Regulation by SoxR of mfsA, Which Encodes a Major Facilitator Protein Involved in Paraquat Resistance in Stenotrophomonas maltophilia

Stenotrophomonas maltophilia MfsA (Smlt1083) is an efflux pump in the major facilitator superfamily (MFS). Deletion of mfsA renders the strain more susceptible to paraquat, but no alteration in the susceptibility levels of other oxidants is observed. The expression of mfsA is inducible upon challenge with redox cycling/superoxide-generating drug (paraquat, menadione and plumbagin) treatments and is directly regulated by SoxR, which is a transcription regulator and sensor of superoxide-generating agents. Analysis of mfsA expression patterns in wild-type and a soxR mutant suggests that oxidized SoxR functions as a transcription activator of the gene. soxR (smlt1084) is located in a head-to-head fashion with mfsA, and these genes share the -10 motif of their promoter sequences. Purified SoxR specifically binds to the putative mfsA promoter motifs that contain a region that is highly homologous to the consensus SoxR binding site, and mutation of the SoxR binding site abolishes binding of purified SoxR protein. The SoxR box is located between the putative -35 and -10 promoter motifs of mfsA; and this position is typical for a promoter in which SoxR acts as a transcriptional activator. At the soxR promoter, the SoxR binding site covers the transcription start site of the soxR transcript; thus, binding of SoxR auto-represses its own transcription. Taken together, our results reveal for the first time that mfsA is a novel member of the SoxR regulon and that SoxR binds and directly regulates its expression.


Introduction
Stenotrophomonas maltophilia is an aerobic, Gram-negative, opportunistic pathogen that is ubiquitous in the environment. This microorganism is of great interest because it is considered to be an emerging nosocomial pathogen with intrinsic antimicrobial resistance to a variety of antibiotics. S. maltophilia is reported to be associated with infections of cystic fibrosis patients and immunocompromised individuals, particularly those who have been admitted in hospitals for a long period. This pathogen causes meningitis, endocarditis, bacteremia and infections of the respiratory, urinary and gastrointestinal tracts. Blood stream infection of S. maltophilia usually results in a substantially high mortality rate [1,2].
The major facilitator superfamily (MFS) consists of membrane transport proteins that are attributed to active efflux of compounds, including antibiotics, toxins and xenobiotics through a cation/substrate antiport mechanism [3]. The prokaryotic MFS encompass at least 58 families [4], and MFS efflux pumps contribute to multidrug resistance in several Gram-positive and Gram-negative pathogenic bacteria [5,6,7,8]. The most well-characterized MFS in term of function and gene regulation is a tetracycline efflux protein (TetA) [9,10,11]. The expression of tetA is regulated by the TetR transcription repressor. In the absence of tetracycline, TetR binds the tetA operator and represses its transcript. Once exposed to tetracycline, the drug-Mg 2+ complex binds the TetR repressor and changes the conformation of TetR, leading to a release of TetR from the tetA operator, and thereby allowing the transcription process of tetA.
SoxR, which is a MerR-type transcription factor, is a [2Fe-2S] cluster-containing transcriptional regulator that senses and responds to superoxide stress, redox cycling drugs/superoxide generators and reactive nitrogen species through one-electron oxidation of its [2Fe-2S] 1+ reactive center to [2Fe-2S] 2+ [12,13]. In E. coli, the SoxR homodimer binds to a dyad-symmetric sequence that is located between the -35 and -10 elements of the soxS promoter. Under normal physiological conditions, binding of reduced SoxR represses soxS transcription, while, under redox stress, binding of oxidized SoxR activates soxS gene expression [14]. The synthesized SoxS then up-regulates other genes in the regulon that are involved in stress alleviation and damage repair. While SoxR is highly conserved in diverse bacteria, SoxS is absent from most non-enteric bacteria, and SoxR directly regulates genes in the regulon [15,16,17,18,19,20,21].
