The agr P2 operon in Staphylococcus aureus codes for the elements of a density-sensing cassette made up of a typical two-component signalling system and its corresponding inducer. It is postulated that the autoinducer, a post-translationally modified octapeptide generated from the AgrD peptide, interacts with a receptor protein, coded by agrC, to transmit a signal via AgrA regulating expression of staphylococcal virulence genes through expression of agr RNA III. We show by analysis of PhoA fusions that AgrC is a transmembrane protein, and confirm using Western blotting that a 46 kDa protein corresponding to AgrC is present in the bacterial membrane. This protein is autophosphorylated on a histidine residue only in response to supernatants from an agr+ strain, and can also respond to the purified native octapeptide. A recombinant fusion protein where most of the N-terminal region of AgrC is replaced by the Escherichia coli maltose-binding protein is also autophosphorylated in response to stimulation by agr+ supernatants or purified octapeptide. We conclude that AgrC is the sensor molecule of a typical two-component signal system in S. aureus, and that the ligand-binding site of AgrC is probably located in the third extracellular loop of the protein.
Microbial pathogenicity in Staphylococcus aureus is a complex phenomenon involving a number of virulence factors, primarily exotoxins that damage host cells and interfere with immune responses, and cell wall-associated proteins involved in adhesion and protection against host defences (Janzon and Arvidson, 1990; Kornblum et al., 1990). The expression of most of these factors is orchestrated by at least one regulatory locus, the accessory gene regulator agr (Janzon and Arvidson, 1990; Kornblum et al., 1990; Novick et al., 1993), which comprises two divergent operons (Peng et al., 1988; Kornblum et al., 1990). One of these forms a density-sensing cassette similar to the comAP operon of Bacillus subtilis (Magnuson et al., 1994). It contains four open reading frames (ORFs) of which two, agrA and agrC, might encode a two-component sensory transduction system, whereas the other two, agrB andagrD, are needed for producing the activation signal for AgrC. Expression of agrABCD appears to have the sole result of increasing transcription both from their own promoter, P2, in an autocrine fashion and from the divergent promoter P3 of the agr locus, which results in the production of a 517 nt transcript called RNA III (Kornblum et al., 1990; Novick et al., 1995). The untranslated RNA III is the actual effector of the agr system, and downregulates expression of exponential-phase proteins such as protein A and coagulase while upregulating expression of post-exponential proteins such as enterotoxins B, C and D, toxic shock syndrome toxin-1, haemolysins, proteases and leucocidins (Janzon and Arvidson, 1990; Kornblum et al., 1990; Novick et al., 1993).
AgrC possesses several key amino acid motifs typical of well-characterized histidine-protein kinase sensors and is therefore the potential sensor of the two-component system encoded within the agr−P2 operon. The autoinducing signal is a modified octapeptide, derived from the agrD gene product, probably through modification by the agrB product (Ji et al., 1995; 1997). The modification consists of a thioester bond between an internal cysteine and the C-terminal carboxy group (Ji et al., 1997; T. Muir, unpublished). The linear peptide (produced synthetically) is without activity (Ji et al., 1995). As agr-null cells of S. aureus containing a cloned copy of agrC titrated out the signalling octapeptide, whereas the same agr-null strain containing other agr components did not, AgrC has been thought to be the target of the octapeptide (Ji et al., 1995; 1997). AgrA, which is also required for agr activation and corresponds to the response regulator, does not appear to bind directly to the agr promoters (Morfeldt et al., 1996a) but may interact with SarA, the product of a separate genetic locus, to mediate activation (Morfeldt et al., 1996b).
In this paper we show that the agrC locus encodes a 46 kDa protein that is present in the bacterial membrane, as shown by PhoA fusions and by Western blotting with specific antisera. This molecule responds specifically to stimulation with natural supernatants containing the AgrD-derived octapeptide or with the purified octapeptide, by autophosphorylation on a residue whose chemical susceptibility is consistent with histidine. These results confirm that AgrC is the sensor for the proposed two-component signalling system in S. aureus.
