Chapter 106

Morphological and Physiological Changes during Stationary Phase

GJALT W. HUISMAN, DEBORAH A. SIEGELE, MARÍA M. ZAMBRANO, and ROBERTO KOLTER

INTRODUCTION

Upon depletion of essential nutrients from the medium, the growth rate of a bacterial culture slows down and evenually reaches zero. At this point the culture has entered stationary phase. The term "stationary phase" should be understood as an operational term indicating that the culture is displaying no net increase in cell number. However, the physiology of cells in a stationary-phase culture is heterogeneous and depends completely on the conditions that resulted in the cessation of growth. Starvation for an essential nutrient, perhaps most often the carbon source, has been used experimentally as a tool to study the physiology of nongrowing cells. Changes resulting from carbon source starvation, however, should not be assumed to hold for other forms of entry into stationary phase. When the term stationary phase is used here, the reader should keep in mind that the observation described may only apply for the particular conditions used in each experiment and that generalizations may not be valid.

The transition from rapid growth with a generation time of less than 20 min to stationary phase is accompanied by adaptation of the culture to the new conditions. During growth, energy-demanding processes convert nutrients primarily to nucleic acids, proteins, and other cell constituents to enable cell growth and cell division. Rapidly growing cells rely on inducible responses to express defense mechanisms against particular stress conditions. In stationary phase, energy generation is usually limited (79), making inducible responses less immediate and less effective. It thus makes sense for organisms such as Escherichia coli and Salmonella typhimurium (Salmonella enterica serovar Typhimurium) to develop increased resistances against many potential environmental assaults at the onset of starvation. Such a program is advantageous to the cell because it can cope with environmental stresses instantaneously without requiring new protein synthesis. In this way, precious energy equivalents are preserved and chances for survival are not impaired.

A possible consequence of the development of a highly resistant state is that resumption of growth may take more time since the protective measures have to be dismantled. Within a population, cells that have not fully expressed their increased resistances might respond more quickly to conditions supporting growth and could have an advantage in resuming growth. The identification of mutants that take over stationary-phase cultures supports this possibility (178). One potential result is that a culture that is repeatedly switched from starvation to growth conditions loses its clonal identity and evolves into a mixture of cell genotypes. Some conditions appear to select mutants that can take over the population (178), while other conditions select for balanced mixtures of cells that complement each other (135). The observations that starved cultures can quickly lose their clonality should undoubtedly warn workers in this field to exercise extreme care when storing bacterial strains in agar stabs and plates.

This chapter describes the changes that E. coli and S. typhimurium undergo when starved. We summarize the changes (schematically represented in Fig. 1) that occur in the different subcellular compartments and in the genome and describe genes that have been found to be essential to survive in stationary phase. Several additional reviews describing bacteria in stationary phase and gene expression in slowly growing cells have appeared recently (65, 72, 81, 101, 102, 145).

Fig. 1Summary of phenotypic changes in stationary-phase E. coli and S. typhimurium.

Source:

PHENOTYPIC CHANGES DURING STATIONARY PHASE

The Envelope: Inner Membrane

The inner membrane of gram-negative cells forms the most important barrier between the cytoplasm and the extracellular medium. This membrane is also a key component of the transport mechanisms that allow the selective flow of molecules into and out of the cell and is the main facilitator in maintaining the proton motive force. The membrane protects the cell from extracellular chemical and physical assaults and is the major scaffold on which the cell shape-determining peptidoglycan molecule is synthesized. It is therefore not surprising that this cellular component undergoes major changes when cells enter stationary phase.

The fatty acid composition of the membrane changes dramatically in response to starvation. The membranes of both E. coli and S. typhimurium contain monounsaturated fatty acids, predominantly palmitoleic acid (C_16:1) and oleic acid (C_18:1) (chapter 37, this book). The relative abundance of these unsaturated fatty acids declines more than 10-fold during starvation, from almost 50% to less than 5% (42). Until recently, no polyunsaturated fatty acids had been observed in E. coli or S. typhimurium; they had only been analyzed during exponential phase or early stationary phase. A recent study of fatty acid content after 72 h of carbon starvation indicated the presence of linoleic acid (18:2), a polyunsaturated fatty acid (126); however, this report could not be confirmed (26). The role of unsaturated fatty acids in the membrane during stationary phase remains unknown, although an unsaturated fatty acid auxotrophic mutant was shown to develop extreme sensitivity to low osmolarity exclusively when starved (140).

