Characterization of the Bacillus subtilis YxdJ response regulator as the inducer of expression for the cognate ABC transporter YxdLM

In their biotope, bacteria often experience drastic changes in environmental conditions. An appropriate response can be developed by a bacterium that possesses a system ensuring the detection and subsequent transduction of a stimulus generated by modification of the external medium. In many cases, two-component regulatory systems (TCSs), usually composed of a sensor kinase and a response regulator, are involved in this adaptive response (Parkinson & Kofoid, 1992). After detection of the signal, the sensor kinase autophosphorylates on a histidyl residue. The phosphoryl group is then transferred to a conserved aspartyl residue of the cognate response regulator. Once phosphorylated, the regulator can modulate the transcription of specific genes, eliciting an appropriate cellular response to the original stimulus.

Genomic sequencing of several micro-organisms has revealed great diversity of the TCS repertory in many species. This fact reflects the capability of some organisms to respond to a wide range of environmental changes. Bacillus subtilis, a Gram-positive spore-forming soil bacterium, possesses more than 30 such TCSs (36 sensor kinases and 35 response regulators; Fabret et al., 1999). The largest group, the IIIA/OmpR family, comprises 14 systems, to only four of which have known functions been attributed. The PhoP/PhoR and ResD/ResE systems participate in the response to phosphate starvation (Hulett et al., 1994), the latter system playing a central role in aerobic and anaerobic respiration (Sun et al., 1996). The YycF/YycG system is an essential two-component regulator of B. subtilis growth that modulates ftsAZ operon expression (Fukuchi et al., 2000), and the bceRS (formerly ytsAB) system plays a role in resistance to bacitracin (Bernard et al., 2003; Mascher et al., 2003; Ohki et al., 2003).

As exporters or importers of a wide variety of compounds across the membrane (Ames, 1986; Higgins et al., 1986), ABC (ATP-binding cassette) transporters play a key role in the response of bacteria to environmental changes. The prototypic ABC transporter comprises two membrane-spanning domains and two cytoplasmic nucleotide-binding domains (NBDs) that bind and hydrolyse ATP to provide energy for the transport. The inventory and classification of B. subtilis ABC transporters indicated that among the 59 systems predicted as ABC transporters more than 60 % are of unknown function (Quentin et al., 1999).

We recently demonstrated genetic and functional relationship between some members of the IIIA/OmpR family of TCSs and of subfamily 9 of ABC transporters in B. subtilis (Joseph et al., 2002), the TCS structural genes, yxdJK, yvcPQ and bceRS (formerly ytsAB), controlling the expression of the cognate ABC transporter genes yxdLM, yvcRS and bceAB (formerly ytsCD), respectively. In addition, the BceR/BceS TCS, together with the BceA/BceB ABC transporter, were shown to participate in bacitracin resistance of this bacterium (Bernard et al., 2003; Mascher et al., 2003; Ohki et al., 2003).

We have focused our study on the yxd locus. The operon encoding these ABC transporter structural genes contains an additional gene, yxeA, which encodes an 80 aa peptide conserved in several bacteria of the Bacillus/Clostridium group. The goal of the present work was to characterize the promoter region of the yxdLMyxeA operon and to identify other genes regulated by YxdJ. We showed that YxdJ directly interacts with DNA upstream of the yxdL gene, but this is the only strongly regulated transcript which we detected.

General molecular biology techniques.
All molecular biology procedures not presented in detail were carried out as described by Sambrook & Russell (2001). DNA-modifying enzymes were used as recommended by the manufacturer (New England Biolabs). DNA fragments were purified using either Microcon-30 (Millipore) or Qiaquick nucleotide removal kit (Qiagen). DNA cloning was done in Escherichia coli DH5α. PCRs were done in 50 or 100 µl final volume using Yellow Star polymerase (Eurogentec) according to the manufacturer. Primers are listed in Table 1.

Table 1. Primers

Bacterial strains, plasmids and growth conditions.
The bacterial strains and plasmids used in this study are listed in Table 2. E. coli strains were grown in LuriaBertani broth medium and B. subtilis in either LuriaBertani or Spizizen's minimum salt media (Spizizen, 1958). When needed, IPTG, arabinose and glucose were added to the culture at 1 mM, 0·2 % and 0·2 % respectively. The following antibiotics were used: ampicillin at 50 µg ml¹ and chloramphenicol at 25µg ml¹ for E. coli, and kanamycin at 20 µg ml¹ for B. subtilis. β-Galactosidase assays were carried out as described by Miller (1972).

