MRSA Isolates from Patients with Persistent Bacteremia Generate Nonstable Small Colony Variants In Vitro within Macrophages and Endothelial Cells during Prolonged Vancomycin Exposure

ABSTRACT

Staphylococcus aureus (especially methicillin-resistant S. aureus [MRSA]) is frequently associated with persistent bacteremia (PB) during vancomycin therapy despite consistent susceptibility in vitro. Strategic comparisons of PB strains versus those from vancomycin-resolving bacteremia (RB) would yield important mechanistic insights into PB outcomes. Clinical PB versus RB isolates were assessed in vitro for intracellular replication and small colony variant (SCV) formation within macrophages and endothelial cells (ECs) in the presence or absence of exogenous vancomycin. In both macrophages and ECs, PB and RB isolates replicated within lysosome-associated membrane protein-1 (LAMP-1)-positive compartments. PB isolates formed nonstable small colony variants (nsSCVs) in vancomycin-exposed host cells at a significantly higher frequency than matched RB isolates (in granulocyte-macrophage colony-stimulating factor [GM-CSF], human macrophages PB versus RB, P < 0.0001 at 48 h; in ECs, PB versus RB, P < 0.0001 at 24 h). This phenotype could represent one potential basis for the unique ability of PB isolates to adaptively resist vancomycin therapy and cause PB in humans. Elucidating the molecular mechanism(s) by which PB strains form nsSCVs could facilitate the discovery of novel treatment strategies to mitigate PB due to MRSA.

INTRODUCTION

Staphylococcus aureus is a leading cause of bacteremia worldwide, especially in endovascular infections (1). This relationship is especially relevant to antimicrobial-resistant strains such as methicillin-resistant S. aureus (MRSA), which is associated with poor clinical outcomes despite seemingly appropriate therapy (2–4). Findings from a multinational clinical trial of S. aureus bacteremia demonstrated two distinct categories of MRSA isolates emerging during antibiotic therapy, persistent and resolving bacteremia (PB and RB, respectively) (4). Patients with persistent bacteremia were defined as having MRSA-positive blood cultures for at least 7 days despite vancomycin therapy. In contrast, patients with resolving bacteremia immediately cleared their bloodstream of MRSA (5–7). Remarkably, infection with such clinical PB and RB isolates in infective endocarditis animal models recapitulates the clinical categories experienced by the original-source patients (2, 3), suggesting that a specific adaptive mechanism(s) emerging in vivo likely underpins the PB versus RB phenotypes.

Several studies have aimed to identify important differences between PB and RB isolates to understand the underlying mechanisms that allow PB strains to cause persistent infections (3, 6–11). Of note, PB isolates are less susceptible to host antimicrobial peptides (3), display enhanced biofilm formation in response to vancomycin (12), and have elevated expression of purine biosynthesis genes compared to RB MRSA isolates (10). However, to date, no single phenotypic characteristic sufficiently explains the PB phenotype. Guérillot et al. (13) recently demonstrated that S. aureus undergoes unstable chromosomal rearrangements associated with persistent phenotypes in vivo. Such nonclassical adaptive mechanisms involved transient and asymmetric chromosomal inversions, allowing rapid switching of phenotypes associated with immune evasion. Moreover, there is increasing evidence that multivariable phenotypic and genotypic signatures differentiate PB from RB isolates in human clinical infections. Conceivably, these differential signatures of PB versus RB isolates arise in response to complex combinations of stimuli in vivo. For example, the intersection of a distinct clinical strain with host immune cells and vancomycin within specific host microenvironments is hypothesized to be relevant to endovascular infection establishment and outcomes. In support of this concept, Chan et al. (14, 15) demonstrated that macrophages represent an integral component of host defense via innate immune memory in protection against S. aureus infection. Likewise, Mba Medie et al. (6) and Chang et al. (5) discovered epigenetic signatures of host susceptibility to PB versus RB outcomes in human S. aureus bacteremia. Hence, there are undoubtedly numerous host-MRSA interactions that strongly influence PB versus RB outcomes in MRSA bacteremia when treated with antibiotics appropriate to their susceptibilities. Understanding these interactive responses would likely provide key insights into the mechanisms that allow PB bacteria to persist within the host.

In vivo, macrophages represent a critical immune gate, but also a tipping point, during S. aureus bloodstream infection (16). While macrophages engulf and normally kill ingested microbes, previous work has demonstrated that a subset of S. aureus not only survives killing but eventually begins to grow while contained within mature phagolysosomes despite being exposed to a multitude of stressors within this compartment (16–18). Along with reversible chromosomal rearrangements (13), S. aureus engages two-component regulatory systems such as GraRS to adapt to this intracellular niche, especially to pH and antimicrobial “cues” (18). Additionally, during bloodstream infection, bacteria may bind to and invade endothelial cells (ECs) (19). This strategy is likely to enable immune evasion of MRSA and potentially serve as an intracellular reservoir for continuous hematogenous seeding. In addition, the endothelium acts as a barrier that can selectively control the passage of pathogens from intravascular to extravascular spaces; thus, disruption of endothelial barrier integrity can facilitate infection of deeper parenchymal tissues (20). This strategy may also be important during antibiotic therapy of bloodstream infections, as the intracellular EC niche could represent an environment protected from antibiotic therapy, thus enabling persistent infection (19, 21–23).

