6–7 weeks old female C57BL/6J mice for in vivo experiments were obtained from Charles River Laboratories (Sulzfeld, Germany). Gp91 phox (nox2^−/−) (45) and NOX4 knockout (nox4^−/−) mice (46) were kindly provided by Ralf Brandes (Goethe University Frankfurt). NOX1 knockout mice (47) were kindly provided by Karl-Heinz Krause (University of Geneva) and nmf333 mice harboring the Y121H (p22^Y121H) point mutation in the gene encoding p22phox (48), were obtained from J. Woo (Stanford University Medical Center, Stanford).

All mice were backcrossed at least 10 times to the C57BL/6J background. Mice were kept under specific pathogen-free conditions at the animal facilities of the Medical Center of the University of Cologne. All animal experiments have been carried out with local ethical committee approval (AZ 84-02.04.2014.A013) and adhering to the guidelines of German jurisdiction and the guidelines for the welfare and use of animals in research. All efforts were made to minimize suffering of the animals.

Staphylococcus aureus strain MW2, a community acquired MRSA strain also known as USA 400, was obtained from the Network on Antimicrobial Resistance in Staphylococcus aureus (www.narsa.net). S. aureus MW2 constitutively expressing GFP under control of the spa-derived constitutive promoter was constructed by transformation of MW2 with plasmid pCN-F7-GFP (49). Wildtype MW2 was used for ROS measurements, all other experiments were conducted with MW2-GFP. MW2-GFP was selected on 5 μg/ml erythromycin on agar plates and during overnight (ON) cultures. For experiments S. aureus were inoculated 1:100 from ON culture into fresh Luria- Bertani (LB) broth and grown at 37°C to an OD₆₀₀ of 0.3. Bacteria were harvested, washed, and the concentration was adjusted to 1 × 10⁹ CFU/ml in PBS.

For in vitro differentiation of bone marrow cells into bone marrow-derived macrophages (BMDM), bone marrow was prepared from the tibias and femurs of mice from the C57/BL6 J Background since this is the background of used KO mouse lines. The erythrocytes were lysed with Tris-buffered ammonium chloride (8.3% NH₄Cl, 0.1 M Tris). Bone marrow cells were cultured in VLE RPMI 1640 medium (Biochrom), supplemented with 10% FCS, penicillin (100 U/ml) and streptomycin (100 μg/ml), 2 mM HEPES, 200 nM sodium pyruvate and 10 ng/ml recombinant M-CSF (Peprotech) to differentiate bone marrow cells into BMDM. BMDM were used for experiments at day 8 of in vitro differentiation unless specified otherwise. More than 90% of these cells were F4/80⁺/Cd11⁺ BMDM as determined by flow cytometry. Antibiotics were removed 16 h prior to infection of BMDM (50).

For ROS measurements BMDM were seeded in a density of 1 × 10⁵ cells /well in sterile white 96 well LumiNunc plates (Thermo Scientific) in antibiotic free medium 16 h before measurement. After washing cells twice with cold HBSS containing magnesium sulfate (200 mg/L) and calciumchloride (185.4 mg/L) (Sigma), either viable non-opsonized or opsonized S. aureus MW2 [5% normal mouse serum (NMS)] were added to respective wells using MOI50 or MOI10 as indicated. Cells treated with 5% normal mouse serum (NMS; Innovative research) in HBSS served as non-infected control. Upon addition of bacteria or mouse serum, infection was synchronized by centrifugation at 840 g at 4°C in a swing bucket rotor for 5 min. Subsequently supernatant was aspirated and fresh HBSS was added to the cells. 2x Luminol/HRP mix in HBSS was added to each well. Finally, 10 μM PMA (Sigma) as positive control or 100 μg/ml Pam2CSK4 (Invivogen) as control for TLR2 dependent ROS production in HBSS was added immediately before measurement. ROS production was measured in a plate reader preheated to 37°C in 60 s intervals over 60 min (Tristar, Berthold Instruments). For evaluation first cell-free values were subtracted from each sample, subsequently values of non-infected or untreated samples were subtracted. Plotted was the corrected RLU per minute in mean and SD of triplicates.

