Bacitracin and nisin resistance in Staphylococcus aureus: a novel pathway involving the BraS/BraR two-component system (SA2417/SA2418) and both the BraD/BraE and VraD/VraE ABC transporters


  • Aurélia Hiron,

    1. Institut Pasteur, Biology of Gram-Positive Pathogens, Department of Microbiology, 25 rue du Docteur Roux, 75724 Paris Cedex 15, France
    2. CNRS, URA 2172, F-75015 Paris, France
    Search for more papers by this author
  • Mélanie Falord,

    1. Institut Pasteur, Biology of Gram-Positive Pathogens, Department of Microbiology, 25 rue du Docteur Roux, 75724 Paris Cedex 15, France
    2. CNRS, URA 2172, F-75015 Paris, France
    Search for more papers by this author
  • Jaione Valle,

    1. Instituto de Agrobiotecnología, Universidad Pública de Navarra-CSIC, Pamplona, 31006, Spain
    Search for more papers by this author
  • Michel Débarbouillé,

    1. Institut Pasteur, Biology of Gram-Positive Pathogens, Department of Microbiology, 25 rue du Docteur Roux, 75724 Paris Cedex 15, France
    2. CNRS, URA 2172, F-75015 Paris, France
    Search for more papers by this author
  • Tarek Msadek

    Corresponding author
    1. Institut Pasteur, Biology of Gram-Positive Pathogens, Department of Microbiology, 25 rue du Docteur Roux, 75724 Paris Cedex 15, France
    2. CNRS, URA 2172, F-75015 Paris, France
      E-mail; Tel. (+33) 1 45 68 88 09; Fax (+33) 1 45 68 89 38.
    Search for more papers by this author

E-mail; Tel. (+33) 1 45 68 88 09; Fax (+33) 1 45 68 89 38.


Two-component systems (TCSs) are key regulatory pathways allowing bacteria to adapt their genetic expression to environmental changes. Bacitracin, a cyclic dodecylpeptide antibiotic, binds to undecaprenyl pyrophosphate, the lipid carrier for cell wall precursors, effectively inhibiting peptidoglycan biosynthesis. We have identified a novel and previously uncharacterized TCS in the major human pathogen Staphylococcus aureus that we show to be essential for bacitracin and nisin resistance: the BraS/BraR system (Bacitracin resistance associated; SA2417/SA2418). The braRS genes are located immediately upstream from genes encoding an ABC transporter, accordingly designated BraDE. We have shown that the BraSR/BraDE module is a key bacitracin and nisin resistance determinant in S. aureus. In the presence of low antibiotic concentrations, BraSR activate transcription of two operons encoding ABC transporters: braDE and vraDE. We identified a highly conserved imperfect palindromic sequence upstream from the braDE and vraDE promoter sequences, essential for their transcriptional activation by BraSR, suggesting it is the likely BraR binding site. We demonstrated that the two ABC transporters play distinct and original roles in antibiotic resistance: BraDE is involved in bacitracin sensing and signalling through BraSR, whereas VraDE acts specifically as a detoxification module and is sufficient to confer bacitracin and nisin resistance when produced on its own. We show that these processes require functional BraD and VraD nucleotide-binding domain proteins, and that the large extracellular loop of VraE confers its specificity in bacitracin resistance. This is the first example of a TCS associated with two ABC transporters playing separate roles in signal transduction and antibiotic resistance.


Staphylococcus aureus is a major Gram-positive human pathogen responsible for a broad spectrum of infections, ranging from superficial skin abscesses to more serious diseases such as pneumonia, meningitis, endocarditis, septicaemia and toxic shock syndrome (Lowy, 1998). The heterogeneity of these diseases and the unique ability of S. aureus to adapt to a great variety of environments have made it one of the main causes of hospital- and community-acquired infections. Treatment of staphylococcal infections is becoming increasingly difficult due to the worldwide emergence of multiple antibiotic resistance determinants (Lowy, 2003), emphasizing the need to further our understanding of the molecular mechanisms involved in antimicrobial resistance.

Bacitracin, a branched cyclic dodecylpeptide antibiotic synthesized by Bacillus licheniformis and some strains of Bacillus subtilis (Johnson et al., 1945; Azevedo et al., 1993; Ishihara et al., 2002), is active against Gram-positive bacteria and used to prevent and treat skin and ophthalmic infections. Bacitracin binds very tightly to undecaprenyl pyrophosphate (UPP) (Stone and Strominger, 1971), the lipid carrier responsible for translocation of cell envelope precursors from the cytosol to the extracellular surface of the cytoplasmic membrane, effectively inhibiting peptidoglycan biosynthesis and leading to cell lysis.

Gram-positive bacteria have developed several mechanisms of bacitracin resistance (Cain et al., 1993; Chalker et al., 2000; Bernard et al., 2005). Among these, the B. subtilis BceSR two-component system (TCS)/BceAB ABC transporter is the most efficient and well studied, constituting the prototype for TCS/ABC transporter modules, where the genes are genetically linked (Ohki et al., 2003). TCSs are one of the principal means through which bacteria sense, respond and adapt to changes in their environment, and are typically composed of a membrane histidine kinase (HK), acting as a signal sensor/transducer through phosphorylation of its cognate response regulator (RR), which acts as a transcription activator or repressor. The BceS HK belongs to the so-called 'intramembrane-sensing kinase' (IMSK) subfamily, recently defined as conserved in low G+C % Gram-positive bacteria and characterized by a very short amino-terminal sensing domain, composed of two transmembrane helices separated by a small loop of only a few amino acids, which is thought to be buried in the cytoplasmic membrane (Mascher, 2006). These structural features suggest a kinase sensing process occurring at or from within the membrane interface, and members of this family appear to be involved in responding to cell envelope stress, very often exerted by cell wall active antibiotics (Mascher, 2006; Mascher et al., 2006; Jordan et al., 2008). BceSR control the synthesis of the ABC transporter composed of BceA, the nucleotide-binding domain (NBD) protein, and BceB, the membrane-spanning domain (MSD) protein or permease, with has 10 transmembrane domains and a long extracytoplasmic loop (202 aa) (Mascher et al., 2003; Ohki et al., 2003; Rietkötter et al., 2008). In B. subtilis this transporter was shown to be essential for bacitracin sensing by the BceSR TCS, highlighting a new ABC transporter-dependent mechanism for TCS-mediated signal transduction (Bernard et al., 2007).

Most S. aureus strains are endowed with 16 sets of genes encoding TCSs, with an additional one present in the staphylococcal cassette chromosome mec of MRSA, linked to induction of methicillin resistance (Kuroda et al., 2001). This sophisticated arsenal of environmental monitoring proteins could, in part, explain the highly adaptive nature of S. aureus. Although many of these systems remain to be characterized, several have been shown to play a role in virulence (AgrCA, SaeSR) (Giraudo et al., 1999; Novick, 2003), antibiotic resistance (VraSR, GraSR) (Gardete et al., 2006; Herbert et al., 2007; Meehl et al., 2007) or cell wall metabolism (WalKR) (Martin et al., 1999; Dubrac and Msadek, 2004; Dubrac et al., 2007; Dubrac et al., 2008; Dubrac and Msadek, 2008). Only two of the S. aureus TCSs with IMSKs form genetically linked TCS/ABC transporter modules sharing some characteristics with the BceSR/BceAB system: GraSR/VraFG, playing an essential role in antimicrobial peptide resistance (Herbert et al., 2007; Li et al., 2007; Meehl et al., 2007), and the uncharacterized SA2417/SA2418 TCS whose genes are located upstream from those of a putative ABC transporter (SA2416/SA2415). While the existence of an inducible bacitracin resistance mechanism in S. aureus had been suggested as early as 1949 (Stone, 1949), no specific orthologue of the BceSR/BceAB module has been described in S. aureus. However, mutants lacking the VraDE ABC transporter, closely related to BceAB but not genetically linked to TCS-encoding genes, were shown to have increased bacitracin sensitivity (Sass et al., 2008; Pietiäinen et al., 2009).

In this study we identified a novel TCS encoded by the SA2417/SA2418 genes as a key regulatory element allowing bacitracin and nisin resistance in S. aureus, which we have named BraSR. While this manuscript was under review, independent reports have since referred to this system as BceSR or NsaSR (Blake et al., 2011; Yoshida et al., 2011), but we believe the BraSR nomenclature to be more appropriate (see Discussion). In the presence of bacitracin or nisin, we show that BraSR activate transcription of the braDE and vraDE operons, encoding two ABC transporters, which play distinct and original roles in bacitracin and nisin resistance. Indeed, BraDE is only involved in antibiotic sensing and signalling through BraSR, whereas VraDE acts specifically as a detoxification module and is sufficient to confer bacitracin and nisin resistance when expressed on its own. This is the first example of a TCS associated with two ABC transporters involved in separate functions in signal transduction and antibiotic resistance.


Identification of a new TCS/ABC transporter module involved in bacitracin and nisin resistance in S. aureus

Of the 16 TCSs present in S. aureus, only two have IMSKs with the characteristic features of the Bce-like family: GraSR, involved in cationic antimicrobial peptide resistance (Herbert et al., 2007) and the uncharacterized SA2417/SA2418 system. Indeed, the genes encoding these two TCSs are adjacent to those encoding the VraFG and SA2416/SA2415 ABC transporters, respectively, and both the GraS and SA2417 kinases have extremely short extracellular loops (seven and three amino acids in length respectively).

