agr RNAIII divergently regulates glucose-induced biofilm formation in clinical isolates of Staphylococcus aureus

Staphylococcus aureus biofilm formation is an important factor in the pathogenesis of central venous catheter-associated bacteraemia and infections related to the use of medical prostheses (Vuong et al., 2000). Despite this knowledge, the composition of S. aureus biofilms and the associated regulatory network for their production have not been fully clarified thus far. The ica locus, which encodes proteins involved in the biosynthesis of the polysaccharide intercellular adhesion or poly-N-acetylglucosamine (PIA/PNAG) is present in the genome of both S. aureus and Staphylococcus epidermidis. Deletion of ica in S. aureus strain ATCC 35556 has been shown to inhibit glucose-induced biofilm development on polystyrene surfaces, under static conditions (Cramton et al., 1999). However, ica-independent mechanisms of biofilm formation have also been reported, and recently it was shown that a SasG-expressing mutant of S. aureus strain SH1000 formed a glucose-induced biofilm on the surface of fibronectin- or fibrinogen-coated plates (Corrigan et al., 2007). Studies by O'Neill et al. (2007) indicated that glucose-induced biofilm formation in meticillin-resistant S. aureus (MRSA) was mainly ica-independent and probably mediated by protein(s). More recently, the same researchers showed that FnbpA and B were involved in glucose-induced biofilm development of meticillin-susceptible S. aureus (MSSA) and MRSA clinical isolates, under static and flow conditions (O'Neill et al., 2008).

SarA (a global transcriptional regulator of S. aureus) was found to be a positive regulator of both ica-dependent and independent biofilm formation (Valle et al., 2003; Beenken et al., 2003; Toledo-Arana et al., 2005; O'Neill et al., 2007). It was shown that mutation in arlRS in S. aureus strain 15981 led to an increased biofilm accumulation even when ica was deleted. The elimination of the major S. aureus quorum-sensing (QS) locus, agr, had no effect on ica-independent biofilm repressed by ArlRS. Indeed, no difference was observed when those studies were carried out under static or flow conditions (Toledo-Arana et al., 2005). Beenken and colleagues, studying 13 agr-null constructs of S. aureus, proposed that mutation in agr had only a modest impact on the biofilm development induced by both glucose and sodium chloride on the surface of fibronectin-precoated plates. Nevertheless, they also reported that the only exception was with S. aureus strain RN6390B, which regularly formed a biofilm only after agr deletion (Beenken et al., 2003). Similarly, it has been demonstrated by others that removing agr from S. aureus strains 15981 and V329 had no impact on biofilm formation in the presence of glucose, under flow conditions (Valle et al., 2003). In spite of this, Yarwood et al. (2004) showed that the biofilm modulation promoted by agr in S. aureus could diverge under different environmental conditions. In fact, other studies found biofilm was enhanced in the agr-null strain RN6911 and in the agrC-transposon mutant mut6, suggesting that agr downregulates biofilm formation induced by glucose on the surface of polystyrene plates (Vuong et al., 2000).

It is important to note that most of the studies cited here were carried out with a small number of constructs or with NCTC 8325-4 derivatives, including RN6911. More recently, using multiple agr mutants, 1 % glucose induction and microtitre plate assays for biofilm evaluation, it was concluded that agr elimination increased ica-independent biofilm formation in five constructs and had no significant effect on the 16 others (O'Neill et al., 2007). Using an in vivo model to study biofilm, Balaban et al. (2007) reported that either TRAP^– or agr-null constructs derived from strain 8325-4 were deficient in biofilm formation. Thus, these authors suggested that RIP (an RNAIII-inhibiting peptide) could be used as a therapeutic agent to prevent biofilm development. In contrast, Tsang et al. (2007) found that traP mutations in S. aureus UAMS-1 and USA 300 did not affect agr or biofilm formation induced by glucose on the surface of polystyrene, under static conditions.

Clearly, the role played by agr on S. aureus biofilms is at best variable, and remains to be better understood. To address this, we analysed 12 MRSA and MSSA agr-null constructs derived from contemporary clinical isolates to further investigate the influence of the RNAIII transcript (presence or absence) on biofilm formation/accumulation by S. aureus. In addition, we examined biofilm development in a variant of MRSA ST239-SCCmecIIIA, which displays natural sarA/agr attenuation.

