σS (RpoS) is the master regulator of the general stress response in Escherichia coli. Several stresses increase cellular σS levels by inhibiting proteolysis of σS, which under non-stress conditions is a highly unstable protein. For this ClpXP-dependent degradation, the response regulator RssB acts as a recognition factor, with RssB affinity for σS being modulated by phosphorylation. Here, we demonstrate that RssB can also act like an anti-sigma factor for σSin vivo, i.e. RssB can inhibit the expression of σS-dependent genes in the presence of high σS levels. This becomes apparent when (i) the cellular RssB/σS ratio is at least somewhat elevated and (ii) proteolysis is reduced (for example in stationary phase) or eliminated (for example in a clpP mutant). Two modes of inhibition of σS by RssB can be distinguished. The ‘catalytic’ mode is observed in stationary phase cells with a substoichiometric RssB/σS ratio, requires ClpP and therefore probably corresponds to sequestering of σS to Clp protease (even though σS is not degraded). The ‘stoichiometric’ mode occurs in clpP mutant cells upon overproduction of RssB to levels that are equal to those of σS, and therefore probably involves binary complex formation between RssB and σS. We also show that, under standard laboratory conditions, the cellular level of RssB is more than 20-fold lower than that of σS and is not significantly controlled by stresses that upregulate σS. We therefore propose that antisigma factor activity of RssB may play a role under not yet identified growth conditions (which may result in RssB induction), or that RssB is a former antisigma factor that during evolution was recruited to serve as a recognition factor for proteolysis.
The σS (RpoS) subunit of RNA polymerase (RNAP) is the master regulator of the general stress response in Escherichia coli. Rapidly growing cells contain little if any σS, but the σS level strongly and rapidly increases in response to a variety of stresses, including starvation, hyperosmolarity, high or low temperature and acidic pH. Several of these stresses inhibit proteolysis of σS, which in non-stressed cells is rapidly degraded (for a review of σS function and regulation, see Hengge-Aronis, 1996). For σS turnover, three factors are essential: ClpXP protease (Schweder et al., 1996), the response regulator RssB (Bearson et al., 1996; Muffler et al., 1996a; Pratt and Silhavy, 1996; Bouchéet al., 1998) and a cis-acting element within σS, the turnover element (Muffler et al., 1996b; Schweder et al., 1996; Becker et al., 1999). RssB is a direct recognition factor that interacts with the turnover element within σS in a manner that is dependent on the phosphorylation of the N-terminal RssB receiver domain (Becker et al., 1999). Stress may thus control RssB affinity for σS and consequently σS proteolysis by affecting the phosphorylation of the RssB receiver.
σS interacts with at least two different partners: (i) RNAP core enzyme, which results in σS-dependent gene expression and (ii) RssB, which leads to σS degradation. The question therefore arises whether RssB can affect σS binding to RNAP core. If there were conditions under which interaction between RssB and σS would not immediately lead to σS proteolysis, RssB could then interfere with σS activity, i.e. RssB would act like an antisigma factor according to the original definition of antisigma factors (for recent reviews, see Hughes and Mathee, 1998; Helmann, 1999).
More recently, proteins (such as the phage T4 protein AsiA; Colland et al., 1998) have been found that can inhibit or alter the activity of sigma factors even when bound to RNAP core, and these factors are now also seen as antisigma factors (Hughes and Mathee, 1998; Helmann, 1999). In view of this broader definition, it is interesting to note that the binding site for RssB in σS, i.e. the turnover element, is directly adjacent to region 2.4 in σS, which is crucial for recognition of the − 10 promoter element, and even overlapping with region 2.5 (Becker et al., 1999), which as in σ70 may be involved in recognition of an extended − 10 promoter region (Barne et al., 1997; Bown et al., 1999). Binding of RssB, if it can take place in the holoenzyme complex, could thus affect the ability of σS to interact with promoters.
