The SOS response promotes qnrB quinolone-resistance determinant expression

Abstract

The qnr genes are plasmid-borne fluoroquinolone-resistance determinants widespread in Enterobacteriaceae. Three families of qnr determinants (qnrA, B and S) have been described, but little is known about their expression and regulation. Two new determinants, qnrC and qnrD, have been found recently. Here, we describe the characterization of the qnrB2 promoter and the identification of a LexA-binding site in the promoter region of all qnrB alleles. LexA is the central regulator of the SOS response to DNA damage. We show that qnrB2 expression is regulated through the SOS response in a LexA/RecA-dependent manner, and that it can be induced by the quinolone ciprofloxacin, a known inducer of the SOS system. This is the first description of direct SOS-dependent regulation of an antibiotic-resistance mechanism in response to the antibiotic itself.

Introduction

Quinolones are synthetic broad-spectrum antibiotics used widely to treat a variety of infectious diseases. Bacterial resistance to quinolones emerged rapidly after their widespread use. In Gram-negative bacteria, quinolone resistance was for a long time considered to be entirely mediated by mutations in chromosomal genes encoding quinolone targets (that is, DNA gyrase and topoisomerase IV) and/or in regulatory genes of outer membrane proteins or efflux pumps (Jacoby, 2005). Recently, plasmid-encoded quinolone resistance has been described along with three mechanisms: (i) a quinolone-protective mechanism encoded by the qnr genes; (ii) a modifying enzyme, aac(6′)-Ib-cr (Robicsek et al, 2006b); and (iii) an efflux pump encoded by the qepA gene (Perichon et al, 2007; Yamane et al, 2007).

Since the identification of the first plasmid-mediated quinolone-resistance determinant, qnrA, a decade ago (Martinez-Martinez et al, 1998), many other plasmid-borne qnr determinants have been found, comprising three families of Qnr proteins (QnrA, QnrB and QnrS), for which, respectively, 6, 20 and 3 alleles have been described (Jacoby et al, 2008). Recently, two more qnr determinants have been identified: qnrC and qnrD (Cavaco et al, 2009; Wang et al, 2009a). The qnr determinants encode pentapeptide-repeat proteins, which bind to DNA gyrase, protecting it from quinolone inhibition. Several studies worldwide have shown that qnr genes are widespread in the Enterobacteriaceae often associated with extended-spectrum β-lactamases (Robicsek et al, 2006a). The qnr determinants are found on conjugative plasmids that often carry other antibiotic-resistance determinants, and all qnrA and most qnrB genes are located in complex sul1-type integrons downstream from the insertion sequence, ISCR1 (insertion sequence common region 1; Nordmann & Poirel, 2005; Garnier et al, 2006). So far, the qnr promoter has been identified for only two determinants, qnrA1 and qnrB19 (Mammeri et al, 2005; Cattoir et al, 2008). Little is known about qnr gene expression and its regulation, except that qnrA1 expression can be induced by ciprofloxacin, through an unknown mechanism (Rodriguez-Martinez et al, 2006).

We recently reported the presence of the qnrB2 gene on a complex sul1-type integron in Salmonella enterica serovar Keurmassar (Garnier et al, 2006). Here, we characterized the qnrB2 promoter and identified a LexA-binding site upstream from the start codon. LexA is the central regulator of the SOS response to DNA damage. We showed that qnrB2 expression is repressed by LexA, and that ciprofloxacin, a known inducer of the SOS response, upregulates the expression of its own resistance determinant. This regulation seems to be a common trait of qnrB determinants, as the LexA site is conserved in all qnrB allele promoter regions described so far.

Results

Characterization of the qnrB2 promoter

We identified the qnrB2 promoter by using the 5′ rapid amplification of cDNA ends technique (Frohman, 1993). The qnrB2 transcription initiation site (+1) was mapped to position –28, upstream from the start codon (Fig 1). The –35 and –10 promoter elements (TTGACG and TACCAT, respectively), separated by 18 bp, were identified upstream from the +1 start site (Fig 1). This organization is in agreement with the recent description of the promoter of the qnrB19 allele, albeit with a difference in the spacer between the –35 and –10 boxes (18 bp in qnrB2, 17 bp in qnrB19; Cattoir et al, 2008; Fig 1). Only six other potential qnrB promoter regions are available among all qnrB sequences deposited in databases. Their alignment with the qnrB2 and qnrB19 promoter sequences showed that the −10 box is strictly conserved among the various qnr alleles and is separated by 17–18 bp from the −35 box, of which two sequences can be found (Fig 1). Our results confirm that the second ATG initiation codon, common to all known qnrB gene variants, is the initiation codon for all qnrB alleles, as postulated by Cattoir et al (2008).