Bacterial strain, plasmid and construction.
S. aureus strain RN450 was used for phage propagation (Kreiswirth et al., 1983). S. aureus strain RN6911(Δagr : : tetM) is a derivative of strain RN6390B (Nesin et al., 1990; Novick et al., 1993). The agr-null constructs were obtained by allele replacement of the entire agr locus. Mutation Δagr : : tetM was transduced from strain RN6911 into MSSA or MRSA clinical isolates using phage 80α (Novick, 1967). The presence of RNAIII transcript in all wild-types (WT) was confirmed by RT-PCR and Northern blotting. Experiments of RNAIII trans-complementation were carried out by introducing pRN6848 (pRN5548 : : rnaIII-cat) into the agr-null MRSA or MSSA, also through 80α phage transduction (Novick, 1967; Novick et al., 1993). The transductants were selected on antibiotic-supplemented plates (agr-null with 5 µg tetracycline ml^–1 and rnaIII-complemented mutants with 10 µg chloramphenicol ml^–1). The agr deletions and rnaIII complementations were confirmed by PCR, dot-blot and Northern-blot assays using an rnaIII-specific probe. The genetic backgrounds of the mutants and respective WT were checked by PFGE, as described below. The MRSA isolates studied belonged to hospital- or community-associated international MRSA lineages currently isolated on the American continent. Strains RN6911 and RN450, and pRN6848, were kindly provided by Richard Novick, Skirball Institute of Biomolecular Medicine, NY, USA. The WT isolates and the corresponding isogenic mutants are listed in Table 1. In addition, some isolates of MRSA lineage ST239-SCCmecIIIA (Brazilian epidemic clone; BEC) were also included in this study. All bacterial strains and constructs were stored in 12 % glycerol at –70 °C.

Table 1. Effect of agr on biofilm formation of MRSA and MSSA isolates

Molecular characterization.
The SCCmec typing of the MRSA clinical isolates was performed as described by Oliveira & de Lencastre (2002). PFGE of the SmaI-fragmented DNA was carried out as described by Teixeira et al. (1995). The criterion used for classifying the PFGE patterns was suggested by Tenover et al. (1995), and was based on comparisons with the PFGE patterns displayed by representatives of international MRSA lineages (Ribeiro et al., 2007). Experiments of multilocus sequence typing (MLST) were also performed and the sequence type (ST) was assigned with reference to the MLST database ().

PCR.
Specific primers for rnaIII (Novick et al., 1993), sarA (5'-ggcaaatgtatcgagcaagatg-3' and 5'-gtatcatctatcaaacttcacc-3') and 16S rrna (5'-aacgcattaagcactccgc-3' and 5'-gtgtgtagcccaaatcataa-3') were used for PCR amplifications and probe preparations in the experiments involving DNA–DNA or DNA–RNA hybridizations. DNA preparations were obtained by phenol extraction and used in standard PCRs (Sambrook et al., 1989). The annealing temperature for all PCR programmes was 55 °C. The primers for sarA and 16S rrna were designed based on conserved regions of the respective sequences deposited at GenBank. Multiple alignments, using CLUSTAL W (Thompson et al., 1994), were performed to check for in silico primer specificity.

RT-PCR.
Total RNA was obtained using the RNeasy Mini kit (Qiagen). Purified RNA preparations were quantified using GeneQuant RNA/DNA Calculator (GE Healthcare). Prior to performing RT-PCR, 1 µg of the RNA preparation was treated with 1 U amplification-grade DNase I as recommended for RT-PCR (Invitrogen). All DNase reactions were controlled by PCR in the absence of reverse transcriptase. RT-PCR (SuperScript III One-Step RT-PCR System with Platinum Taq DNA Polymerase; Invitrogen) was performed as recommended by the manufacturer using 1 µg of the template RNA. cDNA synthesis was carried out at 50 °C for 30 min, denatured at 94 °C for 2 min, and amplified in 40 cycles: 94 °C for 15 s (denature), 55 °C for 30 s (anneal) and 68 °C for 30 s (extend); followed by a final extension at 68 °C for 5 min (one-step). The rnaIII-specific primers used have been described previously (Novick et al., 1993).

Dot-blotting.
Genomic DNA of the agr-null mutants and of the rnaIII-complemented constructs was denatured by boiling and spotted onto the surface of a nylon membrane. The membrane (Hybond-N+; GE Healthcare Bio-Sciences) was fixed by baking, then hybridized with labelled rnaIII probe and detected using Alkphos Direct Labelling with CDP STAR (GE Healthcare), as recommended by the manufacturer.

Northern blotting.
Ten nanograms of the total RNA, prepared as above, were run in an electrophoresis system in a 1.2 % RNA-agarose gel, as recommended by the RNeasy kit manufacturer (Qiagen). The RNA was transferred to a nylon membrane (Hybond-N+, GE Healthcare) using vacuGene XL (GE Healthcare). Hybridization and detection of chemoluminescence signals were carried out using the kit Gene Images Alkphos Direct Labelling and Detection System with CDP STAR (GE Healthcare).

