Genes rcsC and rcsB form a two-component system in which rcsC encodes the sensor element and rcsB the regulator. In Escherichia coli, the system positively regulates the expression of the capsule operon, cps, and of the cell division gene ftsZ. We report the identification of the promoter and of the sequences required for rcsB-dependent stimulation of ftsZ expression. The promoter, ftsA1p, located in the ftsQ coding sequence, co-regulates ftsA and ftsZ. The sequences required for rcsB activity are immediately adjacent to this promoter.
Evidence for post-transcriptional regulation of ftsZ has been reported. Gene stfZ produces an ftsZ antisense RNA, and inhibits cell division when cloned on a high copy number plasmid (Dewar and Donachie, 1993). ftsZ mRNA is cleaved by endonuclease Rnase E, and this significantly affects the level of transcripts (Cam et al., 1996). In Caulobacter crescentus, the stability of FtsZ protein is cell cycle regulated (Kelly et al., 1998). However, the number of promoters found upstream of ftsZ leaves a great potential for regulation of the gene at the level of transcription initiation. Despite the fact that the three kilobase pairs of DNA upstream of ftsZ only account for 34% of the transcription of this gene in rich medium (Flärdh et al., 1998), at least five promoters have been identified in this region (Aldea et al., 1990; Cam et al., 1996; Ballesteros et al., 1998). Two of these (ftsQ2p, ftsQ1p) are within the ddlB coding sequence and three others within the ftsA coding sequence (ftsZ4p, ftsZ3p, ftsZ2p). A sixth promoter (ftsAp), located in the ftsQ coding sequence, has been proposed but not mapped precisely (Dewar and Donachie, 1990; Flärdh et al., 1997). The most distal promoter, ftsQ2p, is activated by SdiA, a transcription factor homologous to quorum-sensing system activators (Wang et al., 1991; Garcia-Lara et al., 1996; Sitnikov et al., 1996). The next promoter, ftsQ1p, is recognized by σs, the stationary phase-specific sigma factor (Sitnikov et al., 1996; Ballesteros et al., 1998). No regulatory elements have been reported for ftsZ2-4p promoters.
Gervais et al. (1992) showed that gene rcsB suppresses the temperature-sensitive mutation ftsZ84, and that the suppression results from increased transcriptional output of the ftsZ2-3p region. Gene rcsB is part of a two-component system, rcsC–rcsB, initially reported to regulate the synthesis of colanic acid. RcsB is thought to be activated through the transfer of a phosphate group from either its cognate sensor, RcsC, or from another protein RcsF. The signal(s) recognized by either RcsC or RcsF have not been identified (reviewed by Gottesman, 1995). Sledjeski and Gottesman (1996) reported that an osmotic shock could transiently induce the expression of colanic acid genes (genes cps) in an rcsC–rcsB-dependent fashion. Overproduction of DjlA, a transmembrane protein with a DnaJ-like domain, can also activate the synthesis of the capsule via the rcsC–rcsB pathway (Clarke et al., 1997; Kelley and Georgopoulos, 1997). Because this activation is abolished by point mutations in the DjlA transmembrane segment, Clarke et al. (1997) suggested that it might require a direct contact between DjlA and the membrane protein RcsC. Stimulation by RcsB of genes involved in capsule synthesis is potentiated by another protein, RcsA, whereas Gervais and Drapeau (1992) reported that the ability of RcsB to suppress the growth defect of the ftsZ84 mutation does not require RcsA, suggesting that RcsA does not stimulate ftsZ expression. The sequences required for RcsA/RcsB activation of the Erwinia amylovora cps genes have been reported (Kelm et al., 1997; Wehland et al., 1999), those for RcsB activation of the ftsZ gene have not yet been defined.
In this study, we report the identification of the division operon promoter regulated by RcsB, and define the sequence required for RcsB-dependent activation.
An engineered hexahistidine-tagged RcsB protein mimics wild-type RcsB activity
Overexpression of transcriptional factor RcsB increases the expression of capsule synthesis, cps, and cell division, ftsZ, genes. Increased expression of cps genes can be detected either by observing the mucoid phenotype of colonies on solid media, or by measuring the expression of a cps–lacZ operon fusion (see Gottesman, 1995). The effect of RcsB on ftsZ expression can be monitored using a ‘ddlB-ftsQ-ftsA-ftsZ’–lacZ operon fusion (Gervais et al., 1992). Based on these reporter gene systems, an engineered form of RcsB, generated by the addition of a hexahistidine peptide to the N-terminal part of the protein (plasmid pHis6rcsB, see Experimental procedures) showed an activity similar to that of the wild-type RcsB protein (data not shown). The data reported below were obtained using the modified version of RcsB.