In this study, we tested whether conditions could be found in vivo under which σS proteolysis is reduced and a potential ability of RssB to affect σS activity in transcription initiation would become visible. The experiments were performed in vivo because standard in vitro transcription assays tend to not correctly reflect in vivo activity of σS, which is crucially dependent on intracellular salt conditions (Ding et al., 1995), global and local DNA supercoiling (Kusano et al., 1996) as well as additional promoter-associated proteins (Marschall et al., 1998; Hengge-Aronis, 1999). We report here two sets of results obtained under different experimental conditions which indicate that RssB can indeed interfere with σS activity in vivo, i.e. can act like an antisigma factor.
RssB together with Clp protease can interfere with σS activity in stationary phase cells
In vivo activity of σS can be monitored by measuring the expression of various σS-dependent genes, which can easily be assayed using lacZ reporter fusions. For the present investigation, we used strains (Becker et al., 1999) that carry a lacZ fusion in a σS-dependent gene (which can be osmY, bolA, csiD or otsB), as well as the wild-type rssB allele and a rpoS::Tn10 null mutation in the chromosome. The rpoS mutation is complemented by σS being expressed from pBAD18 (a moderate copy number expression vector that makes use of the arabinose-inducible pBAD promoter; Guzman et al., 1995). In the absence of inducer, this system yields σS levels that correspond to ≈ 30% of the σS level in a wild-type strain (i.e. the RssB/σS ratio in these strains is about threefold elevated; Becker et al., 1999). Not only β-galactosidase activities were determined but also cellular σS levels (assayed by immunoblot), as the latter could vary up to twofold in the different strains compared. This allowed the determination of a ‘relative in vivo activity’ of σS by normalizing the β-galactosidase activities, i.e. the expression of various σS-dependent genes, for actual σS levels. Data were obtained 2 h after entry into stationary phase because induction of σS and the σS-regulated genes used here is continuous at least up to this time, and RpoS levels can easily be quantified by immunoblot.
The expression of osmY::lacZ was found to be significantly higher in rssB or clpP mutant backgrounds than in the rssB+clpP + strain (all strains expressing wild-type σS; Fig. 1A). When normalized for σS levels (Fig. 1B), the values obtained for the rssB or clpP mutant backgrounds were more than sixfold higher (Fig. 1C). This indicated that RssB and Clp protease can reduce σS activity. Interestingly, both factors have to be present for this interference to occur. Moreover, this phenomenon could only be seen in stationary phase cells. In exponentially growing cells, the expression of osmY::lacZ is proportional to σS levels, no matter whether or not RssB and/or ClpP are present (data not shown).
Similar data indicating an inhibition of σS by RssB (Fig. 2A) and ClpP (data not shown) were also obtained when fusions to the σS-regulated genes bolA, csiD and otsB were used to monitor σS activity (Fig. 2A). Taken together, these data provide evidence that RssB and Clp protease not only promote σS degradation but, under conditions where proteolysis is downregulated, i.e. in stationary phase, are still able to interact with σS and reduce its ability to activate gene expression.
In contrast, no such inhibition by RssB/Clp was observed for σS carrying the K173E mutation, no matter whether osmY, bolA, csiD or otsB were used to test for σS activity (Figs 1 and 2B). This is consistent with our previous demonstration that this mutant σS is not recognized by RssB (Becker et al., 1999). The E174T variant of σS, which can bind to RssB, exhibited an inhibition by RssB/Clp similar to that observed with wild-type σS (Figs 1 and 2C).
The intracellular level of RssB protein is much lower than the σS level
Inhibition of σS by RssB/Clp would require a stoichiometric interaction either between σS and RssB or between σS and Clp. The latter interaction may be equivalent to terminal sequestration of σS to Clp protease, which could occur even if RssB levels are significantly lower than σS levels, if the transfer of σS from RssB to Clp protease would be an efficient one-way reaction. Our finding that ClpP is required for inhibition of σS to occur (the clpP mutant still contains RssB but exhibits high expression of osmY; see Fig. 1) tentatively suggested that σS sequestration to ClpP might be the basis for this inhibition.