Biofilm.
Biofilm assays were performed in 96-well Nunclon microtitre inert polystyrene plates (Nunc), using trypticase soy broth (TSB) (Difco) supplemented with 1 % (w/v) glucose (TSB-1 % Glc) as described previously (Amaral et al., 2005). Briefly, bacteria were grown in TSB-1 % Glc, in a shaker (250 r.p.m.) at 37 °C for 18 h. Cultures were diluted 1 : 100 in TSB-1 % Glc and 200 µl was inoculated into each well. The microtitre plate was incubated at 37 °C for 20 h. Supernatants were removed from each well and biofilms were gently washed twice with 0.85 % NaCl, then dried and fixed at 65 °C for 1 h. Finally, the plates were stained with crystal violet (Gram-stain), gently washed twice, and the absorbance was read in a microplate reader at 570 nm (A_biofilm). In parallel, the A₅₇₀ of 1/100 cultures in TSB-1 % Glc (incubated in static conditions at 37 °C for 20 h) was also determined (A_growth). The biofilm unit (BU) was calculated using the following formula: BU=A_biofilm/A_growth. Based on the results obtained for the negative control Streptococcus pyogenes 75194 (BU=0.115), the isolates were classified as non-producers (BU≤0.230), weak producers (BU>0.230 and ≤0.460), moderate producers (BU>0.460 and ≤0.920) and strong producers (BU>0.920). Because the development of biofilm is subject to phase variation, tests were repeated eight times. At least two independent experiments were carried out for each test. The mean BU value was used for the statistical calculation.

In addition, to confirm the differences between biofilm phenotypes, as determined by BU values, confocal laser scanning microscopy (CLSM) was employed to record and contrast structural images of the biofilms formed for representatives of clinical isolates displaying different biofilm phenotypes and also for the agr mutants derived from isolates HC474 and NY19335. The biofilm assays were performed as above, but after being fixed, all cells within the biofilm were stained with 25 nM SYTO 9 DNA-intercalating stain (Invitrogen) for 15 min in the dark. The stain was gently removed and biofilm was visualized using a Zeiss LSM510 metalaser scanning confocal microscope. The microscope was inverted and configured with one laser (argon 458 nm/477 nm/488 nm/514 nm). Images were captured at random with a Plan-Neofluar 40x/0.6 Korr objective. Filters were set to a band pass of 500–530 nm.

Statistical calculations.
Student's t-test (unpaired data) was used to compare the mean BU values. The null hypothesis (H₀: µ=µ₀) was rejected at level α=0.05 (Dunn, 1964).

agr interference using conditioned supernatant.
It was shown that the effect of a conditioned supernatant (CS) and purified heterologous Agr autoinducing peptide (AIP) in inhibiting agr activation was identical for all agr types tested (Ji et al., 1997; Jarraud et al., 2000; Novick et al., 2000). Although we carried out RNAIII trans-complementation studies to check our genetic analysis, we also performed the same simple test as an additional control of the experimental system. Thus, 0.1 vol. CS (obtained from a stationary phase culture sterilized by filtration) was added to the bacterial inoculum (previously diluted 1 : 100) in TSB-1 % Glc. Then 200 µl of this mixture was placed in each well and assayed for biofilm formation using microtitre plates, as described above. CS type I (prepared using isolate BMB9393) was used to impair the Agr system of the agr type II isolates USA 100 and MSSA cultures. CS type II (prepared from isolate USA 100) was used to impair agr of the BEC isolates BMB9393 and GV69. In parallel, the experiment was performed exactly as described above but using filtered culture supernatant containing homologous AIP, for controlling possible effects of other bacterial products that might interfere in the system.

Haemolytic activity on blood agar plates.
The δ-haemolysin (Hld), encoded by the hld gene, is encoded within the rnaIII region (Janzon & Arvidson, 1990). Thus, preliminary, BEC isolates (ST239-SCCmecIIIA) were screened for haemolytic activity on sheep red blood agar plates. A fresh culture of strain RN4220 (grown at 37 °C for 18 h on trypticase soy agar; TSA) was stripped in the central region of the plate. Then test isolates (also grown on TSA) were inoculated perpendicular to RN4220, in each half of the plate. The plate was incubated at 37 °C for 18 h and examined for a haemolytic zone surrounding the bacterial growth. Because β- and δ-haemolysins lyse sheep red blood cells synergistically, the production of the δ-haemolysin frequently results in an arrow-tip-like zone where δ-haemolysin overlaps with β-haemolysin produced by RN4220 (Adhikari et al., 2007). Representatives of BEC isolates that produced very weak or undetectable haemolytic activity on blood plates were selected for Northern-blotting experiments with the RNAIII-specific probe.

DNA sequencing.
Purification of PCR products was carried out using the QIAquick PCR Purification kit (Qiagen), as recommended by the manufacturer. DNA sequencing was carried out by primer walking using a MegaBACE 1000 automatic sequencer (GE Healthcare). Sequences were multiply aligned with the most similar sequences available at GenBank; CLUSTAL W was used to align the sequences.