An RcsB-activated promoter is located within the ftsQ coding sequence
Gervais et al. (1992) established that an RcsB-activated promoter is present within the ‘ddlB-ftsQ-ftsA-ftsZ’ chromosomal moiety of pTGV15 (Fig. 1). To identify this promoter precisely, various fragments from the region were fused to lacZ in ColEI-derived cloning vectors, and β-galactosidase synthesis was monitored in the absence or presence of His6-rcsB expression. The results (Fig. 1) indicated that a 67 bp fragment extending from co-ordinates 1050–1116 was sufficient to confer RcsB-dependence (pFAB9). Typically, LacZ activity was nearly 15 units with the cloning vector alone, 300 units for pFAB9 in the absence of lacZp-His-rcsB induction and 15 000 units when rcsB was induced. Deletion of sequences downstream of co-ordinate 1089 (pFAB3) reduced LacZ activity to the vector level, indicating that the promoter was destroyed by this deletion. Conversely, deletions removing 5′ sequences down to co-ordinates 1060, 1065 or 1073 (pFAB10, pFAB11 and pFAB5 respectively; Fig. 1) abolished RcsB-dependent activation, but did not affect the basal level of transcription. Taken together, these results indicate that the 1073–1116 segment is sufficient for transcription.
The exact location of the promoter was determined by primer extension analysis, which indicated the presence of a transcript starting at an A residue at position 1111 (co-ordinate 103 561 of E. coli K-12 sequence, version M52). The transcript was detected only when rcsB was overexpressed (Fig. 2). This result suggested that the basal promoter consisted of a −35 box, TTGAAT, separated by 17 nucleotides from a −10 box, TGGAAT (Fig. 1). In agreement with this assignment, mutagenesis of the putative −35 box (pFAB7, Fig. 1) reduced LacZ activity to 70 units, indicating promoter inactivation. This promoter will be referred to hereafter as ftsA1p. Basal LacZ activity produced from the fusion containing the ‘ddlB-ftsQ-ftsA-ftsZ’ region is about 6000 units (pTGV15), whereas that arising from ftsA1p alone is about 300 units. Therefore the unstimulated ftsA1p promoter is responsible for about 5% of the ftsZ transcription initiated in this region.
The sequence required for RcsB activation is immediately adjacent to the −35 box
The deletion analysis, shown in Fig. 1, indicated that a 25 bp segment upstream of the ftsA1p−35 box was sufficient for RcsB activity. Stimulation was no longer observed when this segment was reduced to 15 bp or less (pFab10, pFab11, pFab5, Fig. 1). As indicated in Fig. 1, this 25 bp segment is constituted by an almost perfect 10 bp inverted repeat separated by five bases. This feature served as a guide for the mutational analysis.
Changing the three most promoter-distal bases from CCT to GGA had no effect on ftsA1p activation (pFAB17, Fig. 3). Therefore, a 22 bp segment upstream of ftsA1p is sufficient for RcsB activity. In contrast, changing the three most proximal bases from TGG to ACC abolished rcsB-dependent stimulation (pFAB33). In order to identify the bases important for RcsB activity, single complementary base substitutions were introduced between positions −37 and −59 (Fig. 3). Substitutions affecting RcsB activity could be divided into three classes. Class I mutations reduced activity to 10–50% that of wild type. Class II substitutions left less than 10% of the wild-type activity. Class III substitutions increased stimulation above that of wild-type RcsB.
Five mutations, at positions −38, −41, −43, −44 and −48 (pFAB31, pFAB28, pFAB8, pFAB25 and pFAB41 respectively), belonged to class I. Four class II mutations affected bases at positions −39, −40, −46 and −47 (pFAB30, pFAB29, pFAB24 and pFAB42 respectively). A single mutation at position −50 (pFAB39) belonged to class III. Classes I and II ‘down’ mutations indicate that critical bases for activation are located between positions −38 and −48. However, the isolation of the ‘up’ mutation at position −50 indicates that bases beyond −48 can limit RcsB activity. Furthermore, although single substitutions beyond position −50 had no effect, two results indicated that the sequence required for RcsB activity extends beyond position −50. First, a five-base insertion between positions −51 and −52 abolished RcsB activity (pFAB35, Fig. 3). Second, the deletion analysis (Fig. 1) showed that RcsB activity is lost in constructs deleted beyond position −52 (pFAB10), indicating that the sequence required for RcsB-dependent activation goes beyond that position.
