Multiple stress signal integration in the regulation of the complex σS-dependent csiD-ygaF-gabDTP operon in Escherichia coli



The csiD-ygaF-gabDTP region in the Escherichia coli genome represents a cluster of σS-controlled genes. Here, we investigated promoter structures, sigma factor dependencies, potential co-regulation and environmental regulatory patterns for all of these genes. We find that this region constitutes a complex operon with expression being controlled by three differentially regulated promoters: (i) csiDp, which affects the expression of all five genes, is cAMP-CRP/σS-dependent and activated exclusively upon carbon starvation and stationary phase; (ii) gabDp1, which is σS-dependent and exhibits multiple stress induction like σS itself; and (iii) gabDp2[previously suggested by Schneider, B.L., Ruback, S., Kiupakis, A.K., Kasbarian, H., Pybus, C., and Reitzer, L. (2002) J. Bacteriol. 184: 6976–6986], which appears to be Nac/σ70-controlled and to respond to poor nitrogen sources. In addition, we identify a novel repressor, CsiR, which modulates csiDp activity in a temporal manner during early stationary phase. Finally, we propose a physiological  role  for σS-controlled  GabT/D-mediated γ-aminobutyrate (GABA) catabolism and glutamate accumulation in general stress adaptation. This physiological role is reflected by the activation of the operon-internal gabDp1 promoter under the different conditions that also induce σS, which include shifts to acidic pH or high osmolarity as well as starvation or stationary phase.


The general stress sigma factor σS (or RpoS) is strongly induced upon exposure of Escherichia coli cells to various stress conditions, such as starvation, hyperosmolarity, pH downshift or non-optimally high or low temperature (for a recent review on σS regulation, see Hengge-Aronis, 2002a). Many σS-controlled genes just follow this regulatory pattern, i.e. they are activated as soon as σS accumulates in the cell. Other σS-dependent genes, however, exhibit highly specific regulation, with a narrow window of expression under essentially one sort of stress condition.

The best-studied example of this second type of σS-controlled genes is the csiD gene, which is induced exclusively by carbon starvation because cAMP-CRP acts as an essential activator for σS-containing RNA polymerase (EσS) at the csiD promoter (csiDp) (Weichart et al., 1993; Marschall et al., 1998; Germer et al., 2001). This specific regulatory requirement of the csiD promoter explains its lack of activation in response to hyperosmotic shift or nitrogen starvation conditions (Marschall et al., 1998) despite strong accumulation of σS under these conditions (Gentry et al., 1993; Muffler et al., 1996).

While studying csiD regulation, we noticed that csiD is followed by several open reading frames in an arrangement that may indicate a complex operon. These are the ygaF gene and the gabD-gabT-gabP region. The latter has been known to be an operon encoding gene products involved in uptake (GabP) and metabolism of γ-amino butyric acid (GABA; GabD and GabT, i.e. succinic semialdehyde dehydrogenase and GABA transaminase, respectively, which together produce succinate from GABA) (Dover and Halpern, 1972; 1974; Metzer et al., 1979; Niegemann et al., 1993).

In order to learn more about the regulation and the physiological relevance of the csiD-ygaF-gabDTP region, we investigated promoter structures, sigma factor dependencies, potential co-regulation and environmental regulatory patterns for all of these genes. We demonstrate that this entire region represents a complex operon with two σS-dependent, but highly differentially regulated promoters (located upstream of csiD and gabD). In addition, we identify a novel repressor, CsiR, which controls the activity of the csiD promoter. We also propose, that the physiological significance of the gabDTP genes is not ‘narrow’ as even recently assumed (Schneider et al., 2002), but that these genes rather play a pleiotropic role in various stress responses, which is precisely reflected by their multiple stress activation from the operon-internal gabD promoter.


The csiD-ygaF-gabDTP region constitutes a σS-dependent operon with an operon promoter (csiDp) and an internal promoter (gabDp)

The genes csiD, ygaF, gabD and gabT directly follow each other on the E. coli chromosome without any apparent terminator sequences in between. Only gabT and gabP are separated by an intergenic region of 237 bp, which contains three partially overlapping imperfect repetitive extragenic  palindromic  (REP)  sequences  that  may  have  a regulatory effect. In order to test for co-regulation and regulatory dependencies on specific upstream sequences, we constructed a series of lacZ reporter gene fusions in each of these genes as shown in Fig. 1 (see Experimental procedures for details). For ygaF::lacZ and gabD::lacZ, constructs were made pairwise such that they either contained the csiD promoter or not. For gabT::lacZ and gabP::lacZ, constructs were made in a similar way such that they contained the gabD promoter or not (for the exact localization of the gabD promoter, see below). All reporter fusions were assayed in a merodiploid situation after integration in single copy at the att(λ) site in the chromosome.

