Evolution of Staphylococcus aureus under Vancomycin Selective Pressure: the Role of the Small-Colony Variant Phenotype

Abstract

Staphylococcus aureus small-colony variants (SCVs) often persist despite antibiotic therapy. Against a 10⁸-CFU/ml methicillin-resistant S. aureus (MRSA) (strain COL) population of which 0%, 1%, 10%, 50%, or 100% was an isogenic hemB knockout (Ia48) subpopulation displaying the SCV phenotype, vancomycin achieved maximal reductions of 4.99, 5.39, 4.50, 3.28, and 1.66 log₁₀ CFU/ml over 48 h. Vancomycin at ≥16 mg/liter shifted a population from 50% SCV cells at 0 h to 100% SCV cells at 48 h, which was well characterized by a Hill-type model (R² > 0.90).

TEXT

Staphylococcus aureus is a virulent pathogen responsible for a myriad of infections ranging from minor community-acquired skin and soft tissue infections to severe nosocomial infections (1). While the current IDSA guidelines recommend vancomycin as the primary agent for treatment of methicillin-resistant S. aureus (MRSA) infections, the utility of the drug has been brought into question due to increasing reports of heterogeneous resistance, treatment failure, and nephrotoxicity (2–4). Despite the global decrease in vancomycin susceptibility, the exact mechanism by which S. aureus develops resistance is not well understood (5). It has been suggested that S. aureus adapts by utilizing an array of genotypic alterations that arise stepwise during the selective pressure of antimicrobial therapy (6, 7).

One pathway that S. aureus may exploit during the evolution of antimicrobial resistance is the development of small-colony variants (SCVs) that grow slowly relative to strains of the normal phenotype (NP) (8–10). In vitro testing and macrophage models have confirmed that the SCV phenotype is less susceptible to vancomycin (11, 12). Studies with other antibiotics also suggest that SCV subpopulations may cooperate with NP S. aureus to attenuate antimicrobial activity (13). At present, it is unknown whether SCVs alter vancomycin pharmacodynamics through interactions with NP S. aureus or how the selection of a vancomycin regimen influences the relationship between the two phenotypes. The objective of the current study was, therefore, to utilize reconstructive population biology to determine how the interplay of both phenotypes alters vancomycin pharmacodynamics.

The MRSA strain COL (NP) and its isogenic hemB knockout Ia48 (COL hemB::ermB, a stable SCV phenotype) were utilized. The creation of the mutant strain and its features were previously characterized (14). Prior to each experiment, a solution of vancomycin was prepared using analytical-grade powder (Sigma Chemical, St. Louis, MO) in the following concentrations: 0, 0.5, 1, 2, 4, 8, 16, 32, 64, and 128 mg/liter. Brain heart infusion (BHI) broth supplemented with magnesium (12.5 mg/liter) and calcium (25 mg/liter) was used for every experiment. SCV and NP cell suspensions were volumetrically titrated to achieve 5 different starting compositions, with a total bacterial load of 10⁸ CFU/ml. Two experiments were conducted exclusively investigating the NP or the SCV phenotype (0% SCV/100% NP cells and 100% SCV/0% NP cells), and three mixed populations were investigated with starting inoculums of approximately 1% SCV/99% NP cells, 10% SCV/90% NP cells, and 50% SCV/50% NP cells adapted as previously described (30). Time-killing experiments were conducted over 48 h as previously described (11). In mixed-population experiments, bacteria were plated on both plain BHI agar and BHI containing gentamicin (2.0 mg/liter) to select for the SCV phenotype.

An integrated pharmacokinetics/pharmacodynamics (PK/PD) approach was utilized as defined previously (11). Using the linear trapezoid rule, the area under the CFU curve in a 48-h time span (AUCFU_0–48) was calculated for each concentration and a growth control. The log ratio area was then calculated as the logarithm of the AUCFU_0–48 of drug divided by the AUCFU_0–48 of the growth control (equation 1).

$$\text{log ratio area}={\text{log}\hspace{0.17em}}_{10}\left(\frac{{\text{AUCFU}}_{\text{drug}}}{{\text{AUCFU}}_{\text{control}}}\right)$$ (1)

$$E=E_0-\frac{E_{\text{max}\hspace{0.17em}}\times {(C)}^H}{{({EC}_{50})}^H+{(C)}^H}$$ (2)

A four-parameter concentration-effect Hill-type model was fit to the effect parameter using Systat (version 12; Systat Software Inc., San Jose, CA) (11). In equation 2, E (dependent variable) represents the log ratio area, E₀ is the effect at a vancomycin concentration of 0, E_max is the drug's maximum effect, C is the concentration of vancomycin, EC₅₀ is the vancomycin concentration displaying 50% of the maximum effect, and H is the sigmoidicity constant. The overall model fits were assessed based on coefficients of determination (R²).