The fusion of pFAB34 contains five bases inserted between position −37 and the −35 box of ftsA1p. The LacZ level of this fusion upon induction of rcsB expression was only 6% that of the wild-type fusion (Fig. 3). This suggests that the position of the RcsB activating sequence with respect to the promoter is critical to the activation process.
RcsB stimulates ftsA1p promoter in vitro
In order to demonstrate the direct effect of RcsB on the ftsA1p promoter, we set up a mixed templates in vitro transcription assay, with one template bearing ftsA1p promoter and another lacUV5p promoter (see Experimental procedures). When tested separately, these ftsA1p and lacUV5p templates supported the synthesis of transcripts of the expected size, i.e. 104 nt and 55 nt respectively (data not shown). Transcripts with the same size were also observed in the mixed template assay with RNA polymerase alone (Fig. 4, lane 1). The efficiency of the transcription of ftsA1p by the RNA polymerase holoenzyme increased when increased amounts of purified RcsB protein were added to the reactions, whereas that of lacUV5p did not (Fig. 4, lanes 2–4). With 106 μM RcsB, transcription from ftsA1p was stimulated fivefold (Fig. 4, lane 4). Therefore, RcsB specifically and directly stimulates the promoter ftsA1p. An additional transcript of 113 nt (asterisk in Fig. 4) increased in the same fashion as the 104 nt ftsA1p run-out product. The origin of this transcript is unknown.
rcsB–rcsC is a two-component system involved in the activation of capsule synthesis, cps, and cell division, ftsZ, genes. We have identified a promoter of ftsZ, ftsA1p, whose activity can be stimulated at least 50-fold by RcsB in vivo. This promoter, located in the ftsQ coding sequence, explains the transcription activity already reported but not completely characterized by Dewar and Donachie (1990) and Flärdh et al. (1997). We found that basal transcription from the promoter ftsA1p contributes about 5% of the ftsZ transcription originating from the ftsQ–ftsA region, in agreement with the estimate of Flärdh et al. (1997). If one considers that the contribution of that region to total ftsZ transcription is about 20% (Flärdh et al., 1997, 1998), this represents about 1% of total ftsZ transcription. When RcsB is overexpressed, overall transcription from the ftsQ–ftsA region increases five to sixfold (Gervais et al., 1992). This level of stimulation was obtained with a lacZ fusion either in single copy (λTGV3, Gervais et al., 1992; this work, data not shown) or on a multicopy vector (pTGV15, Fig. 1), indicating that, in these experiments, RcsB is not limiting for ftsA1p. Therefore, upon induction of rcsB expression, total transcription of ftsZ should increase at least twofold. Wang et al. (1991) have shown that an 1.7 increase of FtsZ concentration is sufficient to make a cell resistant to the cell division inhibitor MinCD. Ward and Lutkenhaus (1985) showed that 4.2% of the cell population presents polar divisions when FtsZ concentration increases 2.7-fold. Therefore, an RcsB-dependent twofold increase of ftsZ expression might be very significant for the cell physiology. However, to address fully the contribution of ftsA1p to ftsZ regulation and to the cell physiology, it will be necessary to study ftsA1p regulation under conditions in which the RcsC–RcsB system has been activated by its natural signal, which has not yet been identified. Because ftsA1p is located in the ftsQ coding sequence, both ftsA and ftsZ are co-regulated by RcsB, and the intracellular ratio of the corresponding proteins should thus stay constant upon induction. The constancy of the ratio is important for correct cell division (Dai and Lutkenhaus, 1992; Dewar et al., 1992; Flärdh et al., 1998). Begg et al. (1998) reported that increased levels of both proteins stimulate cell division, leading the cells to divide earlier in the cell cycle. Indeed, we have observed that when rcsB is overexpressed, cells become smaller, suggesting that they divide earlier in the cell cycle (data not shown).
Mutations at nine positions in the ftsA1p upstream sequence resulted in a significant reduction of RcsB stimulation. These bases are clustered between positions −38 and −48 relative to the ftsA1p transcription start, and might define an RcsB binding site. However, the left border of the binding site is beyond the −48 position and includes at least position −52, because (i) an up mutation at position −50 has been isolated, (ii) a 5 bp insertion between positions −51 and −52 decreases RcsB activity dramatically, and (iii) a deletion leaving only 15 bases upstream of −35 also strongly affects RcsB activity. As far as the right-hand border of the binding site is concerned, the effect of mutations on activation is difficult to distinguish from that on RNA polymerase binding (e.g. pFAB7, Fig. 1), and thus an extension to the right is not ruled out (see below).
