Staphylococcus aureus RNAIII and the endoribonuclease III coordinately regulate spa gene expression

Eric Huntzinger, Sandrine Boisset, Cosmin Saveanu, Yvonne Benito, Thomas Geissmann, Abdelkader Namane, Gérard Lina, Jerome Etienne, Bernard Ehresmann, Chantal Ehresmann, Alain Jacquier, François Vandenesch, Pascale Romby

Author Affiliations

  1. Eric Huntzinger1,,
  2. Sandrine Boisset2,,
  3. Cosmin Saveanu3,
  4. Yvonne Benito2,
  5. Thomas Geissmann1,
  6. Abdelkader Namane4,
  7. Gérard Lina2,
  8. Jerome Etienne2,
  9. Bernard Ehresmann1,
  10. Chantal Ehresmann1,
  11. Alain Jacquier3,
  12. François Vandenesch*,2 and
  13. Pascale Romby*,1
  1. 1 UPR 9002 du CNRS, Institut de Biologie Moléculaire et Cellulaire, Strasbourg Cedex, France
  2. 2 Faculté de Médecine Laennec, National Reference Center for Staphylococci, INSERM E0230, Lyon Cedex, France
  3. 3 URA2171‐CNRS‐Génétique des Interactions Macromoléculaires, Paris Cedex, France
  4. 4 Plateforme protéomique, Institut Pasteur, Paris Cedex, France
  1. *Corresponding authors: UPR 9002 du CNRS, Institut de Biologie Moléculaire et Cellulaire, 15 rue René Descartes, 67084 Strasbourg Cedex, France. Tel.: +33 388 41 70 51; Fax: +33 388 60 22 18; E‐mail: p.romby{at}ibmc.u-strasbg.frFaculté de Médecine Laennec, National Reference Center for Staphylococci, INSERM E0230, IFR62, 7 rue Guillaume Paradin, 69372 Lyon Cedex 08, France. Tel.: +33 478 77 86 57; Fax: +33 478 77 86 58; E‐mail: denesch{at}
  1. These authors contributed equally to this work

View Full Text


Staphylococcus aureus RNAIII is one of the largest regulatory RNAs, which controls several virulence genes encoding exoproteins and cell‐wall‐associated proteins. One of the RNAIII effects is the repression of spa gene (coding for the surface protein A) expression. Here, we show that spa repression occurs not only at the transcriptional level but also by RNAIII‐mediated inhibition of translation and degradation of the stable spa mRNA by the double‐strand‐specific endoribonuclease III (RNase III). The 3′ end domain of RNAIII, partially complementary to the 5′ part of spa mRNA, efficiently anneals to spa mRNA through an initial loop–loop interaction. Although this annealing is sufficient to inhibit in vitro the formation of the translation initiation complex, the coordinated action of RNase III is essential in vivo to degrade the mRNA and irreversibly arrest translation. Our results further suggest that RNase III is recruited for targeting the paired RNAs. These findings add further complexity to the expression of the S. aureus virulon.


Virulence gene expression in most bacteria is tightly regulated, and many bacterial pathogens turn on virulence genes under appropriate conditions allowing the bacteria to survive within the host and to evade the host defense system. Recent studies performed on pathogenic bacteria point to the importance of RNAs in virulence control. Regulatory RNAs might exert their activity in trans through antisense properties or by sequestering proteins (for a review, see Johansson and Cossart, 2003; Storz et al, 2004). One of the most attractive examples is the Staphylococcus aureus 514‐nucleotide regulatory RNAIII, which regulates the expression of several virulence factors. S. aureus virulence factors are precisely controlled in response to cell density (quorum sensing), energy availability, and various environmental signals (reviewed in Novick, 2003). Signal receptors are the primary regulatory mediators for the expression of the virulon in S. aureus. Among these receptors, the agr system, composed of two divergent transcription units, functions as a sensor of the population density. Its operon combines a density‐sensing cassette (agrD and B) and a two‐component sensory transduction system (agrA and C). The P2 operon expression is required for transcriptional activation of the P3 operon coding for RNAIII, the intracellular effector of the agr regulon (Novick, 2003). The expression of RNAIII is maximal in the late logarithmic and stationary phase cultures.

RNAIII has the unique property of acting as a messenger RNA encoding hld (delta‐hemolysin), and having multiple regulatory functions: repression of the expression of surface proteins such as protein A during the exponential phase, and activation of the expression of extracellular toxins and enzymes during the postexponential phase (Janzon and Arvidson, 1990; Kornblum et al, 1990; Novick et al, 1993). The secondary structure of RNAIII is characterized by 14 stem–loop structures and by two long helices, which close off structural domains (Benito et al, 2000). Only two specific domains of RNAIII have been shown to control the expression of different targets. The 5′ end of RNAIII positively regulates the translation of hla (encoding alpha hemolysin) by competing directly with an intramolecular RNA secondary structure that sequesters the hla ribosomal binding site (Novick et al, 1993; Morfeldt et al, 1995). The 3′ domain of RNAIII is necessary and sufficient for repression of protein A synthesis (Novick et al, 1993; Benito et al, 1998, 2000). In this latter case, RNAIII is believed to function either directly or indirectly at the transcriptional level (Patel et al, 1992; Novick et al, 1993; Cheung et al, 1997). However, one cannot exclude an alternative mechanism in which RNAIII would alter the stability of spa mRNA (Novick et al, 1993; Benito et al, 2000). Indeed, sequence complementarity between RNAIII and spa mRNA has been suggested previously (Novick, 2003).

