SoxR protein of Escherichia coli governs a global response against superoxide‐generating agents (such as paraquat) or nitric oxide, and provides broad antibiotic resistance. A redox signal activates SoxR post‐translationally to trigger transcription of a second regulatory gene, soxS. Activated and non‐activated SoxR bind the soxS promoter with the same high affinity, but only the activated protein stimulates soxS transcription. SoxR acts by an unusual mechanism of positive control: the protein binds the soxS promoter between near‐consensus −10 and −35 elements that are separated by an unusually long 19 bp (versus the optimal 17 bp). We have constructed and analyzed site‐specific deletions that alter the promoter element spacing. Reducing the spacer length to 16‐18 bp dramatically elevated basal expression of soxS in vivo and in vitro, and nearly eliminated additional activation by SoxR in response to paraquat. More strikingly, shortening the spacer converted SoxR from an activator into a repressor regardless of paraquat treatment. Gel mobility‐shift assays show that repression by SoxR of the promoters with 17 and 16 bp spacers is due to interference with binding by RNA polymerase. Thus, activated SoxR remodels the unusual configuration of the wild‐type soxS promoter into a highly active form, probably by compensating for the suboptimal distance between the −10 and the −35 elements.
In order to initiate transcription, Escherichia coli RNA polymerase (RNAP) recognizes and binds specific promoter sequences in DNA occupying a stretch of 50‐60 bp determined by both genetic and chemical approaches (Siebenlist et al., 1980; Pivec et al., 1985). Analysis of individual promoters (Jaurin et al., 1982) and comparison of many E.coli promoter sequences (Siebenlist et al., 1980; Hawley and McClure, 1983; Pivec et al., 1985; Gralla and Collado‐Vides, 1996) revealed the existence of two conserved regions (the −10 and the −35 boxes) to which the sigma 70 (σ70) subunit of RNAP binds, separated by a non‐conserved spacer of 17 ± 1 bp (Russell and Bennett, 1982). An A/T‐rich region upstream of the −35 box, the UP element, is also present in some strong promoters (Ross et al., 1993) and interacts with the carboxy‐terminal domains of the RNAP α subunits. Activator proteins enhance binding or open complex formation by RNAP at promoters that deviate from the consensus, whereas repressors can impair such a process (Collado‐Vides et al., 1991; Gralla and Collado‐Vides, 1996). For both positive and negative regulation, the position of regulatory sites in different E.coli promoters enables communication with RNAP. With very few exceptions, the majority of repressor binding sites (operators) overlap the promoter elements (Collado‐Vides et al., 1991; Gralla and Collado‐Vides, 1996). Studies with the lac operator revealed that the degree of repression varies with the operator position, and that the highest repression was achieved when the repressor binding site was placed in the −10/−35 spacer region (Lanzer and Bujard, 1988). The MerR/SoxR family of transcriptional activators constitutes a surprising exception to this rule: their binding sites are located between the −10 and the −35 boxes (Summers, 1992; Hidalgo and Demple, 1996b), but they exert strong positive regulation.
The soxRS locus of E.coli regulates an oxidative stress response that includes at least 10 genes involved in counteracting oxidative damage and in providing resistance to multiple antibiotics (Hidalgo and Demple, 1996b). The soxRS response is triggered when cells are exposed to nitric oxide (NO●) or to superoxide (O2●−)‐generating agents such as paraquat (PQ) (Amábile‐Cuevas and Demple, 1991; Wu and Weiss, 1991; Nunoshiba et al., 1992; Nunoshiba et al., 1993). SoxR protein is evidently the sensor for this response (Nunoshiba et al., 1992; Wu and Weiss, 1992). Activated SoxR protein, a homodimer of 17 kDa subunits containing a pair of [2Fe‐2S] clusters (Hidalgo et al., 1995; Wu et al., 1995), triggers expression of the soxS gene, whose protein product, SoxS, activates transcription of all the regulon genes (for a review, see Hidalgo and Demple, 1996b).