Figure 1.

Sequence alignment of the qnrB2 promoter and qnrB alleles. The –35 and –10 promoter elements are indicated; the +1 start site is represented by an arrow; the start of the qnrB coding sequence is indicated by a dashed-open frame and the consensus sequence of the LexA-protein-binding site is boxed. qnrB allele sequences were aligned using the Geneious software. Sequence accession numbers EF682134, AM234698, EF683583, EF667294, EU052800, EU523120 and AB379831 for sequences with promoter regions for qnrB1, qnrB2, qnrB4, qnrB6, qnrB10, qnrB19 and qnrB20, respectively; for the remaining determinants, see Jacoby et al (2008).

Interestingly, we also identified the CTGTATAAAAAAACAG sequence between the +1 start site and the initiation codon of qnrB2. This sequence is homologous to the gammaproteobacteria LexA-protein-binding site consensus, CTGTN₈ACAG (Erill et al, 2007), suggesting that qnrB2 might be regulated by LexA. Furthermore, alignment of all available qnrB allele sequences upstream from the ATG (18 out of 20) showed that this potential LexA-binding site is fully conserved (Fig 1). The LexA protein is a member of the SOS regulatory network that represses a set of genes, the products of which are involved in several cellular processes (Little & Mount, 1982; Fernandez De Henestrosa et al, 2000). By binding specifically to a 16-bp palindromic motif typically located near or within the RNA-polymerase-binding site, the LexA dimer blocks gene transcription by interfering with RNA polymerase activity (Erill et al, 2007). The SOS response is induced in response to DNA damage, leading to RecA protein activation, which in turn promotes LexA autocatalytic cleavage and thereby derepresses LexA-controlled genes. To verify the functionality of the LexA site identified upstream from the qnrB2 gene, we performed electrophoresis mobility shift assay (EMSA) experiments with purified Escherichia coli LexA protein and showed that LexA binds to the identified motif, but not a form mutated at crucial site positions (Fig 2). This strongly suggested a role for LexA in the regulation of qnrB2 expression.

Figure 2.

LexA binds to the qnrB2 promoter. (A) Sequence of the qnrB2 promoter region with the LexA-binding site and the mutated version. The LexA-binding site sequence is boxed, the –35 and –10 promoter sequences are shown in bold, the mutated bases in the LexA site are grey-shaded and the qnrB2 5′ sequence indicated in a dashed open frame. (B) Electrophoresis mobility shift assay with the native (lanes 1–6) or mutated (lanes 7–8) qnrB2 LexA-binding site (PqnrB2 and PqnrB2lexA*, respectively) in the presence or absence of LexA purified protein (amounts in nanograms are indicated). Competition experiments were performed with an excess of cold probe PqnrB2 (lane 5) or PqnrB2lexA* (lane 6). F, free DNA; R, retarded complex.

Negative regulation of qnrB2 by LexA

To investigate whether qnrB2 gene expression is regulated in a LexA-dependent manner, we constructed a lacZ reporter transcriptional fusion with the qnrB2 promoter (PqnrB2). The recombinant plasmid, pPqnrB2–lacZ, was introduced into E. coli K12 MG1656 (a lac- derivative of MG1655), its lexA- and recA-deleted derivatives (MG1656ΔlexA and MG1656ΔrecA, respectively), and E. coli K12 DM49, a strain carrying the lexA3(Ind-) allele encoding a non-cleavable LexA protein (Mount et al, 1972). Measurements of β-galactosidase activity showed that PqnrB2 exhibited a basal level of expression in the MG1656 parent strain, its recA- derivative and in the DM49 lexA3(Ind-) strain, which was sixfold higher in the lexA- derivative (Fig 3). A fivefold increase was also observed (Fig 3) with strain MG1656(pPqnrB2lexA*–lacZ), which carries changes in the crucial positions of the LexA-binding site that impede LexA binding (Fig 2). Complementation experiments in which LexA was overexpressed by the arabinose-inducible plasmid pBad-lexA showed that, when induced in the MG1656ΔlexA(pPqnrB2–lacZ) strain, LexA repressed PqnrB2, nearly restoring its expression level to that of MG1656(pPqnrB2–lacZ; supplementary Fig S1 online).

