RsaC sRNA modulates the oxidative stress response of Staphylococcus aureus during manganese starvation

Abstract

The human opportunistic pathogen Staphylococcus aureus produces numerous small regulatory RNAs (sRNAs) for which functions are still poorly understood. Here, we focused on an atypical and large sRNA called RsaC. Its length varies between different isolates due to the presence of repeated sequences at the 5′ end while its 3′ part is structurally independent and highly conserved. Using MS2-affinity purification coupled with RNA sequencing (MAPS) and quantitative differential proteomics, sodA mRNA was identified as a primary target of RsaC sRNA. SodA is a Mn-dependent superoxide dismutase involved in oxidative stress response. Remarkably, rsaC gene is co-transcribed with the major manganese ABC transporter MntABC and, consequently, RsaC is mainly produced in response to Mn starvation. This 3′UTR-derived sRNA is released from mntABC-RsaC precursor after cleavage by RNase III. The mature and stable form of RsaC inhibits the synthesis of the Mn-containing enzyme SodA synthesis and favors the oxidative stress response mediated by SodM, an alternative SOD enzyme using either Mn or Fe as co-factor. In addition, other putative targets of RsaC are involved in oxidative stress (ROS and NOS) and metal homeostasis (Fe and Zn). Consequently, RsaC may balance two interconnected defensive responses, i.e. oxidative stress and metal-dependent nutritional immunity.

INTRODUCTION

Staphylococcus aureus is a major human opportunistic pathogen responsible for a wide variety of infections, from superficial to deep-seated infections with high mortality (e.g. infective endocarditis, osteomyelitis and necrotizing pneumonia). Professional phagocytes (i.e. neutrophils and macrophages) represent the first line of host defence and are important components of innate immunity against S. aureus (1,2). After migration to the infection site, immune effectors deploy an arsenal of offensive and defensive strategies including phagocytosis, prevention of S. aureus dissemination by extracellular traps (ETs) (3), and synthesis of a battery of antimicrobial effectors such as reactive oxygen species (ROS) (1). ROS include superoxide anion (O₂^.−), hydrogen peroxide (H₂O₂) and hydroxyl radical (^.OH), which are also natural products of metabolism formed when oxygen becomes partially reduced during aerobic respiration (4). They have the potential to damage nucleic acids, proteins and lipids. Neutrophils also locally secrete the chelating agent calprotectin to sequester critical ions for bacterial growth such as manganese (Mn) and zinc (Zn) (5).

To evade immune defences and colonize numerous organs, S. aureus produces a wide range of virulence factors and stress-response proteins. They notably impede phagocytosis, escape ETs, extract nutrients from host cells and detoxify ROS (2,3). For example, superoxide dismutases (SOD) and catalases are produced to convert O₂^.− to H₂O₂, and then to water and oxygen (6). SODs are metalloenzymes classified according to their loaded cofactor: copper/zinc (Cu/Zn-SOD), manganese (Mn-SOD), nickel (Ni-SOD) or iron (Fe-SOD) (7). In S. aureus, two SODs are involved in oxidative response (8,9). The major SOD enzyme SodA is strictly Mn-dependent while the cambialistic enzyme SodM is active when loaded with Mn or Fe and is therefore important in manganese limiting conditions (10). Remarkably, SodM is specific to S. aureus as coagulase-negative staphylococci lack sodM gene (9,10). There are conflicting reports regarding the role of SODs in virulence. An early study suggested no correlation between SOD activity and lethality in a mouse infection model (11). However, SOD activity was found higher in S. aureus strains isolated from infected patients (12). Deletion of sodAM reduced virulence in abscess or retro-orbital infection models (5,13). More recently, the contribution of each Sod enzyme was deciphered during mice infection (10). SodA plays a crucial role when Mn is abundant while SodM becomes the main enzyme assuring the resistance to oxidative stress when Mn is scarce. Nevertheless, the regulation of their synthesis is still poorly understood.

The production of S. aureus virulence factors is regulated by a combination of transcriptional regulators and regulatory RNAs (sRNAs) (14). Many sRNAs form base-pairings with their mRNA targets to control their expression and play key functions in adaptive processes and virulence (15). The most studied sRNA in S. aureus is RNAIII, one of the main intracellular effectors of the quorum sensing agr system. RNAIII represses the expression of virulence factors such as protein A, immunoglobulin-binding protein Sbi, coagulase, and of the repressor of toxins Rot while it activates the synthesis of hemolysin α (16–21). In this study, we have analyzed the function of another very peculiar and large sRNA, named RsaC, which was found highly expressed in acute osteomyelitis, a bone infection usually caused by S. aureus (22). Here, we identified sodA mRNA as a primary target of the 3′-UTR derived sRNA RsaC, which is mainly produced upon Mn starvation. By repressing the translation of the Mn-containing enzyme SodA, RsaC avoids the synthesis of a non-functional enzyme. Concurrently, RsaC indirectly enhances the oxidative stress response mediated by the cambialistic enzyme SodM, which in absence of manganese, uses Fe as a cofactor to restore the ROS detoxification pathway. We also demonstrated that RsaC might have a broader role in oxidative stress response (ROS and NOS) and in metal homeostasis (Mn, Fe and Zn). The consequences of this regulation on S. aureus pathogenicity will be discussed.

MATERIALS AND METHODS

Strains, plasmids and growth conditions

Staphylococcus aureus strains and plasmids used in this study are listed in Supplementary Table S1. Escherichia coli strain DH5α (Promega) was used as host strain for plasmid constructions. Staphylococcus aureus strain RN4220 (23) or E. strain DC10B (24) were used as host strains for plasmid amplification before electroporation in S. aureus HG001. Transformation of E. coli DH5α or DC10B was performed by heat shock and S. aureus strains were transformed by electroporation (Bio-Rad gene Pulser).

Escherichia coli strains were cultivated in lysogeny broth (LB). Staphylococcus aureus strains were isolated on glucose peptone (GP), tryptic soy (TS) or blood agar plates. Overnight cultures in brain-heart infusion (BHI) or in tryptic soy broth (TSB) ± erythromycin (5 μg/ml) were diluted 50-fold in fresh medium and grown at 37°C (180 rpm, 5:1 flask-to-medium ratio). To induce Mn starvation, cells were grown in BHI-chelex medium as followed. We added 5% Chelex-100 resin (Sigma) to BHI medium and mix 6h at 20°C. The medium was sterile filtered and, according to Kehl-Fie et al. (25), we complemented it with metal ions at the following concentrations: 100 μM CaCl₂, 25 μM ZnCl₂, 1 mM MgCl₂, 1 μM FeSO₄. As a control, BHI-chelex medium was also supplemented with 25 μM MnCl₂.

Deletion of rsaC₁₁₁₆ gene (ΔrsaC) in HG001 strain was performed using pMAD (26). Chromosomal regions upstream and downstream of the rsaC sequence were amplified by PCR (RsaC-PCR1-Rev/RsaC-PCR1-For and RsaC-PCR2-Rev/RsaC-PCR2-For, respectively) (Supplementary Table S2) and then cloned into pMAD vector. The resulting plasmid (pLUG1110) was electroporated into RN4220 recipient strain and then transferred to HG001. Growth at non-permissive temperature (44°C) was followed by several subcultures at 30°C and 37°C to favor double crossing over as previously described (26). As control, we verified that deletion of rsaC does not alter the expression of surrounding genes. Indeed, rsaC gene is located between mntC and nhaK_1 genes, which encodes a subunit of a manganese ABC transporter and a proton pump, respectively (Supplementary Figure S1A). Using qRT-PCR experiments, we have verified that no difference existed in the expression level of downstream and upstream genes between wild type and mutant strains (Supplementary Figure S1B). To complement ΔrsaC strain (rsaC+), rsaC₁₁₁₆ gene was amplified using PstI-RsaC₁₁₁₆-For/BamHI-RsaC₁₁₁₆-Rev oligonucleotides (Supplementary Table S2). The amplicon RsaC₁₁₁₆ was digested by PstI and BamHI restriction enzymes and inserted into a PstI/BamHI digested pCN51-P3 plasmid. The quorum sensing-dependent P3 promoter allows an overexpression of the downstream gene in stationary phase of growth. To generate the RsaC₁₁₁₆mut construct (deletion of nucleotides +1000 to +1006), two independent PCR reactions were performed with the following oligonucleotides: PstI-RsaC₁₁₁₆-For/RsaCmut-Rv and RsaCmut-For/BamHI-RsaC-Rev. The two PCR products were then mixed to serve as template for a third PCR (PstI-RsaC₁₁₁₆-For/BamHI-RsaC₁₁₁₆-Rev). The amplicon RsaC₁₁₁₆mut was digested by PstI and BamHI restriction enzymes and inserted into a PstI/BamHI digested pCN51-P3 plasmid. To obtain a pJET-T7-sodA, a PCR fragment was amplified using oligos T7-sodA-For/T7-sodA-Rev (Supplementary Table S2) and then cloned into pJET1.2/blunt vector (ThermoFisher). For MAPS experiment, MS2 aptamer was fused to the 5′ end of RsaC₅₄₄ using oligonucleotides PstI-MS2-RsaC-For and BamHI-RsaC-Rev (Supplementary Table S2). Obtained fragment was cloned into PstI/BamHI-digested pCN51-P3. Total DNA and plasmid DNA were prepared using DNAeasy tissue Kit and Qiaprep Miniprep respectively (Qiagen).

