INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Staphylococcus aureus is a major cause of human disease. The organism causes a variety of clinical manifestations, ranging from localized skin infections to severe sepsis, and is a leading cause of hospital-acquired infection (3). Despite advances in antibacterial chemotherapy, S. aureus strains have demonstrated resistance to all currently available antibiotics. Due in part to the immense clinical importance of this organism, an enormous amount of effort has been directed toward identifying the genes and regulatory mechanisms associated with S. aureus pathogenesis. Collectively, this work has demonstrated that the organism's pathogenesis can be attributed to its capacity to produce a variety of virulence factors (29).

The identification of virulence factors and the regulatory networks that influence their expression has been facilitated by the observation(s) that many, if not most, virulence genes are expressed in laboratory cultures. While there is currently a substantial list of staphylococcal virulence factors, it is likely that this list is incomplete and is skewed by the limitations of the experiments used to identify them. Virulence factors that have already been identified generally include (i) bacterial surface proteins that are involved in processes such as adhesion and evasion of the host immune response and (ii) secreted exoproteins that degrade host tissue(s) and inactivate host defensive mechanisms (29).

The genes encoding most virulence factors belong to an extensive regulon that is coordinately regulated in response to a variety of intra- and extracellular signals (1, 5, 21). Octapeptide signaling molecules that are produced as laboratory cultures increase in cell density are the best-studied "stimuli" of the virulence response, mediating a transition in expression of the genes encoding virulence determinants from a predominance of cell surface-expressed genes to a predominance of genes encoding exoproteins (26).

The density-dependent regulation of most virulence factors is mediated by regulatory loci such as the accessory gene regulator (agr) and staphylococcal accessory regulator (sarA) loci (9, 19). In the laboratory setting, post-exponential growth activates the agr locus, which contains two divergent promoters, P2 and P3, that direct expression of RNAII and RNAIII transcripts, respectively (19). RNAII encodes four proteins, AgrB, AgrD, AgrC, and AgrA, all of which are required for agr-mediated virulence factor regulation (19, 27). AgrD and AgrB act to generate an octapeptide quorum-sensing molecule (autoinducing peptide [AIP]), which, after reaching an extracellular threshold concentration, stimulates activation of AgrC and AgrA, the sensor and regulator, respectively, of a two-component regulatory pathway (17-19, 27, 28). Activated AgrA results in the upregulation of RNAII and RNAIII production; the latter is the effector molecule of the agr response (25, 27, 28). RNAIII expression is, in part, responsible for the upregulation of exoprotein production and the downregulation of cell surface protein transcription during the late-log and stationary phases of growth. The manner in which RNAIII regulates target genes is not yet understood, but it is clear that the molecule acts mainly at the level of transcription. RNAIII also encodes the structural gene for delta-hemolysin, although the transcript, rather than the protein product, acts as a global virulence factor regulator (28).

The S. aureus accessory regulatory (SarA) protein also influences both exoprotein and cell surface protein expression (9). The sarA locus contains three overlapping transcripts designated sarA, sarC, and sarB, each of which has a common 3' end encoding SarA (2). SarA binds to conserved regions termed Sar boxes within promoter regions of genes encoding cell surface proteins (spa, encoding protein A), genes encoding exoproteins (hla, encoding alpha-hemolysin), and agr (12). SarA binding to agr promoter elements augments both RNAII and RNAIII transcription and therefore contributes to virulence factor regulation indirectly (7, 13). SarA has also been shown to regulate expression of spa and hla in an agr mutant background, indicating that SarA controls regulation of certain virulence factors directly, in an agr-independent manner.

Because RNAIII and/or SarA influences the transcription of most known virulence factors, it was hypothesized that additional, previously uncharacterized potential virulence factors may be identified by analyzing the S. aureus genome for open reading frames (ORFs) that are expressed in an agr- and/or SarA-dependent manner. Furthermore, such an analysis may begin to unravel the overlapping RNAIII and SarA regulatory contributions to known virulence factors. Using an S. aureus GeneChip (Affymetrix), we have identified genes that produce transcript patterns expected of genes regulated by the S. aureus agr and/or sarA locus. In addition to confirming the predicted expression patterns of a number of known virulence factors, and thus validating the methodology, we identified a set of putative virulence factors as being regulated by each effector. Moreover, expression patterns of genes involved in a number of biological processes that are difficult to reconcile as contributing directly to S. aureus pathogenesis produced expression profiles expected of genes regulated in an agr- and/or SarA-dependent, density-dependent manner. These results suggest that, in addition to regulating virulence factors, each effector is a more-general transcriptional regulator than previously recognized.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and culture conditions. The S. aureus strains used in this study are listed in Table 1. Strains were grown overnight in 25 ml of brain heart infusion (BHI) medium at 37°C with aeration. Overnight cultures were used to inoculate (1:100 dilution) 1.5 liters of fresh BHI medium. Cultures were incubated with vigorous aeration at 37°C and aliquots were removed at the indicated growth phases as determined by growth phase analysis (data not shown). Cells from each aliquot were pelleted by centrifugation at 5,000 × g for 10 min at 4°C in a Beckman JA-10 rotor and were stored at 80°C.