To assess uptake and clearing of bacteria in BMDM, cells were harvested and adjusted to 5 × 10⁶ cells/ml. Assays were performed in triplicates using a total of 2.5 × 10⁶ cells/ml. BMDM were infected with S. aureus MW2 GFP (MOI15) opsonized with 5% NMS in suspension. After addition of bacteria, samples were mixed and infection was synchronized by centrifugation at 840 g at 4°C for 2 min, subsequently incubated for 5 min at 37°C rotating end over end, and thereafter the pellets were resuspended. This scheme was performed thrice. Remaining extracellular S. aureus were removed with Lysostaphin (Sigma; 2.5 mg/ml) in a final concentration of 0.2 mg/ml on ice for 20 min, followed by washing with PBS twice. The pellets were resuspended in 1 ml of culture medium without antibiotics and 50 μl of samples were analyzed directly (t0) and at additional time points (30 min, 1 h, 6 h and 24 h p. i.) for FACS analysis. Bacterial clearing was determined by calculating the amount of GFP positive macrophages (percentage within M1; see supp. Supplementary Figure 1) of respective time points in percent of t0. Samples were incubated rotating end over end at 37°C. After the 6 h time point Gentamycin (Sigma; 50 μg/ml) was added to the samples, to enable assessment of viability 24 h post infection. BMDM viability was assessed by trypan blue exclusion using the Countess (Life Technologies) and was calculated as percent of viable non-infected BMDM.

BMDM on day 8 of differentiation were infected with S. aureus-GFP using MOI30, which results in close to 100% loading of BMDM (Supplementary Figure 1B). After Lysostaphin treatment, BMDM were washed and 2.5 × 10⁶ cells in 300 μl were administered intravenously into the tail vein of mice. 5 h post infection the amount of S. aureus present in infected cells was determined after lysis of BMDM in H₂O at pH11 and subsequent plating in 10x serial dilutions on agar plates. Control mice were injected with 4 × 10⁶ plain S. aureus corresponding to the CFUs retrieved from S. aureus infected BMDM.

For determination of the bacterial load, mice were euthanized and sacrificed by cervical dislocation 24 h, 72 h and on day 7 post injection. Organs were homogenized in 0.1% Triton X-100 in gentleMACS M tubes using the predefined program for protein isolation using the gentleMACS (Miltenyi Biotec) and plated in 10-fold serial dilutions on Mueller-Hinton agar plates using the Eddy Jet spiral plater (IUL Instruments) in mode log50. After overnight incubation of the plates at 37°C, the CFUs were counted using the Countermat flash plate reader (IUL Instruments).

In vitro results and log transformed CFU values of in vivo experiments were analyzed by unpaired two-sided students t-test or 1-way ANOVA with Bonferroni post-test between selected pairs (viability data in Figures 2E–G). Contingency table analysis was performed using fisher' exact test. Statistical analysis and plotting of data was performed using GraphPad Prism 5.01.

Macrophages do not always succeed in S. aureus killing, thereby providing a niche for persistence, which fosters continued infection (36). It has been speculated for a long time that professional phagocytes might serve as Trojan horses for S. aureus and lead to the frequently observed dissemination from the primary focus of infection (15). Whereas the role of neutrophils has been evaluated in greater detail, macrophages, although being a source of bacterial persistence, have been far less studied. A prerequisite for bacterial persistence is a perfect symbiosis of intracellular S. aureus and host cells, which can only be achieved when the bacteriocidal activities of macrophages are kept at bay. One of the most important bacteriocidal mechanisms of macrophages, the oxidative burst is generated by the NOX family of NADPH oxidases. Next to NOX2, at least two additional NOX isoforms, namely NOX1 and NOX4, are expressed by macrophages and might be involved in ROS generation. Indeed, wildtype BMDM strongly responded with ROS production to opsonized S. aureus (Figure 1A, left). The kinetics of ROS production was comparable to that of BMDM stimulated pharmacologically with 10 μM PMA used as positive control (Figure 1A, right). Treatment with the NOX inhibitor DPI led to a significant inhibition of ROS production (Figure 1A). We next scrutinized the relative contribution of NOX-isoenzymes 1, 2 and 4 to the oxidative burst of macrophages in response to S. aureus. To this end, BMDM with genetically defined deficiencies for individual NOX isoforms were employed, including nox1^−/−, nox2^−/− and nox4^−/− as well as p22^Y121H, a non-functional mutant of the NADPH oxidase subunit p22^phox that is common to isoforms NOX1,−2,−3, and−4. The oxidative burst induced by S. aureus in BMDM from nox1^−/− and nox4^−/− mice was MOI-dependent and comparable to that of wildtype BMDM (Figure 1B). By contrast, macrophages isolated from nox2^−/− or p22^Y121H mice showed completely abolished production of S. aureus-induced ROS. These findings demonstrate that mainly the NADPH oxidase NOX2 is responsible for ROS production by BMDM upon S. aureus infection and raised the question about the control of NOX2 activation.

