Introduction

The core RNA polymerase of E. coli may interact with seven different types of sigma subunits, each responsible for the transcription of a specific set of genes (Maeda et al. 2000). The principal sigma factor is σ70, which recognizes the majority of E. coli promoters, is particularly active in exponentially growing cells (Paget and Helmann 2003). The second most important sigma subunit, σS, accumulates in the cell during the stationary phase and under diverse stress conditions (Hengge-Aronis 2002). Many genes transcribed by the RNA polymerase holoenzyme coupled to σS (EσS) are related to bacterial protection from stresses. Some examples are katE, that encodes catalase and the genes osmB, osmC and osmY, which are involved in the cell response to hyperosmotic shock (Schellhorn and Hassan 1988; Vijayakumar et al. 2004; Weber et al. 2005). σS sometimes interacts with other transcription factors, such as cAMP/CRP, Lrp and IHF to control transcription (Colland et al. 2000; Lange et al. 1993; Taschner et al. 2006).

The accumulation of σS in the stationary phase and under stress leads to the competition between this sigma factor and σ70 for the binding to the core RNA polymerase, which results in partial inhibition of transcription of σ70-dependent genes (Farewell et al. 1998; Maeda et al. 2000). This form of negative regulation is at the core of the SPANC (self-preservation and nutritional competence) equilibrium, a trade-off pattern in which the transcription of genes related to protection against stress comes at the expense of the expression of growth-related genes (Ferenci 2005). Part of the σS/σ70 trade-off is an inverse relationship between stress resistance and metabolic capability, which varies in strains with different σS levels (King et al. 2004). For example, a strain with a high level of σS exhibits poorer growth on acetate as well as with many other carbon sources.

Since σS negatively affects the expression of σ70 and other sigma factors dependent genes, and since the cellular level of σS varies considerably among different E. coli strains (King et al. 2004), it is likely that the level of expression of a σ70-dependent gene would be a good reporter for σS variation among different strains. A good candidate is the gene phoA, which encodes the periplasmic enzyme alkaline phosphatase (AP). AP expression is strongly inhibited by the presence of rpoS (Taschner et al. 2004) and is sensitively detected using chromogenic substrates.

phoA belongs to the PHO regulon, which comprises more than 40 genes and operons that are induced in response to phosphate (Pi) shortage (Wanner 1996). Genes that belong to the PHO regulon possess one or more PHO-box sequences at their promoter regions instead of the conventional −35 (Wanner 1996). PHO genes are regulated by a two-component system formed by the sensor protein PhoR and the regulator PhoB. When the concentration of Pi in the medium falls below a critical level, PhoR autophosphorylates and transfer its Pi moiety to PhoB, which in turn binds to the PHO-boxes and interacts with Eσ70 initiating the transcription of the PHO genes.

Here we show that both the basal and induced levels of AP vary according to the level of σS in different E. coli strains. That variation in the levels of σS among different strains is easily detected by plating the bacteria on plates containing the AP chromogenic substrate 5-bromo-4-chloro-3-indolyl-phosphate (X-P). We also show that cells growing in a chemostat under Pi-limiting conditions accumulate mutations in rpoS and that the rate of acquisition of these mutations is dependent on the intrinsic σS level of each strain.

Materials and methods

Strains and growth conditions

The strains used in this study were MC4100 (F− araD139 (argF-lac)U169 rpsL150 deoC1 relA1 thiA ptsF25 flbB5301 rbsR) and its rpoS::Tn10 mutant BW3323; MG1655 (E. coli K-12 wild-type) and its rpoS::Tn10 mutant BW3708; M534 (EHEC isolate) and its rpoS mutant; ZK126 (W3110 ΔlacU169 tna2) and its rpoS::Tn10 mutant ZK1171. BW2952 is a derivative of strain MC4100 containing a malG::λplacMu55 insertion, which was used in chemostat evolution experiments (Maharjan et al. 2006). Strains BW3767, BW4002, BW4001, BW4003 and BW4029 evolved from the starting strain BW2952 in a glucose-limited chemostat. The former two are rpoS mutants and the three other strains are rpoS +. ECOR 13, 14, 24, 25, 64 and 70 are E. coli isolates from different locations (Ochman and Selander 1984).