We have constructed a complete collection of S. aureus mutants inactivated for each of the TCS-encoding gene pairs (J. Valle, A. Hiron and M. Débarbouillé, unpubl. results) in the HG001 reference strain (Herbert et al., 2010). Mutants lacking the VraFG, VraDE and SA2416/SA2415 ABC transporters were also constructed. Indeed, both the VraDE ABC transporter and the VraSR TCS, whose kinase belongs to the LiaS-like IMSK family (Mascher, 2006), have been reported to be involved in bacitracin resistance in S. aureus (Pietiäinen et al., 2009). The sensitivity of each mutant strain to bacitracin was tested (Table 1), allowing us to identify SA2417/SA2418 as a novel TCS essential for bacitracin resistance, which we have accordingly renamed BraS/BraR (Bacitracin resistance associated Sensor/Regulator). As shown in Table 1, mutants lacking the SA2416/SA2415 genes, encoding an ABC transporter, also displayed hypersensitivity to bacitracin, and were renamed braD and braE respectively. As expected, the vraDE mutant and, to a lesser extent, the vraSR mutant were also more sensitive to bacitracin in our genetic background, whereas neither the absence of the GraSR TCS, which is the most closely related to the B. subtilis BceS/BceR bacitracin resistance TCS, nor the absence of the associated VraFG ABC transporter affected bacitracin resistance (Table 1).

Table 1.  Role of S. aureus two-component systems and ABC transporters in bacitracin and nisin resistance.
StrainRelevant genotypeMIC bacitracin (µg ml−1)MIC nisin (µg ml−1)
HG001NCTC 8325 rsbU+48>128

To determine the substrate specificity of the previously uncharacterized BraSR TCS, sensitivity of the ΔbraRS mutant strain to a wide spectrum of antimicrobial agents affecting different cellular processes was tested. No difference in vancomycin, fosfomycin, oxacillin, colistin, capreomycin, viomycin or daptomycin sensitivity was observed between the wild-type and mutant strain (data not shown) but the ΔbraRS mutant exhibited highly increased sensitivity to nisin (Table 1). We also determined the nisin minimal inhibitory concentration (MIC) values for mutants lacking the other TCS/ABC modules and showed that both the ΔvraDE and ΔbraDE strains are highly sensitive to nisin (Table 1), as well as the ΔgraRS and ΔvraFG mutants, as previously described (Li et al., 2007).

Genetic structure and transcriptional organization of the bra and vraDE loci

Close examination of the braRS locus indicates that the two genes are located 25 bp downstream from a gene annotated as encoding a 66-amino-acid hypothetical protein (SA2419) and 107 bp upstream from the braDE genes (Fig. 1A). We performed RT-PCR analysis on total RNA extracted from cells grown in TSB with 0.5 µg ml−1 bacitracin, indicating that SA2419 is co-transcribed as an operon with the braRS genes and that the braDE genes are co-transcribed as a separate operon (Fig. 1A and data not shown). A ΔSA2419 mutant strain was constructed and we showed that this gene does not play a role in bacitracin resistance (data not shown). Primer extension experiments were performed to identify the transcription start sites of the braDE and vraDE operons (Fig. 1B), which were found to be preceded by appropriately spaced −10 and −35 regions sharing similarities with the consensus sequence of promoters recognized by the vegetative form of RNA polymerase, EσA, particularly the −10 sequences, whereas the −35 regions were less conserved (Fig. 1C).

Figure 1.

Genetic structure and primer extension analysis of the braRS, braDE and vraDE loci. A. Genetic organization of the braRS and braDE operons. Genes are indicated by grey arrows, operon transcripts by double-headed horizontal arrows and the braDE operon transcription initiation site by a horizontal arrow. B. Primer extension analysis of braDE and vraDE mRNAs. Total RNA isolated from strain HG001 grown in TSB with 0.5 µg ml−1 bacitracin was used as a template for primer extension experiments using braD- and vraD-specific radiolabelled oligonucleotides. The corresponding Sanger dideoxy chain termination sequencing reactions were carried out on a PCR fragment corresponding to the respective operon upstream regions (lanes C, T, A, G). C. Nucleotide sequences of the braDE and vraDE operon promoter regions. The −35 and −10 promoter sequences are boxed, the transcription initiation start sites are labelled +1 and the ATG translation initiation codons are indicated in bold.

Expression of the braDE and vraDE operons is induced by bacitracin or nisin

In the BceRS/BceAB module of B. subtilis, the bceAB genes encoding the ABC transporter are strongly induced in the presence of sublethal concentrations of bacitracin (Ohki et al., 2003). Since the braDE, vraDE and vraFG operons are all involved in bacitracin and nisin resistance, we investigated their expression during growth in the presence of a wide range of subinhibitory concentrations of these two antibiotics. Transcriptional lacZ fusions with the three operon promoters were constructed using the pSA14 vector and introduced into S. aureus strain HG001 (see Experimental procedures). β-Galactosidase activities were measured during mid-exponential growth (OD600 = 2) with increasing sublethal bacitracin and nisin concentrations (ranging from 0.06 to 2 µg ml−1), and are shown as fold induction with respect to the reference value of braDE–lacZ or vraDE–lacZ expression in bacteria grown in TSB without antibiotic as an inducer (approximately 15 Miller units per mg protein for braDE and 5 for vraDE). We showed that expression of the braDE genes is significantly induced in the presence of low bacitracin concentrations (ninefold at 0.5 µg ml−1) but not by nisin (Fig. 2A). Expression of vraDE was induced by both bacitracin and nisin, even at low concentrations (Fig. 2B), but the level of induction was much stronger with bacitracin than with nisin (500-fold and 7-fold at 0.5 µg ml−1 respectively). Expression from the vraFG promoter on the other hand was not induced by either of the two antibiotics under these conditions (data not shown).

Figure 2.

Induction of vraDE and braDE expression as a function of bacitracin and nisin concentrations. Antibiotic concentrations (µg ml−1) are shown on the x-axis and plotted against fold induction values (y-axis). The reference expression value is that of bacteria grown in TSB without antibiotic. Black bars represent induction in the presence of bacitracin and grey bars in the presence of nisin. β-Galactosidase assays were performed as described in Experimental procedures. Results are presented as the average of three independent assays. The strains used for promoter activity measurements are (A) ST1245: pSA14-PbraDE–lacZ and (B) ST1208: pSA14-PvraDE–lacZ.

The BraSR TCS controls inducible expression of the braDE and vraDE operons but does not regulate its own synthesis

As shown above, the BraSR, VraSR and GraSR TCSs are each involved to various extents in bacitracin and nisin resistance. We therefore wished to determine which of these systems might be involved in controlling braDE and vraDE expression in response to these antibiotics. Plasmids carrying braDE–lacZ and vraDE–lacZ transcriptional fusions were introduced into the ΔvraSR, ΔgraRS and ΔbraRS mutant strains, and β-galactosidase assays were performed during growth in TSB with or without 0.5 µg ml−1 bacitracin or nisin. Expression of braDE and vraDE was no longer induced by bacitracin in the ΔbraRS mutant while no significant difference with the parental strain was observed in the ΔvraSR and ΔgraRS mutant strains (Fig. 3A and B). Similar results were obtained for vraDE expression in the presence of nisin (Fig. 3B). These data clearly demonstrate that induction of the vraDE and braDE operons in response to bacitracin or nisin is mediated by the BraSR TCS but is not controlled by VraSR or GraSR. We also verified that expression of vraFG was not affected in the ΔbraRS mutant in the presence or absence of bacitracin (data not shown).

Figure 3.

Inducible expression of the braDE and vraDE operons is controlled by the BraSR TCS. β-Galactosidase assays (A) and (B) and qRT-PCRs (C) were performed as described in Experimental procedures. Cells were grown until mid-exponential phase (OD600 = 2). β-Galactosidase activities and qRT-PCR products on the y-axis are represented as fold induction against the reference value, i.e. that of bacteria grown in TSB in the absence of bacitracin or nisin. Black and grey bars represent expression levels in the presence of 0.5 µg ml−1 bacitracin or 0.5 µg ml−1 nisin respectively. A. Effect of braRS, vraSR and graRS deletions on PbraDE expression. The strains used are (from left to right): ST1245 (pSA14-PbraDE–lacZ), ST1087 (ΔbraRS pSA14-PbraDE–lacZ), ST1092 (ΔvraSR pSA14-PbraDE–lacZ), ST1091 (ΔgraRS pSA14-PbraDE–lacZ). Results are presented as the average of three independent assays. B. Effect of braRS, vraSR and graRS deletions on PvraDE expression. The strains used are (from left to right): ST1208 (pSA14-PvraDE–lacZ), ST1088 (ΔbraRS pSA14-PvraDE–lacZ), ST1092 (ΔvraSR pSA14-PvraDE–lacZ), ST1089 (ΔgraRS pSA14-PvraDE–lacZ). Results are presented as the average of three independent assays. C. Relative levels of vraDE transcripts were measured by qRT-PCRs. Expression levels were normalized using 16S rRNA as an internal standard and are indicated as n-fold change, expressed as the means and standard deviations of quadruplicate experiments. The strains used are (from left to right): HG001 (reference strain), ST1179 (ΔbraRS) and ST1184 (ΔbraRS pMK4-Pprot braRS).