Northern blot analysis

Total RNAs were prepared from different volumes of S. aureus cultures taken at 2 h (OD_600nm∼0.4; 15 ml), 4 h (OD_600 nm ∼2.5; 10 ml) and 6 h (OD_600 nm ∼5; 5 ml). After centrifugation, bacterial pellets were resuspended in RNA Pro Solution (MP Biomedicals). Lysis was performed with FastPrep apparatus (MP Biomedicals). RNA purification followed strictly the procedure described for the FastRNA Pro Blue Kit (MP Biomedicals).

Electrophoresis of total RNA (10–20 μg) was performed on 1% agarose gel containing 25 mM guanidium thiocyanate. After migration, RNAs were vacuum transferred on Hybond N+ nitrocellulose membrane (GE Healthcare Life Sciences). Hybridization with specific digoxygenin (DIG)-labelled probe complementary to RsaC or 5S sequence followed by luminescent detection were carried out as described previously (16). For mntA, we used a radiolabelled DNA probe (Supplementary Table S2). The labeling of the 5′ end of the oligonucleotide was performed with T4 polynucleotide kinase (Fermentas) and [γ³²P] ATP. For determination of molecular weights, we used the Dynamarker® Prestain Marker for RNA high (CliniSciences).

MS2-affinity purification coupled with RNA sequencing (MAPS)

The MS2 aptamer, which is recognized by the MS2 coat protein, was fused to the 5′ end of RsaC_544. Here, we used a shorter form of RsaC due to nonspecific interaction of RsaC₁₁₁₆ with the affinity column. RsaC₅₄₄ construct carries the whole 3′ domain of RsaC (447 nt) and 97 nucleotides from the 5′ part, which enables to preserve the secondary structure of the 3′ domain of RsaC (Figure 1B). This construct was expressed in vivo under the control of the quorum sensing-dependent promoter P3 and confirmed as functional when compared to the full-length RsaC₁₁₁₆ and MS2-RsaC₁₁₁₆ (Supplementary Figure S2). Indeed, the overexpression of the three RNAs induces similar ROS accumulation, revealing that the 3′ domain of RsaC is sufficient to modulate the oxidative stress response.

Figure 1.

RsaC is a large and highly structured sRNA. (A) Northern blot analysis of RsaC sRNA in three different S. aureus strains (HG001, MW2 and Newman). Total RNA was extracted after 2, 4 and 6 h of growth in BHI at 37°C. 5S rRNA was used as loading control. Data are representative of three independent experiments. (B) Secondary structure model of RsaC. The structure of RsaC (from HG001 strain) was mapped using selective 2′-hydroxyl acylation by benzol cyanide and the reactivity was analyzed by primer extension (SHAPE). Data were analyzed, and each modification was quantified using QuSHAPE software. The secondary structure was drawn using RNA structure and VARNA considering the reactivity of each ribose. The regions, which were not mapped, are shown in blue. The color code for each reactivity is given on the right side of the secondary structure model, red is for the highest reactivity while green is for the lowest reactivity. A black arrow indicates the +1 of RsaC₅₄₄.

Crude extracts from ΔrsaC strain expressing MS2-RsaC₅₄₄ or from WT strain expressing MS2 tag alone (control) were harvested after 6 h of growth at 37°C in BHI medium. MS2-affinity purifications were performed as previously described (27,28). All steps were performed at 4°C. Poly-Prep chromatography column (Biorad) containing amylose resin (300 μl; NEB) were washed three times with 10 ml of Buffer A (20 mM Tris–HCl pH 8, 150 mM KCl, 1 mM MgCl₂, 1 mM DTT), before addition of 3600 pmol of MS2-MBP. Crude bacterial extracts (4 ml) were directly loaded, followed by three washes with 10 ml of Buffer A. Finally, RNAs were eluted from the column with 1 ml of Buffer E (20 mM Tris–HCl pH 8, 150 mM KCl, 1 mM MgCl₂, 1 mM DTT, 0,1% Triton X-100, 12 mM maltose). Eluted RNAs were extracted with phenol:chloroform:isoamylalcohol (25:24:1 (v/v), Roth) and precipitated with 3 volumes of cold absolute ethanol in the presence of 0.3 M sodium acetate. RNA samples were then treated with DNase I prior to RNA-seq analysis. MS2-affinity purifications were performed in duplicates.

RNA quality and quantity assessments were performed on Agilent Nano Chip on Bioanalyzer 2100. RNA samples were ribo-depleted (Ribo-Zero rRNA Removal Kit (Bacteria) Illumina) and cDNA libraries were prepared using the Random Hexamer approach and sequenced by Fasteris (Switzerland). Libraries were sequenced using Hi-Seq 2500 system (Illumina). RNA-seq analysis was performed according to Lalaouna et al. (2018) (27). Reads were processed and aligned on HG001 genome (29) using the Galaxy platform (30). We used DEseq2 to estimate enrichment values for MAPS (n = 2) (P-value < 0.05; Fold change (FC) > 2) (Table 1). The whole list of RNA enriched with MS2-RsaC₅₄₄ is available in Supplementary Table S3. MAPS data that support our findings are available in the GEO database with the accession code GSE128741.

Table 1.

List of RNAs significantly co-purified with MS2-RsaC₅₄₄

ID	Gene name	Function	Fold change MS2-RsaC/Ctrl	P-value
HG001_00569	-	RsaC sRNA	9.981	3.33E-23
HG001_01477	sodA	Superoxide dismutase SodA	8.088	2.44E-18
HG001_01627	gloB_1	Hydroxyacylglutathione hydrolase	5.406	4.20E-13
HG001_02152	fba	Fructose-bisphosphate aldolase	4.082	4.84E-10
HG001_00556	sarA	Transcriptional regulator SarA	3.156	1.88E-06
HG001_00934	ptsI	Phosphoenolpyruvate-protein phosphotransferase	3.122	5.80E-07
HG001_00252	-	FtsX-like permease family protein	3.093	5.72E-07
HG001_02633	envR	DNA-binding transcriptional regulator EnvR	2.877	7.08E-06
HG001_00671	fecD_1	Fe(3+) dicitrate transport system permease protein FecD	2.721	4.27E-05
HG001_00933	ptsH	Phosphocarrier protein HPr	2.692	1.30E-05
HG001_01479	znuB	High-affinity zinc uptake system membrane protein ZnuB	2.642	4.16E-05
HG001_00967	-	Cysteine-rich secretory protein family protein	2.613	2.67E-05
HG001_01416	srrB	Sensor protein SrrB	2.576	2.71E-05
HG001_02376	-	Hypothetical protein	2.491	8.80E-05
HG001_01478	zur	Zinc-specific metallo-regulatory protein Zur	2.446	0.000206
HG001_01220	msrR	Regulatory protein MsrR	2.443	9.65E-05
HG001_01721	menE	2-succinylbenzoate–CoA ligase	2.351	0.000382
HG001_02519	-	Hypothetical protein	2.222	0.000486
HG001_00968	-	Hypothetical protein	2.21	0.000569
HG001_01480	znuC	High-affinity zinc uptake system ATP-binding protein ZnuC	2.176	0.001228
HG001_02411	-	hypothetical protein	2.146	0.001107
HG001_01710	yvgN	Glyoxal reductase	2.124	0.001455
HG001_01657	ccpA	Catabolite control protein A	2.101	0.001521
HG001_00540	ywqN_2	Putative NAD(P)H-dependent FMN-containing oxidoreductase YwqN	2.065	0.001444
HG001_01626	nhaX	Stress response protein NhaX	2.056	0.001481
HG001_01845	lytN_3	putative cell wall hydrolase LytN precursor	2.056	0.001999
HG001_01925	cobQ	Cobyric acid synthase	2.035	0.002683
HG001_00932	-	Hypothetical protein	2.029	0.001941
HG001_02069	rex	Redox-sensing transcriptional repressor Rex	2.023	0.001773
HG001_00568	nhaK_1	Sodium, potassium, lithium and rubidium/H(+) antiporter	2.005	0.002142
HG001_02589	-	Virus attachment protein p12 family protein	2.002	0.004538