Table 2. Strains and plasmids

RNA isolation.
Culture aliquots (2 ml) were harvested and submitted to a brief centrifugation (13 000 r.p.m. for 3 min). The bacterial pellet was then frozen immediately and stored at 20 °C. Total RNA was isolated using the High Pure RNA isolation kit (Roche) according to the supplier's recommendation. To avoid genomic DNA contamination, which was tested by PCR with each RNA preparation, two DNase treatments and column purifications were done instead of one.

Primer extension analysis.
Primer extension reactions were carried out with 550 µg of total RNA, 10 pmol of one of the ³³P end-labelled primers YxdL_EA1 or YxdL_EA2, and 200 units of SuperScript II RNaseH reverse transcriptase (Gibco-BRL) in 20 µl of 1x first-strand buffer (Gibco-BRL) containing 10 mM dithiothreitol, 0·5 mM each deoxyribonucleotide triphosphate (Amersham Pharmacia Biotech). 5' end labelling of oligonucleotide was done using T4 polynucleotide kinase (Amersham Pharmacia Biotech) and [γ-³³P]ATP (Amersham Pharmacia Biotech). Primer extension products were analysed by electrophoresis on a 6 % polyacrylamide/6 M urea gel, alongside a sequencing ladder (lanes C, T, G and A) obtained with the same end-labelled primer and the yxdL promoter PCR fragment as template. Sequencing reactions were done with Sequenase (United States Biochemical).

Deletion analysis of the yxdLMyxeA promoter.
DNA fragments were PCR-amplified using genomic B. subtilis DNA as template and were cloned upstream of lacZ in plasmid pGE593 (Eraso & Weinstock, 1992). The PCR products were generated using a common primer (YxdKL_right) and one of the following specific primers: for fragment A (281, 159), YxdKL_left_A; for fragment B (94, 159), YxdKL_left_B; for fragment C (69, 159), YxdKL_left_C; and for fragment D (30, 159), YxdKL_left_D. yxdJ was amplified with Yxdj_dir and Yxdj_rev and then cloned in pBAD33 (Guzman et al., 1995) under the control of an arabinose-inducible promoter. Both recombinant plasmids were introduced into E. coli. β-Galactosidase activities were then measured in the recombinant bacteria grown in the presence of either glucose (repression of yxdJ expression) or arabinose (induction of yxdJ expression).

Overexpression and purification of His-tagged YxdJ protein from B. subtilis.
yxdJ was amplified from B. subtilis genomic DNA as template using Yxdj_dir and Yxdj_rev primers. The DNA fragment obtained was purified and cut with EagI. It was then cloned in a modified version of pET22b+ (Novagen) linearized with PmlI. To get the modified pET22b+, an adapter, obtained by hybridization of the two primers pET_his_1 and pET_his_2, was cloned between the NdeI/EcoRI sites of pET22b+. This construction introduces a PmlI cloning site into this vector and also allows addition of several codons at the 5' end of the cloned gene, creating a His-tag at the N-terminus of the protein to be produced. The resulting plasmid was introduced into E. coli BL21/DE3. The recombinant His-tagged YxdJ protein was thus overproduced and purified on Ni-NTA resin (Qiagen). Resin-bound His-tagged YxdJ was washed with buffer A (HEPES 10 mM, NaCl 150 mM, pH 7·4) containing 30 mM imidazole. The protein was eluted with buffer A containing 300 mM imidazole. After dialysis against buffer A containing 10 % (v/v) glycerol, the tagged protein was stored at 80 °C.

Gel mobility shift assay.
DNA fragments were obtained by PCR amplification using one of the YxdKL_left primers (A, B, C or D) and the YxdKL_right primer with B. subtilis genomic DNA as template. A nonspecific DNA fragment, chosen within the yxdL gene, was PCR-amplified with primers Yxdl1 and Yxdl2. Fifty nanograms each of DNA fragment (A, B, C or D) and nonspecific DNA fragment were mixed together with variable amount of purified His-tagged YxdJ protein in 50 mM Tris/HCl (pH 8), 1·25 mM EDTA, 0·25 M sucrose and 0·025 % bromophenol blue. The mixture (4 µl final volume) was incubated for 30 min at room temperature and loaded on a native 12·5 % acrylamide Phast gel (PhastSystem from Amersham Pharmacia Biotech). After migration the gel was soaked in an aqueous ethidium bromide solution (0·5 µg ml¹) for 5 min and then observed on a UV transilluminator.