The current study was designed to investigate whether the intracellular environments within macrophages and/or ECs play a role in persistence adaptations of PB versus RB MRSA clinical isolates. The studies here demonstrate that PB and RB MRSA localize to mature lysosome-associated membrane protein-1 (LAMP-1)-positive phagolysosomes in macrophages and LAMP-1-positive endolysosomes in ECs. Moreover, we find that vancomycin exposure of PB-infected macrophages and ECs uniquely induces increased formation of nonstable small colony variants (nsSCVs) compared to RB-infected cells. These findings suggest that PB isolates of MRSA can differentially adapt to their specific intracellular microenvironments in the context of vancomycin exposure by converting to an adaptive SCV morphotype possessing the ability to survive host cell responses and vancomycin treatment.

RESULTS

PB isolates exhibit decreased intracellular replication in macrophages.

Viable counts for all MRSA clinical isolates obtained from RB outcomes increased in infected macrophages by 6- to 10-fold from 1.5 to 12 hours postinfection (hpi) (Fig. 1A). This pace of replication was similar to that observed for control S. aureus strain LAC-USA300 (Fig. 1A) (17). In contrast, all MRSA clinical isolates obtained from PB outcomes displayed significantly reduced growth within macrophages, similar to the control graS mutant of S. aureus LAC-USA300, previously shown to have replicative defects within cultured macrophages (17).

FIG 1.

Persistent bacteremia strains have decreased intracellular replication in macrophages compared to resolving strains and USA300. Macrophages were infected with USA300, USA300 ΔgraS, 3 representative PB strains, and 3 representative RB strains. (A) Growth of the indicated strains, as determined by gentamicin protection assay, is shown. The data are expressed as fold change in CFU per milliliter, where bacterial counts for a given strain at 12 h postinfection are divided by the number of bacteria present at 1.5 h postinfection for the same strain. (B) Graph depicting the percentage of macrophages that contain replicating USA300, USA300 ΔgraS, PB, and RB strains as determined by fluorescence-based proliferation assays at 12 h postinfection. The data are the mean ± standard deviation of three independent experiments. Ordinary one-way ANOVA (Tukey’s posttest) was used to determine statistical significance. **, P < 0.01; ns, not significant.

A fluorescence-based proliferation assay (24) was used to corroborate these observations and to determine the percentage of macrophages containing replicating MRSA. This analysis showed that significantly fewer macrophages contained replicating PB MRSA than those infected with RB isolates (Fig. 1B; see Fig. S1 in the supplemental material). Taken together, these data suggest that PB strains display slower growth, relative to RB MRSA, within macrophages.

PB and RB isolates of MRSA proliferate within macrophage phagolysosomes.

After initial interaction with macrophages, S. aureus LAC-USA300 is rapidly internalized and trafficked to phagosomes, which then mature into phagolysosomes. These intracellular environments are characterized as lysosome-associated membrane protein-1 (LAMP-1)-positive membrane-constrained compartments in which S. aureus is exposed to oxidative and nonoxidative antimicrobial effectors (17). To ascertain whether the fate of PB and RB isolates differed from that of the LAC-USA300 control, we performed LAMP-1 immunostaining of PB- versus RB-infected RAW 264.7 macrophages at 12 hpi. This time point was selected because it is a time when S. aureus USA300 is known to display intracellular growth in a subset of macrophages (17, 18). This analysis revealed that >96% of the replicating (i.e., green fluorescent protein [GFP]-positive, eFluor 670-negative) PB or RB bacteria were constrained by LAMP-1-positive membranes at 12 hpi (Fig. S2C). To confirm the MRSA was indeed actively replicating within phagolysosomes, we performed fluorescent dextran pulse-chase experiments as previously described (17). This analysis revealed that GFP-positive, eFluor-670-negative RB and PB isolates (i.e., replicating bacteria) colocalize with dextran (Fig. S2D), confirming these bacteria are within phagolysosomes and that PB and RB did not differ from LAC-USA300 in this regard.

Antibiotic promotes increased frequency of nsSCV formation in PB isolates within macrophages.

Given the potential role that vancomycin exposure could play in PB formation, we sought to characterize how vancomycin treatment of infected cells affects intracellular MRSA. We found that in the presence of extracellular fluid-phase vancomycin, the burden of S. aureus for both PB and RB bacteria was significantly reduced after 2 consecutive days of vancomycin exposure (Fig. 2A). Moreover, inspection of the resulting plated bacterial colonies recovered from 48-h vancomycin-treated macrophages revealed microcolonies that are characteristic of S. aureus small colony variants (SCVs) (25) (Fig. 2B and C, arrows). The SCVs were characterized by small colony size, reduced pigmentation, and a paucity of hemolysis on blood agar. To begin to characterize these colonies, we restreaked them onto tryptic soy agar (TSA) and blood agar plates and observed that they regrew virtually identically to the parental strains, with normal colonial morphologies, pigmentation, and hemolytic pattern. This finding indicated that this morphotypic switch within macrophages likely represents the emergence of a nonstable SCV phenotype (nsSCV).

FIG 2.