Differential proteomic analysis on cytoplasmic proteins

Label free spectral count analysis was performed in triplicate on cytoplasmic proteins extracts prepared from tested strains using nanoLC–MS/MS. Cells were harvested after 6 h of growth in BHI (Supplementary Table S4) or in BHI-chelex supplemented or not with 25 μM of MnCl₂ (Supplementary Table S5). Proteins were extracted from cytoplasm of tested strains and were digested with sequencing-grade trypsin (Promega, Fitchburg, MA, USA) as previously described (28). The samples were analyzed by nanoLC-MS/MS either on a NanoLC-2DPlus system (with nanoFlex ChiP module; Eksigent, Sciex, Concord) coupled to a TripleTOF 5600 mass spectrometer (AB Sciex) or on a QExactive+ mass spectrometer coupled to an EASY-nanoLC-1000 (Thermo-Fisher Scientific, USA). Data were searched against S. aureus HG001 database using Mascot algorithm (version 2.5, Matrix Science, London, UK) through ProlineStudio 1.4 package (http://proline.profiproteomic.fr/). Peptides were validated on Mascot pretty rank equal to 1, an ion score cut-off equal to 25, and 1% FDR on both peptide spectrum matches (PSM) and protein sets (false discovery rate-based ion score). The total number of spectra was recorded for all the proteins, allowing to quantify these proteins across all the samples (quantification by spectral count). Data were submitted to a negative-binomial test using an edgeR GLM regression through R (R v3.5.0), normalized according to a median-to-ratio method to calculate fold changes (FC) and P-values. The mass spectrometric data that support our findings are available via ProteomeXchange with identifier PXD013225.

Preparation of RNAs for in vitro experiments

PCR fragments containing T7-RsaC₅₄₄, T7-RsaC₅₄₄mut, T7-ldh1 (from –167 to +418), T7-rex (from –103 to +205), T7-sarA (full-length; from –147 to +375), T7-sodM (full-length; from –28 to +651), T7-sufC (full-length; from –100 to +762), T7-sufD (from –97 to +626), T7-znuB (full-length; from –60 to +786), T7-znuC (from –22 to +646) and T7-zur (full-length; from –212 to 412) were used for in vitro transcription with T7 RNA polymerase (Supplementary Table S2). We used HG001 genomic DNA as DNA template for most constructs. For T7-RsaC₅₄₄ and T7-RsaC₅₄₄mut, we used pCN51-P3-MS2-RsaC₅₄₄ and pCN51-P3-RsaC₁₁₁₆mut, respectively. The transcription of full-length mRNA sodA was performed from a DraI/XhoI-digested pJET-T7-sodA plasmid (from –28 to +600). RNAs were then purified using a 6% polyacrylamide–8 M urea gel electrophoresis. After overnight elution in Elution Buffer (0.5 M ammonium acetate pH 6.5, 1 mM EDTA and 0.1% SDS), RNAs were precipitated in cold absolute ethanol and washed with 80% ethanol. RNAs were dephosphorylated using Fast AP (ThermoFisher), according to the manufacturer protocol. The labeling of the 5′ end of dephosphorylated RNA was performed with T4 polynucleotide kinase (Fermentas) and [γ³²P] ATP as previously described (16).

Gel retardation assays

5′-radiolabeled RsaC₅₄₄ (15000 cpm/sample, concentration <1 pM) and cold mRNAs were renaturated separately by incubation at 90°C for 1 min in Buffer GR- (20 mM Tris–HCl pH 7.5, 60 mM KCl, 40 mM NH₄Cl, 3 mM DTT), cooled 1 min on ice, and incubated 15 min at RT in presence of 10 mM MgCl₂. For each experiment, increasing concentrations of cold mRNA were added to the 5′ end labelled RsaC₅₄₄ in a total volume of 10 μl containing Buffer GR+ (20 mM Tris–HCl pH 7.5, 60 mM KCl, 40 mM NH₄Cl, 3 mM DTT, 10 mM MgCl₂). Complex formation was performed at 37°C during 15 min. After incubation, 10 μl of glycerol blue was added and samples were loaded on a 6% polyacrylamide gel containing 10 mM MgCl₂ under non-denaturing conditions (8 h, 300 V, 4°C). Quantification of data corresponding to RsaC₅₄₄ and to RsaC₅₄₄/mRNA complex was done with ImageQuant TL software (GE Healthcare Life Sciences). Under these conditions where the concentration of the labeled RNA is negligible, the Kd dissociation constant can be evaluated as the concentration of the cold RNA that showed 50% of binding. The same protocol was used with 5′-radiolabeled RsaC₅₄₄mut.

Toe-printing assays

The preparation of S. aureus 30S ribosomal subunits, the formation of a simplified translational initiation complex (constituted by S. aureus 30S ribosomal subunit, initiator tRNA^fMet and sodA mRNA), and the extension inhibition conditions were performed as described previously (31). Increasing concentrations of RsaC₅₄₄ (25–400 nM) were used to monitor their effects on the formation of the initiation complex (250 nM of 30S S. aureus; 4 μM of tRNA^fMet) with sodA mRNA (50 nM).

Primer extension assays

Primer extension assays were performed using 30 μg of total RNA, incubated with 5′-radiolabeled oligonucleotide (PE-RsaC-Rev, Supplementary Table S2), 2.5 mM dNTPs and AMV reverse transcriptase (4 units, NEB). RNA template was removed by addition of 3 µl of 3 M KOH and 20 µl of Destroy Buffer (50 mM Tris–HCl pH 8.0, 0.5% SDS, 7.5 mM EDTA) for 3 min at 90°C, followed by 1 h at 37°C. Samples were finally precipitated and migrated on a denaturing 10% polyacrylamide gel, next to sequencing ladder. The sequencing ladder was obtained with a DNA template (PCR with oligonucleotides PE-RsaC-For and PE-RsaC-Rev, from genomic DNA). Here, ddNTP were added to stop the reaction performed by the Vent (exo-) DNA polymerase (2 units, NEB). As an example, the A reaction is composed of 0.25 mM ddATP, 0.025 mM dATP, 0.05 mM dGTP, 0.05 mM dCTP and 0.05 mM dTTP. 0.1% Triton was also added to the reaction. PCR was performed as follows: 1 min at 95°C, 1 min at 52°C and 1 min at 72°C (25 cycles). The primer PE-RsaC-Rev is located at nucleotides +42 to +59 (from the +1 of RsaC).

Footprinting assays

RsaC₅₄₄-sodA mRNA complex formation was carried out at 37°C for 15 min in Native Buffer (20 mM Tris–HCl pH 7.5, 10 mM MgCl₂ and 150 mM KCl). Enzymatic hydrolysis was performed on the unlabelled and renatured RsaC sRNA (50 nM) in absence or in presence of sodA mRNA (75, 150 or 300 nM). The reaction was performed in Native Buffer with RNase V1 (0.5 U) or with RNase T2 (0.05 U) at 20°C for 5 min in presence of 1 μg of total tRNA. The reactions were stopped by phenol extraction followed by RNA precipitation. The enzymatic cleavages in RsaC were detected by primer extension with AMV reverse transcriptase according to Fechter et al. (32).