Figure 1.

lacZ reporter fusions in different genes of the csiD-ygaF-gabDTP region. Fusions were constructed as detailed in the Experimental procedures section. Black bars indicate the regions of chromosomal DNA present in the various fusions, arrows denote the positions of the promoters present upstream of csiD and gabD. For all experiments, strains were used (strain names are given at the left margin), in which these lacZ fusions were present on λRS45 integrated in single copy at the att(λ) site in the E. coli chromosome.

Expression of all these reporter gene fusions was tested along the growth cycle in LB medium (Fig. 2). As previously observed (Weichart et al., 1993; Marschall et al., 1998; Germer et al., 2001), csiD::lacZ exhibited stationary phase induction (Fig. 2A). It should be noted that the apparent biphasic induction pattern with a transient cessation of expression for about two hours during the transition phase before cells finally stop growing completely, is reproducible and also has been seen before (Becker and Hengge-Aronis, 2001; Germer et al., 2001). ygaF::lacZ shows a very similar expression pattern, which is dependent on the presence of csiDp (Fig. 2B), indicating that ygaF is the second gene in an operon. The lacZ reporter fusions in all gab genes also exhibit stationary phase induction (Fig. 2C–E). gabD::lacZ expression is reduced by approximately 60% when the construct does not carry the csiD promoter (Fig. 2C). The remaining activity indicates the presence of a second promoter upstream of gabD, i.e. somewhere within the ygaF open reading frame. With the gabT::lacZ and gabP::lacZ fusion pairs, this promoter could be roughly localized, as expression of these fusions was basically eliminated when only the shorter 5′-regions were present (Fig. 2D and E). Taken together, these data show that the csiD-ygaF-gabDTP region is a complex operon with the upstream csiD promoter acting as the operon promoter, and an additional promoter upstream of gabD, such that the gabDTP region actually constitutes an operon within an operon.

Figure 2.

Contributions of the csiD and gabD promoter regions to the expression of all genes in the csiD-ygaF-gabDTP region. Strains carrying single copy lacZ fusions (see Fig. 1) were grown in LB medium and OD578 (open symbols) and specific β-galactosidase activities (closed symbols) were measured along the growth curve.
A. Expression of csiD::lacZ (strain JG17).
B. Expression of ygaF::lacZ in the presence (JG24; squares) and absence (JG111; circles) of csiDp.
C. Expression of gabD::lacZ in the presence (JG18; squares) and absence (JG19; circles) of csiDp.
D. Expression of gabT::lacZ in the presence (MM45; squares) and absence (MM44; circles) of gabDp.
E. Expression of gabP::lacZ in the presence (MM47; squares) and absence (MM46; circles) of gabDp.

σS regulation of the csiD promoter has been demonstrated before, and as expected from ygaF being in an operon with csiD (Fig. 2), stationary phase expression of the ygaF::lacZ fusion is strongly σS-dependent (Fig. 3). In addition, the lacZ fusions in gabD, gabT and gabP, when present in constructs that contained gabDp but not csiDp, also exhibited strong σS dependency under the same conditions (Fig. 3). This indicates that also the gabD promoter is activated by σS-containing RNA polymerase, consistent with its stationary phase induction pattern (Fig. 2). These data were also confirmed by whole genome expression analysis on microarrays, which identified csiD, ygaF, gabD, gabT and gabP as σS.-regulated genes (H. Weber and R. Henge, unpubl. results).

Figure 3.

Expression of all genes in the csiD-ygaF-gabDTP region is σS-dependent. Strains JG17, JG24, JG19, MM45 and MM47 (see Fig. 1), which carry single copy lacZ fusions in csiD, ygaF, gabD, gabT and gabP, respectively (black bars), as well as their rpoS::Tn10 derivatives (white bars) were grown in LB medium. Specific β-galactosidase activities were measured in stationary phase (data given represent the average of three independent cultures each).

For the identification of the transcriptional start site at the gabD promoter by primer extension experiments, we made use of our observation that the gab genes are not only stationary phase-inducible (Fig. 2) but also activated by shift to low pH (see below, Fig. 5). This allowed clear detection of a reverse transcript (Fig. 4) that localized the gabDp start site downstream of an appropriately located putative − 10 region (CTACGCT), in which five out of seven nucleotides (among them the most conserved ones) correspond to the σS (−10) consensus sequence, i.e. CTATACT (Espinosa-Urgel et al., 1996; Becker and Hengge-Aronis, 2001; Lee and Gralla, 2001; Hengge-Aronis, 2002b). No reverse transcript was obtained if the cells used for RNA preparation were growing at pH 7 or when a rpoS mutant background was used (Fig. 4).

Figure 5.