The present study analyses the regulatory mechanism by which RNAIII inhibits the expression of the surface protein A, one of the major virulence factors. Our results indicate that spa mRNA is repressed at the translational level by RNAIII and, in turn, is rapidly degraded by the double‐strand‐specific endoribonuclease III (RNase III). The mechanism of RNAIII action relies on an antisense activity, which is located in the 3′ domain and, is mediated by base pairings with spa mRNA.


spa regulation occurs both at the transcriptional and post‐transcriptional levels in vivo

Sequence complementarity between RNAIII (nucleotides 384–445) and spa mRNA (nucleotides 2–58) suggested that the 3′ domain of RNAIII can repress spa expression at the translational level, through annealing to spa mRNA (Novick, 2003). In order to test the in vivo relevance of such a mechanism, we fused the spa and lacZ genes in expression vectors. The spa promoter region (−454 to +1) alone or containing part of the coding sequence (−454 to +12) were cloned into the pTCV‐lac shuttle vector (Poyart and Trieu‐Cuot, 1997). We also used an agr‐independent promoter (PrpoB) fused to the coding sequence of spa mRNA (+1 to +63), which contains the sequence complementarity to RNAIII. These constructs were designed in order to distinguish between transcriptional and post‐transcriptional control. β‐galactosidase activity was assayed in the S. aureus strains RN6390 (agr+) and WA400 (ΔrnaIII). Experiments were also carried out on the strain WA400, complemented with different sets of plasmids expressing either the wild‐type RNAIII (LUG581) or RNAIII‐Δ1 (LUG451) deleted of residues 430–437, which are complementary to the Shine and Dalgarno sequence of spa mRNA (see Figures 1 and 2A).

Figure 1.

β‐galactosidase activity detected from different gene fusions. (A) PspalacZ and Pspa (+1/+12)∷lacZ fusions in S. aureus RN6390 (agr+, WT), WA400 (ΔrnaIII) and RN6390‐Δrnc (LUG774, Δrnc), and PrpoB∷lacZ fusion in S. aureus RN6390 and WA400. (B) PrpoB∷spa (+1/+63)‐lacZ fusion in different S. aureus strains: RN6390, WA400, WA400+RNAIII (LUG581, ΔrnaIII/pLUG298), WA400+RNAIII‐Δ1 (LUG451, ΔrnaIII/pLUG302) and LUG774. The β‐galactosidase activity is expressed in arbitrary unit per milligram of protein. The results represented a mean of three independent experiments.

Figure 2.

RNAIII binds efficiently to spa mRNA in vitro. (A) The secondary structure model of RNAIII from Benito et al (2000), and the 3′ domain are given. The two mutants, which carry deletion in loop 13 (RNAIII‐Δ1), and the whole deletion of hairpin 13 (RNAIII‐Δ2) are squared. Nucleotides complementary to spa mRNA are printed in italic. (B) Determination of the apparent dissociation constant for RNAIII/spa mRNA complex. The 5′‐end‐labeled spa mRNA was incubated alone (−) or with various concentrations of unlabeled wild‐type RNAIII (0.1, 0.5, 1, 5, 10, 50 and 100 nM) and mutant RNAIII (50 and 100 nM). The fraction of labeled spa mRNA associated with RNAIII was calculated from the counts in the corresponding band relative to the total counts in the lane. The apparent Kd value was determined as the concentration of RNAIII allowing 50% of spa mRNA binding. (C) Binding rate constant for the RNAIII/spa mRNA complex as determined from three independent experiments. The 5′‐end‐labeled RNAIII (0.1 nM) was incubated with unlabeled spa mRNA (1 nM) at 37°C. Aliquots were withdrawn at 0, 1, 2, 4, 8, 16, 32 and 60 min. The percentage of free RNAIII was plotted as a function of time.

We observed that the minimal spa promoter element, which responds to the agr regulation, required the first 12 nucleotides of the spa gene in agreement with recent data showing that agr response involved nucleotides −137 to +7 (Gao and Stewart, 2004; Figure 1A). As expected, under the control of the spa promoter, the expression of the fusion protein strongly decreased in the wild‐type strain as compared to the mutant strain WA400 (Figure 1A). These data suggest that RNAIII is required for the transcriptional control of spa expression.

When the rpoB promoter was fused with the lacZ gene, the β‐galactosidase synthesis was identical in RN6390 (agr+) and WA400 (ΔrnaIII) showing that the rpoB promoter activity is agr independent (Figure 1A). The addition of the first 63 nucleotides of spa mRNA fused in frame with lacZ deleted of its ribosomal binding site (ΔRBS) reduced the β‐galactosidase activity by two‐fold in the wild‐type strain as compared to WA400 (Figure 1B). The β‐galactosidase levels decreased more drastically in WA400 carrying the multicopy plasmid pE194 producing high level of the wild‐type RNAIII (LUG581) (Figure 1B). Conversely, the expression of RNAIII‐Δ1 in WA400 (LUG451) had no effect on β‐galactosidase reduction (Figure 1B). Taken together, these results indicate that spa expression is controlled at both transcriptional and post‐transcriptional levels.

RNAIII interacts with spa mRNA in vitro

The in vivo data prompted us to examine whether RNAIII might efficiently bind to spa mRNA. Duplex formation between RNAIII and spa mRNA was investigated by gel shift analysis. In vitro‐synthesized labeled spa mRNA (<1 nM) was incubated with increasing concentrations of unlabeled RNAIII in the presence of 5 mM MgCl2 at 37°C for 15 min. The reactions were analyzed by gel electrophoresis using nondenaturing polyacrylamide gels (Figure 2B). RNAIII was found to bind spa mRNA with an apparent dissociation constant of 31 (±5.5) × 10−9 M. The deletion of nucleotides C430 to A437 (RNAIII‐Δ1) or of the whole hairpin 13 (RNAIII‐Δ2) in RNAIII (Figure 2A) had a deleterious effect on complex formation (Figure 2B). Thus, RNAIII formed a moderately stable complex with spa mRNA and mutations, altering the complementarity with spa mRNA, impaired complex formation.