The ΔbraRS mutant was complemented by expressing the braRS genes from the pMK4-Pprot multicopy plasmid (Archambaud et al., 2005), fully restoring resistance to bacitracin and nisin (Table 2), as well as bacitracin-dependent induction of the vraDE genes as determined by quantitative real-time PCR (qRT-PCR) (Fig. 3C).

Table 2.  Effects of vraDE and braDE complementation on bacitracin and nisin resistance.
StrainRelevant genotypeMIC bacitracin (µg ml−1)MIC nisin (µg ml−1)
HG001NCTC 8325 rsbU+48> 128
ST1216ΔbraDE pMK4-Pprot64
ST1187ΔbraDE pMK4-Pprot braDE48> 128
ST1188ΔbraDE pMK4-Pprot vraDE16> 128
ST1232ΔbraDE pMK4-Pprot braR32> 128
ST1217ΔvraDE pMK4-Pprot64
ST1192ΔvraDE pMK4-Pprot braDE64
ST1191ΔvraDE pMK4-Pprot vraDE16> 128
ST1233ΔvraDE pMK4-Pprot braR64
ST1238ΔbraRS pMK4-Pprot64
ST1185ΔbraRS pMK4-Pprot braDE64
ST1186ΔbraRS pMK4-Pprot vraDE16> 128
ST1184ΔbraRS pMK4-Pprot braRS48> 128
ST1121ΔbraRS pMK4-Pprot braR32> 128
ST1239ΔvraDEΔbraDE pMK4-Pprot64
ST1240ΔvraDEΔbraDE pMK4-Pprot braDE64
ST1218ΔvraDEΔbraDE pMK4-Pprot vraDE16> 128

As many TCSs are known to positively autoregulate their own synthesis, a transcriptional lacZ fusion with the SA2419/braRS operon promoter region was constructed. The SA2419/braRS–lacZ fusion was expressed constitutively in the absence of bacitracin (approximately 400–500 Miller units per mg protein), and this expression did not vary in the presence of bacitracin concentrations ranging from 0.0625 to 2 µg ml−1 or in the ΔbraRS mutant strain ST1179, indicating that BraSR do not control their own synthesis (data not shown).

Identification of a conserved palindromic sequence as the likely BraR binding site

We have shown that expression of the vraDE and braDE operons is controlled by BraSR in response to antibiotics. An in silico approach was used to identify potential regulatory DNA motifs in the promoter regions of these two co-regulated operons. A highly conserved imperfect 14-base-pair palindromic motif (CTTTCAA NN T/CTGTAAG) was identified, located 66 and 64 bp upstream from the braDE and vraDE transcription start sites respectively (Fig. 4A). To test whether this conserved sequence is necessary for expression of the two operons in response to bacitracin and/or nisin, a series of truncated promoter fragments were generated and used to construct transcriptional lacZ fusions using plasmid pSA14 (Fig. 4B). Removal of the braDE and vraDE upstream sequences up to the inverted repeat had no effect on promoter activity; however, removal of half or all of the inverted repeat completely abolished bacitracin- and nisin-induced expression (Fig. 4B and C). Three point mutations were introduced by site-directed mutagenesis through PCR in one of the inverted repeats, effectively destroying the palindromic operator sequence and leading to the complete loss of induction of expression from the braDE and vraDE promoters by bacitracin or nisin (Fig. 4A–C). Together, these results indicate that the conserved inverted repeat upstream from the vraDE and braDE promoters is essential for the induction of promoter activity in the presence of bacitracin and nisin, strongly suggesting it is the likely binding site for the BraR RR. Genome scanning analysis, using the AureoList relational database (, indicates that no S. aureus genes other than the braDE and vraDE operons are preceded by this sequence.

Figure 4.

A highly conserved imperfect palindromic sequence within the braDE and vraDE promoters is required for bacitracin-dependent promoter activation. A. Alignment of the braDE and vraDE promoter region nucleotide sequences. The highly conserved palindromic sequence is shaded and boxed and indicated by inverted arrows. Point mutations introduced in the braDE′ and vraDE′–lacZ fusions are indicated below the alignment by stars. Nucleotide positions are given with respect to the transcription initiation sites. B and C. β-Galactosidase assays of braDE (B) and vraDE (C) promoter transcriptional fusions. The braDE and vraDE promoter regions were deleted up to the inverted repeat (ΔB), half way through the inverted repeat (ΔC), just after the potential BraR binding site (ΔD) or carried 3 point mutations destroying the inverted repeat (ΔE). Positions are numbered with respect to the transcription initiation site. S. aureus strains carrying the indicated transcriptional fusions were grown in triplicate in TSB with 0.5 µg ml−1 bacitracin (black bars) or 0.5 µg ml−1 nisin (grey bars). β-Galactosidase activities are given as the average of three independent assays and represented as fold induction with respect to the reference value, i.e. braDE–lacZ or vraDE–lacZ activity of bacteria grown in TSB without antibiotic as an inducer.

The BraDE ABC transporter is involved in antibiotic detection

In B. subtilis, the BceAB ABC transporter has been shown to be required for its own synthesis (Bernard et al., 2007). As shown above, two separate ABC transporters, VraDE and BraDE, are both involved in S. aureus bacitracin and nisin resistance. In order to determine their respective contributions in signal perception and antibiotic resistance, braDE–lacZ and vraDE–lacZ expression was measured in ΔvraDE and ΔbraDE mutant strains. As shown in Fig. 5A and B, the BraDE ABC transporter is essential for bacitracin and nisin-dependent transcription from both the braDE and vraDE promoters, whereas the VraDE ABC transporter was not required. To confirm these results, expression of the braDE and vraDE genes was uncoupled from BraSR-dependent regulation using the pMK4-Pprot plasmid, and used to complement the ΔbraDE and ΔvraDE mutant strains respectively. Expression of vraDE was measured by qRT-PCR in the wild-type strain HG001, the ΔbraDE mutant strain (ST1097) and the ΔbraDE complemented strain (ST1187), grown in the presence of bacitracin (0.5 µg ml−1). As shown in Fig. 5C, the BraDE ABC transporter is essential for induction of vraDE expression by bacitracin. Taken together, these results indicate that only the BraDE ABC transporter is essential for bacitracin and nisin detection by BraSR.

Figure 5.

BraDE is required for bacitracin-dependent induction of braDE and vraDE expression. β-Galactosidase assays (A and B) and qRT-PCRs (C) were performed as described in Experimental procedures. Cells were grown until mid-exponential phase (OD600 = 2). β-Galactosidase activities and qRT-PCR products are represented as fold induction against the reference value, i.e. bacteria grown in TSB without antibiotic as an inducer. Black and grey bars represent expression levels in the presence of 0.5 µg ml−1 bacitracin or 0.5 µg ml−1 nisin respectively. A. Effect of braRS, braDE and vraDE deletions on PbraDE expression. The strains used are (from left to right): ST1245 (pSA14-PbraDE–lacZ), ST1087 (ΔbraRS pSA14-PbraDE–lacZ), ST1189 (ΔbraDE pSA14-PbraDE–lacZ), ST1195 (ΔvraDE pSA14-PbraDE–lacZ). B. Effect of braRS, braDE and vraDE deletions on PvraDE expression. The strains used are (from left to right): ST1208 (pSA14-PvraDE–lacZ), ST1088 (ΔbraRS pSA14-PvraDE–lacZ), ST1190 (ΔbraDE pSA14-PvraDE–lacZ), ST1230 (ΔbraD pSA14-PvraDE–lacZ), ST1196 (ΔvraDE pSA14-PvraDE–lacZ). C. Relative levels of vraDE transcripts were measured by qRT-PCR. Expression levels were normalized using 16S rRNA as an internal standard and are indicated as n-fold change with respect to those of bacteria grown in TSB in the absence of bacitracin, expressed as the means and standard deviations of quadruplicate experiments. The strains used are (from left to right): HG001, ST1097 (ΔbraDE) and ST1187 (ΔbraDE pMK4-Pprot braDE).

The VraDE ABC transporter acts as a resistance module

We then addressed the question of the role of the BraDE and VraDE ABC transporters in antibiotic detoxification. Complementation by the pMK4-Pprot braDE plasmid fully restored bacitracin and nisin resistance to the ΔbraDE mutant and the pMK4-Pprot vraDE plasmid restored significant levels of bacitracin resistance and full resistance to nisin to the ΔvraDE mutant strain (Table 2). To explain this partial complementation of bacitracin resistance in the ΔvraDE mutant, vraDE and braDE mRNA levels were compared by qRT-PCR in the wild-type and complemented strains. In comparison with the wild-type strain in the presence of bacitracin, expression levels of braDE from the pMK4-Pprot plasmid were threefold higher, whereas plasmid-based expression of vraDE was 3.5-fold lower in the ΔvraDE complemented strain. This result is consistent with our observation that expression of the vraDE genes is much more highly induced in the presence of 0.5 µg ml−1 bacitracin than that of braDE (approximately 500-fold and 7-fold respectively) (Fig. 1).

The pMK4-PprotbraDE and pMK4-PprotvraDE plasmids were introduced into different mutant strains as shown in Table 2 and sensitivity to bacitracin and nisin was compared. Interestingly, constitutive expression of vraDE on its own bypasses the requirement for BraSR and BraDE (Table 2) restoring bacitracin and nisin resistance. In the reverse experiment, expression of braDE could not complement the ΔbraRS and ΔvraDE mutant stains (Table 2).