The reduction of monounsaturated fatty acids observed at the onset of starvation is accompanied by a corresponding increase in the levels of cyclopropyl fatty acid derivatives (157). This increase of cyclopropyl derivatives from 10% during growth to nearly 50% of total fatty acid is accounted for by C₁₇ and C₁₉ cyclopropane fatty acids (174). The conversion of unsaturated fatty acids to their cyclopropyl derivatives is catalyzed by cyclopropyl fatty acid synthase, a membrane-bound enzyme encoded by the cfa gene (157, 168). Synthesis of cyclopropyl fatty acids appears to be oxygen sensitive because increased aeration leads to a decreased cyclopropyl fatty acid formation (116). Mutants lacking cfa have been constructed and subjected to many different extreme environments and treatments, such as prolonged incubation in stationary phase or exposure to drying, detergents, heavy metals, low pH, or high salt concentrations (157). Such mutants showed no clear phenotype except for being less resistant to repeated cycles of freezing and thawing (63).

The phospholipid head group composition has also been found to change in response to starvation. The levels of phosphatidylglycerol and phosphatidylserine decrease in stationary phase, whereas cardiolipin levels increase (1, 104). A gene important for cardiolipin synthesis (cls) has been cloned from E. coli and analyzed (113). As is the case for mutants lacking cfa, no growth phenotypes have been found for mutants lacking cls, however, such strains still had low levels of cardiolipin, probably synthesized by phosphatidylserine synthase (encoded by pss). Attempts to introduce a cls mutation into a pss-null strain have been unsuccesful, suggesting that cardiolipin is essential for suvival (113).

While no firm conclusions can be drawn regarding the physiological consequences of these growth phase-dependent changes in membrane composition, fluorescence polarization studies suggest that membranes from stationary-phase cells are in a highly ordered state that may render them more damage resistant (149, 150). It has been suggested that this highly ordered state may result not only from changes in lipid composition of the membrane, but also from starvation-induced increases in the levels of polyamines (151) and allysine cross-links among membrane proteins (106).

The Envelope: Outer Membrane

The composition of the outer membrane also changes at the onset of stationary phase; the amount of exposed lipopolysaccharide on the outer surface of the outer membrane increases (75), thus increasing the charge at the surface of the cell and thereby possibly increasing its resistance to membrane-damaging agents. Additionally, more compounds containing sulfhydryl substituents are directed to the outer membrane and eventually excreted (93, 94, 120, 147). Whether this phenomenon serves as a protective mechanism remains to be elucidated.

There is an overall decrease in the amount of proteins in the outer membrane (6); however, synthesis of some outer membrane-associated proteins is induced during stationary phase. Important new outer membrane proteins that are synthesized include flagellin (10), the curli subunit CsgA (118), and the Pap-related pili encoded by htrE and ecpD (127). Flagellin synthesis peaks at the onset of stationary phase, but then declines rapidly (10). The synthesis of attachment organelles and increased motility can be seen both as important strategies to search for alternative nutrient sources and to escape conditions that do not support growth. Chemotaxis is by far much more energy-consuming than surface attachment. The cells’ diminished proton motive force decreases their ability to generate ATP (79), which may be the reason why flagellar synthesis is only transiently turned on at the onset of stationary phase in batch cultures (10).

In contrast, surface attachment organelles such as curli pili can provide the cell, at a lower cost in energy, with a physically secure location in which to survive long periods of nutrient scarcity. Attachment to host cells, in conjunction with toxin production or cell entry, can also be a way to increase the local nutrient concentration. However, conditions within the host cannot be directly compared to those in batch cultures. It is possible that to successfully contact and invade eukaryotic cells, organisms such as S. typhimurium alter their surface composition by alternating the expression of genes, some of which are expressed during growth in culture (82, 98) while others are induced during stationary phase (46, 64, 114).

At the periplasmic side of the outer membrane, the amount of lipoprotein bound to the peptidoglycan layer increases during the transition from logarithmic growth to stationary phase (170). Recently, a gene encoding a new lipoprotein (nlp) was described (74, 87) which, interestingly, is organized in an operon with rpoS, the gene encoding the stationary phase-specific sigma factor (156). The increased connection between wall and outer membrane may reflect the cell’s need to protect its outer structure under harsh environments. In doing so it may sacrifice some flexibility that could be useful during rapid growth, but gain in the overall stability of the envelope.

The Envelope: Peptidoglycan and Periplasm

The peptidoglycan of E. coli and S. typhimurium consists of a single molecule in which an oligosaccharide is cross-linked by specific oligopeptides (chapter 6). During growth the peptidoglycan wall consists of two to three layers (representing 0.75% of the cellular weight) which are increased to four to five layers during stationary phase (90, 105). In addition to the increase in thickness, composition also changes during stationary phase. The length of the glycan strands decreases and the amounts and type of peptide constituents change. Free tri- and tetrapeptides decrease with a corresponding increase in peptide-peptide and peptide-lipoprotein cross-linking (122, 170).