Differential multiple stress regulation in the csiD-ygaF-gabDTP operon. Strains carrying single copy lacZ fusions (see Fig. 1) were exposed to pH downshift or osmotic upshift (for details, see Experimental procedures) and OD578 (open symbols) and specific β-galactosidase activities (closed symbols) were measured. Data for unstressed cells (squares) and cells exposed to the respective stress conditions (circles) are shown.
A. Expression of csiD::lacZ (strain JG17).
B. Expression of gabD::lacZ (JG19).
C. Expression of gabT::lacZ (MM45).
D. Expression of the gabT::lacZ construct with a 5′-deletion upstream of the − 35 region of gabDp (MM55).
E. Expression of gabP::lacZ (MM47).

Figure 4.

Determination of the transcriptional start of the gabD promoter.
A. For the primer extension experiment, either strain MC4100 containing pMM4 (carrying the same csiD-ygaF-gabD′ insert as chromosomally present in JG19; see Fig. 1) or its rpoS::Tn10 derivative RH90 carrying the same plasmid were grown in LB medium to an OD578 of 0.7, and an aliquot of the cultures was shifted to pH 5. Samples for RNA preparation were taken 30 min after pH downshift or at an OD578 of 0.8 (log) or 4 (stat) for the cultures at neutral pH. Primer extension was performed as detailed in the Experimental procedures.
B. The relevant parts of the DNA sequence upstream of gabD are shown, with the gabD promoter identified in (A) indicated in the top line, whereas the downstream promoter (lowest line) and a putative Nac binding site were reported by Schneider et al. (2002). Δ1 and Δ2 indicate the 5′-ends of the non-coding regions present upstream of the gabT::lacZ fusions in strains MM55 and MM44 respectively.

Differential stress regulation of the csiD and gabD promoters

The csiD promoter exhibits highly specific activation under carbon starvation or stationary phase conditions suggesting that the gene products of csiD and ygaF constitute a functional unit only required under these specific conditions. The observation that the gab genes are not under the control of csiDp alone, but in addition can be activated by the operon-internal gabDp, indicates that the gab gene products, and therefore GABA catabolism, plays a more general role in E. coli physiology. This hypothesis, together with the σS-dependency of the gabD promoter, lead us to investigate the expression of the gab genes under σS-inducing stress conditions other than carbon starvation.

A downshift in external pH (from pH 7 to 5 in rich medium) as well as osmotic upshift (by the addition of 0.3 M NaCl to cells growing in minimal medium) are known as strongly σS-inducing conditions (Lee et al., 1995; Muffler et al., 1996). Both conditions do not activate csiD::lacZ expression (Fig. 5A; note that round symbols represent data from cultures exposed to the respective stress conditions). By contrast, gabD::lacZ as well as gabT::lacZ are both induced under these conditions (Fig. 5B and C). As the two constructs used contain only gabDp, but not csiDp (see Fig. 1), this regulatory pattern reflects gabDp activity. Thus, the expression of gabD and gabT essentially seem to reflect the increase in cellular σS content under these conditions.

For gabT::lacZ, we also tested a construct in which the entire 5′-region upstream of the − 35 region of gabDp was deleted, i.e. in which no activating regions upstream of the core promoter could be present. This minimal promoter construct exhibits acid and osmotic induction as well as stationary phase induction (Fig. 5D; stationary phase induction can be seen in the pH downshift control culture which remained at pH 7). Thus, multiple stress activation of the gabD promoter does not require putative activation sequences upstream of the promoter, but is likely to be a function of the core promoter alone, whose activity just reflects the cellular level of σS-containing RNA polymerase.

Interestingly, gabP::lacZ expression does not follow this pattern, even though gabP expression is dependent on the gabD promoter (Fig. 2E) and is σS-dependent (Fig. 3). Rather, gabP::lacZ activities remain at the baseline, without any apparent induction upon acid or osmotic shift (Fig. 5E). In addition, gabP::lacZ expression is (as already visible in Fig. 2) at least tenfold lower than the expression of gabD::lacZ and gabT::lacZ, even though the genetic constructs were made identically. These data indicate that there is some gabP-specific negative regulation superimposed on the control exerted by the gabD promoter, which results in generally lower expression of gabP and a lack of acid and hyperosmotic inducibility. We suspected that this downregulation of gabP could be a function of the palindromic (REP) sequences between the gabT and gabP open reading frames. Therefore, we introduced multiple point mutations in these putative hairpin regions that eliminated either single REP sequences or all of them in combination. However, even the elimination of all putative hairpin sequences increased gabP::lacZ expression by maximally 60% (data not shown). We conclude that these imperfect palindromic sequences in the intergenic space between gabT and gabP are not responsible for the much weaker expression of gabP. In order to test, whether this operon-internal down-regulation between gabT and gabP is due to post-transcriptional regulation, the originally translational gabT::lacZ and gabP::lacZ reporter fusions were converted into transcriptional fusions (see Experimental procedures for details). We indeed observed similar expression levels for the two transcriptional reporter fusions (data not shown). As post-transcriptional regulatory mechanisms were beyond the scope of the present study, this downregulation of gabP was not investigated in further detail.

