We first constructed an hfq mutant in the S. aureus avirulent strain RN4220, in which the region corresponding to the hfq ORF was replaced by a cassette conferring Cm resistance (see Methods and Fig. 1A). Gene inactivation was verified by Southern blot (data not shown) and the hfq mutation was then transferred by phage transduction into three virulent strains that differ in their virulence regulation pathways, RN6390, Newman and COL. Gene transfer was confirmed by PCR and RT-PCR (Fig. 1B).

In E. coli, Hfq is one of the most abundant intracellular proteins, whose level in exponential growth phase is reportedly around 30,000 to 60,000 molecules per cell [24]. Hfq levels further increase during the transition to stationary phase [25]. These data suggest that Hfq likely plays a crucial role in controlling expression of numerous growth-related genes. In view of the described properties of Hfq, we asked whether hfq expression levels in S. aureus were as high as in E. coli. We first followed hfq expression during bacterial growth in BHI medium, by northern blot experiments, using an hfq-specific fragment as probe. No transcript corresponding to hfq mRNA was detected in RN6390, Newman or COL WT backgrounds (from total RNAs extracted from both exponential and stationary phase cultures; data not shown). In these experiments, an E. coli hfq strain containing a plasmid carrying the S. aureus hfq gene under control of a tetracycline inducible promoter system was used as positive control of hfq mRNA expression. We then compared hfq expression using RT-PCR, which is a more sensitive detection method than northern blotting. Steady state levels obtained from RN6390 WT and hfq post-exponential RNA extracts are represented on Figure 1B. A band corresponding in size to the hfq amplicon (which is comparable in size to the fragment amplified from chromosomal DNA) was detected in the WT profile, suggesting that hfq is expressed in S. aureus. As expected, no amplification from both hfq mutant strain chromosomal DNA (not shown) and cDNAs (Fig. 1B) was observed, which confirms the hfq ORF deletion. Positive amplification from cDNAs of the RN6390 hfq strain was visualized with PCR using spa specific primers, confirming the quality of these extracts (Fig. 1B). Similar results were obtained with RT-PCR experiments realized with Newman strain.

These results indicate that hfq gene is expressed in S. aureus in the experimental conditions tested but at a very low level, as no transcript could be visualized by northern blot experiments. In view of these data, we speculate that Hfq protein is probably less abundant in S. aureus than in E. coli.

We first compared growth of hfq mutants of RN6390, COL and Newman, with their respective WT strains in aerated BHI rich medium or in aerated RPMI synthetic medium at 37°C. No growth differences were noticed in any of the genetic backgrounds tested (data not shown) in either media. As Hfq is reportedly involved in stress tolerance in many pathogens [8,10,14], we also tested the ability of staphylococcal hfq mutants to resist to several environmental stresses: Under osmotic (NaCl), oxidative stress (H₂O₂) or heat-shock (44°C) conditions, the hfq mutant grew as well as the WT strain in all genetic backgrounds (data not shown). In L. monocytogenes, the contribution of Hfq in stress response depends on the alternative sigma factor σ^B pathway, as σ^B controls hfq stress-inducible transcription [14]. In S. aureus, Hfq does not seem to be implicated in stress resistance, regardless of the presence or absence of the σ^B pathway, as the same results were obtained in both σ^B pathway functional (COL, Newman) and defective (RN6390) strains.

In vitro studies showed that Hfq specifically binds the spa mRNA regulatory region, and more tightly, RNAIII [19]. spa encodes the Staphylococcal protein A (Spa), a major surface protein. Its expression is known to be repressed by RNAIII at the transcriptional level by an undetermined mechanism, but additional studies have shown that RNAIII also inhibits spa expression by affecting its translation and stability [19]. To elucidate the involvement of Hfq in the stability of RNAIII and spa mRNA in vivo, we compared their steady state levels in WT and hfq mutant strains. No significant differences were observed in RNAIII expression in the hfq mutant, compared to the WT strain, whether in the RN6390 or Newman context (Fig. 2A). When we compared spa RNA transcript profiles in RN6390, Newman and COL backgrounds, no differences between Δhfq and WT were detected (Fig. 2B). Variation of spa mRNA expression visualized between strain backgrounds reflects differences in virulence regulatory pathways, as previously described [23]. Amounts of protein A were also analysed in staphylococcal strains by western blot experiments [26] using a monoclonal anti-protein A antibody (Sigma, Saint Louis, MO, USA); again no differences were observed (data not shown). The results described above show that Hfq does not affect the quantities of RNAIII or of spa mRNA in vivo. In view of these observations, we speculate that in the natural context of the bacterium, Hfq does not contribute to the quantities of RNAIII and spa mRNA and likely has no effect on the formation of the RNAIII-spa mRNA complex. This is relevant to previous in vitro results showing that increasing concentrations of Hfq have no effect on the formation of RNAIII-spa mRNA complex [19].