Prolonged antibiotic exposure promotes the formation of intracellular nsSCVs. (A) Graph showing the CFU per milliliter from USA300, PB, and RB strains recovered from 1.5 h (post-gentamicin protection), 24 h, and 48 h postinfection. (B) Representative images showing the nsSCV recovered from the prolonged infections (48 hpi). (C) Representative images of blood tryptic soy agar used to phenotypically confirm the presence of nsSCVs recovered 48 h postinfection. (D) Graphs depicting the percentage of nsSCV formation at 24 h and 48 h postinfection from bacteria grown within RAW 264.7 cells (left), M-CSF-derived human macrophages (middle), and GM-CSF-derived human macrophages (right). The plates were kept for 48 h to 72 h at 37°C for SCV quantification. In panels A and D, ordinary one-way ANOVA (Tukey’s posttest) was used to determine statistical significance; **, P < 0.01; ****, P < 0.0001; ns, not significant. The data are the mean ± standard deviation of three independent experiments.

Next, we sought to quantify the magnitude of nsSCV formation within macrophages. To this end, infected RAW 264.7 macrophages were cultured in the presence of vancomycin. Recovered nsSCV colonies (defined by the small colony morphotype, reduced pigment, and hemolysis) were enumerated and plotted as a percentage of the total intracellular bacterial population. This analysis revealed that at 48 hpi, and only when macrophages were treated with vancomycin, PB isolates form nsSCVs at a heightened frequency compared to RB isolates (Fig. 2D, left). Indeed, PB isolates formed nsSCVs at nearly twice the frequency as RB strains, S. aureus LAC-USA300 or its graS mutant. We next determined the frequency of nsSCV formation for MRSA isolates within human macrophages compared to murine-derived macrophages. Infection of macrophage colony-stimulating factor (M-CSF)-derived macrophages correlated with ~6% of the intracellular PB population forming nsSCVs. In contrast, only 3% of the infecting RB strain, or LAC-USA300, adopted the nsSCV phenotype (Fig. 2D, middle). A similar trend resulted from PB strains recovered from granulocyte-macrophage colony-stimulating factor (GM-CSF)-derived macrophages, where PB adopted the nsSCV morphotype twice as frequently as their RB counterparts (Fig. 2D, right). These observations confirm that vancomycin treatment fosters intracellular murine or human macrophage conditions or enters these cells, favoring an intracellular phenotypic switch of MRSA to nsSCV phenotype with greater propensity in PB versus RB clinical isolates.

In vitro growth in the presence of sub-MIC vancomycin and macrophage microbicidal effector-like conditions does not differentiate PB from RB MRSA.

To determine whether this phenomenon of generating nsSCVs, which differentiated PB and RB MRSA, required the intracellular compartment, we investigated whether any combination of in vitro stress conditions, mimicking what is encountered in macrophage phagolysosomes (i.e., acidic pH, antimicrobial peptides, and reactive oxygen species [ROS]), in association with vancomycin treatment, could result in differences in the frequency of nsSCV formation between PB and RB MRSA. Checkerboard assays were performed in a chemically defined medium (RPMI 1640) or in tryptic soy broth (TSB) at an acidic pH, and we identified the highest concentrations of an antimicrobial peptide (polymyxin B) and reactive oxygen species (hydrogen peroxide [H₂O₂]), along with sub-MIC vancomycin, that allowed growth of the individual strains (Tables S3 and S4). Under these growth conditions, we then assessed the ability of PB and RB MRSA to form nsSCVs. In RPMI 1640 medium at 24 h, nsSCVs were identified; however, no differences in the frequency of their occurrence between PB and RB isolates were observed (Fig. S3A). In this medium, a second subculture into fresh stressor-containing RPMI 1640 does not support growth. When these experiments were performed in TSB as the base growth medium, we were able to grow the cells upon subculture, and thus, the cultures were assessed for nsSCV formation at 24 h and upon further subculture into the same medium out to 48 h. Again, no differences in the frequency of the occurrence of nsSCVs could be identified between PB and RB isolates. Note that the graS mutant did not grow in these experiments, as it is sensitive to the antimicrobial peptide. Together, these data support the hypothesis that it is the intracellular compartment along with antibiotic treatment, and not simply in vitro growth, that specifically differentiates PB from RB in terms of their abilities to generate nsSCVs.

Extracellular antibiotic permits infected macrophages to maintain viability despite intracellular MRSA.

Having established that PB and RB MRSA can replicate within macrophage phagolysosomes, we sought to determine whether the bacteria displayed differences in their ability to intoxicate infected phagocytes. To this end, we again performed gentamicin protection assays and utilized the vital dye propidium iodide (PI) to identify infected cells with compromised plasmalemma integrity. We observed that macrophages containing either nonreplicating or replicating PB or RB MRSA were devoid of PI at 12 hpi without antibiotic added to cell culture (Fig. S4A). Note that soon after the 12-hpi time point, bacteria begin to emerge from macrophages and intoxicate the entire monolayer. In experiments where vancomycin is added to the cell culture, macrophages containing intracellular bacteria can be maintained and are viable for at least 48 hpi (Fig. S4B), indicating that the bacteria were primarily replicating in macrophages that were not necrotic. Moreover, we observed no differences in these phenotypes between PB and RB MRSA. Thus, a differential ability to intoxicate macrophages in these experiments is not a phenotype that differentiates these phenotypically different groups of MRSA.

Prolonged antibiotic exposure promotes the formation of intracellular nsSCVs.