Bacterial constructs
The presence of RNAIII transcript in the WT isolates (seven MRSA and five MSSA) was confirmed by RT-PCR and Northern blotting. The agr locus was entirely deleted in these isolates by allele replacement. MRSA agr-null constructs (Δagr : : tetM) were obtained from representatives of USA 100 (five isolates), USA 200 (one isolate) and USA 400 (one isolate) lineages. agr elimination was confirmed by resistance to tetracycline, absence of PCR amplification using rnaIII-specific primers, and absence of chemiluminescence signals after dot hybridization with the rnaIII probe. Among these 12 agr-null constructs, four were amenable to complementation experiments by transferring pRN6848 (pRN5548 : : rnaIII-cat) using phage transduction. The rnaIII-complemented mutants were confirmed by concomitant tetracycline and chloramphenicol resistance, PCR amplification of the rnaIII-specific fragment and dot-blot hybridization using the rnaIII probe. PFGE experiments confirmed the relatedness of each set of WT, agr-null and -complemented mutants (data not shown). In addition, trans-complemented mutants were able to transcribe RNAIII as showed by the Northern-blotting experiments. All the isolates and constructs used in this study are listed in Table 1.

Variability in the biofilm phenotype among S. aureus clinical isolates
We have previously reported that variants within the same MRSA clone (ST239) can display different biofilm phenotypes (Amaral et al. 2005). Thus, we decided to evaluate biofilm formation/accumulation in the clinical isolates of S. aureus, in order to select representatives of each different phenotype for the experiments involving genetic manipulations.

Previous studies demonstrated that the in vitro, under static or flow conditions, and in vivo analysis of biofilm formation correlated quite well (Toledo-Arana et al. 2005; Beenken et al. 2004; O'Neill et al., 2008). Thus, because multiple isolates were tested here and based on the observation above, the semiquantitative method chosen for measuring biofilm formation was primarily carried out on polystyrene microtitre plates and validated by CLSM experiments. The isolates were classified as strong, moderate, weak or non-biofilm producers based on bacterial cell density as determined by spectrophotometry.

Using these criteria, three of the clinical isolates selected were classified as biofilm non-producers (MSSA: HC410, HC296; USA 100: NY19339) and nine as biofilm producers. Among the producers, four were classified as strong (MSSA: HC474, HC642; USA 100: NY17859; USA 400: WB81), with BU values varying from 1.01 to 2.81, and five as weak producers (USA 100: NY17896, NY22735, NY19335; MSSA: HC569; USA 200: W7749). The BEC isolates BMB9393 and GV69 included in this study were classified as strong and moderate biofilm producers, respectively (Table 1). Fig. 1 shows the three-dimensional topography of S. aureus biofilms and the topometry values obtained from CLSM experiments for representatives of different biofilm phenotypes. The strong biofilm producer BMB9393 (BU=2.78) formed a dense lawn 47.1 µm thick (Fig. 1a), the moderate biofilm producer GV69 (BU=0.90) displayed a more irregular film 26.6 µm thick (Fig. 1b) and the weak producer NY22735 (BU=0.38) developed a biofilm 18.4 µm thick (Fig. 1c). These data confirmed our previous findings (Amaral et al., 2005) that the ability of S. aureus clinical isolates to accumulate glucose-induced biofilm on polystyrene surface can vary significantly.

(28K): Fig. 1. Topographical images obtained by CLSM of different biofilm phenotypes expressed by S. aureus isolates, representative of two separate experiments. (a) Strong biofilm phenotype formed by isolate BMB9393 (47.1 µm thick). (b) Moderate biofilm phenotype formed by the isolate GV69 (26.6 µm thick). (c) Weak biofilm phenotype formed by the isolate NY22735 (18.4 µm thick).

Effect of agr on biofilm development
The fact that most previous studies addressing the effect of agr on S. aureus biofilm were conducted using a small number of well-known laboratory mutants motivated us to examine the impact of agr elimination on contemporary clinical isolates of MSSA/MRSA displaying different biofilm phenotypes. Given that ST239 isolates are resistant to most antimicrobial drugs, agr mutation Δagr : : tetM was carried out only in MRSA lineages that displayed tetracycline susceptibility (Tc^s).

The deletion of agr in all four Tc^s isolates classified as strong biofilm producers exerted a decrease in the biofilm density, varying from about two- to eightfold reduction when compared with WT (P<0.01 to P<0.001; Table 1). We succeeded in trans-complementing RNAIII in one of these agr-null constructs, the MHC474 mutant. The biofilm defect shown by MHC474 (BU=0.36) was reverted (BU=0.95) by the introduction of the complementing plasmid pRN6848 into this construct (Fig. 2a, panels 1–3 and Fig. 2b, wells 1–3). Fig. 2(c) shows the result of RNAIII transcription for the set of WT, agr-null and rnaIII-complemented constructs. These data suggest that RNAIII upregulates the strong biofilm accumulation developed by these isolates on the static surface of polystyrene plates.

