Investigation of the Staphylococcus aureus GraSR Regulon Reveals Novel Links to Virulence, Stress Response and Cell Wall Signal Transduction Pathways

The GraS/GraR two-component system has been shown to control cationic antimicrobial peptide (CAMP) resistance in the major human pathogen Staphylococcus aureus. We demonstrated that graX, also involved in CAMP resistance and cotranscribed with graRS, encodes a regulatory cofactor of the GraSR signaling pathway, effectively constituting a three-component system. We identified a highly conserved ten base pair palindromic sequence (5′ ACAAA TTTGT 3′) located upstream from GraR-regulated genes (mprF and the dlt and vraFG operons), which we show to be essential for transcriptional regulation by GraR and induction in response to CAMPs, suggesting it is the likely GraR binding site. Genome-based predictions and transcriptome analysis revealed several novel GraR target genes. We also found that the GraSR TCS is required for growth of S. aureus at high temperatures and resistance to oxidative stress. The GraSR system has previously been shown to play a role in S. aureus pathogenesis and we have uncovered previously unsuspected links with the AgrCA peptide quorum-sensing system controlling virulence gene expression. We also show that the GraSR TCS controls stress reponse and cell wall metabolism signal transduction pathways, sharing an extensive overlap with the WalKR regulon. This is the first report showing a role for the GraSR TCS in high temperature and oxidative stress survival and linking this system to stress response, cell wall and pathogenesis control pathways.


Introduction
The opportunistic pathogen Staphylococcus aureus is both a commensal and a major Gram-positive pathogen, causing a variety of infections ranging from superficial skin abscesses to more serious diseases such as pneumonia, meningitis, endocarditis, septicemia and toxic shock syndrome [1]. The ubiquitous nature of this pathogen stems mostly from its capacity to survive a large variety of environmental conditions as well as an impressive ability to resist host innate immune defense mechanisms and produce systemic infections, often in healthy humans [2,3]. This unique adaptive potential has made S. aureus one of the major causes of nosocomial infections today, compounded by the rapid emergence of multiple antibiotic-resistant strains over the past few decades [4], particularly methicillin and vancomycin-intermediate resistant strains (MRSA and VISA). Until recently, vancomycin had remained the weapon of last resort, but the recent appearance of the enterococcal vanA vancomycin-resistance gene cluster in S. aureus highlights the growing threat this bacterium poses to human health and the urgent need for developing novel therapeutic approaches [5,6].
Cationic antimicrobial peptides (CAMPs) are an important component of host innate immunity and understanding the molecular mechanisms involved in resistance is a key factor in staphylococcal treatment research. CAMPs have both cationic and amphipathic properties and are classified according to their length and secondary structure [7]. They are produced by certain immune, skin and epithelial cells in all living kingdoms, as defenses against microbial proliferation, and many are known to act by forming pores in the cell membrane, through interactions with bacterial cell envelope components [8]. However, recent work has shown that several CAMPs, including indolicidin and colistin, can also kill by inhibiting intracellular processes such as protein and DNA synthesis as well as septum formation and division [9].
To counteract CAMP antimicrobial activity during infection, Gram-positive bacteria have developed several resistance mechanisms, including degradation, sequestration or electrostatic repulsion [10]. In Bacillus subtilis and related Gram-positive bacteria, Dalanylation of wall teichoic acids (WTAs) and lipoteichoic acids (LTAs), mediated by the DltABCD enzymes, as well as MprFdependent lysylination of phosphatidylglycerol, prevent CAMPbinding by increasing the bacterial surface positive charge [10,11].
Two-component systems (TCSs) play an important role in these mechanisms by coordinating the expression of resistance genes, when CAMPs are detected at the cell surface. TCSs 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. Most S. aureus genomes have a sophisticated arsenal of sixteen sets of environmental monitoring TCS genes, with an additional one present in the staphylococcal cassette chromosome mec element of MRSA strains [12]. Among these systems, the well-studied AgrCA peptide quorum-sensing TCS controls the expression of several virulence genes [13] and VraSR was shown to be responsible for resistance to cell wall-damaging compounds, including b-lactam antibiotics and some CAMPs [14].
The main regulatory pathway controlling CAMP resistance in staphylococci, however, is the GraSR (Glycopeptide Resistance Associated) TCS, aka ApsSR (Antimicrobial Peptide Sensor), which has been extensively studied over the past five years [15][16][17][18]. First discovered in S. aureus as a locus whose overexpression led to increased vancomycin resistance, the GraSR TCS was shown to be required for resistance of S. aureus and S. epidermidis to several CAMPs, by controlling expression of mprF and the dlt and vraFG operons [15][16][17][18][19]. Additionally, the first gene of the graRS operon encodes GraX, a protein of unknown function that also plays a role in CAMP resistance [16][17][18][19]. Missense mutations in the graRS locus have been linked to CAMP sensitivity of certain S. aureus strains [20], and the system also plays a role in biofilm formation [21,22]. GraS was shown to play a role in survival of S. epidermidis and S. aureus within neutrophils [23] and the GraSR system has been implicated in S. aureus virulence in several experimental models [18,[24][25][26].
In this study we set out to further define the GraSR regulon in S. aureus. We identified a highly conserved palindromic sequence as the likely GraR binding site, and showed that the GraSR TCS is required for growth of S. aureus at high temperature and resistance to oxidative stress. Using a combination of genome-based predictions and transcriptome analysis, we revealed several novel GraR target genes as well as unsuspected links with the AgrCA and WalKR TCSs.