Purification of SoxR protein revealed important characteristics of its DNA‐binding and transcriptional activity. Two forms of the protein were isolated that contained Fe (Fe‐SoxR) or lacked detectable metals (apo‐SoxR), and both forms bound the soxS promoter with equal affinity (Hidalgo and Demple, 1994). However, only Fe‐SoxR stimulated transcription in vitro from the soxS promoter, up to 100‐fold over the basal transcription seen with RNAP alone (Hidalgo and Demple, 1994; Hidalgo et al., 1995). These observations led to our current model for soxRS regulation in vivo (Hidalgo and Demple, 1996b). In this model, the first step in this two‐stage transcriptional cascade involves the NO●‐ or O2●−‐mediated activation of pre‐existing SoxR, by a mechanism that could involve both insertion of the iron‐sulfur centers and their oxidation (Hidalgo et al., 1995; Hidalgo and Demple, 1996a). Activated, Fe‐SoxR then induces transcription of the soxS gene. The limiting step in this transcription activation event would not be the binding of SoxR to the soxS promoter, but an allosteric effect that promotes open complex formation by RNAP (Hidalgo et al., 1995). The proposed binding of non‐activated SoxR to the soxS promoter in vivo requires experimental confirmation.
Two striking features of this SoxR‐soxS interaction are the unusually long spacing between the −10 and −35 promoter elements and the position of the SoxR binding site. For σ70‐dependent E.coli promoters, changes in the −35/−10 spacing in different promoters can have important regulatory functions (Pivec et al., 1985). The constitutively weak soxS promoter includes −10 and −35 sites that are very close to the E.coli consensus, but with an unusual 19 bp spacing (Hidalgo and Demple, 1994). Furthermore, SoxR strongly binds (Kd∼10−10) and footprints a region that is centered in the spacer region upon a perfect 9 bp inverted repeat (Hidalgo and Demple, 1994). All the members of the MerR/SoxR family of transcriptional activators seem to share this unexpected feature in the interaction with their target promoters. In the case of MerR, it has been proposed than mercury‐dependent activation of the target merT promoter brings the −10 and −35 promoter elements into better helical alignment through a MerR‐mediated DNA untwisting effect (Ansari et al., 1992). Promoter element spacing does play a role in the low‐level expression of the merT promoter: small deletions in the spacer region conferred an up‐promoter phenotype (Lund and Brown, 1989; Parkhill and Brown, 1990).
The intriguing mechanism by which transcriptional activation occurs for the members of this family, as well as an important difference between the SoxR and MerR systems (the repression effect over merT expression exerted by uninduced MerR is absent in the soxS‐SoxR system), prompted us to analyze the effect of promoter element spacing in the constitutive activity of the soxS promoter. We conclude from these studies that the 19 bp spacing is required for effective SoxR‐mediated activation, and that this unusually long distance prevents SoxR from being a repressor.
Cloning of the wild‐type soxS promoter into pBluescript and the lacZ‐based vector, pRS551
Previous studies showed that soxS promoter sequences as short as ∼180 bp are enough for full activation by RNAP and SoxR in vitro (Hidalgo and Demple, 1994; Hidalgo et al., 1995; Wu et al., 1995). However, the shortest version of the soxS promoter that we have assayed for activity in vivo by fusion to the lacZ gene is a 500 bp long segment that encompasses the whole soxRS intergenic region together with ∼40 and ∼65% of the soxR and soxS genes, respectively (Nunoshiba et al., 1992). In order to construct soxS operator mutants, we decided to use a PCR‐mediated mutagenesis method. For this purpose, a 149 bp fragment of the wild‐type soxS promoter was amplified and subcloned into pBluescript to generate the plasmid pEH44 (see Materials and methods). The same PCR‐generated fragment subcloned into pRS551 generated an operon fusion to lacZ in plasmid pEH40, and the fusion was inserted into the chromosome of the soxRS+ strain GC4468, yielding strain EH40 with the fusion present in a single copy. As analyzed by β‐galactosidase assay (data not shown), the single‐copy fusion in EH40 could be induced by PQ to the same level as the previous reporter strain TN530 with lacZ fused to the 500 bp fragment (Nunoshiba and Demple, 1993). Interestingly, the basal (uninduced) level for the fusion for EH40 was significantly lower (2‐ to 3‐fold) than that reported previously for TN530 (data not shown), although the use of slightly different cloning vectors for the construction of the two fusions could explain this difference.
