|Home | About | Journals | Submit | Contact Us | Français|
Cell density-dependent gene regulation by quorum-sensing systems has a crucial function in bacterial physiology and pathogenesis. We demonstrate here that the Staphylococcus aureus agr quorum-sensing regulon is divided into (1) control of metabolism and PSM cytolysin genes, which occurs independently of the small regulatory RNA, RNAIII, and (2) RNAIII-dependent control of additional virulence genes. Remarkably, PSM expression was regulated by direct binding of the AgrA response regulator. Our findings suggest that quorum-sensing regulation of PSMs was established before wide-ranging control of virulence was added to the agr regulon, which likely occurred by development of the RNAIII-encoding region around the gene encoding the PSM, δ-toxin. Moreover, the agr regulon in the community-associated methicillin-resistant S. aureus MW2 considerably differed from that previously determined using laboratory strains. By establishing a novel, two-level model of quorum-sensing target gene regulation in S. aureus, our study gives important insight into the evolution of virulence control in this leading human pathogen.
Many bacteria undergo substantial physiological rearrangements with increasing population density (Camilli and Bassler, 2006). The corresponding gene regulatory changes are controlled by quorum-sensing systems. In pathogenic bacteria, these systems have first been described as virulence regulators (Passador et al., 1993; Recsei et al., 1986), whereas it has been shown more recently that quorum-sensing control is also aimed to respond to changing environmental conditions via metabolic adaptations (Cassat et al., 2006; Dunman et al., 2001; Schuster et al., 2003; Yao et al., 2006).
S. aureus is a major human pathogen and among the most common causes of bacterial infections in the community and the hospital. Many strains are resistant to a wide spectrum of antibiotics including methicillin (methicillin-resistant S. aureus, MRSA), making therapeutic intervention often extremely difficult. Moreover, recent outbreaks of MRSA in the community form a novel major challenge for the public health system (Diep and Otto, 2008). This situation has sparked considerable interest in the analysis of staphylococcal molecular biology with the goal to find novel drug targets. Particularly, global regulators of virulence such as quorum-sensing regulators have been studied intensively, based on the notions that interfering with such systems would eliminate production of many virulence factors at a time (Alksne and Projan, 2000) and drugs interfering with virulence would not lead to resistance at the same speed as drugs that kill bacteria (Clatworthy et al., 2007).
The agr (accessory gene regulator) system represents a prototype of quorum-sensing regulators in Gram-positive bacteria (Kleerebezem et al., 1997). It has been recognized early as a pivotal regulator of virulence factor expression (Recsei et al., 1986) and a potential target for therapeutic intervention (Otto, 2004). The agr system consists of two adjacent transcriptional units, RNAII and RNAIII, which are transcribed in opposite directions (Figure 1). RNAII comprises the messages for the AgrD precursor of the extracellular peptide-based quorum-sensing signal and its maturation and export protein AgrB. Furthermore, it encodes a classical bacterial two-component signal transduction system, composed of the sensor histidine kinase AgrC and its cognate response regulator AgrA. Upon binding of the quorum-sensing signal to AgrC, AgrA is activated and binds to the promoter regions in the agr system (P2 for RNAII and P3 for RNAIII). This positive feedback loop, typical of quorum-sensing systems, ascertains that target gene expression is altered swiftly at a certain threshold level of bacterial cell density (Novick, 2003).
According to current knowledge, modulation of target gene expression by agr is achieved via RNAIII (Novick et al., 1993), which represents a paradigmatic example of virulence factor control via a small regulatory RNA (sRNA) (Romby et al., 2006) and the central role of sRNAs in quorum-sensing (Toledo-Arana et al., 2007). Recent findings indicate that RNAIII generally acts by an antisense base pairing mechanism (Boisset et al., 2007), and regulates many target genes via control of a repressor protein gene called rot, a member of the sarA family of transcriptional regulators (Boisset et al., 2007; McNamara et al., 2000; Said-Salim et al., 2003).