RNA extraction. Total bacterial RNA was extracted from samples by resuspending each cell pellet at a concentration of 10⁹ CFU per ml in TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) containing 50 µg of lysostaphin (Amicon)/ml. Resuspensions were incubated at 37°C for 60 min, and 1 × 10¹⁰ to 5 × 10¹⁰ CFU was then applied to a Qiagen RNeasy Maxi column, according to the manufacturer's recommendations for prokaryotic RNA isolation. RNA concentrations were determined by spectrophotometry (an optical density at 260 nm of 1.0 equals 40 µg of RNA/ml). To assess RNA integrity and for Northern blot analysis, 2 µg of each RNA sample was electrophoresed in a 1.2% agarose-0.66 M formaldehyde gel, according to Qiagen RNA electrophoresis recommendations.

Northern blot analysis. RNA samples were transferred from formaldehyde-containing agarose gels to Biodyne B membranes (Gibco-BRL) by standard molecular procedures. High-stringency hybridization was performed in Kapak/Scotchpak heat-sealable pouches containing 12 ml of PerfectHyb Plus hybridization buffer (Sigma) and 10 µl of radiolabeled randomly primed DNA synthesis products from PCR-generated templates at 42°C. Membranes were subjected to high-stringency washes and were analyzed by phosphorimaging (Bio-Rad Phosphorimager) or densitometry (Bio-Rad). The following PCR primers were used to generate PCR templates: for 16S rRNA (478 nucleotides [nt] of GenBank accession no. Y15856), 5'-AAATCTTGACATCCTTTGACAACTC-3' and 5'-CTAGCTCCTAAAAGGTTACTCCACC-3'; for spa (1,999 nt of GenBank accession no. M18264), 5'-AAGATTTAATTGAAACAATCCACCA-3' and 5'-GACCAGGTTTGATCATGTTTTTATC-3'; for hla (769 nt of GenBank accession no. X01645), 5'-TAGGTTCCATATTGATGAATCCTGT-3' and 5'-ATATTTGTTTGTTGTTTGGATGCTT-3'; and for RNAIII (480 nt of GenBank accession no. AF288215), 5'-GGGGCTCACGACCATACTTA-3' and 5'-GGAGTGATTTCAATGGCACA-3'. All PCR products were subjected to restriction enzyme analysis and were gel purified prior to labeling reactions.

mRNA enrichment, fragmentation, and biotinylation. For mRNA enrichment reactions (240 µl), 200 µg of total bacterial RNA was mixed with 0.7 µM (final concentration) rRNA-specific oligonucleotide mix (5'-GATACGGCTACCTTGTT-3', 5'-TCAACCTTGCGGTCGTACTC-3', 5'-TCCGGATAACGCTTGCCACC-3', 5'-AGCACTTATCCCGTCCACAC-3', 5'-CTACAGTAAAGCTCCACGGG-3', and 5'-TCCCCATCACAGCTCAGCCT-3'). Aliquots (30 µl) were transferred to 0.5-ml thin-walled Eppendorf tubes (Perkin-Elmer), and solutions were incubated at 70°C for 5 min in a thermocycler (Perkin-Elmer). Reverse transcriptase reaction mixture (50 mM Tris-HCl [pH 8.3], 10 mM MgCl₂, 75 mM KCl, 10 mM dithiothreitol, 0.5 mM each deoxynucleoside triphosphate, 735 U of RNAguard [Amersham Pharmacia Biotech], and 3,000 U of Moloney murine leukemia virus reverse transcriptase [Epicentre]) was then added. Solutions were incubated at 42°C for 25 min, followed by 45°C for 20 min. For rRNA removal, 0.5 U of RNase H (Epicentre) was added and mixtures were incubated first at 37°C for 45 min and then at 65°C for 5 min. DNase I (0.12 U/µl; Amersham Pharmacia Biotech) and 0.225 U of RNAguard were added, and mixtures were incubated at 37°C for 20 min. Reactions were terminated by addition of 10 mM EDTA. Samples were applied to a Qiagen RNeasy minicolumn, according to manufacturer recommendations for RNA clean-up. For RNA fragmentation, 20 µg of an mRNA-enriched sample was mixed with fragmentation buffer (1× T4 polynucleotide kinase buffer [New England Biolabs], 70 mM Tris-HCl [pH 7.6], 10 mM MgCl₂, 5 mM dithiothreitol) and incubated at 95°C for 30 min and then cooled to 4°C. Samples were then 5'-thiolated by addition of 0.1 mM -S-ATP and 1 U of T4 polynucleotide kinase (New England Biolabs) and were incubated at 37°C for 50 min, followed by 65°C for 10 min. Samples were then ethanol precipitated and subsequently biotinylated by addition of 2 mM polyethylene oxide (PEO)-iodoacetyl-biotin and 30 mM morpholinepropanesulfonic acid (MOPS), pH 7.5. Unincorporated biotin molecules were removed by passing RNA mixtures through a Qiagen RNA/DNA minicolumn according to manufacturer recommendations.