These results suggest that the VraDE ABC transporter plays a direct role in the bacitracin detoxification process whereas BraDE is only required for bacitracin sensing and/or signal transduction through BraSR. To confirm this last point, the braR gene was expressed alone using the pMK4-Pprot plasmid. Overexpression of RR genes is well known to bypass the requirement for the cognate HK, leading to constitutive activation due to other phosphate donors such as acetyl phosphate or aspecific kinase activity within the cell (Kobayashi et al., 2001). As shown in Table 2, the pMK4-PprotbraR plasmid restores significant levels of bacitracin resistance to both the ΔbraRS and ΔbraDE mutant strains but was not able to complement the ΔvraDE mutant (Table 2). These results confirm that, contrary to VraDE, the BraDE ABC transporter only appears to be required for sensing and BraSR-dependent signal transduction, leading to phosphorylation of the BraR RR and antibiotic resistance.

Bacitracin sensing and resistance requires functional BraD and VraD ABC transporter NBD proteins

As described above, the BraDE and VraDE ABC transporters play specific roles in bacitracin detection and resistance respectively. In order to determine whether these processes are energy-dependent, the braD or vraD genes, encoding the ABC transporter NBD proteins were deleted. Both NBD mutant strains were as sensitive to bacitracin as the ΔbraDE and ΔvraDE mutant strains (Table 3). As shown in Fig. 5, inducible expression of vraDE and braDE by bacitracin and nisin was abolished in the ΔbraD mutant strain (Fig. 5B). Since the braDE and vraDE genes are co-transcribed as an operon, removing the first gene might have a polar effect on expression of the downstream permease gene. To rule out that possibility, we complemented the ΔbraD and ΔvraD mutants using the pMK4-Pprot plasmid, fully restoring bacitracin resistance (Table 3). We also note that both ATPases were highly specific to their cognate permease as no cross-complementation between vraD and braD was observed (data not shown).

Table 3.  BraD- and VraD-dependent bacitracin resistance requires a functional ATP-binding domain.
StrainRelevant genotypeMIC bacitracin (µg ml−1)
ST1210ΔbraD pMK4-Pprot6
ST1197ΔbraD pMK4-Pprot braD48
ST1206ΔbraD pMK4-Pprot braD*WB6
ST1231ΔvraD pMK4-Pprot6
ST1198ΔvraD pMK4-Pprot vraD16
ST1205ΔvraD pMK4-Pprot vraD*WB6

To investigate whether ATP hydrolysis by BraD or VraD is critical for their role in antibiotic resistance, single-point mutations were introduced by site-directed mutagenesis through PCR in the Walker B motif-coding sequences of braD (E168Q) and vraD (E168Q). This mutation has been described to abolish ATP hydrolysis, but not binding, while simultaneously stabilizing the NBD dimer and maintaining the native ATP-bound structure (Moody et al., 2002). The mutated alleles were expressed using the pMK4-Pprot plasmid (pMK4-PprotbraD*WB and pMK4-PprotvraD*WB respectively), and were both unable to complement the respective ΔbraD and ΔvraD mutants in contrast to the wild-type alleles (Table 3).

Taken together, these results strongly suggest that both the BraDE and VraDE ABC transporters require ATP hydrolysis to exert their respective functions in antibiotic sensing and resistance.

The extracellular loop of the VraE permease confers its specificity for bacitracin resistance

As shown above, VraDE plays a central role in bacitracin resistance. The predicted membrane topology and primary structure of this ABC transporter are very similar to those of VraFG (62% amino acid sequence identity between VraD and VraF and 39% between VraE and VraG), whereas the BraDE ABC transporter is not closely related (Fig. 6). The VraFG system is involved in resistance to cationic antimicrobial peptides and its synthesis is controlled by the GraSR TCS (Li et al., 2007; Meehl et al., 2007). We have shown that neither GraSR nor VraFG is involved in bacitracin resistance (Table 1) and also that vraFG expression is not controlled by BraSR and is not induced in the presence of bacitracin (data not shown). As shown in Fig. 6B, sequence alignment of the VraE and VraF permeases indicated that the most dissimilar sequence between the two proteins corresponds to a large extracytoplasmic loop of respectively 193 and 198 amino acids in length, located between transmembrane regions 7 and 8 (Fig. 6A). Amino acid sequence identity between these two loops is only 20% whereas the rest of the two proteins share 53% identity (Fig. 6B). In B. subtilis, the extracellular loop of BceB is thought to be necessary for bacitracin sensing and resistance (Rietkötter et al., 2008; Coumes-Florens et al., 2011).

Figure 6.

Topology and sequence comparison between ABC transporter membrane-spanning domain proteins. A. Membrane topology of VraE, VraG and BraE. The graphical representation of the membrane topology prediction was performed using the TopPredII program and von Heijne's algorithm (Claros and von Heijne, 1994). The extracytoplasmic loop between the predicted membrane-spanning helices 7 and 8 has a respective length of 193, 198 and 231 amino acids for VraE, VraG and BraE. B. Alignment of the VraE, VraG and BraE amino acid sequences. Sequence alignment was performed using the clustal program (Higgins et al., 1992). Identical residues are shaded. The predicted extracellular loops of the VraE and VraG permeases that were exchanged by domain-swapping to construct the VraFG*VraE chimerical permease are boxed.

To investigate the role of the VraDE extracellular loop in S. aureus, we took advantage of the close similarities between VraE and VraG in order to construct a chimeric ABC transporter by domain-swapping, exchanging the extracellular loop of VraG for that of VraE and expressing the recombinant gene using the pMK4-Pprot plasmid. The resulting plasmid, pMK4-PprotvraFG*vraE, was introduced into the ΔvraDE mutant strain and was as efficient in restoring bacitracin resistance as the pMK4-PprotvraDE complementing plasmid (Table 4). We verified that the pMK4-PprotvraFG plasmid did not confer resistance to bacitracin but was able to fully complement a ΔvraFG mutant strain for resistance to colistin, a cationic antimicrobial peptide (Table 4). These results demonstrate that the extracellular loop of the VraE permease is the determinant of its specificity in bacitracin resistance. Interestingly, we also showed that the pMK4-Pprot vraFG*vraE plasmid did not restore colistin resistance to a ΔvraFG mutant strain (Table 4), confirming the importance of the extracellular loop in the specificity of substrate recognition.

Table 4.  The extracellular loop of VraE is essential for bacitracin detection.
StrainRelevant genotypeMIC (µg ml−1)
  1. vraG*vraE, vraG with the extracellular loop-coding sequence of vraE; braE*vraE, braE with the extracellular loop-coding sequence of vraE.

HG001NCTC 8325 rsbU+48
ST1217ΔvraDE pMK4-Pprot6
ST1191ΔvraDE pMK4-Pprot vraDE16
ST1193ΔvraDE pMK4-Pprot vraFG6
ST1194ΔvraDE pMK4-Pprot vraFG*vraE16
ST1234ΔvraDE pMK4-Pprot braDE*vraE6
ST1216ΔbraDE pMK4-Pprot6
ST1235ΔbraDE pMK4-Pprot braDE*vraE6
ST1120HG001 pMK4-Pprot700
ST1150ΔvraFG pMK4-Pprot100
ST1151ΔvraFG pMK4-Pprot vraFG700
ST1177ΔvraFG pMK4-Pprot vraFG*vraE200

We also constructed the pMK4-PprotbraDE*vraE plasmid, expressing a chimeric system with the BraE extracellular loop replaced by that of VraE. This plasmid could not complement the ΔbraDE or ΔvraDE mutant strains for bacitracin resistance (Table 4). Although we cannot exclude that the chimeric BraE*VraE permease may be non-functional because of the greater diversity between BraE and VraE, this result could suggest that unlike VraDE, the extracellular loop of the permease is not the sole determinant involved in bacitracin sensing and signalling by the BraDE ABC transporter.


In this study, we identified a novel TCS/ABC transporter module, BraSR/BraDE, as the main regulatory element controlling bacitracin and nisin resistance in S. aureus. We showed that together with VraDE, the BraSR/BraDE module constitutes an original mechanism for antibiotic sensing and resistance.

The BraSR TCS responds to low bacitracin concentrations by inducing expression of the braDE and vraDE genes, encoding two ABC transporters (Fig. 7). The ΔbraRS, ΔbraDE and ΔvraDE mutant strains are all highly sensitive to bacitracin (Table 1), suggesting that the BraSR/BraDE/VraDE system could be functionally equivalent to the BceSR/BceAB module of B. subtilis (Ohki et al., 2003). Among the panel of antibiotics tested in this study, the BraSRDE/VraDE multicomponent system displayed restricted substrate specificity since it appears to only be involved in bacitracin and nisin resistance.

Figure 7.

Signal transduction network controlling bacitracin sensing and resistance in S. aureus. Bacitracin is sensed by BraDE and the signal is transduced to BraSR through a mechanism that remains to be elucidated, but which likely involves interaction between BraE and BraS. Activation of the BraSR system leads to BraS-dependent phosphorylation of BraR and subsequent transcription activation of the braDE and vraDE operons. The VraDE transporter then removes bacitracin. As the direction of bacitracin transport by VraDE, i.e. import or export, is not known, it is indicated by a vertical double-headed arrow, but we have demonstrated that the extracellular loop of VraE is a determinant of its specificity and activity.