The recently discovered phenol-soluble modulins (PSMs) play a key role in S. aureus immune evasion owing to their considerable leukocidal activity, and significantly impact the outcome of infections by S. aureus, especially community-associated (CA-) MRSA (Wang et al., 2007). There are 7 PSM peptide genes in S. aureus, 4 forming the psmα operon encoding the PSMα peptides PSMα1 through PSMα4, 2 forming the psmβ operon encoding the PSMβ peptides PSMβ1 and PSMβ2, and the hld gene, which encodes the δ-toxin and is embedded within RNAIII (Wang et al., 2007) (Figure 1). Most likely, δ-toxin is not involved in the RNAIII-dependent regulation of agr target genes, although this has been a matter of debate (Janzon and Arvidson, 1990; Novick et al., 1995). While the psm genes are present in all S. aureus strains whose genome has been sequenced, their expression may differ significantly (Wang et al., 2007). Among the PSM peptides, the PSMα peptides have the most pronounced chemotactic, pro-inflammatory, and leukolytic activity (Wang et al., 2007). PSM peptides are secreted by a yet unknown mechanism as the direct translational product with the N-terminal formyl-methionine modification (Wang et al., 2007), which is typical for bacteria. However, due to the activity of bacterial N-formyl-methionine deformylase, a certain percentage of PSMs are secreted in a N-deformylated form, dependent on media and growth phase (Somerville et al., 2003). It is not known how N-terminal formylation impacts biological activity of the PSM peptides.
Intriguingly, production of PSM peptides is under exceptionally strict control by agr (Wang et al., 2007), but the molecular details of PSM regulation are unknown. Moreover, it is not understood why the hld gene encoding the PSMδ-toxin is embedded within RNAIII. Therefore, here we re-evaluated the role of RNAIII within the agr regulatory system, with a focus on PSM regulation. By discovering RNAIII-independent regulation of agr target genes, our findings establish a novel mechanism of target gene control by quorum-sensing in S. aureus and give insight into the evolution of quorum-sensing systems with regard to the connection of metabolism and virulence gene regulation.
Our knowledge of agr target genes is predominantly based on studies with the laboratory strain RN6390, which has a mutation in the rsbU global regulatory gene. The limited predictive value of those studies for clinically important strains has recently become apparent (Blevins et al., 2002). Thus, we chose the S. aureus MW2 strain for the current study, because it represents a community isolate with major clinical significance and has an intact rsbU gene (Baba et al., 2002; CDC, 1999). This strain is the prototype CA-MRSA strain and was reported to cause fatal community-associated infections in children (CDC, 1999). In accordance with its success as a pathogen in the community, it has revealed hyper-virulence in animal infection models (Voyich et al., 2005). Most importantly for the current study, strain MW2 has an active agr system (Wang et al., 2007). Here, to reevaluate the mechanism of target gene regulation by agr, we first determined the agr regulon of strain MW2. To that end, we performed genome-wide transcriptional profiling analyses with microarrays specific for the MW2 strain (Li et al., 2007), comparing gene expression in the MW2 wild-type and isogenic agr deletion strains during post-exponential growth phase (at maximal expression of agr).
In accordance with the classical notion of agr-dependent gene regulation (Novick, 2003), the microarray results demonstrated that agr in strain MW2 up-regulates a series of exotoxin genes, such as those coding for several leukocidins including the Panton-Valentine leukocidin (PVL), enterotoxins, and α-toxin. Furthermore, genes encoding degradative exoenzymes (serine and cysteine proteases, lipase), and the methicillin resistance gene mecA, are under positive regulation of agr (Figure 1, Supplemental Table 1). The most strongly down-regulated transcript was that coding for the immune evasion molecule protein A. Interestingly, other down-regulated agr target genes in MW2 mostly code for metabolic enzymes, possibly indicating that the up-regulation of toxins occurs at the price of reduced metabolic activity (Figure 1, Supplemental Table 1).
We were particularly interested in agr regulation of psm genes. Because the newly identified psmα genes are too small to meet the threshold length previously used for gene annotation in S. aureus genomes, including that of MW2 (Baba et al., 2002), the MW2 chip that we used only contains probes for the larger psmβ, but not the smaller psmα genes. Thus, we used quantitative real-time PCR (qRT-PCR) to measure expression of the psmα operon, in addition to confirmatory analysis of the psmβ operon (Figure 2A). Together, the microarray and qRT-PCR data showed that the psmα and psmβ operons are under very strong positive regulation by agr.
The agr regulon of the CA-MRSA clinical isolate MW2 revealed several discrepancies to findings achieved using the laboratory strain RN6390. First, the current model of agr regulation comprises down-regulation of surface binding proteins by agr. It is believed that they are specifically needed during the establishment of an infection, while increased agr activity down-regulates their expression later during infection (Novick, 2003). In contrast, these proteins were not regulated by agr in strain MW2, except for protein A, which does not have a role in host matrix protein binding but in immune evasion (Forsgren and Nordstrom, 1974; Gomez et al., 2004). Second, the SarA paralog Rot has been attributed a central role in the regulation of agr target genes (Boisset et al., 2007; McNamara et al., 2000; Said-Salim et al., 2003). However, the only sarA paralog to be regulated by agr in strain MW2 was sarH1, also termed sarS gene. It is tempting to speculate that sarH1 might functionally substitute for rot in strain MW2.