Construction and in vitro characterization of soxS operator mutants
Using the PCR‐based mutagenesis method described in Materials and methods, we constructed a series of mutant soxS operators and subcloned them into pBluescript (see Table I and Figure 1). Two series of deletion mutations were engineered that converted the 19 bp promoter element spacing of soxS to 18 bp. Series 50 consisted of a set of plasmids and strains with a 1 bp deletion in the 3 bp segment (TTA) between the inverted repeat and the −10 box whereas, in series 70, 1 bp was deleted near the center of the inverted repeat. Series 60 and series 80 contained, respectively, 2 or 3 bp deletions between the −10 and the inverted repeat, yielding spacings of 17 and 16 bp, respectively. Finally, series 90 combined the series 70 and series 80 mutations to remove 4 bp and generate a suboptimal spacing of 15 bp.
We next tested whether SoxR in vitro could bind the modified operators, as a prelude to determining the in vivo effect of SoxR on expression directed by the mutant promoters. As seen in Figure 2, the in vitro affinity of Fe‐SoxR for most of the promoter fragments was similar to the affinity for the wild‐type soxS promoter. However, for the series 90 construct (4 bp deletion), a ≥5‐fold higher SoxR concentration was required to yield binding similar to that with the wild‐type promoter.
In previous studies, we demonstrated the ability of Fe‐SoxR to activate transcription in vitro from a 180 bp wild‐type promoter sequence (Hidalgo et al., 1995). We now carried out a similar analysis of the 149 bp promoter fragments using primer extension with a new oligonucleotide to quantify the amount of soxS transcript synthesized in the reactions. Plasmids pEH44 (wild‐type, 19 bp spacing; series 40) or pEH55 (1 bp deletion, 18 bp spacing; series 50) were incubated with RNAP alone (Figure 3, lanes 1 and 4) or with the active Fe‐SoxR (Figure 3, lanes 2 and 5) or the inactive apo‐SoxR (Figure 3, lanes 3 and 6). The results demonstrate that, although Fe‐SoxR was required for activation of transcription from the wild‐type promoter, active SoxR was not required for transcription from the 18 bp spacing mutant promoter. Although apo‐SoxR did not affect transcription from the pEH55 promoter detectably, there was slight activation by Fe‐SoxR.
Activity in vivo of mutant soxS promoters fused to the lacZ gene
The in vitro transcription assay described above had a limited sensitivity compared with previous experiments (Hidalgo et al., 1995) and did not address the in vivo situation. We initially tried to construct lacZ fusions to the mutant promoters by using the same strategy described above for the wild‐type soxS promoter. This effort was unsuccessful, probably due to the high constitutive activity of the deletion mutant promoters and the resulting toxicity of excess β‐galactosidase, as has been reported before for other strong promoters (Podkovyrov and Larson, 1995). In fact, all the constructs obtained had modified or deleted part of the lacZ gene, or harbored additional mutations that restored the sub‐optimal 19 bp spacing or altered some other component of the soxS promoter (data not shown). One of these mutant promoters fused to lacZ (series 51, containing an A→G mutation in the −35 box, along with the 1 bp spacer deletion; Figure 1) was retained for further study.