The changes in peptidoglycan structure are not accompanied by a net increase in wall components, but are due to a turnover process in which the penicillin-binding proteins (PBPs) play an important role (17). During starvation the translocation of cell wall precursors across the membrane as isoprenoid lipid-linked intermediates slows down (50). The activities of some peptidoglycan biosynthetic enzymes and the levels of PBPs change (105, 163). The specific activities of the d-glutamic acid and d-alanyl-d-alanine-adding enzymes vary little with growth phase, whereas the specific activities of UDP-N-acetylglucosamine transferase and UDP-N-acetylglucosamine-enolpyruvate reductase and the diaminopimelic acid-adding enzymes decrease 20 to 50% in late stationary phase (105). PBP 6 is thought to be a d-alanine carboxypeptidase I that functions in the stabilization of peptidoglycan specifically during stationary phase (20, 163).

The expression of proteins involved in peptidoglycan alterations at the onset of stationary phase is regulated by different factors, including cyclic AMP (cAMP)-CAP, ppGpp, and the stationary phase-specific sigma factor σ ^S (4, 18, 28, 40, 85). For example, expression of PBP 2 is regulated by ppGpp (165), but its link to the septation apparatus appears to be mediated by a factor whose expression is cAMP-CAP dependent (28). Expression of both PBP 3 and PBP 6 is σ ^S regulated. When cells enter stationary phase, levels of PBP 3 decrease, whereas PBP 6 levels increase 2- to 10-fold (40). The increase in PBP 6 is particularly relevant since it has been proposed to play a role in stabilization of stationary-phase peptidoglycan analogous to the sporulation-specific PBP 5a from Bacillus subtilis (163).

Other stationary phase-dependent changes in the periplasm include the accumulation of membrane-derived oligosaccharides (83, 138) and trehalose (164), both thought to act as osmoprotectants (15). Expression of the genes encoding enzymes involved in trehalose metabolism is σ ^S dependent (67). Trehalose is synthesized by the otsAB gene products and degraded by trehalase encoded by treA (19, 58, 78). Other periplasmic proteins that are expressed as part of the σ ^S regulon are acidic phosphatase, AppA (31, 160), and the osmY gene product, whose function remains unknown (175, 176).

Extracellular Environment

When E. coli enters stationary phase, a number of metabolites are excreted into the medium. Acetate is the major metabolite that is generally found at the end of exponential phase when cells are grown on glucose and which is rapidly used at the onset of stationary phase (43). Other metabolites that have been found to accumulate when cells cease growing are reduced glutathione and other thiol compounds (93, 94, 120). Recently, the transient accumulation of 1,5-anhydroglucitol early in stationary phase was observed (144). The metabolic pathways for biosynthesis and degradation of this molecule remain unknown, but 1,5-anhydroglucitol has been found to be produced by many different organisms, including humans, suggesting it may be a ubiquitous signaling molecule (144). Another interesting finding is that glucose-grown cya or crp mutants excrete l-glutamate into the medium upon starvation (91), demonstrating the intricacies of changes in metabolic pathways at the onset of stationary phase.

Several lines of evidence suggest that an external metabolite may trigger the expression of a starvation response by activation of the σ ^S regulon. Different reports have suggested that weak acids such as acetate and benzoate induce expression of rpoS when growing cells are resuspended in medium lacking a carbon source (109, 141). However, a signaling role for acetate or benzoate in cells entering stationary phase by nutrient depletion has never been shown. In contrast, homoserine lactone has recently been shown to act as an inducer of the σ ^S regulon (73). This may be analogous to the inducer role of extracellular acylated homoserine lactone derivatives in a number of cell density-dependent phenomena such as bioluminescence in Vibrio fischeri and Vibrio harveyi, Ti plasmid transfer in Agrobacterium tumefaciens, elastase synthesis in Pseudomonas aeruginosa, and antibiotic and exoenzyme biosynthesis in Erwinia carotovora (for a recent review, see reference 51). Unmodified homoserine lactone in E. coli may be an intracellular switch to signal its nutritional status, or a modified form may be excreted to signal starvation to the population. However, no excreted form of homoserine lactone has been characterized from E. coli supernatants yet.

The Cytoplasm: DNA/Nucleoid

During growth in rich medium, E. coli has on average more than four chromosomes per cell (111). During the final cell divisions, new rounds of replication are no longer initiated, and at the onset of stationary phase the cells are typically found to have a single chromosome (100). The synthesis of several DNA-binding proteins, including H-NS (or H1a), IHF, and Dps, which regulate gene expression and are involved in compacting or protecting DNA against physical or chemical attack, has been shown to be growth phase regulated (7, 38, 39).