As mentioned above, and as illustrated in Fig. S4, MRSA-infected macrophage cultures can be maintained for longer periods of time if antibiotic is included in the culture media, presumably, and as we have previously shown (26) because some of the antibiotic is pinocytosed from the fluid phase, and those antibiotic-containing endosomes may fuse with bacteria-containing phagolysosomes, thus impacting bacterial growth/outgrowth. We tested whether exposure to gentamicin from the gentamicin-protection assays, despite this treatment being transient (i.e., 1 h), could result in the formation of nsSCVs. Indeed, a previous report showed the isolation of SCVs from endothelial cells where high concentrations of gentamicin were left on for the duration of the experiment (19). Leaving a lower concentration of gentamicin in the macrophage cell cultures after infection with MRSA had a significant effect on controlling intracellular growth (Fig. 3A), as macrophage monolayers remained intact. In contrast, not including gentamicin resulted in the death of macrophages within 12 to 15 hpi and bacteria replicating extracellularly in the culture medium (Fig. 3A). Importantly, in this scenario where bacteria are coming out of the macrophages within 12 to 15 hpi, we failed to observe the generation of nsSCVs despite the bacteria subsequently being exposed to the extracellular gentamicin over the course of the remaining 48 h of the experiment (Fig. 3B). Additionally, as shown in Fig. 3B, although this treatment resulted in the generation of nsSCVs, it did not differentiate PB from RB MRSA in terms of the frequency with which they generate nsSCVs.

FIG 3.

Prolonged antibiotic exposure promotes the formation of intracellular SCVs. (A) Graph showing the fold change in CFU per milliliter of USA300, PB, and RB strains recovered from 1.5 hpi (i.e., the standard time where gentamicin is normally removed) versus 48 h postinfection, comparing leaving gentamicin (5 μg/mL) on the cells for the 48 h duration (left side) or washing it off at 1.5 hpi and incubating cells in the absence of antibiotic for the remainder of the 48 h (right side). (B) Graph depicting the percentage of SCV formation at 48 h postinfection from cultures grown as described for panel A. Blood agar plates were incubated for 48 h to 72 h at 37°C for SCV quantification. In panels A and B, ordinary one-way ANOVA (Tukey posttest) was used to determine statistical significance, **, P < 0.01; ****, P < 0.0001; ns, not significant. The data are the mean ± standard deviation of three independent experiments.

Exposure of intracellular bacteria to vancomycin differentiates PB from RB MRSA.

We next sought to address whether it was the exposure of intracellular bacteria to vancomycin that promoted the differential generation of nsSCV between the two morphotypes. We sought to rule out any effects the gentamicin exposure could exert during the transient treatment (1 h) employed our assays, as gentamicin exposure to S. aureus has previously been shown to generate SCVs, both in vitro and in vivo (19, 27–30). To do so, we infected macrophages, but 10 min before adding gentamicin to kill the extracellular bacteria, we added the inhibitors PIK-III and Dynasore to the infected cultures. These inhibitors have previously been used to inhibit pinocytosis and endocytosis (26) and block the uptake of the fluid phase (i.e., culture media), including the gentamicin contained within. Gentamicin and the inhibitors were kept on the infected macrophages for 1 h before being washed away. Vancomycin was then added to the cultures as above, and the infected macrophages were incubated until 48 hpi. In doing so, we were able to demonstrate (Fig. 4) that the exposure of cells containing intracellular bacteria to vancomycin promoted the differential generation of nsSCVs between PB and RB MRSA, as we had previously observed (see Fig. 2). Together, our data demonstrate that gentamicin promotes formation of nsSCV from intracellular USA300 and PB and RB bacteria, and at approximately the same frequency (Fig. 3B). However, we also demonstrate that it is exposure of infected cells to extracellular vancomycin that generates a higher frequency of nsSCV in PB than RB strains (Fig. 4B).

FIG 4.

Blockage of pinocytosis and endocytosis during gentamicin treatment does not affect nsSCV formation. (A) Graph showing the fold change in CFU per milliliter from USA300, PB, and RB strains recovered at 1.5 hpi (i.e., post-gentamicin protection) versus 48 hpi. Fluid-phase gentamicin uptake was blocked with PIK-III and Dynasore (left side) or left untreated with inhibitors (right side). At 1.5 hpi, wells were washed with PBS to remove gentamicin and inhibitors and vancomycin (5 μg/mL) was added, and cultures were incubated out to 48 hpi. (B) Graph depicting the percentage of nsSCV formation at 48 hpi from cultures grown as described for panel A. Blood agar plates were incubated for 48 h to 72 h at 37°C for SCV quantification. The plates were kept for 48 h to 72 h at 37°C for SCV quantification. In panels A and B, ordinary one-way ANOVA (Tukey’s posttest) was used to determine statistical significance; ****, P < 0.0001; ns, not significant. The data are the mean ± standard deviation of three independent experiments.

PB and RB isolates of MRSA proliferate within EC endolysosomes.

