Verweile doch! Du bist so schön!

– Goethe. Faust

This review is focused on the physiological and evolutionary strategies of the processes occurring during the entry of microbial cells into stationary phase and the subsequent period of stasis. The molecular mechanisms adapting microorganisms from exponential growth to a static state involve activation and complex regulation of the stationary factor Sigma-S, which directs RNA polymerase to the specific promoters. As a result the static cells acquire general resistance (simultaneous tolerances) to different environmental stresses. In parallel with the physiological adaptation to stasis, diverse genetical processes are aimed towards resuming the growth of the static cells. Different types of mutagenesis occur: (i) in cells entering stasis and (ii) in static cells (adaptive mutagenesis). Cessation of growth induces the transient hypermutator state resulting in the accumulation of random mutations in the subpopulation of the static cells. If by chance, one or a few of such mutations lead to resumption of division, the growing cell will return to a normal mechanism of spontaneous mutagenesis.  
Another mechanism for generating genetical variability in stressed cells involves transposons and conjugative plasmids. Stresses can stimulate the excision of some transposons, which, in turn, can generate chromosomal mutations and activate intracellular mechanisms of mutagenesis. Under stress some conjugative plasmids activate genes encoding antirestriction proteins that repress restriction-modification systems of the recipient cells. Moreover, under stress special cellular mechanisms decrease (alleviate) the activity of restriction-modification systems which, in turn, enhance the probability of gene transfer into the stressed cells.
Under stress, the efficiency of inter-species genetical barriers also decreases. This, stimulates inter-species gene transfer and may lead to a burst of incipient speciation in the population of non-growing cells. After resumption of growth the genetical barriers leading to isolation will be restored.
In general, the cessation of growth "switches on", and resumption of growth "switches off", a set of special processes that are responsible for generating bursts of genetical variability in populations of microorganisms.

1. Introduction

A wealth of knowledge including many of the most fundamental concepts of molecular genetics has been obtained using Escherichia coli cells that were grown rapidly under laboratory conditions. But in the natural environment bacteria only seldom encounter conditions that permit exponential growth. Rather, microbial life is characterized by long periods of nutritional deprivation punctuated by very short periods that allow fast growth, a feature that is commonly called the feast-or-famine life style (review, Kolter et al 1993). This is why there is an accelerating interest in global physiological and genetical phenomena elicited by changing environmental conditions. The natural habitats of microorganisms such as soil, sediments, sea- and fresh water are frequently characterized by nutrient limitation. In laboratory conditions this situation much better resembles the starvation of bacteria in stationary phase under culture conditions. In general, the physiology of starving bacteria traditionally can be divided into three stages: (i) entry into stationary phase, (ii) maintenance of viability and, (iii) exit from stationary phase (reviews, Roszak and Colwell 1987; Kolter et al 1993).

The objective of this article is to show that there is an additional stage which must be placed between the second and third stages. It is called "the stage of the generation of genetical diversity in the populations of non-growing (or static, or resting) cells". It seems that in a majority of cases only the successful realization of this stage can provide for the appearance and subsequent selection of beneficial mutation(s) and, as the result, entry into the stage of growth.

2. Entry into the static state: everything is foreseen

Limitation of a single nutrient does not necessarily result in a transition into the stationary phase but usually induces alternative scavenging systems, for example: the cAMP/CAP system, the NTR system, and the PHO regulons (Hengge-Aronis 1993; Kolter et al 1993). Once these systems are induced, growth of the cells can continue. Total starvation for any essential substrate, however, induces entry into stationary phase. This transition is a highly odered process which involves dramatic changes in cellular physiology and morphology. What are the strategies behind these changes?

2.1 Starving cells decrease the activities of anabolic enzymes, but retain energy generating systems
ready for action

Several genes/proteins responding to carbon starvation are themselves involved in reorganizing and modulating catabolic fluxes, while others form an integral part of a defence system directed at avoiding the damaging effects of ongoing respiratory activity. A significant fraction of the energy of maintenance is required to prevent the denaturation and spontaneous ageing of proteins during stasis (Morita 1988; Nystrom and Gustavsson 1998). As a result, the metabolic activity of starving cells decreases to zero, redox potential decreases drastically, energetic status becomes lower and starvation specific proteinases, which degrade proteins that have become unnecessary, are induced. Amino acids released from proteolysis are used for the synthesis of new specific proteins. In starving cells enzymes of anabolic pathways are not found, while enzymes and proteins of energy generating systems remain intact and ready for action when the environmental conditions will favour growth once again (Morita 1988; Matin 1991).

2.2 Starving cells make proteins that increase viability during the period of stasis

At the onset of starvation E. coli undergoes a temporally ordered programme of gene expression involving 40–80 genes. Some hours later this yields cells possessing, in the non growing state, an enhanced resistance to different environmental stresses (general resistance). Two classes of genes are induced upon carbon starvation: the cst genes, requiring cAMP, and the pex genes, not requiring cAMP. The cst genes are not involved in the development of the resistant state and are concerned with escape from starvation, while pex gene induction appears to be associated with general resistance. Many of the latter are induced in response to a variety of starvation conditions (Matin 1991). What is the source of the energy which is required for protein synthesis in the absence of exogenous substrates?

2.3 Starving cells generate energy for maintenance from endogenous carbon sources

It seems that these carbon sources are fatty acids. Under conditions of growth arrest, specific alterations in the membrane lipid–fatty acid composition are required for survival of the cell and, concurrently, the membrane lipids are suggested to serve as endogenous reserves providing carbon/energy for maintenance requirements. It appears that the global regulator FadR is needed for both of these activities to be performed properly and that the FadR regulon is interconnected to the universal stress response of E. coli (DiRusso and Nystrom 1998). Thus, cells entering into the stage of non-growth rearrange themselves to survive and overcome the different stresses to which they may be exposed in the future. What is the mechanism of this rearrangement?

2.4 Small molecules of the starvation alarm

2.4a cAMP: Of the 30 carbon starvation proteins whose induction is important for starvation survival of E. coli, 20 of them (Cst) are not induced in cya or crp deletion mutants at the onset of carbon starvation. Most of the Cst proteins were synthesized in the D cya mutant if exogenous cAMP was added at the onset of starvation. Thus, two-thirds of the carbon starvation proteins require cAMP and its receptor protein for induction; the rest do not. The rest, nearly all Pex proteins, are also induced during nitrogen starvation and during phosphate starvation (Schultz et al 1988; Blum et al 1990).

2.4b ppGpp (guanosine 3¢ ,5¢ -bispyrophosphate): ppGpp is a global regulator of bacterial RNA synthesis. It inhibits transcription initiation from stable RNA promoters, as well as synthesis of (bulk) mRNA. Inhibition of stable RNA synthesis occurs mainly during slow growth of bacteria when cytoplasmic levels of ppGpp are high. In contrast, inhibition of mRNA occurs mainly during fast growth when ppGpp levels are low, and it is associated with a partial inactivation of RNA polymerase (review, Bremer and Ehrenberg 1995). Indeed, the entry of E. coli into glucose starvation is connected with the induction of both the stringent response and the general stress response. The changes in the concentration of ppGpp affects the level of Sigma-S which is the master regulator of the reaction of the cells on the starvation and multiple stresses (§ 3.1). However, during transients with slow continuous changes of the nutrient availability, this concerted reaction of ppGpp and Sigma-S is less apparent, indicating the specific importance of these regulators for the adaptation of the cells to fast changes of environmental parameters (Teich et al 1999).

2.5 Starving cells preserve their DNA by condensation

Upon sensing an impending saturation level of their population density E. coli cells enter into the stationary phase. The chromosome undergoes topological changes consistent with the reduction of gene expression observed in starved cells. After several hours in stationary phase the nucleoid condenses; changes in the negative superhelical density of plasmids (if they are present) become apparent. The nucleoid is subjected to marked compactization. What proteins are involved in this condensation? The major DNA-binding proteins, in the exponential-phase nucleoid Fis, HU and HF-1(Hfq) are replaced by a single stationary-phase protein Dps (see § 2.5b), thereby compacting the nucleoid and ultimately leading to silencing of DNA functions. The transcription apparatus is modified by replacing the major promoter recognition subunit Sigma-70, with Sigma-S (see § 3.1). A stationary-phase protein, Rsd (regulator of Sigma D), with the binding activity of Sigma-70, is involved in the efficient replacement of Sigma-70 and/or the storage of unused Sigma-70. Together, these effects may result in the preferential transcription of stationary-phase specific genes. The translation machinery is modulated in stationary phase, by the formation of translationally incompetent 100S ribosomes. A small stationary-phase protein, RMF (ribosome modulation factor), is involved in the dimerization of 70S ribosome monomers (Ishihama 1999).

2.5a IHF: The DNA-binding protein IHF was found to be required for starvation survival and for the induction of 14 proteins of the glucose starvation stimulon. Many of these proteins have been shown previously to be general responders to diverse stress conditions (Nystrom 1995).

2.5b Dps: This starvation-inducible DNA-binding protein is abundant in starved cells. In vitro Dps forms extremely stable complexes with DNA, without apparent sequence specificity. When complexed with Dps, DNA is rendered DNAase resistant. Mutant cells lacking Dps show dramatic changes in the pattern of proteins synthesized during starvation. The mutants also fail to develop starvation-induced resistance to hydrogen peroxide, an agent that can cause oxidative damage to DNA in vivo (see § 4.4). In general, Dps plays an important role both in gene expression and DNA protection during stationary phase (Almiron et al 1992).

2.5c DNA gyrase and topoisomerase I: These enzymes which are responsible for DNA supercoiling are also involved in the protection of DNA during stress. E. coli cells could be protected against seawater-induced loss of culturability by increasing their DNA-negative supercoiling in response to environmental factors. Inactivation of either one of the subunits A and B of DNA gyrase, or topoisomerase A, which leads to important DNA relaxation, was accompanied in both cases by an increased loss of culturability of conditional mutants after transfer to seawater (Gauthier et al 1992).

2.6 Starving microorganisms "prefer" to enter into
a static state with the largest possible number
of the cells in population

Morphological changes that are brought about by starvation are apparent through both light and electron microscopic examination: the familiar rod shape of growing
E. coli is lost in stationary phase because cells become much smaller and almost spherical as the result of several cell divisions without an increase in cell mass. It seems that such a dramatic size reduction may improve a chances of a population survival by increasing cell number (Lange and Henage-Aronis 1991).

2.7 Starving cells "prefer" to be aggregated and fixed on a surface

Changes in the cell envelope that result from starvation reflect the need for protection and insulation from stressful environments. On their surface starved cell are covered with more hydrophobic molecules that favour adhesion and aggregation. Membranes may become less fluid and less permeable as fatty acids compositions changes (Kjelleberg et al 1987). In responses to starvation at low temperatures or to low osmomolarity, E. coli produces curli – a fibronectin-binding filament that may also be involved in aggregation (Olsen et al 1993). It seems that one of the possible benefits of such an aggregated state is that it can increase the chances for gene transfer (see § 9 and 10). After prolonged persistence in the non growing state, the cells could enter into a viable but non culturable VNC (or VBNC) state (Velkov 1996).

2.8 VNC: more dead than alive, or more alive
than dead?

Numerous studies have shown that bacteria which are normally culturable form large populations of nonculturable cells when subjected to adverse environmental conditions, for example, in sterile soil or water. Thus was born the concept of viable but non culturable bacteria (reviews, Roszak and Colwell 1987; Relman and Fa1kow 1992; Velkov 1996; Golovlev 1998). Strains of Vibrio cholerae, E. coli, Salmonella eneteriditis, Shigella sonnei and Legionella pneumophila are in the list of microbes known be capable of entering a state in which they failed to show up on nutrient agar yet took up substrates and signalled in other ways that they are certainly not "completely dead". The combination of enzymatic nucleic acid amplification techniques with 16S rRNA-based molecular phylogeny has enabled the identification of microbial species that can not be cultivated in the laboratory and has revealed the existence of a far greater microbial diversity than has been so far appreciated with culture-dependent methods. PCR-based studies of aqua-tic environmental microbial communities have already reached similar conclusions. PCR-based approach made it plausible that in reality, the diversity of species selected from the environment by selective enrichment is less than 1% of that actually present in many niches. Significantly, evidence suggests far greater microbial diversity even among human pathogens than currently appreciated with culture dependent methods (Relman and Falkow 1992).

VNC cells, although not growing, can demonstrate:
(i) metabolic activity, (ii) induction of the enzyme synthesis and (iii) DNA synthesis. According to cytochemical reactions, metabolic activity of VNC bacteria could be detected over 250–300 days (Gribbonn and Barer 1995). The synthesis of b -galactosidase could be induced in VNC cells of E. coli and S.eneriditis by isopropylthiogalactoside (Nwoguh et al 1995). Increasing incorporation of thymidine into E. coli was observed during two weeks of the persistence of the strain in sea water (Garcia-Lara et al 1993). Although an obvious explanation of these results is that they represent the residual activities of the "irreversibly" dead cells, the alternative has been advanced that these cells are viable but unable to grow until they have been "resucitated"; they are said to be in the VNC state. Indeed, it was shown that viable cells of Micrococcus luteus secrete a factor which promotes the resuscitation and growth of dormant, non-growing cells of the same organism. The resuscitation- promoting factor (Rpf) is a protein which in picomolar concentrations increases the viable cell count of dormant M. luteus cultures at least 100-fold and can also stimulate the growth of viable cells. Rpf also stimulates the growth of several other high G + C Gram-positive organisms, including Mycobacterium avium, Mycobacterium bovis (BCG), Mycobacterium kansasii, Mycobacterium smegmatis, and Mycobacterium tuberculosis (Mukamolova 1998).