T-salts (Echols et al. 1961) plates contained 0.2% glucose, 1 mM KH2PO4 and 40 μg/ml X-P. LB plates were as described (Miller 1992). For the batch cultures, cells were grown overnight in liquid T-salts supplemented with 0.2% glucose and 5 mM KH2PO4 for the Pi-excess medium or 0.2 mM KH2PO4 for the Pi-limited medium at 37°C. Liquid T-salts supplemented with 0.2% glucose and 30 μM KH2PO4 were used to set up an 80 ml chemostat culture as described (Notley-McRobb et al. 2002). The dilution rate was set to 0.1 h−1. Daily samples were taken for AP activity measurements and also for rpoS status of members of the population. AP activity was assayed directly on the chemostat sample. The rpoS status was determined by diluting the culture, spreading on LB plates and staining with iodine (see below).

AP assays

For the quantitative AP assay p-nitrophenyl-phosphate (p-NPP) was used as substrate as described (Spira et al. 1995). Briefly, overnight cells were washed and resuspended in unsupplemented T-salts; 0.1 ml of the cell suspension was mixed with 0.9 ml p-NPP (1 mg/ml in Tris pH 8) and incubated at 37°C until a yellow colour developed; the reaction was then stopped by the addition of 1 M Na2HPO4. The reaction products were briefly centrifuged to pellet cell remains and read in a spectrophotometer at 410 nm. AP activity units are defined as the increase in absorbance at 410 nm min−1 optical cell density at 600 nm−1.

Alkaline phosphatase was qualitatively assayed by observing the colour development of bacterial patches growing on T-salts plates supplemented with 40 μg/ml X-P.

Detection of rpoS status

The level of rpoS was qualitatively assessed by staining glycogen with iodine solution as described (Hengge-Aronis and Fischer 1992). Patches of bacteria or diluted chemostat samples were grown overnight on LB plates, stored at 4°C for 24 h and then flooded with iodine. The intensity of the brown colour varies according to the level of σS in the cell (Notley-McRobb et al. 2002). rpoS + strains, show, therefore, different intensities of brown, while rpoS mutants always remain very light.

Results and discussion

AP activity in different E. coli strains

To test if differences in the AP level found among E. coli strains are related to the intrinsic σS concentration in these bacteria, we assayed the AP activity of strains MC4100, MG1655, ZK126 and M534 and their isogenic rpoS::Tn10 derivatives. These strains were chosen despite having identical rpoS sequences, as they display significant differences in σS concentration; in both acetate and glucose-limited chemostats the level of σS was the highest in the strains M534 and MC4100, followed by ZK126 and MG1655 (King et al. 2004). The cells were grown overnight in minimal medium under Pi-limited and Pi-excess conditions and assayed for AP. Under Pi starvation conditions, the highest AP levels were observed for MG1655 (0.53 ± 0.06 units of activity) and ZK126 (0.56 ± 0.04), followed by M534 (0.33 ± 0.02) and MC4100 (0.18 ± 0.01). Disruption of the rpoS gene increased the level of AP in all four strains to similar levels (2.11 ± 0.23 for MG1655 rpoS; 1.77 ± 0.14 for ZK126 rpoS; 1.9 ± 0.03 for M534 rpoS and 1.92 ± 0.1 for MC4100 rpoS). However, the net increase in AP due to rpoS inactivation was 10.8-fold for strain MC4100, 6.7-fold for M534, fourfold for MG1655 and 3.2-fold for ZK126. Therefore, strains that naturally produce high levels of σS showed a stronger effect on AP expression, as high levels of σS strongly compete with σ70, thus producing less AP (Taschner et al. 2004).

In the cells growing under Pi abundance conditions (AP basal level), no clear effect of rpoS on the level of AP was observed (not shown), with the possible exception of MC4100 in which disruption of rpoS increased the AP level by 2.2-fold (from 0.004 ± 0.001 to 0.009 ± 0.0007). The lack of a negative effect of rpoS on the non-starved cells could be due to the fact that expression of phoA under these growth conditions is so low that it is barely influenced by the σS/σ70 competition or because the assay with p-NPP is not sensitive enough to detect discrepancies in the basal level of AP.