In contrast, the VraSR TCS plays a minor role in bacitracin resistance (Table 1) but the ΔvraSR mutant is sensitive to a large range of antimicrobial compounds (Pietiäinen et al., 2009). VraSR is thought to sense cell wall damage to co-ordinate a general cell envelope stress response in S. aureus (Belcheva and Golemi-Kotra, 2008) and thus could be considered as the equivalent of the B. subtilis LiaSR system (Mascher et al., 2004). Two distinct pathways contributing to bacitracin resistance can therefore be distinguished in S. aureus: the highly efficient, sensitive and specific BraSR/BraDE/VraDE multicomponent system, and the damage-sensing VraSR system, less effective but responding to general envelope stress conditions.

The BraSR and GraSR TCSs share some common features: they both have IMSK family kinases and control the synthesis of closely related ABC transporters (VraDE and VraFG) (Fig. 6B), and they are both essential for nisin resistance. A transcriptome analysis previously suggested that expression of the vraDE operon is controlled by GraSR in response to indolicidin (Li et al., 2007). Indeed, they observed an induction of approximately ninefold in vraDE expression upon addition of 0.5 µg ml−1 indolicidin, which fell to threefold in a ΔgraRS mutant, suggesting the existence of an additional unidentified regulatory system. In order to test whether this residual regulation was due to BraSR, suggesting a possible co-regulation of vraDE expression by the BraSR and GraSR TCSs, we measured vraDE expression in the presence of indolicidin. We saw no induction of vraDE–lacZ expression in our genetic background using indolicidin concentrations ranging from 0.5 to 40 µg ml−1 (data not shown). Moreover, vraDE expression did not vary between the WT and ΔgraRS strains with or without bacitracin, nisin, colistin or indolicidin (Fig. 3B and data not shown). Taken together these results demonstrate that vraDE expression is clearly not controlled by the GraSR TCS under our conditions and that no cross-regulation between the Bra and Gra TCSs exists. In the previous report (Li et al., 2007) no independent validation of the transcriptome data was reported and other transcriptome experiments did not identify vraDE as a member of the GraSR regulon (Herbert et al., 2007).

Contradictory results have also been reported regarding sensitivity of the S. aureusΔvraDE mutant towards antibiotics. Indeed, VraDE was first reported as being involved in resistance to a large number of antibiotics and/or antimicrobial peptides (daptomycin, bacitracin, vancomycin, hBD3, LL37, nisin, pep5) (Sass et al., 2008) and then as only playing a role in bacitracin resistance, using a similar selection of antibiotics as the previous study (Pietiäinen et al., 2009). In our genetic background, we found that, among the panel of tested antibiotics (vancomycin, fosfomycin, oxacillin, colistin, capreomycin, viomycin, daptomycin, indolicidin), the VraDE ABC transporter is only involved in resistance to bacitracin and nisin, two cyclic polypeptides that target lipid II precursor or lipid II (Siewert and Strominger, 1967; Breukink and de Kruijff, 2006), which could explain that this system has evolved as a common defence system against these two compounds in S. aureus. We cannot, however, exclude that other lantibiotics besides nisin that target lipid II may also act as inducers for the BraSR TCS. Indeed, both actagardine and mersacidin were recently shown to be inducers of the B. subtilis BceSR TCS whereas nisin does not, although the resistance conferred by the system to the two former compounds is much lower than to bacitracin (Staron et al., 2011).

Induction of vraDE expression by very high vancomycin concentrations (10- and 5-fold MIC) has also been reported in two transcriptomic studies (Kuroda et al., 2003; Pietiäinen et al., 2009). We performed β-galactosidase assays with sublethal vancomycin concentrations ranging from 0.25 to 1 µg ml−1 (MIC = 2 µg ml−1) and saw no effect on vraDE expression (data not shown). These results indicate that if vraDE induction by vancomycin does occur, it is only at very high concentrations and independent of the BraRS/BraDE regulatory pathway, which is highly sensitive and responds to low concentrations of bacitracin or nisin (Fig. 2).

The BraSR TCS has attracted a considerable amount of attention recently. Indeed, during the past several months while different versions of this manuscript were under review, several independent reports appeared in the literature, studying various aspects of the system. In agreement with the results reported here, separate studies reported BraSR as linked to bacitracin resistance in S. aureus strain MW2 (Matsuo et al., 2010; Yoshida et al., 2011), where the system was referred to as BceSR (Yoshida et al., 2011), while missense mutations within braS were described as leading to increased nisin resistance in the SH1000 genetic background, and the TCS was designated NsaS/NsaR (Nisin susceptibility associated) (Blake et al., 2011).

More recently, a transcriptome analysis was performed in the SH1000 background using a strain where the braRS operon promoter drives expression of the intact braR gene and that of a truncated copy of braS, while a truncated copy of braR and an intact copy of braS are still present downstream from the plasmid insertion, albeit separated from the operon promoter (Kolar et al., 2011). The transcriptome analysis was performed by comparing expression levels of genes in this mutant, where expression of braS is thought to be reduced, with those in the parental SH1000 strain (Kolar et al., 2011). However, the analysis was carried out using cells grown in the absence of any inducing antibiotic, making it difficult to judge the physiological relevance of the results. Indeed, TCSs are known to be essentially inactive in the absence of the signal detected by the kinase and as we show here, expression of genes controlled by BraSR such as the vraDE operon is approximately 500-fold lower in the absence of bacitracin, remaining at background levels. In fact, in their analysis, Kolar et al. failed to identify vraDE as being controlled by BraSR (Kolar et al., 2011), although it is the most highly regulated member of the regulon as we show here. Furthermore, Kolar et al. saw no effect of their braS mutation on sensitivity to nisin or bacitracin, in contrast to our results and those of Yoshida et al. (Kolar et al., 2011; Yoshida et al., 2011) and they report that expression of braRS is induced by nisin, whereas we show here that expression of braRS is not induced by bacitracin or nisin, and that BraSR do not autoregulate their own synthesis. Part of the problem may come from the fact that their study was carried out in the S. aureus SH1000 strain, derived from strain 8325-4, both of which are known to contain multiple mutations in comparison with the reference NCTC 8325 strain (O'Neill, 2010), whereas we used the HG001 strain, a rsbU-repaired variant of strain NCTC 8325, which has never been mutagenized or subjected to UV irradiation (Herbert et al., 2010).

The appearance of these recent reports raises the issue of nomenclature for this TCS, as it has been referred to variously as NsaS/NsaR (Blake et al., 2011) or BceS/BceR (Yoshida et al., 2011), and in our study as BraS/BraR. In our view, the Nsa appellation is inappropriate since we show here that nisin does not induce expression of the BraSR-dependent braDE operon, and only induces expression of vraDE approximately sevenfold, whereas bacitracin induces both target operons significantly (over 500-fold for vraDE). Additionally, the system was originally identified by ourselves and Matsuo et al. as linked to bacitracin resistance (Matsuo et al., 2010). The BceSR designation is equally inappropriate since BraSR is not the orthologue of the B. subtilis BceSR system (see above); BceSR-dependent gene expression in B. subtilis is not induced by nisin, nor does the Bce system confer resistance to nisin (Staron et al., 2011); and the BraSR system differs significantly in its mechanistics from the BceSR system in that it uses two distinct ABC transporters, one for sensing (BraDE) and one for detoxification (VraDE) as we show here. For all of these reasons, we believe that the Nsa or Bce nomenclatures would be misleading and instead propose that the system be designated BraSR (Bacitracin resistance associated), following the previously established S. aureus nomenclature for related systems (Gra, Vra).

The main part of this study was devoted to the characterization of the BraSR/BraDE/VraDE system. Following bacitracin detection, we have shown that BraSR strongly activates transcription of the braDE and vraDE operons (Fig. 3). BraR is a member of the OmpR subfamily of RRs, with a typical winged helix–turn–helix domain (Martinez-Hackert and Stock, 1997; Mizuno and Tanaka, 1997) extending from amino acid residues 172 to 202. Although OmpR-type RRs are known to bind to short direct repeats with a spacing placing them on the same face of the DNA helix (Blanco et al., 2002; Dubrac and Msadek, 2004), several members of this family such as BceR of B. subtilis or Streptococcus mutans were in fact shown to bind to inverted repeat sequences (Ohki et al., 2003; Ouyang et al., 2010).

In this study we have identified a conserved imperfect palindrome (CTTTCAA NN T/CTGTAAG) in the promoter regions of the braDE and vraDE operons and through a genetic approach, combining deletions and point mutations within this sequence, shown that it is essential for bacitracin-dependent induction and activation by BraSR (Fig. 4). Although the sequence of this operator site is strikingly similar to that reported for BceR of S. mutans, strongly suggesting it is indeed the BraR binding site, it shares no similarities with the binding site for BceR of B. subtilis (Ohki et al., 2003; Ouyang et al., 2010). The position of the likely BraR binding site, approximately 60 bp upstream from the transcription initiation sites of the braDE and vraDE operons (Fig. 4A), suggests that BraR is probably a Class I transcription activator, interacting with the α subunit of RNA polymerase (Ishihama, 1993).

