Staphylococcus aureus Persisters Tolerant to Bactericidal Antibiotics

Bacterial persister cells were described by Joseph W. Bigger [1944] who studied the action of penicillin against staphylococci. Addition of the antibiotic to liquid cultures led to the sorting out of a small number of survivors he termed persisters. Upon reinoculation, these cells gave rise to a culture that again was susceptible to the antibiotic, forming a new subpopulation of persisters. Decades later, comparable observations were made [Gunnison et al., 1964; Mayhall et al., 1976; Newsom, 1970], confirming that a fraction of Staphylococcus aureuscells within a common culture survived antibiotic treatment although the original strains were susceptible to it. These cells were neither spheroplasts nor L-forms, they did not arise in response to the antibiotic and tolerance was not transferred to the progeny. Recently staphylococcal persisters were revisited, providing insights into the distribution of S. aureus and S. epidermidis persisters in biofilms and planktonically grown cells [Keren et al., 2004a; Möker et al., 2010; Shapiro et al., 2011]. Persisters have been identified in an increasing number of microorganisms, including Pseudomonas aeruginosa[Brooun et al., 2000; Harrison et al., 2005; Möker et al., 2010; Spoering and Lewis, 2001], Escherichia coli[Keren et al., 2004a, b; Shah et al., 2006], Mycobacterium tuberculosis[Keren et al., 2011] and the unicellular fungus Candida albicans[Lafleur et al., 2010; reviewed in Lewis, 2010a]. To identify genetic mechanisms underlying the persister state, Moyed and colleagues [Moyed and Bertrand, 1983; Moyed and Broderick, 1986; Scherrer and Moyed, 1988] carried out a selection for E. coli mutants with increased survival to ampicillin, which resulted in the identification of a mutant carrying the hipA7 allele producing elevated amounts of persisters. Persisters are presumably formed due to stochastic variation in the levels of persister gene expression irrespective of antibiotic action [Balaban et al., 2004]. At the same time, the probability of persister formation can be strongly increased by stress. It was reported that the majority of E. coli persisters under DNA-damaging conditions are produced by expression of the TisB toxin, which is under the control of the SOS response [Dörr et al., 2009]. By definition, persisters are not mutants, but are regarded as phenotypic variants that are isogenic to normally growing cells within a common culture [Lewis, 2007, 2010a, b; Wiuff et al., 2005]. Persister levels in most species examined increase sharply during late exponential growth, reaching ∼1% at the stationary phase [Keren et al., 2004a]. Unraveling the molecular mechanisms of persister formation in species other than E. coli has only recently begun, reflected by recent studies in P. aeruginosa or M. tuberculosis[Keren et al., 2011; Möker et al., 2010]. Regarding S. aureus, the observation that a stationary culture of strain ATCC 55585 was completely tolerant to the fluoroquinolone ciprofloxacin had suggested that stationary cells may be equivalent to persisters in this organism [Keren et al., 2004a]. We examined this possibility in the present study and found that stationary cultures of S. aureus, similarly to other species examined, represent a mixture of regular cells and persisters. We also considered persister formation by S. aureus small colony variant (SCV) strains that are intrinsically tolerant to selected antibiotics. SCVs are often isolated in hospitals and contribute to antibiotic recalcitrance of infections [von Eiff, 2008]. SCVs from both clinical isolates and genetically manipulated mutants are frequently impaired in the respiratory chain. They may carry mutations in hemB that is involved in hemin biosynthesis [von Eiff et al., 1997], or menDrequired for menaquinone production [Bates et al., 2003], or may be auxotrophic for thymidine. Notably, their smaller size pertains to the colony but not on the single-cell level [Proctor et al., 2006]. Whether SCVs are also affected in persister production is an important question which we examined in this study.

