Polymorphisms in rpoS are common in Escherichia coli. rpoS status influences a trade-off between nutrition and stress resistance and hence fitness across different environments. To analyze the selective pressures acting on rpoS, measurement of glucose transport rates in rpoS+ and rpoS bacteria was used to estimate the role of Fnc, the fitness gain due to improved nutrient uptake, in the emergence of rpoS mutations in nutrient-limited chemostat cultures. Chemostats with set atmospheres, temperatures, pH's, antibiotics, and levels of osmotic stress were followed. Fnc was reduced under anaerobiosis, high osmolarity, and with chloramphenicol, consistent with a reduced rate of rpoS enrichment in these conditions. Fnc remained high, however, with alkaline pH and low temperature but rpoS sweeps were diminished. Under these conditions, Fsp, the fitness reduction due to lowered stress protection, became significant. We also estimated whether the fitness need for the gene was related to its regulation. No consistent pattern emerged between the level of RpoS and the loss of rpoS function in particular environments. This dissection allows an unprecedented view of the genotype-by-environment interactions controlling a mutational sweep and shows that both Fnc and Fsp are influenced by individual stresses and that additional factors contribute to selection pressure in some environments.
MUTATIONS sweep through bacterial populations during periodic selection events (Atwood et al. 1951; Notley-McRobb and Ferenci 2000). The beneficial mutations enriched in an environment are seldom identified, even in experimental evolution systems studied in detail (Elena and Lenski 2003). Recently, some of the genes affected in Escherichia coli populations subjected to prolonged periods of nutrient limitation have been described and the earliest mutation sweeping glucose-limited populations is in rpoS (Notley-McRobb et al. 2002, 2003; Zinser and Kolter 2004). This gene is of particular interest because it encodes the σ-subunit σS, the master regulator of the general stress response, which positively regulates the expression of hundreds of genes involved in resistance to various environmental stresses (Hengge-Aronis 2002; Patten et al. 2004; Weber et al. 2005). RpoS is elevated under stressful conditions in addition to the RpoD σ-factor (σ70), which is responsible for the expression of housekeeping and growth-related genes (Jishage and Ishihama 1995). A second reason why the experimental evolution of rpoS mutations is of interest is because natural populations of E. coli and Salmonella contain a high frequency of polymorphisms in this gene (Ferenci 2003) and some bacterial strains have a higher propensity for accumulating rpoS mutations than others (King et al. 2004; Ferenci 2005). Using an experimental system in which the acquisition of rpoS mutations can be reproducibly studied, this study reports on the impact of variations in the bacterial environment on the mutational sweeps leading to enrichment of rpoS mutations.
A hypothesis to explain polymorphism in rpoS has been formulated (Ferenci 2003). The enrichment of rpoS mutations was proposed to be due to the competition between the σ-factors RpoS and RpoD (Zhou and Gross 1992; Farewell et al. 1998; Jishage and Ishihama 1999), which reduces the fitness of bacteria in some situations. rpoS mutations accumulate in stationary-phase batch cultures (Zambrano et al. 1993) and in steady-state glucose-limited chemostat populations (Notley-McRobb et al. 2002) of E. coli largely to alleviate this competition and to improve nutrient scavenging by increasing expression of RpoD-dependent genes (Ferenci 2003). Consistent with this notion, E. coli strains with high endogenous RpoS levels accumulated rpoS mutations at a higher rate during growth under nutrient limitation and disruption of rpoS indeed improves nutritional capability with many poor substrates of E. coli (King et al. 2004). Of course, the trade-off in losing RpoS function is that rpoS mutants exhibit reduced resistance to encountered stresses such as prolonged starvation, high pressure, high osmolarity, and low pH (Cheville et al. 1996; Waterman and Small 1996; Robey et al. 2001; Dodd and Aldsworth 2002). Survival in the gastrointestinal tract is reduced (Price et al. 2000) and rpoS mutants also exhibit a reduced virulence potential in organisms in which virulence gene expression is under RpoS control (Wiedmann et al. 1998; Nadon et al. 2002). The balance between self preservation and nutritional competence balancing (Ferenci 2005) hence represents an interesting example of antagonistic pleiotropy (rpoS mutations being beneficial in some settings but not in others).