The rate of complex formation between RNAIII and spa mRNA was further analyzed by gel shift assays (Figure 2C). The initial rate of binding (kapp) was estimated by a time‐course analysis using labeled RNAIII and an excess of the unlabeled spa mRNA and was determined to be 3 (±0.5) × 105 M−1 s−1. The rate of binding was proportional to the concentration of the RNAs, as expected for a second‐order reaction (result not shown). This constant is probably underestimated since complex formation is reversible and some dissociation could occur within the gel. The value is, however, close to that obtained for OxyS‐fhlA mRNA (Argaman and Altuvia, 2000) and in the range of the association rate constants of 105–106 M−1 s−1, determined in a variety of natural fully complementary antisense–target RNA systems using analogous methodology (for a review, see Wagner et al, 2002). This rapid binding between antisense and target RNA is thus compatible with RNAIII exerting its inhibitory effect on spa expression by binding to spa mRNA in vivo.

RNAIII sequesters the ribosome binding site of spa mRNA

The conformation of the first 200 nucleotides of spa mRNA either free or bound to RNAIII was probed using different enzymes and chemicals under conditions where less than one cleavage/modification per molecule takes place. Enzymes used were RNase T1 (specific for single‐stranded guanines), RNase T2 (specific for unpaired nucleotides with a preference for adenines) and RNase V1 (specific for double‐stranded regions). Spa mRNA was also subjected to chemical modifications by using DMS (A(N1), C(N3)) and CMCT (U(N3), G(N1)). Experiments are shown in Figure 3 and the reactivity levels are indicated on the secondary structure of the RNAs in Figure 4.

Figure 3.

RNAIII binds to the SD region of spa mRNA. (A, B) Enzymatic hydrolysis of 5′‐end‐labeled spa mRNA, free (−RNAIII) or in the presence of an excess of RNAIII (+RNAIII). (C) Chemical probing of unlabeled spa mRNA, free (−) or in complex with RNAIII (+). Lanes U, G, C and A: dideoxy‐sequencing reactions performed on spa mRNA. (D) Enzymatic hydrolysis of 3′‐end‐labeled RNAIII free (−spa mRNA) or in the presence of an excess of spa mRNA (+spa mRNA). T2, T1 and V1: RNase T2, RNase T1 and RNase V1, respectively. Lanes T and L: RNase T1 under denaturing conditions and alkaline ladders, respectively. Lanes C: incubation controls in the absence of the probe. Arrows denote the main reactivity changes induced by complex formation.

Figure 4.

Secondary structure models of free and complexed RNA. A summary of the probing results is represented on the secondary structure of spa mRNA (A) and RNAIII (B). Enzymatic cleavages are given as follows: RNase T1 (Graphic), RNase T2 (Graphic) and RNase V1 (Graphic) moderate and (Graphic) strong cleavage. Chemical modifications of cytosines at N3, and adenines at N1 towards DMS, and of uridines at N3 and guanines at N1 towards CMCT: full and dashed circled nucleotides are for strong and moderate reactivity, respectively. Small dots are for not determined due to unspecific cleavages or pauses of RT in the incubation control. Reactivity changes induced by complex formation are indicated as follows: black or empty circles denote strong and moderate protection, respectively; enhancements are represented by asterisks; new RNase V1 cleavages are shown by red arrows. Blue arrows show RNase III cleavages in the free RNA. Potential noncanonical base pairs are denoted by NoN. (C) Secondary structure model of the RNAIII/spa mRNA complex summarizing the enzymatic cleavages and chemical reactivities.

For RNAIII, the data were in good agreement with the secondary structure previously derived from chemical and enzymatic probing (Bénito et al, 2000; Figure 2A). The 3′ domain (containing hairpins 13 and 14) was shown to form an intrinsic structural domain, highly conserved among Staphylococcus species (Bénito et al, 2000). The existence of hairpin 13 (the main spa mRNA binding site) was supported by previous data: nucleotides within loop 13 were reactive at their Watson–Crick position despite their weak accessibility towards RNase T2, while the helix was cleaved by RNase V1 (Bénito et al, 2000; Figure 4B). For spa mRNA, the probing data were well correlated with the existence of hairpins I and II (Figure 4A). An alternative structure in spa mRNA involving nucleotides A29 to A47 is suggested due to the presence of several RNase V1 and RNase T2 cleavages (inset in Figure 4A).

Binding of RNAIII induced reactivity changes in a restricted region of spa mRNA, which covers the ribosome binding site. Protections against RNase T1 and T2 hydrolysis were mainly observed in loop I (Figure 4A). In addition, reactivity of adenines 29–33, and 35 towards DMS, and of uridines 42, 43, 46–47, 51–52 towards CMCT was significantly decreased upon RNAIII binding. Several RNase V1 cuts were also reduced at positions 18–20 and 26–27, whereas a new cleavage appeared at position 29. The reactivity pattern was unchanged in the other regions of spa mRNA.

Binding of spa mRNA induced correlated reactivity changes in hairpin 13 of RNAIII, encompassing nucleotides 390–440 (Figure 4B). RNase T2 cleavages located at positions U392–U394 and U421–U424 were reduced, whereas increased cleavages were observed at A425–A428. New or increased RNase V1 cleavages also occurred at positions U420–U424, and C430–A434. The complex was also formed by annealing the two RNA molecules at high temperature followed by slow cooling to 37°C. The reactivity pattern was strictly identical to that of the complex formed at 37°C (results not shown).

These data show that RNAIII anneals to spa mRNA, and induces changes restricted to the complementary region that overlaps the ribosome binding site.