The csiR (ygaE) gene product encodes a repressor that acts at the csiD promoter

Downstream of gabP, there is yet another open reading frame, ygaE (csiR), which is transcribed in the same direction. This gene encodes a regulatory protein with clear similarity to the gluconate repressor (encoded by gntR; Peekhaus and Conway, 1998). Because of the immediate vicinity of ygaE to the csiD-ygaF-gabDTP region, we tested whether its gene product is involved in the control of either the csiD promoter or the gabD promoter. We observed that a ygaE::kan insertion mutation strongly increased the expression of the csiD::lacZ fusion in stationary phase. The general pattern of stationary phase induction of csiD, however, is not altered (Fig. 6A). Thus, the ygaE gene product has a stationary phase-specific repressing effect on the csiD promoter. More precisely, in the ygaE mutant the biphasic induction pattern of csiD disappeared, i.e. the ygaE gene product may be responsible for the transient cessation of csiD-ygaF expression after the first phase of induction during entry into stationary phase.

Figure 6.

The regulator encoded by ygaE affects expression from csiDp but not from gabDp. Strains carrying the csiD::lacZ fusion (JG17; A) or the gabD::lacZ fusion (in a construct with just gabDp present; JG19; B) as well as their csiR(ygaE)::kan derivatives (MM27 and MM33, respectively) were grown in LB medium and OD578 (open symbols) and specific β-galactosidase activities (closed symbols) were measured. Expression of the reporter fusions in the csiR+ strain are shown with squares, the corresponding fusion activities in the csiR::kan derivatives are shown with circles.

By contrast, the mutation in ygaE had no effect on the expression of a gabD::lacZ fusion in a construct where only the gabD promoter was present (Fig. 6B). Thus, the gabD promoter is not a target for regulation by the ygaE gene product.

We conclude that the specific role of the ygaE-encoded regulator is to modulate the activity of the csiD promoter and thereby mainly CsiD and YgaF expression under carbon starvation conditions, whereas gabDTP-specific control via the operon-internal gabD promoter is not affected. Therefore, we suggest ‘csiR ’ as an appropriate, i.e. function-reflecting novel designation for ygaE.

Because of its location downstream of gabP (Fig. 1), also csiR (ygaE) could in principle be part of the csiD-ygaF-gabDTP operon. However, the expression of csiR::lacZ reporter gene fusions (see Fig. 1) indicates that this is not the case (Fig. 7): (i) the presence or absence of the gabD promoter did not affect the expression of csiR::lacZ, and (ii) csiR expression from a fusion construct, which carries only 344 nucleotides upstream of the csiR open reading frame, is stationary phase-induced but not σS-dependent and not autoregulated. In addition, csiR::lacZ expression is not stimulated by osmotic upshift or pH downshift (data not shown). Therefore, we conclude that csiR is not part of the csiD-ygaF-gabDTP operon, but must have a promoter of its own, which most likely is activated by σ70-containing RNA polymerase.

Figure 7.

σS-independent stationary phase activation of csiR expression. Strains carrying the csiR::lacZ reporter gene fusion with different 5′-upstream extension in different genetic backgrounds were grown in LB medium and OD578 (open symbols) and specific β-galactosidase activities (closed symbols) were measured. The following constructs were assayed: csiR::lacZ fusions carrying the entire gabDTP coding regions in their 5′ region, either with the gabD promoter (squares) or without gabDp (circles); a csiR::lacZ fusion carrying only 344 nucleotides in the non-coding 5′region (diamonds) as well as the corresponding rpoS::Tn10 (triangles), csiD::cat (inverted triangles) or csiR::kan (crosses) derivatives.