A central finding of this study is the involvement of two separate ABC transporters, BraDE and VraDE, playing distinct roles in antibiotic resistance: the first being dedicated to signal perception and the second to detoxification. Indeed, complementation experiments with constitutively expressed braDE and vraDE genes show that VraDE is sufficient to confer resistance to bacitracin and nisin when expressed on its own, whereas BraDE is not. Conversely, BraDE is required for expression of the braDE and vraDE operons, whereas VraDE is not, indicating that the VraDE ABC transporter is only involved in the detoxification process. Complementation of the ΔbraDE mutant by overproducing BraR confirmed that this ABC transporter is only required for activation of the RR through phosphorylation and has no role in the resistance process itself. Our results strongly suggest these two distinct functional roles are both energy-dependent, since the BraD and VraD NBD proteins with intact ATP hydrolysis domains are each essential for bacitracin resistance (Table 3). Although BraE and VraE are similar in their predicted membrane topology (Fig. 6A) the proteins share only 33% amino acid identity overall. We demonstrated that the extracellular loop of VraE is the specificity determinant for bacitracin recognition, but sequence alignment with BraE did not reveal any particularly similar or dissimilar regions (Fig. 6B) in contrast to VraG and VraE, making it difficult to predict a common bacitracin recognition domain between BraE and VraE. The functional specialization of BraDE and VraDE in the S. aureus bacitracin detoxification process is quite unexpected since in B. subtilis the BceAB transporter is believed to carry out both roles, in sensing and transport (Bernard et al., 2007; Rietkötter et al., 2008). Although a role for BceAB in bacitracin sensing has been shown, there is no evidence that it is directly involved in the detoxification process and the bacitracin sensitivity of a B. subtilis bceSR mutant constitutively expressing bceAB was not tested. However, we were unable to identify any VraDE equivalent in B. subtilis, suggesting that the involvement of two distinct ABC transporters with separate functions in bacitracin resistance might be unique to Staphylococcus species.

Our results allowed us to propose a model for antibiotic sensing and resistance through this pathway in S. aureus (Fig. 7), where the antibiotic is first detected by the BraDE ABC transporter, as we have shown that a braDE mutant strain no longer responds to the presence of extracellular bacitracin or nisin (Fig. 5). This is fairly unusual since typically in TCSs the signal is sensed by the external loop of the HK (Hoch and Silhavy, 1995). However the BraS HK, like BceS of B. subtilis, belongs to the IMSK family. Because of their short extracellular loop (only three amino acids in BraS), it has been suggested that these kinases detect their stimuli from within or at the membrane interface (Mascher, 2006; Mascher et al., 2006). Our results support the hypothesis that the stimulus is sensed by the BraDE ABC transporter and then transferred to BraS, which in turn activates the BraR RR, as suggested for the BceSR/BceAB module in B. subtilis and S. mutans (Bernard et al., 2007; Ouyang et al., 2010).

In this model, several important questions remain to be answered. First, how does the BraDE ABC transporter sense bacitracin? The membrane topology of the BraE permease (Figs 6A and 7) is similar to that of BceB in B. subtilis, characterized by the presence of a large extracellular loop, which has been suggested to be essential for bacitracin detection by BceB (Rietkötter et al., 2008; Coumes-Florens et al., 2011). It is likely that in BraE, this extracellular loop is also essential for signal recognition and could be involved in direct interaction with the antibiotic. Indeed, we show here that replacing the extracellular loop of BraE with that of VraE prevented bacitracin sensing by the chimeric BraDE*VraE ABC transporter (Table 4). The second question is how do BraDE and BraSR interact? One possibility is that antibiotic binding and/or transport by BraDE leads to a conformational change of the permease structure, promoting interaction with BraS and subsequent phosphorylation of BraR.

Another unanswered question is how the VraDE ABC transporter confers bacitracin resistance. The direction of bacitracin transport by VraDE remains unknown. This is also the case for the bacitracin transporter BceAB of B. subtilis. Some studies favour the hypothesis of an export system based in particular on the lack of a substrate-binding protein (Bernard et al., 2007) while others suggest an import function due to the general architecture of BceAB (Rietkötter et al., 2008). Based on our results and the importance of the large extracytoplasmic domain that appears to be required as a substrate-binding protein, we favour the hypothesis that VraDE acts as an importer. Bacitracin could be bound by the extracellular loop of the permease and then transferred to the cytoplasm, away from its site of action. Further work will be required to unravel the mechanisms involved in detection and activation of the BraSR/BraDE/VraDE resistance module to confirm these hypotheses.

Experimental procedures

Bacterial strains and growth procedures

Escherichia coli K12 strain DH5α™[F- (φ80dlacZΔM15) Δ(lacZYA-argF) U169 recA1 endA1 hsdR17 (rk−, mk+) phoA supE44 λ-thi-1 gyrA96 relA1] (Invitrogen) was used for cloning experiments. S. aureus strain HG001, a rsbU+ derivative of strain NCTC 8325 (Herbert et al., 2010) and its derivatives were used in all experiments. S. aureus strains and plasmids used in this study are listed in Table 5. Plasmids were constructed in E. coli and passaged through the restriction modification deficient S. aureus strain RN4220 (Kreiswirth et al., 1983) before transformation into strain HG001 and its derivatives. E. coli strains were grown in LB medium and ampicillin (100 µg ml−1) was added when required. S. aureus was grown in Trypticase Soy Broth (TSB; Difco) with shaking (200 r.p.m.) at 37°C, with chloramphenicol (10 µg ml−1) or erythromycin (2 µg ml−1) added for plasmid selection when required. E. coli and S. aureus strains were transformed by electroporation using standard protocols (Sambrook et al., 1989) and transformants were selected on LB or Trypticase Soy Agar (TSA; Difco) plates, respectively, with the appropriate antibiotics.

Table 5.  Strains and plasmids used in this study.
Strain or plasmidRelevant genotype or descriptionSourcea
  • a. 

    Arrows indicate plasmid introduction by electroporation.