Demonstration of the agrC gene product
Although predicted by sequence analysis and by its biological activity, AgrC protein has not thus far been directly detected. Accordingly, we constructed a translational fusion between the E. coli maltose-binding protein (MBP) and the C-terminal part of AgrC from residue 173 to the stop codon, purified this product from E. coli lysates using an amylose column and used it to prepare rabbit antiserum. A Western immunoblot using this antiserum is shown in Fig. 1. Total-cell extracts from the E. coli TB1 strain producing the fusion protein showed a single reactive band with an estimated size of 66 kDa (Fig. 1, lane A), as expected from the sequence of the expression vector. Total-cell extracts from S. aureus RN6390B showed a reactive band at 46 kDa (lane B), which is consistent with the size of the AgrC peptide predicted from the DNA sequence. This band was never observed, even after 10-fold concentration, in the cytoplasmic (soluble) fractions or supernatants of RN6390B (Fig. 1, lanes D and F), or in either cytoplasmic or membrane (insoluble) extracts from RN6911, the agr-null strain of S. aureus (Fig. 1 lanes G and H). Both cytoplasmic and membrane fractions from agr+ strains of S. aureus showed a second reactive band with an apparent size of 49 kDa that was also present in the agr− strain. This was proved to be due to non-specific recognition of the rabbit IgG by protein A, because it was more intense in the agr− strains (in which protein A is constitutively overexpressed), and absent in extracts from an agr+, protein A-null strain (Fig. 1, lanes I and J). The observation of a significant quantity of protein A in the RN6390B soluble fraction is probably due to contamination by material released from the cell wall by lysostaphin during the cell breakage and fractionation procedures. Migration of the 46 kDa peptide was not altered by the addition of 6 M urea, suggesting that conformation does not unduly influence our estimate of size (not shown).
Transmembrane nature of AgrC
Sequence analysis plus previous results suggesting that AgrC could titrate the agr-activating peptide (Ji et al., 1995), suggested that AgrC is a transmembrane histidine protein kinase serving as the receptor for a two-component signal transduction pathway that activates the agr system. A Kyte–Doolittle (Kyte and Doolittle, 1982) hydrophobicity plot for the RN6390B prototype and several naturally occurring variants of AgrC (Ji et al., 1997) is shown in 2Fig. 2A. The hydrophobicity analysis suggests that there are six or seven potential transmembrane helices in the N-terminal domain (Kornblum et al., 1989). The predicted transmembrane topology of AgrC was partially confirmed by construction of AgrC–PhoA fusions at various points in the protein. PCR products with different primers were cloned into a phoA fusion vector to give AgrC–PhoA fusion points at AgrC amino acid positions 33 (AP1/AP2), 105 (AP1/AP3), 142 (AP1/AP4) and 176 (AP1/AP5). E. coli cells containing cloned AP1/AP2, AP1/AP3 and AP1/AP5 were alkaline phosphatase positive; AP1/AP4 was negative, suggesting that AgrC amino acid residues 33, 105 and 176 are outside and 142 is inside the inner membrane. After osmotic shock of E. coli containing the cloned AP1/AP2 fragment, the AgrC–PhoA fusion protein was released, suggesting that the first predicted transmembrane segment of AgrC is actually outside and therefore so is the N-terminus. The status of helices II and III has not been resolved. The hydrophobic region corresponding to helix II contains only 12 amino acid residues, which would not be sufficient to span the membrane, and a computer analysis (Geourjon and Deleage, 1994; 1995) of the structure in this region predicts that hydrophobic regions II and III comprise a single helix. The predicted transmembrane topology of AgrC derived from a combination of these results with computer analysis of AgrC hydrophobicity is shown in 2Fig. 2B.
Autophosphorylation of AgrC
The capacity of this 46 kDa membrane protein to autophosphorylate upon autocrine stimulation (Balaban and Novick, 1995; Ji et al., 1995) was tested by incubating membrane fractions from agr+ or agr− strains of S. aureus with early post-exponential culture supernatants from the same two strains in the presence of radiolabelled ATP. Incorporation of radioactivity occurred into the 46 kDa band only when the membrane fraction from an agr+ strain was incubated with supernatant from an agr+ strain in the presence of [γ-32P]-ATP (Fig. 3, lane A). In contrast, membrane fractions from agr− strains never incorporated radioactivity at the corresponding size (Fig. 3, lanes G and H), nor did membrane fractions from the agr+ strain when incubated with supernatant from the agr− strain (Fig. 3, lane C). When [α-32P]-ATP instead of [γ-32P]-ATP was used in the reaction mixtures labelling was completely abrogated (Fig. 3, lane B). To confirm that the phosphorylated protein observed in Fig. 3 lane D corresponds to AgrC, the 32P-labelled proteins obtained by incubation of agr+ membrane fraction plus agr+ supernatant were immunoprecipitated by an AgrC antiserum and then separated using SDS–PAGE as shown in Fig. 3, lane J; the labelled protein corresponds to the 46 kDa AgrC. Moreover, preincubation of the membrane fraction with AgrC-specific antibody completely abrogated autophophorylation of the 46 kDa band (Fig. 3, lane L).