Proponents of the VNC concept often lace their papers with warnings that their findings could mean that standard microbiological methods are inadequate to protect public health. For example, strains of E. coli, after exposure to sunlight and entering the VNC state, as well as culturable E. coli, retained pathogenicity, i.e., produced enterotoxin (Pommepuy et al 1996; Bloomfield et al 1998). However, the presence of residual culturable cells confounds such studies, as they would grow in response to the added nutrients and give the illusion of resuscitation. Indeed, a new approach, called the mixed culture recovery (MCR) method, has been developed to determine whether recovery of culturable bacterial cells from a population of largely nonculturable cells is due to resuscitation of the nonculturable cells from a viable but nonculturable state or simply to growth of residual culturable cells. In repeated experiments with strains of E. coli, Klebsiella pneumoniae, Enterococcus faecalis, Enterobacter aerogenes, and Salmonella choleraesuis, only cells of the culturable strain were recovered after application of various resuscitation techniques. These results suggest that the nonculturable cells were dead and that the apparent resuscitation was merely due to the growth of the remaining culturable cells (Bogosian et al 1998). Moreover, a recent critical review of the evidence relating to putative VNC cells demonstrates that most of the reports claiming a return to culturability have failed to exclude the regrowth of a limited number of cells which had never lost culturability (Kell et al 1998). In any case, in dealing with the concept of VNC state the key questions are: is there a check point (biocheimical and/or genetical), after which the non growing cells are "irreversibly" dead? How could this point be recognised? And what do the cells do in order not to cross this last border?

3. How the entry into a static state is regulated

3.1 What is Sigma-S?

The Sigma-S subunit of DNA dependent RNA polymerase is a master regulator in a complex regulatory network that governs the expression of many inducible genes: (i) in stationary phase and (ii) by different environmental stresses in exponential phase. Physiologically Sigma-S plays the role of a general stress sigma factor and acts predominantly as a positive effector, but it has a negative effect on some genes (reviews, Hengge-Aronis 1996, 1999; Loewen et al 1998). A mutation in the gene encoding Sigma-S abolishes transcription of some genes in stationary phase, and also causes superinduction of other stationary phase-induced genes. In the opposite situation, overproduction of Sigma-S markedly reduced stationary phase expression of one, and perhaps more, Sigma-70 dependent promoters. It seems that Sigma-S and Sigma-70 factors compete for a limiting amount of RNA polymerase during stationary phase. Under various stress conditions, the two sigma subunits of RNA polymerase, Sigma-S and Sigma-70, coexist in E. coli cells. In vivo, these sigma factors clearly control different genes. They are structurally and functionally very similar and basically recognise the same promoter sequences. Sigma factor specificity at stress-activated promoters is affected by the interplay of the two RNA polymerase holoenzymes with additional regulatory factors, such as H-NS, Lrp, CRP, IHF, that differentially affect transcription initiation by Sigma-S or Sigma-70 in a promoter-specific manner (Farewell et al 1998; Hengge-Aronis 1999).

The rpoS(katF) gene coding Sigma-S consists of a 1086 base pair (bp) open reading frame which corresponds to a 362 amino acid protein with an apparent size of 38 kDa. Comparison of the rpoS sequence to the sequence of rpoD, which encodes the Sigma-70 subunit of RNA polymerase, revealed a 181 bp region with 65% homology and a 38 bp segment that was 87% homologous. A 62 amino acid region of Sigma-S was found to be 85% homologous to the corresponding sequence of Sigma-70, including a segment implicated in core polymerase binding. Homology was also observed with the heat shock Sigma-32 encoded by htpR (Mulvey and Loewen 1989).

With the usage of a combination of primer extension experiments and 5¢ deletion analysis of the region upstream of rpoS, it was shown that rpoS transcription is mainly driven by a single promoter (rpoSp1) located within the nlpD gene upstream of rpoS (the two relatively weak nlpD promoters contributed to the low level of
rpoS expression during early exponential phase). The expression of both transcriptional and translational rpoS : :  lacZ fusions as well as the level of rpoS mRNA originating at rpoSp1 was strongly reduced in ppGpp-deficient relA spoT mutants. However, experiments with the 5¢ deletion constructs indicated that a lack of ppGpp does affect transcriptional elongation rather than initiation (Lange et al 1995).

3.1a How the level of Sigma-S is regulated; multiple ways to control a single protein: The synthesis and the level of Sigma-S are controlled by mechanisms affecting transcription, translation, proteolysis, and the formation of the active holoenzyme complex (E-Sigma-S).

3.1b Cellular content of Sigma-S is highly variable:
In exponentially growing cells which are not subjected to any particular stress, Sigma-S levels are very low. But the levels are increased up to 20-fold during entry into stationary phase or by a variety of different stresses, namely: (i) depletion of carbon, nitrogen or phosphorous sources, (ii) high osmomolarity, (iii) high and low temperature and (iv) acid pH. Environmental stress conditions (or their alleviation) lead to a rapid increase (or decrease) in cellular Sigma-S level and, consequently, to the activation (or inactivation) of Sigma-S regulated genes (Hengge-Aronis 1996).

3.1c General principles of Sigma-S regulation: The regulation of the cellular content of Sigma-S is extraordinarily complex and is of special interest because it reflects the necessity of sensing and integration of a multitude of very different stress signals. Stress conditions (with the exception of entry into stationary phase) hardly affects rpoS transcription, but various stress signals differently challenge either rpoS translation or Sigma-S turnover, or both. According to the time needed for the response, there are two ways of Sigma-S regulation:
(i) slow – by the modulation of transcription (in parallel with the gradual process of the entry into starvation) and, (ii) rapid – by the "immediate" response to "sudden and unexpected" stresses – by the modulation of translation and/or protein stability which could take place on the background of exponential growth (Hengge-Aronis 1996). Although Sigma-S does not control its own transcription, it is apparently indirectly involved in a negative feedback control that operates on the post-transcriptional level. At least five different signals (cAMP, ppGpp, a cell density signal, an osmotic signal, and a starvation signal) are involved in the control of all these processes that regulate rpoS expression (Lange and Hengge-Aronis 1994).

3.2 Molecular mechanisms modulating
the level of Sigma-S

3.2a Transcriptional control of rpoS: RpoS transcription is inversely correlated with growth rate and is positively controlled by ppGpp and negatively by cAMP-CRP.

(i) cAMP: In rich medium rpoS transcription is stimulated during entry into stationary phase, whereas in minimal media, it is not significantly induced. cAMP-CRP inhibits RpoS transcription (Lange Hengge-Aronis 1994). According to the study of expression of Sigma-S level in glucose limited continuous culture the pattern of induction of RpoS-dependent activities could be separated from those regulated by cAMP, and the induction occurred at extreme glucose limitation. E. coli turns to a protective stationary phase response when nutrient levels fall below approximately 10^–7 M glucose, which is insufficient to saturate scavenger transporters regulated by cAMP plus endoinducers, and this response is optimally expressed at 10^–6 M glucose (Notley and Ferenci 1996).

(ii) ppGpp: The level of rpoS mRNA is strongly reduced in ppGpp deficient relA spoT mutants. Moreover, such mutants are defective in Sigma-S synthesis as cells enter stationary phase in a rich medium. In general, ppGpp can activate Sigma-S synthesis under conditions of nutrient sufficiency as well as during entry into stationary phase (Gentry et al 1993). Experiments with 5¢ deletion constructs indicate that a lack of ppGpp affects transcriptional elongation rather than initiation (Lange et al 1995).

(iii) Homoserine lactone: A variety of Gram-negative bacteria produce membrane permeant, acylated homoserine lactone (HL) pheromones that acts as cell density cues. The function of HL derivatives in many cell density-dependent phenomena suggest that its synthesis is a general signal of starvation (review, Faqua and Greenberg 1998). It was found that HL induces the expression of Sigma-S (Huisman and Kolter 1994).

(iv) UDP-glucose plays an inhibitory role in the post-transcriptional control of the Sigma-S level (Bohringer et al 1995). The mechanisms of the actions of HL and UDP-glucose are to be elucidated.

3.3 Translational control of rpoS

Translation of rpoS mRNA is stimulated during late exponential phase above a certain cell density (when cells are growing in minimal media) by osmotic shift and temperature downshift. rpoS secondary structure plays a role in the regulation of its own translation; the transcriptional initiation region may be base-paired and therefore not accessible for ribosomes. Under inducing conditions, an increased frequency of initiation may result from alterations in this secondary structure. The alterations could be realised by two ways: (i) by the actions of the RNA-binding protein HF-1 (Hfq), of the regulatory protein LeO and/or, (ii) by the action of small DsrA-RNA, all altering the secondary structure of rpoS mRNA. The action of HF-1 and DsrA-RNA is controlled by the protein HN-S and by the small OxyS RNA. In general, translation of rpoS mRNA is controlled by a cascade of interacting factors including HF-1 (Hfq), H-NS, dsrA RNA, LeuO, and OxyS RNA that modulates the stability of a region of secondary structure in the ribosome-binding region of the gene’s mRNA.

3.3a Proteins controlling rpoS translation

(i) HF-1 is a positive regulator: It is an hfq-encoded abundant ribosome-associated RNA-binding small protein of 102 amino acids, which has been known previously only as a host factor for the replication of phage Qb RNA (Kajitani and Ishihama 1991). Also, HF-I binds to both supercoiled DNA and linear DNA and this binding seems to be sequence-nonspecific (Takada et al 1997). HF1 is one of the growth-related proteins. The synthesis rate of HF-I at the exponential-growth phase is higher than at the stationary phase, and it increases concomitantly with the increase in cell growth rate. The intracellular level of HF-I is about 30,000 to 60,000 molecules per cell, the majority being associated with ribosomes as one of the salt wash proteins (Kajitani et al 1994). HF-1 is the essential factor for rpoS mRNA translational control. An hfq null mutant exhibits strongly reduced Sigma-S levels and is deficient for growth phase-related and osmotic induction of Sigma-S and is impaired in rpoS transla-tion (Muffler et al 1996). HF-1 binds to rpoS mRNA. A large deletion from the 5¢ end of the rpoS transcript that still retains 220 bp upstream of the rpoS ATG codon, including a proposed antisense element inhibitory for rpoS translation, was no longer regulated by HF-I (Cunning et al 1998).

(ii) H-NS is a negative regulator: H-NS is an abundant DNA-binding protein that is able to condense DNA in vitro and in vivo in a way that is similar to eukaryotic histones (review, Atlung and Ingmer 1997). Three distinct functional domains were found in H-NS which appear to be responsible for DNA-binding, transcriptional repression and protein-protein interaction (dimerization and/or oligomerization), respectively. Mutations in the C-terminal domain resulted in a loss of its DNA-binding ability, suggesting that this domain is directly involved in its binding to DNA. The N-terminal domain was suggested to be involved in the ability to repress transcription (Ueguchi et al 1996). H-NS is a global inhibitor of expression (global transcriptional silencer). Although H-NS is known to influence the transcription of a number of apparently unlinked genes on the chromosome, it also is involved in maintaining of a low level of Sigma-S in exponentially growing non-stressed cells. H-NS inhibits the expression of Sigma-S itself by a mechanism that acts at the post-transcriptional level. The relief of repression by H-NS plays a role in stationary-phase induction as well as in hyperosmotic induction of rpoS translation (Barth et al 1995). Maybe the activity of HF-1 could also be controlled by H-NS. HF-1 and H-NS can associate with each other in vitro (Kajitani and Ishihama 1991) which raises the possibility that in vivo, H-NS may interfere with the activity of HF-1 by direct protein-protein interaction (Muffler et al 1996c).

(iii) LeuO is a negative regulator: The LysR-like regulator protein LeuO reduces rpoS translation mainly at low temperature by inhibiting expression of small DsrA-RNA. It seems that LeuO represses dsrA and thereby reduces rpoS translation at low temperature. LeuO does not contribute to temperature regulation of dsrA since its own expression is rather low and not temperature dependent. In a mutant deficient for N-HS, however, LeuO is strongly derepressed (Klauck et al 1997).

3.3b Small RNAs controlling rpoS translation

(i) DsrA RNA is a positive regulator: dsrA encodes a small, an 87-nt untranslated RNA which regulates both the transcription of rpoS by overcoming transcriptional silencing by the nucleoid-associated H-NS protein, and translation, by promoting efficient translation of the RpoS mRNA. When overexpressed, DsrA antagonises the H-NS-mediated silencing of numerous promoters. Activities responsible for the overcoming of the H-NS mediated transcriptional silencing and for the promoting of the RpoS mRNA translation can be separated: the first of three stem-loops of DsrA RNA is necessary for RpoS translation but not for anti-H-NS action, while the second stem-loop is essential for anti-silencing and is less critical for RpoS translation (Majdalani et al 1998). Level of DsrA-RNA is high at low temperature which caused by an induction of transcription of the gene and by increased frequency of transcriptional termination at the correct position; incorrectly terminated DsrA-RNA is inactive (Sledjeski et al 1996). In general, DsrA acts as a riboregulator via specific RNA : RNA base pairing interactions at the hns locus to antagonise H-NS translation. Negative regulation of hns by DsrA is achieved by the RNA : RNA interaction blocking translation of hns RNA. Positive regulation of rpoS by DsrA occurs by the formation of an RNA structure that activates a cis-acting translational operator (Lease et al 1998).

(ii) OxyS RNA is a negative regulator: It is induced in response to oxidative stress. It acts as a global regulator to activate or repress the expression of as many as 40 genes, including rpoS. OxyS RNA inhibits translation of the rpoS mRNA; this repression is dependent on HF-I. OxyS RNA binds HF-1 and, as a result, represses rpoS translation by altering HF-1 activity (Zhang et al 1998). In conclusion, rpoS translation is controlled by a regulatory network that includes rpoS mRNA, HF-1(Hfq), DsrA RNA, OxyS RNA, H-NS and LeuO. In this network H-NS plays a dual role: (a) by interfering with rpoS translation in general and, (b) via LeuO, influencing the synthesis of its own low-temperature antagonist, DsrA-RNA.

3.4 Turnover of Sigma-S

The rate of Sigma-S proteolysis is tightly regulated: the Sigma-S half-life varies between 1·5 min and more than 40 min, depending on particular stress conditions. Proteolysis is more, or more strongly, inhibited in response to: (i) starvation, (ii) high osmolarity, (iii) heat shock and (iv) acidic pH. Osmotic shift is only known stress that at the same time stimulates rpoS translation and almost completely inhibits Sigma-S proteolysis (Hengge-Aronis 1996). An element required for Sigma-S degradation (the target for proteolysis, or "turnover element") is encoded between nucleotides 379 and 742 of the rpoS coding sequence (Muffler et al 1996a). Amino acid residues 173 to 188 of Sigma-S may directly or indirectly serve as at least part of the target for the ClpX protease (Schweder et al 1996). According to the recent results the "turnover element" is a small element (around the crucial amino acid lysine-173) directly downstream of the promoter-recognizing region in RpoS. It represents a unique proteolysis-promoting motif and is a site of interaction with RssB (Becker et al 1999; see § 3.4a).