Several drug resistance determinants in S. maltophilia have been investigated and reported, but, to our knowledge, none belong to the MFS family. Here, we report the contribution of mfsA, which encodes a MFS protein, to the drug resistance of S. maltophilia. The transcription of mfsA is controlled by SoxR, which is a redox cycling drug/superoxide sensor and transcriptional regulator. Redox cycling drugs such as paraquat, plumbagin and menadione are compounds capable of undergoing enzymatic one-electron reduction to form a transient radical that can be reoxidized by molecular oxygen (O 2 ) to produce superoxide anion [22]. Thus, mfsA expression is increased in response to the presence of redox cycling drugs. Because mfsA and soxR have overlapping promoter sequences, binding of SoxR to a single SoxR binding site controls the gene expression of mfsA and soxR, but in different manners.

Results and Discussion
MfsA belongs to a 14 transmembrane domain subfamily of the major facilitator superfamily At least 37 putative MFS are identified in the annotated genomes of S. maltophilia [23,24]. Among these coding DNA sequences (CDS), Smlt1083 (hereafter referred to as mfsA) is of particular interest because it is located next to a divergent gene, smlt1084, encoding a SoxR homolog, which is a superoxide sensor and transcriptional regulator. A linkage between mfs and soxR has been previously reported in Xanthomonas campestris pv. campestris [18]. Analysis of the available bacterial genomes unveiled that the mfs and soxR gene organization is found in some proteobacteria (including Xanthomonas axonopodis and Agrobacterium vitis) and actinobacteria (including Gordonia bronchialis, Mycobacterium abscessus, Rhodococcus erythropolis, Rhodococcus opacus, Renibacterium salmoninarum and Nocardia farcinica). Thus far, no functional connection between mfs and soxR has been reported.
The putative mfsA encodes a 473-amino-acid protein [23,24]. Our primer extension results (in the promoter analysis section) strongly suggest that a starting codon (ATG) should be located 51 bp downstream of the original ATG that is annotated in the genome sequence database. Thus, mfsA encodes a 456-amino-acid protein with a theoretical molecular mass of 49.4 kDa. Analysis of its deduced amino acid sequence using several algorithms for protein topology prediction, viz. TMPred, TMHMM 2.0, SOSUI and TopPred II, strongly suggest that S. maltophilia MfsA possesses 14 helical transmembrane domains (L28-T50, R59-R81, P91-G113,  R119-Y141, V149-I171, W179-R201, N216-Q237, A242-G264, S274-L296, S306-L325,  T343-V365, W370-L391, A403-L425 and T433-A453) and exposes its N-and C-termini to the cytoplasm (Fig 1). Generally, transmembrane transport through a MFS is energized by the electrochemical gradient of protons in an antiport process [25]. Most prokaryotic MFS proteins typically contain either 12 or, to a lesser extent, 14 transmembrane α-helical domains [4]. Based on the classification of MFS proteins that was proposed by Pao et al. [26] and phylogenic tree analysis, S. maltophilia MfsA is suggested to be an efflux pump that belongs to a subfamily of the drug:H + antiporter with 14 transmembrane domain subfamily (DHA14). The representative example of this group is the drug efflux pump QacA, which is an MFS that mediates resistance to antiseptics and disinfectants of Gram-positive Staphylococcus aureus [27]. However, the MfsA primary structure shares only 17.0% identity with S. aureus QacA and 16.1 and 14.4% identity with the methyl viologen resistance protein, SmvA, from Salmonella enterica serovar Typhimurium [28] and Acinetobacter baumannii [29], respectively. Furthermore, individual members of the MFS subfamily commonly share low amino acid sequence identity and similarity [4].