Figure 3.

Expression of qnrB2 is regulated by LexA and the SOS response. β-Galactosidase activity was measured without drugs (light grey), in the presence of mitomycin C (medium grey) or ciprofloxacin (dark grey) in various genetic backgrounds: the parent Escherichia coli strain MG1656 (MG), its derivatives MG1656ΔlexA and MG1656ΔrecA, as well as strain DM49 lexA3(Ind-). These strains contained a lacZ-fusion plasmid with either the native or the mutated qnrB2 promoter, pPqnrB2–lacZ and pPqnrB2lexA*–lacZ, respectively. The error bars represent the standard deviation of at least four independent assays.

We also performed real-time PCR assays with MG1656, MG1656ΔlexA and DM49 lexA3(Ind-) strains that carry a plasmid allowing QnrB2 expression from its own promoter (pPqnrB2–qnrB2; supplementary Table S1 online). We observed a similar number of transcripts for strains MG1656 and DM49 lexA3(Ind-), and a 5.3-fold increase in the MG1656ΔlexA strain. A 3.6-fold increase was consistently observed with the MG1656(pPqnrB2lexA*–qnrB2) strain containing a mutation in the LexA-binding motif.

Together, these results confirmed that LexA is involved in qnrB2 negative regulation.

Involvement of the SOS response in qnrB2 regulation

To confirm the involvement of the SOS response in qnrB2 regulation, we measured the β-galactosidase activity of MG1656(pPqnrB2–lacZ) after treatment with mitomycin C (a commonly used in vitro SOS activator) or ciprofloxacin (a quinolone known to induce the SOS response; Beaber et al, 2004; Bisognano et al, 2004). As shown in Fig 3, the addition of either mitomycin C or ciprofloxacin induced a fivefold increase in β-galactosidase activity in MG1656(pPqnrB2–lacZ) cultures, but not in MG1656ΔrecA(pPqnrB2–lacZ) cultures, indicating that mitomycin C and ciprofloxacin regulation of PqnrB2 is RecA dependent. As expected, neither drug had a significant effect on MG1656ΔlexA(pPqnrB2–lacZ) nor on MG1656(pPqnrB2–lexA*–lacZ) cultures, and there was no induction in DM49 lexA3(Ind-) (Fig 3).

These results confirmed that qnrB2 expression is regulated by the SOS response in a LexA/RecA-dependent manner.

Effect of qnrB2 regulation on quinolone resistance

To investigate the influence of the SOS-dependent regulation on the level of qnrB2-encoded quinolone resistance, we determined the minimal inhibitory concentration (MIC) of ciprofloxacin for each strain (Table 1). As ciprofloxacin itself is a well-known inducer of the SOS response, we expected to observe an induction of ciprofloxacin resistance with strain MG1656(pPqnrB2–qnrB2). Indeed, the MIC was 32-fold higher with MG1656(pPqnrB2–qnrB2) than with MG1656, and was comparable with the MIC of MG1656ΔlexA(pPqnrB2–qnrB2). Conversely, the MIC remained low for both the DM49 lexA3(Ind-)(pPqnrB2–qnrB2) and MG1656ΔrecA(pPqnrB2–qnrB2) strains. Furthermore, both strains DM49 and MG1656ΔrecA carrying plasmid pPqnrB2lexA*–qnrB2, in which the LexA-binding site is mutated, showed the same MIC (0.32 μg/ml) as MG1656ΔlexA(pPqnrB2–qnrB2).

Table 1.