S. aureus GeneChip design. Preliminary genomic sequence data of the S. aureus COL strain were obtained from The Institute for Genomic Research (TIGR) website (http://www.tigr.org). The sequence consisted of approximately 2,000 contigs, which were concatenated into a single sequence alternating with an 18-nt sequence containing stop codons in all 6 frames. ORF predictions were made using a combination of GLIMMER 1.0 (32) and GeneMark.hmm (23), with a minimum ORF size of 75 nt. In all, 4,528 ORFs, 12 tRNAs, and 3 rRNAs were tiled, with an average of 25 probe sets per ORF. Due to the preliminary state of the genome sequence, many genes are represented on the chip as partial, duplicate, or overlapping fragments. Recent analyses based on the updated sequences of the genomes of COL, NCTC 8325 (OU-ACGT), and MRSA (Sanger Centre) suggest that these 4,528 tilings represent approximately 2,700 to 2,900 individual genes and that >86% of the S. aureus COL genome is represented on the chip.

GeneChip hybridization and washing. Prior to RNA hybridization, S. aureus GeneChips (Affymetrix) were brought to room temperature, washed once with hybridization buffer (100 mM N-morpholinoethanesulfonic acid [MES], 1 M Na⁺, 20 mM EDTA, 0.01% Tween 20), loaded (250 µl) with hybridization buffer, and incubated at 45°C for at least 10 min. Hybridization mixture (1.5 µg of biotinylated mRNA-enriched RNA, 50 pM oligonucleotide B2 [Affymetrix], 0.1 mg of herring sperm DNA/ml, 0.5 mg of acetylated bovine serum albumin [BSA; Gibco-BRL], and 1× control spike-in cocktail [Genetics Institute] in 1× hybridization buffer) was denatured by incubating at 95°C for 5 min, followed by 45°C for 5 min, and was centrifuged at high speed on a tabletop microcentrifuge for 20 min at room temperature to pellet all nonsolubilized material. A 200-µl volume of hybridization mixture was then loaded onto a GeneChip and incubated for 15 h at 45°C. Following hybridization, the RNA-containing mixture was removed and stored at 80°C. GeneChips were then washed 20 times with nonstringent buffer (6× SSPE [1× SSPE is 0.18 M NaCl, 10 mM NaH₂PO₄, and 1 mM EDTA {pH 7.7}], 0.01% Tween 20, 0.005% antifoam) at 25°C. Chips were then subjected to 60 washes with stringent buffer (100 mM MES, 0.1 M Na⁺, 0.01% Tween 20) at 45°C. Following washing, the hybridized RNA was treated with primary stain (10 µg of streptavidin, 2 mg of acetylated BSA) in stain buffer (100 mM MES, 1 M Na⁺, 0.05% Tween 20, 0.005% antifoam) for 10 min at 25°C and was washed 40 times with nonstringent buffer. Next each GeneChip was subjected to a secondary stain containing 2 mg of acetylated BSA, 0.1 mg of goat immunoglobulin G, and 5 µg of biotinylated anti-streptavidin (Vector Laboratories) in stain buffer for 10 min at 25°C. The chip-bound RNA molecules were labeled with 10 µg of streptavidin-phycoerythrin conjugate (Molecular Probes)/ml in stain buffer for 10 min at 25°C. Excess label was removed by 100 nonstringent washes at 25°C.