During bloodstream infections, S. aureus directly interacts with the EC barrier and can exploit these host cells to evade antibiotic and immune clearance. Thus, intra-EC survival is a biologically plausible step toward developing persistent endovascular infection (19, 20, 22). Therefore, we next investigated the intracellular fate of PB and RB isolates within ECs. We performed EC invasion assays using human microvascular ECs (HMEC-1) as representative host ECs. Uptake of S. aureus into ECs is dependent on exponential-growth-phase expression of fibronectin-binding proteins and fibronectin-dependent engagement of host cell integrins (8, 31, 32). We infected HMEC-1 cells with LAC-USA300 parent or its isogenic fnbAB mutant, which is impaired for EC invasion. Although the PB and RB isolates had similar levels of adhesion to EC monolayers (Fig. 5A), RB strains displayed a notably enhanced capacity to invade HMEC-1 ECs versus PB strains; further, the fnbAB mutant displayed no invasion of ECs and served as a relevant negative control for these assays (Fig. 5B). To investigate the ability of PB and RB MRSA to replicate inside HMEC-1 ECs, infected cells were incubated after gentamicin treatment until 6 hpi at which time they were lysed and plated to determine intra-EC bioburdens. This analysis revealed that, in the time frame investigated, PB and RB MRSA displayed similar intracellular growth profiles within HMEC-1 ECs (Fig. 5C). This observation was also confirmed using fluorescence-based proliferation assays, which revealed these isolates were indeed growing intracellularly (Fig. S5). We also demonstrated that PB and RB MRSA isolates began to replicate within LAMP-1-positive compartments within ECs and that the majority (86%) of replicating bacteria also colocalized with dextran previously accumulated within EC lysosomes (Fig. S6).

FIG 5.

PB and RB strains grow inside infected endothelial cells. (A) Graph showing the percentage of bacterial CFU that adhered (30 min postinfection, prior to gentamicin treatment) to endothelial cell monolayers normalized to the CFU present at the inoculum. (B) Graph depicting the percentage of bacterial CFU that invaded ECs (post-gentamicin treatment for 30 min) normalized to the number of adherent bacterial CFU for that same strain. (C) Intracellular growth of the indicated strains. The data are expressed as a fold change in CFU per milliliter where bacterial counts for a given strain at 6 h postinfection are divided by the number of bacteria present at 1 h postinfection for the same strain. In panels A to C, ordinary one-way ANOVA (Tukey’s posttest) was used to determine statistical significance. **, P < 0.01; ns, not significant.

Vancomycin treatment promotes nsSCV formation in PB bacteria within ECs.

The above observations that identified nsSCV formation from within the macrophage phagolysosome prompted us to examine whether related phenomena occurred within nonprofessional phagocytes such as ECs. To explore this, confluent HMEC-1 ECs were infected with LAC-USA300, its isogenic ΔfnbAB mutant, or PB versus RB clinical isolates, as above. Over the course of up to 96 h in the presence of vancomycin, intracellular CFU recovered from intact ECs increased over time (Fig. 6A). In parallel, we assessed the kinetics of nsSCV formation over time. All S. aureus strains produced nsSCVs within ECs, in agreement with previous observations (21). Importantly, substantially more nsSCVs were observed for PB than the other strains tested over the entire 96-h observation period (P < 0.001 to 0001 at 24 to 72 h growth), in agreement with the macrophage data described above (Fig. 6B).

FIG 6.

nsSCVs arise from ECs infection but remain constant during the infection. HMEC-1 cells were infected with USA300, USA300 ΔfnbAb, 3 PB, and 3 RB strains and kept under vancomycin treatment for 96h. (A) Graph showing the growth of the indicated strains monitored for 96 h in endothelial cells. (B) Graph depicting the total number of nsSCVs at 24 h, 48 h, 72 h, and 96 h postinfection (left), as well as the same data plotted as a percentage of the entire bacterial population recovered from the intracellular compartment (right). Note that since the number of nsSCVs does not change (left), while the bacterial population increases over time (A), the percentage of nsSCVs decreases over time (B, right). The plates were kept for 48 h to 72 h at 37°C for SCV quantification. In panels A and B, two-way ANOVA (Tukey’s posttest) was used to determine statistical significance; *, P < 0.01; **, P < 0.001; ***, P < 0.0001; ns, not significant. The data are the mean ± standard deviation of three independent experiments.

DISCUSSION

We investigated the intracellular fate of clinical MRSA bloodstream isolates derived from patients with either PB or RB, and several interesting observations emerged. In murine-derived macrophages, or M-CSF- and GM-CSF-derived human macrophages, as well as human-derived ECs and PB and RB strains, equivalently trafficked to LAMP-1-positive compartments where they began to replicate. This fate did not differ from other S. aureus strains such as LAC-USA300 (a cause of epidemic, community-acquired S. aureus infections in North America) (17). We observed several significant differences between PB and RB strains recovered from infected cells in response to vancomycin exposure. Indeed, PB isolates showed an increased capacity to form nsSCVs within vancomycin-exposed macrophages and ECs compared to matched RB isolates. In comparison, we could not identify such differences when in vitro exposed the strains to a combination of low pH, cationic peptide, hydrogen peroxide, and vancomycin. These data would indicate that while we cannot conclude that these stressors are not important, it is likely that they synergize in a combinatorial manner within the intracellular milieu to drive differential nsSCV formation comparing PB to RB. These findings suggest that PB strains undergo distinct adaptations to the intracellular milieu compared to RB strains, fostering their persistence. Moreover, this PB outcome is selectively induced by vancomycin treatment. Taken together, these data identify two important themes. First, antibiotic exposure of infected cells can significantly prolong the life span of the infected cell culture. Thus, without the inclusion of vancomycin in the culture medium, intracellular bacteria grow out of infected cells within ~8 to 12 h; moreover, the extracellular S. aureus rapidly grows in the tissue culture medium and rapidly kills host cells through expression of bicomponent leukocidins (33). Second, antibiotic exposure, in conjunction with microenvironmental conditions encountered within macrophages, induces these strains to adapt as SCVs that revert to parental phenotype upon subsequent culture in vitro. Of note, previous studies have observed SCV formation in response to antibiotic exposure, both in vitro and in vivo (19, 27 to 30). These changes have important implications for our understanding of this clinically significant but incompletely understood condition.