On the basis of the above hypothesis, the rate at which rpoS mutations sweep populations in a particular environment is predicted to depend on two factors. The first of these is Fnc, the fitness gain improving nutrient uptake, nutritional competence, and growth. The second input is Fsp, the fitness loss due to RpoS involvement in stress resistance in the same environment. Under aerobic glucose limitation at neutral pH and 37°, when no stress protection is required, Fnc ≫Fsp and the rate of takeover by rpoS mutations, SrpoS, is high. The type of rpoS allele enriched will also influence the magnitude of Fnc and Fsp changes in a sweep. Generally, null mutants are enriched but attenuation or partial loss of function was found in some experiments (Zambrano et al. 1993; Notley-McRobb et al. 2002) so the Fnc and Fsp changes are potentially less than with null mutants. By testing the rate of rpoS mutation accumulation and type of mutation, as well as by estimating Fnc and Fsp, the aim of this study was to dissect the selective pressures on bacteria in various environments and to define the effects on rpoS sweeps in E. coli populations.
The magnitude of both Fnc and Fsp is expected to be sensitive to the surroundings as Fnc depends on RpoS accumulation (high with high RpoS levels and a strong negative effect on RpoD-dependent expression—to zero if no RpoS is expressed). Fsp is also likely to be environment specific, as the contribution of RpoS in resistance to different stresses is unlikely to be constant. We can measure the rate of displacement of wild-type bacteria under nutrient limitation with and without additional stresses and assume that the rate of rpoS sweeps, SrpoS, will be dependent on (Fnc–Fsp). A direct estimation of Fnc is possible with bacteria in a chemostat culture because in an environment with limiting glucose, transport of limiting nutrient determines fitness. Hence the rpoS influence on Fnc is measurable from a comparison of glucose uptake rates in bacteria with and without intact rpoS, comparing isogenic strains differing only in rpoS. As demonstrated earlier, the inactivation of RpoS and the removal of σ-factor competition increases expression of all genes involved in glucose uptake, which are RpoD dependent (Notley-McRobb et al. 2002; Seeto et al. 2004). By measuring glucose transport in various modified environments in rpoS and rpoS+ bacteria, the repressive effect can be revealed and the positive selection pressure for loss of rpoS quantitated.
Measuring Fsp directly is more problematic, but can be approximated by difference if both SrpoS and Fnc are known. Also, an indirect indication of Fsp may be obtained by measuring the extent of induction of rpoS under stress conditions. This of course assumes that the magnitude of RpoS levels in an environment is indicative of the need for RpoS in stress resistance. We therefore measured the extent of RpoS accumulation under all the environmental conditions used to test the correlation between presence and need. Several environmental signals and reduced growth rate are known to induce high levels of RpoS protein (Gentry et al. 1993; Lange and Hengge-Aronis 1994; Jishage and Ishihama 1995; Bearson et al. 1996; Jishage et al. 1996; Muffler et al. 1996, 1997; Sledjeski et al. 1996). However, a comparison of the extent of induction by different environmental stimuli is impossible from published data, given the differences in culture media, growth rates, and growth phases used in earlier studies of individual stresses. Here we use a single strain subject to constant growth rates in glucose-limited chemostats to avoid many of these variables. It should be noted, however, that glucose limitation itself results in elevated RpoS levels, so the combined environmental effects of both limitation and additional stress are being studied together, with the possible regulatory complications that this entails.
In summary, by studying the rate of rpoS mutation accumulation in defined environments as well as Fnc and Fsp, it is feasible to study genotype-by-environment influences on a mutational sweep in more detail than is available for any other gene or experimental evolution system. Of course, the Fnc/Fsp model may be an oversimplification and miss other influences of the environment (such as changed mutation rates, the protein synthesis load in adapting to environments, unknown effects in regulation, and σ-factor competition), but estimating Fnc and Fsp and their influence on rpoS sweeps should identify deviation from the simple model and the involvement of other factors in specific environmental situations. Overall, the results illustrate the complex influence of antagonistic pleiotropy in determining evolutionary choices in different environments and the magnitude of Fnc as a major factor in rpoS sweeps.