GraX, GraS and GraR are required for Staphylococcus aureus colistin resistance
In an effort to fully define the GraSR regulon of Staphylococcus aureus, and determine the roles of GraXSR in colistin resistance, we constructed mutant strains ST1036 (DgraRS) and ST1070 (DgraX) in the S. aureus HG001 background [27] by removing the entire coding sequences of the genes (DgraRS), or by an in-frame deletion (DgraX; see Materials and Methods). GraX, predicted to be a cytoplasmic protein, presents weak similarities to sugar epimerases, and has been shown to be involved in CAMP resistance along with the GraSR TCS, but its specific role remains to be established.
Minimal inhibitory concentration (MIC) values for resistance to colistin, a bacterial CAMP, were determined by following growth in TSB at 37uC over a 12 h period, using a Biotek Synergy Microplate reader, with decreasing concentrations of colistin ( Table 1). The DgraRS and DgraX mutants displayed acute sensitivity to colistin compared to the parental strain. However, the DgraX mutant appeared to be more resistant to colistin than the DgraRS mutant. We therefore analyzed graR expression in the DgraX mutant by quantitative RT-PCR (qRT-PCR), showing that graR expression is increased approximately 2-fold compared to the parental strain (data not shown). This is likely through stabilization of the graRS transcript due to increased proximity with the operon promoter in the DgraX mutant, suggesting that CAMP sensitivity of the DgraX mutant may in fact be underestimated. We also observed that the DgraRS and DgraX mutants were highly sensitive to nisin (data not shown) as previously observed [17,18].
In order to complement the DgraRS mutant, an intact copy of the graR gene was introduced on a multicopy plasmid, resulting in strain ST1116 (DgraRS pMK4-Pprot-graR). Complementation of the DgraRS mutant by constitutive expression of the graR gene fully restored colistin resistance (Table 1). Indeed, it is well known that response regulator overexpression can complement the absence of the cognate kinase, due to its phosphorylation by other phosphate donors such as acetyl phosphate or aspecific kinase activity within the cell [28,29].
GraXSR do not autoregulate their own synthesis As shown in Fig. 1A, the graXRS operon is located directly upstream from the vraFG operon, encoding an ABC transporter [19]. To define the graXRS operon promoter region, we first analyzed its expression using primer extension experiments. Total RNA was extracted from strain HG001 during mid-exponential growth in TSB at 37uC and used for primer extension experiments. We identified a unique transcriptional start site in the graXRS promoter region (Fig. 1B), and the preceding nucleotide sequence revealed appropriately spaced potential 210 and 235 regions, sharing strong similarities with the consensus sequences of promoters recognized by the vegetative form of RNA polymerase holoenzyme, Es A (Fig. 1C).
Several two-component systems are known to positively autoregulate their own synthesis, such as the VraSR cell envelope stress response and AgrCA peptide quorum-sensing virulence regulatory systems of S. aureus [13,30]. In order to test whether this was also the case for the GraSR system, a transcriptional fusion was constructed between the graXRS operon upstream region and the lacZ gene of E. coli using the pSA14 plasmid (see Materials and Methods). To study graXRS operon expression, the graX'-lacZ fusion was introduced into strains HG001, ST1036 (DgraRS) and ST1070 (DgraX), and b-galactosidase activity was measured during mid-exponential growth at 37uC in TSB after a 30 mn treatment with or without 200 mg ml 21 colistin (Fig. 2). No significant differences in graX'-lacZ expression levels were observed between the three strains, or in the presence or absence of colistin, indicating that GraXSR do not autoregulate their own synthesis, and that their cellular levels are not induced by the presence of CAMPs such as colistin.
Identification of potential GraR-binding sites in the promoters of known GraR regulated genes Although several genes involved in CAMP resistance are known to be controlled by GraSR (mprF, dlt and vraFG operons), the specific nucleotide sequence constituting the GraR operator sequence remains unknown. In order to identify potential GraRbinding sites upstream from the coding regions of these genes, we first identified their transcription initiation sites through primer extension analysis. Total RNA was extracted from strain HG001 during mid-exponential growth in TSB at 37uC after treatment with 200 mg ml 21 colistin and used for primer extension experiments. We identified a unique transcriptional start site in the mprF and vraFG promoter regions (Fig. 3A) and two initiation sites for the dltXABCD operon. The first (not shown here), is located 30 bp upstream from the dltX (SAOUHSC_00868) translation initiation codon, and was previously identified in S.  aureus SA113 [31] whereas the second (Fig. 3A), 110 bp further upstream, had not been reported. We identified appropriately spaced potential 210 and 235 regions upstream from all three transcription initiation sites (Fig. 3B). The mprF and dltXABCD 210 regions share strong similarities with the consensus sequence recognized by the vegetative form of RNA polymerase holoenzyme, Es A . However, all three 235 regions as well as the 210 region of the vraFG operon showed only weak similarities with RNA polymerase Es A consensus promoter sequences, consistent with the existence of a positive transcriptional regulator [32].
GraR is a member of the OmpR subfamily of response regulators, with a typical winged helix-turn-helix domain [33] extending from residues 173 to 203. Although response regulators belonging to the OmpR family are known to bind to short direct repeats [34], orthologs of GraR such as VirR of Listeria monocytogenes and BceR of Bacillus subtilis were in fact shown to bind to inverted repeat sequences [35,36]. We failed to identify any significant direct repeat sequences in the upstream regions of mprF and the vraFG and dlt operons. However, a global study aimed at identifying response regulator binding sites in low G+C Gram-positive bacteria [37] reported the presence of an imperfect palindromic sequence (59 AAGTGACA-N4-TGTCATTT 39) within the end of the graS coding region, upstream from the vraFG operon which is known to be controlled by the GraSR system [16,17,18,19]. We were able to identify this palindromic sequence as also being present upstream from the 235 sequences of mprF and the upstream dlt operon promoter. The three inverted repeats are highly conserved, allowing us to align them in order to produce a potential GraR-binding site consensus sequence (Fig. 3B). In agreement with our results showing that GraXSR do not autoregulate their own synthesis, the potential GraR operator sequence is not present in the graXRS operon upstream promoter region.