3.4a ClpX protease is responsible for Sigma-S degradation: In growing cells ClpXP degrades the part of Sigma-S molecules which in exponential phase is made in excess. In case of an abrupt stress it is quicker to provide an adequate level of Sigma-S by the cessation of its degradation than by the initiation of its synthesis. Indeed, in mutants deficient in ClpXP, Sigma-S levels of exponential-phase cells increase to those of stationary-phase wild-type cells. But in stationary phase, Sigma-S becomes more resistant to this protease (Schweder et al 1996).

(i) RssB protein is a negative regulator of ClpXP: RssB protein (also called SprE and MviA) belongs to the class of two-component response regulators whose activity is mediated by a C-terminal "output" domain and controlled by an N-terminal "receiver" domain. The receiver domain is similar in all response regulators and is phosphorylated or dephosphorylated in response to signals transmitted by sensory histidine kinases. With few exceptions, response regulators act as transcription factors with the C-terminal DNA-binding domain (Volz 1995). rssB null mutants exhibit nearly constitutively high levels of Sigma-S. Sigma-S is stable in rssB mutants, indicating that RssB is essential for Sigma-S turnover. A unique C-terminal output domain of RssB is the first known response regulator involved in the control of protein turnover (Muffler et al 1996b; Zhou and Gottesman 1998). Thus, RssB is functionally unique among response regulators as a direct recognition factor in ClpXP dependent RpoS proteolysis. Activity of RssB is modulated by acetyl phosphate.

(ii) Acetyl phosphate (AcP): Its intracellular level is highly variable, depending on the nutritional status of the cells. Growth on glucose, pyruvate or rich medium results in a high level of AcP and the production of acetate, whereas growth on glycerol or entry into stationary phase results in lower AcP levels (McClearly and Stock 1994). There are increased in vivo half-lives of Sigma-S in acetyl phosphate-free (pta-ackA) deletion mutants, even though no sensor kinase is eliminated. The in vivo data indicate that AcP acts through the response regulator RssB. Via the phosphorylation, AcP may thus modulate RssB (Bouche et al 1998). Binding of RssB to RpoS is stimulated by phosphorylation of the RssB receiver domain, suggesting that environmental stress affects RpoS proteolysis by modulating RssB affinity for RpoS. Initial evidence indicates that lysine-173 in RpoS, besides being essential for RpoS proteolysis, may play a role in promoter recognition. Thus the same region in RpoS is crucial for proteolysis as well as for activity as a transcription factor (Becker et al 1999).

Polyphosphates could participate in modulation of Sigma-S level also. If inorganic polyphosphate [poly(P)] levels in E. coli were reduced to barely detectable concentrations, the synthesis of catalase HPII (which is positively controlled by Sigma-S) was greatly diminished. If the level of poly(P) was restored, the level of HPII was restored also (Shiba et al 1997).

In general, the regulatory cascade AcP – RssB – ClpXP – Sigma-S responds to the metabolic status of the cells reflected in the highly variable cellular AcP concentration. But how does Sigma-S become resistant to degradation under stress?

3.4b DnaK is a negative regulator of ClpXP: ClpX protease is inhibited by the chaperone DnaK whose level increases drastically during different stresses (Loewen et al 1998, see also § 4.1a).

3.5 Sigma-S dependent promoters: specific – 35 box, DNA bending, low superhelicity

Three different kinds of promoters are defined according, among other things, to their dependence on the growth rate of the cell: (i) the ‘house-keeper’ promoter of many metabolic genes, (ii) the stringent promoter found at several rRNA and ribosomal protein genes, and (iii) promoters of the genes whose products are required at higher relative amounts at lower growth rates (Vicente et al 1991). With the use of directed mutagenesis it was shown that Sigma-70 and Sigma-S can recognise the same – 10 sequences, but DNA sequences in the – 35 region of Sigma-S dependent promoters function as part of a discriminator mechanism to shift transcription between Sigma-70 and Sigma-S (Wise et al 1996). Promoters recognised by Sigma-S are located in regions where DNA shows intrinsic curvatures. This feature does not appear in a stationary phase-induced promoter which is not controlled by Sigma-S. It seems that DNA bending may help in recognition and/or binding of Sigma-S to stationary phase induced promoters (Espinosa-Urgel and Tormo 1993). The transcription directed by Sigma-S is enhanced with the use of templates with low superhelical density. This is in good agreement with the decrease in DNA superhelicity in the stationary growth phase. The selective transcription of stationary-specific genes by E-Sigma-S holoenzyme would seem to require either a specific reaction condition(s) or a specific factor(s) such as template DNA with low superhelical density (Kusano et al 1996).

The activity of Sigma-S may also be modulated by glutamate, which activates holoenzyme formation and promotes holoenzyme binding to certain promoters. In the study of in vitro transcription of the osmoregulated promoters of the E. coli osmB and osmY genes, it was shown that under conditions of low ionic strength, the osmB and osmY promoters were initiated by both Sigma-70 and Sigma-S. Addition of up to 400 mM potassium glutamate (K⁺ glutamate), mimicking the intracellular ionic conditions under hyperosmotic stress, specifically enhanced transcription at these promoters by Sigma-S but inhibited that by Sigma-70. In general, the RNA polymerase with Sigma-S can itself sense osmotic stress by responding to changes in intracellular K⁺ glutamate concentrations and altering its promoter selectivity in order to recognise certain osmoregulated promoters (Ding et al 1995).

The current view of Sigma-S promoter recognition specificity is that it is very similar or practically the same as that of Sigma-70. Indeed, under various stress conditions Sigma-S and Sigma-70 coexist in cells. Although they are structurally and functionally very similar and in vitro they basically recognise the same promoter sequences, in vivo these sigma factors clearly control different genes. Sigma factor specificity at stress-activated promoters is affected by the interplay of the two RNA polymerase holoenzymes with additional regulatory factors, such as H-NS, Lrp, CRP, IHF, that differentially affect transcription initiation by Sigma-S or Sigma-70 in a promoter-specific manner (see review Hengge-Aronis 1999).

3.6 Genes down-regulated by Sigma-S

The increased level of Sigma-S markedly reduces stationary phase expression of a Sigma-70 dependent promoter because both sigma factors compete for a limiting amount of RNA polymerase during stationary phase. By this mechanism Sigma-S could act as a negative transcriptional regulator (Farewell et al 1998).

3.6a MMR (methyl directed mismatch repair): The MutS, MutL, and MutH of MMR proteins play major roles in several DNA repair pathways (see review Radman et al 1995). The cellular amounts of MutS and MutH decreased by as much as 10-fold in stationary phase cultures. The levels of MutS, MutL, and MutH are regulated by two global regulators, Sigma-S and HF-I (Hfq); mutations in hfq and rpoS reversed the stationary phase down-regulation of the amounts of MutS and MutH.

(i) MutS: The amount of MutS in exponentially growing cells is post-transcriptionally regulated by HF-1, but not by RpoS.

(ii) MutH: HF-1 regulation of the amount of MutH is mediated only through RpoS.

(iii) MutL: The amount of MutL remained unchanged in rpoS, HF-1, and rpoS⁺; hfq⁺ strains in exponentially growing and stationary-phase cultures and served as a control. In general, the levels of MutS and MutH may be adjusted in cells subjected to different stress conditions by an RpoS-dependent mechanism (Tsui et al 1997, see also § 10.1b).

Many different environmental factors regulates the activity of the one protein – Sigma-S. At least, five different stages (which determine the activity of Sigma-S) are involved in accepting and integrating the corresponding regulatory signals. These stages are: transcription of rpoS, translation of rpoS mRNA, degradation of rpoS protein, changing of the superhelicity of the corresponding promoters and activation or inactivation of RNA polymerase Sigma-S holoenzyme on specific promoters by specific compounds. And what is the result? The result is that bacteria are most overprovident beings. If starvation is coming, bacteria ‘prepare’ to be shocked. If they are confronted with an one shock, they ‘prepare’ to meet many others.

4. Shocks induce Sigma-S; Sigma-S induces
shock tolerance

Many Sigma-S regulated starvation proteins are common to those induced by other stresses. Glucose- or nitrogen-starved cultures of E. coli exhibited enhanced resistance to heat (57°C) or H₂O₂ (15 mM) challenge, compared with their exponentially growing counterparts. The degree of resistance increased with the time for which the cells were starved prior to the challenge. Starved cultures also demonstrated stronger thermal and oxidative resistance than did growing cultures adapted to heat, H₂O₂, or ethanol prior to the heat or H₂O₂ challenge. Subsets of the 30 glucose starvation proteins were also synthesized during heat or H₂O₂ adaptation; three proteins were common to all three stresses (Jenkins et al 1988).

4.1 The heat shock

Heat shock increases Sigma-S levels by inhibition of its proteolysis due to the increased steady-state level of the heat shock induced DnaK chaperone machine which directly (or indirectly) protects Sigma-S against degradation. The findings that not only Sigma-32, but Sigma-S is heat-induced and that heat shock protein DnaK is involved in opposite ways in the control of the two sigma factors suggests connection between the stress responses mediated by these two sigma factors. Despite its induction by temperature up-shift, Sigma-S does not seem to contribute to heat adaptation but may induce cross-protection against different stresses (Muffler et al 1997).

4.1a DnaK, the single protein for multiple tolerances: DnaK is essential for starvation-induced resistance to heat, oxidation, and reductive division. It is also required for starvation-induced osmotolerance, catalase activity, and the production of the Sigma-S controlled Dps protein. During carbon starvation, DnaK deficiency reduced RpoS levels 3-fold, while DnaK excess increased RpoS levels nearly 2-fold. DnaK deficient mutant phenotypes closely resemble those of rpoS mutants (Rockabrand et al 1998). The cellular content of DnaK is increased also in response to osmotic upshift and does not decrease as long as osmolarity is high. Also DnaK protein is required directly or indirectly for the maintenance of K⁺ transport at high osmolarity (Meury and Kohiyama 1991).

4.1b Trehalose: otsA and otsB genes, which encode trehalose-phosphate synthase and trehalose-6-phosphate phosphatase, respectively, also are involved in Sigma-S dependent stationary phase thermotolerance. Neither Sigma-S nor trehalose, however, is required for the development of adaptive thermotolerance in growing cells, which might be controlled by Sigma-32 (Hengge-Aronis et al 1991).

4.2 Cold shock

At low temperature the level of Sigma-S is enhanced by the stimulation of rpoS translation via action of DsrA-RNA (see § 3.3b).

4.3 Osmotic shock

Cells growing at high osmolarity already exhibit increased levels of Sigma-S during the exponential phase of growth. RpoS expression is itself osmotically regulated by a mechanism that operates at the post-transcriptional level. Sigma-S also acts as a global regulator for the osmotic control of gene expression, and actually does so in exponentially growing cells. Stimulation of rpoS translation and a change in the half-life of sigma(s) from 3 to 50 min both contribute to osmotic induction (Muffler et al 1996a).

4.3a Osmotolertance: E. coli can adapt and recover growth at high osmolarity. Adaptation requires the deplasmolysis of cells previously plasmolyzed by the fast efflux of water promoted by osmotic upshift. Deplasmolysis is essentially ensured by a net osmo-dependent influx of K⁺. Stationary-phase E. coli cultures also showed enhanced osmotic resistance as compared with cultures in mid-logarithmic growth or preadapted to osmotic stress. Of the 22 polypeptides induced during osmotic shock, five were also starvation proteins (Jenkins et al 1990; Hengge-Aronis et al 1993).

4.4 Oxidative shock and oxidative tolerance

Tolerance to oxidation can be induced by hydrogen peroxide, but also by glucose- or nitrogen-starvation. Exposure of E. coli to hydrogen peroxide induces the transcription of a small OxyS RNA. The OxyS regulatory RNA integrates the adaptive response to hydrogen peroxide with other cellular stress responses and protects against DNA damage (Altuvia et al 1997; Zhang et al 1998) (see § 3.3b).

4.5 Tolerance to visible light

E. coli loses its ability to form colonies in marine environments when exposed to artificial continuous visible light. Survival of illuminated bacteria during the stationary phase is drastically reduced in the absence of Sigma-S (Gourmelon et al 1997).

4.5a High magnetic field induce Sigma-S: When E. coli was aerobically grown under inhomogeneous 5·2–6·1 Tesla magnetic fields in a superconducting magnet biosystem (SBS), the cell number in the stationary phase after the growth was about three times higher than that under a geomagnetic field. When E. coli defective in the rpoS gene was cultivated in SBS, the enhancement of cell survival was significantly reduced. With the use of a rpoS-lacZ fusion a significant increase in the activity of beta-galactosidase was observed in stationary phase under high magnetic field. These data suggest that enhancement of rpoS transcription in stationary phase is involved in the higher survival of the cells in a magnetic field (Tsuchiya et al 1999).

4.6 Tolerance to toxic agents

Stationary-phase E. coli cells are more resistant to exposure to the toxic electrophile N-ethylmaleimide (NEM) than exponential-phase cells. RpoS and the Dps aid the survival of both exponential- and stationary-phase cells against NEM. Alterations in the level of Sigma-S in exponentially growing cells correlate with the degree of NEM sensitivity. A slower growing E. coli strain was also found to accumulate Sigma-S and had enhanced resistance to NEM (Ferguson et al 1998).

4.7 Viability and culturability

When grown in rich medium, E. coli exhibits a drastic reduction of the number of viable cells at the beginning of stationary phase. The decline of cell viability was retarded by disruption of the ssnA gene, which was identified as a gene subject to RpoS-dependent negative regulation. Moreover, ssnA expression was induced at the time of decline of cell viability at early stationary phase. The viability decline was augmented in the rpoS background, and this augmentation was suppressed by ssnA mutation. Cloning of the ssnA gene in a multicopy plasmid, pBR322, caused small colony formation and slow growth in liquid medium. Cells harbouring the ssnA clone showed aberrant morphology that included enlarged and filamentous shapes. It was concluded that ssnA is expressed in response to a phase-specific signal(s) and that its expression level is controlled by RpoS, by a mechanism which may contribute to determination of cell number in the stationary phase (Yamada et al 1999). Also, Sigma-S possibly via ppGpp regulation, positively influenced the culturability of E. coli in oligotrophic seawater. This influence closely depended, however, upon the growth state of the cells and the conditions under which they were grown prior to their transfer to seawater. The protective effect of RpoS was observed only in stationary-phase cells grown at low osmolarity. A previous exposure of cells to high osmolarity (0·5 M NaCl) also had a strong influence on the effect of RpoS on cell culturability in seawater (Munro et al 1995).