MATERIALS AND METHODS
All strains were E. coli K-12 derivatives. BW2952 is an MC4100 derivative [genotype, F− araD139 Δ(argF-lac)U169 rpsL150 deoC1 relA1 thiA ptsF25 flbB5301 rbsR malG∷λplacMu55 Φ(malG∷lacZ)] (King et al. 2004). RO91 contains a translational fusion that harbors the turnover element susceptible to proteolysis [genotype, MC4100 (λRZ5:rpoS742∷lacZ)hybr] (Muffler et al. 1996). Strain BW3709 has an rpoS∷Tn10 transposon insertion in the BW2952 background (Notley-McRobb et al. 2002). The mgl-constitutive strain BW3245 (with a nonsense mutation in mglD) and its rpoS∷Tn5 transductant BW3522 were also constructed in the MC4100 background (Notley-McRobb et al. 2002).
Growth media and culture conditions:
The basal medium used in all chemostats was minimal medium A (MMA; Miller 1972). The medium pH was adjusted to 7 in standard cultures but to 8.5, 8.0, 6.0, or 5.5 in pH stress experiments by adjusting the ratio of K2HPO4 to KH2PO4. The carbon source in all cases (unless specified) was glucose, which was present at 0.02% w/v in the feed medium. Standard Luria broth (LB) (Miller 1972) was 25-fold diluted into the MMA medium to act as a limiting carbon source. Where specified, chloramphenicol was added to the feed medium at a concentration of either 0.15 μg ml−1 or 0.3 μg ml−1. Sucrose was added to chemostats at a final concentration of 5% w/v. Eighty-milliliter chemostat cultures were set up as described previously (Notley-McRobb et al. 2002) and were run at 37° unless specified. A gas flow rate of 40 cc min−1 was used to flush chemostats with 80% CO2:20% O2 and 100% N2 (BOC Gases Australia, Sydney) where specified. All chemostats were operated at a dilution rate of 0.1 hr−1. The culture densities were between 1.9 × 108 and 2.1 × 108 ml−1. The pH and culture densities of each chemostat were checked after 4 days and in every case were found to be stable over the course of the experiments. Batch cultures were grown for 24 hr at 37° in MMA supplemented with 0.02% glucose for stationary phase and diluted into the same medium for exponential cultures.
Detection of rpoS mutants:
rpoS partial and null mutants were distinguished from the wild type by staining colonies on Luria agar plates. Plates were incubated for 24 hr at 37° and then left at 4° for 24 hr before being flooded with concentrated iodine. Iodine staining is dependent on glycogen, whose synthesis is affected by rpoS status (Hengge-Aronis and Fischer 1992). Dark-brown colonies were wild type, while pale brown or white colonies indicated partial or null mutants with different levels of glycogen. Examples of pale or white colonies from day 4 chemostat samples were sequenced in the rpoS gene to confirm mutations as previously described (Notley-McRobb et al. 2002).
The initial rate of uptake of 0.5 μm d-[U-14C]-glucose by chemostat-grown bacteria was determined using bacteria directly removed from chemostats, the OD580 measured and the transport assay conducted as described previously (Death and Ferenci 1993). The rate of transport was measured in units of picomoles of sugar transported min−1 10−8 of bacteria.
Five-milliliter samples from chemostats were removed, and β-galactosidase activity was measured as described by Miller (1972) by using sodium dodecyl sulfate- and chloroform-treated cells.
Environmental conditions and enrichment of rpoS mutants:
The rate of acquisition of rpoS mutations was monitored in nutrient-limited chemostat cultures of an E. coli K-12 strain exposed to a variety of suboptimal environmental settings. The genotype of the strain used itself has a major influence on the scale of rpoS sweeps (King et al. 2004; Ferenci 2005) so all strains used here had the same genetic background. On top of nutrient limitation, the secondary environmental stresses chosen for this study included exposure to a range of temperatures (25°–44°), pH stress (5.5–8.5), osmotic stress (nonutilizable sucrose), exposure to sublethal antibiotic (chloramphenicol) concentrations, modified atmospheres (80% CO2:20% O2 and 100% N2), and variation of the nutrients in the growth media. The environments chosen were not lethal or extreme enough to affect viability (as measured by viable counts), growth, and steady-state population size (∼2 × 1010) in the chemostat at the dilution rate of 0.1 hr−1 (doubling time, 6.9 hr) used in the experiments.