GraSR-dependent gene expression requires GraX, CAMPs and the consensus binding site
To determine the roles of GraX and this potential GraRbinding site in CAMP resistance, we constructed transcriptional lacZ fusions with the vraFG operon and mprF gene promoters, using the pSA14 vector, with or without the potential GraR operator sequence (vraF'-lacZ and DAvraF'-lacZ or mprF'-lacZ and DAmprF'-lacZ, respectively). The fusions were introduced into strains HG001, ST1036 (DgraRS) and ST1070 (DgraX) and b-galactosidase activity was measured during mid-exponential growth at 37uC in TSB with or without a 30 mn treatment with 50 mg ml 21 colistin (Fig. 4A). In the absence of GraX, GraSR or the potential GraR-binding site (strains ST1052 DgraX vraF'-lacZ; ST1041 DgraRS vraF'-lacZ; and ST1040 DAvraF'-lacZ), expression of vraF'-lacZ was strongly lowered, even in the presence of colistin (Fig. 4A). Comparable results were observed using the mprF'-lacZ and DAmprF'-lacZ fusions, although mprF clearly displays a higher basal level of expression in the absence of GraSR and GraX (Fig. 4B). We also note that there is a significant level of GraSRdependent expression from both the vraF and mprF promoters in the absence of colistin, indicating that the GraSR system is at least partially active in the absence of inducer, or that it is responding to some other signal under these conditions.
As a control, we introduced a transcriptional fusion with the constitutively expressed promoter of the TU elongation factor tufA gene into S. aureus strain HG001, and no difference in tufA'-lacZ expression was observed after a 30 mn treatment with or without 200 mg ml 21 colistin (see Supplementary Material Fig. S1A). In order to rule out a potential colistin effect independent of its function as a CAMP, we measured b-galactosidase activities of the two fusions (vraF'-lacZ and mprF'-lacZ) in strain HG001 following 30 mn incubation in the presence or absence of 5 mg ml 21 Figure 3. Identification of potential GraR-binding sites in the promoters of known GraR-regulated genes. (A) Primer extension analysis of mprF, dltXABCD and vraFG transcripts was carried out using total RNA extracted from S. aureus strain HG001 treated with 200 mg ml 21 colistin during mid-exponential growth at 37uC in TSB, using specific oligonucleotides for mprF, dltX and vraF (lanes 1 to 3 respectively). The corresponding Sanger dideoxy chain termination sequencing reactions (GATC) were carried out on PCR-generated DNA fragments corresponding to the respective upstream regions (see Table 5 To further investigate the role of the GraR-binding site we have shown to be required for GraR-dependent regulation, we constructed two fusions of the same length between the vraFG operon promoter sequence and the lacZ gene in the pSA14 vector. The two fusions only extend nineteen base pairs upstream from the potential GraR binding site, and differ by the introduction of seven point mutations in the inverted repeat operator sequence by site-directed mutagenesis through PCR, effectively destroying the palindromic sequence (Fig. 5A). The native and mutated promoter fusions (vraF2'-lacZ and vraF2*'-lacZ respectively) were introduced into strain HG001 and b-galactosidase activity was measured during mid-exponential growth at 37uC in TSB with or without a 30 mn treatment with 200 mg ml 21 colistin (Fig. 5B). Expression of vraF2'-lacZ was induced by colistin (Fig. 5B, strain ST1168), whereas in strain ST1169 (HG001 vraF2*-lacZ) the mutations destroying the inverted repeat strongly diminished vraF2'-lacZ expression and induction by colistin was lost, leaving only a low basal level of expression similar to that measured in the absence of GraR or its binding site (Fig. 4A).
Taken together, these results show that GraSR, GraX and the inverted repeat sequence are all required for the expression of GraR-regulated genes and their induction in response to CAMPs, strongly suggesting that this sequence constitutes the GraR operator sequence. Although the exact function of GraX in this regulatory pathway remains to be elucidated, we show here that it acts as a cofactor of GraSR-dependent transcription activation.