The spacing mutant promoters fused to lacZ were finally constructed by transforming ligation mixtures of pRS551 and the various mutant inserts from the pBluescript derivatives (Table I) into MJC97, a pcnB mutant strain. This mutation reduces the copy number of ColE1‐based plasmids (Lopilato et al., 1986; Xu and Cohen, 1995). The resulting fusions were then transferred to the chromosome of GC4468 (a soxRS+ strain) as described in Materials and methods, yielding the strains indicated in Table I. The in vivo activity of the wild‐type and mutant promoters in these SoxR+ strains was assessed by measuring β‐galactosidase activity in cells grown in the presence or absence of PQ (Figure 4A). Interestingly, among the altered promoters, the 18 bp spacing mutants (strains EH50 and EH70) showed the highest constitutive and PQ‐induced activities, followed closely by the 17 bp mutant (EH60). The 16 bp mutant (EH80) yielded a significantly lower activity than the 18 or 17 bp spacer mutants under both induced and uninduced conditions (Figure 4A). As expected, the 15 bp spacing mutant (EH90) showed a lower basal level of transcription than the other mutants, although this promoter had ∼5‐fold higher activity with or without PQ treatment than the wild‐type promoter under non‐inducing conditions. Mutant EH51 (containing a mutation in the −35 box) had very low basal expression and showed only slight induction by PQ. Notably, PQ seemed to induce significant transcription only from the wild‐type promoter (up to 40‐fold), but not from the mutant promoters (1‐ to 1.5‐fold).
In order to study the effect of SoxR on the intrinsic activity of the different operators, we analyzed expression of the lacZ fusion constructs in the chromosome of DJ901, carrying a deletion of the entire soxRS locus (Greenberg et al., 1990). The pattern of soxS expression in the ΔsoxR strain (Figure 4B) was clearly different from that seen in the soxR+ strain (Figure 4A). As seen in Figure 4B, the constitutive β‐galactosidase levels effected by the 17 (EH66) and 16 bp (EH86) spacer mutations in the ΔsoxR strain were increased by 4‐ to 5‐fold over those observed in the soxR+ strain (EH60 and EH80 in Figure 4A), whereas the rest of the strains retained a comparable level of activity in both the ΔsoxR and the soxR+ backgrounds (with the exception of the expected lack of PQ inducibility in the case of the wild‐type promoter in the absence of SoxR in strain EH46). When we transferred the fusions to a different ΔsoxRS background [in strain BW829 (Tsaneva and Weiss, 1990)], the derepression of the 17 and 16 bp spacing mutants was also observed (data not shown).
Additional evidence for repression exerted by SoxR on the 17 and 16 bp spacer mutants was obtained by varying the expression of the protein. For this purpose, the ΔsoxRS strains containing the various fusions were transformed with plasmid pSXR, a SoxR expression vector derived from pSE380 (Amábile‐Cuevas and Demple, 1991), or with the vector alone. We confirmed overexpression of SoxR in the strains carrying pSXR by immunoblotting (data not shown), and we measured the transcriptional activity of the fusions by β‐galactosidase assay (Figure 5). For the wild‐type soxS promoter, elevated SoxR expression from the plasmid restored normal constitutive expression and actually enhanced induction by PQ by ∼50% (Figure 5A). In the series 50 and 70 promoters (both with 18 bp spacing), increased SoxR levels did not significantly affect the basal or induced levels of fusion expression (Figure 5B and C), although SoxR bound in vitro to these promoters with normal affinity (Figure 2). In contrast, for the series 60 and 80 constructs, SoxR exerted strong repression compared with the ΔsoxR strain (15‐fold for series 60 and 34‐fold for series 80; Figure 5D and E). Relative to the soxR+ strain, SoxR overexpression repressed the 17 and 16 bp spacer derivatives ∼5‐fold (Figure 5D and E). There was no significant effect of SoxR on the expression of the 15 bp spacer derivative (series 90, Figure 5F), consistent with the low affinity of the protein for this operator observed in vitro (Figure 2).