To clarify these issues, we determined the cellular levels of RssB protein in the standard laboratory strain MC4100. Cells growing in minimal glucose medium were used as well as cells exposed to carbon starvation (stationary phase) and high osmolarity, i.e. conditions that are known to strongly increase σS levels (Lange and Hengge-Aronis, 1994). Quantitative immunoblotting of both RssB and σS was performed with purified proteins as standards (Fig. 3). It was found that RssB levels are significantly lower than σS levels. In cells growing on glucose or glycerol as carbon sources, the actual RssB concentrations calculated were 0.5 and 0.34 fmol μg− 1 cellular protein respectively. Cellular concentrations of σS under these conditions were 12 and 8.5 fmol μg− 1, i.e. the cellular RssB/σS ratio is ≈1:25. Whereas RssB levels do not change much in response to glucose starvation (0.8 fmol μg− 1) and osmotic stress (0.45 fmol μg− 1), induction of σS to 18 fmol μg− 1 and 39 fmol μg− 1 respectively, was observed under these conditions. This also means that the RssB/σS ratio remains practically constant during entry into stationary phase, but changes to ≈1:85 upon osmotic upshift. The values for σS reported here are in good agreement with previously published data (Jishage et al., 1996).
Although in the strains constructed for the present study σS contents are threefold lower than in MC4100, RssB levels would still be approximately eightfold lower than σS levels. This indicates that the inhibition of σS by RssB/Clp as demonstrated above cannot be due to direct stoichiometric interaction between RssB and σS, but rather involves a catalytic role of RssB in the formation of a complex between σS and Clp.
When present in stoichiometric amounts, RssB can inhibit σS activity in the absence of Clp protease
As stable interaction between purified RssB and σS alone occurs in vitro (Becker et al., 1999), it was interesting to ask whether RssB would also be able to directly inhibit in vivo activity of σS in a binary complex, i.e. when RssB was expressed in quasistoichiometric amounts with σS in a background deficient for Clp protease. The strains used for these experiments carry the wild-type chromosomal allele of rpoS, the osmY::lacZ fusion as a reporter for in vivo activity of σS and rssB::Tn10 and clpP::cat null mutations in the chromosome. Now, RssB was produced from the pBAD18 plasmid, i.e. the cellular level of RssB could be strongly increased by induction with arabinose (Fig. 4). As the presence of high arabinose concentrations slightly reduced σS levels under some conditions (Fig. 4), the osmY::lacZ activities used as a measure for σSin vivo activity were again normalized for σS levels. The experiments were performed during exponential growth, during the onset of starvation and 2 h after entry into stationary phase. The exponential phase data could be obtained because in the absence of Clp protease σS levels and the expression of σS-dependent genes are high also in growing cells.
Overproduction of RssB was indeed found to strongly reduce the expression of osmY::lacZ (Fig. 4A1, B1 and C1). The relative in vivo activities of σS obtained by normalization for σS levels was reduced to 20–25% upon full induction of RssB (Fig. 4A3, B3 and C3). Although high concentrations of arabinose, even in the absence of RssB (i.e. in the strain just carrying the vector), somewhat reduced σS levels (Fig. 4A2 and B2) and therefore also osmY expression not yet corrected for σS levels (see arrows in Fig. 4A1, B1 and C1), it had no or little effect on relative σS activity (as calculated for Fig. 4A3, B3 and C3). We conclude that RssB, when expressed in sufficient amounts, can interfere with σS activity also in the absence of Clp protease, i.e. probably by forming a binary RssB–σS complex. This conclusion is the same for growing and stationary phase cells. Interestingly, RssB induction is stronger in stationary phase cells (about tenfold less arabinose is required to achieve half-maximal induction; Fig. 4A2, B2 and C2). This is probably owing to cAMP-CRP activation of the pBAD promoter and increased plasmid copy number in stationary phase cells. Nevertheless, half-maximal induction of RssB always resulted in approximately the same decrease of in vivo activity of σS, independently of growth phase.