The specific intracellular microenvironmental determinants of nsSCV formation within macrophages are the subject of current investigations. Our findings are consistent with recent findings from Guérillot et al. (13) detailed above and are very reminiscent of SCV morphotyping that occurs in the absence of key growth nutrients that S. aureus requires, such as hemin, menadione, and thymidine (25, 34, 35). Of interest, Guérillot et al. (13) demonstrated that unstable chromosomal inversions occur in SCVs recovered following intracellular invasion, and such strains have phenotypes consistent with those of the PB isolates studied herein. Likewise, Vesga et al. (21) previously documented the ability of S. aureus strains to undergo SCV morphologic switching within ECs. The latter study differed from ours in several key ways, including (i) there was a very high rate of SCV formation (~10⁻³ frequency of recovered cells), (ii) a methicillin-susceptible S. aureus (MSSA; rather than MRSA) strain was utilized, (iii) bovine ECs were employed, (iv) SCVs isolated from ECs were relatively stable on multiple passages in artificial media, and (v) auxotrophy to menadione and/or hemin (nutrients critical for generation of the electron transport chain and which are in limited supply within ECs) was noted.

Previous studies have noted that PB isolates exhibited earlier-onset activation of several key global regulators (such as agr) than RB isolates, which, in turn, could affect adaptation to the phagolysosomal environment when paired with vancomycin exposure (9). Other physiological differences, such as increased expression of purine biosynthetic pathway genes (10) or the production of (p)ppGpp (11), have also been observed in PB strains, each of which could collectively affect how PB bacteria respond to the intracellular environment. In contrast, SCVs caused by chromosomal inversion exhibited significant downregulation of agr-mediated virulence or immune evasion determinants, including hla (alpha-hemolysin), cspA (pigment production), spa (protein A) and chp (chemotaxis inhibitory factor [CHIPS]) (13). Ostensibly, the differences in the ability of PB isolates to sense and respond within the context of immune and/or antibiotic stressors underpins the plausibility of intracellular formation of SCVs. Hence, such a phenotype is consistent with the hypothesis that PB isolates of MRSA exploit host cells as a persistence reservoir in which they can evade immune clearance and survive in the face of antibiotic treatment.

In vivo persistence by S. aureus capable of switching to the SCV morphotype has been documented in patients with chronic or recurrent infections such as osteomyelitis and cystic fibrosis-associated pulmonary syndromes (29, 36–38). SCVs often display reduced metabolic activity (especially in cytochrome system enzymes), low growth rates, reduced pigment production, and altered virulence factor expression compared to non-SCV phenotypes (25). Moreover, S. aureus SCV infections are often difficult to treat clinically, as the bacteria persist intracellularly and often display heightened antibiotic resistance in vivo despite susceptibility to that antibiotic in vitro (21). Recent machine learning studies based on outcomes in human and experimental models of infection predicted that MRSA growth rates and immune evasion in vivo are key determinants of persistent bacteremia outcomes (39). Supporting this prediction, parallel investigations reported that nsSCVs arise from bacteria that remain in a lag-phase state (40) and represent a large subpopulation of persister cells. Notably, these authors defined persisters as nongrowing bacteria that are able to survive high concentrations of antibiotics, to which they remain fully susceptible while growing (40), a definition more akin to tolerance than persistence. At present, it is unclear which specific intraphagosomal signals or stresses, alone or combined with antibiotic challenge, cause PB isolates to form nsSCVs at a heightened frequency uniquely within the cell types used in the current study. Presumably, uptake of membrane-impermeant bioactive molecules such as antibiotics, including vancomycin, is dependent on the pinocytic and endocytic function of host cells that can deliver material from the fluid-phase environment to phagocytosed S. aureus residing inside endolysosomal compartments (26). It is conceivable that vancomycin or other antibiotics enter or alter host cells in ways that are not yet appreciated but which promote intracellular SCV formation. Thus, underlying host and/or antibiotic mechanisms driving such phenotype switching are under investigation in our laboratories.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

Bacterial strains and plasmids are listed in Table S1 in the supplemental material. The PB versus RB clinical designations were independently made, as outlined above by the SABG investigators, based on clinical case report form data. PB isolates cause bacteremia after 7 days post therapy, and for RB isolates, bacteremia resolved immediately after the initial set of positive blood cultures. A summary table describing previously documented phenotypic and genotypic characteristics of one each of the PB and RB strains used in this study can be found in Table S2. In this study, all strains were routinely cultured at 37°C in tryptic soy broth (TSB) with shaking at 200 rpm. Strains were also cultured on TSB agar (TSA) plates or TSA supplemented with 5% (vol/vol) human blood (TSAB) in specific experiments. Strains carrying plasmids of interest were cultured in the presence of 10 μg/mL chloramphenicol. USA300-LAC was previously demonstrated to grow within cultured macrophages; its isogenic ΔgraS derivative does not grow within cultured macrophages (17, 18).