Effect of SoxR on binding of RNAP to mutant promoters in vitro
The positioning of the binding site for SoxR in the soxS promoter between the −10 and the −35 boxes is reminiscent of the sites for repressors that inhibit binding of RNAP (Collado‐Vides et al., 1991; Gralla and Collado‐Vides, 1996). However, for the wild‐type soxS promoter, binding of activated SoxR in vitro increases the affinity of RNAP modestly (Hidalgo and Demple, 1994). For the mutated soxS promoters of series 60 and 80 (17 and 16 bp spacing), the in vivo results (Figures 4 and 5) suggested that SoxR might still bind to the operator sites with high affinity (which was seen in vitro, as depicted in Figure 2) but impair the RNAP interaction with the −10 box (see Figure 1). We compared the in vitro binding of RNAP to the different soxS promoters in the absence (Figure 6, lanes 1‐4 in all panels) or presence (Figure 6, lanes 5‐8 in all panels) of saturating amounts of SoxR. As seen previously (Hidalgo and Demple, 1994), binding of Fe‐SoxR to the wild‐type soxS promoter improved the subsequent binding of RNAP nearly 5‐fold (Figure 6A), as determined by densitometric analysis (see Materials and methods). Prior binding of SoxR to the series 50 promoter (18 bp spacer) had no significant effect on subsequent RNAP binding (Figure 6B), and only a 2‐fold enhancement of RNAP binding by SoxR was observed for the 18 bp series 70 (Figure 6C). In contrast, SoxR clearly impaired the binding of RNAP to the promoters of series 60 and 80 (17 and 16 bp spacers; Figure 6D and E), by ∼3‐ and 10‐fold, respectively. These effects are consistent with the repression exerted by SoxR on these promoters in vivo. RNAP binding to the promoter of series 90 (15 bp spacer; Figure 6F) was also reduced 2‐fold by SoxR; this blocking effect might be stronger if SoxR could bind this promoter with high affinity relative to RNAP.
The first transcriptional event in the response to superoxide or nitric oxide stress in E.coli involves the interaction of activated SoxR with the promoter of the second regulatory gene, soxS. We have demonstrated with the construction and analysis of five soxS promoter mutants that the 19 bp spacing between the −10 and the −35 boxes of the soxS promoter is essential for normal induction by SoxR in the response to PQ, and that deletion of the spacing from 19 bp down to 17 or 16 bp converts SoxR from an activator into a significant repressor. The data presented here suggest that the induction of soxS transcription by activated SoxR is achieved by reorienting the −10 and −35 regions to compensate for the 19 bp spacer.
Studies of many E.coli promoters have suggested that spacer length, rather than specific nucleotide sequence, can have important regulatory functions, and that the intrinsic activity of otherwise similar promoters depends only on fluctuations of the spacer distance (Russell and Bennett, 1982; Hawley and McClure, 1983; Pivec et al., 1985). MerR and SoxR constitute the only known examples in which activation of their target genes takes place by compensating for overwinding between the −10 and the −35 elements in σ70‐RNAP recognition sites. In the case of the soxS promoter, the in vivo activities of the different mutant promoters in the strains lacking SoxR systematically revealed the quantitative effects of five different spacing sizes on transcription from an otherwise standard E.coli promoter. Although we have not confirmed that the same origin of transcription was used in all the promoter mutants, previous studies demonstrated that the transcriptional initiation site in a synthetic promoter is determined by the −10 region sequence (Rossi et al., 1983), which has been conserved in all our constructs. As suggested for other promoters (Gralla and Collado‐Vides, 1996), 17 bp is evidently the optimal spacing for the soxS promoter elements; 1 bp deviations from this distance, leading to 18 and 16 bp spacers, reduced the promoter activity by about half. A 15 bp spacer still allows a considerable amount of transcriptional activity, exceeding by ∼10‐fold the level for the 19 bp spacer.
