Toxin Mediates Sepsis Caused by Methicillin-Resistant Staphylococcus epidermidis

Bacterial sepsis is a major killer in hospitalized patients. Coagulase-negative staphylococci (CNS) with the leading species Staphylococcus epidermidis are the most frequent causes of nosocomial sepsis, with most infectious isolates being methicillin-resistant. However, which bacterial factors underlie the pathogenesis of CNS sepsis is unknown. While it has been commonly believed that invariant structures on the surface of CNS trigger sepsis by causing an over-reaction of the immune system, we show here that sepsis caused by methicillin-resistant S. epidermidis is to a large extent mediated by the methicillin resistance island-encoded peptide toxin, PSM-mec. PSM-mec contributed to bacterial survival in whole human blood and resistance to neutrophil-mediated killing, and caused significantly increased mortality and cytokine expression in a mouse sepsis model. Furthermore, we show that the PSM-mec peptide itself, rather than the regulatory RNA in which its gene is embedded, is responsible for the observed virulence phenotype. This finding is of particular importance given the contrasting roles of the psm-mec locus that have been reported in S. aureus strains, inasmuch as our findings suggest that the psm-mec locus may exert effects in the background of S. aureus strains that differ from its original role in the CNS environment due to originally “unintended” interferences. Notably, while toxins have never been clearly implied in CNS infections, our tissue culture and mouse infection model data indicate that an important type of infection caused by the predominant CNS species is mediated to a large extent by a toxin. These findings suggest that CNS infections may be amenable to virulence-targeted drug development approaches.


Introduction
Bacterial sepsis is a frequent cause of death in hospitalized patients. Coagulase-negative staphylococci (CNS) are the leading cause of nosocomial sepsis, especially in neonates [1][2][3]. CNS sepsis most often originates from the infection of indwelling medical devices, such as in catheter-related bloodstream infections (CRBSIs) or central line-associated blood stream infections (CLABSIs) [4]. Most prominent among CNS infections are those due to the skin commensal Staphylococcus epidermidis [5]. However, the bacterial factors contributing to the development of sepsis, in particular in CNS, are poorly understood.
Given that toxins have long been assumed to be widely absent from CNS [6], sepsis caused by S. epidermidis and other CNS, similar to other Gram-positive bacteria, has so far been believed to be due predominantly to an overwhelming immune reaction directed against invariable, pro-inflammatory cell surface molecules, such as teichoic acids and lipopeptides [7]. Recently, the notion that CNS do not commonly produce toxins had to be revised with the discovery of the pro-inflammatory and cytolytic phenol-soluble modulin (PSM) staphylococcal toxin family [8]. However, due to the difficulties associated with genetic manipulation of S. epidermidis and other CNS, the roles of PSMs in CNS infections, including most notably sepsis, have hitherto remained unexplored.
Most S. epidermidis blood infections are caused by methicillin-resistant strains (MRSE), with methicillin resistance rates even exceeding those found among S. aureus [9]. Methicillin resistance is encoded on so-called staphylococcal chromosome cassette (SCC) mec mobile genetic elements, which are believed to have originated from CNS, from where they were transferred to S. aureus [10]. While other PSMs are core-genome encoded [8], one PSM toxin, called PSM-mec, is encoded within SCCmec elements of subtypes II, III, and VIII [11,12]. The psm-mec gene is embedded in a short regulatory (sr) RNA, which in S. aureus has been reported to down-regulate the production of other PSMs and thereby decrease virulence [13,14]. While this effect has been claimed to generally explain lower virulence of hospital-associated as compared to community-associated MRSA strains [13], it is quite moderate and extensively strain-dependent [11,13]. Recently, the psm-mec locus has been introduced on a plasmid into some CNS that naturally lack psm-mec, and was reported to trigger gene regulatory changes [15]; but the roles that the PSM-mec peptide or the psm-mec srRNA naturally play in CNS including S. epidermidis are unknown.
Here we analyzed the role of the psm-mec locus in S. epidermidis sepsis by using tissue culture and animal infection models. Our findings show for the first time that a toxin can have a strong impact on CNS sepsis, setting the stage for anti-virulence strategies directed against this frequent and deadly infection.