Microbial cells are well armed to overcome starvation and stresses and to survive them. But is it all arms and tools the cells have to overcome the unfavourable conditions of life? Do the cells really prefer only to conserve themselves and to wait an indefinitely long time for good days to come, gradually loosing their viability?

5. Mutagenesis under stress

5.1 A double-edged sword of mutations

Mutations are a double-edged sword: they are the ultimate source of genetic variation upon which evolution depends, yet most mutations affecting fitness appear to be harmful. Only deleterious mutations of small effect can escape quick elimination by natural selection, and can accumulate in small populations by drift. The key question is: what is the ratio between deleterious, neutral and beneficial mutations? According to theoretical considerations it is usually argued that because most newly arising mutations are neutral or deleterious, the mutation rate has evolved to be as low as possible (reviews, Miller 1998; Horst et al 1999). But what do the experimental results say? Are there any possibilities to increase the genetical diversity of a microbial population and, with increased probability, to overcome conditions which are not favourable for growth?

5.1a To be a mutator: beneficial over short times and dangerous in the long run?: In early experiments a population of a mutT strain of E. coli which was deficient in MMR was maintained in a chemostat for 2,200 generations and the rate of mutations conferring resistance to three antibiotics was determined. It was found that the resulting strain had a distinctly reduced mutability after long-term cultivation compared with the original strain. Nevertheless the mutability was still much higher than that of a wild-type strain. After transduction of the mutT gene into another genetic background the transductants showed the same mutability as the original strain indicating that the mutT allele itself had not changed. It was concluded that although under new environmental conditions mutator strains have an advantage due to their more efficient production of beneficial mutations, after optimal adaptation there is selection against high mutation rates due to the increased deleterious mutational load in the mutator population (Trobner and Piechocki 1984). The study of another E. coli mutator (dam^–) confirmed this conclusion. In competition experiments between dam^– and dam⁺ strains it was found that dam^– mutator strains are negatively selected (Trobner and Piechocki 1985). Recent estimations of the genomic deleterious mutation rate for total fitness in a microbe show that the per-microbe rate of deleterious mutations is in excess of 0·0002 (Kibota and Lynch 1996). Is this rate the highest possible that is consistent with evolution and that is not self destructive? Could it be that under special environmental conditions increased mutation rates are of evolutionary advantage? Indeed, when 12 independently propagated clonal populations of E. coli, were serially cultured over 10,000 generations in a glucose-limited environment and thereby subjected to alternating periods of growth and stasis, most of them retained the ancestral mutation rate, but three populations displayed mutation rates that were between one and two orders of magnitude higher that those in the ancestor. The ancestral mutation rate was fully restored in these mutator strains only by the presence of wild-type alleles of the genes of MMR and uvrD. The three strains evolved to a mutator phenotype which was due to defects in MMR genes (Sniegowski et al 1997). In recent experiments the frequency of incidence of mutator strains was studied in laboratory cultures of S. typhimurium. It was found that subpopulations of mutators, residing in normal populations at a finite frequency, can be culled from the culture by strong selection for a required phenotype (LeClerc et al 1998). A theoretical study confirmed that in asexual, clonal populations of E. coli, the ability to generate mutator alleles can lead to an increase in mutation rate on account of which increasingly fitter individuals arise in the population. Such models demonstrates that strong mutator genes (such as those that increase mutation rates by 1,000-fold) can accelerate adaptation, even if the mutator gene remains at a very low frequency (for example, 10^–5). Less potent mutators (10- to 100-fold increase) can become fixed in a fraction of finite populations (Taddei et al 1997a). This is consistent with the finding that up to 1% of natural bacterial isolates are mutator clones that have high mutation rates.

5.1b One per cent of natural bacterial isolates are mutators: Up to 1% of natural bacterial isolates are mutator clones that have high mutation rates and these high rates might play an important role in adaptive evolution (Matic et al 1995).

(i) The frequency of mutators among pathogenic strains is 1%: What is of serious clinical concern is that the incidence of mutators among isolates of pathogenic
E. coli and S. enterica is high, over 1%. Among 212 strains of E. coli, 9 were mutators, among 137 strains of S. eneterica 17 were mutators. Putative mutators displayed at least a 50-fold increase in mutation frequency as compared with controls. Of 9 independently derived hypermutable strains, 7 contained a defective mutS allele (LeClerc et al 1996; Matic et al 1997). Because mutant alleles of MMR increase the mutation rate and enhance recombination among diverse species (see § 10.1), these studies may explain both the rapid emergence of antibiotic resistance and the penetrance of virulence genes within the prokaryotic community (Taddei et al 1997b).

5.1c The dark side of the Moon: cancer cells are mutators: In eukaryotes selection of advantageous mutations underlies tumorigenesis. The growth of a tumour is therefore a form of evolution at the somatic level in which the population is comprised of individual cells within the tumour. Some cancers display a "mutator phenotype", probably leading to faster growth. The core components of eukaryotic MMR systems are highly homologous to their bacterial counterparts. In humans, defects in four MMR genes is associated with an increased mutational burden and predisposition to certain malignancies (see review Kolodner and Alani 1994). The large number of mutations reported in tumour cells cannot be accounted for by the low mutation rates observed in normal non-dividing somatic cells; rather, it must be a manifestation of a mutator phenotype present early during the tumorigenic process. The interaction between increased mutagenesis and clonal selection provides a mechanism for the selection of cells with increased proliferative advantage (Loeb 1998).

In general, the activity of MMR is responsible for the constant maintenance of genome stability and its faithful transmission from one generation to the next. However, without genetic alteration species would not be able to adapt to changing environments. In microorganisms programmed and reversible inactivation of MMR in response to environmental stress could lead to an evolution, but in eukaryotic somatic cells the sporadic inactivation of MMR leads to a cancer (Radman et al 1995; Cairns 1998).

It seems that only 1% of the cells of natural populations undergo the risk to be mutators. Is it the result of the events occurring on the populational level? The result of the selection of the randomly occurred mutator strains? Or, may be, there are another evidences? Or could it be the evidence that in some circumstances microorganisms "deliberately" increase the rate of mutagenesis on the intracellular level, but not on the populational?

5.2 What are the environmental conditions that increase the rate of mutagenesis?

5.2a Stringent response: ppGpp: Different isogenic strains of E. coli K-12 differing only in mutations which inactivates the stringent response (relA) were examined with respect to ppGpp levels and reversion rates of a leuB allele. A positive correlation was established between reversion rates and the steady-state concentration of ppGpp during exponential growth. A summary of reversion rates in leucine-limited strains relA⁺ and relA^– is: the reversion rate of the relA⁺ strain was 6·7-fold higher than that of the relA^– strain. Reversion rate of ArgH in relA⁺ strain was 28-fold higher than that of relA^– strain. The proposed explanation is: the enhanced transcription stimulated by ppGpp on the ppGpp activable promoters would increase the concentration of single stranded DNA in specific areas of the genome; the single stranded DNA is uniquely vulnerable to mutagenes. Thus, genetic or environmental conditions favouring high ppGpp levels would result in increased transcription of the genes in a susceptible operon and higher mutation rates in the genes of that operon. This mechanism should operate selectively in nutritional stress genes activated by the stringent response (Wright 1996; Wright and Minnick 1997).

5.2b SOS response without DNA lesions: non-targeted mutagenesis: The cellular response to DNA damage that has been most extensively studied is the SOS response of E. coli. Analysis of the SOS response has led to new insights into the transcriptional and post-translational regulation of processes that increase cell survival after DNA damage as well as insights into DNA-damage-induced mutagenesis, i.e., SOS mutagenesis. SOS mutagenesis requires the operation of a specialized system involving the UmuD, UmuC, RecA and DNA polymerase III proteins, which allows translesion synthesis. Because of the considerable decrease of the fidelity of DNA replication under SOS response this reparation is considered as error prone. Translesion of potentially lethal mutation which are non-consistent with normal DNA replication generates mutations which are consistent with DNA replication and which are potentially non-lethal (see review, Smith and Walker 1998). But what would be the situation if SOS response is induced in the absence of DNA lesions? In fact, induction of the SOS functions by SOS constitutive mutation in the recA730 gene promotes a SOS mutator activity which generates mutations in undamaged DNA which, in turn, increases the level of spontaneous mutation. The number of such recA mutant induced mutations is greatly increased in MMR deficient strains in which replication errors are not corrected, which suggests that the majority of these mutations (90%) arise through correctable, i.e., non-targeted, replication errors (Caillet-Fauquet and Maenhaut-Michel 1988). Dealing with the specificity of non-targeted SOS induced mutations it was shown, that in the strains of E. coli in which SOS system is continuously induced in the absence of mutagenic treatment, it stimulates specifically G : C ®  T : A and, to a lesser extent, A : T ®  T : A transversions (Miller and Low1984). If SOS response is induced by the environmental factors, but not by DNA damaging agents, it will increase the rate of spontaneous mutagenesis. What are these factors?

5.2c Stresses that induce SOS-response

(i) Ageing on the agar plates and cAMP: SOS induction and mutagenesis were observed in ageing E. coli colonies on the agar plates in the absence of exogenous sources of DNA damage. These mutations which are not observed in ageing liquid cultures, accumulate linearly with the age of the colonies. The observed SOS induction and mutagenesis were shown to be controlled by the LexA repressor and are RecA^– and cAMP-dependent (Taddei et al 1995). This cAMP-dependent mutagenesis occurring in so called resting organisms in a structured environment (ROSE) is unaffected by a umuC mutation and therefore differs from both targeted UV mutagenesis and recA730 (SOS constitutive) untargeted mutagenesis. As a recB mutation has only a minor effect on ROSE mutagenesis it also differs from both adaptive mutations (see § 6.1). Besides its recA and lexA dependence, ROSE mutagenesis is also uvrB and polA dependent. In general, ROSE mutagenesis might offer a good model for bacterial mutagenesis in structured environments such as biofilms and for mutagenesis of quiescent eukaryotic cells (Taddei et al 1997c).

(ii) pH stress induces SOS response: Alkalinization of intracellular pH causes an increase in UV resistance in wild-type and pH-sensitive mutant cells of E. coli. The effects of pH on cells functions may involve the lexA product of the SOS system (Schuldiner et al 1986).

(iii) Heavy metals can induce SOS response: Cobalt chloride, when present in the plating medium, was able to block mutagenesis and lysogenic induction promoted by UV irradiation. It was found that CoCl₂ blocked protein synthesis. On the other hand, if the cells were treated for a short time with CoCl₂, in the absence of Mg, CoCl₂ per se promoted lysogenic induction as well as enhanced the phage reactivation induced by UV irradiation. It seems, that depending on conditions, cobalt chloride may act either as an inhibitor or as an inducer of the SOS functions (Leitao et al 1993). Also SOS response could be induced by some transposons (see § 7.3) and by conjugative plasmids (see § 8.3a and 8.3c).

Are there other ways in which the stressed cells can enhance the rate of mutagenesis? Emerging evidence suggests the existence of a number of other stress-inducible pathways that also affect the fidelity of replication. The most provocative recent findings are:
(i) UVM, an SOS-independent damage-inducible mutagenic pathway and (ii) a new recA dependent but umuD/C independent pathway that appears to be provoked by translational stress (see review Humayun 1998).

5.2d UVM response: the new way of stress induced mutagenesis: UVM (ultraviolet modulation of mutagenesis, or May Day response) is a recA independent, inducible mutagenic phenomenon in which prior UV irradiation of E. coli cells strongly enhances mutation fixation (Humayun 1998).

5.2e Factors inducing UVM

(i) UV irradiation: Prior UV irradiation of delta recA cells, in which the SOS pathway does not function, enhances mutagenesis. This indicates the existence of a mechanism which is manifested during gap-filling DNA synthesis as well as during normal DNA replication (Palejwala et al 1994).

(ii) Alkylating agents: 1-methyl-3-nitro-1-nitrosoguanidine, N-nitroso-N-methylurea (MNNG) and dimethylsulphate, but not methyl iodide, are potent inducers of UVM. MNNG induction of UVM is independent of ada, alkA and alkB genes and define UVM as an inducible mutagenic phenomenon distinct from the E. coli adaptive and SOS responses (Wang et al 1995).

(iii) H₂O₂ is a potent inducer of UVM and the induction of UVM by H₂O₂ does not require OxyR-regulated gene expression. UVM induction by H₂O₂ appears to be mediated by DNA damage (Wang and Humayun 1996).

(iv) On the mechanism of UVM response: Since UVM is observed in cells in which SOS induction should not occur, UVM may represent a novel, SOS-independent, inducible response. Three hypothetical mechanisms for UVM were proposed: (i) UVM results from a recA-independent pathway for the induction of SOS genes thought to play a role in induced mutagenesis, (ii) UVM results from a polymerase switch in which replication in treated cells is carried out by DNA polymerase I (or DNA polymerase II) instead of DNA polymerase III, and (iii) UVM results from transient depletion of a MMR that normally acts to reduce mutagenesis (Palejwala et al 1995).

(v) UMV: no relations with SOS and MMR systems: UVM is observable in delta recA cells, in lexA3 (non-inducible SOS repressor) cells, in LexA-overproducing cells, and in delta umuDC cells. Furthermore, UVM induction occurs in the absence of detectable induction of dinD, an SOS gene. It is unlikely that UVM results from a recA independent alternative induction pathway for SOS gene (Palejwala et al 1995). Moreover, normal UVM induction was observed in cells defective for MMR as well as in the cells overexpressing MutH, MutL, or MutS.

Taken together, all these results indicate that UVM represents a generalized cellular response to a broad range of chemical and physical genotoxicants, and that DNA damage constitutes the most likely signal for its induction. Molecular mechanism(s) of the UVM response is independent of known SOS and MMR systems and may thus represent a previously unrecognised misrepair or misreplication pathway (Murphy et al 1996). Another newly discovered mechanism of mutagenesis is the pathway inducible by translational stress.