The antibiotic activity and pharmacodynamics of vancomycin against each population are depicted in Fig. 1. Vancomycin concentrations of >16 mg/liter achieved bactericidal activity against the exclusively NP population by 24 h (4.99 log₁₀ CFU/ml reduction by 48 h) and a maximal reduction against the exclusively SCV population of 1.66 log₁₀ CFU/ml by 48 h. Despite the presence of SCVs in the mixed populations, at 48 h, vancomycin concentrations of >16 mg/liter achieved maximal reductions of 5.39, 4.50, and 3.28 log₁₀ CFU/ml in the experiments initially containing 1%, 10%, and 50% SCV cells, respectively.

FIG 1.

Vancomycin time-kill experiments involving exclusively strains of the NP at an inoculum of 0% SCV/100% NP cells (A) and exclusively strains of the SCV phenotype at an inoculum of 100% SCV/0% NP cells (B), as well as three mixed-population experiments consisting of 1% SCV/99% NP (C), 10% SCV/90% NP (D), and 50% SCV/50% NP (E) cell inoculums. The pharmacodynamic relationship between log ratio area and vancomycin concentration are also displayed for 0% SCV/100% NP (F), 100% SCV/0% NP (G), 1% SCV/99% NP (H), 10% SCV/90% NP (I), and 50% SCV/50% NP (J) cells. R², E_max, and EC₅₀s are listed next to the corresponding Hill plots (percent standard errors are listed parenthetically).

Overall, the Hill-type model displayed excellent fits to the log ratio areas, with R² values exceeding 0.99 for each experiment. The model-fitted parameters E_max, EC₅₀, and R² are listed for each time-killing experiment in Fig. 1. When comparing the exclusively NP and SCV populations, the E_max for the NP population was nearly double the E_max of the SCV population (E_max of NP = 2.17, E_max of SCV = 1.15). In the mixed-population experiments, the E_maxs for the three populations were 1.96, 1.92, and 1.67, which corresponded to starting inoculums containing 1%, 10%, and 50% SCV subpopulations, respectively. As the proportion of the SCV in the starting inoculum increased, the activity of vancomycin decreased in a manner that was consistent with the trend observed in colony counts.

In addition to performing the PK/PD analysis, we plotted the maximum percentages of SCV cells observed for each vancomycin concentration for the three mixed populations, and these are depicted in Fig. 2. The concentration plots were also fitted with a Hill-type function to reveal the relationship between drug concentration and the proportion of SCVs. In all three experiments, vancomycin concentrations of >8 mg/liter resulted in a decrease in the total bacterial population, with a concurrent rise in the ratio of SCV to NP S. aureus cells. After exposure to vancomycin at >16 mg/liter for 48 h, the inoculum initially containing 50% SCV cells was completely dominated by 100% SCV cells. Similarly, vancomycin concentrations of >16 mg/liter raised a 10⁶-CFU/ml SCV subpopulation from about 1% of the starting inoculum to approximately 33% of the total population.

FIG 2.

Colony counts used to track SCV subpopulations. In mixed-culture experiments, samples were plated on plain BHI agar as well as BHI containing 2.0 mg/liter gentamicin. Gentamicin plates permitted only the growth of SCVs, and the corresponding CFU plots derived from the drug plates are displayed for 1% SCV/99% NP (A), 10% SCV/90% NP (B), and 50% SCV/50% NP (C) cell inoculums. Additionally, the maximum percentages of the SCV detected over 48 h for each vancomycin concentration are plotted and fit to a Hill-type function for 1% SCV/99% NP (D), 10% SCV/90% NP (E), and 50% SCV/50% NP (F) cell inoculums.