To isolate S. aureus persister cells, we wanted to identify conditions under which (1) the bulk of cells of a culture is rapidly eradicated after the addition of an antimicrobial compound; (2) a low percentage of cells within this culture is still capable of forming colonies on nonselective rich media after up to 48 h, and (3) the colony-forming unit (CFU) count of the liquid culture does not increase to a sizeable extent after the antibiotic challenge, but exhibits a plateau or a slow decrease during the course of the experiment. Such outcomes are reflected by biphasic killing curves when CFU counts are plotted against time [Balaban et al., 2004; Dhar and McKinney, 2007]. During extensive in vitro antibiotic challenge experiments, we tested model laboratory S. aureus strains (table 1) exhibiting either normal growth or an SCV phenotype in liquid rich medium by five different antibiotics and one disinfectant applied at 1-, 10-, and 100-fold the minimal inhibitory concentration (MIC). These compounds and respective MIC values determined at the onset of this study are summarized in table 2. In a first approach, time-dependent eradication of exponentially grown cells was monitored. Results explicated in the following are restricted to conditions under which the trajectories of resulting killing curves indicated either persister isolation or other interesting phenomena. When ciprofloxacin was added to the cultures at 10-fold MIC, a rapid decrease in CFU counts was observed ranging from 94% (menD) to 99% (SA113) reduction after 2 h, followed by rather stable values for up to 24 h, indicating the presence of ciprofloxacin-tolerant persisters (fig. 1a). SCV strains were more tolerant to the drug than strains HG001, HG002, and HG003 (in the following referred to as ‘HG strains’ due to mostly uniform behavior in this study) and SA113, which were killed by ciprofloxacin at comparable rates. In the presence of 100-fold MIC of rifampin, live counts of SA113 decreased drastically, whereas killing of HG strains and particularly SCV strains was considerably slower during the first 3 h (fig. 1b). When cultures were challenged by 10-fold MIC of tobramycin, live counts of all strains decreased by three or four orders of magnitude during the 1st hour and leveled off within the next 7 h in non-SCV strains (fig. 1c). Colonies grown from HG cells surviving this treatment gave rise to two different morphotypes. About 78% of them had the size of the parent strains, whereas ∼22% had a stable, non-reverting SCV appearance. Treatment with tobramycin at 100-fold of the MIC resulted in rapid killing of the entire hemB and menD population, whereas CFU values of SA113 decreased by about four orders of magnitude after 1 h, followed by a rather slow decrease during the next 5 h (fig. 1d). Inexplicable fluctuations in case of HG strains precluded unequivocal conclusions under these conditions. Treatment of hemB and menD with 10-fold MIC of daptomycin resulted in a slightly biphasic killing pattern (fig. 1e), whereas the majority of SA113 cells were killed within the 1st hour and the remainder was eradicated completely after 4 h. Again, respective values were less meaningful with HG strains. The activity of daptomycin is strongly increased in the presence of calcium ions [Steenbergen et al., 2005], and indeed applying 100- or 10-fold the MIC of daptomycin with 50 µg/ml Ca²⁺ eradicated all cultures within up to 4 or 5 h, respectively. The complete set of killing curves with exponential-phase S. aureuscultures, including treatments with additional antimicrobials listed in table 2, is provided in online supplementary figure 1A–G (for all online supplementary material, see www.karger.com?doi=10.1159/000342449).

In order to reappraise whether all stationary growth phase S. aureus cells actually represent persisters, overnight cultures were treated with the same set of antimicrobials (except penicillin G) applied at identical concentrations as before. In contrast to treatment with higher concentrations of the quaternary ammonium compound benzalkonium chloride, which resulted in the complete eradication of all strains tested within 3 h or less (online suppl. fig. 1G), a high degree of drug tolerance irrespective of strains or antibiotic concentration was observed in most cases. Notably, tobramycin or daptomycin challenge resulted in selective eradication of cells. CFU values of most strains declined slowly upon treatment with 100-fold MIC of tobramycin whereas live counts of menD cells decreased by 94.6% within the 1st hour or by 99.9% within 4 h, respectively, followed by an 84% reduction between time points t = 4 h and t = 8 h (fig. 2a). Challenging SA113 with 100-fold MIC of daptomycin yielded a killing curve that obeyed a very distinct biphasic pattern (fig. 2b). The bulk of cells (99.98%) were eradicated within 1 h, followed by a subsequent CFU decrease by only about 13%/h for the next 7 h or by about 4.3%/h for the next 23 h on average, respectively. Of note, killing curves observed with 100-fold MIC of daptomycin/Ca²⁺ were biphasic not only in case of SA113 but clearly also for the menDstrain which had been unaffected by daptomycin alone (fig. 2c). CFU counts of HG strains declined rather continuously whereas the hemB population remained relatively stable. Addition of 1 µg/ml hemin to the hemBculture resulted in killing kinetics almost indistinguishable to those of SA113 (online suppl. fig. 2). The complete set of killing curves observed with stationary phase S. aureus cultures can be found in online supplementary figure 3.