Our data can be compared with others in which RcsB is implicated in transcription activation. RcsB-dependent activation of the ams genes of E. amylovora requires RcsA as a co-activator. Wehland et al. (1999) found that a 23 bp fragment, located within the segment defined by the former study of Kelm et al. (1997), is sufficient for binding of the RcsA–RcsB heterodimer to ams DNA. Using an electrophoretic mobility shift assay with various mutated templates, they restricted the sequence to 18 bp (underlined in Fig. 5) and defined an optimized sequence BDTRVGAAWAWTSYGRGR (B = C, G or T, D = A, G or T, R = A or G, V = A, C or G, W = A or T, S = C or G, Y = C or T). The same group also identified a homologous RcsA/RcsB binding site upstream of the P. stewartii cps operon (underlined in Fig. 5; Wehland et al., 1999). Ebel and Trempy (1999) reported that E. coli RcsA activates its own expression. This activation is potentiated by RcsB. By deletion analysis, they defined a 25 bp ‘RcsA box’ (underlined in Fig. 5) required for the rcsA activation. This RcsA box shows homology to the E. amylovora ams and P. stewartii cps regulatory sequences (Fig. 5). Stout (1996) showed that a 500 bp fragment containing the E. coli cps gene promoter region is able to titrate cellular RcsA and RcsB proteins. A sequence homologous to the regulatory sequence described above was found in this 500 bp fragment (Kelm et al., 1997; Ebel and Trempy, 1999).
In contrast to these four systems, activation of fts genes by RcsB is independent of RcsA (Gervais et al., 1992; F. Carballès, J. P. Bouché and K. Cam, unpublished data), and one may expect to find within the ftsA1p binding site some features of symmetry characteristic for the binding of an RcsB homodimer. Indeed, the 15 base pairs defined by point mutations and deletions align well with the ams core sequence (9/15 identities, 12/15 identities with the ams optimized sequence). Down mutations affecting ftsA1p activation are arranged symmetrically around a position that superimposes the middle of the ams optimized core sequence. Therefore, we propose that the ftsA1p RcsB binding site consists of two symmetrical 9 bp half sites, with a common sequence WWSHAGATK (W = A or T, S = 0, C or G, H = A, C or T, K = G or T). The mixed RcsA/B ams sequence agrees with this consensus in 14/18 positions, whereas the agreement between the cps or rcsA sequence and the consensus is more limited (12/18 or 11/18; randomness: 7/18). These differences may reflect differences between the optimal recognition sequences of RcsA and RcsB, or the fact that the proposed consensus is based on a unique, non-optimized sequence.
Transcription activators favour the formation of an active transcription initiation complex in promoters deficient at one step or another of initiation. This may be achieved either by modifying the DNA topology or by direct contact between the activators and the transcription machinery (see review by Rhodius and Busby, 1998). The closeness of the RcsB activation site to the −35 box suggests that the formation of an active complex might require contact between the activator and the transcription machinery. Although this work has demonstrated the direct effect of RcsB on ftsA1p stimulation, detailed analysis will be required to understand the mechanism of activation.
Strain and cloning vectors
The study was performed using Escherichia coli strain MC1061 Δ(lacIZYA)X74. Transcriptional fusions were cloned in vectors pRS550 and pRS551 (Simons et al., 1987). rcsB was cloned in pIM10, which allows gene transcription from the lacUV5p promoter and translation from the Shine–Dalgarno sequence of the T7 phage φ10 gene. Vector pIM10 is a p15A derivative and confers resistance to spectinomycin and kanamycin (Cornet et al., 1994)
Construction of hexahistidine-tagged rcsB expression plasmid (pHis6rcsB)
The fragment for cloning was generated using PCR amplification with MC1061 DNA as template. The upstream primer (5′ GGAATTCCAT-ATG-CAT-CAC-CAT-CAC-CAT-CAC-ATG-AAC-AAT-ATG-AAC-G) added codons for one Met and six His residues upstream of the rcsB ATG codon. The primer is complementary to the rcsB sequence from co-ordinate −1 to + 16, and provides an additional NdeI restriction site at the 5′ end. The sequence of the downstream primer (5′ CGCGGATCCGGATAAGCGTTTACGCG) was complementary to the rcsB sequence between co-ordinates 828 and 845. The primer contains an additional BamHI restriction site at its 5′ end. The PCR generated fragment was cloned between the NdeI and BamHI restriction sites of pIM10.