Minimal inhibitory concentrations of ciprofloxacin

E. coli strains	Plasmids
	pSU38Δtot	pPqnrB2– qnrB2	pPqnrB2lexA*– qnrB2
MG1656	0.02†	0.64	0.64
MG1656ΔlexA	0.02	0.32	0.32
MG1656ΔrecA	0.0025	0.005	0.32
DM49 lexA3(Ind-)	0.01	0.01	0.32

^†Ciprofloxacin minimal inhibitory concentration (μg/ml).

These results showed that, in vivo, ciprofloxacin induced qnrB2-mediated quinolone resistance through LexA.

Discussion

Our results show that qnrB2 has all the characteristics of an SOS-response LexA-regulated gene, as defined by Kelley (2006). These include: (i) induction in wild-type strains when exposed to DNA-damaging agents, and basal expression without drug exposure; (ii) no induction in recA-negative strain (no LexA derepression) or DM49 lexA3(Ind-) strain (uncleavable LexA); and (iii) constitutive induction in ΔlexA strains. Furthermore, we performed EMSA assays definitely proving that the identified LexA-binding site is functional. This study shows that qnrB2 expression is regulated directly by LexA and induced by SOS-response activators including ciprofloxacin (Fig 4).

Figure 4.

Regulation of qnrB2 gene expression. In the uninduced state, the LexA protein is bound to its site at the promoter region of qnrB2. The qnrB2 gene is expressed at a basal level (dashed-line arrow). On induction of the SOS response, by ciprofloxacin for example, single-stranded DNA (ssDNA) is produced and the co-protease activity of the RecA protein is activated by binding to ssDNA. The interaction between LexA and the nucleoprotein filament RecA/ssDNA results in autoproteolytic cleavage of LexA and subsequently in qnrB2 derepression (thick-line arrow). Induced expression of qnrB2 leads to an increase in the ciprofloxacin minimal inhibitory concentration.

When we analysed the available allele promoter sequences of the other qnr determinants (accession numbers AY070235 and EU495238 for qnrA1 and qnrA3, DQ460733 and AB187515 for qnrS1 and qnrS2, EU917444 for qnrC and FJ228229 for qnrD), we did not see any evidence for a LexA-binding site around the initiation codon of qnrA, qnrS and qnrC alleles, but we identified a potential LexA-binding site upstream from the qnrD gene. Interestingly, phylogenetic analysis showed that qnrB and qnrD are closer to one another than to the other qnr determinants (Baquirin & Barlow, 2008; Cavaco et al, 2009; Wang et al, 2009a). Furthermore, QnrB2 protein has been shown to protect DNA gyrase more efficiently than QnrA, and to inhibit DNA gyrase supercoiling activity at high concentrations (Jacoby et al, 2006). Other gyrase-protecting proteins, such as MfpA and GyrI, have also been shown to inhibit DNA gyrase supercoiling activity, and it has been suggested that this feature could have a cost in terms of bacterial fitness (Robicsek et al, 2006a). Furthermore, it has been speculated that Qnr proteins would have physiological functions other than quinolone resistance, such as resistance to naturally occurring toxins that inhibit DNA gyrase, such as CcdB and MccB17 (Ellington & Woodford, 2006). Indeed, it has been shown that McbG, a Qnr-like protein, protects DNA gyrase against MccB17, which also induces the SOS response (Herrero & Moreno, 1986). Thus, direct qnrB regulation through the SOS response might have two benefits, first by limiting the fitness cost of qnrB expression, and second by providing protection against natural stressors and toxins that are deleterious for DNA gyrase.

The presence of qnr determinants only leads to slightly increased resistance to fluoroquinolones, but these determinants considerably facilitate the emergence of higher-level resistance (Jacoby, 2005). In E. coli, this latter effect depends on the increased mutation ability conferred by the non-essential polymerases Pol II, Pol IV and Pol V on LexA-cleavage-mediated derepression of their respective genes (polB, dinB and umuDC; Cirz et al, 2005). Thus, qnrB-mediated quinolone resistance and increased mutation ability are two events triggered by the same signal, namely the SOS response. Upregulation of the quinolone-resistance gene qnrB by ciprofloxacin in a RecA/LexA-dependent manner is thus one more example of the involvement of the SOS system in the evolution of bacterial antibiotic resistance. Indeed, the SOS system is known to be induced by a variety of antibiotics (ciprofloxacin, rifampicin, β-lactams and trimethoprim) and has been implicated in the spread of antibiotic resistance by promoting horizontal dissemination of antibiotic-resistance genes (Beaber et al, 2004) or mutations (Cirz et al, 2005). All these observations identify LexA as a potential target for the development of inhibitor molecules to delay the emergence of multidrug-resistant bacteria.