Determination of RNA structure with SHAPE

To elucidate the secondary structure of RsaC sRNA, SHAPE reactivity assays were performed as previously described by Rice et al. (2014) (33). RNA folding was probed with benzoyl cyanide (BzCN), which modified 2′-hydroxyl of riboses. 1 pmol of RNA was denatured and renatured in SHAPE-BzCN buffer with 2 μg of yeast tRNA. Then, BzCN was added with different final concentrations (10-50-150 mM). After 1 min at 20°C, RNA was precipitated. Thereafter, reverse transcription was performed with VIC-labelled oligonucleotides (Supplementary Table S2). In parallel, ddGTP sequencing of 1 pmole RNA was made with the same oligonucleotide labelled with NED fluorophore. cDNA samples were analysed by capillary electrophoresis and data processed by QuSHAPE software (34). SHAPE data (Supplementary Table S6) were used to constrain structure prediction by RNAstructure (35). The secondary structure of RsaC was finally visualized using VARNA (36) (Figure 1B).

RT-PCR

RT-PCR was performed using OneStep RT-PCR Kit (Qiagen) according to the manufacturer's instructions. Total RNA extracted after 4 or 6 h of growth in BHI medium was used as template (after DNase I treatment). Primers are described in Supplementary Table S2. The same experiment was performed without RT enzyme as a control.

Determination of the 5′ end status of RsaC

To determine if the 5′ end of RsaC is mono- or tri-phosphorylated, we used Terminator™ 5′-Phosphate-Dependent Exonuclease Terminator (Epicentre) according to manufacturer's protocol. Total RNA (10 μg) was incubated with Terminator Exonuclease (1U) and Terminator 1X Reaction Buffer A at 30°C for 1 h. Subsequently, we performed phenol extraction and ethanol precipitation. Northern blot analysis was performed on 1% agarose gel containing 25 mM guanidium thiocyanate.

Superoxide stress assays

Sensitivity to methyl viologen (MV) was assessed by viability test (13). Fresh BHI broth was inoculated with an overnight culture to an initial OD_600nm of 0.05 and cultivated at 37°C with shaking (200 rpm). Cells were challenged with MV (10 mM) in early-exponential phase (2 h). Samples were collected at appropriate intervals post-challenge. Viability of bacteria was determined by plating onto GP agar supplemented with 5 μg/ml erythromycin. Experiments were performed in triplicate.

SOD activity assays

SOD activity assays were performed as previously described (37). Basically, when exposed to light, riboflavin produces O₂^.−, which reduces tetrazolium to the dark blue formazan dye. The presence of SOD enzymes hinders this reaction and can be visualized by an achromatic area (negative staining). Samples were collected after 8 h of growth in absence or presence of methyl viologen (10 mM; BHI) or after 6h in BHI-chelex ±25 μM MnCl₂. Cells were then washed in PBS and lysed (8). Total protein concentration was determined by Bradford assay (Bio-Rad). SodA and SodM activities were determined by negative staining on native polyacrylamide gel (9). Equal amounts of protein (20 μg) were resolved on 12% polyacrylamide gels by electrophoresis in glycine buffer without sodium dodecyl sulfate. After the run, the gel was soaked in 1.225 mM of nitro blue tetrazolium (NBT) solution for 45 min and then washed with sterile distilled water. Subsequently, the gel was soaked in a solution containing 0.028 mM riboflavin and 28 mM TEMED (tetramethylethylene diamine) for another 45 min. The gel was exposed to light to initiate the photochemical reaction. The SOD activity was monitored as a clear zone surrounded by a dark blue background.

Endogenous ROS production

Cells were harvested (1 ml at OD_600nm = 1) after 6h of growth and washed twice with PBS. The probe 2′,7′-dichlorofluorescein diacetate (Sigma) was then added to cell suspensions to a final concentration of 10 μM. Fluorescence intensity (FI) was acquired during 5 h using 96-well spectrofluorimeter (Ex/Em = 485/538 nm). The FI value was normalized to the OD_600 nm of each culture. The experiment was performed in triplicate.

RESULTS

RsaC is a long and structured RNA whose length varies between isolates

We first analyzed by Northern blot the expression of RsaC in three different strains (HG001, MW2 and Newman) during growth in BHI at 37°C (Figure 1A). RsaC is mainly produced in stationary phase and its length varies depending on tested strains. To understand this variation, the 5′ end of the longest form of RsaC sRNA (in HG001) was mapped using primer extension. The experiment shows a RT stop corresponding to a fragment with a length of 1116 nt (Supplementary Figure S3).

By performing a comparative multiple sequence alignment with MW2 and Newman strains, we observed that the 3′ domain of RsaC (comprising 447 nt) is highly conserved (>95%) (Supplementary Figure S4A), while its 5′ domain differs in strains due to variation in the number of repeats (from 2 to 4; 133 nt) (Supplementary Figure S4B). A truncated repeat of 58 nt marks the end of the 5′ part of RsaC. Therefore, RsaC is 850 nt-long (2 repeats) in Newman, 984 nt-long (3 repeats) in MW2 and 1116 nt-long (longest form; four repeats) in HG001. We expanded our analysis to all bacterial genomes using a sequence similarity search program (BLASTn; Supplementary Table S7) and observed that RsaC is restricted to S. aureus and the closely related S. argenteus species. The length of RsaC sRNA is ranging from 584 nt (a single truncated repeat) to 1116 nt (four repeats and a truncated repeat). This highlights the wide variability in length and composition of the 5′ part of RsaC.

We probed the secondary structure of RsaC (from HG001 strain) using selective 2′-hydroxyl acylation of riboses with benzoyl cyanide (BzCN) (33). Resulting 2′-O-adducts were detected by primer extension and data were analysed and quantified using QuSHAPE software (34). The BzCN reactivity (Supplementary Table S6) provides information on the local nucleotide flexibility and the highest reactivities of riboses are indeed mainly located in unpaired regions. The proposed secondary structure model, which takes into account the ribose reactivities, shows that RsaC is a highly structured sRNA characterized by two large and independent domains (Figure 1B). The 5′ part is constituted of four repeat motifs, each of them carrying four distinct well-characterized stem-loop structures. The region from nucleotides +920 to +1070 was also probed with various ribonucleases specific either for unpaired nucleotides (RNase T2) or for double-stranded regions (RNase V1). These data support the proposed secondary structure model since the RNase T2 cleavages were preferentially located in the apical loops while major RNase V1 cuts occurred in paired regions (Figure 2D). The 3′ part of RsaC is characterized by two large hairpin structures interrupted by internal loops and by several interhelical unpaired regions (Figure 1B).

Figure 2.

RsaC binds to sodA mRNA and blocks its translation. (A) Gel retardation assays using RsaC₅₄₄ sRNA and full-length sodA mRNA. 5′ end-radiolabeled RsaC₅₄₄ (*) was incubated with increasing concentrations of sodA mRNA (from 0 to 100 nM). (B) The RsaC:sodA pairing site predicted using IntaRNA software is indicated in red. The AUG start codon is in bold. Deleted nucleotides (Δ_1000–1006) in RsaCmut sequence are underlined. (C) Footprint of sodA mRNA on RsaC. Incubation controls (Ctrl) were performed in absence of RNases on RsaC either free (–) or bound to sodA (50 nM, +). Reactions with either RNase V1 (V1) or RNase T2 (T2) were performed with RsaC (50 nM) in absence (lanes 1 and 5) or in presence of increasing concentrations of sodA mRNA (lanes 2–4 and 6–8; 75, 150 or 300 nM, respectively). U, C, A: sequencing ladders. *Lane 7 should not be considered due to a technical glitch. On the left side, a bar denotes changes in cleavages as the result of sodA binding. (D) Schematic representation of sodA pairing site on RsaC sRNA revealed by footprinting assays. RNase V1 cleavages are denoted by grey triangles for strong cut and empty triangles for moderate cut. RNase T2 cleavages are shown by filled arrows for strong cut and dashed arrows for weak cut. New RNase V1 cleavages are labelled by stars. The predicted pairing site by IntaRNA software is indicated in red. (E) Toe-printing assay monitoring the effect of RsaC on the formation of the ternary ribosomal initiation complex comprising S. aureus 30S subunit, initiator tRNA^fMet and sodA mRNA (50 nM). Lanes 1–4: incubation controls (lane 1: sodA mRNA alone; lane 2: sodA mRNA with 30S subunit; lane 3: sodA mRNA with 30S subunit and initiator tRNA^fMet (tRNAi); lane 4: sodA mRNA with RsaC); lanes 5–8: formation of the initiation complex in presence of increasing concentrations of RsaC (25, 50, 100, 400 nM). U, C, G, A: sequencing ladders; SD: Shine-Dalgarno sequence; +1: start codon; +5: RT stop induced by RsaC binding; +16: toe-printing signal. Data are representative of two independent experiments.