(98K): Fig. 2. Divergent effect of RNAIII on S. aureus biofilms. (a) CLSM image reconstructions representative of three independent experiments. Biofilms were formed on inert polystyrene surface and treated with SYTO 9. The vertical (red) line indicates the location of the Y plane from which the cross-section was taken. The horizontal (green) line indicates the location of the X plane from which the cross-section was taken. The blue lines indicate the slice of the biofilm from which the XY image was taken. Panels: 1, HC474 (WT, MSSA); 2, MHC474 (Δagr : : tetM); 3, CMHC474 (Δagr : : tetM, pbla-rnaIII); 4, NY19335 (WT, ST5-SCCmecII); 5, MNY19335 (Δagr : : tetM); 6, CMNY19335 (Δagr : : tetM, pbla-rnaIII). (b) Biofilm formed on microtitre plates and stained with crystal violet. Wells: 1, HC474 (WT, MSSA); 2, MHC474 (Δagr : : tetM); 3, CMHC474 (Δagr : : tetM, pbla-rnaIII); 4, HC474 treated with 0.1 vol. conditioned supernatant (CS); 5, NY19335 (WT, ST5-SCCmecII); 6, MNY19335 (Δagr : : tetM); 7, CMNY19335 (Δagr : : tetM, pbla-rnaIII); 8, NY19335 treated with 0.1 vol. CS. (c) Northern-blot experiments using specific RNAIII probe. Lines: 1, HC474 (WT, MSSA); 2, MHC474 (Δagr : : tetM); 3, CMHC474 (Δagr : : tetM, pbla-rnaIII); 4, empty space; 5, NY19335 (WT, ST5-SCCmecII); 6, MNY19335 (Δagr : : tetM); 7, CMNY19335 (Δagr : : tetM; pbla-rnaIII); 8, empty space.

Consecutively, we examined the effect of agr elimination on the biofilm phenotype of eight weak or non-producer isolates (Table 1). For three of these isolates, the elimination of agr caused no significant alteration in the biofilm phenotype when compared with WT (USA 100: NY17896 and NY22735, weak producers; MSSA: HC410, biofilm non-producer). These results suggested that the inability of some S. aureus isolates to form or accumulate biofilm on polystyrene surface was not influenced by agr. Nevertheless, the Δagr : : tetM mutation in three isolates displaying a weak biofilm phenotype (MSSA: HC569; USA 100: NY19335; USA 200: W7749) had a positive impact on biofilm accumulation, corresponding to about two- to fourfold increase in the BU values (P<0.01 to P<0.001; Table 1). We were able to introduce pRN5548 : : rnaIII-cat into two of these agr-null mutants. As a result, the strong biofilm phenotype achieved by the construct MNY19335 (BU=1.28) dropped to the level of the isogenic WT (BU=0.33) after the RNAIII trans-complementation (BU=0.3; Fig. 2a, panels 4–6 and Fig. 2b, wells 5–7). Similarly, the trans-complementation of the MSSA mutant MHC569 (BU=0.69) caused a reduction in biofilm density (BU=0.42) closer to the levels of the corresponding WT (BU=0.35; Table 1). Thus, these data show that agr downregulated biofilm accumulation in three isolates displaying a weak biofilm phenotype. In addition, the removal of agr induced biofilm formation in two non-producer isolates (MSSA: HC296; USA 100: NY19339). The impairment of biofilm formation caused by the presence of the RNAIII transcript was confirmed by trans-complementation. The introduction of pRN5548 : : rnaIII-cat into MNY19339 restricted biofilm formation to the level of the WT (Table 1). These results indicate that RNAIII can block the development of biofilm in some isolates of S. aureus.

It is important to mention that the biofilm development of isolates within the same lineage (quite similar PFGE pattern, the same ST5 and SCCmecII) was found to be either up- or downregulated by agr, or even unaffected. For example, the biofilm accumulation of the USA 100-related isolate NY17859 (strong producer) was upregulated by agr, whereas RNAIII downregulated biofilm accumulation of the USA 100 isolate NY19335 (weak producer) and inhibited biofilm formation of the isolate NY19339 (normally a non-producer). However, RNAIII did not affect the biofilm phenotype of the USA 100-related isolates NY17896 and NY22735, both weak producers (Table 1).

Agr interference using conditioned supernatant
Previous studies have provided strong evidence that the only agr-activating component in the supernatant of S. aureus isolates that are Agr⁺ is related to the presence of AIP (Novick et al. 2000; Shaw et al. 2007). Thus, beside RNAIII complementation, the agr interference using conditioned medium (Ji et al., 1997) could provide a simple and straightforward experiment to quickly recheck the effect of agr in S. aureus biofilm formation on microtitre plates. In fact, the treatment with 0.1 vol. CS of four WT isolates displaying strong (HC474, BU=1.12) or weak (HC569, BU=0.35 and NY19335, BU=0.33) biofilm formation, and a biofilm non-producer (NY19339, BU=0.21) consistently reproduced the effect of agr deletion on biofilm development. Thus, the BU values obtained after agr knockout and from CS experiments were HC474, 0.36 and 0.39; HC569, 0.69 and 0.62; NY19335, 1.28 and 1.13; and NY19339, 0.61 and 0.73. Fig. 2(b) shows the data of the agr-interference experiments for two of these isolates (Fig. 2b, well 2, MHC474 and well 4, HC474+CS; well 6, MNY19335; and well 8, NY19335+CS). Taken together, the results from agr-null mutants, RNAIII trans-complementation and also from 0.1 vol. CS were all consistent with the participation of RNAIII regulation on biofilm formation/accumulation of S. aureus isolates.