The mfsA mutant showed increased susceptibility to paraquat To evaluate the function of mfsA in S. maltophilia, the ΔmfsA mutant was constructed by gene deletion using double crossover and the Cre-loxP system [30]. We determined the susceptibility of the ΔmfsA mutant against methyl viologen (paraquat), and, as illustrated in Fig 2, the ΔmfsA mutant showed increased sensitivity to paraquat. The mutant was 10 3 -fold more susceptible to paraquat when cells were grown in the presence of 400 μM paraquat compared with the wild-type control. The altered phenotype of the ΔmfsA mutant was fully restored in the complemented strain (the ΔmfsA mutant containing a transposed mini-Tn7-mfsA for the single-copy expression of mfsA (ΔmfsA::msrA)). Because paraquat is a redox cycling drug that generates intracellular superoxide anions in the presence of O 2 , we tested whether increased paraquat susceptibility was due to lowering of superoxide dismutase (SOD) activities, as SOD is a group of enzymes that is responsible for dismutation of superoxide anions to H 2 O 2 . Total SOD activity was monitored as previously described [18], and the results showed that SOD activity in the ΔmfsA mutant (98.8 ± 7.4 U mg -1 protein) was comparable to that of wild-type K279a (95.5 ± 6.3 U mg -1 protein) and the mfsA complemented strain (96.5 ± 8.0 U mg -1 protein). Thus, increased susceptibility of the ΔmfsA mutant to paraquat is not due to alteration of total SOD activity.
Next, the sensitivity levels of the ΔmfsA mutant against other redox cycling drugs, i.e., menadione and plumbagin, were determined. Fig 2 illustrates that no significant changes in the sensitivity levels against menadione and plumbagin were observed in the ΔmfsA mutant relative to wild-type K279a. The fact that the phenotype of the ΔmfsA mutant that was sensitive to redox cycling drugs was specific to paraquat indicates that MfsA-mediated protection against paraquat toxicity does not involve superoxide anions. Together with its topology analysis, we proposed that MfsA functions as an efflux pump for paraquat. MfsA shares relatively low identity (14-16%) with the previously characterized paraquat resistance protein SmvA [28,29] in other Gram-negative bacteria. Thus, MfsA is likely a new paraquat resistance MFS protein. Indeed, S. maltophilia possesses an open reading frame, Smlt2678, that shares 33.9% identity with SmvA from Salmonella Typhimurium [28]. The contribution of Smlt2678 to paraquat resistance warrants further investigations.
Experiments were extended to test if deletion of mfsA affected the sensitivity levels to other oxidants, such as a disinfectant, H 2 O 2 , or organic hydroperoxides. The results revealed no The resistance levels of the S. maltophilia ΔmfsA mutant toward redox cycling drugs. A plate sensitivity assay was conducted to determine the resistance levels to oxidants. Ten-fold serially diluted, exponential-phase S. maltophilia K279a wild-type, ΔmfsA mutant and complemented (ΔmfsA::mfsA) cells were spotted onto plates containing 400 μM paraquat, 3 mM menadione or 400 μM plumbagin. The percent survival was defined as the percentage of CFUs on plates containing oxidant over the CFUs on control plates without oxidant.
doi:10.1371/journal.pone.0123699.g002 significant differences in the sensitivity levels to H 2 O 2 , cumene hydroperoxide or t-butyl hydroperoxide in the ΔmfsA mutant relative to wild-type K279a (data not shown). MfsA clearly has a major role in paraquat resistance, but not in general oxidative stress protection of S. maltophilia.
The expression of mfsA is induced upon challenge with redox cycling drugs Real-time reverse transcription polymerase chain reaction (qRT-PCR) was performed to measure the expression of mfsA in response to stresses. As illustrated in Fig 3, exposure to sublethal concentrations of redox cycling drugs, i.e., menadione, paraquat and plumbagin, enhanced the mfsA transcriptional level by 4.4 ± 0.5, 4.0 ± 0.2 and 3.9 ± 0.2-fold, respectively. Challenging the bacteria with H 2 O 2 or organic hydroperoxides did not induce mfsA transcription (Fig 3). The pattern of mfsA gene expression clearly shows that superoxide generators/redox cycling drugs are potent inducers of gene expression; and this result couples with the gene organization of mfsA, which is located immediately upstream and in the reverse orientation to soxR (smlt1984), and strongly suggests a contribution of SoxR in the regulation mfsA expression.