Inhibition of σS activity is relieved by mutations in RssB that eliminate receiver phosphorylation
In order to be sure that inhibition of σS by overproduction of RssB was not some unspecific effect, we wanted to test whether this inhibition would be dependent on the physiological activity of RssB as controlled by the status of its receiver domain. Therefore, we tested a series of rssB mutants with random exchanges in codon 58 encoding the aspartate residue, which is the site of phosphorylation in the RssB receiver domain (Bouchéet al., 1998). Interestingly, not all mutations eliminated RssB activity in promoting σS proteolysis completely, as determined by assaying σS levels as well as the activity of the translational rpoS742::lacZ fusion (which encodes a hybrid protein that is degraded just as σS itself). Although not as active as wild-type RssB, the RssBD58P protein could still support σS degradation to some extent, as apparent from relatively low σS levels and fusion activities (Fig. 5). In contrast, RssBD58Q and RssBD58R had more strongly reduced activities. The low residual activity observed with the D58stop mutant (compared with the result with a strain completely devoid of RssB; Fig. 5A, column 2) is probably owing to occasional readthrough.
The partially active D58P and inactive D58R variants of RssB were tested for their ability to inhibit in vivo activity of σS upon overproduction 5(Fig. 6). Mutant RssB levels obtained by arabinose induction were similar to those observed for wild-type RssB (data not shown). In comparison with wild-type RssB (Fig. 4), RssBD58P showed somewhat reduced ability to inhibit σS, with relative in vivo activity of σS going down to 55% upon full induction of RssBD58P (Fig. 6B and D). RssBD58R could only slightly inhibit σS activity when fully overproduced (87% activity retained; Fig. 6A and C). These data confirm that the ability of RssB to reduce σS activity upon overproduction is as dependent on the status of the RssB receiver domain as its ability to support σS proteolysis when present in low amounts. We therefore conclude that inhibition of σS by quasistoichiometric amounts of RssB is related to the biological properties of RssB and therefore could be of physiological importance.
RssB has the potential to inactivate σS, as well as promoting σS proteolysis
In rapidly growing cells, the response regulator RssB acts as an essential direct recognition factor for σS proteolysis. σS bound to RssB is then transferred to ClpXP protease, where it is rapidly and completely degraded. RssB affinity for σS is modulated by phosphorylation of the RssB receiver domain, which suggests that stress(es) that result(s) in stabilization of σS act by somehow triggering dephosphorylation of RssB (Becker et al., 1999). In the present study, we demonstrate that in vivo interaction of σS with RssB can be uncoupled from degradation under certain conditions, and then results in an inactivation of σS. RssB thus has the potential to act like an antisigma factor for σS.
We do not yet know whether RssB binds only to free σS, in which case RssB may inhibit RNAP holoenzyme formation, or whether RssB is also able to interact with and inhibit σS within the holoenzyme. Both possibilities, however, would conform to the current definition of an antisigma factor (Hughes and Mathee, 1998; Helmann, 1999). In E. coli, the phage T4 protein AsiA binds to σ70 region 4 within the holoenzyme complex, and thereby interferes with recognition of the − 35 promoter region by the RNAP holoenzyme (Colland et al., 1998; Severinova et al., 1998). AsiA-bound RNAP can still transcribe promoters that depend on polymerase contacts in an extended −10 region alone (Colland et al., 1998; Severinova et al., 1998). RssB binds to a part of σS, the turnover element, which overlaps with region 2.4/2.5, i.e. a region crucial for σS binding to the −10 promoter region (Becker et al., 1999). If RssB interacts with σS-containing RNAP holoenzyme, it can be expected to inhibit σS-mediated initiation at all promoters because interaction with the −10 promoter region is essential for transcription initiation, especially for σS-dependent promoters (Hiratsu et al., 1995; Tanaka et al., 1995; Colland et al., 1999).