It is intriguing that the σ70 subunit of RNAP is able to bind and trigger promoter activity with such different distances between the two known DNA contacts, the −10 and −35 boxes. Stefano and Gralla (1982) described the ‘untwist and melt’ model as a mechanism of open complex formation by RNAP. According to this model, RNAP forms unstable closed complexes with promoters due to its inability to make productive contacts with the −10 and the −35 regions. In fact, Werel et al. (1991) demonstrated that increasing the flexibility of the spacer DNA by single nucleoside deletions enhances the binding affinity of RNAP. These results are consistent with the hypothesis that favorable orientation of the binding contacts may be achieved by untwisting the DNA between them. The probability of achieving this reoriented state must depend on the length of the spacer, with the consensus spacer length of 17 bp being optimal. Open complex formation would result when the torsional stress exerted by RNAP is relieved by DNA melting around the transcriptional start site (Stefano and Gralla, 1982). In the case of soxS activation, the proposed underwinding mechanism exerted by SoxR could both reorient the promoter elements appropriately for RNAP binding, and provide torsional stress that could be relieved by open complex formation by RNAP. Footprinting experiments with DNase I or Cu(I)‐orthophenanthroline have already indicated DNA distortion by activated SoxR (Hidalgo and Demple, 1994; Hidalgo et al., 1995), and other experiments are underway. However, a role for protein‐protein contacts between SoxR and RNAP has not yet been eliminated and should be addressed.
An important difference between the merT‐MerR (Summers, 1992) and soxS‐SoxR systems is that non‐activated MerR is a repressor of the merT promoter: MerR is required for repression of the merT gene (Ross et al., 1989). However, in vitro (Frantz and O'Halloran, 1990) and in vivo (Heltzel et al., 1990) footprinting analysis indicates than RNAP still binds the promoter in the presence of uninduced MerR. Such a repression effect has never been observed in the soxS‐SoxR system. Differences between the proteins or the structures of their target promoters could explain this distinction. First, the MerR‐binding site consists of a 7 bp inverted repeat interrupted by 4 bp (Summers, 1992), whereas in the case of SoxR we find a perfect 9 bp inverted repeat (Figure 1) which could differentiate the mode of protein‐DNA interaction in these systems. Second, a DNA bend thought to be comprised of two kinks was detected for MerR bound in its inactive state to the merT promoter, and this bend is partially released by Hg‐dependent activation of the protein (Ansari et al., 1995). The positions of the MerR‐induced kinks were identified as two DNase I‐hypersensitive sites (Ansari et al., 1995). Although two similar sites are seen for SoxR (Hidalgo and Demple, 1994), which shows no repression activity at the wild‐type soxS promoter, preliminary experiments suggest that SoxR produces a much less dramatic bend of the soxS promoter (J.E.Bradner, E.Hidalgo, A.Ansari and B.Demple, unpublished results). Third, the location of the 18 bp activator site relative to the respective promoter elements is slightly different, the MerR site being shifted 1 bp toward the −10 box compared with the SoxR binding site. It would be of interest to determine the effect of relocating the SoxR operator 1 bp downstream within the spacer region, although a similar move of the MerR operator reduced merT activation efficiency (Parkhill and Brown, 1990).
The repression effect exerted by SoxR over the 17 and 16 bp mutant promoters is evidently a consequence of steric constraints in the promoter (Figure 7) and is thus mechanistically different from the repression described above in the MerR‐merT system. SoxR clearly impairs RNAP binding to the shortened promoters, an interference that does not occur for non‐activated MerR at the merT promoter. In the case of the wild‐type soxS promoter with 19 bp spacing, and in the 18 bp spacer promoters, the SoxR dimer evidently can bind without significantly diminishing the RNAP‐DNA contacts. RNAP binds one face of the promoter DNA across the region from −44 to −13 (Siebenlist et al., 1980); this structure requires SoxR to contact DNA along the DNA face opposite RNAP. Because of this physical proximity, moving the SoxR binding site even closer to the RNAP contact at −10 produces a steric clash that interferes with binding of the polymerase to cause repression.