An overnight culture of SA113 was challenged simultaneously with 100-fold MIC each of daptomycin and tobramycin, to combine both conditions which had resulted in pronounced isolation of persisters before. Killing kinetics observed were clearly different compared to those of cultures treated with either of the drugs alone (fig. 3). After a 97% decline in CFU values during the first 3 h, the remaining population was entirely eradicated between time points t = 3 h and t = 4 h, which indicates a synergistic effect in antibacterial activity exerted by these two antibiotics.

In order to assess the stability of persister levels over longer periods of time, CFU values of cultures treated with 100-fold MICs of antibiotics for 21 days were monitored. All cultures underwent considerable decreases in live counts within the first 7 days or earlier, whereas untreated cultures maintained high CFU levels during the entire course of the experiment. Formation of colonies on nonselective plates ceased, when cultures had been challenged with daptomycin or tobramycin for 3 days, with rifampin for 7 days or with ciprofloxacin for 21 days (fig. 4). The recurring CFU increase of cultures in the presence of penicillin G is rationalized by a concomitant constant decrease in antimicrobial activity within up to 10 days, whereas all other compounds tested were active up to the time of complete persister eradication or beyond (online suppl. fig. 4).

Exponential cultures of SA113 were repeatedly sorted out for persisters by treatment with 100-fold MIC of tobramycin or 10-fold MIC of daptomycin to reveal whether antibiotic tolerance of S. aureus persisters is inherited over many generations. As depicted in figure 5a and b, respective killing curves after up to four cycles still resembled those of the original populations. The fact that persister levels were not elevated in later cycles indicates that the persister state of S. aureus is not vertically transmitted.

Although staphylococcal persisters had been described more than 60 years ago [Bigger, 1944], a profound analysis of this drug-tolerant and putatively dormant state for this genus is lacking. We hence addressed a number of different questions regarding S. aureus persisters: (1) Do all stationary phase cells of S. aureus reside in the persister state? (2) Do exponential phase cultures also contain persister cells and, if so, to what extent? (3) Is there a correlation between persister levels and strain and genotype backgrounds? (4) Does the kind and concentration of an antimicrobial agent applied to comparable cultures influence persister levels? To establish a method for isolating drug-tolerant S. aureuspersisters, antibiotics were applied which act on most of the common bacterial target structures, such as the cell envelope (penicillin, daptomycin), ribosome (tobramycin), RNA polymerase (rifampin), or topoisomerase (ciprofloxacin). Sorting for persisters was observed independently of the chemical nature or the target of the drug in accordance with previous studies on S. epidermidis [Shapiro et al., 2011] and other bacteria [Keren et al., 2011; Spoering and Lewis, 2001]. Generally, S. aureus cells in stationary growth phase were less vulnerable by antibiotics than those in exponential phase, arguably due to elevated persister levels [Wiuff et al., 2005]. Under most of the conditions we tested, the survival rates of stationary phase cells were either unaffected by antibiotic treatment (particularly upon challenge with 1- or 10-fold MICs of drugs) or CFU counts decreased rather constantly during 24 h (see online suppl. fig. 3). By contrast, the strong recurring increase in CFU counts of exponential phase SA113 treated with 100-fold MIC rifampin (fig. 1b) indicated genetically determined changes in susceptibility. These can be achieved by single amino acid exchanges in RNA polymerase β-subunit (RpoB) of S. aureus [Aubry-Damon et al., 1998]. Indeed, rifampin-challenged cells forming colonies on nonselective plates also grew well when restreaked on rifampin-supplemented plates (data not shown). During the antibiotic challenge procedure, a higher selective pressure for drug resistance was presumably exerted on populations mainly composed of rapidly killed cells. Accordingly, recovery of CFU counts was not observed with the more tolerant hemB and menD strains within 8 h. Eradication of exponential phase cells was most rapidly achieved with 10- and 100-fold MICs of the cyclic lipopeptide daptomycin, and this effect was augmented by the addition of Ca²⁺ cations [Steenbergen et al., 2005]. Indeed, daptomycin/Ca²⁺ is effective against a number of S. aureus strains including hemB[Baltch et al., 2008; Begic et al., 2009; Mascio et al., 2007], and killing kinetics of S. aureus ATCC 29213 cultures challenged with super-MICs of daptomycin/Ca²⁺ had previously also displayed biphasic trajectories [Mascio et al., 2007], albeit not as pronounced as in the present study. Treatment of HG cells with 10-fold MIC of tobramycin in exponential growth phase isolated persisters that gave rise to both normally growing and small-colony variants in agreement with previous studies using the related aminoglycoside compound amikacin against S. aureus [Singh et al., 2009]. Tiny staphylococcal colonies had also been discovered by Bradley et al. [1978] when clinical isolates were tested for tolerance towards oxacillin, cephalothin, or gentamicin. These had resumed the same size, pigmentation, and hemolytic characteristics as the parent strain after 24 h, which is in contrast to the phenotypically stable SCVs we obtained (fig. 1c). Whether the latter are the result of nongenetic heterogeneity [Tuchscherr et al., 2011] or of mutations [von Eiff et al., 1997; Wiuff et al., 2005] may be clarified by analyzing possible hemin, thymidine, thiamine, or menadione auxotrophies due to defects in hemB, menD, or other factors.