DNase I protection assay.
The DNA fragment obtained by PCR amplification of B. subtilis genomic DNA with primers (YxdKL_right_1 and YxdKL_F) was cloned into SmaI-linearized pBluescript KS (Promega) and the sequences of the recombinant clones were verified. Labelling of the DNA fragment used for DNAse I footprinting was done as follows. PCR-amplified fragment obtained with primers YxdKL_F and PBS-X was 5'-end-labelled with [γ-³²P]ATP (4000 Ci mmol¹, 150 TBq mmol¹; NEN) and T4 polynucleotide kinase (Promega). Unincorporated nucleotides were removed using Nucleotide Removal Kit (Qiagen) following the recommendation of the manufacturer. Once purified, the labelled DNA fragment was digested with BamHI and subjected to treatment with the Qiaquick PCR purification kit (Qiagen). Then 5x10⁴ c.p.m. purified labelled DNA fragment diluted to a concentration of 1·5 nM was incubated with His-tagged YxdJ for 30 min at room temperature in 50 µl binding buffer containing 10 mM Tris/HCl pH 7·5, 50 mM NaCl, 2·5 mM MgCl₂, 0·5 mM dithiothreithol, 4 % glycerol and 1·5 µg poly(dI-dC).poly(dI-dC). The DNAprotein complexes were treated with 1 unit DNase I (Amersham Pharmacia Biotech) for 1 min at room temperature. The reaction was stopped by addition of a solution containing 192 mM sodium acetate, 32 mM EDTA, 0·14 % SDS and 64 µg yeast RNA ml¹. The samples were extracted by a phenol/chloroform treatment and, after ethanol precipitation, they were resuspended in conventional loading buffer. After denaturation, the samples were loaded on a 6 M urea/8 % polyacrylamide gel together with a G+A Maxam and Gilbert reaction done on the same labelled DNA fragment.

Global transcriptional analysis.
Fluorescently labelled cDNA was synthesized during reverse transcription of BSmrs112 and BSmrs139 RNA (10 µg) using Cyanine-modified dCTP. The reaction mixture contained (in 40 µl): 20 µg random primers (GibcoBRL); 1x first-strand buffer (GibcoBRL); 10 mM dithiothreitol (Amersham Pharmacia Biotech); 100 µM (each) dATP, dTTP and dGTP; 50 µM dCTP; 25 µM Cy (3 or 5)-dCTP (Amersham Pharmacia Biotech) and 200 units Superscript II (GibcoBRL). cDNA synthesis was carried out at 42 °C for 1 h and continued for a further 1 h after another addition of 200 units Superscript II. RNA was subjected to alkaline hydrolysis by adding NaOH to a final concentration of 0·05 M, and incubating at 70 °C for 10 min. The mixture was neutralized with HCl (0·05 M final concentration). Unincorporated Cy (3 or 5)-dCTP was removed using Microcon-30 (Millipore). The labelled cDNA was diluted in 450 µl water, and then loaded on the Microcon-30 filter. The filter was washed three times by adding 540 µl water followed by centrifugation (8 min at 12 000 r.p.m.). cDNA was recovered by inverting the filter and centrifuging it for 3 min at 12 000 r.p.m.

Fluorescently labelled cDNAs from the two strains were mixed and used to hydridize to the same microarray (Eurogentec). Microarrays were first pre-hybridized for 1 h at 42 °C in 25 µl Dig Easy buffer (Roche) containing 10 µg salmon sperm DNA (Sigma) previously denatured for 5 min at 95 °C. Hybridizations were done in a final volume of 25 µl containing a mixture of the labelled cDNA at a final concentration of 1 µg ml¹. Each slide was incubated in water-bath using a waterproof hybridization chamber (Corning). Slides were scanned on a ScanArray 4000 (Packard Bioscience) and hybridization signals were quantified with QuantArray software version 2.1 (Packard Bioscience).