GeneChip analysis. To increase reproducibility, each experiment was performed in duplicate (RNA extraction from various phases of growth for each strain) and each RNA sample was hybridized to two separate GeneChips. The work here constitutes a total of 16 GeneChips for each strain analyzed. Each GeneChip was scanned at 570-nm, 3-µm resolution in an Affymetrix GeneChip scanner. Affymetrix algorithms calculated signal intensities (average difference) and made present or absent determinations for each gene, as previously described (22). Next, to normalize for global systematic variations that could be caused by inconsistencies in loading, each average difference value was divided by the median average difference for a given GeneChip. Normalized intensity values were then averaged for each gene at each growth phase. GeneSpring, version 3.2.11, software (Silicon Genetics, Redwood City, Calif.) was used to plot normalized intensity values across growth phases. GeneSpring algorithms also determined signal strength values for each gene within a strain as the product of the average normalized signal intensity value of the gene of interest and the median normalized signal intensity of the gene taken throughout growth phases. To identify genes that are below the detection limit of the system, the signal strengths indicative of genes with profiles at the level of noise were determined for each strain as the average signal strength of genes considered absent (via GeneChip algorithms) plus 2 standard deviations.

Identification of Sar boxes. Intergenic regions were identified within the S. aureus N315 genome by comparing ORF nucleotide coordinates (20). Putative Sar boxes (12) were identified using the Genetics Computer Group (Madison, Wis.) program (Wisconsin Sequencing Package).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Identification of S. aureus virulence factors. In the laboratory setting, expression of many S. aureus virulence factors follows a predictable pattern. Cell surface virulence factors, such as protein A (encoded by spa), are predominately expressed during early-log-phase growth, but as the density of a growing culture increases, transcription of these genes decreases. Genes encoding extracellular virulence factors, such as alpha-toxin (encoded by hla), demonstrate a reciprocal phenotype; they tend to be transcribed at basal levels during early-log-phase growth and are upregulated at higher cell densities. This density-dependent regulation of most virulence determinants can be controlled by a product of the agr locus, RNAIII, and/or SarA.

agr regulation by SarA. To validate the transcription profiling methodology used, we first investigated whether the agr locus transcripts, RNAII and RNAIII, produced expression patterns mimicking previously reported data. It is well established that agr is temporally regulated in a growth phase-dependent manner. Basal levels of RNAII and RNAIII are detected during early-log-phase growth; transcription subsequently increases as cells progress through growth phases (16, 34). Additionally, Cheung and colleagues have shown that SarA interacts with agr promoter regions and facilitates RNAII and RNAIII transcription (7, 15).

View larger version (21K): [in this window] [in a new window]   FIG. 1.   Transcription profiles of known RNAIII and SarA-regulated genes. Transcript levels for wild-type (), agr mutant (), sarA mutant (), and agr sarA double-mutant () cells were measured during the early-, mid-, and late-log and stationary phases of growth (x axis). Data points were plotted as relative intensity values (y axis) (as described in Materials and Methods). (A) Average signal intensities for genes constituting RNAII transcripts (agrB, agrD, agrC, and agrA). (B) Signal intensities of RNAIII transcripts. (C) Profiles of alpha-toxin (hla) transcript titers. (Inset) agr and agr sarA mutant results. (D) Profiles of protein A (spa) transcript titers.

View larger version (47K): [in this window] [in a new window]   FIG. 2.   Northern blot analysis of known agr-regulated genes. Shown are levels of RNAIII (A), hla (B), spa (C), and rRNA (D) transcripts during the early-, mid-, and late-log and stationary phases of growth of wild-type (RN27) or agr mutant (RN6911) cells. Signal intensities were determined by densitometry and normalized to rRNA signals. Normalized relative signal intensities are shown below each panel.