When inoculated with wild-type E. coli, glucose-limited chemostat cultures with optimal settings (37°, pH 7, aerobic, no inhibitor) displayed a rapid takeover by rpoS mutants and elimination of rpoS+ bacteria [Figure 1 “control” and Notley-McRobb et al. (2002)]. By the fourth day of culture, wild-type rpoS+ bacteria were a small minority of the population. Mutations in rpoS were followed in a phenotypic screen using glycogen staining (Lange and Hengge-Aronis 1991; Notley-McRobb et al. 2002). In some environments, two levels of rpoS defect were detectable; most mutations abolished staining but particular selections resulted in partial staining, as noted in Figure 1. As confirmation, examples of isolates with altered staining properties were analyzed for DNA sequence changes in rpoS as shown in Table 1. In each case tested, the changed iodine-staining phenotype was the result of a mutation in the rpoS gene.
A comparison of rpoS sweeps in various environments:
The nature of the media and secondary stresses affected the kinetics of mutation accumulation in rpoS. In exponential cultures growing with excess glucose, rpoS mutants are not enriched (Notley-McRobb et al. 2002) and SrpoS is zero. Interestingly, growth limited by low concentrations of complex media resulted in an even more rapid acquisition of rpoS mutations than growth under glucose limitation at the same growth rate. rpoS mutants were already present within 24 hr of limitation with LB medium (Figure 1). In results not shown, mutations in rpoS were also rapidly enriched under other limitation conditions, including with casamino acids, in minimal media with carbon sources other than glucose such as maltose, as well as under ammonia limitation (Notley-McRobb et al. 2002). Given the generality of rpoS loss in nutrient-limited populations, secondary environmental stresses were applied only to glucose-limited cultures as representative of the situation with nutrient limitation.
With prolonged culture, rpoS mutations accumulated at every pH used (Figure 1). Consistent with this study, Farrell and Finkel (2003) reported that rpoS mutations accumulate under acidic, neutral, and basic conditions in stationary-phase batch cultures of E. coli. From a comparison of the rate of enrichment of rpoS mutants in chemostats or SrpoS, the overall selection pressure for the acquisition of rpoS mutations was highest at neutral pH, followed by acidic pH, and lowest under alkaline pH. This is consistent with a postulated role of rpoS in alkaline resistance (Small et al. 1994). Attenuated rpoS mutants with a partial phenotype were found at alkaline pH as well as at pH 5.5 and 6 shown earlier (Notley-McRobb et al. 2002). Partial mutants provide a compromise solution in maintaining some resistance to acid or alkaline pH, while reducing the σ-factor competition under nutrient limitation, suggesting that Fsp is significant under these conditions. The partial mutants are similar to attenuated mutants found under prolonged stationary-phase selections (Finkel et al. 2000) and have intermediate increases in general stress sensitivity (Notley-McRobb et al. 2002).
Mutations in rpoS became common in cultures at all temperatures tested, with the selective pressure to acquire mutations highest at 37°, followed by 44°, and approximately equivalent at 42° and 30°, with the slowest mutant takeover found at 25° (Figure 1). A response to heat stress does not appear to be an important role for RpoS and low levels of RpoS at 44° may decrease competition for other σ-factors (RpoE and RpoH), which control the cell envelope and cytoplasmic heat-shock responses, respectively (Yura et al. 2000).
A modified atmosphere containing CO2 allowed enrichment of rpoS mutations, although at a slightly lower initial rate than that of the control (Figure 1). Mutations were not at all observed under anaerobic conditions (100% N2; King and Ferenci 2005) and this represents the only nutrient-limited situation where SrpoS is reduced to zero.
With the addition of 5% w/v sucrose to increase medium osmolarity, SrpoS was lower than with normal media (Figure 1). This was consistent with RpoS playing a role in osmoadaptation by increasing the transcription of genes involved in the synthesis and transport of osmoprotectants (Bianchi and Baneyx 1999; Hersh et al. 2004).
The rate of acquisition of rpoS mutations decreased with higher chloramphenicol concentration (Figure 1). Most of the rpoS mutations with chloramphenicol resulted in partial phenotypes, suggesting that some RpoS function was needed with the higher antibiotic concentration and that Fsp was contributing to the reduction of SrpoS. This is an interesting finding, because expression of efflux genes like acrAB is not RpoS dependent (Rand et al. 2002).