Results and Discussion
To analyze the impact of the psm-mec locus on S. epidermidis sepsis, we produced isogenic psm-mec deletion mutants (Δpsm-mec) in two MRSE strains, a clinical isolate (SE620) and the genome-sequenced strain RP62A. PSM-mec production in these strains is representative of clinical PSM-mec-positive MRSE (S1 Fig), which we determined in a clinical S. epidermidis strain collection from Norway to occur in~2/3 (59/91) of the~50% (91/180) methicillin-resistant S. epidermidis. We also introduced a point mutation in the start codon of the psm-mec gene in the genome of strain SE620 to differentiate between effects mediated by the PSM-mec peptide versus those due to the psm-mec srRNA (psm-mec Ã ). Notably, the stability of the psm-mec RNA was not significantly altered by introduction of the 1-basepair start codon mutation (S2 Fig).
We first analyzed those mutants in a murine sepsis model. Mortality was significantly reduced in the Δpsm-mec mutants of both strains (Fig 1A and 1B). There was no significant difference between the Δpsm-mec mutant and the psm-mec Ã start codon mutant (Fig 1A). Furthermore, CFU in the blood and the kidneys were strongly reduced in the Δpsm-mec mutants of both strains and the psm-mec Ã start codon mutant (Fig 1C-1F). These results demonstrate a strong contribution of the PSM-mec toxin to bacteremia and mortality due to S. epidermidis sepsis, while the psm-mec srRNA did not show any impact.
We showed previously that synthetic PSM-mec peptide is strongly pro-inflammatory and has moderate to strong cytolytic capacity [12]. To analyze the contribution that the psm-mec locus has to pro-inflammatory and cytolytic capacity in the S. epidermidis background, we measured cytokine concentrations during experimental murine sepsis and determined cytolytic capacity of the bacterial strains toward human neutrophils in vitro. Cytokine concentrations during sepsis are the result of a systemic reaction due to several immune cell types, and are thus best determined in vivo, while cytolytic capacity can be most accurately measured in vitro.
The PSM-mec peptide, but not the psm-mec srRNA, had a strong and significant impact on the production of cytokines during murine sepsis (Fig 2). At 12 h after infection, the mouse IL-8 homologue CXCL1 was significantly reduced when mice were infected with the Δpsmmec or psm-mec Ã start codon mutant of strain SE620, to about half the concentration measured in mice infected with the wild-type strain (Fig 2A). Concentrations of IL-1β and TNF-α were even more strongly reduced to levels not significantly different from those measured in mock (PBS) infected animals (Fig 2B and 2C). In the RP62A background, the phenotypes were similar, with differences being more pronounced during earlier stages of the infection (measured at 2 versus 12 h) (Fig 3). These results showed that the cytokine storm that commonly accompanies bacterial sepsis is strongly dependent on the PSM-mec toxin in S. epidermidis.
In addition to being pro-inflammatory, the PSM-mec toxin has pronounced cytolytic capacity [12]. Cytolysis by PSMs is believed to be most important for infection when bacteria are engulfed in the phagosome of neutrophils and other phagocytes [16,17]. Survival of bacteria when incubated with human neutrophils and survival in whole human blood was significantly higher with the S. epidermidis wild-type strain than with Δpsm-mec or psm-mec Ã start codon mutants, as was killing of neutrophils when incubated with whole bacteria (Fig 4), emphasizing the role of the PSM-mec toxin in evasion of neutrophil killing and resistance to the strong bactericidal capacities of immune defense mechanisms in human blood. Together, these results indicate that the both the pro-inflammatory and cytolytic capacities of the PSMmec peptide contribute to the development of S. epidermidis sepsis.
In S. aureus, the psm-mec locus has also been implicated in biofilm-forming capacity, although effects were generally minor and highly strain-dependent [12]. Similar to S. aureus, biofilm formation in S. epidermidis was affected only slightly by the psm-mec locus, and as this was seen only in one strain, similarly strain-dependent (Fig 5). In that strain, SE620, the effect was due to the PSM-mec peptide, not the psm-mec srRNA. These findings indicate that during indwelling medical device-associated blood stream infections by S. epidermidis, the impact of PSM-mec generally is by contributing to the development of sepsis, as we have shown here, rather than by promoting biofilm formation on the device itself.
Our results showed that the psm-mec srRNA is not involved with sepsis or other relevant virulence phenotypes in S. epidermidis. As a previous study suggested that the psm-mec srRNA leads to gene regulatory changes in S. epidermidis [15], based on the introduction of a psm-mec expressing plasmid into S. epidermidis, we also directly investigated whether the psm-mec locus has a gene regulatory impact in S. epidermidis. The most important regulatory effect of the psm-mec locus in S. aureus, by which the sometimes negative impact of the psm-mec locus on virulence in S. aureus was explained, has been reported to consist in the alteration of the expression of other, core genome-encoded PSMs [18]. PSM expression was altered only to a very low extent in the psm-mec-negative as compared to the wild-type S. epidermidis strains, with changes only significant for some PSMs and never exceeding a factor of~1.5 (Fig 6). This demonstrates that there is only a very minor effect of the psm-mec srRNA on PSM expression when analyzed directly in the S. epidermidis background. Furthermore, we analyzed genome-wide gene expression in the psm-mec mutants of both strains by microarray analysis (Tables 1 and 2). For microarray analysis, strains were grown to the maximum of PSM-mec expression as determined by qRT-PCR (10 h) (Fig 7). While we observed gene regulatory changes that were due to the psm-mec srRNA, they mostly comprised metabolic (e.g., riboflavin and purin/pyrimidine synthesis) rather than virulence genes, and were inconsistent between the two strains. Notably, the results of the previously claimed impact of the psm-mec locus on virulence would be negative [13,15,18], contrasting the positive effect we observed in the mouse sepsis model. Such a gene regulatory mechanism can thus be ruled out as underlying psm-mec-mediated development of S. epidermidis sepsis.
Our results may explain the highly inconsistent phenotypes that have been attributed to psm-mec in S. aureus [11][12][13], inasmuch as the psm-mec locus may exert effects in the background of S. aureus strains that differ from its original role in the CNS environment. One such possibility that remains to be investigated is that the highly expressed psm-mec mRNA interferes with other DNA or RNA sequences in S. aureus. Furthermore, the psm-mec srRNA barely exceeds the limits of the psm-mec gene [14], which contrasts the only other case of an srRNA with an embedded peptide toxin in staphylococci, namely the well-described regulatory RNAIII of the staphylococcal accessory gene regulator (Agr) system. RNAIII significantly exceeds the boundaries of the embedded PSM peptide gene, hld [19]. Together, these observations suggest that the psm-mec srRNA does not serve a well-defined general purpose in virulence gene regulation. In conclusion, our study reveals that sepsis due to MRSE is mediated to a large extent by the PSM-mec peptide toxin, representing the first example of a toxin being made responsible for the development of CNS sepsis. Our study was largely based on the investigation of isogenic psm-mec mutants in clinical strains of S. epidermidis, using tissue culture and animal infection models. Future clinical work is needed to assess whether PSM-mec and/or other toxins contribute to sepsis in humans. Importantly, our results suggest that CNS sepsis may be amenable to virulence-targeted therapeutic approaches, such as those targeting the quorum-sensing system Agr [20], which strictly regulates PSM expression [21], or monoclonal antibody-based therapy directed against the toxin.