agr and SarA regulation of alpha-toxin. To further validate the methodology and extend our knowledge about the transcription of the well-studied agr-regulated exoprotein alpha-toxin (encoded by hla), we investigated whether our transcript profiles correlated well with published hla expression patterns. It has been shown that RNAIII production promotes hla expression, with maximum levels being reached at the post-exponential phase of growth (6, 34). As shown in Fig. 1C, transcription profiling results demonstrated a 39-fold increase in hla transcription in wild-type (RN27) cells as they transitioned from the early-log to stationary phase of growth, and these results were in good agreement with the 20.9-fold increase seen in our Northern analysis (Fig. 2B). In contrast, the increase in transcription was reduced to 2.2-fold in RN6911 cells (Fig. 1C), confirming that a product of the agr locus mediates hla upregulation during the post-exponential phase. These results also indicate that an additional regulatory mechanism(s) may account for the slight (2.2-fold) increase in hla expression in agr mutant cells. Indeed, SarA binding sites have been identified within the hla promoter region, and SarA has been shown to bind directly to the promoter region and induce hla expression in an agr-independent manner (6, 12). To determine whether this modest increase in expression could be attributed to SarA, as opposed to an additional virulence regulator(s), the profiles of hla transcripts were compared for various growth phases of sarA and agr sarA mutant cells. As expected, hla expression was marginally induced in post-exponential sarA mutant cells (Fig. 1C). However, because SarA is necessary for complete expression of agr, the strain is effectively agr deficient and the observed reduction in hla expression may be a direct consequence of decreased production of RNAIII (10, 12). Yet hla was constitutively transcribed during various phases of growth in agr sarA mutant (ALC842) cells (Fig. 1C), confirming that SarA both directly and indirectly regulates alpha-hemolysin transcription. Interestingly, GeneChip analysis (see Materials and Methods) determined that hla transcripts were present in all phases of ALC842 growth (data not shown), implying that neither regulator is required for basal levels of alpha-hemolysin transcription. In addition, these results suggest that both RNAIII and SarA are required for wild-type levels of hla transcription during post-exponential aerobic growth.

agr and SarA regulation of protein A. It has previously been shown that wild-type S. aureus transcribes spa during early-log-phase growth and that its transcription is reduced when RNAIII is produced. Conversely, spa is transcribed at a high level throughout all phases of growth in an agr-null strain (34). The spa promoter contains SarA boxes (12). Both Northern analysis and spa-reporter fusion studies have shown that SarA is a repressor of protein A transcription (4, 10). To determine whether transcription profiling could identify genes known to be downregulated by each effector, we compared the protein A (spa) transcript profiles of each strain. As shown in Fig. 1D, wild-type cells maximally expressed spa at early- and mid-log phases of growth, and transcription decreased 6.8-fold as cells entered stationary phase. A similar phenotype (a 7.2-fold decrease) was observed by Northern analysis (Fig. 2C). Comparison of signal strengths (see Materials and Methods) demonstrated that agr mutant cells expressed high levels of spa throughout all phases of growth (average normalized signal strength value, 1,031.5, as opposed to 170.8 in wild-type cells), confirming that the agr locus is a potent repressor (>5-fold) of spa transcription. Northern blot analysis (Fig. 2C) confirmed those observations. Transcription profile analysis also revealed that, like agr mutant cells, spa transcription is significantly increased, across all growth phases, in the sarA mutant (Fig. 1D). This finding is in good agreement with previously published spa expression levels in sarA mutant cells (8).

agr-stimulated ORF identification. Given the level of correlation between the transcription profiling data obtained in this study and previously reported results for RNAII, RNAIII, hla, and spa expression, we felt comfortable searching for genes that were previously unrecognized as being regulated by the agr and/or sarA locus. To identify genes upregulated by a product of the agr locus, we took advantage of the observation that transcription of most identified agr-upregulated virulence genes, such as hla, increases with cell density in wild-type cells but does not increase in a density-dependent manner in agr mutant cells. Using GeneSpring software, we were able to compare transcript profiles of genes by analyzing RNA samples taken during various phases of growth from wild-type cells and agr (RN6911), sarA (ALC488), and agr sarA (ALC842) mutant cells. Genes demonstrating transcript patterns expected of those upregulated in an RNAIII-dependent manner were identified by querying for genes that were upregulated at least 2- and 1.5-fold as wild-type and ALC488 cells, respectively, moved from the early-log to the stationary phase of growth but were not upregulated 2-fold or were expressed at background levels in RN6911 and ALC842 cells. Alternatively, genes that were constitutively expressed in wild-type cells but were expressed at twofold-lower levels in both RN6911 and ALC842 cells were likely to be stimulated by a product of the agr locus. In each case the gene also had to be determined to be present in at least one wild-type stationary-phase sample, according to Affymetrix GeneChip software restrictions (Table 2). As expected, genes constituting the agr RNAII transcript (agrB, agrD, agrC, and agrA) produced the looked-for transcript profiles. Likewise, RNAIII (hld, encoding delta-hemolysin) appeared to be induced in an agr-dependent manner. Several known RNAIII-inducible genes were also identified, including splA, splB, splD, and splF (encoding serine protease) and hla (encoding alpha-toxin) (30, 31). Other extracellular virulence factors, such as hlgB and hlgC (encoding gamma-hemolysin) and geh (encoding lipase), were also identified.