Binding of RNAIII to spa mRNA prevents the formation of the translational initiation complex

Since RNAIII binds to the Shine and Dalgarno sequence of spa mRNA, we tested the possibility that RNAIII binding is sufficient to prevent the formation of the ternary initiation complex formed by the Escherichia coli 30S subunit, the initiator tRNAMet and spa mRNA. The formation of the ternary complex blocked the elongation of a cDNA primer by reverse transcriptase, and produced a toeprint 16 nucleotides downstream of the initiation codon (Hartz et al, 1988; Figure 5A). Since RNAIII contained an effective ribosomal binding site that could interfere with ribosome binding at spa mRNA, we also used the two RNAIII mutants (RNAIII‐Δ1, RNAIII‐Δ2), which have lost the capacity to bind spa mRNA but are still recognized by the ribosome. The data revealed that the addition of increasing concentrations of RNAIII significantly decreased the yield of the toeprint (Figure 5A, lanes 6–8), 50% of inhibition being observed at the highest RNAIII concentration (Figure 5B). As the ternary 30S‐mRNA–tRNA complex forms irreversibly, the inhibition was only observed when the initiator tRNA was added after RNAIII. The addition of RNAIII‐Δ1 or RNAIII‐Δ2 slightly altered ribosome binding since less than 20% of inhibition was observed at the highest concentration of the RNAIII mutants (Figure 5B). Thus, the specific RNAIII inhibition of ribosome binding to spa mRNA did not mainly result from a sequestration of the 30S subunit by RNAIII but rather from a direct interaction between spa mRNA and RNAIII. The binding of RNAIII to spa mRNA was sufficiently rapid to inhibit the formation of the translational initiation complex, at a step preceding the formation of the ternary complex.

Figure 5.

RNAIII prevents the formation of the ternary complex 30S‐spa mRNA–tRNA. (A) Formation of the ternary complex between spa mRNA (15 nM), 30S ribosomal subunits (500 nM) and initiator tRNA (1 μM) was monitored in the absence (lane 5) or in the presence of wild‐type RNAIII (lanes 6–8), RNAIII‐Δ1 (lanes 9–11) and RNAIII‐Δ2 (lanes 12–14). Concentrations of RNAIII species were 15, 75 and 150 nM. The toeprint at position +16 is indicated. Lanes U, G, C and A: dideoxy‐sequencing reactions performed on spa mRNA. (B) Quantification of the data determined for three independent experiments. Relative toeprinting (toeprint band over full‐length RNA+toeprint) was calculated by scanning of the gel with the Bio‐imager Analyzer BAS 2000 (Fuji). The inhibition, given in %, represents the ratio between the relative toeprint in the presence of either the wild‐type RNAIII or the mutant RNAs divided by the relative toeprint in the absence of RNAIII.

S. aureus RNase III is essential for efficient regulation of protein A expression in vivo

Many antisense‐target systems are processed by the double‐strand‐specific RNase III (for a review, see Wagner et al, 2002). To test whether RNase III is involved in the regulation of spa expression, we constructed a strain derived from RN6390 in which the RNase III gene (rnc) was inactivated (LUG774). Codons 56–200 of the rnc gene, encompassing the RNA binding domain and the catalytic site, were deleted and replaced by the cat gene. Allelic replacement of the rnc gene in the strain RN6390 (forming LUG774) was analyzed by PCR using different sets of oligodeoxyribonucleotides, allowing detection of the expected sizes of the generated PCR fragments (data not shown). In contrast to Bacillus subtilis (Herskovitz and Bechhofer, 2000), the rnc gene is nonessential in S. aureus.

The translational fusion of the spa gene (1 to +63) with the lacZ gene expressed from the agr‐independent PrpoB promoter was then transformed into LUG774 in order to analyze the post‐transcriptional regulation of spa in the absence of RNase III. The β‐galactosidase levels were identical to those measured in the mutant strain WA400 (ΔrnaIII) (Figure 1B), showing that the absence of either RNAIII or RNase III abolished spa repression. In contrast, low β‐galactosidase levels were detectable in LUG774 carrying the transcriptional Pspa∷lacZ fusion (Figure 1A). These data suggest that the regulation of spa expression at the post‐transcriptional level depends on both the presence of RNAIII and the double‐strand‐specific RNase.

S. aureus RNase III and RNAIII promote rapid degradation of repressed spa mRNA

We then tested the steady‐state level of spa mRNA in the different S. aureus strains. Disappearance of spa mRNA was observed in RN6390 (agr+, Figure 6A, lane 1) as well as in WA400/pE194 expressing the wild‐type RNAIII (LUG581, Figure 6A, lane 4). The level of spalacZ fusion mRNA (expressed under agr‐independent rpoB promoter) was also strongly decreased in WA400/pE194 expressing the wild‐type RNAIII (results not shown). Conversely, the level of spa mRNA was significantly increased in WA400 (ΔrnaIII) (Figure 6A, lane 2), and also in WA400/pE194 expressing RNAIII‐Δ1, which has lost the capacity to bind spa mRNA (Figure 6A, lane 5). We also detected spa mRNA in the Δrnc strain (LUG774) (Figure 6A, lane 6).

Figure 6.

Analysis of spa mRNA and RNAIII levels in different strains. (A) Northern blot analysis on spa mRNA. RNAs from postexponential phase cultures were hybridized with probes corresponding to RNAIII and spa mRNA. Lane 1, RN6390 (WT, agr+); lane 2, WA400 (ΔrnaIII); lane 3, WA400+3′ domain (LUG484: ΔrnaIII/pLUG324); lane 4, WA400+RNAIII (LUG581: ΔrnaIII/pLUG298); lane 5, WA400+RNAIII‐Δ1 (LUG451: ΔrnaIII/pLUG302); lane 6, RN6390‐Δrnc (LUG774: Δrnc). (B) Measurements of the half‐life of spa mRNA in the presence of rifampicin (300 μg/ml) in the strains WA400, RN6390‐Δrnc and RN6390. For comparison, the half‐life of spalacZ mRNA (agr‐independent PrpoB promoter) in WA400 is given. (C) Measurements of the half‐life of RNAIII in the presence of rifampicin in the strains RN6390 (WT) and RN6390‐Δrnc. The half‐life of 5S rRNA was measured on all the membranes in parallel, as an internal control. The quantity of spa mRNA or of RNAIII was normalized using the quantity of 5S rRNA at each time. The half‐life was given as the time where 50% of the RNA was degraded. The experiments were reproduced four times.