Bacterial strains and growth conditions
Strain RP62A is a genome-sequenced clinical MRSE isolate [22]. Strain SE620 is an MRSE clinical isolate from Norway [23]. Isogenic Δpsm-mec deletion mutants and the psm-mec Ã start codon mutant were produced with the constructs previously used for S. aureus [12,14], using Error bars show ±SEM. $, Residual PSM-mec production in the SE620 psm-mec start codon mutant is likely due to strong gene expression and usage of a non-canonical start codon, as we previously found also in S. aureus psm-mec start codon mutants [14]. a strategy with the allelic exchange vector pKOR1 [24]. The psm-mec locus and adjacent DNA do not differ between S. aureus and S. epidermidis [25]. For construction of the psm-mec Ã mutant, the start codon mutation was created by introducing a ClaI restriction site (introducing ATC instead of the ATG start codon) using primer PSMEClarev GAGGGTATGCATATC GATTTCACTGGTGTTATTACAAGC and primer PSMECladir (reverse complement of PSMEClarev). Two PCR fragments were amplified using those primers and primers psmEatt1 and psmEatt2, respectively [12], cut with ClaI, ligated, and cloned into pKOR1. The resulting plasmid was used for allelic replacement as described [24]. Growth patterns of the mutants were indistinguishable from those of the wild-type (S3 Fig). Strains were grown in tryptic soy broth (TSB), unless otherwise noted.