5.2f Miscoding that decreases proofreading: E. coli cells bearing a mutA or mutC allele display a UVM-constitutive phenotype. These alleles stimulate transversions; the A.T ®  T.A and G.C ®  T.A. Both mutA and mutC result from changes in the anticodon in one of four copies of the same glycine tRNA, at either the glyV or the glyW locus. This change results in a tRNA that inserts glycine at aspartic acid codons. The mistranslation of aspartic acid codons is assumed to occur at approximately 1–2%. The most reasonable explanations are: (i) the mutator tRNA effect is exerted by generating a mutator polymerase and, (ii) the epsilon subunit of DNA polymerase, which provides a proofreading function, is the most likely target (Slupska et al 1996). To test these assumptions a wild-type or a mutant glyV gene were placed under the control of a heterologous inducible promoter on a plasmid vector. And E. coli cells expressing the mutant glyV gene were displayed: (i) missense suppression of a test allele, (ii) a mutator phenotype and, (iii) a UVM-constitutive phenotype (Murphy and Humayun 1997). To verify that the proofreading subunit of polymerase III epsilon is a likely target for the aspartic acid-to-glycine change that leads to a lowered fidelity of replication, 16 altered mutD genes were constructed by replacing each aspartic acid codon, in series, with a glycine codon in the dnaQ gene that encodes epsilon. Indeed three of these genes confer a strong mutator effect. The altered epsilon subunits resulting from this substitution (approximately 1% of the time) are sufficient to create a mutator effect (Slupska et al 1998). But are there any evidences, that under the stresses the normal tRNA will be provoked for miscoding and subsequent mutator effect?

(i) Cells exposed to translational inhibitors display a mutator phenotype: Exposure to streptomycin, an antibiotic known to promote mistranslation, induces a recA and umuDC-independent mutator phenotype detected as enhanced mutagenesis (Ren et al 1999). It seems that starvation and different environmental stresses can induce SOS response (and, consequently, non-targeted mutagenesis) and May Day response. Further studies will show as to what kind of stresses (if any) can induce the "translational miscoding" mutagenesis.

In general, these inducible mutator effects are expected to contribute to the adaptation of bacterial populations to the adverse life conditions by increasing their genetic variability. The evolutionary impact of all these mutator systems would be even greater if they were also induced under conditions common in nature, such as in resting bacterial populations. But what about 1% of mutator strains isolated from natural populations? Are they borne by the populational event – by the selection? Or by the intracellular mechanisms? Indeed, if there are environmental stressful conditions then they must act on all viable cells of the populations. To be a mutator is too risky for the cells. And if all cells of a population become mutators it will be risky for the population. Is it the reason why only a few representatives of the populations "decide to be the mutators"? If so, are these representatives the best – the most viable and with resources enough to mutate, or the worst – nothing to loose? Is there a compromise between a risk to die because of the long period of non growth and the risk to be a mutator?

6. To be a transient mutator: mutagenesis
in static cells

In fact non growing cells are not static in the narrow sense of the word. They can evolve. This evolutionary process was studied by using E. coli cultures incubated for prolonged periods of time in stationary phase. The populations of surviving cells were shown to be highly dynamic, even after many months of incubation. And evolution proceeded along different paths even when the initial conditions were identical. As cultures aged, the take-over by fitter mutants was incomplete, resulting in the coexistence of multiple mutant forms and increased microbial diversity (Finkel and Kolter 1999).

Adaptive mutagenesis – this unexpected way of generation of mutations in non growing bacteria has amazed the scientific community and challenged the concepts of
the genetic mechanisms behind evolution. Discovery and study of the molecular mechanism of these so called "adaptive" (or directed, or stationary phase induced) mutations (or adaptive reversions) occurring in non growing cells (and as it was believed – in Lamarckian fashion or in direct response to the selection) is rapidly revealing a surprising and novel molecular mechanism, and it is altering the understanding of how mutations form in nondividing cells (see reviews Cairns 1998; Bridges 1998; Rosenberg et al 1998). Adaptive mutations are induced and realized in microorganisms during periods of prolonged stress in non-dividing or very slowly dividing populations. These mutations are almost invariably those that enable the cell to resume growth (Hall 1998a).

6.1 On the mechanism(s) of adaptive mutagenesis

There are two basic assumptions: (i) an endogenous mutagen and (ii) DNA turnover in non growing cells. These are responsible for adaptive mutagenesis in the following manner:

6.1a An endogenous mutagenic substance accumulates in static cells: DNA lesions resulting from the action of these endogenous mutagens may give rise to RNA transcripts with miscoded bases; if these confer the ability to initiate DNA replication (to resume the growth), the DNA lesions may have an opportunity to miscode during replication and thus could give rise to apparently adaptive mutations.

6.1b Turnover of DNA in resting cells: There is a significant turnover of chromosomal DNA in static bacteria. This could permit polymerase errors to lead to mutations in non-dividing cells. Such cryptic DNA synthesis, which may essentially replace existing DNA rather than duplicating it, could, in principle, act as an additional source of variability on which selection may act, initially in the absence of cell division (Bridges 1997). The next question is: do the static cells make the endogenous mutagen? It seems, yes.

6.1c Mutations in static cells: the role of endogenous DNA damage: Singlet oxygen may be an important endogenously produced mutagen in resting cells. In fact reversion to prototrophy of certain amino acid auxotrophs of E. coli that occurs when the bacteria are starved of a required amino acid results from the accumulation of oxidative damage to guanine residues in DNA. Thus adaptive mutagenesis was approximately 4-fold more frequent in a sodA sodB strain than in the superoxide dismutase-replete parental strain and this mutagenesis was suppressed under anaerobic conditions. Moreover, a cell permeant manganic porphyrin, capable of catalysing the dismutation of O₂^–  , diminished the rate of occurrence of these mutations (Benov and Fridovich 1996). The introduction of a plasmid specifying the production of singlet oxygen scavengers (carotenoids) in stationary phase cells led to a roughly 2-fold reduction in adaptive mutants yield (Bridges and Timms 1998). Repair of oxidative damage
to DNA, in the non-dividing cells, appears to provide the opportunity for adaptive mutagenesis. But what is the direct mutagenic agent? It seems oxidized guanine
(8-oxo-7,8-dihydro-guanine or 8-oxoG) is the direct against 8-oxo-G is potent mutagen because of its ambiguous pairing with cytosine and adenine and so gives rise to G
to T transversions. It is responsible for almost half the G to T transversions arising in static repair proficient strains. The MutT protein specifically hydrolyses both
8-oxo-deoxyguanosine triphosphate (8-oxo-dGTP) and 8-oxoguanosine triphosphate (8-oxo-rGTP), which are otherwise incorporated in DNA and RNA opposite template A. In vivo this cleaning of the nucleotide pools decreases both DNA replication (and transcription) errors. The effect of MutT mutation on transcription fidelity was shown to depend on oxidative metabolism. Such control of replicational and transcriptional fidelity by the ubiquitous MutT function has implications for adaptive mutagenesis and functional maintenance of non dividing cells. And the inactivation of MutT will increase the mutagenic action of 8-Oxo-G (Bridges 1996; Taddei et al 1997d).

But the most exciting and confusing question concerning the nature of adaptive mutations was:

6.1d Darwinian or Lamarckian?: It seems that the controversial issue is resolved recently – the mutations
are Darwinian, not Lamarckian (see review, Rosenberg 1997). Also, it was shown, that adaptive mutagenesis
is not related to bacterial sex and is not associated exclusively with the episomes (Rosenberg 1997) as was suggested earlier (Radicella et al 1995). According to
the recent findings, adaptive mutations occur genome-wide but only in a hypermutable subpopulation of stressed cells (Rosenberg 1997). One of the most studied examples of adaptive mutation is a strain of E. coli FC40, that cannot utilize lactose (Lac^–) but that readily reverts to lactose utilization (Lac⁺) when lactose is its sole carbon source. Adaptive reversion to Lac⁺ occurs at a high rate when the Lac^– allele is on an F¢ episome and conjugal functions are expressed. It was previously believed that nonselected mutations on the chromosome did not appear in the Lac^– population while episomal Lac⁺ mutations accumulated, but it remained possible that nonselected mutations might occur on the episome. To investigate this possibility, a second mutational target was created on the Lac-episome by mutation of a Tn10 element, which encodes tetracycline resistance (Tet^r), to tetracycline sensitivity (Tet^s). Reversion rates to Tet^r during normal growth and during lactose selection were measured. The results clearly demonstrated that nonselected Tet^r mutations do accumulate in Lac^– cells when those cells are under selection to become Lac⁺. In addition, the results suggested that during lactose selection, both Lac⁺ and Tet^r mutations are created or preserved by the same recombination-dependent mechanism (see § 6.1f). In general, reversion to Lac⁺ in FC40 does not appear to be adaptive in the narrow (Lamarckian) sense of the word (Foster 1997). A reversion of four mutants Tet^r genes on pBR322-based plasmids were examined also. Lac⁺ adaptive revertants, and Lac^– unstressed and stressed cells derived from strain carrying each plasmids were tested on the frequency of Tet^r revertants. The hypermutation was seen only among adaptive Lac⁺ revertants but not amongst Lac^– cells from the same starved cultures. This suggests, that a subpopulation of cells exposed to starvation on lactose experienced mutability that can affect a gene in another replicon. Similar results were obtained in a study of the unselected hypermutation in multiple genes in the bacterial chromosome. The data demonstrated a strong correlation between Lac⁺ adaptive reversions and reversion of an unselected gene which could be located in different replicons. Moreover, in a special experiment it was shown that majority of unselected mutation formed coincidentally with Lac⁺ adaptive reversions but not during subsequent growth of Lac⁺ colonies. The frequency of Lac⁺ revertants represented approximately 10^–6 of the population. The mutation rates for unselected genes are similar to that for selected (Torkelson et al 1997).

6.1e Mutator state of static cells is transient: Indeed most adaptive mutants do not possess a heritable stationary phase mutator phenotype, although a small proportion of heritable mutators was found. Amongst 55 Lac⁺ adaptive revertants with associated unselected mutations 49 were non-mutators and 6 were mutators. There are similarities between these well-studied systems and several recent examples of adaptive evolution associated with heritable mutator phenotype in a similarly small proportion of survivors of selection in nature and in the laboratory have been found (see § 8.3a). It was proposed that a transient mutator state was a predominant source of adaptive mutations in these latter systems, the heritable mutators being a minority; heritable mutators may sometimes be a product of, rather than the cause of, hypermutation that gives rise to adaptive mutations (Rosenberg et al 1998). A broad mutational screen, loss of motility (motility could be affected by mutations in as many as 55 genes which represent 1% of the E. coli genome) was used to compare the frequency of non-selected mutations in starved Lac^– cells, in selected Lac⁺ cells, and in those few Lac⁺ revertants that carried an additional mutation. This procedure allowed the estimation of both the size of the hypermutating subpopulation and the magnitude of its increase in mutation rate. It was found that the hypermutating subpopulation makes up approximately 0·06% of the population and its mutational rate is elevated approximately 200-fold. From these numbers it was calculated that the hypermutators are responsible for nearly all multiple mutations and that they produced only 10% of the adaptive Lac⁺ revertants (Rosche et al 1999). If the static cell, undergoing the adaptive mutagenesis resumes growth because of the fixation of the beneficial mutation then the process of adaptive mutagenesis will be switched off.

6.1f Proteins which must be active for adaptive mutagenesis

(i) recA and recBCD: The genetic requirements for adaptive mutation are those which are needed for homologous recombination in the RecBCD pathway. Recombination deficient recA and recB null mutant strains are deficient in adaptive mutations. A hyper-recombinogenic recD strain is hypermutable, and its hypermutation depends on functional recA and recB genes. Genes of subsidiary recombination systems are not required. These results clearly indicate that the molecular mechanism by which adaptive mutation occurs includes recombination. But no such association is seen for spontaneous mutation in growing cells (Harris et al 1994; McKenzie et al 1998).

(ii) polIII: The adaptive mutations decrease in strains with an antimutator DNA polymerase III allele. The latter finding could imply that DNA PolIII itself makes adaptive mutations. Alternatively, normal DNA PolIII errors could saturate post-synthesis mismatch repair during adaptive mutation. If so the antimutator strain would produce fewer adaptive mutations because it possesses greater capacity for mismatch repair which could correct errors made by a polymerase other than DNA PolIII. In fact the antimutator PolIII allele decreases adaptive mutation even in mismatch repair-defective cells. This supports a direct role for DNA PolIII in recombination dependent adaptive mutation (Harris et al 1997a). In addition the sequences spectrum of adaptive reversions of a lac frameshift mutation in E. coli resemble DNA polymerase errors (Rosenberg et al 1995).

6.1g The proteins which must be inactive for adaptive mutagenesis: the programmed failure of MMR: The adaptive mutagenesis is stimulated in the absence of MMR. Loss of MMR activity in E. coli mutS^– strain increases the rate of adaptive mutations 100-fold (Foster and Cairns 1992). As was mentioned, the cellular amounts of MutS and MutH decreased drastically in stationary phase cultures (see § 3.6a and 10.1b). The growth-dependent mutation spectrum can be made indistinguishable from adaptive mutations by disallowing MMR during growth (Longerich et al 1995).

6.2 Nature of the mutation occurring in static cells

Adaptive reversions of lac frameshift mutation are – 1 deletions in small mononucleotide repeats, whereas growth-dependent reversions are heterogeneous. In fact adaptive reversion of a + 1 frameshift mutation is shown to occur by – 1 deletions in regions of small mononucleotide repeats. This pattern makes improbable recombinational mechanisms for adaptive mutation in which blocks of sequences are transferred into the mutating gene, and it supports mechanisms that use DNA polymerase errors. The pattern appears similar to that of mutations found in yeast cells and in hereditary colon cancer cells that are deficient in mismatch repair. These results suggest a recombinational mechanism for adaptive mutation that functions through polymerase errors that persist as a result of a deficiency in post-synthesis mismatch repair (Rosenberg et al 1994). Also, as was mentioned, the sequences of adaptive reversions of a lac frameshift mutation in E. coli resemble DNA polymerase errors, and the adaptive reversions decrease in strains with an antimutator DNA polymerase III allele (Harris et al 1997a). But a comparison of the spectra of spontaneous growth dependent and adaptive mutations in the another gene – ebgR shows that both spectra are dominated by insertion sequence (IS)-mediated mutations. The difference between growth dependent mutations (61% IS mediated) and adaptive mutations (80% IS mediated) is highly significant. In contrast, the spectra of growth-dependent and adaptive non-IS mediated mutations do not differ from each other and therefore do not provide support for the hypothesis that adaptive and growth-dependent mutations arise by substantially different mechanisms (Hall 1999).

