for the RNA polymerase holoenzyme.

As outlined above, other regulatory factors such as CRP, Lrp, IHF, H-NS, and OmpR are involved in the regulation of stationary-phase-inducible genes. Unfortunately, little information about binding sites for these factors is available. Putative CRP boxes are apparent upstream of the regions containing the multiple transcriptional start sites of cstA (173) and glgS (77). A 5'-deletion analysis of the osmY promoter region suggested the presence of a negative regulatory element between 235 and 87 bp upstream of the transcriptional start site (215). This observation is consistent with the location of a putative binding site for Lrp, which negatively controls the expression of osmY (111). Direct binding of purified Lrp to a DNA segment carrying this region has been shown by gel retardation experiments (M. Barth and R. Hengge-Aronis, unpublished results). Binding sites for IHF and OxyR are present upstream of the single promoter of the dps gene. While IHF is required for σ ^S-dependent stationary-phase induction, OxyR and σ ⁷⁰ are involved in oxidative stress regulation of dps (8). Identical transcriptional start sites are used under both conditions (8, 124).

While most stationary-phase-activated promoter regions analyzed so far seem to be of considerable complexity, there are exceptions to this rule, such as the poxB promoter region. poxB is positively controlled by σ ^S and cAMP-CRP, and transcription starts at a single site. Surprisingly, no CRP box was found, suggesting indirect control by cAMP-CRP (28).

In summary, most stationary-phase-inducible promoter regions seem to contain multiple binding sites for transcriptional factors, be it RNA polymerase containing various sigma subunits, "classical" prokaryotic accessory transcription factors like CRP or OmpR, or abundant histone-like proteins with regulatory as well as chromosome-organizing functions such as Lrp, H-NS, and IHF. This complex architecture is reminiscent of the modular structure of eukaryotic promoter regions (52).

CONCLUSIONS AND PERSPECTIVES

Enteric bacteria like E. coli or S. typhimurium are able to survive prolonged periods of nutrient starvation because of complex physiological and morphological alterations. The identification of σ ^S as a regulator of central importance for these processes (113) has greatly stimulated interest and further research on the molecular details of gene regulation during entry into stationary phase. The finding that other global regulators such as cAMP-CRP, Lrp, IHF, and H-NS are also involved indicates that we are dealing with a complex regulatory network that accounts for intricate fine regulation of the many genes involved. Within this network, regulatory cascades and connections to other regulatory circuits are now becoming apparent. Until recently, the molecular signals that trigger the stationary-phase response and the induction of σ ^S in particular have remained mysterious. Evidence is now accumulating that ppGpp (63), UDP-glucose (20), and a homoserine lactone (85) may be such signals, although the molecular mechanisms of signal transduction have yet to be elucidated.

The analysis of stationary-phase gene expression also holds promise for a more general understanding of fundamental processes in prokaryotic gene regulation. The solution of the present paradox that σ ^S and σ ⁷⁰ are structurally closely related and in vitro recognize the same promoters, but differentially control gene expression in vivo, will shed new light on the role of sigma and other transcription factors in transcriptional initiation. Furthermore, rpoS and rpoH are subject to similar translational control mechanisms exerted by mRNA structures in the translational initiation regions that may be resolved in response to certain environmental signals (114). Further elucidation of these processes, as well as of the mechanism of σ ^S turnover control, will also allow a better understanding of posttranscriptional control mechanisms in general.

On the other hand, there are medical and biotechnological aspects of stationary-phase gene regulation. σ ^S was shown to be required for the expression of the Salmonella spv virulence genes (54, 108, 142), and a wider role for σ ^S in pathogenicity seems likely. Starvation-activated promoters may be used for efficient production of heterologous proteins in bacteria and in bioremediation systems (130). It appears that the study of stationary-phase gene regulation has definitely come of age and will provide us with some exciting years to come.

ACKNOWLEDGMENTS

I thank many colleagues for communicating results prior to publication. Work by me and my coworkers mentioned in this review was performed in the laboratories of Winfried Boos, whose continuous interest is gratefully acknowledged, and was supported by the Deutsche Forschungsgemeinschaft (SFB156).