Identification of agr-repressed ORFs. Table 3 lists 64 genes that produced transcript patterns expected of genes downregulated in an agr-dependent manner. These genes appeared to be transcribed at levels at least twofold higher in stationary-phase agr mutant cells than wild-type cells. Additionally, genes were considered present in at least one wild-type stationary-phase sample, according to GeneChip software restrictions. As expected, known RNAIII-downregulated genes, such as spa (34), were detected. Several additional cell wall-associated proteins, such as dlt (involved in teichoic acid linkage to cell wall and resistance to defensins), isaA (cell wall secretory protein), and mnhA, mnhF, and mnhG (constituting a heterologous surface receptor) are potentially repressed by agr.

Effects of SarA on agr-regulated genes. As shown in Fig. 1A and B, SarA is required for wild-type levels of agr transcription. More specifically, RNAII and RNAIII transcripts were decreased ~2.6-fold (±0.6-fold) in sarA mutant (ALC488) cells. Extending this observation, we identified potential agr-stimulated genes (Table 2) that demonstrate a similar fold reduction (within 2 standard deviations) in ALC488 cells. Genes meeting this criterion are presumed to be indirectly regulated by SarA and are indicated in Table 2. Likewise, genes suspected to be downregulated by agr (Table 3) that demonstrated a similar fold increase in transcript levels in ALC488 cells may be indirectly regulated by SarA, as indicated in Table 3. However, it should be kept in mind that RNAIII production has different regulatory effects on responsive genes (i.e., hla is dramatically upregulated in the presence of RNAIII, whereas expression of other genes, such as tst [encoding toxic shock syndrome toxin 1], is stimulated to a lesser extent) (26). Therefore, caution should be exercised in further interpreting these results. Yet comparison of genes determined to be directly regulated by SarA (described below) is in good agreement with our indirect SarA analysis.

Identification of SarA-upregulated genes. A series of transcription profile comparisons were performed to identify genes expected to be upregulated by SarA in a density-dependent manner. First, genes upregulated at least twofold as wild-type cells transition from early-log to stationary phase were identified. Additionally, transcripts had to be considered present within wild-type stationary-phase samples by Affymetrix GeneChip analysis. From this list, genes that were expressed at twofold-lower levels in stationary-phase sarA mutant cells than in wild-type cells were identified. We rationalized that this decrease in expression within the sarA mutant could most likely be due to either (i) a direct loss of SarA activator function or (ii) an indirect consequence of the sarA mutation, which decreases production of RNAIII and/or other transcriptional activators and reduces transcription of an agr-responsive gene. To distinguish between these two possibilities, transcription profiles (of genes fitting the criteria above) of agr and agr sarA mutant cells were compared. Genes directly activated by SarA could be identified as transcripts showing twofold reduction in stationary-phase agr sarA double-mutant cells relative to agr mutant cells. Several ORFs, including known SarA-activated genes, demonstrated background expression levels in agr mutant cells and were considered potentially activated by SarA (Table 4). Additionally, genes transcribed at twofold-higher levels in stationary-phase agr mutant cells are listed in Table 4. As expected our analysis identified several known SarA-upregulated virulence genes, including agrB, agrD, agrC, agrA, hld, and hla (6, 8, 13, 15). Additionally, other virulence factors, such as gamma-hemolysin, were determined to be upregulated by SarA. Potential SarA contributions, as defined by the above criteria, to the agr-regulated genes are listed in Tables 2 and 3.

Identification of SarA-downregulated genes. SarA-downregulated genes were identified as producing at least twofold-higher transcript levels in stationary-phase sarA mutant cells than in wild-type cells. Additionally, genes were considered present in at least one wild-type stationary-phase sample, as determined by GeneChip analysis. Genes with transcript levels meeting these criteria could include (i) genes that are directly repressed by SarA, in which case the absence of SarA allows for increased gene expression, or (ii) genes repressed by RNAIII, whereby decreased agr transcription allows derepression of gene expression. To distinguish between these two possibilities, transcript patterns of these genes were further compared in agr and agr sarA mutant cells. Genes directly repressed by SarA could be identified as producing elevated transcript levels in stationary-phase double-mutant cells compared to agr mutant cells (or at background levels). Table 5 lists genes that were determined to be downregulated by SarA in a cell density-dependent manner.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Using oligonucleotide microarray technology for genome-wide transcription profiling, we have identified S. aureus genes with expression patterns expected of genes either up- or downregulated in a cell-density agr- and/or SarA-dependent manner. Given that most known S. aureus virulence factors are temporally regulated by one, if not both, of these effectors, it is likely that the genes identified within this study include previously unrecognized virulence determinants.