H-NS is a DNA-binding protein that was discovered as a regulator for the expression of several genes (35, 68). Mutants lacking H-NS have extremely pleiotropic phenotypes (69). H-NS expression is autoregulated during logarithmic growth, but this repression ceases when cultures enter stationary phase and σ ^s-independent transcription of hns is induced 10-fold (38, 45, 161). H-NS compacts DNA both in vivo and in vitro (162). At the lacUV5 promoter, H-NS stimulates transcription initiation directly, decreasing the rate of open complex formation (152).

There appears to be a connection between H-NS DNA binding and the differential promoter recognition by RNA polymerase with σ ⁷⁰ and σ ^S. At subsaturating concentrations, H-NS has specificity for curved DNA (119, 158, 171), and some σ ^S-dependent promoters have intrinsic curvature (44). There are at least two examples of σ ^S-dependent promoters, P_csgA (12, 117) and P_mcc (108), whose σ ^S dependence can be relieved by null mutations in hns. This suggests that H-NS may keep these promoters in a conformation that prevents σ ⁷⁰-RNA polymerase from recognizing them. This may explain why, when using linear templates in in vitro transcription assays, promoter discrimination by Eσ ⁷⁰ and Eσ ^S has not been clearly observed (156). The proU promoter may be another example of the interplay between H-NS and σ ^S since H-NS regulates the level of P_proU supercoiling (68) and proU expression is controlled by σ ^S in stationary phase (99).

Integration host factor (IHF) was originally identified because of its role in λ integrative recombination, and was later shown to be a pleiotropic regulator of gene expression (49). In contrast to H-NS, IHF binds to specific sites in the E. coli genome and induces bending of the DNA (60). In contrast to this low number of specific IHF-binding sites, Ditto et al. (39) have reported that the IHF copy number during growth is around 10,000 to 15,000, which increases six times when cells enter stationary phase. It may be that IHF, like H-NS, occupies specific sites at low concentrations but loses specificity at higher concentrations. Several reports have indicated that IHF can compensate for the absence of another highly expressed DNA-binding protein, HU (39). HU is involved in multiple cellular processes (136), and its expression with respect to growth phase regulation has not been reported.

Dps was discovered as a protein synthesized to high levels after E. coli had been in stationary phase for over 3 days (7). Dps appears to bind DNA nonspecifically (possibly as a dodecamer) and may prevent oxidative damage caused by hydrogen peroxide (155). However, in addition to this protective role, a mutation in dps has a pleiotropic effect on the expression of several proteins as judged by two-dimensional gel electrophoresis. dps is expressed in stationary phase as part of the σ ^S regulon, but its expression also depends on IHF (8). Generally dps is not expressed during growth but is induced after hydrogen peroxide treatment in an OxyR-dependent manner (8).

The Cytoplasm: Protein Synthesis

At the transition into stationary phase, changes occur in the transcriptional and translational processes that have a direct impact on the physiology of the cell. Besides the appearance of the new sigma factor, σ ^S, the core of RNA polymerase is modified, apparently by phosphorylation (121). Reduction in the number of ribosomes may be a consequence of their being scavenged for nutrients, and starvation also results in temporarily empty acceptor sites at the ribosomes, which initiates the relA-dependent synthesis of ppGpp (chapter 92). In addition, carbon starvation causes a relA-independent increase in ppGpp, probably due to the action of the spoT gene product. Increased ppGpp levels induce the stringent response and affect many cellular processes including increasing σ ^S levels (55). Perhaps to preserve ribosomes, E. coli synthesizes a ribosome modulating factor, encoded by rmf, that is involved in the dimerization of ribosomes (167). Absence of the 55-amino-acid Rmf protein reduces the viability of the strain in stationary phase. Expression of rmf increases 15-fold upon entry into stationary phase but is rpoS independent (172).

As the rate of protein synthesis decreases during starvation, key signal molecules may be synthesized as a natural consequence of imbalances in the pools of free amino acids and the transient accumulation of intermediates in amino acid biosynthesis (73). Several tRNA synthetases have been shown to bind such intermediates and generate cyclic compounds (76). While these noncognate reactions have been studied from the perspective of proofreading or editing during the charging reaction, it is possible that they serve a physiological role in generating starvation-induced signals. For example, methionine-tRNA synthetase binds homocysteine, adenylates it, and releases AMP and homocysteine thiolactone (76) (Fig. 2). Similarly, isoleucine-, valine-, and lysine-tRNA synthetases bind homoserine (an intermediate in methionine, threonine, and isoleucine biosynthesis [chapters 27, 32, and 33]) and generate homoserine lactone in vitro (76) (Fig. 2). This latter observation is particularly significant because acylated homoserine lactone derivatives are ubiquitous high-cell-density signaling molecules in bacteria (51), and homoserine lactone itself has been shown to be involved in induction of rpoS expression (73).

Fig. 2Synthesis of homocysteine thiolactone and homoserine lactone is catalyzed by tRNA synthetases.