Biofilm development of BEC isolates GV69 and BMB9393
We have found that some BEC isolates can display a natural agr impairment, as first revealed by screening of haemolytic activity on blood agar and confirmed by Northern-blot hybridizations (Fig. 3a, b). Thus, we selected the isolate GV69 (agr attenuated) to compare its ability to form/accumulate biofilm with that of BMB9393, which had stronger RNAIII signals in the Northern blotts (Fig. 4a, upper panel). Our results showed that BMB9393 developed a strong biofilm phenotype (BU=2.78; Fig. 3c, well 1), threefold more dense than that of GV69 isolate (BU=0.90; Fig. 3c, well 2). However, despite the fact that RNAIII expression was very low in GV69 compared with BMB9393 (Fig. 4), the biofilm developed by GV69 could still accumulate on a polystyrene surface (Fig. 1b). Because BEC isolates are tetracycline resistant (Tc^r), we tested the effect of 0.1 vol. CS in the biofilm developed by these two isolates. The CS experiment with the isolate GV69 resulted in further decrease in the BU value to 0.48 (Fig. 3c, wells 3 and 4). Similarly, the biofilm formed by BMB9393 significantly reduced to BU=0.78 after CS treatment (Fig. 3c, wells 5 and 6). Analogous to what was observed for the agr knockouts derived from strong biofilm producers, these data corroborate the evidence that the stronger biofilm accumulation by some isolates of S. aureus seems to be upregulated by agr.

(37K): Fig. 3. (a) Haemolytic activity on sheep blood agar. Vertical strip, β-haemolysin-producing strain RN4220; horizontal strips, BEC isolates. Note that agr naturally attenuated isolates displayed absence of an arrow-tip-like zone on the intersection with the RN4220 strip and weak haemolytic activity. (b) Northern-blot experiments showing lower RNAIII expression for BEC isolates GV12, GV20, GV24, GV69 and GV91. S. aureus isolates BMB109, BMB120 and RN6390B, displaying strong RNAIII signals, were used as controls. (c) Biofilm expression of the BEC isolates (ST239-SCCmecIIIA). Wells 1, BMB93932 and 2, agr naturally attenuated GV69. Wells 3, untreated BMB9393 and 4, BMB9393 treated with 0.1 vol. CS. Wells 5, untreated GV69 and 6, GV69 treated with 0.1 vol. CS. (d) Investigation of the importance of carbohydrates and of protein(s) in biofilm structural composition of BEC strains. Wells 1 and 2, untreated BMB9393; 3 and 4, BMB9393 after treatment with sodium metaperiodate; 5 and 6, untreated BMB9393; 7 and 8, BMB9393 after treatment with proteinase K.

(20K): Fig. 4. (a) Northern-blot experiments with BEC isolates BMB9393 and GV69 (ST239-SCCmecIIIA) using rnaIII, sarA and 16S-rrna specific probes. (b) Multiple alignment of the 3' region of the agr RNAII showing a point mutation in the BEC strain GV69 and BMB9393 (agr type I), RN6390B (agr type I), RN9107 and D22 (GenBank accessed sequences of the S. aureus isolates RN9107 and D22, both agr type I). Grey shading shows the point mutation, the replacement of a guanine by thymine. Asterisks represent the consensus sequence.

Previous studies have suggested that S. aureus biofilm under glucose induction is composed mainly of protein (O'Neill et al., 2007). In order to validate this observation, we treated the pre-formed biofilm of isolate BMB9393 with 100 µl 10 mM sodium metaperiodate (to oxidize carbohydrates) or proteinase K (1 mg ml^–1) for 2 h at 37 °C, as suggested by Chaignon et al. (2007). We also found metaperiodate treatment had only a moderate effect on the strong biofilm accumulated by this isolate (Fig. 3d, wells 1–4). However, proteinase K removed virtually all biofilm formed by BMB9393 (Fig. 3d, wells 5–8), confirming the importance of protein component(s) in this process.

agr and sarA sequencing of BEC isolates
In view of the difference in RNAIII expression between the BEC isolates (Fig. 4a), the whole agr locus of isolate GV69 was sequenced and compared with that of BMB9393 and RN6390B (also an agr type I isolate). Sequencing revealed that the agr-attenuated isolate GV69 had a unique point mutation in the sequence encoding the 3' end of RNAII, corresponding to a substitution of guanine for thymine (dbSNP build130: ss# 8610943). Despite the fact that this mutation was not found in the RN6390B sequence, the same alteration was detected in the isolate BMB9393 (Fig. 4b), which displayed a strong RNAIII signal. Thus, the alteration in RNAII could not be implicated in the agr attenuation detected in GV69 and might be a common characteristic of the lineage ST239. We can infer that the defect would be located upstream of agr.