Materials and methods
Cloning of wild‐type and mutant soxS promoters in pBluescript
To amplify the wild‐type soxS promoter sequence from pBD100 (which contains the whole soxRS locus; Amábile‐Cuevas and Demple, 1991) by PCR, we used primers X (5′‐ATGAATTCTGCGTTTCGCCACTTCG‐3′) and Y (5′‐GCGCGGATCCTCTTTTCAGTGTTGT‐3′), which correspond to regions −114 to −90 and +41 to +17 of the soxS promoter (+1 being the transcriptional start site; Figure 1), and include respectively an EcoRI and a BamHI site (both underlined). The fragment was purified with a PCR purification kit (Qiagen), digested with BamHI and EcoRI, and purified again after electrophoresis in a 1% agarose gel, using a gel purification procedure to extract the DNA from the gel slices (Qiagen). The BamHI‐EcoRI fragment was subcloned into the pBluescript KS plasmid (Stratagene), to yield pEH44 (see Table I), with an insert size of 149 bp. The sequences of all the PCR‐amplified fragments subcloned into pBluescript and described herein were confirmed by sequencing using pBluescript primers T3 (5′‐ATTAACCCTCACTAAAG‐3′) or T7 (5′‐AATACGACTCACTATAG‐3′).
In order to construct most of the mutant operators, pEH44 was used as a template for sequential steps of PCR. Pairs of mutagenic and complementary primers, containing the deletion of interest, were synthesized and used with either the T3 or the T7 primer in a first round of amplification. T3 and a mutagenic primer for the opposite strand were used to amplify the appropriate fragment (182‐186 bp) from pEH44; in parallel reactions, T7 primer and a mutagenic primer complementary to that in the first reaction were used to amplify a second, overlapping fragment (120‐126 bp). Products of the first two reactions (1 μl of each) were combined in a second round of PCR amplification using the T3 and T7 primers. The resulting PCR products (286‐288 bp) containing the complete and mutated soxS operator sequences were purified with a PCR purification kit (Qiagen), digested with EcoRI and BamHI, and purified through agarose gels as described above. The 146‐148 bp fragments were then subcloned into pBluescript KS. Following the mutagenesis protocol described above using pEH44 as a template, the plasmids were constructed: pEH55 [using primers del1 (5′‐CTTGAGGATTATACTCCCC‐3′) and del2 (5′‐GGGGAGTATAATCCTCAAG‐3′)], pEH64 [from primers del3 (5′‐ACTTGAGGTTATACTCCCC‐3′) and del4 (5′‐GGGGAGTATAACCTCAAGT‐3′)], pEH74 [from primers del7 (5′‐GTTACTTGAGGAATTATACT‐3′) and del8 (5′‐AGTATAATTCCTCAAGTAAC‐3′)] and pEH84 [from primers del9 (5′‐ACTTGAGGTATACTCCCC‐3′) and del10 (5′‐GGGGAGTATACCTCAAGT‐3′)]. In order to generate the insert for plasmid pEH94, pEH74 was used as the template and del9 and del10 as mutagenic primers.
Construction of soxS'::lacZ operon fusions
The BamHI‐EcoRI‐digested, 149 bp wild‐type soxS promoter fragment described above was ligated into BamHI‐EcoRI‐digested pRS551, a pBR322‐derived vector for lac operon fusions (Simons et al., 1987). The ligation mixture was transformed into DH5α cells, and pEH40 (Table 1) was isolated. The same procedure was used in attemps to subclone the mutant operators, but the resulting plasmids contained not only the designated deletions but also additional mutations. One of these, pEH51, contained an A→G mutation in the −35 box and was retained for further study. In order to construct fusions of the deletion mutant promoters to lacZ, the BamHI‐EcoRI‐digested mutant fragments from the pBluescript derivatives (pEH55, pEH64, pEH74, pEH84 and pEH94) were purified after gel electrophoresis and ligated into BamHI‐EcoRI‐digested pRS551. The ligation mixtures were transformed into MJC97 [F− Δ(lacX74) galE galK thi rpsL ΔphoA(PvuII) pcnB24‐1 zad1::Tn10]; the pcnB mutation reduces the copy number of pBR322 derivative plasmids, therefore decreasing the possible toxicity of β‐galactosidase overexpression (Podkovyrov and Larson, 1995). The pRS551‐derived plasmids were named pEH50, pEH60, pEH70, pEH80 and pEH90 (Table I). Bacteriophage λRS45 was used to transfer the soxS′::lacZ fusions from MC4100 (F−lacU169 araD139 thiA rpsL relA) carrying pEH40 or pEH51, or MJC97 carrying pEH50, pEH60, pEH70, pEH80 or pEH90, into the chromosome of GC4468 (soxRS+; Greenberg et al., 1990) as described by Simons et al. (1987). The resulting GC4468‐derived lysogens, EH40, EH51, EH50, EH60, EH70, EH80 and EH90, were selected by their resistance to kanamycin, and confirmed by their Lac+ AmpS phenotype (Simons et al., 1987). The resistance to the kanamycin selection procedure could not be used to insert the fusions into the chromosome of DJ901 [as GC4468 but Δ(soxR‐zjc2205)zjc2204::Tn10]. Instead, we used UV induction (Silhavy et al., 1984) to generate phage lysates of the different fusion derivatives already inserted into GC4468. Lysogens were then obtained by infecting DJ901 with the appropriate fusion phage lysate (Simons et al., 1987) and identification on McConkey lactose plates, to yield strains EH46, EH57, EH56, EH66, EH76, EH86 and EH96 (Table I). The DJ901‐derived strains were transformed with the SoxR expression vector pSXR and the control vector pSE380 (Amábile‐Cuevas and Demple, 1991).