Genome-based prediction of the GraSR regulon
We then used the sequence of the identified likely GraR binding site to search for new potential GraSR-regulated genes in S. aureus. For this purpose, we used the restricted consensus 59-ACAAAWKTGT-39 to scan the S. aureus NCTC 8325 genome using the SearchPattern function of the ARTEMIS software [38]. To select GraSR-regulated candidate genes, we excluded inverted repeats lying more than 500 pb upstream from the annotated translational initiation site of each gene. We identified potential GraR-binding sites on either strand upstream from 29 genes or operons ( Table 2). Among these, 13 had already been described in S. aureus and another 10 genes encode putative proteins whose potential function can be deduced from sequence similarities and are suggested to be involved in different cellular pathways, whereas the remaining genes are of unknown function (Table 2).
Interestingly, 15 genetic loci encode proteins that can be classified in major functional groups. The first includes known and putative antimicrobial resistance-associated proteins: MprF, DltABCD, VraFG, and a b-lactam antibiotic modifying enzyme named PnbA [39]. The second group corresponds to transport proteins: the oligopeptide ATP-binding transporter OppF, a putative glucaric acid transporter (SAOUHSC_02815) and two genes located directly downstream from the vraFG operon, encoding a putative inorganic phosphate transporter (SAOUHSC_00669 and SAOUHSC_00670) known as PitAB in E. coli [40]. In the third group, involved in cell envelope modification, we found the tarM operon involved in teichoic acid glycosylation [41] and genes encoding the cell wall amidase SsaA, a probable autolysin regulator (AtlR-like), and the SpsB signal peptidase. The fourth class of potentially GraR-regulated genes is linked to oxidoreduction processes, including the qoxABCD quinol oxidase operon, ald, an alanine dehydrogenase gene, and genes encoding the YrkE-like protein containing multi-redox domains and the YtbE-like protein probably involved in 2,5-didehydrogluconate reduction.
Given the functional coherence of the identified loci, we investigated the relevance of this newly defined GraSR regulon by alignment of the 29 identified potential GraR-binding sites using the WebLogo website (http://weblogo.berkeley.edu/), generating a perfect 10 bp inverted repeat consensus sequence with a high degree of conservation, constituting the likely GraR operator: 59 ACAAA TTTGT 39 (Fig. 6).

Expression profiling of the GraSR regulon
Having defined a consensus GraR operator sequence and several new potential members of the GraSR regulon, we wished to validate these predictions in vivo. Using the BaSysBio Sau T1 chip, a NimbleGen 385K feature tiling array designed to cover both strands of the entire Staphylococcus aureus NCTC 8325 genome (see Materials and Methods), we examined global expression changes between the parental HG001 strain and the DgraRS mutant (ST1036), grown to mid-exponential phase in TSB with . GraSR-dependent gene expression requires GraX, colistin and the consensus binding site. vraFG (A) and mprF (B) expression was followed using transcriptional lacZ fusions, with or without the upstream GraR operator sequence (vraF'-lacZ, mprF'-lacZ and DAvraF'-lacZ, DAmprF'-lacZ, respectively). The fusions were introduced in S. aureus strains HG001, ST1036 (DgraRS) and ST1070 (DgraX). Expression was measured during mid-exponential growth in TSB at 37uC (grey bars) or after treatment with 50 mg ml 21 colistin (black bars). b-Galactosidase assays were performed as described in Materials and Methods. Means and standard deviation values are presented from three independent experiments. doi:10.1371/journal.pone.0021323.g004 50 mg ml 21 colistin. A total of 424 genes were found to be significantly differentially expressed in the DgraRS mutant compared with the parental strain, with a $ 1.8-fold change in transcriptional levels and a P-value (Z-test) #3.5610 24 . Among these, 235 were positively controlled by GraSR and 189 were repressed (Table S1 and Table S2, respectively). Interestingly, among the positively controlled genes, the most highly regulated encode major virulence factors or regulators, while the remaining genes belonged to the principal categories uncovered by our in silico predictions (see Table 2; antimicrobial resistance, transport, cell envelope modification, oxidoreduction processes) as well as stress response genes, and multiple regulatory and metabolic pathways (acetate, purine and pyrimidine, pyridoxal, xanthine) (Tables S1 and S2). We chose to focus our attention on positively controlled genes involved in the classes uncovered by our in silico analysis, as well as virulence, regulation and stress response which are listed in Table 3.
The most strongly regulated genes encode haemolysins, the AgrBDCA peptide quorum-sensing signal transduction pathway, members of the Sar family of virulence regulators, several host interaction proteins and virulence factors, (fibrinogen binding protein, ClfB, CHIPS, haemolysins, Sbi, SdrH), autolysins, as well as quinol oxidases (Table 3). This is the first report linking GraSR and the AgrCA major S. aureus virulence regulatory system.
Among the regulatory genes, we note those encoding the LytSR TCS, involved in autolysis and biofilm formation [42,43]. Most of the genes involved in cell envelope modification encode autolysins, including the AtlA major bifunctional autolysin, the SceD and IsaA transglycosylases, as well as seven genes encoding potential amidases with CHAP domains (Cysteine, Histidine-dependent Amidohydrolases/Peptidases), such as Sle1 or SsaA (Table 3). Interestingly, eight of the GraSR-dependent autolysin genes also belong to the WalKR regulon [44,45] (indicated by an asterisk in Table 3) suggesting a significant regulatory overlap between the two cell envelope signal transduction pathways. Indeed, thirteen other members of the GraSR regulon have also been predicted as belonging to the WalKR regulon as they are preceded by a consensus binding site for the WalR response regulator [44,46], such as the qoxABCD and SAOUHSC_00669-SAOUHSC_00670 operons (indicated by asterisks in Tables 2 & 3, S1 & S2).
In order to validate our microarray data, we chose several relevant genes (qoxA, ssaA, SAOUHSC_00669 and agrB) and compared their relative expression levels in the parental HG001 strain and the DgraRS mutant by qRT-PCR. As shown in Fig. 7, we confirmed by qRT-PCR that all of these genes are positively controlled by GraSR, with factors higher than those seen in the transcriptome analysis, ranging from approximately 3-to 29-fold. Results obtained using the two methods showed a linear correlation ( Fig. S2; see Supplementary Material).
GraXSR are involved in S. aureus resistance to oxidative stress Since GraSR control the expression of genes that appear to involved in oxidoreduction processes, we compared the sensitivity to oxidative stress of the parental HG001 strain with that of the ST1036 (DgraRS) and ST1070 (DgraX) mutants. Cells were grown in TSB in the presence or absence of 40 mM paraquat (methylviologen). No significant difference in growth between the strains was observed in the absence of paraquat (Fig. 8, open  symbols). However, as shown in Fig. 8 (closed symbols), the DgraX and DgraRS mutants were much more strongly affected by the presence of paraquat than the HG001 parental strain. Moreover, similar results were obtained in the presence of H 2 O 2 for the DgraRS mutant (data not shown). These results reveal a novel function for the GraSR system in resistance of S. aureus to oxidative stress.
The GraSR system is required for growth of S. aureus at high temperature The GraSR system is involved in cell envelope modifications through regulation of mprF, the dlt operon and autolysin genes, and also controls the expression of stress response genes (Tables 3, S1 & S2). We therefore tested the ability of the DgraRS mutant to grow