Localization of the yxdLMyxeA transcriptional start point
Transcriptional fusions of yxdK and yxdL to the E. coli lacZ gene, integrated into the B. subtilis chromosome, were tested for expression of β-galactosidase. Regardless of whether cells were grown in LuriaBertani medium or in Spizizen defined medium, the β-galactosidase activity never exceeded 5 or 0·5 Miller units for yxdK or yxdL, respectively (data not shown). Because of low expression, primer extension experiments done with variable amount of RNAs failed to generate detectable products.

We therefore used the strain BSmrs112, in which the overproduction of the YxdJ response regulator mimics the unknown stimulus of the YxdJ/YxdK TCS and triggers the expression of the yxdLMyxeA operon (Joseph et al., 2002). Using an RNA preparation from this strain, the yxdLMyxeA transcription start site was identified; it corresponds to an adenine located 87 bp upstream from the putative translation initiation codon of yxdL (Fig. 1). Upstream of this transcription start, the sequences corresponding to an extended σ^A-binding site (TGXTAATAT), and a 35 region (Helmann, 1995; Jarmer et al., 2001) were found.

(15K): Fig. 1. Primer extension analysis of yxdLMyxeA. Primer extension reactions were done as described in Methods. One-fifth of the reaction (corresponding to 1 µg total RNA before reverse transcription) was loaded on a 6 % polyacrylamide/6 M urea gel, alongside a sequencing ladder (lanes C, T, G and A). The mapped 5' end of the yxdL mRNA and the direction of transcription are indicated by the angled arrow above the sequence. The putative 10 and 35 sequences are underlined.

Deletion analysis of the yxdLMyxeA promoter region
As YxdJ positively controls yxdLM expression, it was of interest to precisely delineate the cis region of the yxdLMyxeA promoter required for this activation. For this purpose, fragments of various lengths of the yxdL regulatory region were cloned in plasmid pGE593 (Eraso & Weinstock, 1992), creating a series of transcriptional fusions with the lacZ reporter gene (plasmids pGE-A, pGE-B, pGE-C and pGE-D). Each of the recombinant plasmids was introduced into a ΔlacZ E. coli strain (DH5α) bearing the compatible pBAD-yxdJ plasmid that contains the yxdJ gene under the control of an arabinose-inducible promoter. Each recombinant strain was grown in medium containing either glucose or arabinose and samples were taken to measure the β-galactosidase activity. As shown in Fig. 2(a), in cells grown in the presence of glucose, the four DNA fragments tested for their promoter activity gave β-galactosidase activities of the same order of magnitude, which did not exceed 50 Miller units. In the presence of arabinose, a strong increase of β-galactosidase activity was observed for fragments A and B, with an induction ratio above 20 in both cases. In contrast, a much lower level of induction was observed with fragment C and no induction could be detected with fragment D. These results indicated that the 69 to 94 region upstream of fragment C contains information required for the YxdJ protein to induce the RNA synthesis from the promoter upstream of the yxdLMyxeA operon.

(30K): Fig. 2. Analysis of the yxdLMyxeA operon regulatory region. (a) Effects of deletion in the yxdL promoter region on the in vivo expression of lacZ. The DNA fragments A, B, C and D were cloned upstream of lacZ in pDE593. yxdJ was cloned in pBAD33 under the control of an arabinose-inducible promoter. Both recombinant plasmids were introduced into E. coli DH5α. Positions of the DNA fragments are indicated with respect to the yxdL transcription start. The yxdK and yxdL genes are indicated by large hatched arrows and 10 and 35 motifs are indicated by white boxes. E. coli cells were grown in the presence of glucose (white bars in histogram) or arabinose (grey bars), both at 0·2 % final concentration. β-Galactosidase activities are in Miller units and represent the mean of three experiments±SD. The ratios correspond to the activities obtained in the presence of arabinose divided by those obtained in the presence of glucose. (b) Gel shift analysis of binding of the purified His-tagged YxdJ to the yxdL promoter: 0·24 pmol of DNA fragment A (441 bp) or 0·33 pmol of DNA fragment B (254 bp) and 0·2 pmol of control DNA fragment (326 bp from within the yxdL gene) were incubated in the presence of various amounts of purified His-tagged YxdJ (lanes 1 and 5, none; lanes 2 and 6, 0·75 pmol; lane 3, 1·5 pmol; lanes 4 and 7, 2·25 pmol).