The half‐life of spa mRNA was then analyzed in different strains (Figure 6B). In rifampicin‐treated cells expressing RNAIII (RN6390, LUG581), spa mRNA was not detectable. Conversely, in rifampicin‐treated cells, which do not express RNAIII (WA400), spa mRNA accumulated and was found to be unusually stable with a half‐life of 15 (±2) min (Figure 6B). A long half‐life was also measured for spalacZ mRNA (expressed under the agr‐independent rpoB promoter), indicating that the 5′ leader conferred a high stability to spa mRNA (Figure 6B). Interestingly, degradation of spa mRNA was even reduced in Δrnc strain (LUG774) since the half‐life of spa mRNA increased to 32 (±2) min (Figure 6B). We also measured the half‐life of RNAIII in the presence of rifampicin in RN6390 and LUG774. In both cases, RNAIII appeared to be highly stable (>45 min), as described previously (Janzon and Arvidson, 1990; Figure 6C), and no stable RNase III‐specific processing intermediates were detectable in the wild‐type strain (results not shown).

Since the spa mRNA half‐life is long, we suggest that decay of spa mRNA significantly contributes to the overall repression mechanism, and that rapid degradation of spa mRNA is dependent on both RNAIII pairing and RNase III.

RNase III recognizes RNAIII, spa mRNA and the RNAIII–spa mRNA complex

Since RNase III is required for the post‐transcriptional regulation of spa expression in vivo, we tested the capacity of the enzyme to cleave the duplex in vitro. The His‐tagged RNase III from S. aureus was overexpressed in E. coli and purified. Mass spectrometry analysis and N‐terminal sequencing did not reveal any contamination by E. coli RNase III. The extent of duplex formation was analyzed by comparing the RNase III‐dependent cleavages under conditions of heat annealing and native RNAIII–spa mRNA complex formation. RNase III hydrolysis was conducted in parallel on the free RNA species. The cleavage positions were mapped using either end‐labeled RNAIII or spa mRNA (Figure 7A and B), and are indicated on the secondary structure model (Figure 4C). The activity of S. aureus RNase III was dependent on Mg2+ ions since the presence of Ca2+ instead of Mg2+ in the reaction buffer inhibited the hydrolysis, as shown for the E. coli enzyme (for a review, see Nicholson, 1996).

Figure 7.

S. aureus RNase III recognizes the free spa mRNA and RNAIII, and RNAIII–spa mRNA complex. (A) RNase III hydrolysis of 5′‐end‐labeled spa mRNA, free (−) or in the presence of an excess of RNAIII (+). (B) RNase III hydrolysis of 5′‐end‐labeled RNAIII, free (−) or in the presence of an excess of spa mRNA (+). Lanes C: incubation control in the absence of RNase III. Lanes T1, L: RNase T1 and alkaline ladders, respectively. Arrows pointed the RNase cleavages occurring either in the free RNA (Graphic) or as the result of spa–RNAIII complex formation (Graphic). (C) S. aureus RNase III binds specifically to RNAIII. Crude extract was loaded onto the free streptavidin beads (−) and on the beads bound to either the 3′ end biotinylated RNAIII (+RNAIII), or a group II intron fragment RNA (Δ52XBA). Elution was carried out with increasing concentrations of NaCl. M is for molecular weight markers (in kDa).

Secondary structure analysis of spa mRNA supported a long and stable hairpin structure II, characterized by a 23 base pair (bp)‐long helix (Figure 4A), which appeared to be a target site for S. aureus RNase III. Two main cleavages were detected at positions A71 and A97 (Figure 7A), even at low enzyme concentration. When labeled spa mRNA was incubated at 37°C in the presence of RNAIII to form the 'native' inhibitory complex, additional cleavages were detected at positions A5, A31 and A39 (Figure 7A).

Using end‐labeled RNAIII, one major cleavage was reproducibly detected in the central domain of the free RNA, at position U321 and to a lesser extent in the 3′ domain, mainly at positions 448, A450 and U452 (Figure 7B). Binding of unlabeled spa mRNA induced new cleavages at positions U413 and to a lesser extent at C415, U418 and U420 within the region of complementarity with spa mRNA (Figure 4C). Concomitantly, cleavage at U321 persisted, while cleavages in hairpin 13 (A448, A450 and U452) were decreased (Figure 7B). The same RNase III‐induced cleavages were detected when the RNA–RNA complex was obtained by heat annealing, indicating that the native complex corresponds to the extended duplex.

In an attempt to identify S. aureus proteins that might specifically interact with RNAIII, we used in vitro‐transcribed RNAIII, biotinylated at its 3′ end and immobilized on streptavidin–agarose beads. Among the proteins that showed specific association with RNAIII, a major band was identified by mass spectrometry as the RNase III. This protein was eluted from the RNAIII beads at high salt concentration only (0.6 M, Figure 7C) as compared with another RNA used as a control. To further confirm the identified interaction, we performed filter binding assays that showed an apparent dissociation constant of 0.5 μM for the RNAIII–RNase III complex, in the presence of Ca2+ to prevent hydrolysis (data not shown).

These data suggested that the spa mRNA–RNAIII complex formed an extended duplex recognized and cleaved by S. aureus RNase III. In addition, we found that spa mRNA alone, and more unexpectedly RNAIII, were specifically recognized by the double‐strand‐specific RNase III in vitro.