SoxR regulates mfsA gene expression
SoxR can either repress or activate transcription of its target gene depending on the redox status of the protein. In many bacteria, SoxR functions as a sensor for superoxide anions/redox cycling drugs or a transcription regulator. Based on phylogenetic tree analysis in a recent report, S. maltophila SoxR is evolutionarily closely related to SoxR from Xanthomonas campestris (with 70.1% identity) [31] and shares 63.8, 61.8 and 59.7% identity with SoxR from Pseudomonas aeruginosa, Streptomyces coelicolor and E. coli, respectively. The four cysteine residues that anchor the [2Fe-2S] cluster are conserved (data not shown).
To test whether SoxR regulates the expression of mfsA, the ΔsoxR mutant was constructed as described in the Methods and its expression profile was determined.  basal level of the mfsA transcript in the ΔsoxR mutant increased moderately (1.3 ± 0.1-fold) relative to the wild-type level. Moreover, expression of soxR in the complemented strain (ΔsoxR:: soxR), in which a single copy of soxR was transposed into the mutant using a mini-Tn7 vector and expressed under the lac promoter, led to a small repression of mfsA basal levels (0.8 ± 0.1-fold, Fig 3). The degree of repression of mfsA by SoxR was fairly low, suggesting that reduced SoxR only has minor role in regulating the basal level of mfsA. The major difference between mfsA expression in the soxR mutant and K279a wild-type was the inability of superoxide generators to induce mfsA expression in the mutant. The superoxide generator-induced expression of mfsA could be recovered in the complemented soxR mutant (ΔsoxR::soxR). The results clearly indicate that oxidized SoxR acts as transcription activator of mfsA expression.
SoxR mediates gene regulation as a transcriptional regulator through binding to the SoxR binding site of target gene promoters. This feature is conserved in all SoxRs that have been studied in enterobacteria, in which SoxR regulates a single gene, soxS (encoding transcriptional activator), or non-enteric bacteria, where SoxR directly controls genes in its regulon [15,17,18,20,21,32]. The promoter of mfsA was then experimentally determined. Primer extension experiments were conducted to map the 5' end of mfsA transcripts. Total RNA samples were prepared from uninduced and paraquat-induced cultures, and primer extension was performed using the 32 P-labeled BT2833 primer. A primer extension product of 53-bp that corresponded to a transcription start site (+1) at a T residue located 39 nucleotides upstream of the translational start codon (ATG) was observed (Fig 4A and 4B). The putative -35 and -10 promoter motifs were identified as TTGACC and CATGCT, respectively, and were separated by 19 bp. The SoxR binding site sequence and its location in relation to the target promoter motifs are remarkably well conserved in diverse groups of Gram-negative bacteria maltophilia K279a using the 32 P-labeled BT2833 primer. Extension products were sized on sequencing gels next to DNA sequence ladders (G, A, T, and C) that were generated using the same labeled primer, BT2833. The arrowhead indicates the putative transcription start site (position +1). (b) The nucleotide sequences of both strands of the mfsA-soxR promoter region are shown. Upper and lower strands represent the mfsA and soxR promoters, respectively. The putative -35 and -10 elements are underlined, and the start codon (ATG) is in bold type. The consensus sequence of the E. coli SoxR binding site is aligned above the sequence line in corresponding letters (gray box), and the homologous nucleotides are marked by asterisks. The arrowhead indicates the putative transcriptional start site (+1).  [14,15,17,18,19,20,21] [32]. Analysis of the nucleotide sequence in the vicinity of the mfsA promoter revealed an inverted repeat sequence, 5'ACCTCAACTGCAGTTGAGGT3', that shared high identity (14 of 18 nucleotides) with the consensus sequence for an E. coli SoxR box (5'CCTCAAGTTAACTTGAGG) [14]. The SoxR box is normally located in the intervening sequence between the -35 and -10 promoter motifs with a slightly imperfect spacer length. Binding of oxidized SoxR to the box leads to alteration in the DNA structure and subsequent activation of transcription of the target gene by RNA polymerases [12,15,18,32]. The mfsA promoter architecture fits well with a typical SoxR-regulated promoter, which has an imperfect spacer length and the SoxR box between the -35 and -10 regions.