ACAAAATTGT… (N 129 ) …ATG
Hypothetical protein a Gene names correspond to the annotation of the S. aureus NCTC 8325 genome sequence [69]. Only the first gene is indicated for operons. b Positions of the inverted repeats of the potential GraR-binding sites are indicated for the given DNA strand with respect to the translation initiation codon. c Known and putative functions for each regulon member based on genome annotations are indicated. Based on these predictions, the potential regulon members were divided in six categories (antimicrobial resistance, transport, cell envelope modification, oxidoreduction processes, other functions and unknown function). *Indicates genes known or predicted to be controlled by the WalKR TCS [44,46]. doi:10.1371/journal.pone.0021323.t002 at high temperatures using a plate spotting assay. Strains ST1120 (HG001 pMK4-Pprot), ST1117 (DgraRS pMK4-Pprot) and the complemented DgraRS mutant, DgraRS-c (Strain ST1116 DgraRS pMK4-Pprot-graR) were grown in TSB at 37uC and diluted to an OD 600 nm of 0.2. Serial dilutions were then carried out, spotted on TSA plates and incubated at 37uC or 44uC for 48 h. As shown in Fig. 9, growth of the DgraRS mutant was strongly impaired at 44uC as compared to the parental strain, whereas no differences were observed between the two strains at 37uC. Resistance to high temperatures was almost completely restored in the complemented DgraRS-c strain carrying the graR gene on a multicopy plasmid, indicating that this phenotype can be compensated by overproducing the response regulator alone. These results demonstrate an important role for the GraSR system in growth of S. aureus at high temperatures.