Persisters of S. aureus, like other bacteria, appear to be physiologically heterogeneous in terms of robustness or fitness [Allison et al., 2011a]. Exposure of identical S. aureus cultures, assumed to contain a reproducible mixture of cells in different physiologic states, to different antibiotics resulted in a variety of killing curves. Some of these exhibited pronounced biphasic trajectories, indicating that the respective drug rapidly sorts for highly robust persisters (e.g. SA113 treated with 100-fold daptomycin in fig. 2b), whereas curves with rather smooth transitions from rapid to slower killing speak in favor of a compound which can be endured for longer periods also by less tolerant persisters. This is exemplified by menD cultures challenged in the stationary growth phase by tobramycin (fig. 2a), which kills nongrowing cells by interrupting translation. Resulting misfolded toxic peptides [Correia et al., 2006] are assumed to influence the persister state [Vazquez-Laslop et al., 2006].

As expected, not even very robust S. aureus persisters seem to be indefinitely inert against antimicrobials. During long-term experiments persisters of SA113 were eradicated completely when exposed to active drugs at high concentrations within 3–21 days (fig. 4; online suppl. fig. 4). Simultaneous treatment of stationary phase SA113 cultures with high concentrations of daptomycin and tobramycin eradicated persister cells completely within few hours, revealing a synergistic bactericidal effect exerted by this pair of drugs. It is conceivable that passive tobramycin entrance into cells was facilitated by daptomycin-mediated membrane damage, similar to the activated uptake of aminoglycosides into persister cells upon generation of a proton motive [Allison et al., 2011b]. In line with our results, Baltch et al. [2008] observed more efficient killing of S. aureus strains SH1000 and Newman upon treatment with a combination of daptomycin/Ca²⁺ and gentamicin than with either of the drugs alone. These results support the strategy to combat bacterial infections by applying suitable combinations of drugs [Cottarel and Wierzbowski, 2007]. It should be noted, though, that our study is generally restricted to provide model conditions for isolating persisters, because conditions of 100-fold the MIC of antibiotics can hardly be achieved during clinical treatment.

The use of different well-established laboratory S. aureus strains and mutants demonstrates the genetic and genomic influence on bacterial persister levels. Generally, among the strains tested, SA113 was most susceptible to the applied antimicrobials, which may in part be due to its short generation time, particularly compared to SCV strains. Except for chromosomal resistance cassettes inactivating hemB or menD, SCVs of this study are isogenic to SA113 [Gaupp et al., 2010] and indeed chemical complementation of a hemB mutant resulted in an eradication pattern almost identical to SA113 (as tested with 100-fold MIC of daptomycin/Ca²⁺, see online suppl. fig. 2). By contrast, genomic comparison between HG strains and SA113, all derived from isolate RN1 [Novick, 1967], is less straightforward. SA113 was obtained by unspecific chemical mutagenesis and its genome sequence is unavailable, whereas HG strains resulted from targeted repairs of the transcriptional regulators RsbU and TcaR [Herbert et al., 2010]. The generally homogenous killing curves among HG strains indicate that neither of these two factors affects S. aureus persister levels. This is particularly remarkable in case of RsbU, which positively regulates the alternative transcription factor σ^B [Giachino et al., 2001], influences the expression of the mazEF toxin-antitoxin system [Donegan and Cheung, 2009; Senn et al., 2005], and downregulates the agr locus [Bischoff et al., 2001] encoding quorum sensing factors [Novick and Geisinger, 2008]. Overexpression of the MazF toxin has been described to induce bacteriostasis in S. aureus[Fu et al., 2009], which may have implications for persister formation. An interesting new finding points towards indole-induced persister formation in E. coli [Vega et al., 2012] and studies by Shapiro et al. [2011] showed that quorum sensing does not alter persister levels in S. epidermidis. Further studies will clarify whether small molecules influence persister levels in S. aureus.