S. aureus autoinducing octapeptide purified from RN6390B supernatants induced phosphorlylation of the 46 kDa protein in membrane extracts from agr+ strain, indistinguishable from that produced by incubation with unfractionated agr+ supernatants (Fig. 4, lane A), whereas an unmodified linear synthetic octapeptide was inactive (Fig. 4, lane D). The phosphorylation induced by the autoinducing octapeptide was not affected by addition of brain–heart medium or supernatant from an agr− strain (not shown).
The recombinant fusion protein between MBP–AgrC(Ser-173), purified by affinity chromatography, was also found to be capable of autophosphorylation in vitro when incubated with [γ-32P]-ATP in the presence of supernatants from the agr+ strain, or in the presence of purified autoinducing octapeptide but not in presence of a synthetic preparation of the unmodified linear octapeptide nor in presence of supernatants from the agr− strain (Fig. 4, lanes F–I). In contrast to MBP–AgrC(Ser-173), the purified MBP–AgrC(Ser-199) (which lacks the whole of the predicted extracellular domain) was spontaneously autophosphorylated in vitro when incubated with [γ-32P]-ATP independently of the presence or absence of purified S. aureus octapeptide (Fig. 5 lanes K–M). Again, no labelling occurred in the presence of [α-32P]-ATP (Fig. 4, lanes E and J).
The transferred phosphate on the 46 kDa molecule was found to resist boiling at neutral pH or overnight incubation with NaOH at pH 10 but to be labile to boiling with 16% trichloroacetic acid (Fig. 5). This suggests N phosphorylation of a histidine, or less probably a lysine, residue.
The mechanisms influencing expression of the important regulatory system controlling production of staphylococcal exoproteins, the agr locus, are not precisely known. It has been postulated, based on amino acid sequence homologies, that the four ORFs in the agr P2 operon encode a density-sensing regulatory unit or cassette consisting of a two-component signal transduction system and an autoinducing ligand (Novick et al., 1995). By this reasoning, the agrC reading frame would be expected to code for the sensor component of the regulatory system. The predicted protein could include hydrophobic regions suitable for membrane insertion, and a conserved asparagine-containing sequence (centred on Asn-337) preceded, some 100 aminoacyl residues earlier, by a histidine (His-237) typical of autocatalytic sensor histidine kinases (Stock et al., 1989). We have expressed a major part of the S. aureus AgrC peptide as a fusion with maltose-binding protein in E. coli, allowing easy purification and the production of a specific antiserum. This serum reacted with a 46 kDa membrane protein that is present only in agr+ strains of S. aureus and that is thus identified as AgrC. The predicted transmembrane topology of AgrC was partially confirmed by PhoA fusions. The peptide was undetectable in the cytosol fraction, suggesting that AgrC is co-translationally transported to the membrane, perhaps with the help of a chaperonin, as are many other membrane components.