Macrophage infections.

Macrophage infections were performed as previously described (17, 18, 41). In brief, S. aureus cells grown overnight in TSB were used to infect macrophages at a multiplicity of infection (MOI) of 10 or 30. Bacteria were washed once and diluted into serum-free RPMI 1640 medium (SF-RPMI). For the synchronization of infection, tissue culture plates were centrifuged for 2 min at 1,500 rpm and then incubated for 30 min at 37°C in a humidified incubator with 5% CO₂. After 30 min, the medium was aspirated and replaced with SF-RPMI containing 100 μg/mL gentamicin for 1 h as previously described (17, 18, 26); note the MIC for gentamicin for all strains used in this study was 0.125 μg/mL. After gentamicin treatment, infected cells were rinsed with 4 mL sterile 1× phosphate-buffered saline (PBS) and replaced with RPMI-containing serum. In some instances, vancomycin (5 μg/mL; i.e., 5× MIC, as previously reported [10, 42]) and gentamicin (5 μg/mL) were used to treat infected macrophages after the initial gentamicin treatment. Once added, vancomycin and gentamicin were maintained throughout the duration of the experiment (48 h). When fluid-phase uptake blockage was required, PIK-III and Dynasore (10 and 100 μM, respectively) were used for 10 min before addition of gentamicin and maintained on the cells along with gentamicin for 1 h. Once the gentamicin protection was done, the cells were washed twice with 1 mL of PBS. To enumerate intracellular bacteria, macrophages were lysed in 0.5 mL PBS containing 0.1% (vol/vol) Triton X-100 and serially diluted and spot plated onto TSA plates or into TSAB. Fold change in CFU per milliliter for each experiment was determined by dividing bacterial counts obtained at different times postinfection by the count obtained at 1.5 h postinfection (hpi) (i.e., immediately after gentamicin treatment).

EC infections.

All EC invasion and infection experiments were performed at a multiplicity of infection (MOI) of 10. For invasion studies, confluent HMEC-1 cells in gelatin-coated 12-well tissue culture plates were used. On the day of infection, cells were washed with PBS and maintained in serum-free (SF) M199 for at least 1 h prior to infection. Bacterial strains of interest were grown overnight in TSB with appropriate antibiotics if necessary. Bacteria were then subcultured at an optical density at 600 nm (OD₆₀₀) of 0.1 and grown in TSB to an OD₆₀₀ of 0.5 (growth phase optimum for expression of fibronectin-binding proteins pivotal in EC invasion). Cells were then pelleted, washed twice with PBS, resuspended in M199, and added to wells of confluent HMEC-1 cells. Plates were pelleted at 1,000 rpm for 1 min and incubated at 37°C and 5% CO₂ for 15 min. Cells were then washed once with PBS, and then fresh SF medium was added for a further 15 min at 37°C and 5% CO₂. Cells were then treated with 150 μg/mL of gentamicin for 30 min at 37°C and 5% CO₂ (to kill all extra-EC organisms), extensively washed to remove the gentamicin, and kept in SF M199 for the desired duration of infection as required. In some instances, vancomycin (5 μg/mL; i.e., 5× MIC as previously reported [42]) was used to treat infected ECs after the gentamicin treatment. Once added, vancomycin was maintained throughout the duration of the experiment. At specific times postinfection, the medium was removed, and cells were lysed in PBS plus 0.1% (vol/vol) Triton X-100 and plated for CFU counting. Quantitative data were expressed as fold change in bacterial counts between 1.5 hpi (i.e., immediately after gentamicin treatment) and 12 hpi.

Mammalian cell culture.

RAW 264.7 macrophages from the American Type Culture Collection (ATCC) were maintained in RPMI 1640 buffered with sodium bicarbonate and 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), supplemented with 5% FBS, and passaged every 2 days. RAW 264.7 cells passaged less than 30 times were used for infection assays. HMEC-1 endothelial cells from the ATCC were maintained in M199 with sodium bicarbonate and HEPES and supplemented with 10 μg/mL heparin, 1 mM sodium pyruvate, 30 μg/mL endothelial cell growth supplement (ECGS), 200 mM l-glutamine, and 20% (vol/vol) fetal bovine serum (FBS) and passaged every 3 to 4 days in flasks coated with 5% (wt/vol) gelatin. HMEC-1 cells passaged less than 30 times were used for infection assays.

Primary human macrophages were derived from peripheral blood mononuclear cells isolated from whole blood taken from healthy volunteers in compliance with the Office of Research Ethics at The University of Western Ontario (protocol number 109059). Mononuclear cell fractions were isolated with Lympholyte-poly (Cedarlane Laboratories) according to the manufacturer’s instructions. Monocytes were purified from mononuclear fractions and were cultured in RPMI containing 10% (vol/vol) FBS supplemented with antibiotic/antimycotic and recombinant human M-CSF at 10 ng/mL. After 5 days of differentiation, adhered cells were washed with sterile PBS, and the culture medium was replaced with fresh medium containing M-CSF and excluded antibiotic/mycotic. Macrophages were differentiated through day 7 and used experimentally until day 11.