In vitro transcription
The ability of RNAP to initiate transcription from two different soxS promoters in the presence and absence of SoxR was analyzed in vitro following a procedure previously described (Hidalgo and Demple, 1996a), except that the template plasmids used were pEH44 (containing the wild‐type soxS promoter) or pEH55 (containing the soxS promoter with a 1 bp deletion between the −10 and the −35 boxes; see Table I). The soxS transcript and a control bla transcript were quantified by primer extension with, respectively, primer T7 (located in pBluescript, 99 bp downstream from the initiation of transcription in the soxS promoter insert) and primer pBR‐1 (Hidalgo and Demple, 1996a).
Plasmids containing either wild‐type or mutated soxS promoters (pEH44, pEH51, pEH55, pEH64, pEH74, pEH84 and pEH94) were digested with BamHI and EcoRI and their inserts isolated by extraction from agarose gel slices after electrophoresis, as described above. The 146‐148 bp fragments were labeled at the 5′ end with [γ‐32P]ATP and T4 polynucleotide kinase (Promega) (Sambrook et al., 1989), and purified from unincorporated nucleotides on a small Sephadex G50 column. The DNA fragments (each at a final concentration of ∼0.1 nM) were incubated with the indicated amounts of SoxR and analyzed by gel electrophoresis (Hidalgo and Demple, 1994). For the RNAP binding experiments, 2.5 ng of Fe‐SoxR were incubated where indicated with the DNA for 10 min at 25°C, followed by the addition of the indicated amounts of σ70‐containing RNAP (Pharmacia) and further incubation at 37°C for 5 min in 20 μl reaction mixtures (Hidalgo and Demple, 1994). Apparent dissociation constants (Kd) were estimated from analysis of the gels by scanning densitometry with a BioImage system (Millipore) as previously described (Hidalgo and Demple, 1994).
soxS induction experiments
Lysogens containing the different operon fusions were inoculated into LB broth (Miller, 1992) with 30 μg/ml of kanamycin and 30 μg/ml of streptomycin and incubated at 37°C for ∼16 h with vigorous shaking. The overnight cultures were diluted 100‐fold into 1 ml of fresh medium in duplicate tubes and incubated at 37°C for exactly 90 min. PQ was then added at a final concentration of 0.25 mM to one of each pair of tubes, and incubation continued for 60 min. The samples were then placed on ice. β‐Galactosidase activity in SDS‐CHCl3‐treated cells was determined as described by Miller (1992)
We especially thank Bernard Weiss and Joe A.Pogliano for providing strains BW829 and MJC97, respectively. We are also very grateful to Ann Hochschild for her careful and critical reading of the manuscript and helpful discussions. This work was supported by grants to B.D. from the National Institutes of Health (CA37831) and the ALS Association. E.H. was supported by a postdoctoral fellowship from the Catalan Government (CIRIT).
- Copyright © 1997 European Molecular Biology Organization