Estimating the selection pressure for loss of rpoS function or Fnc:
The difference in glucose transport rates between wild-type and the rpoS mutant provided a means of measuring the negative effect of RpoS on transporter gene expression and function. In nutrient-excess exponential culture, when rpoS is barely expressed, an rpoS mutation barely influences glucose uptake (Figure 2) and Fnc is minimal. In contrast, a 2.5-fold gain in transport by the rpoS mutant relative to wild type was found in bacteria growing in the nutrient-limited chemostat or “control” environment; this transport difference is the basis of the selection (Fnc) leading to rapid takeover by rpoS mutants in Figure 1. To see if artificially higher Fnc values could be used to verify the relationship between Fnc and rpoS mutant selection, a strain with an mgl–constitutivity mutation (with and without an rpoS mutation) was also compared in Figures 1 and 2. The mgl–con mutation increases expression of the MglBAC glucose/galactose transport system, particularly in an rpoS background (Notley-McRobb et al. 2002). As shown in Figure 2, the presence of RpoS indeed has a greater effect on glucose transport in the mgl–con strain, resulting in an Fnc value more than twice that of the wild-type strain. As shown in Figure 1, the mgl–con strain was swept by rpoS mutations even faster than wild type under identical conditions, with rpoS mutants being a significant proportion of the population even after 1 day of selection. These results confirm a strong correlation between the magnitude of Fnc and SrpoS in the absence of an additional stress.
Each environmental condition involving glucose-limited cultures was also used to test the effect of an rpoS null mutation on glucose uptake. The transport differences measured in the other environments are also shown in Figure 2. The repressive effect of rpoS and hence the Fnc advantage were not constant across environments and were low in anaerobic cultures, with high temperature, chloramphenicol, and increased osmolarity. This could have been due to lowered levels of RpoS reducing σ-factor competition and was tested by measuring RpoS accumulation in these conditions. The low Fnc is consistent with the decreased rates of rpoS sweeps with the conditions. However, the selection pressure derived from Fnc does not explain why no rpoS mutations occur at all anaerobically, while high osmolarity still has rpoS sweeps with a similar Fnc value. Lesser reductions in Fnc were seen with CO2 and pH 5.5, which also had slower sweeps by rpoS mutants. In contrast, Fnc was high at low temperature and high pH, which does not explain the slower sweeps in Figure 1 under these conditions. A possible explanation is that the strong nutritional selective pressure must have been counteracted by the stress-resistance contribution of RpoS or Fsp.
RpoS levels in various environments:
The β-galactosidase activity of an rpoS∷lacZ fusion was measured to help explain why Fnc varied in different conditions and to estimate the stress response role of RpoS in different stress states. The single-copy rpoS∷lacZ fusion contained all the known cis-regulatory elements subject to both transcriptional and post-transcriptional regulation and so is a reporter of protein levels under different growth conditions (Muffler et al. 1996). The fusion was monitored under each environmental condition, always at a dilution rate of 0.1 hr−1. The environmental conditions used for the mutation enrichment in Figure 1 were also used to measure, in day 1 chemostats, the initial level of the rpoS∷lacZ fusion (Figure 3). A fixed growth rate was needed because growth rate itself influences the expression of RpoS (Liu and Ferenci 2001; Ihssen and Egli 2004). In all cultures, the bacterial density was the same (∼2 × 108 bacteria ml−1) so RpoS accumulation was not affected by variations in density (Liu et al. 2000; Ihssen and Egli 2004).
The measured RpoS accumulation was comparable in glucose-limited and LB-limited cultures at the same growth rate (Figure 3), confirming that nutrient limitation itself elevates RpoS. RpoS levels, however, were not uniformly high when other stresses were applied to glucose-limited cultures. The biggest difference in rpoS regulation was observed when the normal N2/O2 atmosphere in chemostat cultures was replaced by sparging with 80% CO2:20% O2 or with 100% N2. Both conditions resulted in an ∼10-fold reduction in the level of RpoS accumulation compared to nutrient-limited growth in air (Figure 3). Anaerobic bacteria also exhibit lower expression of genes, such as katE dependent on RpoS, and exhibit reduced stress resistance (King and Ferenci 2005). The lowered RpoS levels are consistent with the lower Fnc value and lower RpoS/RpoD competition.