This description of the direct SOS-dependent regulation of an antibiotic-resistance mechanism has clinical implications. Indeed, we found that qnrB2-mediated quinolone resistance is induced in response to the antibiotic itself even at sub-inhibitory concentrations. This is probably the case for all qnrB genes given the presence of a conserved LexA-binding site. Indeed, during the reviewing process of this study, Wang et al also identified the LexA-binding site upstream from qnrB alleles and showed by PCR experiments after reverse transcription that expression of several qnrB determinants (qnrB1–4) was increased in response to SOS inducers including ciprofloxacin (Wang et al, 2009b). Thus, a qnrB-containing strain does not express quinolone resistance in non-inducing conditions (silent resistance gene), but this resistance will be activated under selective antibiotic pressure. This is an important observation with respect to preventing the dissemination of resistance genes, and should be taken into account in the management of infectious disease treatments and in future antibiotic policies.

Methods

Bacterial strains and culture conditions. All bacteria were grown in Brain Heart Infusion at 37°C. Antibiotics, when required, were used at the following concentrations: kanamycin 25 μg/ml and ampicillin 100 μg/ml. Mitomycin C was added at a final concentration of 0.8 μg/ml, ciprofloxacin at 0.025 μg/ml, and glucose and arabinose at 1%. MG1656ΔsfiAΔlexA (referred to as MG1656ΔlexA) and MG1656ΔrecA were constructed by three-step PCR as described in Chaveroche et al (2000) and at http://www.pasteur.fr/recherche/unites/Ggb/3SPCRprotocol.html. Bacterial strains and plasmids used in this study are summarized in supplementary Table S2 online.

Plasmid construction. Plasmids were constructed as detailed in the supplementary information online.

5′ rapid amplification of cDNA ends. 5′ rapid amplification of cDNA ends was performed as recommended by the manufacturer (Invitrogen, Cergy Pontoise, France). Total RNAs were extracted from cultures of S. enterica serovar Keurmassar (Garnier et al, 2006). Gene-specific primers GSP1-qnrB2, GSP2-qnrB2 and GSP3-qnrB2 were used (supplementary Table S3 online).

Minimal inhibitory concentration determination. MICs were determined as recommended by CLSI (Clinical and Laboratory Standards Institute; http://www.clsi.org), and were evaluated three times.

β-Galactosidase assay. Overnight (o/n) cultures of cells containing pPqnrB2–lacZ were diluted 1:100 in Brain Heart Infusion supplemented with kanamycin and grown for 2 h before adding mitomycin C or ciprofloxacin. Cells were then grown for a further hour before the assay. β-Galactosidase-specific activity was measured as described in Miller (1992), except that the temperature was set at 37°C.

Electrophoresis mobility shift assay. Overexpression and purification of the E. coli LexA protein encoded by pUA1107 were performed as described earlier (Abella et al, 2004). EMSA probes were obtained by PCR amplification using oligonucleotides qnrB-EMSA-5 and qnrB-EMSA-3 (supplementary Table S3 online), and were end-labelled with [γ³²P]ATP (Amersham, Saclay, France) using T4 polynucleotide kinase (Promega, Charbonnières, France). EMSA experiments were performed as described elsewhere (Abella et al, 2004), using various amounts of purified LexA, 20 ng of one of the radiolabelled DNA probes in the binding mixture and 1.4 μg of unlabelled probe for competition experiments. Samples were separated in 5% non-denaturing Tris–glycine–EDTA polyacrylamide gel, then dried and exposed to storage Phosphor Screen (Perkin-Elmer, Courtaboeuf, France). Images were digitized with a Cyclone scanner. Supplementary information is available at EMBO reports online (http://www.emboreports.org).

Supplementary Material

Supplementary Information