S. aureus Hfq may assist the RNAIII‐dependent regulation

E. coli Hfq protein has been directly linked to the action of several small regulatory RNAs that use base pairings to regulate the expression of target mRNAs (e.g. Zhang et al, 2002). To test for an interaction between RNAIII and S. aureus Hfq protein in vivo, extracts from the wild‐type cells (RN6390) were subjected to immunoprecipitation with either Hfq‐specific serum or a nonspecific antibody. The experiments showed that RNAIII was only co‐immunoprecipitated with Hfq‐specific serum (Figure 8A). Neither the presence of RNAIII nor of Hfq was detected in the cell extracts using a nonspecific antibody. Using bandshift experiments, S. aureus Hfq was shown to bind RNAIII with a higher affinity than the regulatory region of the spa mRNA (Figure 8B and C). However, the addition of increasing concentrations of Hfq had no significant effect on the formation of the RNAIII–spa mRNA complex (Figure 8C).

Figure 8.

S. aureus RNAIII is a target of Hfq protein. (A) RNAIII co‐immunoprecipitates with Hfq in vivo. RNAIII was detected by RT–PCR reaction in the flow through fraction (lane F) and in the precipitate fraction (lane E). The immunoprecipitation assays were also carried out with unspecific Ab, which do not bind Hfq. (B) Gel mobility shift assay with end‐labeled spa mRNA in the presence of increasing concentrations of Hfq, and (C) with end‐labeled RNAIII either free or in complex with spa mRNA in the presence of increasing concentrations of Hfq protein. Labeled RNAIII was first incubated with spa mRNA (10 nM) at 37°C for 20 min and various concentrations of Hfq (0.1, 5 and 1 μM) were added for a further 10 min of incubation. Note that the shift is not pronounced due to the large size of the RNAIII, but was reproducibly found in three independent experiments.


S. aureus RNAIII is a regulatory RNA, which plays a key role in the quorum‐sensing‐dependent central regulatory circuit and coordinately regulates several virulence‐associated genes (for a review, see Novick, 2003). The data presented here indicate that the RNAIII‐dependent spa expression is not only repressed at the transcriptional level (Novick et al, 1993; Gao and Stewart, 2004) but also at the post‐transcriptional level via an antisense RNA mechanism. Protein A is one of the major virulence factors, produced by almost all S. aureus clinical isolates, for which a tight regulation of expression might be critical. The membrane protein interferes with Ig‐mediated opsonization (Patel et al, 1987) and is considered as a surface adhesin since it binds to the von Willebrand factor, a large glycoprotein mediating platelet adhesion at sites of endothelial damage (Hartleib et al, 2000; Gomez et al, 2004). Protein A has also the properties of a B‐cell superantigen (Silverman and Goodyear, 2002) and acts as a natural B‐cell toxin to induce the programmed cell death of susceptible VH‐targeted B cells (Goodyear and Silverman, 2004). It is thus possible that during the infection process, the B‐cell superantigen activity of protein A provides a benefit to the bacteria only when it is produced at low level. Conversely, a higher level of stimulation could trigger the innate immune system and may be deleterious for the bacteria. Such efficient downregulation of protein A during the growth cycle and potentially during the course of infection suggests that the amount and timing need to be restricted to a narrow window for optimal efficiency. These observations might be correlated with the fact that the expression of the virulence factors follows a specific temporal program, suggesting that the adhesins and surface proteins are required earlier than the secreted postexponential enzymes, immunotoxins and cytotoxins in the course of the infection (for a review, see Novick, 2003). Notably, the expression of α‐hemolysin, a postexponential exoprotein, is also regulated at both the transcriptional and translational levels (Novick et al, 1993; Morfeldt et al, 1995). This double control seems to be the rule rather than the exception for the expression of the virulence genes in S. aureus.

Transcription regulation of spa expression involves a complex interplay between positive (SarS, SarT, Rot) and negative (SarA) regulatory proteins (Novick, 2003). Based on the analysis of the regulatory elements of the spa promoter, a model was recently proposed for the regulation of spa transcription (Gao and Stewart, 2004). During early exponential phase of growth, the agr system is turned off and as a consequence SarT and Rot become activated. These two regulatory proteins are required for the activation of SarS, which competes with SarA for binding to the spa promoter region to activate transcription (Gao and Stewart, 2004). In this scenario, one possible role of RNAIII would be to modulate the activity of transcriptional activators (Saïd‐Salim et al, 2003). Alternatively, RNAIII could indirectly affect transcription through inhibition of mRNAs that encode transcriptional regulatory factors.

The 3′ domain of RNAIII causes direct repression of spa mRNA translation and induces its rapid degradation in a process that is dependent on the double‐strand‐specific RNase III. RNAIII acts as an antisense RNA and binds to spa mRNA with an association rate constant of 3 × 105 M−1 s−1. Rapid binding is a prerequisite for efficient control in many antisense regulatory systems, and is usually a consequence of a limited number of initial interactions that involve either a loop–loop or unpaired region–loop interactions (for a review, see Wagner et al, 2002). The deletion of the cytosine residues in loop 13 of RNAIII strongly decreased the association rate constant and impaired regulation in vivo. These data suggest that the initial binding involves a restricted number of nucleotides through a putative loop–loop interaction between loop 13 of RNAIII and loop I of spa mRNA (Figure 4). Probing experiments showed that the initial pairings are subsequently converted to extended interactions involving 35 bp interrupted by several bulged loops (Figure 4C). Since RNAIII sequesters the spa ribosome binding site, we propose that the primary effect of RNAIII is to prevent the formation of the translational initiation complex.