Mouse sepsis model
Female, 6-10 weeks old, C57BL/6NCRl (Charles River) mice were used. The mice were injected via the tail vein with 5 x 10 8 CFU in 100 μl phosphate-buffered saline (PBS) of the indicated bacterial strains grown to mid-exponential growth phase and monitored for disease development every 8 h for up to 120 h. This dosis was determined to be minimally necessary to achieve mortality and production of inflammatory cytokines (S4 Fig). Animals were

Interaction of bacteria with human neutrophils and bacterial survival in human blood
For survival in whole blood experiments, about 10 8 bacteria in 100 μl Dulbecco's PBS from mid-exponential growth phase were added to 500 μl heparinized human blood and mixtures were incubated for 6 h. Aliquots were taken at 2-h intervals, and CFU were determined by plating and incubating plates overnight at 37˚C. For neutrophil interaction experiments, neutrophils were isolated from the venous blood of human volunteers as described [26]. Bacteria from mid-exponential growth phase were mixed with neutrophils at an MOI (bacteria/neutrophils) of 10:1. Bacteria/neutrophil mixtures were incubated at 37˚C, 5% CO 2 , 90% humidity for 6 h. At 2-h intervals, 50 μl of Triton X-100 was added to the 200-μl bacteria/neutrophil suspensions, aliquots were plated, and plates incubated at 37˚C overnight for CFU counting. Alternatively, the rate of neutrophil lysis promoted by the bacteria was determined after 4-h incubation using a lactate dehydrogenase (LDH) assay at an MOI of 100:1.

Biofilm formation
Biofilm formation was assessed in a semi-quantitative 96-well microtiter plate assay as previously described [27], using TSB + 0.5% glucose.

Quantitative real-time (RT)-PCR and microarray analysis
Quantitative RT-PCR was performed as previously described [29] with the following oligonucleotides: psm-mecF, TGCATATGGATTTCACTGGTGTTA, psm-mecR, CGTTGAATATT TCCTCTGTTTTTTAGTTG, psm-mec probe, ATTTAATCAAGACTTGCATTCAG. Expression was measured relative to that of 16S RNA. Cultures were grown to the maximum of psmmec expression as determined by qRT-PCR (10 h). Total RNA and cDNA were prepared as described [30]. Biotinylated S. aureus cDNA was hybridized to custom Affymetrix GeneChips (RMLChip 3) with 100% coverage of chromosomal genes from strains S. epidermidis RP62A and scanned according to standard GeneChip protocols (Affymetrix). Each experiment was replicated 3 times. Affymetrix GeneChip Operating Software was used to perform the preliminary analysis of the custom GeneChips at the probe-set level. Subsequent data analysis was performed as described [30]. The complete set of microarray data was deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and is accessible through GEO Series accession number GSE85265.

Analysis of mRNA stability
To determine psm-mec mRNA stability in strains S. epidermidis SE620 and the psm-mec Ã start codon mutant, bacteria were cultured at 37˚C for 10 hours. At time t = 0 min, rifampicin (50 mg/ml stock in DMSO) was added to the cultures to a final concentration of 100 μg/ml. One-ml aliquots were taken and immediately centrifuged at 4˚C to pellet cells, which were then frozen at -70˚C. Remaining cultures were further incubated at 37˚C with shaking; one-ml aliquots were taken at the indicated times and RNA was subsequently isolated from all cell pellets as described [29]. Samples were analyzed by qRT-PCR using primers psm-mecR and psm-mecF with a SuperScript III Platinum SYBR Green One-Step qRT-PCR kit (Invitrogen) according to the manufacturer's instructions. Expression was measured relative to that of 16S RNA.

Statistics
Statistical analysis was performed using GraphPad Prism Version 6.0. Comparisons were by 1-way or 2-way ANOVA for comparisons of three and more, and by unpaired t-tests for comparisons of 2 groups. Error bars show ±SEM.

Ethics statement
The animal protocol (LB1E) was reviewed and approved by the Animal Care and Use Committee at the NIAID, NIH, according to the animal welfare act of the United States (7 U.S.C. 2131 et. seq.). All mouse experiments were performed at the animal care facility of the NIAID, Building 50, in accordance with approved guidelines. All animals were euthanized by CO 2 at the end of the studies. Human neutrophils were isolated from blood obtained under approved protocols at the NIH Blood Bank or with a protocol (633/2012BO2) approved by the Institutional Review Board for Human Subjects, NIAID, NIH. All subjects were adult and gave informed written consent.