Genetic structure and regulation of the csiD-ygaF-gabDTP operon

We have shown here that the csiD-ygaF-gabDTP region in the E. coli chromosome constitutes a complex operon (Fig. 8). The csiD operon promoter (csiDp) is essential for the expression of csiD and ygaF and also affects the expression of the gab genes, whereas an operon-internal promoter upstream of gabD (gabDp) imposes an additional control onto the gabDTP region (Fig. 2). Under all conditions tested, both promoters were found to be strongly σS-dependent (Fig. 3). Their regulatory patterns, however, are strikingly different. Because of its cAMP-CRP dependency, csiDp is activated exclusively under carbon starvation conditions and stationary phase in LB (Weichart et al., 1993; Marschall et al., 1998; Germer et al., 2001). gabDp, however, is activated by all σS-inducing conditions tested, including hyperosmotic and acidic shifts besides starvation and stationary phase in LB (Figs 2 and 5). Correct regulation of a construct devoid of all sequences naturally present upstream of the − 35 region of gabDp suggests that no upstream activating sequences are involved and the core promoter region is sufficient for this multiple stress induction (Fig. 5D). Thus, regulation of gabDp just follows σS accumulation under these conditions. This is consistent with its close similarity to the typical σS consensus promoter (Espinosa-Urgel et al., 1996;Becker and Hengge-Aronis, 2001; Gaal et al., 2001). In addition, alkaline induction of gabT has been reported (Stancik et al., 2002). Our experiments shown here (Fig. 5) were either done in buffered minimal medium, or when LB was used, the pH remained between 6.8 and 7 during the course of the experiment (data not shown; an overnight culture in LB, however, exhibits a pH of approximately 9). Thus, multiple stress induction of gabDp as shown here is not caused by alkalinization. The mechanism of alkaline induction of the gab genes has yet to be clarified.

Figure 8.

Differential and multiple stress regulation at three different promoters in the csiD-ygaF-gabDTP region. The genetic structure of the csiD-ygaF-gabDTP region is summarized. Filled arrows denote the position of the three relevant promoters (csiDp, gabDp1, gabDp2 from left to right). The upper part of the figure summarizes multiple environmental signal input as well as the relevant regulators that cooperate in the regulation of csiD-ygaF-gabDTP operon. The lower part of the figure indicates the role of the CsiR(YgaE) regulator in the control of csiDp activity. For further details, see Discussion.

σS regulation of the gab genes has also been indicated by two other studies that used chromosomal reporter gene fusions (Schellhorn et al., 1998; Baca-DeLancey et al., 1999), but was disputed by a more recent report (Schneider et al., 2002). However, Schneider et al. (2002) could not identify a gab transcript in vivo (as we could in stationary phase and acid-shifted cells; Fig. 4), and therefore relied on primer extension experiments with a transcript obtained in vitro with σ70-containing RNA polymerase. This σ70-recognized putative promoter region is located 261 bp downstream of the σS-dependent gabDp identified here in vivo (Fig. 4B), and is still present on our gabT::lacZ and gabP::lacZ fusion constructs with the shorter 5′-upstream regions that do not carry the gabD promoter identified here. Our observation that these 5′-deletions strongly reduce expression (Fig. 2D and E) indicates that the σ70-controlled putative promoter identified by Schneider et al. (2002) does not play a role under the gab-inducing conditions that we tested. Also primer extension experiments did not reveal a corresponding transcript under all these conditions (data not shown). However, it is possible that this putative and probably weak promoter is activated under other conditions, such as the presence of certain poor nitrogen sources (Schneider et al., 2002). Also, a potential binding site for the nitrogen regulator Nac is located just upstream and a nac mutation affects the expression of the gab genes (Zimmer et al., 2000; Schneider et al., 2002). In conclusion, we propose that most likely the csiD-ygaF-gabDTP operon is controlled by three promoters (Fig. 8): (i) csiDp, which is cAMP-CRP/σS-dependent and therefore specifically carbon starvation/stationary phase-induced; (ii) gabDp1, which is dependent on σS alone and whose activity thus reflects multiple stress induction of σS; and (iii) a putative weak gabDp2, which seems to be Nac/σ70-controlled and therefore could be activated by certain poor nitrogen sources.

In addition to this complex and differential positive control at three promoters, there is evidence for several negative control mechanisms. One of these mechanisms operates via a repressor, CsiR (YgaE), which is expressed from the gene downstream of gabP (for further discussion, see below). Additional negative control mechanisms seem to affect the expression of ygaF and gabP, as both exhibit significantly lower expression than their respective promoter-proximal genes, i.e. csiD and gabT. This downregulation is particularly apparent for gabT, as this gene also does not exhibit multiple stress induction as the gabD and gabT genes (Fig. 5), although gabP expression requires the gabD promoter (Fig. 2E) and is σS-dependent (Fig. 3). The palindromic (REP) sequences between gabT and gabP are not involved as putative leaky terminator sequences, as their elimination did not result in a strong increase in gabP expression (data not shown). A comparison of the expression of transcriptional and translational fusions, both in gabT and gabP, indicated that this downregulation is a result of post-transcriptional control (our unpublished data), which was beyond the scope of the present study, but which in further investigations may reveal an interesting fine-regulation superimposed on general σS control of an operon. In the case of ygaF, operon-internal downregulation does not alter the general regulatory pattern. Preliminary Northern blot analyses suggest that in this case mRNA processing may play a role (our unpublished data).