HG001NCTC 8325 rsbU+Herbert et al. (2010)
ST1036ΔgraRSJoanne et al. (2009)
ST1068ΔvraFG::spc, SprpMADvraFG→HG001
ST1087ΔbraRS pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→ST1179
ST1088ΔbraRS pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ST1179
ST1184ΔbraRS pMK4-Pprot braRS, CmrpMK4-Pprot braRS→ST1179
ST1121ΔbraRS pMK4-Pprot braR, CmrpMK4-Pprot braR→ST1179
ST1185ΔbraRS pMK4-Pprot braDE, CmrpMK4-Pprot braDE→ST1179
ST1186ΔbraRS pMK4-Pprot vraDE, CmrpMK4-Pprot vraDE→ST1179
ST1187ΔbraDE pMK4-Pprot braDE, CmrpMK4-Pprot braDE→ST1097
ST1188ΔbraDE pMK4-Pprot vraDE, CmrpMK4-Pprot vraDE→ST1097
ST1189ΔbraDE pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→ST1097
ST1190ΔbraDE pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ST1097
ST1235ΔbraDE pMK4-Pprot braDE*vraE, CmrpMK4-Pprot braDE*vraE→ST1097
ST1191ΔvraDE pMK4-Pprot vraDE, CmrpMK4-Pprot vraDE→ST1090
ST1192ΔvraDE pMK4-Pprot braDE, CmrpMK4-Pprot braDE→ST1090
ST1193ΔvraDE pMK4-Pprot vraFG, CmrpMK4-Pprot vraFG→ST1090
ST1194ΔvraDE pMK4-Pprot vraFG*vraE, CmrpMK4-Pprot vraFG*vraE→ST1090
ST1234ΔvraDE pMK4-Pprot braDE*vraE, CmrpMK4-Pprot braDE*vraE→ST1090
ST1195ΔvraDE pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→ST1090
ST1196ΔvraDE pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ST1090
ST1197ΔbraD pMK4-Pprot braD, CmrpMK4-Pprot braD→ST1180
ST1231ΔvraD pMK4-Pprot, CmrpMK4-Pprot→ST1181
ST1198ΔvraD pMK4-Pprot vraD, CmrpMK4-Pprot vraD→ST1181
ST1205ΔvraD pMK4-Pprot vraD*WB, CmrpMK4-Pprot vraD*WB→ST1181
ST1210ΔbraD pMK4-Pprot, CmrpMK4-Pprot→ST1180
ST1211ΔbraD pMK4-Pprot vraD, CmrpMK4-Pprot vraD→ST1180
ST1206ΔbraD pMK4-Pprot braD*WB, CmrpMK4-Pprot braD*WB→ ST1180
ST1230ΔbraD pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ ST1180
ST1217ΔvraDE pMK4-Pprot, CmrpMK4-Pprot→ST1090
ST1233ΔvraDE pMK4-Pprot braR, CmrpMK4-Pprot braR→ST1090
ST1216ΔbraDE pMK4-Pprot, CmrpMK4-Pprot→ST1097
ST1232ΔbraDE pMK4-Pprot braR, CmrpMK4-Pprot braR→ST1097
ST1240ΔbraDEΔvraDE pMK4-Pprot braDE, CmrpMK4-Pprot braDE→ST1182
ST1218ΔbraDEΔvraDE pMK4-Pprot vraDE, CmrpMK4-Pprot vraDE→ST1182
ST1238ΔbraRS pMK4-Pprot, CmrpMK4-Pprot→ST1179
ST1239ΔbraDEΔvraDE pMK4-Pprot, CmrpMK4-Pprot→ST1182
ST1244pSA14-PvraSR–lacZ, CmrpSA14-PvraSR–lacZ→HG001
ST1245pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→HG001
ST1246pSA14-PbraRS–lacZ, CmrpSA14-PbraRS–lacZ→HG001
ST1247ΔbraRS pSA14-PbraRS–lacZ, CmrpSA14-PbraRS–lacZ→ST1179
ST1208pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→HG001
ST1221pSA14-PvraDE–lacZΔB, CmrpSA14-PvraDE–lacZΔB→HG001
ST1222pSA14-PvraDE–lacZΔC, CmrpSA14-PvraDE–lacZΔC→HG001
ST1223pSA14-PvraDE–lacZΔD, CmrpSA14-PvraDE–lacZΔD→HG001
ST1224pSA14-PvraDE–lacZΔE, CmrpSA14-PvraDE–lacZΔE→HG001
ST1241pSA14-PbraDE–lacZΔB, CmrpSA14-PbraDE–lacZΔB→HG001
ST1242pSA14-PbraDE–lacZΔC, CmrpSA14-PbraDE–lacZΔC→HG001
ST1243pSA14-PbraDE–lacZΔD, CmrpSA14-PbraDE–lacZΔD→HG001
ST1092ΔvraSR pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→ST1163
ST1091ΔgraRS pSA14-PbraDE–lacZ, CmrpSA14-PbraDE–lacZ→ST1036
ST1086ΔvraSR pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ST1163
ST1089ΔgraRS pSA14-PvraDE–lacZ, CmrpSA14-PvraDE–lacZ→ST1036
ST1120HG001 pMK4-Pprot, CmrpMK4-Pprot→HG001
ST1150ΔvraFG::spc pMK4-Pprot, Spr, CmrpMK4-Pprot→ST1068
ST1151ΔvraFG::spc pMK4-PprotvraFG, Spr, CmrpMK4-PprotvraFG→ST1068
ST1177ΔvraFG::spc pMK4-Pprot vraFG*vraE, Spr, CmrpMK4-Pprot vraFG*vraE→ST1068
pSA14pMK4 derivative carrying promoterless E. coli lacZ for constructing transcriptional fusions, CmrJoanne et al. (2009)
pMK4-PprotpMK4 derivative carrying a constitutive Gram-positive promoter for gene complementation, CmrArchambaud et al. (2005)
pMADpE194 derivative with a thermosensitive origin of replication for deletion/replacement of genes in Gram-positive bacteria, ErrArnaud et al. (2004)
pMAD braRSpMAD derivative allowing deletion of the braRS genes, ErrThis study
pMAD braDEpMAD derivative allowing deletion of the braDE genes, ErrThis study
pMAD braDpMAD derivative allowing deletion of the braD gene, ErrThis study
pMAD graRSpMAD derivative allowing deletion of the graRS genes, ErrThis study
pMAD vraFGpMAD derivative allowing deletion of the vraFG genes, Err, SprThis study
pMAD vraSRpMAD derivative allowing deletion of the vraSR genes, ErrToledo-Arana et al. (2005)
pMAD vraDEpMAD derivative allowing deletion of the vraDE genes, ErrThis study
pMAD vraDpMAD derivative allowing deletion of the vraD gene, ErrThis study
pMAD SA2419pMAD derivative allowing deletion of the SA2419 gene, ErrThis study
pSA14-PbraRS–lacZpSA14 derivative carrying the promoter region of braRS, CmrThis study
pSA14-PvraSR–lacZpSA14 derivative carrying the promoter region of vraSR, CmrThis study
pSA14-PbraDE–lacZpSA14 derivative carrying the promoter region of braDE, CmrThis study
pSA14-PbraDE–lacZΔApSA14 derivative carrying a truncated promoter region of braDE, CmrThis study
pSA14-PbraDE–lacZΔBpSA14 derivative carrying a truncated promoter region of braDE, CmrThis study
pSA14-PbraDE–lacZΔCpSA14 derivative carrying a truncated promoter region of braDE, CmrThis study
pSA14-PvraDE–lacZpSA14 derivative carrying a truncated promoter region of vraDE, CmrThis study
pSA14-PvraDE–lacZΔApSA14 derivative carrying a truncated promoter region of vraDE, CmrThis study
pSA14-PvraDE–lacZΔBpSA14 derivative carrying a truncated promoter region of vraDE, CmrThis study
pSA14-PvraDE–lacZΔCpSA14 derivative carrying a truncated promoter region of vraDE, CmrThis study
pMK4-Pprot braRSpMK4-Pprot derivative carrying the braRS genes, CmrThis study
pMK4-Pprot braDEpMK4-Pprot derivative carrying the braDE genes, CmrThis study
pMK4-Pprot braDpMK4-Pprot derivative carrying the braD gene, CmrThis study
pMK4-Pprot vraDEpMK4-Pprot derivative carrying the vraDE genes, CmrThis study
pMK4-Pprot vraDpMK4-Pprot derivative carrying the vraD gene, CmrThis study
pMK4-Pprot vraFGpMK4-Pprot derivative carrying the vraFG genes, CmrThis study
pMK4-Pprot vraFG*vraEpMK4-Pprot derivative carrying the vraFG*vraE genes, where the extracellular loop-coding sequence of vraG was replaced by that of vraE, CmrThis study
pMK4-Pprot braDE*vraEpMK4-Pprot derivative carrying the braDE*vraE genes, where the extracellular loop-coding sequence of braE was replaced by that of vraE, CmrThis study
pMK4-Pprot braD*WBpMK4-Pprot derivative carrying the braD gene containing a single-point mutation in the Walker B motif, CmrThis study
pMK4-Pprot vraD*WBpMK4-Pprot derivative carrying the vraD gene containing a single-point mutation in the Walker B motif, CmrThis study

MIC determinations

Minimal inhibitory concentration determinations were performed in a 96-well microtitre plate (100 µl culture volume). A bacterial culture grown in TSB to an OD600 = 1 was used to inoculate wells containing TSB with standard twofold increments of antibiotic concentration. Plates were incubated for 10 h with vigorous shaking at 37°C in a Synergy 2 thermoregulated spectrophotometer plate reader using the Gen5TM Microplate Software (BioTek Instruments, Winooski, VT). All experiments were performed in triplicate. The MIC was defined as the concentration at which growth was prevented.

DNA manipulations

Oligonucleotides used in this study were synthesized by Sigma-Proligo and their sequences are listed in Table 6. S. aureus chromosomal DNA was isolated using the MasterPureTM Gram-positive DNA purification Kit (Epicentre Biotechnologies). Plasmid DNA was isolated using a QIAprep Spin Miniprep kit (Qiagen) and PCR fragments were purified using the Qiaquick PCR purification kit (Qiagen). T4 DNA ligase and restriction enzymes (New England Biolabs), PCR reagents and High-Fidelity Phusion thermostable DNA Polymerase (Roche) were used according to the manufacturer's recommendations. Nucleotide sequencing of plasmid constructs was carried out by Genome Express-Cogenics or MilleGen.

Table 6.  Oligonucleotides used in this study.a
  • a. 

    Added restriction site sequences are shown in italics.

TCS17AbraRS upstream region, deletion mutant5′-GGATCCGCTGCAGAATCAGTAATATT-3′
TCS17CbraRS downstream region, deletion mutant5′-AAGCTTAAACTTTCAATATTGTAAGTATA-3′
OAH071braDE upstream region, deletion mutant5′-CGCGCGGGATCCCTATAAGCCAACAACAACAAATTG −3′
OAH073braDE downstream region, deletion mutant5′-TGCATGCCATGTTGCCGCAATTAACCATGTTTCAGTG-3′
OMD189vraFG upstream region, deletion mutant5′-GGAGGATCCTCGATGCTGGATACACAGCTG-3′
OMD191vraFG downstream region, deletion mutant5′-AGAAGATCTGCTGTTTTTGCAGTGACGGC-3′
OAH075vraDE upstream region, deletion mutant5′-GGGCCCGGATCCGACACCTACGAATACAAATGC-3′
OAH077vraDE downstream region, deletion mutant5′-AGACTATGACGATATCCATTTAACACTTACGATTAAAAG-3′
OAH079braDE promoter region lacZ fusions5′-CCCGTGCTGCAGGGCCAGCGCCAAAGTAACTC-3′
OAH081vraDE promoter region lacZ fusions5′-CCCGCGCTGCAGCCATTTAATACCTTGATTTCC-3′
OAH100braRS-coding sequence, complementation5′-GATAATGGATCCGAAGGAAGAAGATTATAGATG-3′
OAH105braDE-coding sequence, complementation5′-AGAACTGGATCCGAAATGAGGTGCATGCCATGTTGC-3′
OAH103vraDE-coding sequence, complementation5′-TGAAATGGATCCCAAAAGGAGTGAGACTATGACG-3′
MF150vraFG-coding sequence, complementation5′-GGAGGATCCGATAAATTATAGGAGTGTTAAAGTG-3′
OAH137vraFG*vraE-coding sequence, complementation5′-AGGGGTCTCCATTTACTTATAGCAGCAAAGCAAAGAAC-3′
OAH197braD Walker B motif-coding sequence site-directed mutagenesis5′-GTTTTTGAATCAAGTGCACCTGTAGGTTGATCGGCTAG-3′
OAH199vraD Walker B motif-coding sequence site-directed mutagenesis5′-TGCACTTTTCGAGTCGAGTGCGCCTGTTGGTTGATCTGC-3′
OAH213braDE*vraE-coding sequence, complementation5′-TCTGGTCTCCATTTGGTTGTTAATGTCATTGTGAGCG-3′
OAH123braDE intragenic region, qRT-PCR5′-CGGCGGATACTTACTTGG-3′
OAH125vraDE intragenic region, qRT-PCR5′-GCACTAACTATGGCTATGACATC-3′
OAH129braRS promoter region, primer extension5′-AACCATCCTGTTTCAGTCGTATC-3′
OAH127vraDE promoter region, primer extension5′-CCCTTTTCTCAATTTACACAAAC-3′
OMD156Tn554 spc spectinomycin resistance gene amplification, gene deletion5′-AGAAGATCTCACCTAGATCCTTTTGACTC-3′

