Introduction

Staphylococcus aureus is an opportunistic human pathogen that can cause a variety of infections [1]. The expression of virulence factors is to a large extent controlled by the agr quorum sensing (QS) system composed of the response regulator, AgrA and the sensor histidine kinase, AgrC that in response to auto-inducing peptides expresses a regulatory RNA, RNAIII. At high cell density RNAIII mediates the transition from production of host matrix binding and immune evasion proteins to expression of a large number of extracellular toxins including the α-hemolysin encoded by hla [2]. Within RNAIII itself a toxin, the δ toxin, is also encoded [3]. S. aureus QS has been demonstrated to be important for virulence in several animal models of acute infection, including infective endocarditis, skin and soft tissue infections and septic arthritis [4,5,6]. Yet S. aureus QS defective mutants are commonly found in clinical isolates and they are associated with a wide range of infections such as persistent bacteremia; infections of the lungs of cystic fibrosis patients and with higher mortality in general [7,8,9,10,11]. Also, in the laboratory they arise spontaneously at high frequencies [12,13,14]. Recently we showed that exposure to sub-lethal antibiotic concentrations increases the fitness cost of the agr system by inducing RNAIII expression levels [15]. The fitness advantage of the agr mutant over the wild type (WT) strain could not be related to any differences in exponential growth rate and was only detected in competition between the two strains [15]. These observations prompted us to examine the hypothesis that the apparent fitness advantage of the agr mutant cells compared to the WT can be explained by differences in viability.

Main text

Methods

Bacterial strains and growth conditions

Staphylococcus aureus strains (Additional file 1: Table S1) were grown in Tryptic Soy Broth (TSB) containing 2.5 g/l glucose (Sigma-Aldrich) or in Bacto™ Tryptic Soy Broth without glucose, Benton Dickson (BD286220). Blood agar plates contained 1.5% Agar (Difco) and 5% calf’s blood. Antibiotics were added in the following concentrations: tetracycline 2 µg/ml; chloramphenicol 10 µg/ml; erythromycin 10 µg/ml (Sigma-Aldrich). Mutations and plasmids were transferred by transduction using phage φ11 [16, 17]. Transposon mutant clones were obtained from the Network on Antimicrobial Resistance in S. aureus (NARSA). agr activity was assayed on blood agar plates and hemolysis was scored for approximately 100 colonies.

For hld overexpression, pTXΔ-RNAIII, was equipped with an E. coli replication origin and ampicillin selection marker from pUC19 using primers pUC-1 and pUC-2 (Additional file 1: Table S2) through SmaI/SacI restriction sites yielding plasmid pTXΔ-RNAIII-pucori. The hld coding sequence was amplified using primers Hld1 and Hld2 (Additional file 1: Table S2) followed by the substitution of RNAIII for hld using restriction sites BamHI/MluI yielding plasmid pTXΔ-Hld-pucori.

Competition experiment

Five independent cultures of ΔagrA plus WT cells were grown in TSB with or without glucose for 100 generations (10 passages) with 1 × 106 bacteria transferred daily into fresh growth medium. The ratio of the tetracycline-resistant ΔagrA mutant cells to WT cells was determined on TSB agar plates with and without tetracycline after 30, 50 and after 100 generations of growth.

Live/dead staining

Staining was performed with thiazol orange (TO) (Sigma-Aldrich) at 0.168 μM and propidium iodide (PI) (Sigma-Aldrich) at 8 μM, staining 106 cells/ml for 15 min under dark conditions at room temperature. The flow cytometry data was recorded with a BD Biosciences Accuri C6 flow cytometer counting 50,000 cells at a flow rate of 35 μl/min and with a core size of 16 μm. Stained cells were excited with a 488 nm argon laser and emission was detected with the FL1 emission filter at 533/30 nm using FL1 photomultiplier tub and in FL-3 emission filter at 670 nm using FL3 photomultiplier tub.