We here demonstrate that growth phase, strain background, and genotype influence S. aureus persister levels. As in other bacteria, the concentration of these drug-tolerant cells clearly increased from exponential to stationary phase cultures and the persister state does not seem to be inherited [Keren et al., 2004a]. A systematic analysis of persister formation with clinical S. aureusisolates and mutants is now required and so are clinically applicable strategies to kill S. aureus persisters. Recently, interesting approaches exploiting compounds specifically targeting persisters or the generation of proton motive force in persisters have been described [Allison et al., 2011b; Kim et al., 2011]. In the future, we need to learn more about the distribution of drug tolerant S. aureus strains and the genetic mechanisms pertaining to this phenomenon in order to catch up with the state of knowledge regarding genetically drug resistant S. aureus strains and lineages [Chambers and Deleo, 2009].

S. aureus strains used in this study are listed in table 1. Unless stated otherwise, bacteria were cultured at 37°C with aeration in baffle flasks containing tryptic soy broth (TSB; Sigma) at a 1:6 culture-to-flask ratio or on tryptic soy agar (TSA). Liquid cultures were shaken at 150 rpm. To chemically complement the SCV phenotype of the SA113 hemB mutant, hemin (Sigma) was added to a final concentration of 1 µg/ml where indicated.

Daptomycin analytic grade powder (designated ‘Cubicin’) was purchased from Novartis Pharma. Facultatively, Ca²⁺ cations (50 µg/ml final concentration), provided as CaCl₂, were added to daptomycin-treated cultures, to increase antibiotic activity where indicated. Ciprofloxacin was obtained from Fluka. Tobramycin, penicillin G, rifampin, and benzalkonium chloride were purchased from Sigma. Solutions of antimicrobials were prepared freshly prior to each application and were sterilized using a filter of 0.2-µm pore size (Whatman).

The MIC of each antimicrobial compound was determined in triplicate by a conventional broth microdilution technique in TSB. The MIC was defined as the lowest antibiotic concentration that inhibited visible bacterial growth (according to OD₅₇₈ measurements) after 24 h of incubation at 37°C.

To determine the number of persister cells in exponential phase, cells were grown overnight in 8 or 16 ml TSB and were inoculated to 16 ml of fresh media to an initial OD₅₇₈ of 0.07. Cultures were shaken for 1.5–2 h (for normally growing cells) or 2.5–3 h (for SCVs), until an OD₅₇₈ of approximately 0.5 was reached. To determine the number of persister cells in stationary phase, overnight cultures were used. In both cases, 2.24 ml of the overnight or the exponential phase culture was transferred to 14-ml culture tubes (Greiner), antimicrobials were added at 1- to 100-fold MIC as indicated and cultures were shaken for 24 h or for 21 days during long-term experiments. For CFU determinations, 100 µl samples were taken before and during antimicrobial challenge on an hourly basis during the first 8 h and once after 24 h, or after 1, 2, 3, 7, 10, 14, and 21 days during long-term experiments. Cells were washed in 1% NaCl solution and spotted as 10-µl aliquots of serial dilutions onto TSA plates. Colonies were counted after incubation for 24 h (normally growing colonies) to 48 h (SCVs) at 37°C. The lower limit of quantification was 100 CFU/ml. All time-kill experiments were conducted using at least three biological replicates.

2.24 ml of exponential phase cultures were transferred to 14-ml culture tubes and cells were exposed to 100-fold MIC of tobramycin or 10-fold MIC of daptomycin for 3 h. Subsequently, cells were washed in 1% NaCl solution, resuspended in 16 ml fresh TSB and cultured overnight. Dilution of cells in fresh medium and growth to exponential phase finished the first cycle. Cells were then again subjected to antimicrobial treatment as described and the procedure was conducted in a total of four consecutive cycles.

We thank Tanja Hildebrandt for technical assistance and Friedrich Götz for support. This work was supported by grants BE4038/1 and BE4038/2 within the priority program 1316 ‘Host Adapted Metabolism of Bacterial Pathogens’ and by the research training group Graduate College 685 ‘Infection Biology: Human- and Plant-Pathogenic Bacteria and Fungi’ of the German Research Foundation (DFG), as well as by NIH grant R01 GM061162.