Physiological significance of multiple stress activation of the gabDp1 promoter

Even though the biochemical activities of the gabDTP gene products in uptake and catabolism of GABA have been clear for many years, their role in whole cell physiology has remained mysterious, as E. coli K-12 cannot even use GABA as a sole source of nitrogen (unless the cells carry not further characterized ‘gabC  ’ mutations which lead to increased GabD and GabT activities (Metzer et al., 1979, see below). More recently, it was suggested that GABA catabolism may play a role in putrescine degradation, i.e. polyamine catabolism (Schneider et al., 2002).

Based on our observation that gabDTP expression is σS-dependent and multiple stress-induced, we would like to suggest a broader physiological role for the GabT/GabD pathway that includes putrescine/GABA catabolism within the larger context of multiple stress adaptation (Fig. 9). This adaptation may involve the ability to establish and maintain high internal glutamate levels under stress conditions. Glutamate accumulation has long been known for hyperosmotically stressed cells (McLaggan et al., 1994; Yan et al., 1996), high glutamate levels stimulate the activity of σS-containing RNA polymerase (Ding et al., 1995; Ohnuma et al., 2000) and the cellular glutamate levels are lower in rpoS mutants (Tweedale et al., 1998). Three enzymatic systems are involved in glutamate production, but their relative contributions under various stress conditions have not been clarified: (i) glutamate dehydrogenase (GDH), which appears to be the vegetative ‘house-keeping’ enzyme (Csonka et al., 1994; Helling, 1994) (ii) the glutamine synthase/glutamate synthase cycle (GS/GOGAT), which is induced under conditions of nitrogen limitation (Csonka et al., 1994; Helling, 1994), and (iii) the GabT/D system, which is stress-induced under σS-control, but whose role in glutamate production has not been considered so far.

Figure 9.

A model for the role of GABA catabolism and glutamate production by GABA transaminase (GabT) and succinic semialdehyde dehydrogenase (GabD) and/or glutamate dehydrogenase (GDH) under different stress conditions. Metabolic reactions and intermediates that are relevant under various stress conditions (as indicated in A, B, and C) are shown in black, non-relevant parts of metabolism are shown in grey. For a detailed discussion, see text.

Upon pH downshift, glutamate plays a crucial role in proton scavenging via two glutamate decarboxylase isoenzymes (GadA and GadB; Fig. 9A and B). Both decarboxylases as well as a glutamate/GABA exchange carrier (GadC) are induced in a σS-dependent manner (de Biase et al., 1999; Castanie-Cornet et al., 1999; Castanie-Cornet and Foster, 2001; Ma et al., 2002). Two pH downshift situations have to be distinguished. If external glutamate is present at significant concentrations (Fig. 9A), the Gad system essentially operates on its own, i.e. glutamate is taken up, its decarboxylation results in proton scavenging, and the resulting GABA is excreted. The GabT/GabD system and probably also GDH are present but are unlikely to be relevant for glutamate production due to simple mass action (internal glutamate is built up via uptake, GABA is excreted). If, however, glutamate is not provided externally, GabT/GabD could cooperate with the GadA/B decarboxylases in a cyclic extension and partial bypass of the tricarboxylic acid (TCA) cycle which allows proton scavenging (Fig. 9B).

In starved or hyperosmotically shifted cells (Fig. 9C), the same extension of the TCA cycle could operate, but play a different physiological function. Under these conditions, the cellular levels of putrescine are strongly decreased (Tabor and Tabor, 1985; Tweedale et al., 1998). Although the fate of putrescine has not been studied systematically under these conditions, it is likely that this decrease in putrescine concentration involves excretion (Kashiwagi et al., 1992; Schiller et al., 2000) as well as catabolism, which leads to the synthesis of GABA, which then is further metabolised to succinate by the GabT/GabD system. This also results in the production of glutamate, which would accumulate, since the glutamate decarboxylases (GadA/B) are low pH-activated only (Capitani et al., 2003). High glutamate levels, however, are crucial for σS-containing RNA polymerase (Ding et al., 1995; Ohnuma et al., 2000), i.e. by inducing the gab system, σS may boost its own activity.

Based on reduced fitness of a gdhA mutant in competition experiments, a role for GDH in glutamate synthesis under conditions of energy limitation has been proposed (Helling, 1994). We propose that the GabT/GabD system and GDH probably function redundantly in glutamate production, at least initially upon exposure to high osmolarity or carbon starvation. In the longer run, however, GDH is degraded in starved cells (Maurizi and Rasulova, 2002; Weichart et al., 2003), whereas in osmotically stressed cells, it is transcriptionally downregulated (Weber and Jung, 2002). This would explain why gdhA mutants are not impaired in osmotic adaptation (McLaggan et al., 1994), and why, on the other hand, also gabT mutants do not exhibit severe defects in adaptation to acid or hyperosmotic stress (our preliminary results). Experiments with double mutants are currently under way. An interesting difference between glutamate production by the GabT/GabD system and GDH is, that the former can basically recruit carbon atoms from the ‘metabolic periphery’, i.e. from putrescine/GABA, whereas the latter has to withdraw carbon atoms from central metabolism, i.e. the TCA cycle, which at least in carbon starved cells may result in a competitive disadvantage. In order to provide further evidence for these hypotheses, competition experiments will have to be done with gab and gdhA mutants exposed to various stress conditions.