S. aureus has a remarkable ability to evolve in the face of antimicrobial therapy. One mechanism by which S. aureus may persist in difficult-to-treat, deep-seated infections is through the formation of SCVs. Here, we sought to determine the interaction of the SCV phenotype with the NP under vancomycin pressure. Clinically isolated SCVs typically have an inability to synthesize menadione, hemin, or thymidine, resulting in a disrupted metabolism that confers a low growth rate (10). Due to the high rate of reversion from the SCV phenotype back to the NP, clinical SCV isolates of S. aureus are often difficult to study in vitro (14). However, insertion of an ermB cassette into the hemin biosynthesis gene hemB generates a stable SCV phenotype that mimics clinically isolated electron transport-defective SCVs that are auxotrophic for hemin, making hemB mutants the ideal SCVs for the current investigation.

Building upon previous studies, we determined that vancomycin activity against cells with the SCV phenotype was diminished relative to that against cells of the NP at a high bacterial density (11). Moreover, in all of the mixed-population experiments involving an initial inoculum of 10⁸ CFU/ml, attenuation of vancomycin activity was greater among S. aureus populations containing a larger proportion of cells of the SCV phenotype. Findings by Massey and Peacock suggest that cells of the SCV phenotype of S. aureus may interact with cells of the NP to augment the survival of the parent phenotype during antimicrobial treatment (13); in the presence of gentamicin, they observed that the SCV acidified its environment, thereby inhibiting gentamicin activity and protecting NP gentamicin-sensitive bacteria. In the current study, vancomycin retained bactericidal activity regardless of the proportion of the SCV cells in the starting inoculum, which contrasts with the collaborative survival observed by Massey and Peacock during gentamicin exposure. These findings suggest that the two phenotypes do not cooperate during vancomycin therapy.

It is also noteworthy that high concentrations of vancomycin selected for the SCVs in mixed-population experiments. It has been postulated that S. aureus SCV cells exist in a dynamic equilibrium with cells of the NP, and the emergence of SCVs in persistent infections is attributed to the survival advantage of preexisting SCVs compared to the survival of strains with the NP during vancomycin exposure (15, 16). In the current work, the equilibrium was likely shifted in favor of the SCVs through a similar pathway: vancomycin was more effective at eradicating the fast-growing NP cells of S. aureus, thus allowing the slower growing but more resistant SCV cells to expand in number in the overall population.

Although SCVs have been isolated from persistent infections and subsequently analyzed, the population dynamics of S. aureus have yet to be characterized in severe infections, such as endocarditis (17–19). In lieu of in vivo data revealing the interplay between S. aureus phenotypes during antibiotic treatment, our observations suggest that vancomycin will preferentially select for the amplification of SCV subpopulations. Increasing vancomycin exposure against a MRSA population that is not fully eradicated may be counterproductive, as more exposure may result in a bacterial shift toward a variant phenotype that is adept at evading antimicrobial effects and persisting intracellularly (20, 21). Upon the discontinuation of antibiotics, the SCVs may revert to the NP and perpetuate a cycle of treatment failure and recurrent infection (22, 23).

Thus, clinicians should consider vancomycin combination regimens or alternative antimicrobials in patients with severe persistent MRSA infections in which the SCV phenotype may play a role. In vitro analyses have identified fluoroquinolones and oritavancin as retaining high levels of activity against S. aureus SCVs (24, 25). Daptomycin has also been shown to retain more activity than vancomycin in vitro against SCVs (26), and β-lactam combinations with daptomycin may offer a new option for combating SCVs (27, 28). Macrophage models have revealed not only that oritavancin, rifampin, moxifloxacin, and quinupristin-dalfopristin are active against intracellular SCVs (12) but that SCVs may be more susceptible to β-lactams while in the intracellular domain as well (29). Combination therapy including either rifampin or oritavancin appears to be particularly effective at eradicating intracellular SCVs (12). In the clinical setting, combinations integrating either rifampin or fluoroquinolones with other antibiotics have been frequently utilized following the identification of SCVs, and in the case of a device-related infection, surgical debridement is frequently necessary (10). Further investigations exploring intramacrophage persistence, the host immune response, and the extended time course of SCV emergence will help elucidate counterstrategies for eliminating S. aureus SCVs.