To gain a better understanding about the role of RNAIII in the agr regulon, we produced an allelic replacement mutant of RNAIII and compared it to the isogenic MW2 wild-type and agr deletion strains using genome-wide transcriptional profiling with the MW2 microarray and qRT-PCR analysis (Figures 1, ,2A,2A, Supplemental Table 1). Remarkably, we found that many genes were regulated by agr independently of RNAIII. This group comprised many agr-down-regulated genes involved in carbohydrate and amino acid metabolism, and biosynthesis of the yellow pigment staphyloxanthin (Figure 1, Supplemental Table 1). qRT-PCR confirmed this for the most strongly down-regulated hexose phosphate transporter gene MW0197 (Figure 2A).
Most strikingly, genes up-regulated by agr independently of RNAIII almost exclusively comprised the psmα and psmβ genes (Figures 1, ,2A,2A, Supplemental Table 1). Differential expression values of only 2 additional genes in this group were low (Supplemental Table 1). Additionally, there was differential expression of a gene locus, apparently forming an operon, composed of the MW0370/MW0372 genes, which code for proteins of unknown function present only in the MW2, MSSA476, and RF122 strains among the 13 sequenced S. aureus genomes. Interestingly, the qRT-PCR data revealed RNAIII-dependent down-regulation of the psmα, psmβ, and MW00370/0372 operons (Figure 2A), which was opposite to the overall strongly positive effect of agr on PSM production. However, RNAIII-independent up-regulation clearly overrides this additional RNAIII-dependent down-regulation of psm expression. Of note, these findings establish a previously unknown level of agr target regulation that is exclusively dependent on genes encoded in the RNAII transcript encoding the agrA, agrC, agrD, and agrB genes.
To analyze the impact of RNAIII on PSM production at the translational level, we performed reversed-phase high-pressure liquid chromatography/electrospray mass spectrometry (RP-HPLC/ESI-MS) of culture filtrates. There was no detectable PSMα or PSMβ expression in the agr mutant strain, in contrast to considerable PSMα and PSMβ quantities in the wild-type and, importantly, the RNAIII deletion strain (Figure 2B). As mentioned above, the δ-toxin is not produced in the agr or RNAIII deletion strains, because the hld gene is embedded within RNAIII and thus deleted in both these strains. Slight differences that we observed in PSM levels between the wild-type and RNAIII deletion strains were not consistent between the PSMs of the same operon and may be explained by post-translational effects, and our observation that δ-toxin is needed for the co-elution of more hydrophobic PSM peptides. Together, these results showed on the transcriptional and translational level that RNAIII has no considerable impact on PSM expression that would explain the dramatic overall influence of agr on PSM production.
We hypothesized that direct interaction of AgrA with the psmα and psmβ promoters is the most likely mechanism to account for RNAIII-independent control and the exceptionally strong influence of agr on the expression of the psmα and psmβ genes. To evaluate this hypothesis, we used 1) electrophoretic mobility shift assays (EMSAs) with natural and mutated AgrA consensus binding sequences, 2) DNase footprinting analyses, 3) expression studies with vector-expressed psm genes including promoter fragments of different lengths, and 4) measurement of PSM production with constitutive or inducible expression of vector-expressed agrA versus RNAIII genes in agr mutant strains. For experiments 1 and 2, we purified the AgrA protein of strain MW2 using expression in E. coli (Supplemental Figure 1).
Sequence similarity to the agr P2 and P3 promoters (Koenig et al., 2004) indicated the presence of 3 possible AgrA binding sites in both the psmα and psmβ promoter regions (Figure 3A). Using EMSAs with specific probes, we detected AgrA binding to the psmα and psmβ promoter regions and determined the binding region to encompass 2 of the 3 putative binding sites in each promoter (fragments αS2 and βS1, Figures 3A,B). We observed multiple shifts very similar to those previously shown for the P2/P3 promoters (Koenig et al., 2004). In comparison, absence of shifts with the very similar fragments αS1 and βS2 (Figures 3A,B) indicated specificity of AgrA binding. Sequences similar to the binding sites in the psmα and psmβ promoter regions were found upstream of the MW0370/MW0372 locus (data not shown), suggesting that this locus is also under direct control by AgrA.