Characterization of RsaC targetome

To identify RsaC targets, we performed MS2-affinity purification coupled with RNA sequencing (MAPS) in ΔrsaC background complemented with MS2-RsaC₅₄₄. Computer prediction suggested that this 3′ part of RsaC forms an independent and conserved domain, which would adopt the same structure as in full-length RsaC. The MS2-RsaC₅₄₄ construct was expressed in vivo under the control of the quorum sensing-dependent promoter P3 and was confirmed as functional (Supplementary Figure S2). Crude extracts from ΔrsaC strain expressing MS2-RsaC₅₄₄ or from WT strain expressing MS2 alone were harvested after 6h of growth in BHI medium. After MS2-affinity purification and RNAseq analysis, we identified 31 RNAs significantly enriched with MS2-RsaC₅₄₄ (Fold change > 2, P-value < 0.05; Table 1). The best enrichment was for sodA mRNA (reads obtained with MS2-RsaC₅₄₄ versus MS2 control), which encodes a superoxide dismutase. Moreover, other candidates code for proteins involved in oxidative stress response (e.g. sarA and rex), metal homeostasis (e.g. fecD_1, znuB, zur and znuC) and carbon metabolism (e.g. fba, ptsI, ptsH and ccpA).

To unravel RsaC effect on identified candidates, differential proteomic analysis was performed on cytoplasmic proteins extracts prepared from WT and ΔrsaC strains using LC–MS/MS coupled to spectral count analysis. Using stringent criteria (Supplementary Table S4), we did not observe major changes (Fold change <0.5 or >2). However, the level of SodA was found slightly but reproducibly decreased in WT strain as compared to ΔrsaC strain. These minor effects suggested that the expression of RsaC under tested conditions (i.e. rich medium conditions) is not optimal to reveal the full spectra of its regulatory activities.

The combination of data suggests that sodA mRNA is one of the major targets of RsaC and that its regulation would occur at the translational level. Indeed, using qRT-PCR (Supplementary Figure S5), we demonstrated that the steady-state level of sodA mRNA remains equal in WT and mutant strains as well as in ΔrsaC strain complemented with a plasmid overexpressing RsaC (∼10-fold). In the following, we focused our study on sodA mRNA target.

RsaC blocks translation initiation of sodA mRNA

Using gel retardation assays, we first demonstrated that the 5′ end radiolabeled RsaC₅₄₄ forms a stable complex with sodA mRNA in vitro (Kd ≈24 nM) (Figure 2A). This result correlates well with the fact that the seed region of RsaC is located in its 3′ part (Figure 2B) and is accessible for efficient binding (Figures 1B and 2D). Indeed, the in silico prediction of the pairing site (Figure 2B) identified base pairing complementarities involving nucleotides +994 to +1009 of RsaC and –9 to +7 of sodA mRNA.

In agreement with the in silico prediction, enzymatic footprinting experiments (Figure 2C) demonstrated that binding of sodA mRNA induces changes in the unpaired region of RsaC encompassing the predicted pairing site (from A+994 to G+1009, Figure 2C and D). Our data support the existence of a double-stranded region with sodA as revealed by the appearance of new RNase V1 cleavages at positions A+1003, U+1004 and U+1007 (Figure 2D). Concomitantly, nucleotides A+1001 to A+1006 are found protected against RNase T2 hydrolysis. Remarkably, nucleotides A+999 to A+1006 are highly modified by BzCN (Figure 1B), indicating that this region is accessible for efficient pairing with sodA mRNA. Additional changes in the enzymatic cleavage patterns were also observed in distant regions of RsaC including two surrounding and accessible apical loops, encompassing nucleotides +950 to +975, and +1015 to +1030 (Figure 2D and Supplementary Figure S6A). Whether these effects might result from conformational changes of RsaC or from the formation of additional base pairing interactions with sodA mRNA, remained to be demonstrated. Notably, base pairing complementarities are predicted between nucleotides -26 to -23 at the 5′ end of sodA with nucleotides +1027 to +1030, and between nucleotides +362 to +367 in the coding sequence of sodA with nucleotides +957 to +962 of RsaC (Supplementary Figure S6B).

To further probe the importance of the pairing site of RsaC, we deleted the nucleotides +1000 to +1006 in RsaC (RsaCmut; as indicated in Figure 2B) and analyzed the ability of RsaCmut to form a complex with sodA (Supplementary Figure S7). Data showed that RsaC mutation completely abolished RsaC:sodA complex formation. This demonstrates that the sequestration of the initiation codon of sodA is the critical site, which confers high stability to the complex.

Because RsaC sequesters the AUG initiation codon of sodA mRNA, we further addressed the consequences of RsaC binding on sodA translation initiation. Toe-printing assays were used to monitor the formation of the ternary ribosomal initiation complex (Figure 2E). We observed that RsaC efficiently hinders translation initiation to the SD sequence of sodA mRNA. Indeed, the toe-print signal (+16) fades when RsaC and sodA are equimolar and completely disappears when RsaC is supernumerary (Figure 2E, lanes 4–8). Simultaneously, we noticed a RT stop at nucleotide +5 in presence of RsaC, hinting that RsaC pairs near the AUG start codon

All in all, our data showed that RsaC binds to the translation initiation site of sodA mRNA and, consequently, prevents the recruitment of ribosomes to inhibit translation initiation.

RsaC modulates oxidative stress resistance

As SodA is involved in oxidative stress resistance, we analyzed if the expression of RsaC could influence it. Hence, WT/pCN51-P3 (WT), ΔrsaC/pCN51-P3 (ΔrsaC) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) strains were cultured in BHI (2h) and exposed to the ROS-generating reagent methyl viologen (MV) (Figure 3), which is commonly used to mimic the oxidative stress encountered in vivo. Growth (OD_600nm; Figure 3A) and cell viability (Figure 3B) were measured for 24 h. While the deletion of rsaC had no impact on bacterial growth, the overexpression of RsaC induced a significant growth defect and rendered cells highly sensitive to oxidative stress. Hence, in presence of high level of RsaC, bacteria hardly survived to the internal oxidative stress induced by MV.

Figure 3.

RsaC is involved in oxidative stress response via the repression of SOD activity. (A) Growth monitoring of WT/pCN51-P3 (○), ΔrsaC/pCN51-P3 (□) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) (Δ) upon internal superoxide stress (10 mM methyl viologen (MV)). MV was added after 2h of growth. (B) Viability assays of WT/pCN51-P3 (○), ΔrsaC/pCN51-P3 (□) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) (Δ) grown in BHI and challenged with 10 mM MV at early-exponential phase (2 h). Samples were taken thereafter, and viability was determined by performing viable counts on GP plates containing 5 μg/ml erythromycin. All data shown represent mean ± SD of three independent experiments. (C) Monitoring of SOD activity in WT/pCN51-P3 (WT), ΔrsaC/pCN51-P3 (ΔrsaC), ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆mut (rsaC_mut+) after 8h of growth (in presence or absence of MV). SOD activity was determined by negative staining on native polyacrylamide gel (NBT/riboflavin method). Results are representative of two independent experiments. (D) Measurement of reactive oxygen species (ROS) accumulation using fluorescence assays (after 6 h of growth in TSB) in WT/pCN51-P3 (black), ΔrsaC/pCN51-P3 (white) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) (gray). PBS buffer (light grey) was used as a control. Data represent mean ± SD of three independent experiments. Statistical analysis with ANOVA, ** P<0.005. Strains and plasmid constructs are given in Supplementary Table S1.