Discussion
Cationic antimicrobial peptides (CAMPs) are a major component of host innate immune defense systems, produced by all living organisms, and have emerged as promising therapeutic antimicrobial agents [47,48]. Part of the success of some major human pathogens such as Staphylococcus aureus can be attributed to efficient CAMP resistance. One such mechanism involves incorporation of positively charged residues into the envelope, effectively increasing electrostatic repulsion of CAMPs from the cell surface. In S. aureus, this is accomplished through D-alanylation of teichoic acids, mediated by the DltABCD enzymes, as well as by MprFdependent lysylination of phosphatidylglycerol [10]. Expression of mprF and the dlt operon is induced by the presence of CAMPs and specifically controlled by the GraSR TCS, which has attracted growing interest in recent years [16,17,18,19]. GraSR also control expression of the vraFG operon, located directly downstream of the graXRS genes and encoding an ABC transporter also playing a role in CAMP resistance [16,17,18,19].
In this study we wished to further define the GraSR regulon and its function. We determined that unlike other TCSs such as VraSR or AgrCA [13,30], GraSR do not autoregulate their own synthesis. We demonstrated that graX, cotranscribed with graRS, specific to S. aureus and also involved in CAMP resistance, encodes an essential regulatory cofactor of the GraSR signaling pathway, effectively constituting a three-component system.
Noting that an imperfect palindromic sequence upstream from the vraFG operon had been suggested as a potential regulatory target in a multi-genome analysis of low G+C% Gram-positive bacteria [37], we found this inverted repeat as being highly conserved upstream from two other well-studied GraSR target genes/operons, mprF and dltXABCD. Extending our analysis through detailed genome scanning of the S. aureus NCTC 8325 genome, we were able to derive a highly conserved ten base pair perfect palindromic sequence (59 ACAAA TTTGT 39) upstream from 29 potential GraSR regulon members ( Table 2). By a genetic approach combining deletions and point mutations, we were able to conclusively demonstrate that this sequence is essential for transcriptional regulation by GraR and induction in response to CAMPs, indicating it is the likely GraR operator binding site. Despite multiple attempts, we were unable to purify the GraR response regulator in an active form in order to show DNAbinding in vitro. However, our proposed GraR operator binding site is similar to that suggested for the closely related VirR response regulator of Listeria monocytogenes which was shown to bind to DNA [35]. The two proteins share 46% overall amino acid sequence identity, rising to 73% for the winged helix-turn-helix domain, with 9 out of 11 identical residues in the DNA recognition helix, indicating that they must bind to similar DNA sequences.
Of the 29 potential GraR regulon members we identified with this binding site present in their upstream regions, we showed that 13 of these are indeed controlled by the GraSR system in vivo under our conditions. Of these, nine are positively regulated (dltXABCD, mprF, vraFG, tarM, sdrH, ssaA, SAOUHSC_00669, qoxABCD, rplE) and four were found to be repressed (SAOUHSC_00146, SAOUHSC_00991, SAOUHSC_00882, SAOUHSC_02816). This suggests that for the remaining 16 predicted target genes, either the potential GraR binding site is not appropriately located with respect to the promoter in order to allow transcriptional activation/repression, or that additional genetic control mechanisms exist for these genes, preventing their expression under our specific experimental conditions. During a phenotypic analysis, we observed that the DgraRS mutant displayed increased sensitivity to oxidative stress. This may in part be due to positive control by GraSR of the mntABC manganese transporter genes (Table 3), which have been shown to play a role in S. aureus resistance to superoxide radicals [49]. Furthermore, these data suggest that the GraSR system may also respond to other signals in S. aureus, and not only to the presence of CAMPs. The combined sensitivity of the DgraRS mutant to oxidative stress and antimicrobial peptides could explain the important role of this system in staphylococcal survival in human neutrophils [23,24].
Virulence gene expression in S. aureus involves a complex regulatory network, with at least four two-component systems (AgrCA, ArlSR, SaeSR and SrrAB) and several accessory transcription factors (SarA, SarS, SarT, SarR, and Rot) [13]. In this study, our transcriptome analysis allowed us to unveil previously unsuspected connections between the GraSR TCS and the AgrCA signal Figure 6. The GraR operator consensus is a perfect inverted repeat obtained by alignment of regulon gene upstream sequences. The consensus sequence for the GraR-binding site was generated using the WebLogo tool (http://weblogo.berkeley.edu/) by alignment of the upstream sequences of the 29 potential regulon genes identified by in silico analysis (   (Table S2). This is the first report linking the GraSR and AgrCA TCSs, which could in part explain the numerous results implicating the GraSR system in S. aureus virulence using several experimental models [18,[23][24][25][26]. It is likely this connection could not be detected in a previous GraSR transcriptome analysis as it was carried out using strain SA113, an agr mutant, in the absence of CAMPs as an inducer [16,27]. However, comparing the two sets of transcriptome data obtained during exponential growth revealed similar numbers of GraSR-regulated genes using the same cutoff values, although significant differences in the genes controlled were observed. Indeed, only 63 genes were common to the two experiments, and 19 differentially regulated in each condition, which may be attributed to differences in the genetic backgrounds of the two strains (HG001 and SA113) or to a different behaviour of the GraSR system under basal growth conditions or upon induction in the presence of CAMPs. In addition to AgrCA, GraSR also appear to interact with the WalKR TCS, involved in cell wall metabolism and autolysis [45,51], providing increasing evidence for TCS signal transduction networking in S. aureus, as described for the Gram-positive model bacterium B. subtilis [52]. Regulatory overlap with the WalKR regulon is particularly extensive, with at least 21 WalR regulon genes also  Table 3. Cont.
The S. aureus GraS/GraR Regulon controlled by GraSR, suggesting a significant level of interaction between the two cell envelope signal transduction pathways. These include eight autolysin genes (atlA, sceD, isaA, ssaA, sle1, SAOUHSC_00671, SAOUHSC_02576, SAOUHSC_02883) that have all been shown to be transcriptionally controlled by WalKR [45] (Delaune et al., in preparation). Thirteen other members of the GraSR regulon have also been predicted as being controlled by WalKR as they are preceded by a consensus binding site for the WalR response regulator, including the qoxABCD quinol oxidase biosynthesis operon, the SAOUHSC_00669-SAOUHSC_00670 operon, prs, encoding a putative ribose-phosphate pyrophosphokinase, the manA mannose-6 phosphate isomerase gene, the SAOUHSC_00738 and vraFG ABC transporter genes as well as a gene of unknown function, SAOUHSC_ 00060 (Tables 3, S1 & S2) [44,46]. This is reminiscent of genes under multiple regulatory controls, such as the B. subtilis degQ gene, which is preceded by tandemly arranged binding sites for both the DegU and ComA response regulators [53], and it will be interesting to determine the respective contributions of WalKR and GraSR to expression of their co-regulated genes. During this investigation, we also showed that GraSR are required for growth of S. aureus at high temperatures. This may be linked to their role in modification of wall teichoic acids, which are known to be required for growth under these conditions [54]. This is the first report revealing a function for GraX as a regulatory cofactor of the GraSR TCS, and showing a role for this system in staphylococcal high temperature and oxidative stress survival. We have shown that the GraSR system controls genes involved in stress response, cell wall metabolism and pathogenesis control pathways in addition to its primary role in CAMP resistance, significantly enhancing its importance as a major signal transduction pathway in S. aureus.

Bacterial strains and growth procedures
Bacterial strains and plasmids are listed in Table 4. Escherichia coli K12 strain DH5a TM (Invitrogen Life Technologies) was used for cloning experiments. Plasmid constructs were first passaged through the restriction modification deficient S. aureus strain RN4220 [55]   before introduction into S. aureus strain HG001, a rsbU + variant of strain NCTC 8325 [27] and its derivatives. HG001 is a genetically tractable, clinically relevant non mutagenized strain and was used for all genetic studies. E. coli strains were grown in LB medium with ampicillin (100 mg ml 21 ) when required. S. aureus was grown in Trypticase Soy Broth (TSB; Difco) with shaking (180 rpm) at 37uC ; for plasmid selection, chloramphenicol (10 mg ml 21 ) or erythromycin (2 mg ml 21 ) were added as required. E. coli and S. aureus strains were transformed by electroporation using standard protocols [56] and transformants were selected on LB or Trypticase Soy Agar (TSA; Difco) plates, respectively, with the appropriate antibiotics. Colistin sulfate, nisin and indolicidin (Sigma-Aldrich) were used as CAMPs when required.