The constructions used in this study are listed in Table 1. Most of the transcriptional fusions were made into cloning vectors pRS551 or pRS550 opened at the BamHI and EcoRI sites. The relevant features of these plasmids are the presence of four strong transcription terminators upstream of the cloning site and of lacZ as reporter gene. The plasmids confer resistance to kanamycin and ampicillin. Cloned fragments were either generated by PCR amplification with primers to which BamHI or EcoRI restriction sites were added at their 5′ end, or subcloned using natural restriction sites. Base substitutions or insertions were generated by PCR with mutagenic primers. The presence of the mutations was checked by sequencing.
Table 1. . Transcriptional fusions of ‘ddlB–ftsQAZ’ fragments to lacZ gene. a. 5′- and 3′-ends co-ordinates of cloned ′ddlB-ftsQAZ’ fragments. Co-ordinates are the same as those used in Flärdh et al. (1997).b. Artificial restriction sites brought by the PCR primers used for cloning.c. Yes or No means the presence or not on the construction of an active RcsB site.d. A1, Z4, Z3 and Z2 stand for the ftsA1p ftsZ4p ftsZ3p,and ftsZ2p promoters respectively.
All assays were carried out in the MC1061 background. Cultures were made in Luria–Bertani (LB) broth medium at 37°C. Overnight cultures were diluted 1000-fold and grown for five generations, then diluted 20-fold in prewarmed medium containing 500 μM IPTG. Samples for the assay were collected after five additional generations (around OD600nm of 0.5). Specific activities are expressed in Miller units.
RNA preparation and analysis
Cells were grown as for β-galactosidase assays. Total RNA was extracted with RNeasy kit (Quiagen). Primer extension analysis was performed with SuperScriptTMII reverse transcriptase (Life Technologies). Sequencing and reverse transcription reactions were analysed on denaturing polyacrylamide gels (National Diagnostic).
Purification of His6-RcsB protein
Strain MC1061, containing plasmid pFAB1, was grown in LB medium with 500 μM IPTG until OD600nm was ≈ 1. Cells were collected by centrifugation, washed twice with buffer I (20 mM Hepes (pH 7.4), 5 mM MgCl2, 100 mM KCl, 1 mM DTT, 10% glycerin, 0.02% NaN3, 0.2 mg ml−1 PMSF), resuspended in the same buffer, subjected to three cycles of freezing and thawing and disrupted by sonication. The crude extract was loaded onto a Ni–NTA affinity column (Qiagen). The column was washed successively with buffer I, containing 5 mM and 40 mM imidazole respectively, then the His6-RcsB protein was eluted with 200 mM imidazole in buffer I. The collected fractions were analysed by 12% SDS–PAGE. Fractions containing His6-RcsB were pooled, dialysed against buffer I without NaN3 and PMSF, and concentrated with a Centricon 10 microconcentrator spin column (Bio-Rad). Protein concentrations were calculated from the absorbance at 280 nm.
Mixed templates in vitro transcription assay
DNA templates for the transcription assays were generated by PCR and purified by exclusion chromatography (MicroSpin S-300 HR columns, Amersham Pharmacia Biotech). The ftsA1p template is a 260 bp fragment (co-ordinates 955–1214), generating an expected transcript of 104 nt. The other template, serving as an internal control, is a 177 bp fragment, containing the lacUV5p promoter, which should generate an expected transcript of 55 nt. Single-run transcription assays were performed in 15 μl of buffer (50 mM Tris-HCl, pH 7.8, 50 mM KCl, 3 mM MgCl2, 0.1 mM EDTA, 0.1 mM DTT, 25 μg ml−1 BSA) at 37°C. For each reaction, 0.3 pmol of a mixture of the two templates (containing ftsA1p and lacUV5p promoters respectively) and 0.3 units of RNA polymerase (Roche Molecular Biochemicals) were used. Templates were incubated with various amounts of RcsB protein in 7.5 μl of buffer for 5 min, then the RNA polymerase was added. Five minutes later, 7.5 μl of a mixture containing 0.25 mM ATP, GTP, CTP, 0.17 mM UTP, 50 μCi [32P]-UTP and heparin (0.6 μg ml−1) was added. After 5 min incubation, the reaction was stopped with 15 μl of sequence loading buffer. Five microlitres of the reaction was loaded onto 8% denaturing acrylamide gel. Signals were quantified with a Phosphor-Imager.
We are grateful to A. I. Nazer and G. Juillot for technical assistance, to G. Duval-Valentin for advice and to D. Lane for helpful discussions on the manuscript. This work was supported in part by the Université Paul Sabatier.