Source:

In addition to amino acid levels, tRNA concentrations also fluctuate with growth phase. E. coli has five different tRNA genes that specify leucine codons. The leuX tRNA recognizes one of the minor leu codons, and its expression increases with decreasing growth rate (137). Strains that harbor an amber suppressor allele of leuX (supP) have a pleiotropic phenotype in stationary phase. For example, induction of microcin B17 production is greatly decreased (E. I. Vivas and R. Kolter, unpublished data), and such strains grow well within the mouse intestine but cannot be maintained there (112). Interestingly, such supP mutant strains can colonize the intestine. As long as they are increasing in numbers within the animal they behave as wild-type strains. But once bacterial counts no longer increase, the supP strain is lost. These effects are specific for supP; amber suppressor alleles of other tRNAs had no effect in this system. It may be that the frequency and position of rare codons is a general regulatory mechanism for achieving growth phase-dependent control of gene expression (23).

The Cytoplasm: Metabolic Circuits

At the onset of stationary phase the bacterial cell redirects its physiology in order to survive. The proton motive force is the main engine in growing cells for energy generation. At the onset of stationary phase, however, the proton motive force declines slightly and energy generation proceeds at a slower rate (79). Redirecting metabolism to scavenge any potential nutrient from the medium or within the cell may increase survival.

Cells growing aerobically express primarily one cytochrome oxidase, encoded by the cyo locus (14). Continuation of the efficient utilization of oxygen as terminal electron acceptor during stationary phase is governed by the expression of two alternative cytochrome oxidases, encoded by cyd (61) and cyx or appAB (32). Transcription of cyd depends on arcA/arcB (56), and transcription of cyx depends on rpoS (32). The significance of having three different terminal oxidation systems is underscored by the finding that a cyo cyd double mutant is less sensitive to oxygen than a mutant that lacks all three oxidases (32). Apparently, evolutionary pressure has resulted in E. coli maintaining three independently regulated systems for survival under aerobic conditions. While the subtle physiological consequences of expressing different cytochrome oxidases during starvation remain unclear, it is interesting that a mutant defective in the proper assembly of cytochrome d is unable to reinitiate growth under aerobic conditions at 37°C (146).

As discussed above, E. coli excretes metabolites during growth that are taken up again and used at the onset of stationary phase. When E. coli grows on glucose it excretes large amounts of acetate (up to 85 mM) (96). This apparent waste of nutrients is a result of an inefficient tricarboxylic acid (TCA) cycle in which the α-ketoglutarate dehydrogenase complex is inactive (9). Consequently, the TCA cycle is not a cyclic process under those conditions and instead proceeds as two pathways: one from oxaloacetate to succinate and the other from oxaloacetate and acetyl coenzyme A (acetyl CoA) to α-ketoglutarate (9). Apparently, E. coli does not need all the energy that can be obtained from glucose and saves some of it in the growth medium in the form of acetate until later stages where carbon becomes limiting. Upon entrance into stationary phase, E. coli has to conserve energy and carbon sources, and those reactions that waste carbon building blocks are redirected (89). For instance, the complete TCA cycle runs, and to prevent the complete loss of carbon in the form of CO₂ during the conversion of isocitrate via α-ketoglutarate to succinyl CoA the glyoxylate shunt is activated (chapter 21). In this pathway isocitrate is converted to succinate and glyoxylate rather than being converted to α-ketoglutarate by isocitrate dehydrogenase. Glyoxylate then reenters the TCA cycle by condensing with acetyl CoA to form malate. This anapleurotic route saves carbon and thus is expected to be important during stationary phase. The fact that addition of acetate to stationary-phase cultures leads to loss of isocitrate dehydrogenase activity supports this idea (53, 54). Regulation of isocitrate dehydrogenase activity by phosphorylation/dephosphorylation is performed by the aceK gene product (88).

Studies on expression of katE, the gene encoding catalase HPII, suggested that TCA cycle intermediates may play a pivotal role in sensing the growth stage of E. coli (95). This gene is under the control of σ ^s, and its expression increases 10-fold at the onset of stationary phase. However, a similar increase is found when E. coli is grown in minimal medium on TCA cycle intermediates. Whether this increase reflects a relation between growth rate, which is lower on these carbon sources, and σ ^s expression, or is a consequence of the presence of TCA cycle intermediates, remains to be elucidated.

Recently there have been numerous reports of genes whose products are involved in metabolic pathways in stationary-phase cells. In connection to efficient metabolism of acetate, the identification of the gene that encodes pyruvate-formate lyase (pfl) (128) and the gene encoding pyruvate oxidase (pox) (22) as being expressed preferentially in stationary phase underscores the importance of central metabolic pathways in stationary-phase cells.