To confirm specific binding of AgrA, we exchanged one base in both of the two determined binding sequences of the psmα or psmβ promoters (Figure 3A). Binding of AgrA was reduced by a factor of ~ 2.5 to 5 for the psmα and ~ 1.5 to 2 for the psmβ promoter (Figure 3C). Binding to the psmβ promoter fragment was overall stronger, which may explain the smaller effect of the introduced 1 basepair changes compared to the psmα promoter. Additionally, EMSAs with psmα promoter mutated binding sites showed a low shift that was also observed by Koenig et al. when acetyl phosphate was omitted (Koenig et al., 2004), most likely indicative of AgrA binding as an inactive monomer. Moreover, to determine the exact AgrA binding location, we performed DNase footprinting analyses, which pinpointed the AgrA binding sites to regions overlaying (psmβ promoter) or immediately adjacent to (psmα promoter) the consensus sites derived from the comparison with the P2/P3 promoters and the EMSA experiments (Figure 3A, Supplemental Figure 2). These experiments substantiated that AgrA binds specifically to the psmα and psmβ promoters.
To further confirm direct control of psm gene expression by AgrA on a translational level, we focused on the psmα genes as those having the greatest biological importance (Wang et al., 2007). We transformed a psmα deletion strain of MW2 with psmα expression vectors containing different lengths of the psmα promoter region (Fig. 3A). As expected, we observed PSMα production only with the constructs containing the region to which we have demonstrated AgrA binding (Figure 4A). Interestingly, our results also indicated that the psmα genes are under minor negative control of additional regulators that act on the intergenic region upstream of the AgrA binding site. These findings are in accordance with those from our previous study indicating that while agr appears to be the major regulator controlling PSM levels, PSM expression in various S. aureus strains is not entirely correlated with agr expression and might thus be subject to additional regulation other than by agr (Wang et al., 2007).
To substantiate that AgrA rather than RNAIII impacts PSM expression, we cloned the agrA gene or RNAIII-encoding region in plasmids with inducible or constitutive control of expression, and transformed the agr-negative strain S. aureus 601055 mut6 (Vuong et al., 2000) with these plasmids. PSM expression was only achieved when agrA, but not RNAIII, was induced or constitutively expressed (Figure 4B). Values for the δ-toxin and PSMα3 peptides are shown as examples, while the remaining PSM peptide values changed accordingly (not shown). These results confirmed that psm genes are under control of AgrA, but not RNAIII. By using a different strain background, these experiments also ruled out that the regulatory mechanism that we describe here is strain-specific.
While AgrA binding to the psm promoter regions explained RNAIII-independent up-regulation, the mechanism by which down-regulation by agr in an RNAIII-independent manner occurs remained unclear. However, no reasonable number of consensus sequences was found in the S. aureus MW2 genome, possibly indicating that yet unidentified, additional structures are important for AgrA binding, or the respective genes are not regulated by direct binding of AgrA.
As an alternative explanation, we hypothesized that the RNAIII-independent down-regulation of agr target genes may occur indirectly via a regulatory effect of the PSMs or the MW0370/MW0372 proteins. In fact, gene regulatory effects of S. aureus toxins (Vojtov et al., 2002) and contrasting reports on a potential regulatory role of δ-toxin have been published (Janzon and Arvidson, 1990; Novick et al., 1995). However, comparative transcriptional profiling with MW2 and its isogenic psm (Wang et al., 2007) and MW0370/MW0372 gene deletion strains clearly ruled out a regulatory function of those proteins, notably including δ-toxin (GenBank accession number GSE10165).
Quorum-sensing regulation in staphylococci is assumed to be accomplished exclusively by the agr system. An alleged role of the TRAP protein in staphylococcal quorum-sensing was found to be due to an experimental artifact (Adhikari et al., 2007); and a quorum-sensing role of the ubiquitous luxS system in staphylococci is being debated, but unlikely (Doherty et al., 2006; Li et al., 2008). Here we show that there are two distinct subsets of agr target gene regulation: an RNAIII-independent that is primarily in charge of controlling metabolic genes, and an RNAIII-dependent controlling virulence (Figure 1). Additionally, we demonstrate that agr-dependent regulation of the PSM gene family is achieved by direct binding of the AgrA response regulator protein.