Direct interaction of YxdJ with the yxdMLyxeA promoter
The observation that YxdJ positively controls the expression of the yxdLMyxeA operon promoter in E. coli suggested that this response regulator binds directly to this region. To check this, we purified His-tagged YxdJ by nickel affinity chromatography and used it in a mobility shift assay with various yxdL promoter fragments. We first verified that the in vivo activity of YxdJ bearing a His-tag at the amino terminus is similar to that of the wild-type regulators. Indeed, both were shown to give similar levels of β-galactosidase activity when overexpressed in B. subtilis bearing the yxdLlacZ transcriptional fusion (data not shown). As indicated in Fig. 2(b), a change of mobility was observed with DNA fragments A (441 bp) and B (254 bp) using increasing amounts of YxdJ. In both cases, the tagged-YxdJ binding to its DNA target was specific, as no change in mobility was observed with a control DNA fragment. When fragment C was used, no band shift could be detected (data not shown). This result confirms the importance of the 69 to 94 region not only for a full stimulation of the yxdLMyxeA operon (Fig. 2a) but also for efficient binding of the YxdJ regulator.

Characterization of the YxdJ binding site by DNase I protection assay
To define the YxdJ binding site more precisely, DNase I footprint experiments were performed using a fragment corresponding to positions 178 to +160 with respect to the yxdL transcriptional start site. This fragment encompasses fragment B and a part of fragment A (Fig. 2). As shown in Fig. 3, a 38 bp region of the minus-strand, extending from base 41 to 78, was efficiently protected by YxdJ from DNase I digestion. Analysis of the protected region sequence revealed a 9 nucleotide direct repeat, TTAMRAAAA. Spacing of 21 nucleotides between the repeats indicates that they lie on the same side of the DNA helix, a feature that is expected from regulatory regions controlled by the OmpR-subfamily members acting as multimers (Makino et al., 1988; Rampersaud et al., 1989; Tsung et al., 1989). Thus, the YxdJ binding site extends between nucleotides 41 to 78 in the regulatory region of the yxdLMyxeA operon, and contains two direct repeats.

(39K): Fig. 3. DNase I footprints of YxdJ on the yxdLMyxeA promoter. A 338 bp PCR product, corresponding to positions 178 to +160 with respect to the yxdL transcriptional start, was 5' end labelled at the template strand and incubated with bovine serum albumin (lane 1) or with various amounts of His-tagged YxdJ (0·14 µM, lane 2; 0·56 µM, lane 3; and 1·12 µM, lane 4). After treatment with DNase I, the DNA was denatured and subjected to electrophoresis on an 8 % polyacrylamide/6 M urea gel. G+A corresponds to the MaxamGilbert sequencing reaction performed with the same 5'-labelled DNA fragment. The protected area is shown by the vertical bar. In the right panel, the DNA sequence of the protected region is shown by a dark grey box; 10 and 35 motifs are underlined and the transcription start site is indicated by an angled arrow. Boldface letters indicate nucleotides identical to the extended σA-binding site (TGXTAATAT). The arrowheads point to the position of the 5' end of each fragment (B, C and D) used in deletion analysis (Fig. 2a). The 5' end of fragment A is too far away to be indicated. ATG (capital letters) indicates the yxdL translation initiation codon. Dotted arrows indicate the TTAMRAAAA repeats.

Whole-genome transcription analysis of the YxdJ targets
To assess the extent of the YxdJ regulon, microarray hybridizations using RNA isolated from cells overproducing or not overproducing YxdJ were compared. Overproduction of YxdJ in strain BSmrs112 was used to mimic the unknown physiological conditions that normally lead to YxdJ activation. Three repetitive microarray analyses were carried out using RNA isolated from independent cell cultures of strains BSmrs139 and BSmrs112 grown in LuriaBertani medium. IPTG was added at mid-exponential growth and cells were harvested after 30 min further incubation and used for RNA preparation. The growth rates of the two strains were indistinguishable (not shown). After 30 min induction, only seven genes were shown to be induced (stimulation index above 3) and none was repressed (stimulation index below 0·33) (Table 3).

Table 3. Transcriptome analysis RNA was extracted from cells of BSmrs112 and BSmrs139 30 min after induction with 2 mM IPTG. Induction index values are ratios of BSmrs112 signal over BSmrs139 signal. Results are expressed as means±SD based on results of three independent experiments. Only genes with induction ratios >3 are listed (no genes were repressed, i.e. none showed a stimulation index <0·33). Genes are grouped when they belong to the same operon.