The media at non-neutral pH's resulted in different levels of RpoS accumulation. The level of RpoS was significantly below that at neutrality at acid pH but slightly higher at alkaline pH (Figure 3). The differential effect of pH on RpoS accumulation is not readily comparable to previous studies, but suggests that alkaline conditions are better at stimulating the general stress response. As in the case of altered atmospheres, there was no direct correlation between the rate of rpoS mutant enrichment and the level of RpoS in the culture. The RpoS accumulation data are also consistent with the levels of Fnc, which were higher with higher RpoS levels and increasing σ-factor competition.
In chemostats adjusted to various temperatures, RpoS accumulation was quite uniform over the range of 30° to 42° (Figure 3). At the less optimal temperatures, RpoS at 25° was slightly elevated, but was strongly reduced at 44°. Largely consistent with these results, RpoS accumulation in exponential and stationary-phase LB-grown batch cultures increased with decreasing temperatures between 20° and 42° (Sledjeski et al. 1996). At 44°, RpoS accumulation is possibly reduced with the additional high expression of the heat-shock σ-factors σH and σE at high temperatures (Yura and Nakahigashi 1999). As previously suggested by other workers, growth at low temperature does involve RpoS, and DsrA stimulates rpoS expression (Repoila et al. 2003) while Crl enhances the activity of RpoS (Bougdour et al. 2004).
The highest RpoS accumulation was found when medium osmolarity was increased and was 1.5-fold higher than that of the control (Figure 3). This was consistent with the regulation of RpoS levels by osmotic signals (Muffler et al. 1996). The elevated level with sucrose also indicated that the high induction by nutrient limitation (the “control” in Figure 3) is not the absolute maximal level of induction of rpoS. The high RpoS predicted from the fusion data, however, was not consistent with the expectation that high RpoS should increase Fnc through increased σ-factor competition. Possible reasons for these discrepancies are discussed below.
RpoS accumulation did not increase, relative to the control, with addition of either of the sublethal concentrations of 0.15 μg ml−1 or 0.3 μg ml−1 of the chloramphenicol tested (Figure 3). Hence low antibiotic concentrations did not further induce the general stress response and, if anything, decreased it at the higher concentration. The antibiotic effect on Fnc was in line with a lower level of RpoS and reduced σ-factor competition.
Nutrient limitation greatly increases the selection pressure for loss of rpoS functionality in E. coli (Ferenci 2003). The magnitude of Fnc and the advantage in nutrient uptake enjoyed by rpoS cells in comparison to rpoS+ bacteria, was much higher under glucose limitation than under nutrient-excess conditions, particularly when other environmental settings were near optimal. Artificial elevation of Fnc, as in the mgl mutant under glucose limitation, accelerated the selection of rpoS mutations. All this suggests that Fnc contributes to the magnitude of the selection pressure for rpoS mutations, at least when no other stress is applied. When the glucose-limited culture conditions were shifted to suboptimal environmental settings, the magnitude of Fnc was altered to a lesser or greater extent. The differential effects of environments on Fnc, SrpoS, and RpoS concentration are depicted in the environmental footprints shown in Figure 4. There was no obvious match between the magnitude of Fnc in different environments and the rate of takeover by rpoS mutations. If Fnc were the sole determinant driving SrpoS, then the environmental footprints would be much more coincident. As would be expected from the antagonistic pleiotropy associated with rpoS mutations, other influences in addition to Fnc complicate whether rpoS mutations sweep E. coli populations.
The magnitude of Fnc is itself likely to be subject to several influences. Shifts in pH away from neutrality or to low temperature maintained a high Fnc, but other suboptimal environments involving anaerobiosis, high osmolarity, high temperature, and chloramphenicol decreased Fnc and the advantage gained by rpoS mutants. The reduction in Fnc may be explained in two different ways. In some environments (anaerobiosis, high temperature, antibiotic) the level of RpoS was reduced relative to purely nutrient-limited bacteria, and hence reduced σ-factor competition can explain why RpoD-dependent glucose transport is less repressed. At this stage, we cannot explain the regulatory changes that result in reduced RpoS levels in these environments. Also unclear are situations such as with high osmolarity, when the RpoS accumulation is high and therefore σ-factor competition was also expected to be high. Yet Fnc was low with high osmolarity and there was reduced glucose transport in both rpoS and rpoS+ bacteria. A possible explanation is that expression or function of one of the glucose transporters is directly affected by growth in the presence of sucrose and it remains to be investigated whether osmolarity and altered intracellular solutes affect the PtsG or Mgl transport systems independently of rpoS regulation. All the regulatory changes affecting glucose transport are not necessarily fully understood for all environments, and this qualification applies to the interpretation of the other Fnc data as well.