S-Adenosylmethionine (SAM) is perhaps the most important methyl donor in the cell (chapter 33), and its synthesis is of vital importance for the cell. The enzymes involved in SAM biosynthesis represent an example of redundant metabolic functions. Two adjacent genes, metK and metX, encode SAM synthases (139). MetK is preferentially active during growth in rich medium, and MetX is expressed during growth in poor medium and during stationary phase. Although mutants in any one of the two genes are viable, a double mutant cannot be made without the appearance of revertants.

A well-known phenotype of stationary-phase E. coli is its ability to accumulate glycogen as a carbon reserve (30). Glycogen is synthesized from glucose in a three-step pathway involving glucose kinase, glucose-6-phosphate ADP transferase, and glycogen synthase. Branching enzymes can introduce 1,6-glycosidic bonds that result in a highly branched polyglucose molecule. Glycogen synthesis is regulated at several different levels by (i) allosteric interactions of metabolites with glucose-6-phosphate UDP transferase (123, 124); (ii) genetic regulation of the glgCAB operon by cAMP, CRP, and ppGpp (133); (iii) regulation of expression of the putative priming protein GlgS by rpoS (66); and (iv) the product of the csr gene (132). Expression of the first gene in the gluconeogenic pathway, pckA, is also growth phase regulated and regulated by catabolite repression (59).

A second important storage polymer in stationary-phase E. coli is polyphosphate. Polyphosphate kinase catalyzes the elongation reaction in which the γ-phosphate group from ATP is added to an existing polyphosphate molecule (2). A polyphosphatase is involved its degradation (3). Strains deficient in polyphosphate accumulation are impaired in survival during prolonged periods of starvation (27). Polyphosphate is proposed to play an essential role in the uptake of DNA by E. coli in the transformation process as part of a complex with poly(3-hydroxybutyrate) (71, 130). Recently, it has been shown that the polyphosphate/poly(3-hydroxybutyrate) complex functions as a voltage-gated Ca²⁺ channel reminiscent of the Ca²⁺ channels found in eukaryotes (R. N. Reusch and L. Bramble, personal communication). The σ ^S-dependent appearance of this complex in stationary phase also suggests a role for this structure under nongrowing conditions (G. W. Huisman and R. Kolter, unpublished data).

Upon exhaustion of a carbon source, E. coli expresses several systems dedicated to the uptake of sugars such as galactose, ribose, and maltodextrins (5, 34). The induction of such operons, which is due to the accumulation of endogenous inducer and cAMP, improves the sugar-scavenging ability of a nearly starved cell (33). Chemostat cultures operating at low steady-state levels of glucose have been shown to accumulate galactose to levels as high as 0.2 mM, which acts as endogenous inducer of the mgl system. The facts that LamB is involved in the uptake of several sugars (34) and that Mgl and LamB both constitute high-affinity glucose uptake systems illustrate how E. coli initially responds to diminishing availability of a carbon source. Possibly, the transition from exponential growth to stationary phase is a "hunger" phase in which bacteria turn on high-affinity transport systems for many nutrient sources (33).

When E. coli is unable to synthesize amino acids efficiently de novo due to starvation, it degrades endogenous proteins and peptides (129). Degradation of proteins during carbon starvation is accomplished in part by the Clp protease (29). Clp consists of two subunits; expression of the ClpP subunit is growth phase independent, whereas ClpA is expressed when the growth rate slows down (80). This catabolic pathway to convert existing proteins to amino acids and then to new proteins requires energy for which the NADH dehydrogenase I, encoded by the nuo operon, is important (11). Mutations in nuo also result in a competitive disadvantage in mixed cultures during stationary phase (177) and result in poor growth on some amino acids (125). NADH dehydrogenase I activity confers on the cell a greater efficiency in converting the energy potential from reduced NADH into proton motive force (52). The finding that nuo mutants have severe phenotypes during stationary phase, but virtually no phenotype during rapid growth, indicates that efficient energy generation is much more crucial for starved cells than for rapidly growing cells (177).

Because the degradation of proteins to synthesize new proteins is a burden on the energy capacity of the cell, this system needs to be well controlled. The finding that the rpoH regulon, which includes DnaK, is involved in stationary-phase survival suggests that recognition of misfolded and unfolded proteins is important to preserve cell integrity (77). Additionally, DnaK is essential in carbon starvation survival and plays a role in the expression of other starvation-specific proteins (153). During exponential growth, DnaK functions together with DnaJ and GrpE as a molecular chaperone (57). A stationary phase-specific DnaJ homolog (CbpA) has recently been discovered (173). Expression of CbpA is σ ^S dependent and may be important in protein turnover in stationary-phase cells. Additionally, stationary-phase E. coli express an enzyme that recognizes and repairs proteins in which isoaspartyl residues have been introduced (92). The corresponding gene, pcm, has been cloned and characterized and, interestingly, was found to be located just upstream of rpoS, possibly in the same operon (74).