DNA manipulations
Oligonucleotides used in this study were synthesized by Sigma-Proligo and are listed in Table 5. S. aureus HG001 chromosomal DNA was isolated using the MasterPure TM Gram-positive DNA purification Kit (Epicentre Biotechnologies). Plasmids were 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 Pwo thermostable DNA Polymerase (Roche) were used according to the manufacturers' recommendations. Nucleotide sequencing of plasmid constructs was carried out by Genome Express-Cogenics or GATC Biotech.

Plasmid and mutant construction
The thermosensitive shuttle vector pMAD was used for introducing markerless gene deletions [57]. 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, two DNA fragments, of approximately 600 bp, corresponding to the chromosomal DNA regions located directly upstream and downstream from the gene(s) of interest, were generated by PCR, digested with XhoI or BsaI, and ligated using T4 DNA ligase. The ligation product was reamplified using the external primers and purified before cloning into the temperature-sensitive shuttle vector pMAD between the EcoRI/ NcoI or SalI/BglII restriction sites. Nucleotide sequences of the constructs were confirmed by DNA sequencing and the resulting plasmids passaged through S. aureus strain RN4220 and introduced into S. aureus HG001. Integration and excision of the pMAD derivatives and deletion of the chromosomal region of interest was carried out as previously described [57]. Gene deletions in mutant strains were systematically verified by PCR and qRT-PCR.
Plasmid pMK4Pprot, a derivative of vector pMK4 carrying a constitutively expressed Gram-positive promoter sequence [58] was used for gene complementation experiments. Complementation of the ST1036 (DgraRS) strain was carried out using a DNA fragment corresponding to the graR gene coding sequence, amplified with oligonucleotides MF118-MF119, generating BamHI/PstI restriction sites at the extremities, and cloned in the replicative plasmid pMK4-Pprot.
Plasmid pSA14 was used to measure the expression of S. aureus genes by constructing transcriptional fusions between gene promoter regions and the E. coli lacZ reporter gene. The pSA14 plasmid was constructed by cloning a 3.2 kb EcoRI-PstI DNA fragment from plasmid pHT304-18Z [59] between the corresponding restriction sites of the pMK4 shuttle vector [60]. The insert carried the promoterless E. coli lacZ gene fused to the B. subtilis spoVG ribosome binding site [61,62]. In the absence of an upstream promoter, the pSA14 vector displays no detectable b-galactosidase activity, making it a highly useful transcriptional lacZ fusion reporter tool for in vivo expression analysis. For constructing transcriptional lacZ fusions, promoter regions of the mprF gene and the graXRS and vraFG operons, and truncated promoter regions of mprF gene and the vraFG operon were amplified by PCR using oligonucleotides introducing BamHI/PstI restriction sites, except for the vraF2*-lacZ fusion, constructed using a forward oligonucleotide containing seven mismatches generating point mutations (see Table 5). The corresponding DNA fragments were then cloned between the corresponding restriction sites of the pSA14 vector, yielding plasmids listed in Table 4. The pSAtufA plasmid [63] carrying the strong constitutive promoter of the tufA gene was used as a control.

MIC determinations
MIC determinations were performed in a 96-well microtiter plate with a 100 ml culture volume. Bacterial cultures were grown for eight hours in TSB at 37uC, diluted to an OD 600 nm = 0.05 and used to inoculate wells containing TSB with standard two-fold increments of colistin concentration (v/v). Plates were incubated for 12 h with vigorous shaking at 37uC in a Synergy 2 thermoregulated spectrophotometer plate reader using the Gen5 TM Microplate Software (BioTek Instruments Inc., Winooski, VT). All experiments were performed in triplicate.

Extraction of total RNA
Cells were grown until OD 600 nm = 1, and colistin was added to the medium at 50 mg ml 21 or 200 mg ml 21 when required. Growth was pursued during 30 min and cells were harvested by centrifuging 30 ml culture samples (4 min; 5,400 x g) and immediately frozen at 280uC. RNA extractions were then performed as previously described [64], 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 [65] using 30 mg of RNA, 2 pmol of oligonucleotide (previously radiolabeled with [c-32 P] ATP using T4 polynucleotide kinase, New England Biolabs), and 200 U of Superscript II reverse transcriptase (Invitrogen). Oligonucleotides were chosen so as to hybridize downstream from the translation initiation codon (see Table 5). 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).