As expected, the stimulation indices of the yxdJ, yxdL and yxdM genes were found to be very high. This indicated that (i) the response regulator overproducing system is working efficiently; and (ii) the yxdL and yxdM genes are positively controlled by the YxdJ response regulator as we previously described (Joseph et al., 2002). However, a large difference in expression was observed between the three genes within the operon yxdLMyxeA: the stimulation index was 64 for yxdL, 6 for yxdM and below the threshold of detection for yxeA, which might reflect rapid degradation of this polycistronic mRNA. Similar results were obtained when real-time RT-PCR experiments were performed on the same RNA preparations (data not shown).

Previous experiments done by real-time PCR (Joseph et al., 2002) showed that neither the yxdK gene, encoding the histidine kinase partner of YxdJ, nor the yvcR gene, encoding an ABC transporter NBD, was induced upon YxdJ overexpression. These results are totally different from those obtained using the microarray approach and we believe that the latter are artefactual. First, to construct the recombinant plasmid for YxdJ overproduction we used a DNA fragment containing 29 bp that overlap with the 5' end of the yxdK gene. Thus, a large amount of an mRNA containing this overlap is produced when overproducing YxdJ. After reverse transcription and labelling, hybridization of the 29 bp fragment to the yxdK probe might easily occur on the microarray and produce the artefactual signal. Second, among the NBD-encoding genes, yvcR is by far the best yxdL homologue in B. subtilis (62·1 % identity at the DNA sequence level). In addition, the yvcR and yxdL DNA sequences contain long stretches of identical nucleotides. Thus, in the conditions of high-level yxdJ mRNA production, cross-hybridization between the yxdL labelled target and the yvcP probe might occur.

The dltA and dltD genes, involved in alanination of lipoteichoic and teichoic acids, are also induced. Using longer IPTG induction times, we have observed that dltB, dltC and dltE were also induced (data not shown) as well as the ywaA gene, which is predicted to encode a putative branched-chain amino acid aminotransferase. All these genes (dltABCDE and ywaA) are predicted to constitute an operon.

The yxdJK and yxdLM genes, encoding the TCS and ABC transporter systems, respectively, and belonging to separate operons (Joseph et al., 2002; Yoshida et al., 2000) were shown to be functionally related (Joseph et al., 2002). As reported previously (Yoshida et al., 2000) and in the present study, both operons are constitutively expressed at a very low level during the growth phase of B. subtilis. This complicates any studies of their promoter sequence and regulation. In conditions of overproduction of the YxdJ response regulator that mimics YxdJK activation, the level of yxdLMyxeA operon expression is dramatically increased (Joseph et al., 2002), allowing its transcriptional start to be precisely located. Further analyses presented here show that the YxdJ response regulator positively controls the yxdLMyxeA operon transcription by direct interaction with its promoter region. In retardation experiments a complete shift occurred with 15 pmol YxdJ, representing a 50-fold molar excess of the regulator to DNA. This ratio is of the same order of magnitude or even smaller than that needed to gain the complete shift of promoter region by response regulators of the OmpR family, as for instance YycF and its ftsZA promoter in B. subtilis (Fukuchi et al., 2000). The purified YxdJ protein protects from DNase I degradation a 38 nucleotide region located between positions 41 and 78 relative to the yxdL transcription start. Two TTAMRAAAA repeats were found in this protected region. Deletion of one base at position 70 in one repeat together with the flanking protected region (71 to 78) (Fig. 3) results in a complete loss of YxdJ binding to the DNA fragment in a gel shift assay. These results suggest that the two repeats (TTAMRAAAA) are the YxdJ binding site. Thus, this pattern was used to screen the whole B. subtilis genome using bioinformatics tools. Even with weak constraints on the sequence of the repeats and variable distance between them (1, 2 or 3 helixturns), we were unable to identify any region containing this motif apart from that in the yxdLMyxeA promoter region. This result is in concordance with the experimental search for these targets using transcriptome analysis.

Data collected from microarray experiments indicate a very restricted regulon of YxdJ: only four genes show significant change in their transcription level upon conditions mimicking TCS induction. Indeed, after 30 min of IPTG-mediated YxdJ overproduction, we detected only the expression of the cognate ABC transporter genes yxdLM, confirming our previous observation (Joseph et al., 2002), and the induction of the dltA and dltD genes.