In general the molecular mechanisms of adaptive mutations include a requirement for: (i) homologous recombination; (ii) the implication of DNA double-strand breaks as a molecular intermediate needed for action of RecBCD, (iii) a unique sequence spectrum of – 1 deletions in mononucleotide repeats which implies polymerase errors, and also, (iv) implies a failure of postsynthesis mismatch repair on those errors.

6.3 Regulation of adaptive mutagenesis

6.3a Sigma-S and HF-1 are negative regulators of MMR: In fact, the cellular amounts of MutS and MutH decreased by as much as 10-fold in stationary-phase cultures. The levels of MutS, MutL, and MutH are regulated by two global regulators, Sigma-S and HF-I(Hfq) (see § 3.6a and 10.1a). In general, the levels of MutS and MutH may be adjusted in cells subjected to different stress conditions by an RpoS-dependent mechanism. HF-1 directly or indirectly regulates several genes, including MutS an RpoS-independent mechanism that destabilizes transcripts (Tsui et al 1997). Overproduction of MutL inhibits mutation in stationary phase but not during growth. MutS overproduction has no such effect and MutL overproduction does not prevent stationary phase decline of either MutS or MutH. These results imply that MutS and MutH decline to levels appropriate for the decreased DNA synthesis in stationary phase, whereas functional MutL is limiting for mismatch repair specifically during stationary phase (Harris et al 1997b). Although it is generally believed that the failure of MMR is responsible for adaptive mutagenesis, according to the recent data, evidence in support of this hypothesis is lacking. According to this study, the MMR system is no less effective in correcting errors during prolonged selection than it is during growth. Furthermore, MMR proteins supplied in excess reduce both growth dependent and adaptive mutation (Foster 1999). Another type of adaptive mutations is fusions (deletions) formation in stationary cells.

6.3b Sigma-S and H-NS are regulators of the formation of araB-lacZ fusions: Sigma-S is strictly required for the appearance of the stationary phase induced araB-lacZ fusion clones of E. coli MCS2, whereas, H-NS is negative regulator of their emergence. This result clearly shows that genetic changes leading to adaptive mutation in this model system are regulated by physiological signal transduction networks (Gomez-Gomez et al 1997). Indeed the formation of araB-lacZ coding sequence fusions in E. coli is a particular type of chromosomal rearrangement induced by Mucts62, a thermoinducible mutant of mutator phage Mu. Fusion formation occurs after aerobic carbon starvation and requires the phage encoded transposase pA, suggesting that these growth conditions trigger induction of the Mucts62 prophage. The thermal induction of the prophage accelerated araB-lacZ fusion formation, confirming that derepression is a rate-limiting step in the fusion process. But when it was shown that the Mucts62 prophage was derepressed in stationary phase at low temperature, this derepression did not apply to prophages that expressed the Mu wild-type repressor. The derepression was dependent upon the host ClpXP and Lon ATP-dependent proteases and the Sigma-S, but not upon Crp. The maintenance of the derepressed state required the ClpXP and Lon host proteases and the prophage Ner-regulatory protein (Lamrani et al 1999).

6.3c PhoPQ is a positive regulator of adaptive mutagenesis at ebgR: ebgR is a gene that specifies a repressor which controls expression of the ebgAC-encoded Ebg-galactosidase. The phoQ gene encodes the sensor kinase component and phoP encodes the positive regulatory component of a two component regulatory system PhoP/PhoQ that controls expression of at least 50 genes in E. coli, among them the several genes that are induced by starvation for carbon, nitrogen, or phosphorus and are thought to be Pex genes which are involved in the development of resistant state during entry into stationary phase (Groisman et al 1992). Disruption of phoP or phoQ significantly reduces the adaptive mutation rate to ebgR, indicating that the adaptive mutagenesis machinery is regulated, directly or indirectly, by phoPQ (Hall 1998b). The findings that adaptive mutagenesis is regulated implies that this mutagenesis does not simply result from a simple failure of various error correction mechanisms during prolonged starvation.

In general: (i) although the understanding of the mechanisms of adaptive mutagenesis is far from being perfect – this process really exists; (ii) adaptive mutations are induced by the stress(es) but not by the specific selective conditions; (iii) adaptive mutations are random and, thus, could be deleterious also; (iv) only the small subpopulations of static cells undergo the adaptive mutagenesis; (v) the adaptive mutator state is transient, if the static and mutating cell resume the growth (because of the fixation of the random beneficial mutation) – the rate of mutagenesis and the spectrum of mutations returns to normal (as in growing cells).

What is really impressive is the multitude of pathways leading to mutagenesis in static cells: (i) via stringent response; (ii) via cAMP and SOS response; (iii) via recABCD recombination pathway; (iv) via decrease of mismatch repair, and, maybe; (v) via May Day response. Does it mean that for the special stress there is special mutagenesis? With its special frequency and specificity? And are all these multiple pathways of mutagenesis in static cells related, coordinated and cross regulated? What is really exciting – only small subpopulation of static cells undergo adaptive mutagenesis and only subpopulation of natural isolates are mutators. What is the mechanism choosing cells to be the risky mutators? Its Majesty – The Chance?

7. Transposons under stress

Insertion sequences, transposons and other mobile DNA elements are found in all species of bacteria and archaebacteria where they have been sought and are usually considered to be genomic parasites or selfish genes. However, many transposons and other mobile repetitive DNA are remarkably species or phyla-specific, indicating that infection with transposable elements coincides with speciation events and is involved in promoting evolutionary change. Although being selfish, as the side effect of their multiplication and spreading – they can promote a large number of the diverse genetical events: insertions, deletions, fusions, recombinations, inversions, transpositions, translocations, amplifications etc., (see reviews, Stellwagen and Craig 1998; Hallet and Sherratt 1997; Klecner et al 1996). Do transposons have any special genetical programmes which are realized in situations when their host cells undergo the stresses? If yes, how does this programme affect the stressed host cells?

7.1 Under SOS response the transposons excise

The precise excision of Tn5 and Tn10 from the chromosomal insertion sites was induced in E. coli K-12 and B cells, wild-type for DNA-repair, both by the low doses and the high doses of UV-light, respectively. What is of special interest – the precise excision of these transposons was induced by the range of low doses incapable of inducing targeted point mutations. The mechanisms of UV induced transposons excision is too sensitive to provide the excision before the UV irradiation became mutagenic (Aleshkin et al 1998). UV irradiation of E. coli carrying an IS10 element in the plasmid led to an increase of up to 28-fold in IS10 transposition. UV radiation also induced transposition of IS10 from the chromosome. This induction was not dependent on the umuC and uvrA gene product, but it was also not observed in lexA3 and delta recA strains, indicating that the SOS stress response is involved in regulating UV-induced transposition. It was proposed that IS10 transposition, known to increase the fitness of E. coli, may have been recruited under the SOS response to assist in increasing cell survival under hostile environmental conditions (Eichenbaum and Livneh 1998).

7.1a Chemicals inducing SOS response stimulate the excision of transposons: During the study of the ability of 23 chemicals (carcinogens and non-carcinogens) to induce precise excision of Tn10 it was observed that Tn10 precise excision was induced only by potent SOS mutagens. This is in good accordance with data on the SOS dependence of the induction of precise excision of Tn10 (Rusina et al 1992).

7.1b Chemicals inhibiting SOS response repress the excision of transposons: In fact DNA-repair inhibitors: caffeine, ethionine, acriflavine, procaine and cinnamaldehyde are effective inhibitors of precise excision of Tn10 (MacPhee and Hafner 1988). What are the mechanisms of SOS induced excision of transposons? These mechanisms are complex and many different proteins responsible for recombination and reparation are involved.

7.1c Genes involved in the excision of transposons: LexA, RecA; RecBC, RecF; RecG, RecN; uvrB, uvrD, polA, Ruv: In general precise and nearly precise excision of transposon Tn10 occur by host-mediated processes unrelated to transposition. Three mutations that enhance excision of Tn10 and of the structurally analogous transposon Tn5 were selected. All these mutations were unusual alleles of the recB and recC genes that alter but do not abolish RecBC function (Lundblad et al 1984).
UV induction of Tn10 excision is not evidenced in SOS response non inducible lexA3(ind^–) mutant. High levels of RecA synthesized by a recA⁺ gene not repressible by LexA do not relieve the non-inducibility of Tn10 excision in a lexA3(ind^–) background which indicates that the expression of an SOS gene different from recA is necessary for the induction of Tn10 excision. It seems that UV-induced Tn10 precise excision requires SOS induction and that it involves a pathway different from point mutagenesis (Levy et al 1993). A mutant defective in the induced excision of Tn10 was characterized as recN and showed a markedly decreased frequency of excision of Tn10 after treatment with UV or mitomycin C. Thus recN is involved in the induced excision of Tn10 (Chan et al 1994). Ruv mutants which are defective in DNA repair and recombination, showed diminished frequencies of both spontaneous and UV- or mitomycin C (MMC) induced excision of Tn10. RecG mutants, which are also defective in DNA repair and recombination, showed decreased induction of Tn10 excision with MMC, but not after UV treatment. It was proposed that the Ruv proteins, which are known to be involved in the resolution of Holliday junctions, might also be involved in the resolution of putative intermediates generated during the induced precise excision of Tn10 (Nagel et al 1994). Induced excision of Tn10 in the uvrD null mutant depends on the expression of recA rather than on any of the other genes repressed by LexA. A null recF mutation in combination with a uvrD^– mutant was also found to abolish the increased frequencies of this process. The recF mutation increased precise excision of Tn10 induced by MMC in a uvrD⁺ isogenic strain. These observations indicate that recA and recF are involved in the increased frequencies of Tn10 excision exhibited by uvrD mutants or after MMC treatment (Chan and Nagel 1997). Excision and transposition of the Tn5 element in E. coli ordinarily appear to occur by recA-independent mechanisms. However, recA(Prtc) genes, which encode RecA proteins that are constitutively activated (as in SOS response) greatly enhanced excision and transposition and the stimulation of transposition was inhibited by an uncleavable LexA protein. It was suggested that there may be a LexA binding site within the promoter for the IS50 transposase, that activated RecA may cleave the IS50 transposition inhibitor, and that the transposase may be formed by RecA cleavage of a precursor molecule (Kuan et al 1991).
As was mentioned (§ 3.6a), under starvation there is the programmed failure of MMR which stimulates the adaptive mutagenesis.

7.2 The decline of MMR enhances the frequency of transpositions

Among 40 mutants designated tex, which increase the frequency of Tn10 precise excision 3 mutations (texA) have been shown to qualitatively alter RecBC function but 21 tex mutations with a mutator phenotype map to five genes have been identified as components of a MMR: uvrD, mutH, mutL, mutS and dam. So, mutations in MMR genes might enhance transposon excision by a single general mechanism. Alternatively, since mutations in each gene have qualitatively and quantitatively different effects on transposon excision, defects in different MMR genes may enhance excision by different mechanisms (Lundblad and Kleckner 1985). The similar increase in Tn10 transposition frequency was observed in S. typhymurium; uvrB and MutH mutants showed Tn10 frequencies higher than control values. An increase in excision frequency of about 20 or 150 times in 2 different polA mutants, and a smaller increase, of about 2 or 15 times over control values, was detected in mutH and uvrD mutants, respectively (Lorenzo et al 1990). As was mentioned, SOS response can induce excisions of transposons. The opposite situation is true as well.

7.3 Excision of transposons can induce SOS response

Indeed Tn10 transposition induces expression of SOS functions. Lambda prophage induction is increased in lambda lysogens containing increased Tn10 transposase function plus single or multiple copies of an appropriate pair of transposon ends. This increase occurs by the normal pathway for prophage induction, which involves RecA-mediated cleavage of the phage lambda repressor. Tn10 transposes by a nonreplicative mechanism. It seems that the signal for RecA protease activation and SOS induction is generated by degradation of the transposon donor molecule and suggests that SOS induction is biologically important in helping a cell undergoing transposition to repair and/or recover from damage to the transposon donor chromosome (Roberts and Kleckner 1988).

If the excision of transposons could be induced by a stress – does it mean, that being "selfish" they "try to escape" from the troubling cells by hitchhiking into the conjugative plasmid? For example, it is known, that Tn7 prefer to integrate into the conjugative plasmids. Indeed most transposons display target site selectivity, inserting preferentially into sites that contain particular features. And Tn7 possesses the unusual ability to recognise different classes of target sites. Tn7 preferentially inserts into conjugative plasmids. Further Tn7 appears to recognise preferred targets through the conjugation process; this offers Tn7 the ability to spread efficiently through bacterial populations (Wolkow et al 1996). It seems, that under the stresses the best way for transposons is the way into another cell. But "who knows" in which environmental situation an another cell is?

7.4 Excised transposons can induce beneficial mutations in the stressed cell

In fact strains of E. coli carrying Tn10 gain a competitive advantage in chemostat cultures. All Tn10-bearing strains that increase in frequency during competition have a new IS10 insertion that is found in the same location in the genome of those strains. These results show that the IS10-generated insertion increases fitness in chemostat cultures. Transposable elements may speed the rate of evolution by promoting nonhomologous recombination between pre-existing variations within a genome and thereby generating adaptive variation (Chao and McBroom 1985). Also transposable elements could be involved in speciation events by their ability to produce irreversible deleterious mutations that promote escape from evolutionary stasis. A model was developed to investigate the effect of transposon-mediated mutations on the rate of evolution of microorganisms as they compete for resources within an artificial adaptive landscape. In the absence of transposon mutations the seed organisms quickly evolve to occupy the nearest adaptive peak but thereafter evolutionary stasis ensues and adjacent empty peaks are left unoccupied. In the presence of transposon mutations, evolution is again dominated by stasis but is punctuated by bursts of rapid evolution in which consecutive unoccupied adaptive peaks are filled with organisms derived from single transposition events. Rapid evolutionary events leading to founding of new biological species, may be similarly initiated by irreversible deleterious mutations induced by transposition (McFadden and Knowles 1997).