To generate primary human macrophages polarized to the M1 state, adhered monocytes were cultured as above but with 20 ng/mL recombinant human GM-CSF. At day 5, fresh RPMI containing serum, GM-CSF, 10 ng/mL recombinant human interferon gamma (IFN-γ), and 250 ng/mL lipopolysaccharide were added to the cells that were allowed to differentiate through to day 7 and used experimentally until day 11.

Microvascular endothelial cells (HMEC-1) isolated from human foreskin transfected with pSVT vector, a pBR322-based plasmid containing the coding region for simian virus 40A gene product, were obtained from ATCC (CRL-3243). HMEC-1 cells were maintained in M199 supplemented with 30 μg/mL endothelial cell growth factor (ECGS), 100 μg/mL of heparin, 1 mM sodium pyruvate, and 2 mM l-glutamine. Flasks and plates were coated with 1% gelatin for adherence. The cells were seeded in 12-well plates 24 to 48 h before infection and incubated in SF media for 1 h before bacterial infection.

Fluorescence proliferation assays.

To visualize intracellular/subcellular bacterial replication, fluorescence microscopy was employed using methodology previously described (24). When necessary, infected cells were incubated with 1 μg/mL wheat germ agglutinin and tetramethylrhodamine conjugate (TMR-WGA) for 2 min prior to 20 min of fixation in 4% (vol/vol) paraformaldehyde (PFA) at room temperature.

LAMP-1 immunostaining.

Detection of endogenous LAMP-1 was performed as previously described (17). Rat anti-mouse LAMP-1 antibodies (clone 1D4B) and mouse anti-human LAMP-1 antibodies (clone H4A3) were purchased as supernatants from the Developmental Studies Hybridoma Bank (DSHB).

Dextran pulse-chase.

Cells adherent to 18-mm glass coverslips were cultured for approximately 16 h in the appropriate medium containing tetramethyl rhodamine isocyanate (TRITC)-dextran (100 μg/mL) as previously described (17). Infections were performed as described above, and, at specific times postinfection, infected cells were analyzed by live-cell fluorescence imaging.

Viability assay with PI.

To monitor membrane integrity and cell viability, the vital dye propidium iodide was added to the medium at a final concentration of 1 μg/mL. Cells were imaged live except for the positive control where cells were fixed for 20 min in 4% (vol/vol) PFA at room temperature and permeabilized with 0.1% Triton-x 100 in PBS prior to addition of medium containing PI. For each experiment, at least 1,000 cells were counted per strain/condition, and averages were used to calculate the percentages of PI-positive cells.

Fluorescence and light microscopy.

Wide-field fluorescence and differential interference contrast (DIC) microscopy were performed on a Leica DMI6000B inverted microscope equipped with 340 (numerical aperture [NA], 1.3), 363 (NA 1.4), and 3,100 (NA 1.4) oil immersion PL Apo objectives, a Leica 100 W Hg high-pressure light source, and the Hamamatsu Orca Flash 4.0 and Photometrics Evolve 512 D electron-multiplying charge-coupled-device (EM-CCD) cameras.

Checkerboard assays and in vitro nsSCV formation.

Stock solutions of each agent were prepared immediately prior to testing. A total of 100 μL of culture was distributed into each well of microtiter plates. The first substance of the combination was serially diluted along the ordinate, while the second drug was diluted along the abscissa. An inoculum equivalent to an OD₆₀₀ of 0.02 from an overnight culture was prepared from each strain to be tested. Each microtiter well was then inoculated with 100 μL of the bacterial inoculum for a final inoculum of an OD₆₀₀ of 0.01, and the plates were incubated at 37°C for 24 h with shaking. The results from the checkerboard assays are shown in Tables S2 (RPMI) and S3 (TSB). From these results, concentrations of H₂O₂, polymyxin B, and sub-MIC vancomycin were used in the in vitro nsSCV formation assay.

For nsSCV formation in vitro, 14-mL conical tubes containing 3 mL of the respective medium (RPMI or TSB) containing the concentrations summarized in Tables S2 and S3 were inoculated with 10 μL of the bacterial inoculum for a final inoculum of OD₆₀₀ of 0.01. The tubes were then incubated at 37°C for 24 h with shaking. At the 24-h time point, an aliquot of each culture was serial diluted for quantification of nsSCVs, and the OD₆₀₀ was measured to be normalized and reinoculated in fresh culture tubes containing the same concentrations of stressors; these tubes were then incubated for another 24 h at 37°C with shaking. At the end of the 48 h, the cultures were again serial diluted and spread plated for quantification of nsSCVs. For each experiment, fresh solutions were prepared immediately prior to testing, and the percentage of nsSCV formation was plotted as a percentage of nsSCVs over the total number of CFU enumerated for each strain.

Statistical analyses and software.

All statistical analyses were performed using GraphPad Prism software. When appropriate, one-way analysis of variance (ANOVA) and unpaired t tests were performed to determine statistical significance, with a P value cutoff of ≤0.05 to establish significance. Where appropriate, Bonferroni’s and Tukey’s post hoc tests were performed to directly compare experimental means.

Ethics statement.

Blood was obtained, with permission, only from healthy adult volunteers, in compliance with protocol 109059 approved by the Office of Research Ethics at the University of Western Ontario.