Nevertheless, differences in Fnc can explain the changes in the speed of rpoS sweeps, or SrpoS, to a greater or lesser extent in several of the conditions tested. Low Fnc conditions like anaerobiosis, high temperature, antibiotic, and osmolarity were all associated with slower takeover by rpoS mutants. On the other hand, the weaker sweeps by rpoS null mutants were not associated with low Fnc under high pH or low-temperature conditions. The likeliest explanation of the sweep characteristics under the low-temperature conditions is that RpoS has a significant stress-protection role. In other words, Fsp was particularly strong in counteracting high Fnc. These high Fsp environments had high levels of RpoS accumulation consistent with a postulated role involving RpoS-dependent transcription in alkaline resistance (Small et al. 1994) as well as survival at low temperature (Kandror et al. 2002). Even in anaerobiosis, there was a low Fnc, comparable to that with chloramphenicol, but no mutational loss was observed; this is consistent with rpoS having a physiological role under anaerobic conditions that make rpoS mutations detrimental (King and Ferenci 2005).
In the absence of the ability to directly measure Fsp, we assumed that measuring the level of RpoS accumulation in different environments would give an indication of the beneficial need for RpoS. However, when environments were ranked in terms of decreasing RpoS levels, the extent of rpoS-mutational spread in a particular environment did not show anything like the same order. Indeed, as shown in Figure 4, the environmental footprint describing influence on RpoS levels is distinctly different from the pattern of rpoS mutation accumulation. Under anaerobiosis and modified atmospheres, RpoS levels were low but intact rpoS was maintained. In contrast, RpoS levels were high under optimal nutrient limitation conditions, and rates of mutation accumulation were high. This indicates that RpoS accumulation was high in situations when RpoS was not needed for survival (as in “control” cultures) and low when it was (as in anaerobiosis). The findings suggest that it is necessary to be careful in drawing parallels between the levels of gene expression and the actual benefit to the organism in various environments.
Another interesting result was the decreased selection of rpoS mutants in the presence of chloramphenicol. Resistance to antibiotics is not a known role for RpoS so the argument above—that mutations are less selected when RpoS is needed in stress resistance—does not hold readily. However, previous studies have shown that the major outer membrane porin through which chloramphenicol transverses in E. coli is OmpF (Harder et al. 1981; Cohen et al. 1988, 1989). RpoS negatively influences ompF transcription (Liu and Ferenci 2001); therefore loss of RpoS would be a disadvantage as high outer membrane permeability to antibiotics would be a liability and hence provide an indirect Fsp contribution. Another possible complication is that the chloramphenicol was added to cultures dissolved in ethanol, resulting in submillimolar ethanol levels in the culture. Some physiological effects of ethanol at much higher concentrations (50 mm) have been noted (Vulic and Kolter 2002), which may contribute to the observed changes.
In summary, this study has begun to reveal some of the multiplicity of influences that determine whether a mutation is beneficial and has enriched a bacterial population. One lesson from this study is that gene expression cannot be relied on to give a true indication of the benefit of a gene in different environments. Indeed, each condition makes a different contribution to SrpoS, Fsp, and Fnc. These results allow antagonistic pleiotropy in rpoS sweeps to be dissected, but also indicate that mutational sweeps are subject to complex physiological and regulatory influences yet to be uncovered. Understanding and modeling the genotype-by-environment interactions even within a single selection event is a far-from-trivial task.
We thank an anonymous reviewer for constructive suggestions and the Australian Research Council for funding support.
Communicating editor: S. Gottesman
- Received November 23, 2005.
- Accepted February 2, 2006.
- Copyright © 2006 by the Genetics Society of America