Quantitation of RNAIII expression and eDNA levels by real-time qPCR

The SV RNeasy Mini Kit (Qiagen) was used for isolation of RNA; cDNA RT kit (Applied Biosystems) for cDNA synthesis (using an RNase inhibitor, Applied Biosystems) and the FastStart Essential DNA Green Master (Roche) for qPCR in a Lightcycler 96 (Roche) with the primers listed in Additional file 1: Table S2. Data analysis was performed in the LightCycler Application Software, version 1.1 (Roche). Extracellular, chromosomal DNA (eDNA) was quantified by qPCR amplifying the ileS sequence directly on 100-fold diluted and heat-treated culture supernatants. The eDNA concentration in the supernatants were subsequently calculated from a standard curve of purified genomic DNA and normalized using an exogenous DNA spike which was added to all samples. The DNA spike was PCR product of the GFP gene and 10,000 copies were added to each qPCR reaction (primers used are listed in Additional file 1: Table S2.

Growth potential of lysed cells

Staphylococcus aureus cells in TSB were lysed by applying bead-beating (FastPrep®-24, MP Biomedicals). Increasing concentrations of lysate were added to TSB diluted 1:10 with water, inoculated with S. aureus Newman overnight cultures (diluted 1:100) and incubated in a Bioscreen C reader (Thermo Labsystems) at 37 °C for 24 h. Technical quadruplicates and biological triplicates were included for each condition.

Results

Cell death is reduced in agrA and RNAIII mutant cells

Initially we assessed the frequency with which agr negative cells arise in strain Newman by passaging five individual WT colonies in serial batch cultures and determining the frequency of hemolysin negative mutants on blood agar plates. Although only a fraction of agr mutations will eliminate hemolysis it has previously been used as an indicator of agr activity [14]. Non-hemolytic colonies appeared on day 7 and by day 19 haemolytic colonies could only be detected in one lineage (Additional file 1: Figure S1). Thus, in line with findings for other S. aureus strains [14], agr mutants readily arise in cultures of stain Newman during serial passage.

Since the selection for agr negative cells during the serial passage cannot readily be explained by differences in growth rate and is primarily detected in competition with WT cells [14, 15] we examined the viability of agr positive and negative cells. Stationary phase cultures of Newman WT, ΔagrA and ΔRNAIII mutant derivatives were live/dead stained with propidium iodine (PI) and thiazole orange (TO) and analysed by flow cytometry. Interestingly, we observed a significantly higher fraction (p < 0.05) of dead WT cells compared to ΔagrA or ΔRNAIII mutant cells (Fig. 1a) indicating that a fraction of agr positive cells loose viability by a process not taking place in ΔagrA or ΔRNAIII mutant cells.

Fig. 1
figure 1

agr and RNAIII influence cell viability. a Cultures of WT, ΔagrA and ΔRNAIII mutant were stained with PI and TO and the percentage of dead cells determined by flowcytometry (An asterix indicates a p-value of < 0.05 when compared to WT by Student’s T-test). b The percentage of dead cells was determined by staining and flowcytometry at various time points after inoculation of WT cells carrying either vector (pTXΔ) or the RNA overproducing plasmid (pTXΔRNAIII). Bars represent the mean and the standard deviation from biological triplicates

Full size image

RNAIII overexpression induces lysis

To address if RNAIII may be the factor reducing the viability in agr positive cells, we examined populations of WT cells containing an empty vector, pTXΔ, or a plasmid constitutively expressing RNAIII, pTXΔRNAIII. Cultures of both strains were inoculated to a cell density of 5 × 106 colony forming units per ml, growth as well as the live/dead ratio was continuously monitored with flow cytometry. We observed that with the pTXΔRNAIII construct, substantial cell death occurred after 6 h of growth with more than 60% of the population stained as dead cells (Fig. 1b) whereas few dead cells were observed in cells carrying the empty vector, pTXΔ. At this timepoint the cells carrying pTXΔRNAIII overproduced RNAIII tenfold above the level in cells carrying the vector (Additional file 1: Figure S2). Upon progression into stationary growth phase, the number of cells stained as dead decreased with RNAIII overproduction (Fig. 1b) indicating that dead cells lyse and consequently are not detected in the flow cytometer. This observation was confirmed by monitoring the release of chromosomal DNA by qPCR. After 7 h of growth, extracellular DNA (eDNA) could only be observed in the supernatant of cells with RNAIII overproduction and it appeared at a concentration of 13.9 µg/ml (± 4.0) eDNA detected, corresponding to the DNA content of 5 × 109 S. aureus cells/ml. To assess whether the pronounced cell death observed with RNAIII overexpression was due to overproduction of the δ-toxin encoded by hld within the RNAIII transcript, we overproduced δ-toxin from a construct not expressing RNAIII. To this end we removed the entire RNAIII-encoding region and inserted hld with the same Shine Dalgarno sequence as on the native transcript. With this plasmid cell death was reduced to only 10% of that seen with the RNAIII-overproducing plasmid showing that the effect is mediated via RNAIII and not the δ-toxin.