Gel electrophoretic mobility shift assays with purified S. maltophilia SoxR and a mfsA promoter fragment was performed to test the ability of the transcription regulator to bind specifically to a putative mfsA promoter region. SoxR was purified as described in the Methods section. The results in Fig 5 illustrate that purified SoxR was able to bind the mfsA promoter sequence. The specificity of SoxR binding was revealed by the ability of unlabeled mfsA promoter sequence (cold probe, CP) to compete and prevent the binding between the purified SoxR protein and labeled probe, while an excess amount of unrelated, unlabeled DNA (pUC18, UD) did not compete with the labeled probe in the SoxR-mfsA promoter complex (Fig 5A). Additionally, excess of the unrelated protein bovine serum albumin (BSA) was unable to bind the labeled probe. To examine the role of the putative SoxR binding sequences that span the -35 and -10 motifs of the mfsA promoter, site-directed mutagenesis was performed to convert the putative SoxR box, 5'CCTCAACTGCAGTTGAGG3', to 5'AAGACCAGTACTGTGAGG3', and gel mobility shift experiments were repeated using the mutagenized mfsA promoter as a probe. As illustrated in Fig 5B, no binding complexes were detected at any of the SoxR concentrations that were tested, indicating that mutation of the putative SoxR binding site severely affected the binding ability of purified SoxR to the mfsA promoter. Thus, the putative SoxR binding site on the mfsA promoter is required for SoxR-mediated regulation of mfsA. Generally, SoxR in Apo-, reduced, or oxidized form binds to the target promoter sequence with a similar affinity. However, only the oxidized SoxR can activate the transcription of the regulated genes [33,34].
Examination of the mfsA promoter/translation initiation site revealed an atypical arrangement. The transcription start site is located 42 bp from the translation initiation site; thus, the mfsA mRNA contains a relatively long 5' untranslated region (UTR). The role of this 5' UTR in the translational control of mfsA is unknown. Furthermore, mfsA has a poor ribosome-binding site that precedes the translation initiation codon, ATG, which suggests that the mfsA mRNA could be poorly translated.
soxR is autoregulated but constitutively expressed SoxR typically regulates its own expression, and two types of soxR gene expression patterns have been reported. In E. coli as well as xanthomonads, reduced and oxidized forms of SoxR bind to their promoters and repress their own expression; therefore, soxR is constitutively expressed at low levels [14,18]. However, in some bacteria, the expression of soxR is inducible upon exposure to redox cycling drugs [15,16]. In S. maltophilia, soxR is located divergently to mfsA and shares the same promoter region; and the putative SoxR box is situated 7 bp upstream of the ATG initiation codon (Fig 4). This structure suggests that soxR expression is auto-regulated. The expression level of soxR was determined in the K279a wild-type, a ΔsoxR mutant and a ΔsoxR complemented strain (ΔsoxR::soxR). Total RNA was extracted from exponential-phase cultures that were induced with oxidants. qRT-PCR was conducted using BT2787 and BT3713, and the results are shown in Fig 6. soxR was constitutively expressed, and exposure of K279a cultures to redox cycling drugs (menadione, plumbagin and paraquat) and peroxides (H 2 O 2 and cumene hydroperoxide) failed to induce high-level transcription of soxR. A lack of functional SoxR in the ΔsoxR mutant increased its expression by 15.0 ± 4.6-fold relative to the wild-type level (Fig 6), and challenge with oxidants caused no significant alteration in the soxR expression levels. The constitutively high expression of soxR in the ΔsoxR mutant was abolished in the complemented ΔsoxR strain. Thus, the expression pattern of soxR together with the gel mobility shift results in which purified SoxR bound to a putative SoxR box suggest that SoxR in its reduced and oxidized forms acts as a transcriptional repressor. Because the promoter architecture of its target genes is critical for SoxR function, the soxR promoter was characterized. We attempted to map the putative transcriptional start site (+1) of soxR using primer extension. However, no extension product was detected. Therefore, 5'RACE was employed to determine the 5' end of the soxR transcript. The RACE-PCR products were cloned and sequenced, and the majority of the sequences from 7 independent clones mapped the 5' end of the soxR transcript to an A residue located 21 bp upstream of its ATG codon (Fig 4B). The putative -35 and -10 elements were identified as TTGTCA and CAGCAT, respectively, and they were separated by 17 bp (Fig 4B). The putative SoxR binding site was positioned at -4 to +14 of the soxR promoter. The structure of the soxR promoter and the SoxR box fit well with the observed expression studies of soxR expression (Fig 6). The binding of either reduced or oxidized SoxR to the SoxR box interferes with RNA polymerase binding to the promoter, leading to transcription repression.