The prediction that AgrC can respond to specific stimulation by autophosphorylation on a histidine residue is also supported by our experiments. A 46 kDa band, co-localizing with the immunoreactive AgrC band and identified as AgrC by immunoprecipitation, acquires radiolabelled phosphate from [γ-32P]-ATP but not [α-32P]-ATP only in the presence of supernatants from strains capable of producing the autoinducing octapeptide or the purified octapeptide. This conclusion was confirmed by the experiment showing that MBP–AgrC(Ser-173) fusion peptide was also autophosphorylated in response to agr+ but not agr− supernatants and in response to the purified octapeptide. A chemical transfer of phosphate rather than passive adsorption of ATP is strongly suggested by retention of the label on SDS gels and by the absence of radioactive labelling when [α-32P]-ATP rather than [γ-32P]-ATP was used. The phosphate is probably attached to a histidine residue because the phosphoprotein was stable to SDS, heat and alkali treatment, but rapidly destroyed by acid treatment. This is typical of N-phosophorylated histidine, or lysine, but not of other phosphorylated amino acids (Duclos et al., 1991). The specificity of the in vitro reaction in response to the purified octapeptide was suggested by the absence of inhibition by brain–heart medium, agr− supernatants or non-immune serum, and confirmed by the absence of autophosphorylation of MBP–AgrC(Ser-173) in response to a linear synthetic version of the octapeptide and by the inhibition of phosphorylation when AgrC was preincubated with a specific immune serum. The mechanism of this inhibition has not been examined. In contrast to MBP–AgrC(Ser-173), MBP–AgrC(Ser-199) was spontaneously autophosphorylated in the absence of any signal as observed with several other histidine protein kinases when their input domain is deleted (Parkinson, 1992).
Taken together, these results suggest that the ligand binding site in AgrC is located in the 16–18 aminoacyl residues of the final predicted extracellular loop of the AgrC protein. This site cannot be in the C-terminal half of the protein as MBP–AgrC(Ser-199) is spontaneously autophosphorylated and this domain is highly conserved among strains producing radically different AgrD peptides (Ji et al., 1997) that do not cross-activate. We therefore speculate that in the absence of the activating signal (octapeptide) the final extracellular loop of AgrC prevents the spontaneous activity of histidine kinase (Parkinson, 1992). This model must be considered tentative until a possible involvement in the input function of the remaining part of the N-terminal domain, and possible effects of the MBP moiety of MBP–AgrC(Ser-173) have been tested.
The region N-terminal to the putative octapeptide-binding site contains five predicted transmembrane helices, of which we have PhoA fusion evidence for three, and the presence of large quantities of the fusion protein in cytoplasmic extracts of the E. coli used for production, suggests that the N-terminal sequences of AgrC are involved in membrane localization. Several other known bacterial signal receptors with apparent transmembrane topology similar to that of AgrC have been described in Gram-positive bacteria including Clostridium perfringens VirS (Lyristis et al., 1994), Bacillus subtilis ComP (Weinrauch et al., 1990), Streptococcus pneumoniae ComD (Morrison et al., 1996) and signal receptors for bacteriocin production by various lactobacilli (Pestova et al., 1996). These receptors are all activated by autoinducing peptides [with the possible exception of C. perfringens VirS, which is activated by a soluble factor of unknown nature (Imagawa and Higashi, 1992)]. The possibility of a common functional organization for bacterial signal receptors with peptide ligands is intriguing.
We show that the 46 kDa membrane protein, AgrC acts as a sensor for the autocrine signal provided by the autoinducing octapeptide derived from AgrD, and responds by autophosphorylation on a histidine residue. This autophosphorylation reaction may be used to evaluate the capacity of agents to influence agr expression. The central regulatory role of agr in the expression of diverse factors associated with virulence in S. aureus may suggest targets within the agr expression system for novel therapeutic interventions.
S. aureus RN6390B was used as the standard agr+ strain, and RN6911 was used as the corresponding agr-null strain. RN6911 was derived from RN6390B by replacement of the agr locus with tetM (Novick et al., 1993). Autoinducing octapeptide was purified from supernatants of RN7668 (pRN6911), a derivative of RN6911 transformed with a staphylococcal plasmid (pRN6911), which contains agrB-D cloned from RN6390B, transcribed under the control of blaZ promoter (Ji et al., 1995). S. aureus A90159, an agr+ strain spontaneously defective for protein A, was used as control for Western blotting. Staphylococci were grown in brain–heart broth. DNA was cloned and recombinant proteins expressed in either Escherichia coli TB1 (New England BioLabs) or MC106 (Manoil and Beckwith, 1985) grown in LB broth (Difco laboratories). Strains were stored at −80°C until used.
Staphylococcal autoinducing peptide
Supernatants of 6 h cultures of S. aureus RN7668 (pRN6911) were collected, and octapeptide was purified from 30 ml samples using HPLC through a C18 column (Applied Biosystems) as previously described (Ji et al., 1995), then stored lyophilized until used. A linear peptide with the same amino acid sequence as the native octapeptide was synthesized commercially (Yale University, New Haven, CT, USA).