The experimental conditions we used were slightly different from those of Kobayashi et al. (2001) in a similar approach. We used a wild-type strain rather than a yxdK-deleted mutant and a more efficient expression system (Joseph et al., 2001) giving a yxdJ overexpression ratio that reaches 122 (Table 3) whereas Kobayashi et al. (2001) used a 30-fold increase in yxdJ expression (data available at ftp://ftp.genome.ad.jp/pub/kegg/expression/ex0000286.dat). A very restricted gene expression pattern change was obtained in both cases with a common characteristic, a strong positive control of the yxdLM cognate ABC transporter gene expression by the YxdJ response regulator.

Using longer IPTG induction time, we have seen that all the genes of the dlt operon were induced, including the ywaA gene encoding a putative branched-chain amino acid aminotransferase (data not shown). The five dlt gene products are involved in teichoic acid polyalanylation (Perego et al., 1995). The promoter region of the dlt operons did not show detectable DNA binding of His-YxdJ in the gel mobility shift assays (data not shown). This probably indicates indirect regulation of expression of these genes by YxdJ and presumably explains the lack of detection of any YxdJ binding site sequence using bioinformatic approaches. Interestingly, expression of the dlt operon from Streptococcus agalactiae is also controlled by a TCS belonging to the OmpR family, the DltS/DltR system (Poyart et al., 2001). The exact physiological role of the yxd gene cluster remains to be elucidated. From our results it appears that the YxdJ regulon is limited to a set of genes encoding systems responsible for compound efflux, such as the membrane pump of the ABC family, and cell wall biosynthesis/modification. YxdL shows strong similarity to several NBD ABC transporters, such as SalX, involved in salivaricin resistance in Streptococcus salivarius and Streptococcus pyogenes (Upton et al., 2001), VraD and VraF from Staphylococcus aureus (Kuroda et al., 2000), and MbrA, responsible for bacitracin resistance in Streptococcus mutans (Tsuda et al., 2002). It was also shown recently that the bce system (formerly yts), which is paralogous to the yxd system, is involved in bacitracin resistance (Bernard et al., 2003; Mascher et al., 2003; Ohki et al., 2003). The increase in D-alanyl esterification of teichoic acids caused by activation of dlt transcription should result in a neutralization of the negative charge of adjacent phosphoryl residues of this anionic polymer, eventually leading to an increased resistance of the cells to some antibiotics. In fact, mutants lacking D-alanine on teichoic acids displayed an increased sensitivity: of B. subtilis to methicillin (Wecke et al., 1997), of S. agalactiae to several cationic antimicrobial peptides (Poyart et al., 2003) and of S. aureus to gallidermin or nisin (Peschel et al., 1999).

We therefore suggest that the yxdLMyxeA operon gene products might be involved in resistance to an as yet unknown group of antibiotics, and that yxdJK encodes the corresponding signal detector/transducer system.

Related to yxdLM, the gene yxeA encodes a long peptide which is conserved in several Gram-positive bacteria: Bacillus anthracis, Enterococcus faecalis, Lactococcus lactis, Listeria innocua, Listeria monocytogenes and Staphylococcus aureus. This peptide might participate in the proposed antibiotic resistance mechanism as an immunity peptide interacting with and neutralizing the antibiotic. It is predicted to be processed (Nielsen et al., 1997) and exported via the general secretion pathway. In that case the YxdLM ABC transporter might work in conjunction with YxeA as an antibiotic efflux pump. However, one cannot exclude that YxeA might be exported by the ABC transporter YxdLM to protect the cell as indicated above. We are currently studying these different possibilities.

We thank Marc Chippaux, Maryline Foglino, Anne Galinier, Vincent Méjean and Abraham L. Sonenshein for helpful discussions, Athel Cornish Bowden for critical review of the manuscript and Martine Chartier for help with the primer extension analysis. This work was supported in part by grants from CNRS, HMR (FRHMR2/9932), CNRS puce à ADN, an MRT fellowship and a fellowship from the Fondation pour la Recherche Médicale to P. J.

Footnotes

†Present address: Department of Molecular Biology and Microbiology, Tufts University School of Medicine, 136 Harrison Avenue, Boston, MA 02111, USA.