S. aureus secretes numerous virulence factors, which are mainly produced during the post-exponential phase, and whose expression involves the coordination of many regulators, including agr RNAIII (for review, see [17]). To test the possible role of Hfq in the production of secreted virulence factors, we compared exoprotein profiles in RN6390, Newman, and COL WT strains and their Δhfq derivatives. No differences were observed in the various genetic backgrounds tested (data not shown). Hemolysins and proteases are major secreted virulence factors in some S. aureus virulent strains, such as RN6390. To test the effect of Hfq on production of those secreted enzymes, we compared hemolytic and proteolytic activities in RN6390 and Newman WT and hfq mutant strains. Stationary cultures were streaked onto BHI medium containing rabbit blood (which reveals α-hemolytic activity), sheep blood (corresponding to β-hemolytic activity) and horse blood (δ-hemolytic activity), or onto NB medium containing skimmed milk (to reveal proteolytic activities). No differences were observed in halos corresponding to the different enzymatic activities in the two genetic backgrounds tested (data not shown), which suggests that the lack of Hfq does not affect production of either hemolysins or secreted proteases.

Altogether, these results suggest that deletion of hfq does not affect exoprotein expression profiles. It is thus likely that Hfq is not decisive for the production of secreted virulence factors.

PM assay is a relatively new technology in which ~ 2000 phenotypes are tested simultaneously, and has already been used for comparative analysis between strains in several bacterial species [27-29]. Strains are grown in microtiter plate, each well containing different media. This technology may help to determine gene functions. Respiration is used in PM as an indication of bacterial growth, and was compared between strains in media containing different sources of carbon (PM1 to PM2), nitrogen (PM3 and PM6 to PM8), phosphorus and sulfur (PM4), nutrients or cofactors (PM5), or in the presence of different stresses or chemical agents (PM9 to PM20). Consensus PM results of comparative analysis of RN6390 WT and hfq mutant strains are shown on Figure 3. Additional file 1 indicates values of phenotypes gained or lost. Briefly, slight or no differences between strains were observed under any metabolic conditions, and differences in sensitivity to chemical agents observed could not be reproduced by independent experimentation.

Metabolism tests (PM1 toPM8), performed in a synthetic medium, showed negative differences for utilization of several amino acids or dipeptides as nitrogen source (PM3 and PM6 to PM8); although the signal decreases were very weak. These data seemed to suggest that the hfq mutant strain could be defective for nitrogen metabolism. To test this hypothesis, we compared growth of RN6390 WT and hfq mutant strains in a synthetic complete medium or in a synthetic minimal medium (containing the essential amino acids for auxotrophy, glutamate and leucine; Elise Durant, INRA, personal communication) with or without several nitrogen sources, including different combinations of amino acids (including glutamate and/or aspartate) as used in the PM assay. No differences in growth were observed between the two strains (data not shown), suggesting that the hfq mutant uses different nitrogen sources as well as the WT strain. The weak differences detected by PM analysis (PM3, PM6 to PM8) may reflect the use in PM of a standard chemically minimal medium that lacks the essential amino acids, where only residual growth of both strains occurred. In this case, it is likely that phenotypes observed result from residual respiration of surviving bacteria in the medium and not from real bacterial growth.

Stress or chemical agent sensitivity PM tests (PM9 to PM20) were performed in rich medium, similar to LB. Essentially no differences were observed in osmotic (PM9) and pH (PM10) stress panels between the RN6390 WT and hfq mutant strains (Fig. 3), which confirmed some results obtained by conventional stress tests (see above). Positive or negative differences were observed for some chemical agents or antibiotics (PM11 to PM20; see Fig. 3, and Additional file 1) but signals were weak (see Additional file 1). The only real signal concerned Cm resistance of the hfq mutant, a consequence of the strain construction (PM11, F3-4 and PM 14, F4, and derivative thiamphenicol, PM18, A10-12). We compared sensitivity of some chemical agents or antibiotics that gave negative responses (cobalt chloride, PM13, G3; cefotaxime, PM16, A3-4), or positive signals (domiphen bromide, PM15, D6; dodine, PM20, E7) on MH broth by measuring the MIC or on agar plates using disk-specific antibiotics. Again, no significant differences were observed between WT and hfq mutant strains (data not shown).

This comparative analysis showed that among ~ 2000 phenotypes, essentially no reproducible phenotypic differences between S. aureus RN6390 WT and its hfq derivative mutant were detected. For metabolic tests, the few phenotypes that were noted likely have no physiological significance, and may result from the poor growth of both strains. We thus conclude that Hfq has no role on stress response and on chemical agent sensitivity and probably no impact on metabolic pathways in S. aureus.

Bacterial strains and plasmids used in this study are listed in Table 1. S. aureus strains were routinely grown in BHI medium with aeration at 37°C, or at 30°C for the RN4220 strain harboring the thermosensitive plasmid. Defined synthetic medium used for staphylococcal growth was either RPMI medium 1640 (Gibco-Invitrogen, Eugene, USA) or as published [35]. E. coli DH5αZ1 [36], which was used for cloning experiments, was grown in LB broth at 37°C. Antibiotics were added to media as needed at the following concentrations: Cm, 5 μg/ml for S. aureus and 10 μg/ml for E. coli; Ery, 2 μg/ml for S. aureus and Amp, 100 μg/ml for E. coli. X-Gal was added as needed to BHI plates at a final concentration of 150 μg/ml.