To ensure that E. coli has adequate reserve material to resume fatty acid biosynthesis as quickly as possible, it preserves the required cofactor. Acyl carrier protein, which is bound to all fatty acids during their biosynthesis, is protected from degradation during stationary phase by the formation of mixed disulfides with glutathione (131). Similarly, the fatty acid catabolic cofactor CoA is joined to glutathione (93).

The Cytoplasm: Regulatory Changes

The regulatory processes on which all of the phenotypic changes described above rely can be divided into two categories, rpoS dependent and rpoS independent. The rpoS regulon is the subject of the chapter by Hengge-Aronis (chapter 93). The second category involves other pathways through which genes are specifically expressed in stationary phase. A number of stationary phase-induced genes have been found to be expressed in a σ ^S-independent manner; several are regulated by cAMP-CRP (142). Examples of these are the glgCAB operon and cstA (66, 143). The increase in cytochrome d oxidase expression as cells approach stationary phase is dependent on arcA/arcB, which appear to sense the rate of flow of electrons through the electron transport chain (56). Some starvation-induced proteins may respond to increases in ppGpp levels in an rpoS-independent fashion. Microcin B17 is an example of a gene whose induction in stationary phase is not dependent on σ ^S, cAMP/CRP, or ppGpp (18, 25; D. Siegele, unpublished data).

Several other regulatory genes that respond to starvation or regulate starvation-inducible genes have been identified. Among these are uspA, which is induced under every stress condition thus far tested (115); mprA, which in high copy represses growth phase induction of the mcb and mcc operons (36, 37); and csr, which encodes a carbon storage regulator (132). Mutations in csr affect not only the biosynthesis of glycogen, but also metabolism of certain carbon sources and cell adhesive properties (132). Analogous to mprA, the rspAB operon was also identified as a locus that when overexpressed represses the expression of a stationary-phase gene, namely, σ ^S (73). The similarity of the encoded proteins with catabolic enzymes suggested the involvement of small molecules in starvation signaling. Such approaches are expected to give further insights into the signals that sense the state of individual cells or bacterial cultures in general.

A number of procedures have recently been used to identify genes that are induced in stationary phase or proteins that are specifically expressed in this growth phase. Analysis of new proteins synthesized at the onset of stationary phase has resulted in a general idea of how many proteins are specifically expressed (62, 86), and through reverse genetics, these proteins have been identified and characterized (7). Random lacZ fusions have been generated in the E. coli chromosome, and those that were induced in stationary phase were selected (cstA, osmY, and others) (62, 169). A different procedure was recently reported by Chuang et al. and consisted of probing the ordered Kohara phage library with RNA isolated from stationary-phase or stress-challenged E. coli (24).

A characteristic of stationary phase-inducible genes is that often multiple factors are involved in regulating their expression and conditions other than starvation induce their expression. For example, aidB (whose function remains unknown) can be induced by alkylating agents in an ada-dependent manner, but it is also expressed independently of ada in stationary phase (166). Using lacZ fusions, Lange et al. (84) observed that Lrp, IHF, and cAMP-CRP all affected the σ ^S-dependent expression of osmY, a gene that can also be induced by high osmolarity, independent of all these effectors. The dps gene, encoding a DNA-binding protein, is transcribed in stationary phase by RNA polymerase with σ ^S in conjunction with IHF and can be expressed upon oxidative challenge with hydrogen peroxide during growth in an OxyR-dependent manner, presumably by RNA polymerase with σ ⁷⁰ (8). The cross-regulation of gene expression in nongrowing E. coli is further demonstrated by the fact that mprA is a high-copy repressor of both σ ^S-dependent (mcc and proU) and σ ^S-independent (mcb) genes (36, 37).

GENOTYPIC CHANGES IN STATIONARY-PHASE E. COLI CULTURES

Dynamics of Stationary-Phase Cultures

Prolonged incubation of batch cultures of E. coli results in the apparent death (as measured by the loss of plating ability) of some of the cells. Surprisingly, under some conditions the initial death rate slows down after a few days and a fraction of the cells remain able to form colonies for months and even years (154, 159). Studies in which surviving cells of old cultures were mixed with fresh stationary-phase cultures led to the discovery of the GASP (growth advantage in stationary phase) phenotype (178). This phenotype results from mutations that allow some cells to grow when the majority of the population is dying and represents an unexpected strategy of microbial survival during prolonged starvation. Thus, starved E. coli cultures are not necessarily static but can undergo population changes as GASP mutants take over the population. Mutations in rpoS that result in reduced σ ^S activity can confer on cells the GASP phenotype. However, mutations in other genes can also result in the same phenotype.