8. Plasmids under stress

Plasmids are dispensable extrachromosomal autonomously replicating genetical elements which being molecular endosymbionts can provide to their hosts a lot of different selective advantages. Conjugative plasmids are able to transfer one strand of their DNA from the donor cells into the recipient, and (after the synthesis of the corresponding complementary strands in both cells) expand themselves in the microbial world (see review Pansegrau and Lanka 1996). What do plasmids do if the host cells undergo the stresses? Do they help the host? Or they prefer not to risk and to escape into another, more lucky cell? Or they force the cell into the unpredictable and risky way of evolution?

8.1 Under the stringent response the plasmids
decrease their copy number

Indeed at least some plasmids evolved to decrease their copy number if the host cell entered into starvation (see review, Wegrzyn 1999). Maybe, the biological sense of such "altruistic behaviour" is to alleviate the metabolic load of the stressed cell and to increase mutual chances to survive. Under the substrates limitation, inhibition of DNA synthesis or amplification of plasmid DNA may depend on the nature of deprived amino acid. While in almost all cases plasmid DNA replication was inhibited during the stringent response irrespective of the nature of deprived amino acid the wild-type or copy-up mini-P1, mini-F and mini-R1 plasmids replicated in ppGpp deficient relA mutant depending on the kind of starvation (Wrobel and Wegrzyn 1997). It was found that ColE1, oriC, lambda plasmid and pSC101 but not RK2 replicons are sensitive to high ppGpp level. This is the direct evidence that replication of most, but not all, replicons is under stringent control (Herman and Wegrzyn 1995). The next action which some plasmids realized under the starvation did not seem very altruistic.

8.2 Under the stresses some plasmids force the host cells to kill the surrounding competitors

8.2a Colicins: These are plasmid coded toxic exoproteins produced by colicinogenic strains of E. coli and some related species of Enterobacteriaceae. They inhibit sensitive bacteria of the same family. Synthesis of colicins is coded by genes located on Col plasmids (see review Smarda and Smajs 1998).

(i) Factors stimulating synthesis of colicins: The synthesis of colicin E1 is known to be stimulated by the SOS response, anaerobiosis, catabolite repression and is induced when cells reached stationary phase (Eraso et al 1996).

(ii) cAMP-CRP: Catabolite repression affects the kinetics of induction and the rate of induced colicin synthesis. The CRP-cAMP complex was found to bind to two sites 5¢ to the colicin cea promoter and activate it (Salles and Weinstock 1989).

(iii) SOS-response: LexA protein which is repressor of SOS regulon, is also repressor of the colicin E1 gene. LexA protein binds with a high affinity to the promoter of this gene which contain two overlapped "SOS boxes" to which the LexA protein binds in a cooperative manner (Ebina et al 1983; Lloubes et al 1986).

(iv) Supercoiling of the plasmid DNA: Anaerobiosis significantly increases expression of the genes for colicins E1, E2, E3, K, and D. A good correlation was observed between the levels of colicin synthesis and plasmid DNA supercoiling and the degree of aeration of the cultures. Thus the regulation of colicin gene expression in res-
ponse to a change in aeration appears to be mediated by environmentally induced variations in DNA supercoiling (Malkhosyan et al 1991).

8.2b Microcins: Like colisins these are plasmid-coded peptides antibiotics. Microcinogenic strains are immune to the action of the microcin they synthesize (see review Moreno et al 1995).

(i) Factors stimulating synthesis of microcins: These are also produced when the cells enter into the stationary phase. For example, microcins B17 and C7 are produced by E. coli when cells enter the stationary phase of growth (del Castillo et al 1990).

(ii) Sigma-S: In the mutant for rpoS, synthesis of E. coli microcin C51 is absent or extremely reduced. In experiments with a cloned promoter of the microcin operon, the Sigma-S was shown to participate in the regulation of transcription of plasmid genes that determine microcin synthesis (Fomenko et al 1997).

Under nitrogen starvation some plasmids decrease their copy number to help the host cell to survive, but under starvation for energy and/or under the action of SOS response some plasmids force the host cells to kill the surrounding competitors.

8.3 Under stress some plasmids can induce
mutagenesis

8.3a Plasmids inducing SOS-response and non-targeted mutagenesis

(i) Plasmids that contain genes involved in SOS response: muc and recA: rep region of the pR plasmid encodes a function which regulates the expression of
the muc genes. These plasmid genes are analogous to chromosomal umu gene and are under the negative control of lexA and responsible for an increased rate of spontaneous mutagenesis and resistance to UV and chemicals (Battaglia et al 1987). Plasmid pKM101, which carries muc genes protected umu deficient strains of E. coli from UV irradiation. Plasmid pGW16, a derivative of pKM101 selected for its increased spontaneous mutator effect, also gave some protection to the UmuC-deficient strain and to increase both spontaneous and UV induced mutations (Little et al 1991). The R-plasmid, pEB017, restored recombination ability to recA strain and conferred enhanced resistance to UV-radiation and enhanced UV-radiation mutability to wild-type, recA and umuC strains of E. coli K12. Plasmid pEB017 also mediated about a
3-fold enhancement of the SOS induction in a recA strain. It was found that pEB017 has a recA-like gene that mediates the enhanced resistance to UV-radiation and enhanced UV-radiation mutability (Obaseiki-Ebor and Smith 1992).

(ii) Plasmid coded methylase can induce SOS response: The plasmid pFM366 from a virulent S.enteritidis strain was found to code for DNA methylase activity. The expression of the cytosine methylase encoded by pFM366 is negatively controlled by the rpoS gene in E. coli and induces the SOS response (Ibanez et al 1997).

(iii) Mutagenic effect of the plasmid could depend
on cAMP: When stationary (!) phase E. coli K12 trp(amber) cells were exposed to UV irradiation it was found that the recovery of any of the induced Trp⁺ revertants was possible only after the irradiated cultures were first supplied with the Muc⁺ mutation-enhancing IncP plasmid pKM101 (by conjugation). A number of UV-induced Trp⁺ revertants recovered from pKM101⁺ cultures varied quite dramatically depending upon which carbon sources were present in the post-irradiation plating medium. At least one component of the mutational pathway which was shown to operate in UV-irradiated pKM101-containing cells was extremely sensitive to classical cAMP-mediated catabolite repression and thus, is cAMP-CRP dependent (Ambrose and McPhee 1998). On the role of cAMP in mutagenesis, see § 5.2b.

In general, some plasmids contain genes analogous to that of SOS regulon and it seems, these genes can provide to the host cells the additional defence against mutagenes. Is it logical to speculate that under the SOS response inducing stresses these genes will enhance the untargeted mutagenesis? Another possibility for the plasmids to increase the mutagenesis of the host cell is the decreasing of the efficiency of MMR.

8.3b Mismatched base pairs of retrons increase mutagenesis by the declining of MMR: Retrons are genetic elements that encode multicopy single-stranded DNAs called msDNAs. They are clonally distributed in E. coli and retrons in different clones produce msDNAs with different nucleotide sequences. msDNAs consist of a RNA molecule covalently linked to a single-stranded DNA molecule. The latter contains an inverted repeat, resulting in a stem-loop structure. All known retrons except Ec78, have one or more mismatched base pairs in the stem-loop structure. Two retrons, Ec86 and Ec83, when present in high copy numbers are mutagenic. The ratios of mutation frequencies were similar to the ratios observed for a mutant defective in MMR. It is known that some proteins required for MMR bind to mismatched base pairs prior to carrying out repair. The similarity in the mutation frequency ratios suggested that the mutagenesis caused by msDNAs of retrons Ec86 and Ec83 might be due to sequestration of a mismatch repair protein by msDNA (Maas et al 1994). The ppGpp was found to be a positive regulator of retron-Ec107 expression. Its presence is required for starvation-induced transcription of retron-Ec107 and multicopy single-stranded DNA production. It was also found that expression from the retron promoter is independent of the Sigma-S (Herzer 1996). Under starvation the increased copy number of the msDNAs can enhance the mutagenesis of the host cells.

8.3c Single stranded regions in plasmid DNA can induce SOS response

(i) Aberrant initiation of the plasmid DNA replication can induce SOS response: MiniF, a 9·3 kb fragment of the F plasmid, carries genes necessary for its replication and partition as well as for the expression of an SOS signal. The arrest of replication of mini F induced SOS functions. Also, the plasmid determined SOS induction was increased greatly near the stationary phase. During host cell exponential growth, mini F plasmids were lost rapidly, although SOS induction persisted for several cell generations (Sommer et al 1985). P1 Km – the smallest of miniplasmids derived from plasmid P1 can induce the SOS pathway of the cell as shown by increased expression of the recA operon, spontaneous induction of lambda etc. This induction was caused by an aberrant initiation of DNA replication (Capage and Scott 1983).

(ii) UV irradiated plasmids can induce SOS response in the recipient cells: F plasmid and its mini-derivatives efficiently induced cellular SOS genes when they were damaged by UV irradiation and then introduced into a recipient bacterium. To generate an SOS signal, UV light-damaged mini F required bacterial RecBC enzyme. In contrast, UV light-damaged F plasmid produced an SOS signal independently of the activity of the RecBC enzyme. These findings are consistent with a picture in which the SOS signal is constituted by stretches of single-stranded DNA on a replicon (Bailone et al 1985). The pR bat gene is essential for plasmid replication and for spontaneous induction of the SOS response. Mutations preventing single-stranded DNA production, needed for pR plasmid replication, also prevent the induction of the SOS system (Gigliani et al 1993). The plasmids determined SOS-inducing signal appears to be the single-stranded DNA produced during plasmid replication. But does this signal really induce SOS response in natural conditions? Some plasmids have the special genes to prevent this.

8.4 Plasmids that inhibit SOS-response in the
recipient cells: Psi function

During the course of evolution, the Psi (plasmid SOS inhibition) function has been selected in some conjugative plasmids which permit them to transfer single-stranded DNA without generating an SOS signal. In the plasmid R6-5 Psi function is expressed by psiB, a gene located on oriT, the origin of conjugal transfer. PsiB is a coding polypeptide which is transiently expressed by a wild-type F sex factor during its transmission to a recipient. In a F⁺ × F^– cross, PsiB concentration increased at least 10-fold in F recipients after 90 min and declined thereafter; the psiB gene may be repressed when F plasmid replicates vegetatively. When overproduced, PsiB may interfere with RecA protein at chromosomal single-stranded DNA sites generated by discontinuous DNA replication, thus causing SOS inhibitory phenotype. Overproduction of PsiB protein sensitizes the host cell to UV irradiation. It seems that PsiB polypeptide has an anti-SOS action by inhibiting activation of RecA protein, thus preventing the occurrence of LexA controlled functions (Bailone et al 1988; Bagdasarian et al 1992). Psi function is widely distributed in the plasmid world, for example, in 9 of 20 conjugative plasmids of different incompatibility groups, including F and R100 (or R6-5), which contain the two sequences which are homologous to the gene psiB (Golub et al 1988). The Incl1 conjugative plasmid Collb-P9 also carries a psiB gene (Jones et al 1992).

In general, the plasmids PsiB proteins prevent SOS dependent mutagenesis and intra-chromosomal recombination but not recombination following conjugation. Under stress plasmids perform a number of diverse activities to help the host cell to survive and, it seems, to evolve an ability to overcome the adverse environmental conditions.

9. Conjugation under stress

Although rich nutritional conditions stimulate plasmid transfer, a conjugation can take place within a wide range of conditions, even in the absence of nutrients and at low temperatures.

9.1 Factors affecting conjugative transfer

9.1a Density factor and donor-to-recipient ratio: The density of parent cells affects the number of transconjugants, reaching a maximum when the cell density is on the order of 10⁸ CFU per ml. As the donor-to-recipient ratios varied from 10^–4 to 10⁴, the number of transconjugants varied significantly, reaching a maximum with donor-to-recipient ratios between 1 and 10 (Fernandez-Astorga et al 1992).

9.1b Rich media stimulates conjugation: The concentration of total organic carbon in the mating medium affects both the number of transconjugants and the transfer frequency, being significantly higher when the total organic carbon concentration is higher than 1,139 mg of C per litre. However the transconjugants were detected even with less than 1 mg of C per litre. Neither the transfer frequency nor the transconjugant number varied significantly in the range of pHs assayed (Fernandez-Astorga et al 1992). In batch mating experiments with Pseudomonas putida PAW1(TOL) as a donor and P. aeruginosa PAO 1162 as a recipient strain it was shown that the impact of the substrate concentration in the mating medium was highly correlated with the growth history of the donor strain. When the donor strain was from the exponential phase, no impact was observed; when the donor strain was from the stationary phase, however, a strong impact of the substrate concentration was measured: a 10-fold reduction in the substrate concentration decreased the observed plasmid transfer rate only by 55% (Smets et al 1995).

Although conjugative plasmids "prefer" to transfer when the host cells live in rich nutritional conditions, they can also provide a transfer in poor living conditions as well. But can they transfer into non-growing cells? It seems, yes.

9.1c Conjugative transfer into VNC cells: The presence of viable, other than culturable, transconjugants was demonstrated in matings with parental cells of E. coli as well as with recipient cells from survival in river water (under illuminated and non-illuminated systems). In matings with a recipient strain from illuminated systems, culturable transconjugants were not detected after the third day of recipient cell survival. In spite of this, viable transconjugants were detected in numbers that exceeded 10⁵ cells per ml. These results clearly show that a fraction of nonculturable recipient cells is able to receive and express plasmids by conjugation processes and form viable but non-culturable transconjugant cells (Arana et al 1997).

In general the plasmids can conjugate in the wide spectrum of environmental conditions. One of the most strong barriers of microbial gene transfer is the restriction-modification defence of the recipient cells.