Although investigators have successfully used transcription profiling technology to study biological processes in a number of other organisms, we initially set out to validate the methodology used by determining whether the results obtained with the S. aureus GeneChip correlated with our Northern blot data and previous reports. As expected, profile analysis of the effector of the agr response (RNAIII) demonstrated that transcript titers increased in a cell density-dependent manner within wild-type cells but were undetectable in agr mutant strains. Results also confirmed that RNAIII expression is diminished in a sarA mutant, further establishing that SarA is required for wild-type levels of RNAIII production or that the absence of SarA delays the onset of RNAIII expression. Likewise, the transcription profiles of alpha-toxin produced results that are consistent with other reports and confirmed that it is upregulated by each effector. Interestingly, profiling indicated that hla expression in RN6911 cells increased 2.2-fold in a growth phase-dependent manner. This increase was not detected by Northern analysis (Fig. 2B) but has also been reported by others (6), suggesting that the GeneChip technology used may be more reliable than Northern blot analysis. Profiling data for protein A also supported both previous reports, as well as our Northern blot data, and confirmed that genes that are downregulated by each effector can be detected in our system. The protein A transcription profiles indicate that the observed derepression of spa transcription occurs throughout all growth phases of agr and sarA mutant cells, implying that either RNAIII or SarA production may repress expression of an activator of protein A in wild-type cells. Indeed, it has recently been found that either agr or sarA mutant cells produce elevated amounts of SarS (also known as SarHI), an activator of spa transcription (11, 33). More specifically, in those studies it was shown that sarA cells produced more sarS transcripts than did agr mutant cells, which nicely correlates with the spa transcript patterns observed here. In fact, our transcription profiling results indicate that (i) sarS is constitutively expressed at low levels in wild-type cells, (ii) the gene is transcribed above wild-type levels in both RN6911 and ALC842 cells during early growth phases, but transcript levels decrease to wild-type amounts at later growth phases, and (iii) SarS is constitutively expressed at levels 10-fold higher than wild-type levels in sarA cells (data not shown). These results suggest that the mechanisms of sarS derepression differ for sarA and agr cells. This has subsequently been confirmed by additional transcription profiling studies (unpublished data). Results of the present study also demonstrated that the profiles of at least 10 additional genes correlated with the work of other groups, including splA, splB, splC, and splD (31); sarS (11); agrA, agrB, agrC, and agrD (RNAII [7, 13, 19, 27]); and fnbA (35). Collectively these results provide a strong indication that the transcription profiling procedure used correlates well with previous agr and/or sarA studies, and they verify the methodology.

In all, 104 genes were revealed to be upregulated in a cell density- and agr-dependent manner. Among these genes (Table 2) were 20 putative virulence determinants, including 14 genes involved in extracellular factor production. Because their expression corresponds directly with agr transcription, presumably these genes were induced directly in response to RNAIII. In contrast, 34 genes appeared to be downregulated in an agr-dependent manner (Table 3). Two of these genes, spa (protein A) and ORF1941 (cross-reactive antigen), encode putative cell surface virulence factors. One potential extracellular virulence factor, sspA (secretory antigen precursor), was found to be downregulated by agr.

There appears to be a trend in the expression of agr-regulated virulence factors. Collectively our results suggest that an agr determinant upregulates extracellular virulence factors but downregulates cell surface virulence factors. Although this has long been a common hypothesis among investigators, the present body of work provides an unprecedented corroboration of this proposal. An immediate question that remains unanswered is why cell surface virulence factors are expressed during early phases of growth whereas extracellular proteins are produced at higher cell densities.

Traditionally, it has been hypothesized that during early stages of host invasion staphylococci produce MSCRAMMs (microbial surface components recognizing adhesive matrix molecules) and other cell surface proteins, which promote attachment to both host cell and foreign (i.e., catheter) surfaces as well as avoidance of host defensive machinery. Once the organism has reached a critical threshold number, virulence factor expression switches to a more invasive system in which the organism produces extracellular proteins capable of degrading host cells, their products (such as antibacterial agents), or their infrastructure (such as phagocytes). This is an attractive yet, based on our SarA findings, potentially oversimplified hypothesis.