Role of the CsiR (YgaE) regulator

We have also shown here, that the CsiR regulator imposes a negative control specifically on the csiD promoter, whose activity is already restricted to carbon starvation conditions due to cAMP-CRP/σS-co-dependency. Recently, this regulator was also designated as ‘GabC’ (Schneider et al., 2002), as ‘gabC′ mutations were identified more than 20 years ago that affected the activities of the GabD and GabT enzymes (Metzer et al., 1979). However, we propose the designation ‘CsiR′ for this repressor as it controls csiDp and not gabDp (Fig. 6) and therefore is not a regulator specifically controlling GABA metabolism. Moreover, the original gabC mutations were mapped upstream of the gabDTP region (Metzer et al., 1979; Bartsch et al., 1990), whereas csiR (ygaE) is located downstream. The exact nature of these gabC mutations has never been clarified. They may have been promoter-up mutations in csiD or mutations affecting a binding site for CsiR in the csiDp region. csiR (ygaE) is not part of the csiD-ygaF-gabDTP operon, as its expression is not dependent on the gabDp1 promoter nor on σS (Fig. 7). csiR (ygaE) seems to be one of the few known stationary phase-inducible genes expressed by σ70-containing RNA polymerase.

CsiR appears to fine-modulate the expression of CsiD/YgaE in starved cells, as a csiR mutant not only shows increased expression but the reproducible biphasic induction pattern of csiDp disappeared (Fig. 6). CsiR is a homologue of the gluconate repressor (Peekhaus and Conway, 1998) and therefore likely to be allosterically regulated by some small compound. It may be that the cellular content of this substance fluctuates during the several-hour transition period when cells enter into stationary phase. Such a fluctuation may produce the biphasic induction of the csiD promoter. As a mutation in csiD upregulates the expression of the entire csiD-ygaF-gabDTP region (J. Germer and R. Hengge, unpubl. data), the CsiD/YgaF proteins could be involved in the production or catabolism of this compound (YgaF exhibits similarity to various oxidoreductases, whereas the CsiD sequence does not provide any clue to its biochemical function). This substance is obviously not GABA, because the external addition of GABA does not have an effect on the expression of csiD (data not shown) or the gab genes (Schneider et al., 2002; data not shown).

It should also be noted that CsiR is the first regulator identified that seems to fine-modulate the expression of a starvation-induced gene in a specific temporal manner during entry into stationary phase or during the first few hours of carbon starvation. Such fine-modulation and perhaps even a precise temporal genetic expression programme for many proteins was suggested many years ago by the analysis of global protein synthesis on 2D-gels (Groat et al., 1986), but the molecular basis of this temporal control of many proteins during the first hours of starvation has never been studied. The analysis of the regulation and molecular function of CsiR will shed light on these processes and is currently in progress.

Experimental procedures

Bacterial strains and growth conditions

The strains used in this study are derivatives of MC4100 (Casadaban, 1976). Mutant alleles of rpoS, csiR or gabT were introduced into various fusion strains by P1 transduction (Miller, 1972). Specifically these alleles are rpoS359::Tn10 (Lange and Hengge-Aronis, 1991), csiR::kan and gabT::kan (this work). Reporter gene fusions were constructed and introduced into the chromosome as described in detail below.

Standard batch cultures were grown at 37°C under aeration in a rotary shaker in Luria–Bertani (LB) medium. Ampicillin (100 mg ml−1) was used to grow plasmid-containing strains. For osmotic shift experiments, cultures were grown in M9 (Miller, 1972) with glycerol (0.4%) for three or more generations before they were divided into two aliquots to one of which 0.3 M NaCl was added. For pH downshift experiments, cultures were grown in LB for four or more generations before they were divided into two aliquots (at an OD578 of 0.7), to one of which 145 mM 4-N-morpholinethanesulphonic acid (MES) was added. Growth was monitored by measuring the optical density at 578 nm (OD578).

DNA manipulations

For restriction digests, ligation, transformation and agarose gel electrophoresis, standard procedures were followed (Silhavy et al., 1984; Sambrook et al., 1989). Plasmid DNA preparations and recovery of DNA from agarose gels were performed with the Qiagen kits. Bacterial chromosomal DNA was prepared as described previously (Silhavy et al., 1984).