Our data indicate that control of virulence factors by RNAIII and quorum-sensing control of metabolism were once separate regulatory circuits. The connection of RNAIII-dependent with quorum-sensing regulation likely evolved to allow for a quick cell density-dependent change of virulence factor expression during infection, because regulation by an RNA rather than a protein is believed to lead to a shorter response time (Bejerano-Sagie and Xavier, 2007). The inclusion of the PSM toxins– in contrast to other virulence determinants – in the primary core set of agr-regulated genes is most likely due to their presumed original function in the non-infectious lifestyle of S. aureus, i.e. in biofilm development (Kong et al., 2006). While it is difficult to dissect which regulatory circuit existed first, the architecture of the system indicates secondary connection of virulence factor control to quorum-sensing regulation. Formation of the RNAIII-encoding region around the hld gene coding for δ-toxin represents a much more likely event than change of a part of RNAIII to contain a structural gene and simultaneous addition of an AgrA-dependent promoter. Of note, this model is in accordance with the fact that most staphylococcal species have no or only limited pathogenicity and the evolution of virulent species such as S. aureus is believed to be a phylogenetically more recent event.
In addition to giving insight into the mechanism and evolution of target gene regulation by agr, our study for the first time determined the agr regulon of a CA-MRSA strain. The molecular basis of the exceptional virulence of CA-MRSA strains is poorly understood, but believed to contribute significantly to the CA-MRSA pandemic (Diep and Otto, 2008). Notably, the only two factors demonstrated until now to have a major influence on the virulence of CA-MRSA in infection models, α-toxin and α-type PSMs (Wang et al., 2007; Wardenburg et al., 2007), are both regulated by agr (Recsei et al., 1986; Wang et al., 2007). These findings suggest that agr plays a key role in determining the virulence potential of CA-MRSA. In this regard, it is important to stress that our data indicate very strict control of PSM, protease, and overall toxin expression by agr in CA-MRSA. Additionally, we show agr-dependent regulation of the Panton-Valentine leukocidin, which has been suggested to play a role in CA-MRSA infection based on epidemiological data (Gillet et al., 2002), but for which a clear role in CA-MRSA infection models could not be established (Voyich et al., 2006; Wardenburg et al., 2007). Furthermore, in contrast to previous reports on agr regulation, agr targets in the CA-MRSA strain MW2 did not comprise surface binding proteins and the SarA paralog Rot. Additional studies will be needed to analyze whether these features are characteristic for clinically important, especially CA-MRSA strains.
In conclusion, our study reveals a new level of quorum-sensing target gene control in the major human pathogen S. aureus, and exemplifies how gene control by an sRNA may be intertwined during evolution with another global regulatory circuit, adding virulence factor control to a quorum-sensing regulon originally in charge of primarily basic cell physiology.
Determination of PSM production in S. aureus culture filtrates was performed using RP-HPLC/ESI-MS essentially as described (Wang et al., 2007). Samples were taken from cultures inoculated from pre-cultures and grown for 16 h. PSMs accumulate in the culture filtrate after agr expression starts in post-exponential growth phase. Afterwards, PSM concentration in culture filtrates is relatively stable over several hours during stationary growth phase (Wang et al., 2007). PSM concentrations were determined by computing extracted ion chromatograms (EICs) of the most abundant m/z peak for every peptide, and integration of the EIC peaks at the elution times corresponding to the respective peptide, using Agilent QuantAnalysis Software (see Supplemental Figure 4). Values obtained from the peaks representing the N-formylated and the N-deformylated forms of a specific PSM were added.
The AgrA expression clone was constructed by PCR amplification of the agrA gene from chromosomal DNA of strain MW2 and cloning into the expression vector pET9a (New England Biolabs). Purification of AgrA was performed as described (Koenig et al., 2004), except that SOURCE Q material (GE Healthcare) was used for the first chromatographic step and a Resource S 1 ml column (GE Healthcare) for the second, with gradient elution over 20 column volumes. Obtained AgrA was analyzed by SDS-PAGE (Supplemental Figure 1), N-terminal sequencing, and analytical RP-HPLC/ESI-MS, and found to be correct and pure.
EMSAs were performed as previously described (Koenig et al., 2004) with the following modifications. Final DNA concentrations were 25 pM. Binding reactions were performed in 20 μl, incubated at room temperature for 30 min and run on an 8% non-denaturing 0.5 × TBE gel. Acetyl phosphate was added to a final concentration of 50 mM.
The authors thank Larye Parkins for genome pattern searches, Frank Gherardini and Julie Boylan for helpful advice, and Frank DeLeo for critically reading the manuscript. This study was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, NIH.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.