As shown in Tables 4 and 5, 76 genes were found to be upregulated in a cell density- and SarA-dependent manner. Among these genes were 16 putative virulence factors, including 10 extracellular and 6 cell surface virulence determinants. Several known SarA-upregulated genes, such as hla and fnbA, are included in this list (6, 12, 35). A total of 44 genes, including 5 extracellular and 3 cell surface virulence factors, were found to be downregulated in a SarA-dependent manner. These results suggest that SarA-regulated virulence genes do not necessarily follow the reported reciprocal expression patterns that have been established between agr-regulated cell surface and extracellular proteins. Several of the genes identified as being regulated by SarA have been found to harbor putative Sar boxes within their promoter regions, as indicated in each table.

Interestingly, Cheung and colleagues have previously demonstrated that fnbA is regulated by SarA (35). Although attempts were made in that study, no fnbB signal was detected in Northern blot analysis. Conversely, our profiling results suggest that both fnbA and fnbB are expressed in a SarA-dependent manner and indicate that GeneChip technology is more sensitive than Northern blot analysis, in that instance. Similarly, numerous attempts to study the transcription of components of both the arginine deiminase and histidine utilization operons (both of which were identified as potentially being regulated by agr in our transcription profiling studies) by Northern analysis did not demonstrate any signal, further illustrating the sensitivity of the GeneChip technology.

Many of the genes identified as being regulated by agr and/or SarA encode putative virulence factors, yet the majority do not. Although a number of other investigators have proposed that each effector is likely to regulate "nonvirulence genes," data substantiating this hypothesis are sparse. It is likely that because the GeneChip technology used appears to provide far more sensitivity in studying biological processes than conventional approaches, others have not identified many of the genes presented in the present study as being regulated by each effector. Among genes not previously known to be stimulated in an agr- and cell density-dependent manner were genes encoding the members (arcA, arcB, arcC, and arcD) and the activator (arcR) of the arginine deiminase pathway. Activation of this pathway allows Bacillus licheniformis to grow anaerobically in the presence of arginine (24), an ability that could be beneficial to S. aureus during pathogenesis. Additionally, utilization of arginine via the arginine deiminase pathway produces ammonia, which has been shown to protect bacteria from the deleterious effects associated with acidic environments (4). However, further in vivo studies are required to determine whether components of the arginine deiminase pathway contribute to S. aureus pathogenesis.

Members of the histidine utilization pathway (hutG, hutH, hutI, and hutU), which constitute a single transcript in Bacillus subtilis that is induced primarily by the presence of L-histidine, also appeared to be expressed in an agr- and cell density-dependent manner (14). It is difficult to resolve how transcription of this operon contributes directly to S. aureus pathogenesis. Given the number of genes upregulated by each effector that are involved in amino acid metabolism and transport pathways, it is tempting to speculate that each regulator not only mediates virulence factor production but may also poise the cell to scavenge for nutrients. Another alternative is that proteases that are directly produced in response to agr and/or SarA degrade secreted S. aureus factors and activate transport and metabolic processes; as a consequence, these processes may be indirectly inducible by each effector in laboratory cultures. Moreover, it is likely that unrecognized agr- and/or SarA-mediated processes that occur during mid-log-phase growth perturb the cell and affect biological processes, which in turn mask the relevance of biological processes determined to be regulated by each effector in the present body of work.

A number of genes identified as regulated by either agr or SarA or both have not been previously characterized. It is likely that some, if not many, of these genes contribute to S. aureus pathogenesis. Moreover, understanding the functions of these genes may further clarify the biological processes identified here as being stimulated by either effector. Finally, it is important to recognize that pathogenesis is not likely to be a static process; rather, the invading bacterium must cope with fluxes in its immediate environment, such as changes in pH, changes in nutrient levels, differences in cell densities, and interactions with host factors. Therefore, the work presented here should be considered a snapshot of gene expression and a cataloging of the network of genes that are expected to be regulated by each effector in a cell density-dependent manner. We are actively characterizing networks of genes that are regulated by other elements, such as host factors, in an effort to better understand the pathogenic processes of staphylococci.

Because this technology provides transcriptional profiles for each gene tiled onto a gene chip, yet analysis of each individual gene is beyond the scope of this report, we have elected to provide this data to the scientific community, so that others may analyze genes involved in other biological processes. The transcriptional profiling data obtained in these studies can be accessed at the Network on Antimicrobial Resistance in Staphylococcus aureus (NARSA) website (http://narsaweb.narsa.com).