Bacterial lysis releases resources supporting growth

Since dead cells potentially are cannibalized we examined whether lysed S. aureus cells could support growth. For this purpose, we inoculated WT cells in dilute TSB broth (0.1 × TSB) supplemented with varying amounts of staphylococcal lysate obtained from mechanical disruption of S. aureus WT cells and observed that increasing amounts of lysate stimulates growth as observed by a higher final optical cell density (Additional file 1: Figure S3).

Staphylococcus aureus is known to encode a number of autolysins and we speculated that one of these might be responsible for the cell death. However, upon transduction of mutations in lrgB, cidA, lrgA, lytM, tagX, a lysM domain protein (NE1640) and an autolysin (NE1948) from the NARSA transposon library into Newman + pTXΔRNAIII none of the mutations altered the lysis phenotype elicited by RNAIII overproduction (data not shown) indicating that the examined gene products are not responsible for the RNAIII mediated cell death.

Modulation of RNAIII expression eliminates the competitive advantage of agr negative cells

To determine if RNAIII expression levels influences the competition between WT and agr mutant cells we competed ΔagrA with WT cells and observed that ΔagrA cells quickly outcompeted the WT in regular TSB medium (Fig. 2) while this was not the case in TSB lacking glucose where RNAIII expression has been demonstrated to be reduced [18] (Fig. 2). These data suggest that the competitive advantage of being agr negative is associated with less RNAIII expression and less lysis of cells.

Fig. 2
figure 2

Competition between WT and ΔagrA cells. WT and ΔagrA mutant cells were co-cultivated in ratios 1:1; 1:10; 1:100 or 1:1000 of WT to ΔagrA the presence (TSB+) or absence (TSB−) of glucose for either 30 (grey), 50 (dark grey) or 100 (light grey) generations and the ratio of ΔagrA to WT cells was determined based on agar plates with and without tetracycline compared to the inoculum (black). The bars represent the mean and standard deviation obtained from five independent co-cultures

Full size image

Discussion

Here we show that a small fraction of a WT S. aureus population undergoes cell death and that this does not take place in mutant cells lacking the agr QS system. Since both agr positive and negative cells multiply on lysed staphylococcal cells, agr negative cells have an advantage over WT cells. We propose that this phenomenon contributes to the frequent manifestation of agr mutant cells both in vivo and in vitro during serial passage [14]. Our results agree with previous findings that the apparent fitness advantage of agr negative cells is particularly evident in competition assays [15]. Currently we do not know the molecular details of the killing process nor the mechanism behind the stochasticity by which it occurs. However, it has been noted that agr mutant cells are less prone to Triton X-100 mediated lysis and are more resistant to lysis by Penicillin G compared to wild type cells [19] indicating that there is an overall basic physiological difference between agr positive and negative cells.

In P. aeruginosa QS has also been linked with decreased viability as a mutant lacking the las QS system resists autolysis at high cell densities resulting in about tenfold increase in lasR mutant-to-wild-type ratio in mixed cultures [20]. Interestingly QS negative cells of both S. aureus and P. aeruginosa appear under chronic infections experiencing prolonged antibiotic exposure [20]. In S. aureus, some antibiotics are known to increase expression of RNAIII [9, 21]. We speculate that this induction may lead to increased cell death in WT populations of cells and enhanced the appearance of agr mutant cells.

The biological impact of differential death of agr positive cells remain obscure. It may contribute to biofilm formation via the DNA released [22] but it may also serve the purpose of establishing mixed populations of agr positive and negative cells. As the fraction of agr negative cells increases, the induction of agr expression in WT cells will decrease and consequently also the RNAIII mediated lethality. In Salmonella enterica serovar Typhimurium it has been shown that expression of a phenotypically avirulent subpopulation promotes the evolutionary stability of virulence (35) and similar cooperation may take place in S. aureus.

Limitations

While the phenomenon reported in our study is observed for several strains of S. aureus we do not know the extent to which cell death in stationary phase occurs in clinical strains.