Given that SarA is a positive regulator of agr, we speculated that the agr attenuation in GV69 could be due to a mutation in sarA. We carried out sarA Northern blotting and verified, as expected, that sarA was also attenuated in this BEC isolate but not in BMB9393 (Fig. 4a, middle panel). The lower expression of sarA mRNA may therefore explain the agr attenuation detected. Next, we decided to sequence the regions encompassing all sar promoters (P1, P2 and P3). Alignments of the sequences from isolate GV69 (GenBank accession number EU301433) with the equivalent segments from S. aureus sar sequences at GenBank did not show any variation in these regions (data not shown). These data indicated that the responsible defect for agr/sarA attenuation is likely to be upstream of sar.

The accessory gene regulator agr is one of the main global virulence regulators in S. aureus. However, the effective role played by agr in biofilms has not been fully elucidated. Knowing how agr can affect S. aureus biofilms is critical to our understanding of the dynamics involved in the formation, accumulation and detachment of these bacterial films. To further investigate whether the agr RNAIII has a function in biofilm formation/accumulation in clinical isolates of S. aureus, we analysed agr-null mutants derived from contemporary isolates of both MSSA and MRSA. The experiments were designed to have the mutation Δagr : : tetM transferred into isolates representative of different biofilm phenotypes. Essentially, the results obtained from the majority of agr knockouts indicated that agr could either promote or inhibit glucose-induced biofilm on polystyrene inert surfaces, under static conditions. Only with a few isolates did agr deletion not affect biofilm formation or accumulation (3/12; 25 %). Thus, we found that agr modulation could diverge from up- to downregulation in S. aureus isolates displaying different biofilm phenotypes. Yarwood et al. (2004) showed that under some conditions disruption of agr expression could have no detectable influence on S. aureus biofilms, while under others it could either inhibit or enhance biofilm formation.

Here we showed that the phenotype of four strong biofilm producers changed to weak (three isolates) and to moderate (one isolate) when agr was removed. We could restore the biofilm defect by reintroducing rnaIII into one of these knockouts and reproduce the biofilm phenotype of the agr-null mutant by interfering with the agr function using conditioned supernatant (0.1 vol. CS). Therefore, it is reasonable to conclude that agr exerted a positive regulation on biofilm accumulation in these strong biofilm-forming MSSA/MRSA isolates.

Conversely, agr downregulated the biofilm expression in three of five isolates classified as weak biofilm producers. We also found that two isolates classified as non-producers had the ability to form biofilm after agr removal (downregulation). RNAIII-complementation experiments for three of these isolates regenerated the biofilm phenotype to the levels of the corresponding WT. Here again, the results of the interfering experiments with CS reproduced the biofilm impact observed in the agr knockouts. These data are in agreement with previous studies showing agr had a negative impact on biofilm development by S. aureus (Vuong et al., 2000). However, a small impact of agr on biofilm formation was suggested by others, using different substrates for biofilm development. In contrast, these researchers found that sarA deletion caused a significant decrease in the S. aureus biofilm (Beenken et al., 2003; Valle et al., 2003). In fact, there is growing evidence for a role of sarA in the upregulation of biofilm formation by S. aureus (O'Neill et al., 2007).

The contradictory findings for the role of agr on biofilm formation have been attributed to the fact that many studies have used strain NCTC 8325-4 or derivatives, which also have mutations in the sigB operon and in agrA (Kullik et al., 1998; Adhikari et al., 2007). In addition, biofilm formation by 8325-4 varies according to the laboratory source of this strain (Valle et al., 2003). To get around this, O'Neill et al. (2007) also used a number of clinical isolates and found that agr mutations significantly increased biofilm (more than twofold) in 5 of 21 (23 %) clinical isolates but had no significant impact on biofilm formation in the remaining 16.

Biofilm development seems to be a multifactorial process (Otto, 2008). It is well known that agr regulation can target different bacterial products, downregulating surface adhesins and upregulating secreted proteins (Novick et al., 1993). As such, it seems likely that a variable composition of S. aureus biofilms would account for the observed differences in the biofilm phenotypes of the WT and also for the apparent discrepancy in the effect of agr on biofilms. In fact, O'Neill et al. (2008) demonstrated that FnbpAB are involved in glucose-induced biofilm accumulation of S. aureus. Despite the fact that RNAIII downregulates Fnbps (Saravia-Otten et al., 1997), the results of O'Neill and colleagues could not fully explain the role of agr on biofilms since agr deletion did not have a significant impact on 16 of 21 agr-null mutants tested (O'Neill et al., 2007). Studies by Corrigan et al. (2007) showed that the expression of a homologue of the accumulation-associated protein (AAP) SasG by the laboratory strain SH1000 could also promote biofilm formation.

Moreover, not only agr-downregulated proteins but also the positively regulated ones have been implicated in S. aureus biofilm formation and accumulation. Recent studies by Johnson et al. (2008) showed that Eae and Emb secreted-adhesion proteins (both positively regulated by agr) played an important role in iron-regulated biofilm formation in S. aureus. Indeed, it was found that different environmental conditions other than iron could also affect the expression of eae and emb genes. The impairment of the hla gene significantly inhibited the adherence of strain NCTC 8325-4 to plastic surfaces under both static and flow conditions (Caiazza & O'Toole, 2003). Indeed, it was previously suggested that the elimination of the surfactant properties of the RNAIII-encoded δ-toxin was involved in the enhancement of the biofilm formed by the agr-transposon mutant mut6 (Vuong et al. 2000).