Plasmid and mutant construction

Plasmid pSA14, a derivative of the shuttle vector pMK4 (Sullivan et al., 1984), which carries a promoterless E. coli lacZ gene, was used to construct transcriptional lacZ reporter fusions (Joanne et al., 2009). Plasmid pMK4-Pprot, a derivative of vector pMK4 carrying a constitutively expressed Gram-positive promoter sequence (Archambaud et al., 2005), was used for gene complementation experiments. The thermosensitive shuttle vector pMAD was used for introducing markerless gene deletions (Arnaud et al., 2004). Mutant strains of S. aureus HG001 used in this study were obtained by gene deletions, removing the entire coding sequence without the introduction of an antibiotic resistance gene. In a first step, a recombination cassette, consisting of two fused DNA fragments, of approximately 500 bp, corresponding to the chromosomal DNA regions located directly upstream and downstream from the gene of interest was generated by PCR using the gene splicing by strand overlap extension PCR technique (SOE-PCR) (Horton et al., 1989) as previously described for several of the S. aureus TCS genes (Toledo-Arana et al., 2005). Oligonucleotides used for PCR amplifications are listed in Table 6. Amplicons were cloned into the temperature-sensitive shuttle vector pMAD between the BamHI/NcoI, BamHI/PstI or BamHI/EcoRI restriction sites. For the ΔvraFG mutant, a 1021 bp DNA fragment corresponding to the Tn554 spc spectinomycin resistance gene (Murphy, 1985) was cloned between the upstream and downstream amplicons, yielding pMADvraFG (see Tables 5 and 6). Nucleotide sequences of the constructs were confirmed by DNA sequencing. Following introduction into S. aureus HG001, integration and excision of pMAD derivatives with deletion of the chromosomal region of interest were carried out as previously described (Arnaud et al., 2004). Gene deletions in mutant strains were verified by PCR.

For complementation experiments, DNA fragments corresponding to the gene-coding sequences were amplified with oligonucleotides generating either BamHI/PstI or BamHI/SalI restriction sites at the extremities as listed in Table 6 and cloned in the pMK4-Pprot replicative plasmid (Archambaud et al., 2005). To construct the chimeric system VraG MSD protein with the extracellular loop of VraE (VraG*vraE), the predicted membrane topologies of VraG and VraE were determined using the TopPredII program and von Heijne's algorithm (Claros and von Heijne, 1994). The coding sequence of the vraE extracytoplasmic loop was amplified by PCR using oligonucleotides OAH138 and OAH139, carrying BsaI restriction sites (see Table 6). BsaI is a type IIS restriction endonuclease, which cleaves after its restriction site, generating DNA fragments whose tetranucleotide cohesive ends can be defined and designed to be specifically compatible. Two PCR fragments corresponding to the DNA regions flanking the vraG extracellular loop were generated, using oligonucleotides containing BamHI/BsaI restriction sites (MF150 and OAH137; upstream fragment) and BsaI/SalI restriction sites (OAH140 and OAH141; downstream fragment; see Table 6). Following digestion with BsaI, the two vraG fragments were then ligated to the vraE extracellular loop sequence, seamlessly fusing the three fragments together without adding any additional nucleotides. The resulting fragment was re-amplified by PCR using the external oligonucleotides (MF150 and OAH141), and cloned between the BamHI and SalI restriction sites of plasmid pMK4-Pprot, yielding plasmid pMK4-Pprot vraFG*vraE. A similar strategy was used to generate plasmid pMK4-Pprot braDE*vraE, where the coding sequence of the BraE permease extracellular loop was replaced with that of VraE. The sequence of the vraE extracytoplasmic loop was amplified by PCR using oligonucleotides OAH138 and OAH214 carrying BsaI restriction sites (see Table 6) and DNA fragments corresponding to the braE upstream and downstream regions were amplified using oligonucleotide pairs OAH105/OAH213 and OAH215/OAH106 respectively. Single-point mutations in the braD and vraD Walker B motif-coding sequences were introduced by site-directed mutagenesis through SOE-PCR (Ho et al., 1989). Briefly, two DNA fragments were generated by PCR. The first contained the upstream region and the single-point mutation obtained using a reverse oligonucleotide containing 1 mismatch (Table 6, OAH105/OAH197 and OAH103/OAH199 for braD and vraD respectively). The second fragment, corresponding to the region downstream of the mutation site (Table 6, oligonucleotides OAH198/OAH173 and OAH200/OAH196 for braD and vraD respectively), had a 30 bp overlap with the first one, allowing the two to be joined by gene splicing through SOE-PCR, and the resulting DNA fragment was cloned in plasmid pMK4-Pprot, yielding plasmids pMK4-Pprot braD*WB and pMK4-Pprot vraD*WB.

For constructing transcriptional lacZ fusions, promoter regions of the braDE, SA2419-braRS, vraDE and vraSR operons were amplified by PCR using oligonucleotides introducing BamHI/PstI restriction sites (see Table 6). The corresponding DNA fragments were then cloned between the corresponding restriction sites of the pSA14 vector, yielding plasmids PbraDE–lacZ (PSA2415-2416–lacZ), PSA2419-braRS–lacZ (PSA2417-2419–lacZ), PvraDE–lacZ and PvraSR–lacZ. To construct the promoter regions with the three point mutations in the BraR operator site, PbraDE–lacZΔE and PvraDE–lacZΔE, the corresponding DNA fragments were amplified by PCR using forward oligonucleotides containing three mismatches (oligonucleotides OAH209 and OAH210, for braDE and vraDE respectively; Table 6).

β-Galactosidase assays

For β-galactosidase assays, cells were grown in TSB until OD600 = 2. Bacitracin was added to the medium at different sublethal concentrations when required. S. aureus carrying lacZ fusions were harvested by centrifuging 2 ml of culture samples (2 min; 20 800 g). Cells were resuspended in 500 µl of Z buffer (Miller, 1972) with 0.5 mg ml−1 DNase and 0.1 mg ml−1 lysostaphin added extemporaneously, and lysed by incubation at 37°C for 30 min. Cell debris were eliminated by centrifugation (2 min; 20 800 g) and the supernatant was either used directly for assays or stored at −20°C. Assays were performed as previously described and β-galactosidase-specific activities were expressed as Miller units per mg protein (Miller, 1972). Protein concentrations were determined using the Bio-Rad protein assay (Bio-Rad, Hercules, CA) (Bradford, 1976). All experiments were carried out in triplicate.

Extraction of total RNA

Bacteria were grown at 37°C with aeration to an OD600 = 2 in TSB with bacitracin (0.5 µg ml−1) when required. Cells were pelleted (2 min; 5400 g) and immediately frozen at −20°C. RNA extractions were then performed as previously described (Even et al., 2006), followed by a DNase I treatment with the TURBO DNA-free reagent (Ambion, Austin, TX) in order to eliminate residual contaminating genomic DNA.

Primer extensions

Primer extensions were performed as previously described (Chastanet et al., 2001) using 40 µg of RNA, 2 pmol of oligonucleotide (previously radiolabelled with [γ-32P]-ATP using T4 polynucleotide kinase, New England Biolabs) and 200 U of Superscript II reverse transcriptase (Invitrogen). Oligonucleotides were chosen so as to hybridize approximately 30–70 bp downstream from the translation initiation codon (see Table 6). The corresponding DNA sequencing reactions were carried out with the same oligonucleotides and PCR-amplified DNA fragments carrying the respective promoter regions, using the Sequenase PCR product sequencing kit (USB, Cleveland, OH).

cDNA synthesis and qRT-PCRs

cDNAs were obtained using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) according to the manufacturer's recommendations, in a final reaction volume of 20 µl containing 1 µg of total RNA. The mixture was incubated for 5 min at 25°C, 30 min at 42°C and reverse transcriptase was then inactivated by heating the mixture at 85°C for 5 min. For qRT-PCR experiments, primers were designed using BEACON Designer 4.02 software (Premier Biosoft International, Palo Alto, CA) to amplify 200–300 bp amplicons (see Table 6). qRT-PCRs, critical threshold cycles (CT) and n-fold changes in transcript levels were performed and determined as previously described and normalized with respect to 16S rRNA whose levels did not vary under our experimental conditions (Dubrac et al., 2007).


This work was supported by research funds from the European Commission [StaphDynamics (LHSM-CT-2006-019064) and BaSysBio (LSHG-CT-2006-037469) grants], the Centre National de la Recherche Scientifique (CNRS URA 2172), Agence Nationale de la Recherche (ANR GrabIron and NaBab) and the Institut Pasteur (PTR N°256 and PTR N°336). We are grateful to Cécile Wandersman for critical reading of the manuscript. We thank Olivier Poupel for assistance with qRT-PCR experiments and Iñigo Lasa Uzcudun (Universidad Pública de Navarra-CSIC, Pamplona, Spain) in whose laboratory part of this work was carried out. Mélanie Falord received a Young Scientist Fellowship from the Conseil Pasteur-Weizmann.