There are two modes of σS inhibition, depending on the cellular RssB levels
Inhibition of σS by RssB becomes apparent under conditions in which (i) the RssB/σS ratio in the cell is at least somewhat elevated and (ii) proteolysis of σS is reduced or eliminated, either physiologically (in stationary phase) or by a mutation in clpP. Depending on how these two conditions are fulfilled, we were able to distinguish two modes of inhibition of σS, which differ in the role of RssB. The ‘catalytic mode’ (exemplified by the experiments shown in Figs 1 and 2) was observed when the RssB/σS ratio was increased only threefold, which means that the actual concentration of RssB is still approximately eightfold lower than that of σS. Under these conditions, RssB and Clp protease are both required for σS inhibition because mutations either in rssB or in clpP alone are sufficient to increase σS activity (Fig. 1). This mode of inhibition is probably equivalent to sequestration of σS to ClpXP protease. Such inhibition can only be observed in stationary phase as any contact between σS, RssB and ClpXP in exponential phase would result in rapid degradation, i.e. immediate disappearance of σS (see also below). The ‘stoichiometric mode’ of inhibition of σS by RssB was observed in the absence of ClpP, when RssB was overproduced in amounts quasistoichiometric with σS (Figs 4 and 6). These experiments could also be performed with growing cells because the mutation in clpP increases σS levels and therefore the expression of σS-dependent genes in the exponential phase. A previously reported inhibition of σS by RssB (which was observed with a dsrB::lacZ fusion as a reporter and which was about twofold when corrected for σS levels) would probably correspond to stoichiometric inhibition by RssB because it was observed in a clpP-deficient strain that carried rssB on a multicopy plasmid (Zhou and Gottesman, 1998). In our experiments, despite strong overproduction of RssB, its interaction with σS is physiological in the sense that it depended on the status of the RssB receiver domain. Mutant forms of RssB with substitutions in D58, which result in partial or complete defects in σS proteolysis (under conditions of low RssB expression), were defective to similar extents in stoichiometric inhibition of σS activity (i.e. under conditions of RssB overproduction; Fig. 6).
Additional implications for starvation signalling and the role of RssB in the initiation of σS proteolysis
The observation of a catalytic mode of σS inhibition by RssB, which also requires ClpP and occurs in stationary phase, has interesting additional implications. This means that in starved cells RssB/ClpXP can still interact with σS and that, therefore, starvation results in more than just reduced binding of RssB to σS, which may be achieved by dephosphorylation of RssB. Moreover, under these conditions, contact between σS and ClpXP protease is obviously possible without degradation being the immediate consequence. One should keep in mind that σS is stabilized in response to many quite different stresses, which need not act all in the same way. Starvation could be a stress that is not (only) signalled via dephosphorylation of RssB, but may (also) modulate a subsequent step in σS proteolysis, e.g. the activity of ClpXP itself. Also, the finding that similar amounts of (overproduced) RssB reduce σS activity to similar extents in growing and in starving cells would be consistent with starvation stress not affecting RssB phosphorylation. Alternatively, RssB may not only bind σS and present it to ClpXP, but in a ternary σS–RssB/ClpXP complex may also play a role in initiating unfolding. Such a second function of RssB may be more strongly affected by stress-induced dephosphorylation of the RssB receiver domain, whereas binding of non-phosphorylated RssB to σS would only be less efficient (but still lead to enough association of σS with ClpXP to be of measurable consequence if σS transfer from RssB to ClpXP is an efficient one-way reaction). In order to clarify these questions, future studies will have to demonstrate directly what stresses, which affect σS proteolysis, are signalled via dephosphorylation of RssB in vivo.