Construction of the csiR::kan and gabT::kan mutants

For the construction of csiR and gabT knockout mutations we used one-step inactivation of chromosomal genes in Escherichia coli using PCR products (Datsenko and Wanner, 2000). As a template for the PCR reaction pKD4 was used (Datsenko and Wanner, 2000). For the csiR mutant, the PCR product was generated with the primers 5′-GGCAAAAAAC ACCCGT TCATAATACGCGCTGATCATGATCAGGAGTCAC ACCGTGTAGGCTGGAGCTGCTTC-3′ and 5′-GCTAAA AAGCCCGGA TAAAATTTATCCGGGGCAAGTGTTGCGTAT TCCGGCATATGAATATCCTCCTTAGT-3′). For the gabT mutant, the PCR product was generated with the primers 5′-AGAAA TCAAATATATGTGCATCGGTCTTTAACTGGAGA ATGCGAGTGTAGGCTG-3′ and 5′-GGCCGGACAAGA CGCGCAGCGT CGCCTCCGGCA TAGGAGCGGCGCATAT GA-3′. The mutants were constructed and tested as described elsewhere (Datsenko and Wanner, 2000).

Construction of lacZ reporter fusions and their transfer into the chromosome

Different chromosomal lacZ fusions to genes in the csiD region (Fig. 1) were isolated using the fusion vector pJL29 (Lucht et al., 1994), which is a pMLB1034 (Silhavy et al., 1984) derivate carrying the polylinker of pNM481 (Minton, 1984). The inserts were generated by PCR using MC4100 chromosomal DNA as a template, digested with the corresponding enzymes (Table 1) and cloned into the fusion vector, which was digested with the same enzymes. For the conversion of translational fusions into transcriptional fusions, the plasmids containing the former were digested with HindIII (Table 1) and PvuII (which cuts in lacZ), and the fragment containing the fusion joint was replaced by a corresponding fragment from pFS1 (laboratory collection), which contains stop codons in all three reading frames and provides a Shine–Dalgarno sequence (TAAGGAGG) and a start codon for lacZ. PCR-derived parts of all plasmids carrying fusions were sequenced (GATC). All constructs were crossed onto λRS45 or λRS74, followed by lysogenization into MC4100 according to the method of Simons et al. (1987). Single lysogeny was tested by a PCR method (Powell et al., 1994).

Table 1. . Oligonucleotide primers used for construction of lacZ fusions.
gabT::lacZ (Δ > −35) gabD.d-35EcoRI 5′-CGGGAATTCGCCGGGTGCTGC-3′
gabT::lacZgabD p) gabD.d + 10EcoRI 5′-GAAGAATTCGCTGTGCAAAAGCGG-3′
gabP::lacZgabD p) gabD.d + 10EcoRI 5′-GAAGAATTCGCTGTGCAAAAGCGG-3′
csiR::lacZgabD p) gabD.d + 10EcoRI 5′-GAAGAATTCGCTGTGCAAAAGCGG-3′
csiR::lacZgabDTP )like csiR::lacZgabDp) but cloned with native SwaI-site 355 bp upstream of csiR

RNA preparation and primer extension

Total RNA was prepared by hot phenol extraction of cells growing in LB medium. To get a strong primer extension signal, MC4100 (or its rpoS mutant derivative RH90) containing plasmid pMM4 (a pJL29 derivate carrying the gabD::lacZcsiDp) fusion (Table 1) was used for RNA isolation. For the detection of the transcriptional start site upstream of gabD by primer extension, 5′-GGGGAGGGCGCATTGCAGGTG-3′ was used as a primer and the reactions were performed at 43°C for 30 min using 20 µg of total RNA and 10 U avian myeloblastosis virus (AMV) reverse transcriptase (Roche) and alpha-[35S]-thio-dATP (1000 Ci mmol−1; Amersham), followed by a 30-min chase with all four nucleotides. As a reference, double-strand sequencing reactions were performed with the same primer using the T7 Sequencing Kit from USB Corporation. Reactions were stopped by addition of 20 mM EDTA in formamide containing xylene cyanol and bromophenol blue. After heating to 65°C for 2 min, samples were subjected to electrophoresis on 7% sequencing gels, fixed and dried before being autoradiographed and quantified using a FLA-2000G imager (Fuji Photo Film, Japan).

β-galactosidase assay

β-galactosidase activity was assayed using 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).


Financial support for this study was provided by the Deutsche Forschungsgemeinschaft (Gottfried-Wilhelm-Leibniz program and He1556/11–1), the state of Baden-Württemberg (Landesforschungspreis) and the Fonds der Chemischen Industrie.