Recombinant AgrC peptide and antibody production
A fragment of 762 bp was amplified using PCR from the agrC locus of RN6390B using primers AGRC5 (5′-GGAATTCTCGGATGAAGCTAAAG-3′) and AGRC3 (5′-CGGGATCCTTAATGGCTAGTTG-3′), and ligated into the maltose-binding protein (MBP) site in the pMAL-p2 expression vector (New England Biolabs). The cloned fragment extends from Ser-173 to the stop codon of the agrC ORF, and corresponds to the C-terminal portion of the putative input domain together with the whole kinase-like region (Stock et al., 1989; Kornblum et al., 1990).
Another fragment of 705 bp was amplified using PCR from agrC locus of RN6390B using primer agr5-bis (5′-TCTGAATTCCTCCTTAAAGAGATGAAATAT-3′) and AGRC3, and also ligated into pMAL-p2. The cloned fragment which extends from Ser-199 to the stop codon of the agrC ORF lacks the whole of the proposed extracellular domain (Stock et al., 1989; Kornblum et al., 1990). The integrity of the ORFs of these constructions was verified by DNA sequencing of the recombinant plasmids (respectively pLUG131 and pLUG323). The fusion proteins MBP–AgrC(Ser-173) and MBP–AgrC(Ser-199) were purified from supernatants of transfected E. coli TB1 by affinity chromatography on amylose columns acording to the supplier's instructions (New England Biolabs).
New Zealand rabbits supplied by ESD were immunized by an initial intramuscular injection of 100 μg of purified MBP–AgrC(Ser-173) protein in Freund's complete adjuvant (Sigma) followed by two intradermal booster injections of 100 μg of protein in Freund's incomplete adjuvant (Sigma) at 3 weekly intervals. The rabbits were bled 3 weeks after the last injection, and sera were stored at −20°C.
Construction of AgrC–PhoA fusions
To construct translational fusions of alkaline phosphatase (PhoA) at various points within AgrC, PCR products were prepared using a forward primer for 5′agrC, 5′-CGGGGATCCGTGGAATTATTAAATAG-3′ (AP1) paired with four different downstream reverse primers, 5′-CGGGGATCCTTACTGTACTTAATACCAC-3′ (AP2), 5′-CGGGGATCCATTTTGGTGATATATGC-3′ (AP3), 5′-CGGGATCCTATGGTGTGCTAAT-
TTTT-3′ (AP4) and 5′-CCCTGATCATTAGCTTCATCCGAG-3′ (AP5) and plasmid pRN6852 DNA as template (Ji et al., 1995). A restriction enzyme cutting site was included in each primer (underlined). The PCR products were digested with either BamHI (AP1/AP2, AP1/AP3 and AP1/AP4) or BamHI and BclI (AP1/AP5) and cloned into the BglII site of pAWLP-2 (a phoA fusion vector kindly provided by Andrew Wright, Tufts University) respectively. The resulting plasmids were transformed into E. coli MC1061. The junction between agrC and phoA of each construct was verified by DNA sequencing. Alkaline phosphatase activity was assayed on LB agar + X-P as described (Manoil and Beckwith, 1985).
Osmotic shock of E. coli MC1061 containing cloned AP1/AP2 fragment in pAWLP-2 was performed according to the procedure of Weiner and Heppel (1971).
Secondary structure prediction were made using the self optimized prediction method from alignments (Geourjon and Deleage, 1994; 1995) (http://www.ibcp.fr/predict.html).
Cultures of agr+ or agr-null staphylococci were grown to early post-exponential phase (c. 6 h) in brain–heart infusion broth at 37°C with vigorous shaking (200 rpm), then harvested by centrifugation at 3000 × g for 10 min (4°C). Supernatants were filtered through 0.2 μm membranes (Millipore) and pellets were washed three times with cold distilled water; both were stored separately at −20°C. Cells from 50 ml cultures were resuspended in 2 ml of lysis buffer (50 mM Tris-HCl pH 7.6, 50 mM EDTA, 200 μg ml−1 lysozyme and 30 μg ml−1 lysostaphin) for 30 min at 37°C. After ultrasonic disruption for 3 min in 30 s pulses at 4°C in the presence of 15 μg ml−1 of protease inhibitors (PI set, Boehringer Mannheim), the lysed cells were ultracentrifuged at 200 000 × g for 2 h at 4°C. The supernatant (cytoplasmic fraction) and the pellet (membrane fraction) (Heumann et al., 1994), which was washed five times with cold distilled water, were stored separately at −20°C.