To understand the relationship between oxidative stress resistance and the cellular level of RsaC sRNA, we performed SOD activity assays in presence and absence of MV in the same strains (Figure 3C). As it was previously published (38), three bands corresponding to SodM homodimers (top), SodA/SodM heterodimers (middle) and SodA homodimers (bottom) were detected. In tested conditions, the first band was hardly visible in WT background suggesting that SodM homodimer contribution is marginal compared to SodA homodimer. While no clear difference was observed between WT and ΔrsaC mutant, SodA activity was nearly abolished in presence of overexpressed RsaC while SodM activity was enhanced (Figure 3C). Interestingly, the RsaC-dependent regulation of SOD activity was hindered in a strain overexpressing RsaC₁₁₁₆mut, which carries the deletion of sodA mRNA pairing site (from nucleotides +1000 to +1006; Figure 2B). This experiment showed that the repression of SodA synthesis is dependent on RsaC:sodA mRNA interaction in vivo.

As SOD activity assays are semi-quantitative, we quantified internal ROS level using fluorescence assays. We noticed the presence of higher ROS levels in WT strain than in ΔrsaC mutant (Figure 3D) suggesting that SOD activity is enhanced in ΔrsaC mutant. In presence of overexpressed full-length RsaC (Figure 3D) or its 3′ domain (Supplementary Figure S2), the accumulation of ROS was even higher than in WT strain. These results demonstrated that ROS level was correlated to enhance RsaC synthesis.

Because oxidative stress is a defence mechanism used by host cells to kill pathogens, we tested the ability to persist in macrophages (Supplementary Figure S8A) or to resist to opsonophagocytosis mediated by neutrophils (Supplementary Figure S8B). We observed that ΔrsaC mutant strain tends to be more persistent in macrophages and more resistant to opsonophagocytosis than WT strain. Conversely, the rsaC overexpressing strain was less persistent in macrophages and less resistant to opsonophagocytosis. Even if these observations are indirect, they are consistent with a higher SodA activity in absence or limited concentration of RsaC.

RsaC is released from the 3′UTR of mntABC operon

Our data suggest that RsaC is a noxious sRNA within cells, as high RsaC level renders them sensitive to oxidative stress. Hence, RsaC must be only produced under specific conditions and its endogenous expression should be firmly controlled. rsaC gene is located downstream the mntABC operon coding for the major manganese ABC transporter (Figure 4A). MntA is the ATP-binding protein, MntB the permease, and MntC the metal binding protein. The synteny is conserved in S. aureus and in closely related S. argenteus genomes (Supplementary Table S7) and only 113 nt separate mntC and rsaC genes, hinting that RsaC could be part of mntABC operon. By performing RT-PCR with primers in both RsaC and mntC (Figure 4A, B), we found that RsaC is co-transcribed with mntABC operon. We then checked if RsaC is issued from the maturation of the mntABC-RsaC precursor or from an internal promoter. The 5′-phosphate-dependent exonuclease Terminator fully digested RsaC as the large processed rRNAs (Figure 4C). Because this enzyme degrades only processed transcripts, we concluded that the 5′ end of RsaC is monophosphorylated. Previous works revealed that RsaC might be a target of the double-stranded endoribonuclease III (RNase III) (39). Northern blot analysis on RsaC performed in WT and Δrnc (RNase III mutant) strains (Figure 4D, right panel) revealed that in absence of RNase III, RsaC signal drastically drops while a far larger band appears (>3 kb). Using a probe complementary to mntA (Figure 4D, left panel), we confirmed that this very large RNA corresponds to the unprocessed mntABC-RsaC transcript.

Figure 4.

rsaC is part of the polycistronic operon mntABC and is released after RNase III cleavage. (A) Genomic context of mntABC operon. Oligonucleotides used for RT-PCR are indicated with black arrows. (B) RT-PCR using oligonucleotides in both mntC and rsaC genes and total RNA extracted after 4 or 6 h of growth in BHI medium. Same experiment was performed without RT enzyme (–) as a control. (C) Determination of the 5′ end status of RsaC sRNA. 10 μg of total RNA extracted after 6 h of growth in BHI medium were treated with the Terminator™ 5′-Phosphate-Dependent Exonuclease. 16S and 23 rRNAs were used as a positive control and 5S rRNA as a negative control. (D) Northern blot analysis of mntABC-RsaC transcript in wild type strain (WT, HG001) and the isogenic RNase III mutant (Δrnc) strain. Total RNA was extracted after 6 h of growth in BHI at 37°C. We used a probe targeting mntA (left panel) or RsaC (right panel). 5S rRNA was used as loading control. Data are representative of three independent experiments.

Because transcription of mntABC is repressed by the cognate repressor MntR in presence of Mn (40), we verified that RsaC level varies in function of Mn bioavailability (Figure 5A). Tested strains were grown in BHI-chelex medium supplemented or not with 25 μM MnCl₂ for 6 h at 37°C. In absence of Mn, we observed a strong increase of endogenous RsaC level, which is not visible in presence of Mn in WT strain. Only a weak signal was detected after overexposition of the autoradiography (Supplementary Figure S9A). Conversely, when BHI medium was supplemented with 25 or 100 μM MnCl₂ (Supplementary Figure S9B), RsaC synthesis was drastically reduced.

Figure 5.

RsaC is involved in Mn homeostasis. (A) Northern blot analysis of RsaC level in WT/pCN51-P3 (WT), ΔrsaC/pCN51-P3 (ΔrsaC) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) strains. Total RNA was extracted after 6 h of growth in BHI-chelex supplemented or not with 25 μM MnCl₂ at 37°C. 5S rRNA was used as loading control. When RsaC is highly produced, a shorter form is observed. Data are representative of three independent experiments. Differential proteomic analysis of SodA (B) and SodM (C) in presence or absence of 25 μM MnCl₂ (BHI-chelex). Data represent spectral count mean ± SD of three independent experiments. Statistical analysis is based on a negative-binomial test using an edgeR GLM regression through R, *P< 0.05; **P< 0.005; ***P< 0.0005; ns: not significant. (D) Monitoring of SOD activity in WT/pCN51-P3 (WT), ΔrsaC/pCN51-P3 (ΔrsaC) and ΔrsaC/pCN51-P3-RsaC₁₁₁₆ (rsaC+) after 6h of growth (in BHI-chelex ± 25 μM MnCl₂). SOD activity was determined by negative staining on native polyacrylamide gel (NBT/riboflavin method). Results are representative of three independent experiments. Strains and plasmid constructs are given in Supplementary Table S1.

RsaC modulates oxidative stress response upon Mn starvation

SodA is the major SOD enzyme in rich medium but is supplanted by SodM notably when cells are Mn-starved (10). In Figure 3C, we noticed that the overexpression of RsaC completely abolished SodA activity, but at the opposite increased SodM activity. While sodA mRNA was validated as a main target of RsaC, sodM mRNA was not enriched with RsaC in MAPS (Table 1). Moreover, RsaC interacts with sodM mRNA albeit with a very poor affinity in vitro (>500 nM) compared to sodA mRNA (≈24 nM) due to a partial conservation of the pairing region (Supplementary Figure S10). Hence, RsaC-dependent regulation of SodM activity must be indirect.

Beside Northern blot analysis performed in BHI-chelex ± 25 μM MnCl₂ (Figure 5A), the cytoplasmic proteins content was monitored using mass spectrometry (Supplementary Table S5). We observed that MntC is highly produced in absence of Mn (Supplementary Figure S11), confirming that cells are Mn starved. Focusing on SOD enzymes, no variation in SodA level was found in WT and ΔrsaC background in presence of 25 μM MnCl₂ when RsaC is barely detectable (Figure 5A, B). Only the overexpression of RsaC from a Mn-independent promoter had a negative impact on SodA synthesis (Figure 5B). In absence of Mn, SodA synthesis was reduced (∼3-fold) in strains where RsaC was either endogenously produced or overproduced (Figure 5B). RsaC had no significant effect on SodM synthesis although a tendency to increased SodM yield is observed when RsaC is overexpressed (Figure 5C). Remarkably, a clear increase of SodM level was noticed in absence of Mn (∼2.5-fold) (Figure 5C).