SoxR differentially controls mfsA and soxR gene expression E. coli SoxR is a homodimer protein containing a pair of [2Fe-2S] clusters. This feature is believed to be completely conserved in other SoxRs from various bacteria, including S. maltophilia, because the [2Fe-2S] anchor motif is highly conserved [20] (data not shown). Altogether, our results suggest that, under physiological conditions, reduced SoxR binds to the single SoxR binding site that is located on the mfsA-soxR promoter and represses the expression of mfsA and soxR. Expression analysis of mfsA and soxR in the ΔsoxR mutant indicates that SoxR represses its own expression to a greater extent than that of mfsA. The promoter architecture of mfsA, in which the SoxR-binding site is located between the -35 and -10 promoter elements with a 19-bp spacer, is similar to typical, SoxR-regulated promoters, where binding of reduced SoxR impairs the interaction between RNA polymerase (RNAP) and promoters: binding of oxidized SoxR distorts the promoter to a configuration that facilitates initiation of transcription by RNAP, thereby activating transcription [14,35]. However, binding of reduced and oxidized SoxR to the sequence covering the +1 site that corresponds to the soxR promoter occludes its own transcription by interfering with RNAP binding [14]. The scenario of SoxR-mediated activation and repression of two genes from a single binding site is analogous to that reported in E. coli, in which soxR and a divergently transcribed soxS share a promoter sequence that contains a single SoxR binding site. Therefore, oxidized SoxR functions as activator for soxS transcription and as an autorepressor [14]. A similar scenario was observed for S. maltophilia SoxR where it acts as an autorepressor, but as an activator for mfsA expression, which is dependent on oxidized SoxR to increase its low basal level of expression.

Conclusion
We show here that MfsA (Smlt1083), which belongs to a DHA14 subfamily of MFS efflux pumps, plays an important role in mediating resistance to paraquat. In vitro and in vivo evidence indicate that mfsA expression is under the control of SoxR, a transcriptional regulator that is activated by exposure to redox cycling drugs including paraquat, plumbagin and menadione. Paraquat is among the redox cycling drugs that potently induce mfsA transcription, most likely through oxidation of SoxR [2Fe-2S]. Paraquat is also recognized as a herbicide that is widely used in agricultural fields to control unneeded weeds. Because the usual habitat of S. maltophilia is soil, the soxR-mfsA system would directly benefit the survival of this bacterium in paraquat-contaminated environments. In addition, it is likely that plumbagin and menadione are gratuitous inducers of SoxR with regard to mfsA since MfsA does not increase resistance to these oxidative agents. In several bacteria, superoxide dismutases are involved in redox-cycling drug resistance by catalyzing dismutation of superoxide anion. There are reports reveal that SoxR plays a role in sod regulation [15,32,36]. S. maltophilia also contains sod genes, thus the contribution of SoxR to regulation of sod gene expression and the role of sod genes in conferring to redox cycling drug resistance is being investigated.

Bacterial growth conditions
The Stenotrophomonas maltophilia wild-type K279a [24] and mutant strains were grown aerobically in LB medium at 35°C with continuous shaking at 150 rpm. Overnight cultures were inoculated into fresh LB medium to give an optical density at 600 nm (OD 600 ) of approximately 0.1. Exponential phase (OD 600 of 0.5, after 4 h of growth) cells were used in all experiments, as indicated. Bacterial strains and plasmids used in this study were listed in Table 1.

Molecular biology techniques
General molecular genetics techniques, including bacterial genomic DNA preparation, plasmid preparation, PCR, gene cloning, transformation into E. coli and agarose electrophoresis were performed using standard protocols [37]. S. maltophilia was transformed by electroporation using a 0.2-cm electrode gap cuvette with a Gene Pulser electroporator (BioRad) set at 2.5 kV, 200 O and 25 μF.