Membrane or cytoplasmic fraction samples were boiled for 5 min in 1× SDS buffer (125 mM Tris-HCl, 4.4% SDS, 20% glycerol, 2% β-mercaptoethanol; pH 6.8) or a buffer of the same composition plus 6 M urea and clarified by centrifugation. Proteins were separated by SDS–PAGE using a miniProtean II aparatus (Bio-Rad), then electrophoretically transferred to nitrocellulose membranes (BAS-85; Schleicher & Schuell) and blocked by immersion in TTBS (100 mM Tris-HCl, 0.9% NaCl, 0.1% Tween 20; pH 7.5) for 30 min at 20°C. Antigens were detected by incubation for 30 min with a 1:10 000 dilution of rabbit antiserum in TTBS, followed, after two washes in TTBS, by incubation with anti-rabbit IgG peroxidase conjugate (Sigma) diluted 1:3000 in TTBS for a further 30 min. After two washes in complete TTBS and one wash in TBS (TTBS lacking Tween 20), the retained peroxidase activity was revealed by incubation in 5-bromo-4-chloro-3-indolyl phosphate/nitroblue tetrazolium solution as indicated by the supplier (Sigma).
Membrane fraction from 50 ml cultures of agr+ or agr-null S. aureus were vigorously resuspended in 4 ml of distilled water by shaking and repeated ejection from a 1 ml pipette cone to open vesicles and allow access to both the extracellular (ligand binding) and intracellular (APT binding) domains. Incorporation of radiophosphorus from [γ-32P]-ATP (sp. act. 1 μCi mM−1; 10 μCi in 10 μl) was evaluated in the presence of filtered supernatants from agr+ or agr− strains diluted 1:10 in distilled water (50 μl), purified native octapeptide or linear synthetic octapeptide, by incubation of 20 μl of either fresh membrane suspension, recombinant MBP–AgrC(Ser-173) (0.5 μg ml−1) or MBP-AgrC(Ser-199) (0.5 μg ml−1) with an equal volume of phosphorylation buffer (250 mM Tris-HCl, 250 mM KCl, 0.5 mM EDTA, 50 mM MgCl2, pH 7) for 5 min. Control samples were similarly incubated with [α-32P]-ATP of the same specific activity instead of [γ-32P]-ATP. Incorporation was stopped by the addition of 50 μl of 2× SDS buffer, and 20 μl samples were evaluated by autoradiography after SDS–PAGE. Transfers from the same gel to nitrocellulose membrane were tested by Western blotting using specific antibody to AgrC as described above. The specificity of the in vitro reaction was examined by preincubating membrane fractions for 30 min at 37°C with either antiserum against AgrC (1:100), control serum (1:100) from the same animal before immunization, brain–heart medium or agr− supernatants, before phosphorylation experiments. Reactions were investigated using immunoprecipitation. Reactions mixture (100 μl) was incubated at 4°C for 10 min in lysis buffer (20 mM Tris-HCl pH 8, 150 mM NaCl, 5 mM EDTA, 10% glycerol, 1% Triton X-100 and 15 μg ml−1 of protease inhibitors). After 10 min centrifugation at 3000 × g, supernatants were incubated for 2 h at 4°C with protein A-sepharose coated with antiserum against AgrC (1:100), or pre-immunisation control serum (1:100). Pellets were washed four times with lysis buffer. The complexes were dissociated by boiling in 1× SDS buffer for 10 min. Labelled proteins were separated from protein A-sepharose by 1 min centrifugation at 200 g and analysed by autoradiography after SDS–PAGE.
Characterization of the protein–phosphate bond
The 32P-labelled protein was either boiled for 10 min before SDS–PAGE and autoradiography or the radioactive band was excised from the gel, treated either by boiling in 16% trichloroacetic acid for 10 min or by incubation in 100 mM NaOH (pH 10) overnight, then blotted dry and evaluated by autoradiography (Duclos et al., 1991).
We are grateful to A. J. Cozzone, J. Deutscher and A. Galinier for scientific advice.