More recently it was suggested that glucose-induced biofilm was mainly mediated by protein(s) (O'Neill et al., 2007). Corroborating this hypothesis, significant impairment in biofilm formation was found in a knockout for sortase, which anchors LPXTG proteins, many of them negatively regulated by agr (O'Neill et al., 2008). Our results also demonstrate that a protease could totally disrupt the strong biofilm formed by the isolate ST239. Thus, it is possible that differences in protease expression could also be implicated in the variability of biofilm formation/accumulation. Genes encoding extracellular proteases are generally repressed by SarA and positively regulated by RNAIII (Boles & Horswill, 2008). It was found that a clpP protease mutant formed more biofilm compared with WT cells (Frees et al., 2004). In addition, experiments demonstrated that a double mutant aur splABCDEF (encoding the proteases aureolysin and Spl serine proteases, respectively) had minimal protease activity, enhanced biofilm formation and attenuated detachment phenotypes (Boles & Horswill, 2008). Yarwood et al. (2004) have already suggested that agr could influence biofilm dispersion. Taking our data together with the studies cited above, it can be concluded that the process of biofilm development in S. aureus is especially complex, of multifactorial nature and regulated by a coordinate network of regulatory systems including Agr, SarA and bacterial proteases.

Although it is difficult to directly transfer knowledge from experimental laboratory data to what occurs during clinical infections, the results presented here suggest that more detailed studies are needed to clearly evaluate the potential of agr inhibitors, such as RIP (Balaban et al., 2007), as strategy for the development of S. aureus anti-biofilm agents. It would be of interest for future studies to include S. aureus clinical isolates displaying different biofilm phenotypes to test the effect of agr-inhibiting peptides using in vivo models, given that our in vitro studies showed agr knockout could enhance weak biofilms and allow biofilm formation of some non-producer isolates. Our findings that Agr can downregulate biofilm formation in some S. aureus isolates was supported by others (Vuong et al., 2000; O'Neill et al., 2007).

MRSA isolates of ST239 are widespread in Brazil and in many other countries (Teixeira et al., 1995). One of the most striking properties of this lineage is its superior ability to adhere to and invade human airway cells and to form an enhanced biofilm on polystyrene inert surfaces (Amaral et al., 2005). Some ST239 variants are agr/sarA naturally attenuated. We showed that these attenuations were paralleled by more than twofold reduction of the ability to form glucose-induced biofilm. RNAIII attenuation is not a rare phenomenon in S. aureus clinical isolates (Papakyriacou et al., 2000; Yarwood et al., 2007). Although our results indicated that sarA attenuation affected RNAIII expression in the GV69 isolate, we could not detect any alteration in the sequence regions of sar promoters. Preliminary studies using transcriptional profiles indicated that the sarA repressor, sarR, was upregulated in the GV69 stationary phase but not in the BMB9393 (A. M. S. Figueiredo et al., unpublished data). The occurrence of naturally agr-attenuated variants in S. aureus biofilms was recently reported by Yarwood et al. (2007). These studies revealed that sarU (a gene also involved in agr activation) was repressed in the non-haemolytic biofilm variant. All these data together implicate the Sar family of regulators as potential regulatory paths for the agr impairment of some MRSA clinical isolates. It is well recognized that agr and also sarA, as well as other gene regulators, coordinate the global expression of bacterial virulence factors (Novick et al., 2000). Thus, it is logical to suppose that variations in sarA/agr expression among clinical isolates might represent an important mechanism of host adaptability and may contribute to the bacterial fitness in a specific epidemiological scenario.

In conclusion, agr elimination had considerable impact on glucose-induced biofilm formation/accumulation for the majority of clinical isolates tested. This effect could be reversed by trans-complementation of RNAIII into agr knockouts. The effect exerted by agr could be neutral, or up- or downregulation. Further insights into biofilm regulatory pathways involving agr would require the determination of the key molecule(s) involved in biofilm expression in S. aureus. It remains to be clarified if the differences observed in the biofilm phenotypes are due to a multi-component structure of the S. aureus biofilms and/or the modulation of the biofilm-associated gene(s).

This work was supported in part by Conselho Nacional de Desenvolvimento Científico e Tecnológico-CNPq, Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro-FAPERJ. L. R. C., R. R. S. and F. A. F. were supported by fellowships from CNPq. We thank Andréa Cheble de Oliveira and André Marco de Oliveira Gomes from the Microscopia Funcional Multifotônica Unit, Universidade Federal do Rio de Janeiro, for their great help in the CLSM experiments.

Edited by: T. J. Mitchell

Footnotes

^†, Raquel Rodrigues Souza †These authors contributed equally to this work.