Qualitative measurement of AP level in minimal plates

The wild-type strains and their rpoS::Tn10 mutants were patched on minimal medium plates supplemented with excess Pi and the highly sensitive AP-specific chromogenic substrate X-P. Figure 1a shows that the colour intensity developed by MC4100 was the weakest followed by M534, MG1655 and ZK126. At least for M534 and MC4100 the rpoS::Tn10 mutants displayed a stronger blue colour than their wild-type parents. This implies that even under PHO-repressing conditions the negative effect of rpoS on AP synthesis can be discerned and that this effect was not noticed previously because the methods employed, such as AP assay with p-NPP or northern hybridization (Taschner et al. 2004), were not sufficiently sensitive to observe the effect of rpoS under Pi-excess.

Fig. 1
figure 1

AP and glycogen variation in patches of bacteria growing on solid media. a, c and e Patches of bacteria on T-salts plates supplemented with X-P. b, d and f Patches of bacteria on LB plates stained with iodine. g Cultures of different strains mixed together and plated on T-salts supplemented with X-P

Full size image

To confirm the rpoS status of the strains, colonies were patched on LB plates and stained with iodine, which indicates rpoS-dependent glycogen accumulation in the cells (Hengge-Aronis and Fischer 1992; Notley-McRobb et al. 2002) (Fig. 1b). The pattern of colour intensity confirmed previous results (King et al. 2004), i.e., glycogen accumulation was higher in strains MC4100 and M534 than in MG1655 and ZK126, confirming that the former strains produce more σS than the others. Furthermore, all wild-type strains developed a significantly stronger brown colour than the rpoS::Tn10 mutants.

X-P plates can also be used to differentiate between natural E. coli isolates regarding their rpoS status. Patches of ECOR collection strains representing diverse sources of E. coli (Ochman and Selander 1984) were grown on X-P plates (Fig. 1c). Strains ECOR13, 14 and 24 displayed a less intense colour when compared to the other three strains. On the other hand, these same strains stained darker with iodine than strains ECOR25, 64 and 70 (Fig. 1d), suggesting that the latter strains produced less or no σS. The rpoS status of the ECOR strains was confirmed by observing the catalase activity due to the addition of H2O2 to the colonies. While ECOR13, 14 and 24 displayed a strong bubbling, the other three strains did not bubble at all (not shown). This suggests that X-P staining can be used to differentiate between rpoS + and rpoS − phenotypes even in strains of very different genetic backgrounds. It is important to note, however, that other properties besides the rpoS level, such as mutations in the PHO regulatory genes may influence the level of AP in the cells. Therefore, care must be taken when comparing the AP activity of strains from different backgrounds and whenever possible other rpoS-related tests should be performed.

We then tested if plating on X-P plates could be used to follow the evolution of heterogeneity in a glucose-limited chemostat, which selects for rpoS mutants in the population (Maharjan et al. 2006; Notley-McRobb et al. 2002); 6 out of 41 individual clones isolated after 26 days of growth under glucose limitation (Maharjan et al. 2006) were patched on an X-P plate (Fig. 1e). Strains BW2952 (parent strain), BW4001, BW4003 and BW4029 contained no rpoS mutations and stained lightly, while strains BW3767 (stop codon at position 315 of the σS protein) and BW4002 (stop codon at position 247) displayed a considerably darker blue colour. Accordingly, the iodine staining showed an inverse relationship with the AP activity on the plates, as the rpoS mutants were lighter than the rpoS + strains (Fig. 1f).

When the cultures of strains exhibiting different levels of σS were mixed and plated on X-P, it was easy to discern the high σS producers from the other strains as shown in Fig. 1g, where strain MC4100 (light blue colonies) was mixed with its rpoS::Tn10 mutant or with MG1655. Therefore, at least for derivatives of strain MC4100, which expresses high amounts of σS and is used as the starting strain in evolution experiments in this lab, the rpoS − evolved isolates are easily discernible by streaking or patching the colonies on X-P plates. This technique is better than iodine staining because the cells are not killed and can, therefore, be directly isolated from the plates.