NOX2 is activated by opsonophagocytic receptors like FCγR and macrophage-1 antigen (Mac-1) (36, 51). Indeed, opsonized S. aureus led to strong ROS production (Figures 1C,D). Infection with non-opsonized S. aureus did not stimulate any measurable production of ROS at this time after infection (Figure 1D). TLR2-dependent activation of JNK results in inhibition of ROS production upon S. aureus infection and increased bacterial survival (52). Pam₂CSK₄, a synthetic TLR2 ligand, did not induce TLR2-dependent ROS production within the first 30 min (Figure 1C). Furthermore, opsonized S. aureus induced comparable production of ROS in either TLR2-deficient or -proficient BMDM (Figures 1C–E). Thus, the induction of NOX2-mediated ROS production by phagocytic receptors seems to overrule any possibly antagonistic action of TLR2.

To investigate the impact of the reduced oxidative burst in NOX2-deficient BMDM on the survival of both, intracellular S. aureus and host cell, we determined the kinetics of relative numbers of S. aureus-positive BMDM and the viability of host cells. ROS inhibition in BMDM by treatment with DPI significantly impaired the clearance of intracellular S. aureus over the first 24 h (Figure 2A). At 24 h post infection about 25% of DPI-treated BMDM were positive for GFP-expressing S. aureus, whereas only 10% of untreated BMDM carried S. aureus. As expected, NOX2-deficient BMDM, like DPI-treated BMDM, showed a markedly impaired clearance of intracellular S. aureus (Figure 2B). In contrast, nox1^−/− and nox4^−/− BMDM cleared S. aureus like wildtype BMDM (Figures 2C,D), which corresponds to their normal ability to elicit an oxidative burst. Thus, ROS production by NOX2 in BMDM is a prerequisite for the ability to clear phagocytosed S. aureus.

The question arose whether the reduced relative numbers of S. aureus-positive BMDM result from successful ROS-dependent killing of intracellular S. aureus or rather are secondary to killing of the BMDM by S. aureus. Therefore, we investigated the viability of BMDM. DPI-treated as well as nox2^−/− BMDM had a substantial advantage in survival upon infection, despite their higher bacterial burden (Figures 2A–C; right). In fact, the viability of non-infected wildtype BMDM was less than 60% indicating that S. aureus infection led to cell death in 40% of infected BMDM and suggesting that S. aureus may have escaped into the culture supernatant and subsequently killed by gentamycin present in the culture medium. This scenario is consistent with a continuous cycle of phagocytosis, intracellular S. aureus replication, host cell death, bacterial release and re-uptake by macrophages (which may only occur in the absence of antibiotics) as recently proposed by Jubrail and colleagues (37, 53). In contrast, S. aureus induced cell death in only 15% of NOX2-deficient or DPI-treated BMDM (Figures 2A,B; right). Together, these data suggest that the oxidative burst not only reduces the number of BMDM carrying phagocytosed S. aureus but also contributes to death of infected macrophages. Furthermore, a defective oxidative burst likely renders BMDM an intracellular niche for S. aureus survival and replication.

The survival of S. aureus inside BMDM and in particular the prolonged survival of nox2^−/− macrophages observed in our experiments in vitro raised the question about possible functional consequences in vivo, such as enhanced pathogenicity and dissemination of S. aureus in mice. Therefore, we next investigated the course of infection in mice challenged either with S. aureus internalized by macrophages or with the corresponding dose of plain S. aureus injected as a bacterial suspension. Wildtype or nox2^−/− BMDM were infected in vitro with S. aureus and subsequently administered i.v. into wildtype C57BL/6J mice. The dose of plain S. aureus administered as a bacterial suspension corresponded to the number of live S. aureus recovered from infected nox2^−/− BMDM (4 × 10⁶ CFU). Mice challenged with BMDM loaded with intracellular S. aureus showed larger CFU counts in kidneys and liver, and less prominently in spleen, on day 7 after infection than mice challenged with plain S. aureus (Figures 3A–G). Notably, when nox2^−/− BMDM were used as vehicles, they produced by trend the highest bacterial load in kidneys and liver. Intravenous injection of plain S. aureus also induced bacterial dissemination in mice. Animals infected by plain bacteria more efficiently cleared S. aureus from kidney and liver than animals injected with BMDM-associated S. aureus (Figures 3B,C,G–I).