Another finding reported here is interesting with respect to the functioning of response regulator receiver domains. By random mutagenesis of codon 58 of RssB, which encodes the phosphorylated aspartyl residue (Bouchéet al., 1998), we obtained a partially active variant carrying a proline substitution. The three-dimensional structure of response regulators can roughly be divided into two halves connected by β-strand 3, which ends with the phosphorylated aspartyl residue (Stock et al., 1989; Volz and Matsumura, 1991; Volkman et al., 1995; Baikalov et al., 1996; Madhusudan et al., 1997). One could imagine that inserting a proline at this position may somewhat affect the position of the two halves of the receiver relative to each other. The finding that the D58 → P exchange results in partially active RssB suggests that the mutant receiver domain partially mimics a phosphorylated receiver domain. This would be in agreement with NMR chemical shift perturbation studies performed with CheY and NtrC, which indicate that phosphorylation triggers conformational changes mainly in those parts of the receiver that are located downstream of the site of phosphorylation and constitute a surface that acts as an output interface (Lowry et al., 1994; Nohaile et al., 1997).
Is RssB a former antisigma factor that was recruited to serve as a recognition factor for σS proteolysis?
In this study, we have demonstrated that RssB has the potential to act like an antisigma factor for σSin vivo. However, we have also shown that, under standard laboratory conditions, the cellular RssB content is much lower than that of σS, even in growing cells (in minimal medium; in rich medium, σS levels are below detection). This substoichiometry is appropriate for a catalytic role of RssB as a recognition factor in the initiation of σS proteolysis. An antisigma factor, however, would have to act stoichiometrically. Therefore, is there a physiological role for RssB as an antisigma factor for σS? For once, there might be not yet identified conditions that result in strong upregulation of RssB. If, at the same time, Clp protease activity was somehow compromised, the result would be σS inhibition by RssB. Alternatively, RssB may originally have been an environmentally controlled antisigma factor for σS, which during the course of evolution was recruited by the ClpXP machinery to serve as a recognition factor for σS proteolysis.
Bacterial strains and growth conditions
The strains used in this study are derivatives of strain MC4100 described previously (Becker et al., 1999). Different allelic combinations were obtained by P1 transduction (Miller, 1972). Alleles used were rpoS359::Tn10 (Lange and Hengge-Aronis, 1991a), rssB::Tn10 (Muffler et al., 1996a), rssB::cat (kindly provided by F. Moreno) and clpP1::cat (Maurizi et al., 1990). Reporter gene fusions to RpoS-dependent genes were the following: RO151 [MC4100 carrying csi-5(osmY)::lacZ(λplacMu55); Weichart et al., 1993], RH95 [MC4100 carrying λMAV103::bolAp1::lacZYA;Lange and Hengge-Aronis, 1991b], DW12 (MC4100 carrying csi-12(csiD)::lacZ(λplacMu15); Weichart et al., 1993], LB83 [MC4100 carrying otsB::lacZ(λplacMu55) with the otsB fusion derived from FF1112; Giaever et al., 1988]. The reporter fusion rpoS742::lacZ in strain RO91 is a translational fusion in rpoS located at the att site in the chromosome (Lange and Hengge-Aronis, 1994). Fusion strains carrying different combinations of rpoS, rssB and clpP alleles were used as recipients for plasmids derived from pBAD18, which express different variants of RpoS or RssB under the control of the pBAD promoter (Guzman et al., 1995). For experiments that involved arabinose induction, the strains also carried the Δ(ara–leu)7697 deletion.