Not only can mutations in different loci confer the GASP phenotype, but successive rounds of selection can take place during prolonged incubation of a single culture (178). Population changes had been known to occur during continuous growth of E. coli cultures (41). This phenomenon, termed periodic selection, is the result of the successive appearance of mutants with a growth advantage over the parent strain during continuous growth in chemostats (13, 103). However, the population takeovers observed in stationary-phase cultures due to GASP mutations occur much more rapidly than the population shifts reported for continually growing cultures.

Bacteria have often been used as experimental systems to study evolution. The large population numbers and short generation times make them well suited for such studies. Much of the work has focused on the adaptation of bacteria to nutrient-limited conditions. Some of the genetic changes that result in a selective advantage under these conditions include gene duplications (148) and transposition events (107, 110). Continuous subculturing of cells selects for changes that increase cell fitness (16, 70). In most experiments, the fraction of mutants increases slowly and they are able to take over only after prolonged growth, involving hundreds to thousands of generations. This contrasts sharply with the short incubation periods and numbers of generations needed for mutants with a GASP phenotype to take over stationary-phase batch cultures (178). Given that bacteria spend most of their existence under conditions of very low nutrient availability, the discovery of the GASP phenotype should have important implications for studies on microbial ecology and evolution.

The dynamic state of stationary-phase cultures was discovered by mixing differentially marked cell populations. But during prolonged incubation of a "pure" culture, GASP mutants with a competitive advantage replace the original population under the strong selective pressure imposed by starvation. Figure 3 depicts what can be assumed to be occurring in stationary-phase cultures that are seeded with a single population type. An advantageous mutation may occur during growth of the population which allows mutants to utilize more efficiently nutrients released by dying cells. As a result, the original population is rapidly replaced by the fitter strain. In stationary phase the number of viable counts in a culture is the sum of the growing mutants and the survivors of the original population. The kinetics of growth and the resulting population takeover will vary depending on the incubation conditions of the culture, but they have been observed in a variety of media and using many different strains, including fresh clinical isolates and many different bacterial species (A. Tormo, unpublished data; M. M. Zambrano, Ph.D. Thesis, Harvard University, Cambridge, Mass., 1993; E. Zinser and R. Kolter, unpublished data). Variability from culture to culture can also be expected depending on the nature of the particular mutation involved. At a particular point during incubation, a culture is a mixed population of cells determined by the amount of population takeover that has occurred. Additional incubation can bring about successive rounds of selection in which cells with a growth advantage over the parental strain can result in new population changeovers.

Fig. 3Population changes in stationary-phase cultures.

Source:

Postselection Mutations

One of the most exciting, as well as controversial, subjects dealing with stationary-phase E. coli has been the study of the origin of mutants under nonlethal selection conditions. The fluctuation analyses of Luria and Delbrück provided much of the impetus for the dominant hypothesis that mutations arise spontaneously and at random during nonselective growth prior to selective plating (97). However, the number and time of appearance of mutants under nonlethal selections did not fit well with the predictions that all mutations arise prior to selection, and several observations of the unexpected appearance of "late-arising" mutations were reported over the years. In 1988, a paper by Cairns et al. brought the question of the origin of mutations in bacteria back to the limelight (21). The observations presented went directly against dogma, suggesting that mutations could also arise among populations of starved cells and, most importantly, that the process appeared to be "adaptive" in that the mutations that arose under those conditions were those that conferred a growth advantage on the cell. Since the publication of that paper, numerous reports have appeared arguing for and against the adaptive nature of these mutations, and many theories have been offered to explain the phenomenon. It is not the intention here to provide a critical review of this literature; this has been done extensively elsewhere (47). Our attempt here is simply to summarize the current state of the field.

The emphasis has recently switched from attempts to develop theories that explain the phenomenon to characterization of the types of mutations arising postselection and the gene products involved in the process. The results from these efforts make it clear that, in contrast to mutations arising during growth, postselection mutagenesis involves the major recombination pathway (RecABC) and often results in simple deletions in homopolymeric runs (48, 134). This indicates that there are underlying mechanistic differences in the origin of mutations in growing versus nongrowing cells.

What about the "adaptive" nature of these mutations? The jury is still out on that question. Suffice it to end this chapter with an editorial note. Most of the work done in this field thus far has made the implicit assumption that the physiological state of a starved bacterium is unchanged whether or not the selective conditions are present. For example, the overall metabolic state of a Lac^– cell is assumed to be the same during starvation in the presence or absence of lactose. As more studies of the physiology of nongrowing bacteria are carried out in the future, it may become evident that even though the presence of the selective condition (e.g., lactose) does not allow growth, it changes the physiology of the cell. Such changes may help explain why postselection mutations appear adaptive.