We then verified if SOD activities were correlated to proteins level (Figure 5D). Proteins were extracted in the same conditions as above. As expected, SodA activity was maximal when Mn was present, but was strongly impaired by high levels of RsaC. Activities of SodA homodimers and SodA/SodM heterodimers remained very low when RsaC was produced from a plasmid or from the chromosome in response to Mn starvation (Figure 5D). Interestingly, SodM homodimer activity strongly increased in absence of Mn and was significantly enhanced in WT and complemented (rsaC+) strains.

Our data showed that Mn starvation strongly induces RsaC expression, leading to a specific repression of SodA synthesis and to an enhanced activity of SodM, which is most likely occurring in an indirect manner.

RsaC has a broader role in oxidative stress response and metal homeostasis

In the differential proteomic analysis (Supplementary Table S5), we identified additional proteins, such as SufC, SufD and Ldh1, whose synthesis follows a regulatory pattern similar to SodA protein (Supplementary Figure S12). Indeed, their synthesis was significantly repressed in conditions where RsaC level is high (i.e. WT/ΔrsaC (-Mn), rsaC+/ΔrsaC (–Mn), rsaC+/ΔrsaC (+Mn) and WT (–Mn)/WT (+Mn)) and remained equal when RsaC level is low (WT/ΔrsaC (+Mn)). These proteins code for components of the Suf iron-sulfur cluster biosynthetic system and the lactate dehydrogenase 1, respectively (Supplementary Figure S12). None of them were co-purified with MS2-RsaC₅₄₄ (Table 1). However, MAPS experiments were performed after 6h of growth in BHI, in which Mn concentration was not controlled, hinting that some targets could have been missed. Therefore, we performed gel retardation assays to analyze the binding of RsaC to several other mRNAs, an indication that RsaC might regulate the synthesis of corresponding proteins (Figure 6). Our data show that RsaC efficiently binds to sufC transcript (Figure 6A) and ldh1 mRNA (Figure 6B). We suggest that RsaC pairing with sufC might be sufficient to regulate the whole sufCDSUB operon.

Figure 6.

RsaC potentially regulates additional mRNA targets involved in oxidative stress response and metal homeostasis. Gel retardation assays using RsaC₅₄₄ sRNA and several mRNA targets identified by MAPS and/or proteomic analysis. 5′ end-radiolabeled RsaC₅₄₄ (*) was incubated with increasing concentrations of (A) sufC and sufD, (B) ldh1 and rex, (C) znuC, znuB and zur, and (D) sarA mRNA (from 0 to 500 nM). Data are representative of two independent experiments. All transcripts have been described in Material and Methods.

Most of the targets enriched by MAPS were not or only poorly detected in our proteomic analysis (Supplementary Table S5). Nevertheless, we also tested the interaction of RsaC with several enriched mRNA candidates identified by MAPS that are involved in related functions. These mRNAs encode the zinc uptake system and the transcriptional regulators SarA, Zur and Rex, all related to metal homeostasis and oxidative stress (Table 1). Using gel retardation assays, we observed that RsaC binds to every member of znuB-znuC-zur operon and sarA mRNAs, but not with rex mRNA (Figure 6).

Using IntaRNA software (41), pairing sites were predicted between RsaC and each targeted mRNAs validated by gel retardation assays (Supplementary Figure S13). RsaC potentially binds to the ribosome binding site of sufC, ldh1, znuC, znuB, zur and sarA mRNAs, using several exposed C/U-rich sequences located in its 3′ part. Consequently, RsaC could repress the translation of additional targets, mainly involved in oxidative stress response and metal homeostasis.

DISCUSSION

RsaC, a connection node between oxidative stress and manganese bioavailability

In this study, we have shown that RsaC is a long atypical sRNA conserved in all S. aureus strains (Supplementary Table S7), characterized by two structured and independent domains (Figure 1). We showed that its 3′ domain was sufficient to reduce the translation of its major target SodA, identified by a combination of approaches including MAPS. Found in all staphylococci, SodA is the major SOD enzyme which is crucial to protect cells against damage caused by ROS produced upon reduction of O₂ or by immune cells (oxidative burst). We demonstrated that RsaC specifically pairs to the ribosome binding site of sodA mRNA to block its translation (Figure 2). RsaC does not use a classical C-rich sequence to base-pair with sodA mRNA as it was shown for many S. aureus sRNAs (42). Indeed, RsaC binding site overlaps the AUG start codon of sodA mRNA, not its Shine-Dalgarno sequence. This RsaC-dependent repression of sodA might explain why cells expressing a high level of RsaC are very sensitive to internally induced ROS production (Figure 3) and more sensitive to host immune attacks (Supplementary Figure S8). However, these harmful effects were mostly due to artificial overexpression of RsaC. Indeed, we hardly noticed an effect of rsaC mutation on S. aureus oxidative stress resistance when compared to WT. Only opsonophagocytosis assays (Supplementary Figure S8B) showed that endogenously synthesized RsaC reduced cells survival but this cannot be confidently attributed to a greater sensitivity to oxidative stress generated by immune cells.

We showed that RsaC level increased considerably in manganese-starved cells along with a drastic reduction of SodA activity indicating that RsaC exerts its regulatory activities under these specific physiological conditions (Figure 5). In S. aureus, SodM is an alternative and cambialistic SOD enzyme, which can utilize either Mn or Fe as cofactor (10). SodM compensates for the lack of functional SodA in manganese-starved cells and promotes resistance to oxidative stress (10,13). We confirmed that SodM becomes the major SOD enzyme in response to Mn limitation but also to RsaC level (Figures 3C and 5D). Unlike SodA, SodM is restricted to S. aureus and S. argenteus strains and potentially confers them a selective advantage over other staphylococci in human tissue environment. Remarkably, rsaC gene was also acquired exclusively by S. aureus and S. argenteus species. The proposed model is that, upon Mn starvation, RsaC negatively regulates the Mn-containing SodA enzyme to spare Mn and avoid translation of a useless enzyme. The absence of SodA also indirectly favors the formation of SodM homodimers, key players of the alternative ROS detoxification pathway, which is activated by both Mn starvation and RsaC (Figure 7). This situation is reminiscent of iron homeostasis in E. coli (43,44) since RyhB sRNA re-establishes iron homeostasis when iron is limiting. In E. coli, SodA and SodB synthesis are utterly opposed as RyhB negatively regulates the non-essential iron-containing SodB but not the Mn-dependent SodA enzyme (45,46). Consequently, SodA replaces SodB in iron-starved cells. However, the role of RyhB in oxidative stress response is still poorly explored.

Figure 7.

RsaC sRNA modulates oxidative stress response during Mn starvation. SodA is a Mn-dependent superoxide dismutase which is crucial to response to oxidative stress generated by aerobic respiration or by host defence cells (e.g. macrophages, neutrophils). In absence of Mn, RsaC is highly produced and negatively regulates sodA mRNA translation. It enables to spare Mn for essential Mn-containing proteins and avoids the synthesis of a non-functional enzyme. SodM, an alternative SOD enzyme using Fe as co-factor, replaces SodA to re-establish the ROS detoxification pathway. RsaC may indirectly activate SodM synthesis via the repression of the transcriptional regulator SarA. In addition, RsaC could have a broader role in oxidative stress response, notably through the regulation of Fe and Zn homeostasis (transport and storage) and NO^. resistance.

We propose that RsaC directly links oxidative stress response and nutritional immunity, both encounter at infection sites. Nutritional immunity refers to the sequestration of metals (i.e iron, manganese or zinc) by host organism to limit the growth of pathogens, which acquire these metals directly from host organism during infection (47–50). These metals are essential for proteins folding and for numerous enzymatic functions as cofactors (e.g. DNA synthesis, respiration or response to reactive oxygen and nitrogen species). The major innate immune effector involved in Mn depletion is the calprotectin (CP) protein, which sequesters Mn²⁺ with a high affinity (K_d < 1.3 nM) (5). Neutrophils release a large amount of CP (50 μM) to completely restrict Mn bioavailability at infection sites. For example, staphylococcal abscesses are completely Mn-depleted (51). To counteract Mn limitation, S. aureus has two Mn uptake systems, the ABC transporter MntABC and the proton-dependent Nramp homolog MntH (25,52). Both are under the control of the Mn²⁺-sensing transcriptional regulator MntR (52). The MntR-mediated repression of mntABC is released in Mn-starved cells while mntH expression seems to be constitutive and activated by MntR-Mn²⁺. The major manganese transporter MntABC not only promotes resistance to Mn starvation but also contributes to oxidative stress response and virulence (25,53). Indeed, Mn depletion by CP severely impacts SOD activity (5), lowering the resistance of S. aureus to oxidative burst generated by immune cells.