For stress studies, staphylococcal cultures were prepared as follows: serial dilutions of exponential cultures were spotted on BHI plates containing or not an osmotic stress agent (NaCl, 0.5 to 2 M) or an oxidative stress agent (H₂O₂, 1 to 2 mM), and incubated at 37°C for 24 h. Growth at high temperature was examined by plating cells on BHI medium and incubating at 44°C for 24 h.

Plasmid and chromosomal DNA preparations, PCR amplifications, and DNA modifications were performed according to commonly used techniques or suppliers' instructions. Lysis of S. aureus cell suspensions was achieved by treatment with 100 μg/ml of lysostaphin (Sigma-Aldrich, St. Louis, MO, USA) for 1 hour at 37°C, followed by standard methods for DNA preparation. DNA transformation for E. coli [37] and for S. aureus [23] was performed as described.

hfq was inactivated using the pMAD plasmid, which is thermosensitive for replication [38], as described [23]. Briefly, regions up- and downstream of hfq were amplified, fused to flank the pC194 cat amplified gene [39] (see primer list in Table 2) and cloned into the pGEM-T-Easy vector (Promega, Madison, WI, USA). The segment of interest was then subject to appropriate digestions and cloned into pMAD. The resulting plasmid, pMAD Δhfq::cat was introduced into the RN4220 strain and hfq mutant derivatives were selected as described [23]. Gene inactivation was confirmed by Southern blotting [26], after HaeIII digestion of RN4220 WT and hfq chromosomal DNA, using a PCR product corresponding to the Hfq ORF region as a probe (see Table 2 for primers used). The hfq mutation was then transferred into RN6390, Newman and COL S. aureus strains by ϕ11 phage-mediated transduction [23]. Mutation transfer was confirmed by PCR and RT-PCR (see primers list in Table 2).

Overnight cultures of staphylococcal strains in BHI medium were diluted 1000-fold in the same medium and grown at 37°C. Total RNAs of mid-or post- exponential or stationary phase cultures were extracted as previously described [40] or using the RNA spin mini Kit (GE Healthcare, Buckinghamshire, UK). Northern blot experiments were performed using hfq, RNAIII- and spa specific fragments as probes (see Table 2 for primers list), which were α³²-P labelled, as described [26]. RNA transcripts were quantified, using hu mRNA as positive control of constitutive expression. For hfq expression studies, random RT-PCR experiments were performed using the Thermoscript RT-PCR System (Invitrogen, Eugene, USA); primers used for successive PCR are in Table 2.

Overnight cultures of staphylococcal strains in BHI medium were diluted 100-fold in the same medium and grown at 37°C. Supernatant protein extracts from post-exponential or stationary phase cultures were prepared as described [23] and protein profiles were compared by PAGE after staining with Coomassie blue reagent [26]. For hemolytic or proteolytic activities, staphylococcal stationary phase cultures were streaked on BHI medium containing 5% rabbit, sheep, or horse blood, or on NB medium containing 10% skimmed milk. Enzymatic halos were compared after overnight growth at 37°C.

S. aureus RN6390 WT and hfq mutant strains were subject to full profile (20 panels) PM analysis, as described elsewhere [28,29,41]. This extensive analysis contains ~ 2000 tests which include utilization of different sources of carbon (PM1 and PM2; ~ 200 tests), nitrogen (PM3 and PM6 to PM8; ~ 400 tests) phosphorus and sulfur (PM4; ~ 100 tests), nutrient stimulation (PM5; ~ 100 tests) as well as sensitivity to several stresses (PM9 and PM10; ~ 200 tests) and to chemical agents that affect various biological pathways (PM11 to PM20; ~ 1000 tests). Detailed information of PM experimental conditions is available [42]. Briefly, respiration during bacterial growth under different conditions is monitored by NADH production, measured with a tetrazolium dye and evaluated for intensity of color formation. Phenotypic data were recorded over a 24-hour period. Differences in measurements of WT and hfq mutant strains are calculated. The WT strain is recorded as a red tracing, and the hfq mutant as a green tracing. Areas of overlap (no changes) between the two measurements are colored in yellow, whereas differences are colored in red or green for WT or hfq mutant respectively. The consensus profile presented is the result of two independent comparative analyses.

To confirm some of the observations with the nitrogen metabolic arrays (PM3 and PM6 to PM8), growth of RN6390 WT and hfq mutant strains was compared in a chemically defined medium [35] containing or not different sources of nitrogen. Sensitivity to selected chemical agents (cobalt chloride, dodine, domiphen bromide) or antibiotics (cefotaxime) was tested by determining the Minimum Inhibitory Concentration, (MIC) or by using the agar disk diffusion method (Cefotaxime, Fluka). Both of these experiments were performed in Mueller Hinton broth.