9.2 Plasmids can repress the restriction-modification systems of the recipient cells

9.2a ArdA and ArdB – the plasmids coded antirestriction proteins: These proteins are induced during the conjugative transfer. The genes promoting the alleviation of restriction were termed ard (alleviation of restriction of DNA). The ArdA and ArdB genes alleviate DNA restriction by type I and type II restriction endonucleases and promote conjugative transmission of the unmodified plasmid to a restricting host. The IncN plasmid pKM101 (a derivative of R46) and the IncI1 plasmid ColIb-P9 carry genes ardA and ardB. Comparison of the amino acid sequences of the antirestriction proteins of the unrelated plasmids pKM101 and ColIb revealed that these proteins have about 60% identity. Like ColIb Ard, pKM101 ArdA specifically inhibits both the restriction and modification activities of five type I systems of E. coli tested and does not influence type III (EcoP1) restriction or the 5-methylcytosine-specific restriction systems McrA and McrB. However, in contrast to ColIb Ard, pKM101 ArdA is effective against the type II enzyme EcoRI. Like ArdA, ArdB efficiently inhibits restriction by members of the three known families of type I systems of E. coli and only slightly affects the type II enzyme, EcoRI. In contrast to ArdA, ArdB is ineffective against the modification activity of the type I (EcoK) system. It seems that both types
of antirestriction proteins may play a pivotal role in overcoming the host restriction barrier by self-transmissible broad-host-range plasmids (Belogurov et al 1992, 1993). To clarify the ecological role of ardA gene of I1 ColIb-P9 plasmid, its distribution was determined on plasmids from 23 incompatibility groups. The genes with at least 60% identity with ardA, were detected on plasmids belonging to the I complex (IncB, I1 and K), the F complex (IncFV) and the IncN group. The ardA homologues were found to specify an antirestriction phenotype which was enhanced by genetic depression of the plasmid transfer system. ArdA loci map in plasmid leading regions (Chilley and Wilkins 1995).

9.2b How plasmids regulate their antirestriction genes: Activity of antirestriction genes ardA and ardB, coded by IncN plasmid pKM 101 could be controlled by the early stages of mating and by SOS response. Indeed the promoter regions of the ard genes contain the consensus of SOS box, to which repressor LexA can bind (Del’ver et al 1998).

According to their offensive strategy, conjugative plasmids are well armed by antirestriction systems destroying the anti-plasmid defence of the recipient cells. But what about the cells? In which direction have they evolved in dealing with the permanent menace of plasmids invasion? Maybe, they have acquired genes of the antirestriction defence?

9.3 Under stress the cells alleviate their restriction-modification systems and welcome plasmids

This phenomenon is called restriction alleviation (RA), it is widely distributed and could be induced by different stresses.

9.3a Heat shock, pH shift, organic solvents induce the RA: Corynebacterial recipient cells exposed to heat, organic solvents, pH shifts, or detergents show an increased fertility in subsequent interspecific matings with E. coli. This effect is independent of de novo protein biosynthesis and seems to be due to a direct inactivation of a restriction system active against foreign DNA that enters the cell by IncP-mediated conjugation (Schafer et al 1994). RP4-mediated transfer of mobilizable plasmids in intergeneric conjugation of E. coli donors with Corynebacterium glutamicum ATCC 13032 is severely affected by a restriction system in the recipient that can be inactivated by a variety of exogenous stress factors (Schafer et al 1994).

9.3b SOS response induces the RA: UV-induced RA is a SOS function which partially relieves the K-12-specific DNA restriction in E. coli. RA is determined by observing elevated survival of unmodified phage lambda in cells irradiated with UV light prior to infection (Day 1977; Thoms and Wackernagel 1982). UV and gamma-radiation were shown to induce in RA of unmodified T3 and T7 phages (Torosian et al 1987). The activity of the EcoK DNA restriction system of E. coli can be alleviated in wild-type cells, by UV irradiation and expression of the SOS response, so that 10³- to 10⁴-fold increases in lambda phage growth and four-fold increases in pBR322 plasmid transformation occurred with unmodified DNA. RA was found to be a transient (!) effect because induced cells, which initially failed to restrict unmodified plasmid DNA, later restricted unmodified phage lambda (Hiom and Sedgwick 1992).

9.3c Genes of SOS response involved in RA: recA, lexA, recBC, recF, umuC, uvrA: A partial release of K-specific restriction of phage lambda grown in E. coli C was observed when E. coli K strains uvrA⁺ and uvrA^– were irradiated with UV light before infection. The effect occurred in uvrA^– strain at lower UV fluences than it did in uvrA⁺. Little or no release of restriction was observed in recA^– or lexA^– mutants (Thoms and Wackernagel 1984). Induction of this RA depends also on the recBC enzyme and on the recF function (Thoms and Wackernagel 1982, 1983, 1984). Alleviation of EcoK DNA restriction system in UV irradiated wild-type cells is umu dependent (Hiom et al 1991).

9.3d On the diversity of the RA Systems: A new form of RA was demonstrated for phage induced by UV light from dcm strains of E. coli K-12. EcoRII restriction of the induced phage is alleviated, which was the first report of type II RA. Unlike previously reported RA, the increase in phage-plating efficiency was not dependent upon irradiation of the plating host for its induction (Radnedge and Pinney 1991). In the study of the response to UV treatment of the three endogenous modification dependent restriction systems of K-12, McrA, McrBC, and Mrr it was shown that all these resident restriction systems displayed reduced activity following UV treatment, but not in a unified fashion; each response was genetically and physiologically distinct (Kelleher and Raleigh 1994).

In general, under the stress the microbial cells alleviate their restriction systems and welcome the conjugative plasmids to increase the mutual chances of overcoming the stressful conditions. Conjugative plasmids can transfer themselves during the wide spectrum of nutritional conditions and even into non-growing cells. During the transfer they can repress restriction-modification systems of the host cells and induce the SOS response in them. Are there any other special tricks which the conjugative plasmids can demonstrate in severe living conditions?

9.4 Starvation forces the plasmid cells to go in for
homosexual gene transfer

Surface exclusion is the mechanism by which F plasmids prevent the redundant entry of additional F plasmids into the host cell during exponential growth. This mechanism is relaxed during stationary phase and nonlethal selections, allowing homosexual redundant plasmid transfer. Homosexual redundant transfer occurs in stationary-phase liquid cultures, within non-growing populations on solid media, and on media lacking a carbon source. Using genetically marked F¢ plasmids and host strains it was shown that a high level of redundant transfer occurs between these non-growing cells during nonlethal selection (Peters and Benson 1995). A recA null mutation reduced redundant transfer. Conversely, a recD null mutation increased redundant transfer. These results suggest that Rec proteins play an active role in redundant transfer and/or that redundant transfer is regulated with the activation of recombination (Peters et al 1996). In general redundant homosexual plasmid transfer during a period of stress may represent a genetic response that facilitates evolution of plasmid-encoded functions through mutation, recombination, reassortment, and dissemination of genetic elements present in the populations. So under starvation the barriers preventing the homosexual gene transfer could be violated. The barriers preventing inter-species gene transfer could be violated also.

10. Inter-species gene transfer under stress

The barriers to chromosomal gene transfer between bacterial species provide their genetic isolation. These barriers, such as different microhabitats, the host ranges of plasmids and restriction-modification systems, limit gene exchange, but the major limitation is genomic sequence divergence. Three main mechanisms control the inter-species recombination: (i) active MMR prevent recombination between diverged DNAs, inactivation of MMR permits it, (ii) activation of SOS response stimulates inter-species recombination and (iii) natural selection determines the effective recombination frequencies (see review, Matic et al 1996). The ability of related DNAs to undergo recombination decreases with increased sequence divergence. The extent of genetic isolation between enterobacteria is a simple mathematical function of DNA sequence divergence. The function does not depend on hybrid DNA stability, but rather on the number of blocks of sequences identical in the two mating partners and sufficiently large to allow the initiation of recombination. Further, there is no obvious discontinuity in the function that could be used to define a level of divergence for distinguishing species (Vulic et al 1997).

10.1 MMR and SOS response control recombination between the diverged genomes

According to current understanding, genetic barriers can be established, eliminated, or modified by manipulating the systems which control genetic recombination, mismatch repair and SOS response. Recombinating chromosomes interact by means of proteins involved in recombination and in the recognition and repair of mismatched base pairs. Recombination proteins bring homologous chromosomes or chromosomal regions together by facilitating the search for DNA homology and
by catalysing strand exchange between homologous molecules or regions. Mismatch recognition and repair proteins act as editors of recombination and appear to disrupt those DNA associations that contain mismatched base pairs.

10.1a In growing cells the activity of MMR prevents the inter-species recombination: Although the ability of related DNAs to undergo recombination decreases with increased sequence divergence, mutS and mutL mutations confer large increases in recombination between sequences that are divergent by several per cent at the nucleotide level, an effect attributed to a role for products of these genes in control of recombination fidelity. MutS and MutL are proteins involved in the earliest steps of mismatch repair, including mismatch recognition by MutS. Indeed MutS abolishes RecA-catalysed strand transfer between fd and M13 bacteriophage DNAs, which vary by 3% at the nucleotide level. MutL alone has no effect on M13-fd heteroduplex formation, although the protein dramatically enhances the inhibition of strand transfer mediated by MutS. In general, MutS and MutL proteins block branch migration, presumably in response to occurrence of mispairs within newly formed heteroduplex (Worth et al 1994). The detailed genetic analysis of inter-species recombination in E. coli between the linear Hfr DNA from S. typhimurium and the circular recipient chromosome reveals that MutS and MutL mismatch binding proteins prevent homologous recombination between these 16% diverged genomes by at least two distinct mechanisms:
(i) MutH independent mechanisms presumably act by aborting the initiated recombination through the UvrD helicase activity. RecBCD nuclease might contribute to this editing step, presumably by preventing reiterated initiations of recombination at a given locus; (ii) MutH dependent mechanism requires unmethylated GATC sequences, and probably corresponds to an incomplete long-patch mismatch repair process that does not depend on UvrD helicase activity (Stambuk and Radman 1998).

MMR system maintains the mechanisms preventing inter-species recombination. Are there any circumstances in which the cell "prefer" to drop the genetical barrier as it drops its restriction-modification defence under the stresses?

10.1b In the stationary phase MMR is not active: The cellular amounts of MutS and MutH decrease by as much as 10-fold in stationary-phase cultures. In bacteria growing exponentially in enriched minimal salts-glucose medium, about 113 MutL dimers, 186 MutS dimers, and 135 MutH monomers are present per cell. Calculations with the in vitro dissociation constants of MutS binding to different mismatches suggested that MutS is not present in excess, and may be nearly limiting in some cases, for MMR repair in exponentially growing cells. Remarkably, when bacteria entered late stationary phase or were deprived of a utilizable carbon source for several days, the cellular amount of MutS dropped at least 10-fold and became barely detectable. In contrast, the amount of MutH dropped only about 3-fold and the amount of MutL remained essentially constant in late stationary-phase and carbon-starved cells compared with those in exponentially growing bacteria. The amounts of mutS, mutH, and mutL transcripts decreased to undetectable levels in late-stationary-phase cells. In general, the MMR repair capacity is repressed in nutritionally stressed bacteria (Feng
et al 1996; see also § 3.6, 6.3a, 6.1g).

10.1c SOS response: Analysis of inter-species matings between S. typhimurium and E. coli indicated that MMR enzymes act as potent inhibitors of inter-species recombination, whereas the SOS system acts as an inducible positive regulator. Inter-species mating triggers a RecBC-dependent SOS response in female cells that increases recombination mainly through overproduction of the RecA protein. Mismatch repair acts to reduce the mutation rate and recombination between similar sequences, whereas SOS acts to increase both. Indeed while during the inter-species mating between E. coli and S.thyphimurium it was shown, that while inter-species recombination was reduced in the lexA1(Ind^–) strains (with non-inducible SOS response) by 15-fold, intra-species recombination was reduced only by 10-fold. This suggested that some inducible LexA-regulated functions are rate limiting for interspecies recombination. RecA protein alone can restore inter-species recombination proficiency of LexA(Ind^–) strain. The effect of SOS induction on inter-species recombination is much stronger in the mutS, than in mut⁺ strains. Whereas intra-species conjugation induce the SOS response only weakly, a strong SOS induction was observed during interspecies conjugation in 2% of mated recipient bacteria. Clones that had undergone such strong SOS induction had a 20-fold higher inter-species recombination frequency that the noninduced ones. At 60 min after the interruption of conjugation, recA gene induction was 3- to 4-fold higher in inter-species than intra-species conjugation. The observed difference in the level of SOS induction between inter- and intra-species conjugation is RecBC dependent and RecD and recF independent. It seems the signal for SOS response induced by interspecies mating is caused by single stranded stretches of DNA generated by the RecBC helicase action (Matic et al 1995; see also § 8.3c).

The opposing activities of SOS and MMR systems provide a potential for the fine tuning of both the rates of sequences diversification and the extent of genetic isolation, i.e., speciation. A burst of speciation events, perhaps similar to punctuated equilibrium can be understood now on the molecular level. After the period of the environmentally induced high mutation rates caused by the MMR deficiency and by SOS response activation through environmental factors, the return to MMR proficiency and repressed SOS is expected to lead to an immediate establishment of multiple genetic barriers within large highly polymorphic populations (Matic et al 1995; Taddei et al 1997e).

11. Conclusions

In fact all the conclusions which could be drawn from this bird’s eye review may be only preliminary. This new interdisciplinary field (which could be called molecular environmental genetics of microorganisms) is developing very fast and the new data are accumulating exponentially. The practical need promoting this development is due to the problem of the biosafety of the forthcoming releases of genetically modified microorganisms into the biosphere (Velkov 1996). The scientific interest forcing this development is the desire to understand the mechanisms of the evolution of microorganisms in real ecosystems. Indeed what could be said now, according to the current understanding?

(i) In stressful environments which are non favourable for growth, static microorganisms rearrange their metabolism and increase the rate of random mutagenesis. If any mutation(s) resume the growth, the rate of mutagenesis will return to normal. If the cells with the increased rate of mutagenesis will not start growth – they could die because of the accumulation of deleterious mutations. That is why only part (subpopulations) of non-growing cells undergo the increased rate of mutagenesis. The remaining part "prefer" not to risk and to wait for good times.

(ii) In non-favourable environments the stressed cells alleviate the efficiency of their restriction-modification systems which results in the increased chances of accepting in foreign genes, which, probably, can increase the chances to start the growth. If the growth will be resumed – the efficiency of the restriction modification systems will be restored.

(iii) In stressful environments the cells decrease the barrier of inter-species gene transfer which can result in the transient burst of speciation. If after the inter-species recombination, the non-dividing cells resume their growth – the barriers of genetical isolation will be restored.

And the final question. Is the evolution of microorganisms really directed by His Majesty The Blind Watchmaker? It seems, yes. But with one important addition. In adverse environmental conditions it is controlled by microbes – when they "deliberately" accelerate the revolution of the Russian roulette of genetical variability.

Acknowledgements

I apologize to many authors for not citing them owing to limitation of space. I appreciate the assistance given by Vidyanand Nanjundiah in editing the manuscript. Thanks are due to Andrea Rohmert for considerable assistance.

MS received 5 May 1999; accepted 15 September 1999