Determination of resistance levels to oxidants
The resistance levels against oxidants were determined using a plate sensitivity assay, as previously described [38]. Briefly, exponential-phase cells were 10-fold serially diluted in fresh LB medium before 10 μl of each dilution was spotted onto LB plates containing 400 μM paraquat, 3 mM menadione or 400 μM plumbagin. The plates were incubated at 30°C for 18 h before the Table 1. Bacterial strains and plasmids used in this study.

Strain or Plasmid
Relevant characteristic(s) Source

K279a
Wild-type strain [24] K279a:: Regulation by SoxR of Stenotrophomonas maltophilia mfsA numbers of colony forming units (CFUs) were scored. The resistance level to oxidant was expressed as the percent survival, which was defined as the percentage of CFUs on the plates containing oxidant over the CFUs on control plates without oxidant.

Construction of mfsA and soxR mutants
The unmarked gene deletion of mfsA was constructed using the allelic exchange method with the cre-lox marker recycling system [39]. The 3.6-kb DNA fragment spanning the mfsA coding sequence plus the sequences flanking both mfsA termini was PCR amplified from S. maltophilia K279a genomic DNA using primers BT2897 and BT2898 (Table 2) before being cloned into pDrive (Qiagen). The recombinant plasmid was digested with NcoI and SacII to excise the 928-bp sequence of the mfsA coding region prior to ligation with a NcoI-SacII-cut loxP-flanked Gm r cassette (from pUC18Gm [40]), which generated pDrive-ΔmfsA::Gm. This suicide plasmid was introduced into S. maltophilia K279a using electroporation. Putative mutants generated from a double-crossover event were selected from colonies with the Gm r phenotype and negative PCR results using primers located on the deleted sequence of mfsA. The unmarked ΔmfsA mutant was generated using Cre-mediated loxP recombination to excise the Gm r marker and was confirmed using PCR with primers flanking the deletion sequence. The ΔsoxR mutant was constructed. The 978-bp downstream region of soxR was PCR amplified from K279a genomic DNA using primers BT3818 and BT3819 ( Table 2). The PCR product was cloned into pGem-T (Promega, USA) before the NcoI-NotI fragment was subcloned into pUC18Gm, giving pΔSoxR-GmDown. Then, the 1,155-bp upstream sequence of soxR was amplified from K279a genomic DNA using primers BT3817 and BT3713 (Table 2) and cloned into pGem-T prior to subcloning the SacI-ApaI fragment into pΔSoxR-GmDown, which generated pΔSoxR-UpGmDown. This plasmid was introduced into K279a, and the ΔsoxR mutant was selected from colonies with carbenicillin-susceptible (Cb s ) and Gm r phenotypes. Excision of the Gm r cassette from the ΔsoxR mutant was attained through Cre-loxP recombination, and the unmarked ΔsoxR mutant was confirmed using PCR with two primers that flanked the deletion site.
Construction of the ΔmfsA mutant complemented strain (ΔmfsA:: mfsA) A single copy of mfsA was inserted into the chromosome of the S. maltophilia mfsA mutant using a Mini-Tn7 transposon vector. A DNA sequence spanning the full-length mfsA gene and its putative promoter was PCR amplified using the BT3713 and BT2835 primers (see Table 2), and the PCR product was cloned into pGemT-easy (Promega). The inserted DNA was sequenced to assure that no mutation had occurred before the ApaI-SacI fragment containing mfsA was cloned into pUC18 mini-Tn7T-LAC [30] that was cut with the same restriction enzymes. The resulting plasmid pTn7TmfsA and pTNS2 [30], which encoded the TnsABCD proteins that are required for attTn7 site-specific transposition, were transferred into the mfsA mutant. The transformants were selected on plates containing gentamicin. Integration of the mini-Tn7 transposon generally takes place at attTn7, which is located downstream of glmS, which encodes glucosamine-fructose-6-phosphate aminotransferase, an enzyme that catalyzes the formation of glucosamine 6-phosphate. S. maltophilia K279a contains a single glmS (smlt4099) gene and a sequence that shares high identity with the attTn7-associated sequence [41] was identified near the 3' end of the putative glmS gene. Integration of the mini-Tn7 containing mfsA into the mutant genome was assured by PCR amplification of the genomic DNA using a primer pair; one that annealed to the glmS gene (BT3766) and another that annealed to Tn7R (BT3767), as previously described for Xanthomonas campestris [41].