Rate of rpoS loss in a Pi-limited chemostat

As noted above, MC4100 bacteria growing in a chemostat under glucose limitation accumulate mutations in rpoS, which sweep the population leading to the elimination of the rpoS + parent (Maharjan et al. 2006; Notley-McRobb et al. 2002). This radical change in the population profile occurs because σS is a burden for bacteria growing in a carbon-limited environment. In rpoS + bacteria, most of the growth-related genes, such as the ones that encode glucose transporters have their expression negatively affected by σS, due to the competition between σS and σ70 (Farewell et al. 1998; Maeda et al. 2000). A strong selective pressure for the elimination of rpoS + results in the optimisation of the expression of growth-related genes, and these rpoS mutants eventually sweep the population (Ferenci 2005). However, this is not the case for MG1655 populations in a glucose-limited chemostat, which do not enrich rpoS mutations, probably because the relatively low level of σS expressed in these cells does not give rise to a strong selective pressure against rpoS (King et al. 2004).

To test if rpoS also poses a disadvantage to bacteria growing under steady-state Pi-limitation, a short-term evolution experiment using separately both MC4100 and MG1655 as the starting strains was conducted. The cells were inoculated in minimal medium supplemented with excess glucose and 30 μM Pi and grown for 4 days. The proportion of rpoS + cells and the level of AP in the population were followed throughout this period (Fig. 2). In the evolving MC4100 population, after 2 days there was already a 25% decrease in the number of rpoS + cells along with a 2.5-fold increase in the level of AP. On day 4, the rpoS + cells represented only 5% of the population, while the level of AP was fivefold higher than on day 1. For the evolving MG1655, the first rpoS mutants appeared only on day 3, comprising just 11% the population. The AP level barely increased at that point. On day 4, the proportion of rpoS mutants was 64% and the AP level increased 3.3 times as compared to the day 1 population. This demonstrates that the presence of rpoS + is disadvantageous to cells growing in a Pi-limited environment, and that the proportion and rate of rpoS loss is directly related to the intrinsic level of σS. Acquisition of mutations in the rpoS gene is apparently the simplest way to enable an increase in the expression of the PHO genes and in this way to improve the fitness under these conditions. This finding is in agreement with the SPANC trade-off hypothesis (Ferenci 2005) and extends to inorganic nutrients. Therefore, bacteria that express high levels of σS are subjected to similar selection pressures under steady-state Pi-limitation as those growing in a glucose-limited chemostat.

Fig. 2
figure 2

AP activity and rpoS loss in MC4100 and MG1655 cells cultivated in a Pi-limited chemostat. MC4100 (closed symbols) and MG1655 (open symbols) cells were grown for 4 days in a chemostat containing T-salts supplemented with 0.2% glucose and 30 μM Pi at a dilution rate of 0.1 h−1. Total AP activity (squares) in the population and glycogen staining (rpoS status) (triangles) of isolated colonies were measured daily

Full size image

The reason why MG1655 did not acquire rpoS mutations in a glucose-limited chemostat under the same dilution rate as employed here (King et al. 2004) is probably related to the fact that the negative effect of rpoS on the expression of glucose transport genes is milder than its effect on PHO genes expression. For instance, MC4100 rpoS::Tn10 showed a fourfold increase in malG expression when compared to its rpoS + parent (Notley-McRobb et al. 2002). In contrast, the AP level of MC4100 rpoS::Tn10 mutant was 10.8-fold higher than that of its parent and even the MG1655 rpoS::Tn10 mutant showed a fourfold increase in comparison to its rpoS + parent. Therefore, it seems that the PHO genes are more strongly repressed by σS under Pi-limiting conditions than are the glucose transport related genes in bacteria growing in a glucose-limiting chemostat.

In conclusion, it was shown that the activity of AP is inversely related to the intrinsic σS level of individual E. coli strains. The negative effect of rpoS on PHO expression is proportional to the relative amounts of σS in the bacteria, probably because in cells that produce more σS there is more competition with σ70. The use of a newly devised non-destructive method enabled us to find out that inhibition of AP by σS occurs even under conditions of PHO repression, when only basal levels of AP are synthesised. This agrees with the model that sigma factor competition is at the core of the inhibitory effect of rpoS. Under steady-state growth in a Pi-limited medium, high σS producers, such as strain MC4100 rapidly accumulate rpoS mutations as a way of relieving the repression of the PHO genes. Strains that produce lesser amounts of σS, such as MG1655, still acquire mutations in rpoS, but more slowly.