We also show that RNase III is essential for spa repression, and is involved in the degradation pathway of spa mRNA. In the wild‐type strain, spa mRNA was not detectable, while in RNAIII‐deleted strain, spa mRNA had a long half‐life, which was even increased in the RNase III‐deficient strain (Figure 6B). In B. subtilis, the initiation of mRNA decay is hindered by a 5′ proximal ribosome (Sharp and Bechhofer, 2003), and hairpin structures at or near the 5′ end interfere with initial degradation (Hambraeus et al, 2002; Condon, 2003). Thus, the presence of the hairpin I structure at the 5′ end of the spa mRNA, overlapping a strong ribosome binding site, may be responsible for its long half‐life. Furthermore, we found that the long hairpin structure II located downstream of the ribosome loading site is an efficient target for RNase III in vitro, as well as the duplex formed by spa mRNA and RNAIII. Hence, the following mechanism can be proposed: at low cell density, spa mRNA is efficiently translated and protected from degradation; at high cell density, RNAIII synthesis is induced, allowing binding to spa mRNA and occlusion of the ribosome binding site. This inhibition of translation favors a rapid degradation of spa mRNA involving RNase III at the initial step. RNase III‐induced cleavages in the formed duplex and in the long hairpin II of spa mRNA may render the mRNA accessible to other endo‐ or exoribonucleases for rapid degradation.

RNase III cleaves many different fully complementary antisense RNA/target mRNA duplexes, but these cleavages cause alterations in message decay rates in only few cases (Krinke and Wulff, 1990; Gerdes et al, 1997). Nevertheless, in most cases, translation inhibition essentially depends on the rate of the formation of RNA duplexes that remain stable for extended time periods. Thus, cleavage by RNase III of those mRNAs that form fully complementary antisense RNAs near or at the ribosome initiation site of the target mRNA do not significantly contribute to the efficiency of control (Case et al, 1990; Wagner et al, 1992). In the present case, S. aureus RNAIII also sequestered the Shine and Dalgarno sequence of the target spa mRNAs, and the association rate constant for complex formation suggested that inhibition might occur rapidly in a way similar to the fully complementary antisense RNAs. However, in vitro experiments indicated that the resulting complexes were moderately stable as compared to irreversible duplexes obtained from fully complementary antisense RNAs. Thus, we inferred that RNase III cleavage is required for rapid degradation and irreversible repression of spa translation.

Unexpectedly, we also showed that S. aureus RNase III specifically recognized the regulatory RNAIII. The functional significance of this interaction in the regulatory pathway of spa mRNA remains elusive. Given the length of RNAIII, one cannot exclude that RNase III might generate processing intermediates with different regulatory functions. However, RNAIII is highly stable in both the wild‐type and RNase III‐deficient strains, and no stable RNase III cleavage product intermediates were detected. Recently, Massé et al (2003) have shown that the RyhB RNA induced by iron starvation or other small noncoding RNAs induced by stress response, are rapidly degraded as a consequence of pairings with their target mRNAs, and thus act stoichiometrically. Since RNAIII is highly structured, RNase III may facilitate the coupled degradation of RNAIII and spa mRNA following complex formation.

Similar to S. aureus RNAIII, in Streptococcus pyogenes (Kreikemeyer et al, 2001; Mangold et al, 2004), and in Clostridium perfringens (Shimizu et al, 2002), small noncoding RNAs, induced by a two‐component system, were shown to be involved in virulence gene expression. More recently, regulatory RNAs have been involved in the quorum‐sensing cascade that regulates virulence in Vibrio cholerae and bioluminescence in Vibrio harveyi (Lenz et al, 2004). These regulatory RNAs, which target specific mRNAs, require the trans‐acting protein Hfq in agreement with the fact that mutations in V. cholerae hfq gene produced strains defective in virulence (Ding et al, 2004). In E. coli, this Sm‐like protein has been proposed to facilitate base pairings in vitro (Møller et al, 2002; Zhang et al, 2002; Geissmann and Touati, 2004), and to protect RNAs against RNase E‐dependent degradation (Massé et al, 2003). We show here that RNAIII is a potential target of S. aureus Hfq in vivo, and that the protein binds more tightly to RNAIII than to the regulatory region of spa mRNA. The protein Hfq has, however, no detectable effect on RNAIII–spa mRNA complex formation in vitro, probably due to the fact that both RNAs interact with a fast association rate constant. Further experiments are underway to define more precisely the function of S. aureus Hfq in virulence, and the possible implication of the protein in conferring a high stability to RNAIII.

In summary, our data reveal an additional direct mRNA target for this unusually long regulatory RNAIII. How many genes are directly regulated by RNAIII and which mechanisms might be associated with them remains to be investigated.

Materials and methods

Strains, plasmids, construction of transcription–translation fusions

S. aureus RN6390, derived from 8325‐4, is our standard agr+ strain (Table I). S. aureus WA400 (ΔrnaIII) is a derivative of 8325‐4, in which the P2 operon is functional but the P3 operon is deleted and replaced by the chloramphenicol transacetylase gene (cat86) (Janzon and Arvidson, 1990). The deletion/replacement Δrnc∷cat mutant of S. aureus RN6390 (LUG774) was obtained by using pMAD, a thermosensitive plasmid (Arnaud et al, 2004). Plasmid pTCV‐lac is a low‐copy‐number promoter‐less lacZ vector used for the study of promoter gene expression in Gram‐positive bacteria (Poyart and Trieu‐Cuot, 1997). A fragment containing the promoter of either the protein A gene (spa) encompassing nucleotides −454 to +1, −454 to +12, or the beta subunit of RNA polymerase gene (rpoB) encompassing nucleotides −480 to +1, were cloned in this vector. For transcriptional/translational fusions, plasmid pTCV‐lac was modified by deletion of a region encompassing the Shine and Dalgarno sequence and the AUG sequence of lacZ. Detailed description is found in Supplementary data.