Macroscopical examination of explanted livers revealed that all livers from mice injected with infected nox2^−/− BMDM contained one or more subcapsular abscesses of at least 1 mm in diameter (Figure 3I). Therefore, we assessed the frequency of subcapsular liver abscesses at 72 h post infection in the three experimental groups. The incidence of liver abscesses was significantly increased in the group that received nox2^−/− BMDM carrying S. aureus as compared to the group that received wildtype BMDM carrying S. aureus (Figure 3H). No subcapsular abscesses were detected in livers of mice that received plain S. aureus. Histological analysis of HE-stained sections revealed relatively small abscesses (about 1.5 × 10⁴ μm²) for mice that received wildtype BMDM carrying S. aureus (Figure 3I). By contrast, liver sections of mice that received nox2^−/− BMDM carrying S. aureus revealed large abscesses with extensive necrosis and excessive inflammatory infiltrates. In one field of view, in total eight abscesses could be detected, which had an average area of about 3 × 10⁵ μm². As expected, in the livers from mice receiving plain S. aureus, no histopathological signs of infection were detected, consistent with the relatively low CFU count at 72 h post infection. These data suggest that intracellular S. aureus internalized by macrophages induce a disseminated and long-lasting infection in mice, especially when compared to plain bacteria. Moreover, the aggravation of the infection caused by nox2^−/− BMDM-associated S. aureus underscores the pathogenic potential of a weakened but surviving host cell to provide a niche for S. aureus survival and replication.

S. aureus infections are treated with antibiotics like ß-lactams or vancomycin that are extracellularly active, sometimes in combination with intracellularly active antibiotics like clindamycin or rifampin. Our results suggest that BMDM carrying viable S. aureus either die and release S. aureus into the extracellular milieu, where they are exposed to antibiotics. BMDM carrying viable S. aureus are yet also able to survive and initiate and maintain an infection with enhanced severity in vivo. We therefore hypothesized that a combined treatment with an extracellularly active antibiotic together with an intracellularly active antibiotic should be superior to treatment with an extracellularly active antibiotic alone. To test this hypothesis, mice were challenged with S. aureus-containing wildtype BMDM (Figure 4A), S. aureus-containing nox2^−/− BMDM (Figure 4B) or with plain S. aureus (Figure 4C). After 1 h of infection, mice subcutaneously received the extracellularly active antibiotic vancomycin alone, or in combination with the intracellularly active rifampicin. Rifampicin was chosen because of its excellent ability to eradicate intracellular S. aureus in host cells (43). Injection of antibiotics was repeated at 6, 24 and 30 h post infection to account for the limited half-life of vancomycin. At 48 h p.i., livers were explanted and analyzed for bacterial loads. First, we observed that wildtype and nox2^−/− BMDM carrying S. aureus produced 2–3 logs greater numbers of S. aureus (CFU/g) than plain S. aureus (Figure 4, compare A, B, and C), which is probably the result of almost instant exposure of extracellular S. aureus to the bacteriocidal antibiotic vancomycin and to the innate host defense. In the two BMDM groups, vancomycin alone reduced the bacterial load in livers in a statistically significant manner (Figures 4A,B) suggesting that S. aureus must escape BMDM over time and are subsequently eliminated by exposure to vancomycin. When mice were infected with plain S. aureus, vancomycin reduced the bacterial burden to a lesser, although still significant, extent leaving a large residual population of viable S. aureus particularly in the liver (Figure 4C). This finding suggests that a large proportion of plain S. aureus are protected from vancomycin in vivo, probably by internalization into professional phagocytes. Indeed, the combined treatment of mice with vancomycin and rifampicin resulted in complete elimination of S. aureus (Figures 4A–C), which underscores a potential benefit of intracellularly active antibiotics in the treatment of S. aureus infections.