Construction of the ΔsoxR mutant complemented strain (ΔsoxR:: soxR) Single copy complementation of the ΔsoxR mutant was achieved using a mini-Tn7 transposon vector. The full-length soxR gene was PCR amplified using primers BT2787 and BT2788 and K279a genomic DNA as template. The blunt-end PCR product (435 bp) was cloned into pUC18-mini-Tn7T-LAC [30] at the EcoICRI site, the resulting plasmid pTn7TsoxR and pTNS2 were transferred into the ΔsoxR mutant, and the transformants were selected and confirmed, as described for the construction of the ΔmfsA mutant complemented strains. Purification of the S. maltophilia SoxR protein The untagged SoxR protein was highly expressed using the pET-blue2 expression vector and E. coli system (Novagen). The soxR full-length gene was PCR amplified from K279a chromosomal DNA using primers BT3246 and BT3247, and the PCR products were digested with NcoI and XhoI and cloned into the pET-Blue-2 that was cut with the same enzymes, which generated pETsoxR SM . Purification of untagged SoxR was conducted as previously described [15]. Essentially, exponential-phase cultures of E. coli BL21 (DE3)/pLacI harboring pETsoxR SM were induced with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 2 h. SoxR was then purified aerobically using a Whatman P-11 phosphocellulose ion exchange column. Oxidized SoxR protein was eluted with a linear gradient of 0.2 to 1.0 M KCl, and the purity of the SoxR protein was estimated by SDS-PAGE analysis.
Cloning of the mfsA promoter and site-directed mutagenesis A 258-bp DNA fragment spanning the putative mfsA-soxR promoter was PCR amplified from K279a genomic DNA using primers BT2832 and BT2833 and subsequently cloned into the pDrive vector (Qiagen, Germany), which generated pDrivePmfsA. Two-step, PCR-based, sitedirected mutagenesis [18] was performed to change the putative SoxR binding site sequence (5'CCTCAACTGCAGTTGAGG3') to 5'AAGACCAGTACTGTGAGG3' using two mutagenic primers, BT4081 and BT4082 (Table 2), and pDrivePmfsA as a DNA template. The PCR product was cloned into pGemT-easy (Promega, USA), generating pGemP Ã mfsA. To ensure that no unexpected mutation occurred, the nucleotide sequences of the insert DNA in pDri-vePmfsA and pGemP Ã mfsA were determined.

Primer extension
Total RNA was extracted from exponential-phase cultures grown under induction with 200 μM menadione or 100 μM plumbagin or uninduced conditions. The 32 P-labeled primer BT2833 (Table 2) was mixed with 10 μg of DNase I-treated RNA, and reverse transcription was performed using Superscript III reverse transcriptase (Invitrogen) at 60°C for 60 min. Extension products were sized on sequencing gels (9% polyacrylamide-7M urea) next to dideoxy sequencing ladders that were generated using a PCR sequencing kit (Applied Biosystems) with labeled BT2833 and pDrivePmfsA plasmid containing a putative mfsA promoter fragment as templates.

5' rapid amplification of cDNA ends (RACE)
5' RACE was performed using a 5'/3' RACE kit per the manufacturer's recommendation (Roche, Germany). Essentially, DNase I-treated total RNA was reversed transcribed using a specific primer BT2646 (SP1). The first-strand DNA (cDNA) was purified, and poly(A) was added to the 5'-terminus of the cDNA using terminal transferase enzyme. Next, poly(A)-tailed cDNA was PCR amplified using the specific BT3713 (SP2) primer and an anchored oligo(dT) primer. The purified PCR product was cloned into the pGemT vector, and the +1 site was identified from the DNA sequences.