The isolation of pBAD18 derivatives expressing wild-type σS or σS variants with single amino exchanges in the turnover element of σS has been described previously (Becker et al., 1999). pBAD18 derivatives carrying the wild-type rssB or rssAB regions were obtained by cloning corresponding PCR fragments into the EcoRI and XbaI sites of pBAD18 (the primers used were 5′-GCAATAGAATTCCACTATTGAGTAAAGCC-3′ and 5′-GTGGCTACGAATTCATTCCAGGGG-3′ respectively, in combination with the down-stream primer 5′-GCAAATCTAGACCGCGTTATCGTTTGC-3′; with relevant restriction sites given in bold and exchanges to obtain these sites underlined). The PCR-derived regions in the resulting plasmids (termed pRssB18 and pRssAB18) were sequenced.
Cultures were grown at 37°C under aeration in Luria-Bertani (LB) medium or in minimal medium M9 (Miller, 1972) supplemented with 0.1% glucose or 0.4% glycerol. Ampicillin (100 μg ml− 1) was used to grow plasmid-containing strains. For selecting transductants, various antibiotics were added as recommended (Miller, 1972). Growth was monitored by measuring the optical density at 578 nm (OD578).
Isolation of random mutations in the D58 codon of RssB
Random mutations in the D58 codon of rssB were obtained from pRssB18 and pRssAB18, using the Chameleon Double-stranded Site-directed Mutagenesis Kit developed by Stratagene. The selection primer changed a unique ScaI into a MluI site (5′-CTGTGACTGGTGACGCGTCAACCAAGTC-3′; obtained from Stratagene), whereas the mutagenic primer randomized the nucleotides of codon D58 (GAT) of rssB (5′-CCTGATGATATGTNNNATCGCGATGCCACT-3′). The D58P (CCA), D58Q (CAA) and D58stop (TGA) mutations were isolated on pRssB18, whereas the D58R (CGC) exchange was first isolated on pRssAB18. None of the amino-acid-encoding codons isolated was a rare codon. A PCR fragment extending from the 5′ end of rssBD58R beyond a unique BglII site ≈150 bp downstream of the mutated codon was reisolated (using the first primer mentioned above and a primer with the sequence 5′-GATAGAGACAGGCAAAAACCATCTCGCGC-3′), digested with EcoRI and BglII and used to replace the corresponding fragment on pRssB18. The PCR-derived part of the resulting plasmid was sequenced.
SDS–PAGE and immunoblot analysis
Sample preparation for SDS–PAGE (Laemmli, 1970) and immunoblot analysis were performed as has been described previously (Lange and Hengge-Aronis, 1994). Polyclonal sera against σS and RssB and a goat anti-rabbit IgG alkaline phosphatase conjugate (Sigma) were used. For visualization of σS bands, either a chromogenic substrate (BCIP/NBT; Boehringer Mannheim) or chemiluminescent detection (Renaissance Kit, NEN Life Science Products) was used. RssB bands could only be detected using the chemiluminescence procedure with relatively long radiographic film (Amersham Hyperfilm ECL) exposure times. Relative levels of mutant σS protein in cellular extracts were determined in relation to standard samples from strains expressing wild-type σS (always included on the same gels). Absolute amounts of σS and RssB in cellular extracts were quantified using as standards the purified proteins isolated as described (Becker et al., 1999).
β-Galactosidase activity was assayed by use of o-nitrophenyl-β-d-galactopyranoside (ONPG) as a substrate and is reported as micromoles of o-nitrophenol per minute per milligram of cellular protein (Miller, 1972). Relative in vivo activities of σS were determined as specific β-galactosidase activities measured for lacZ fusions to various σS-dependent genes (osmY, bolA, csiD and otsB) normalized for relative cellular σS levels determined by immunoblot analysis (see above).
We thank Susan Gottesmann and Felipe Moreno for bacterial strains and Daniela Fischer for expert technical assistance. Financial support was provided by the Deutsche Forschungsgemeinschaft (Priority programme ‘Regulatory Networks in Bacteria’, He 1556/5; Gottfried Wilhelm Leibniz programme), the State of Baden-Württemberg (Landesforschungspreis) and the Fonds der Chemischen Industrie.