Remarkably, RsaC is co-transcribed with MntABC and, consequently, is repressed by MntR in presence of Mn. We showed that RsaC is released and stabilized from mntABC-RsaC operon after cleavage by RNase III. RNase III is involved in numerous functions including 5′UTR overlapping mRNAs processing, mRNA processing and turnover, and sRNA-mediated regulation (54). Here, we identified a new function of RNase III, which is responsible for sRNA release from its large precursor. The precise mechanism requires further analysis but the large 5′ hairpin structure encompassing the +1 of RsaC might be part of the binding site of RNase III (Figure 1B). Alternatively, two antisense RNAs have been previously identified (39,55) in this region, which can serve as guides for the RNase III-dependent cleavage. Indeed, we previously co-purified a fragment of 246 nt encompassing the +1 of RsaC (from +176 to –71 nt) and an antisense RNA (as_RsaC) using RNase III as bait in S. aureus RN6390 (39). Furthermore, another antisense RNA has been identified in this region in S. aureus SH1000, SbrC (55). This σ^B-dependent ∼310 nt-long RNA also overlaps the +1 of RsaC but predicted positions differ (from +12 to –297 nt). Attempts to validate its extremities were unsuccessful, presumably due to the too low level of SbrC. RsaC is the first example of 3′’UTR-derived sRNA in S. aureus, but in other bacteria several sRNAs were found released from the 3′UTR of precursor mRNAs by the single-strand specific endoribonuclease RNase E (56–58). Notably, SorX sRNA is co-transcribed with a gene encoding an OmpR-like transcriptional regulator in the α-Proteobacteria Rhodobacter sphaeroides (58). SorX counteracts oxidative stress (singlet oxygen and organic hydroperoxides) presumably by regulating a polyamine transporter.

Other physiological consequences of RsaC-dependent regulation

Using MAPS technology, 31 mRNAs were significantly enriched with RsaC (Table 1), suggesting that they could interact with RsaC. Here, not only SodA is involved in oxidative stress response, but also the staphylococcal accessory regulator SarA (59,60) and the redox-sensing transcriptional repressor Rex (61). SarA represses trxB, coding for the thioredoxin reductase, which is involved in ROS scavenging (60). SarA is also a transcriptional repressor of sodM (59). It is tempting to speculate that via sarA translational repression, RsaC could increase SodM level in response to Mn starvation, as suggested by the activation of SodM activity in presence of RsaC (Figures 3C and 5D). Remarkably, we demonstrated that RsaC binds to sarA mRNA in vitro and, according to the in silico prediction, could negatively regulate its translation by base pairing with its SD sequence (Figure 6D and Supplementary Figure S13). Therefore, via the repression of SarA, RsaC might indirectly activate SodM synthesis and, by extent, its cellular activity.

In addition to ROS, phagocytes produce nitrogen-based radicals such as nitric oxide (NO^.) (1). Both SodA and MntABC somehow contribute to NO^. resistance so that S. aureus can adapt its metabolism to limit deleterious effects (62). At high level, NO^. disrupts respiration and, consequently, perturbs the metabolic state, notably NAD+/NADH homeostasis. In response, Rex is inactivated, and cells switch to several metabolic pathways allowing NAD+ regeneration and ATP synthesis (61,63). For example, the lactate dehydrogenase Ldh1, which belongs to Rex regulon, enables the shift to heterolactic fermentation (61). The synthesis of Ldh1 protein appears to be repressed by RsaC (Supplementary Figure S12C). Moreover, we demonstrated that RsaC binds to ldh1 mRNA and potentially block its translation (Figure 6B and Supplementary Figure S13). The level of other proteins involved in NO^. stress resistance and belonging to Rex regulon (Supplementary Figure S14) is slightly decreased in response to Mn depletion in WT strain but does not respond to RsaC (Supplementary Table S5). Although, rex mRNA was copurified with MS2-RsaC₅₄₄, RsaC does not bind to rex mRNA in vitro (Figure 6B). Further studies will be required to better address the involvement of RsaC in redox homeostasis.

MAPS also revealed putative interactions with mRNAs encoding Fe-related proteins (Table 1). Metal ions are essential for all forms of life, but their over-accumulation is toxic to cells. For example, excess of Fe leads to ROS production via Fenton reaction (64). Hence, it is not surprising that the Suf-dependent Fe-S cluster synthesis is negatively regulated by RsaC (SufC and SufD; Supplementary Figure S12). We also validated a pairing of RsaC with sufCDSUB operon (Figure 6A), where RsaC seems to interact with the SD sequence of sufC mRNA (Supplementary Figure S13). Moreover, RsaC potentially controlled a Fe ABC transporter (fecD_1; Table 1). Although we cannot distinguish between positively and negatively regulated targets using MAPS, we can assume that RsaC might suspend iron import and promote iron storage in order to reduce oxidative stress generated by free intracellular iron. Upon manganese starvation, the iron-binding ferritin-like Dps is actively synthetized and YfmC_2, which is part of a Fe³⁺-citrate ABC transporter, is negatively regulated (Supplementary Table S5), supporting our assumption. Here, S. aureus likely uses both RsaC-dependent and -independent regulatory mechanisms to regulate Fe homeostasis. Putative interactions revealed by MAPS also highlighted a potential role of RsaC in the regulation of zinc uptake, i.e. znuC-znuB-zur operon. According to the Irving-Williams series (65), the stability of complexes formed by Mn²⁺ is very poor compared to Zn²⁺. Therefore, an overabundance of Zn is toxic due to competition for binding to Mn-dependent enzymes. Both MntC and MntR are particularly vulnerable to an irreversible Zn²⁺ binding, affecting Mn homeostasis (66,67). Therefore, S. aureus must limit the presence of intracellular Zn²⁺. Although we could not detect the protein involved in Zn uptake (Supplementary Table S5), certainly due to the presence of 25 μM ZnCl₂ in the medium, interactions with znuC, znuB and zur were validated in vitro (Figure 6C). Therefore, RsaC could hinder zinc import by blocking the translation of Znu transporter (Supplementary Figure S13). Besides, the RsaC-dependent control of Zur cellular level could lead to the derepression of its whole regulon. Further investigations will be necessary to understand the role of RsaC in metal homeostasis.

Links between oxidative stress and metal ion homeostasis are tightly regulated by metal-dependent transcriptional regulators such as PerR (68), Fur (69), Zur (70) and MntR (52). Notably, PerR controls oxidative stress resistance and iron storage in response to Mn²⁺ and Fe²⁺ (68,71). Again, this demonstrates the interdependency between oxidative stress response and metal homeostasis, but also the need to balance relative intracellular concentrations of Mn, Fe and Zn.

In this manuscript, we characterized a singular sRNA as a connection node between Mn bioavailability and ROS detoxification. Nevertheless, we expect that RsaC could have a wider role in oxidative stress response, via the regulation of metals homeostasis and NO^. resistance. It also remains to be addressed whether the unusual repeat sequences in the 5′ domain of RsaC might play additional functional roles.

DATA AVAILABILITY

MAPS data have been deposited to GEO under accession GSE128741.

The mass spectrometric data were deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository (72) with the dataset identifier PXD013225.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

labEx NetRNA [ANR-10-LABX-0036 to P.R., ANR-17-EURE-0023 to P.R.] as part of the investments for the future program from the state managed by the French National Research Agency (ANR); ANR [ANR-16-CE11-0007-01 to P.R. and F.V.]; European Union's Horizon 2020 research and innovation programme [753137 to D.L.]; Mass spectrometry instrumentation was funded by University of Strasbourg, IdEx Equipement mi-lourd 2015 and labEx NetRNA. Funding for open access charge: RIBOSTAPH [ANR-16-CE11-0007-01].

Conflict of interest statement. None declared.