b-Galactosidase assays
Cells were grown until OD 600 nm = 1, colistin was added to the medium at 50 mg ml 21 or 200 mg ml 21 when required and growth was pursued for 30 min. S. aureus strains carrying the different lacZ fusions were then harvested by centrifuging 2 ml culture samples (2 min; 20,800 x g). Cells were resuspended in 500 ml of Z buffer [66] with 0.5 mg ml 21 DNase, 5 mM DTT and 0.1 mg ml 21 lysostaphin added extemporaneously, and lysed by incubation at 37uC for 30 min. Cell debris were eliminated by centrifugation (2 min; 20,800 x g) and the supernatant was either used directly for assays or stored at 220uC. Assays were performed as previously described and b-galactosidase specific activity was expressed as Miller units mg 21 protein [66]. Protein concentrations were determined using the Bio-Rad protein assay (BioRad, Hercules, CA) [67]. All experiments were carried out in triplicate. Microarray experiments RNA samples for tiling arrays were prepared as described above using cultures of S. aureus strains HG001 and ST1036 (DgraRS) grown in TSB with 50 mg ml 21 colistin for GraSR induction, with an additional 2-fold dilution step in killing buffer (20 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 20 mM NaN 3 ) before centrifugation. RNA samples were then treated using the RNA Clean-Up kit (Norgen Biotech Corp., Canada) according to the manufacturer's recommendations and eluted in 40 ml of RNAse-free water. RNA preparations were quantified using a spectrophotometer at 260 nm and quality was checked by electrophoregram analysis on a BioAnalyzer (Agilent).
The BaSysBio Sau T1 NimbleGen 385K array was designed with a total of 383,452 features using OligoWiz 2.0 [68], with long iso-thermal probes (45-65 nt) covering the entire genome of Staphylococcus aureus NCTC 8325 (CP000253.1; [69] in 18 nt intervals on each strand (Hanne Jarmer, Technical University of Denmark, Lyngby, Denmark, personal communication). Tiling array experiments were carried out with 20 mg of each RNA sample, sent to Roche NimbleGen (Madison, WI, USA) where it was labelled and hybridized to the BaSysBio Sau T1 chip using the BaSysBio protocol for strand-specific hybridization [70]. All tiling array experiments were performed in triplicate using RNA isolated from independent cultures.
For data analysis, an aggregated expression value was computed for each Genbank annotated CDS as the median log 2 intensity of probes lying entirely within the corresponding region (Pierre Nicolas, MIG INRA Jouy-en Josas, personal communication). To control for possible cross-hybridization artefacts the sequence of each probe was BLAST-aligned against the whole chromosome sequence and probes with a SeqS value above the 1.5 cut-off were discarded (SeqS is 2 for a probe with two exact matches) [71].
Aggregated intensity values of the individual samples were normalized by median scaling using the Rosetta Resolver software (version 7.2.1, Rosetta Biosoftware). Statistical significance of differential expression between the wild type and the mutant strain was then evaluated using the Z-test (ArrayStat software package, GE Lifesciences). Differentially expressed genes were chosen with a ratio between the wild-type and mutant strain $ 1.8 and a Pvalue # 3.5610 24 . The complete MIAME compliant microarray data set is available at the NIH Gene Expression Omnibus (GEO) database under record number GSE26016: (http://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?token=vbcbxmyiykmqotq&acc=GSE26016).
cDNA synthesis and qRT-PCRs cDNAs were synthesized using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) according to the manufacturer's recommendations, in a 20 ml final reaction volume containing 1 mg total RNA. For qRT-PCR experiments, amplicon primers were designed using the BEACON Designer 4.02 software (Premier Biosoft International, Palo Alto, CA) (see Table 5). Quantitative real-time PCRs (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 [45]. All experiments were performed in triplicate.

Oxidative stress
S. aureus strains were treated with paraquat (methylviologendichloride hydrate) or H 2 O 2 (Sigma Aldrich) and growth was followed in a 96-well microtiter plate (100 ml culture volume).
Bacterial cultures were grown in TSB, diluted to an OD 600 nm = 0.05 and used to inoculate wells containing TSB with or without 40 mM paraquat, or 0.004% H 2 O 2 . Plates were incubated for 14 h with vigorous shaking at 37uC in a Synergy 2 thermoregulated spectrophotometer plate reader using the Gen5 TM Microplate Software (BioTek Instruments Inc., Winooski, VT). All experiments were performed at least in duplicate.

High temperature growth
The effect of high temperatures was observed on S. aureus strains ST1120 (pMK4-Pprot), ST1116 (DgraRS-c; DgraRS pMK4-Pprot-graR) and ST1117 (DgraRS pMK4-Pprot), on cells grown in TSB at 37uC and diluted to an OD 600 nm = 0.2. Cultures were serially diluted (10 21 , 10 22 , 10 23 and 10 24 fold) and 10 ml of each dilution was spotted onto TSA plates, which were dried for 10 min at room temperature and incubated at 37uC or 44uC for 48 h. Figure S1 lacZ fusion control experiment expression analysis (A) Expression of tufA'-lacZ is not induced by colistin. Expression of the tufA'-lacZ fusion was measured in strain ST1189 (HG001 tufA'-lacZ) during mid-exponential growth at 37u C in TSB (grey bars) or after treatment with 200 mg ml 21 colistin (black bars). b-Galactosidase assays were performed as described in Materials and Methods. (B) Indolicidin induces expression of the vraFG operon and mprF. Expression of vraF'-lacZ and mprF'-lacZ fusions in S. aureus strain HG001 was measured during mid-exponential growth at 37uC in TSB (grey bars) or after treatment with 5 mg ml 21 indolicidin (black bars). b-Galactosidase assays were performed as described in Experimental Procedures. (TIF) Figure S2 Linear correlation between microarray and qRT-PCR experiments for expression of GraSR-dependent genes. Fold changes in expression as measured by qRT-PCR and transcriptome analysis measured for 4 representative genes in the S. aureus HG001 strain relative to the ST1036 (DgraRS) strain grown in the same conditions were plotted against each other to evaluate their correlation. Data points were analyzed in triplicate by both methods. (TIF)