View this table:
Table 1. Strains and plasmids

Northern blots and measure of mRNA half‐life

Electrophoresis of total RNA was carried out on a 1% agarose gel containing 2.2 M formaldehyde and vacuum transfer to nylon membrane. Hybridizations with specific digoxigenin‐labeled RNA probes and luminescent detection were carried out as described previously (Benito et al, 1998). Staphylococcal cultures at late exponential growth phase were treated with rifampicin (300 μg/ml) and incubated at 37°C. From 0 to 60 min, 1 ml of culture (OD600=1.5) was taken at given times and the RNAs were extracted, and analyzed using specific probes to detect spa mRNA, spa‐lacZ mRNA, RNAIII or 5S rRNA.

β‐galactosidase measurements

S. aureus cells containing lacZ fusions were grown in BHI broth to exponential growth phase, washed and concentrated 10 × (Sambrook et al, 1989). Cells (600 μl) were lysed using a FastPrep® Instrument (QBiogen). Protein concentrations were determined by the Bradford method. β‐Galactosidase specific activity (Miller, 1972) was determined on 100 μl of lysate and expressed in arbitrary units per milligram of protein. All assays were carried out three times on duplicate cultures.

Preparation of Hfq and immunoprecipitation

Recombinant Hfq, fused to maltose binding protein at its N‐terminus, was overexpressed in E. coli and purified by affinity chromatography using an amylose resin. The fusion protein was eluted with 10 mM maltose, and the native Hfq protein was obtained after a specific proteolysis with the Xa factor. After purification on a mono‐Q, Hfq was stored in a buffer containing 50 mM Tris–HCl, pH 7.5, 1 mM EDTA, 250 mM NH4Cl2, 50% glycerol and 2 mM β‐mercaptoethanol. Hfq‐anti‐serum was generated by immunizing two rabbits with purified Hfq protein.

For immunoprecipitation assays, S. aureus crude extract was first passed over a IgG sepharose 6FF column (Pharmacia) to avoid interference with the endogenous protein A. The clarified crude extract (100 μl) was incubated 2 h at 4°C with 5 μl rabbit serum containing S. aureus Hfq‐specific antibody or with 5 μl anti‐mouse antibody, respectively. The antibodies were then bound to 1 ml of protein A sepharose CL‐4B equilibrated with 50 mM Tris–HCl, pH 8. After four washes with 2 ml of buffer containing 50 mM Tris–HCl, pH 8, 150 mM NaCl, 1% Triton X‐100, 0.5% deoxycholate and 0.1% SDS, bound antibodies were eluted with 1 ml of 0.1 M glycine–HCl, pH 3, and neutralized with 100 μl of 1 M Tris–HCl, pH 8. RT–PCR reactions were carried out with two specific primers (RT‐RNAIII5 and RT‐RNAIII6, Supplementary Table S1).

Affinity chromatography using biotinylated RNAIII

In vitro‐transcribed RNAIII (from plasmid pLUG322) and group II intron fragment Δ52XBA (as RNA control) were 3′ end biotinylated using biotin amidocaproyl hydrazide, and then bound to the streptavidin–agarose beads as described previously (Jestin et al, 1997). After incubation of the precleared S. aureus crude extract prepared from RN6390 strain, the proteins bound to the RNA beads were eluted by increasing NaCl concentration, fractionated on SDS–polyacrylamide gel and identified by MALDI TOF mass spectrometry (for details, Supplementary data).

RNase III preparation and RNA hydrolysis

The RNase III coding sequence was amplified from RN6390 gene sequence using primers rnase980/rnase1670 (Supplementary Table S1). The PCR product was digested with BamHI and SmaI, ligated to the linearized pQE30 vector and transformed into E. coli M15. Recombinant RNase III, containing six histidines at its N‐terminus, was purified by affinity chromatography using Ni2+ beads. The purified RNase III was stored in 25 mM Tris–HCl, pH 8.0, 150 mM KCl, 1 mM DTE, 0.1 mM EDTA and 50% glycerol at −20°C.

The 5′‐end‐labeled spa mRNA or RNAIII (3 × 10−8 M) were incubated with a five‐fold excess of the unlabeled RNAIII or spa mRNA, respectively, in 10 μl TMN buffer containing 20 mM Tris acetate, pH 7.5, 10 mM magnesium acetate, 150 mM Na acetate and 1 mM DTT at 37°C for 15 min. Full RNA duplexes between RNAIII and spa mRNA were formed by incubation at 90°C for 2 min followed by slow cooling to 37°C in TMN buffer. Cleavages of end‐labeled RNA, either free or in complex, were performed at 37°C from 30 s to 15 min with 8 pmol of RNase III in the presence of 1 μg of carrier tRNA. Reactions were stopped by phenol extraction followed by RNA precipitation.

Determination of constants of RNAIII/spa complex formation

Binding rate constant of RNAIII/spa complex was measured as described previously (Persson et al, 1988). For details see Supplementary data.

RNA structure probing and toeprinting

Enzymatic hydrolysis and chemical modifications on free or bound RNAs were performed as described previously (Benito et al, 2000). The formation of a simplified translational initiation complex with spa mRNA and the extension inhibition conditions were strictly identical to those described by Benito et al (2000).

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplementary information [emboj7600572-sup-0001.pdf]

Supplementary data [emboj7600572-sup-0002.pdf]


We thank P Fechter, M Possedko, C Gaspin, H Moine, C Brunel for discussions. We are thankful to C Condon, G Wagner and C Isel for discussions and critical reading of the manuscript, to AC Helfer for her skilful technical help and to JC Cortay for technical advises. We are grateful to T Msadek and M Debarbouillé for providing us the pMAD plasmid and to P Trieu‐Cuot and C Poyart for the pTCVlac vector. This work was supported by the Centre National de la Recherche Scientifique, the Ministère de la Recherche (ACI Microbiologie), the Ligue Régionale contre le Cancer and the Fondation pour la Recherche Médicale. SB is supported by a fellowship from the Fondation pour la Recherche Médicale.


View Abstract