Polymorphonuclear leukocytes mediate Staphylococcus aureus Panton-Valentine leukocidin-induced lung inflammation and injury

Abstract

Community-associated methicillin-resistant Staphylococcus aureus (CA-MRSA) is epidemic in the United States, even rivaling HIV/AIDS in its public health impact. The pandemic clone USA300, like other CA-MRSA strains, expresses Panton-Valentine leukocidin (PVL), a pore-forming toxin that targets polymorphonuclear leukocytes (PMNs). PVL is thought to play a key role in the pathogenesis of necrotizing pneumonia, but data from rodent infection models are inconclusive. Rodent PMNs are less susceptible than human PMNs to PVL-induced cytolysis, whereas rabbit PMNs, like those of humans, are highly susceptible to PVL-induced cytolysis. This difference in target cell susceptibility could affect results of experimental models. Therefore, we developed a rabbit model of necrotizing pneumonia to compare the virulence of a USA300 wild-type strain with that of isogenic PVL-deletion mutant and -complemented strains. PVL enhanced the capacity of USA300 to cause severe lung necrosis, pulmonary edema, alveolar hemorrhage, hemoptysis, and death, hallmark clinical features of fatal human necrotizing pneumonia. Purified PVL instilled directly into the lung caused lung inflammation and injury by recruiting and lysing PMNs, which damage the lung by releasing cytotoxic granule contents. These findings provide insights into the mechanism of PVL-induced lung injury and inflammation and demonstrate the utility of the rabbit for studying PVL-mediated pathogenesis.

Community-associated methicillin-resistant Staphylococcus aureus (CA-MRSA), especially the pandemic USA300 clone, has been associated with severe infections and high mortality rates, particularly in patients with fulminant necrotizing pneumonia (1–5). USA300, like other CA-MRSA strains, produces Panton-Valentine leukocidin (PVL), a member of the family of bicomponent β-channel pore-forming toxins. PVL targets phagocytic leukocytes, especially polymorphonuclear leukocytes (PMNs), the first line of defense against S. aureus infections (6–9). Compelling epidemiological data point to PVL as an important virulence factor in S. aureus necrotizing infections, but experimental data in rodent models are inconclusive (10–18).

In contrast to mouse and nonhuman primate PMNs, which are relatively resistant to PVL in vitro, human and rabbit PMNs are susceptible to its cytotoxic effects (6–9). This difference in target cell susceptibility could affect results obtained with different animal species (8, 9). The susceptibility of host PMNs to PVL may be an important consideration when selecting an appropriate animal species to model its role in pathogenesis. Therefore, if PMNs are the relevant target in vivo, the rabbit might be the ideal model to test whether PVL contributes to the pathogenesis of necrotizing pneumonia.

Despite the strong clinical association of PVL with severe necrotizing infections, mechanisms by which this toxin induces tissue necrosis are not known. One possibility is that PVL-induced PMN lysis results in impaired host defenses, interfering with clearance of organisms from the site of infection and allowing unchecked bacterial growth and expression of other tissue-damaging exotoxins (e.g., α-hemolysin). Another possibility is that PVL itself directly or indirectly causes tissue injury. In vitro, PVL activates PMNs to release potent proinflammatory mediators (IL-8 and leukotriene-B4) and granule enzymes (β-glucuronidase, hydrolase, and lysozyme) and to produce reactive oxygen metabolites that may cause tissue injury (7, 19–22). Understanding the mechanisms by which PVL promotes pathogenesis in animal models should clarify its role in human infections and may provide a basis for development of new therapeutic interventions. To these ends, we developed a rabbit model of necrotizing pneumonia to study specific mechanisms of PVL-induced acute lung injury.

Results and Discussion

PVL Is a Potent Cytolytic Toxin for Human and Rabbit PMNs.

As a step toward determining whether the rabbit would be a suitable animal model in which to test a potential role of PVL in USA300 necrotizing pneumonia, we used previously described methods (12, 23–25) to evaluate the ability of purified PVL to form membrane pores or lyse human and rabbit PMNs. Rabbit and human PMNs were highly susceptible to purified PVL (LukS-PV + LukF-PV), as measured by membrane pore formation and cytolysis in vitro (Fig. 1 A and B). PVL from culture supernatants of PVL-expressing SF8300, a USA300 clinical strain, but not from an isogenic lukS/F-PV–negative mutant (Δpvl) strain, also caused pores in plasma membranes of human and rabbit PMNs at low concentration (Fig. 1C).

Fig. 1.

Pore-forming and cytolytic activities of PVL toward human and rabbit PMNs. (A) PMN pore formation [percent ethidium bromide (EtBr)-positive cells] and (B) PMN lysis [percent lactate dehydrogenase (LDH) release] assays using LukS-PV and LukF-PV purified from culture supernatants of a PVL-producing USA300 strain as described in SI Text. Results in A and B are mean of independent experiments with six human and six rabbit blood donors. (C) Pore formation caused by 1:2,000 dilutions of 8-h CCY culture supernatants of USA300 wild-type strain, Δpvl mutant, and compΔpvl strain as indicated. Results in C are mean of independent experiments with 10 human and 4 rabbit blood donors. Statistically significant differences are indicated by asterisk (P < 0.05); all other comparisons are nonsignificant.

PVL Is a Virulence Determinant in a Rabbit Pneumonia Model.

In a dose-escalation study, endotracheal instillation of 1 × 10⁸ and 5 × 10⁸ cfu of SF8300 in outbred rabbits resulted in no mortality, whereas treatment with 1 × 10⁹, 5 × 10⁹, and 10 × 10⁹ cfu caused 40–100% mortality. The higher inoculum, which on a per weight basis corresponds to 10-fold fewer bacteria than used in mouse pneumonia models (11, 13), was selected to compare the virulence of SF8300 wild-type and Δpvl strains.

The wild-type strain caused greater, dose-dependent mortality (Fig. 2A) and more frequent hemoptysis (35/45 vs. 18/43, P < 0.001) than the Δpvl strain. With both the wild-type strain and Δpvl mutants, mortality was associated with high titers of bacteria in the lungs (Fig. 2B). Survival was associated with clearance of organisms, evidenced by 10⁴-fold cfu fewer in the lungs of rabbits killed at 48 h as compared with those that died from infection (open vs. filled symbols). The significantly higher numbers of bacteria recovered from the lungs of rabbits infected with the wild-type strain compared with the Δpvl strain (Fig. 2B) could reflect the rapidly lethal course of infection caused by the wild-type strain that prevented host clearance of these bacteria, a hypothesis explored more fully in the time-course experiments described later.

Fig. 2.

PVL contributes to virulence of USA300 in a rabbit model of necrotizing pneumonia. (A) Kaplan–Meier survival curves for comparison of mortality in rabbits infected via endotracheal instillation with increasing number of cfu of a SF8300 wild-type (wt) strain and its isogenic Δpvl mutant strain (n = 88 total). Fifteen rabbits were randomized to receive one of the six inocula, which were blinded. Two rabbits had anesthesia-related deaths after randomization but before bacterial inoculation; these rabbits were excluded from subsequently analysis. Log-rank test was used to compute P values for comparison of wild-type and Δpvl mutant strains at the same inocula. (B) Comparison of bacterial densities and (C) LW/BW ratio. (D) Kaplan–Meier survival curves for comparison of mortality in rabbits randomized for infection with blinded inocula containing wild-type, Δpvl mutant, or compΔpvl strains (n = 45 total). (E) Corresponding data on bacterial densities in the lung, (F) LW/BW ratio, (G) IL-8 per lung, and (H) MCP-1 per lung. In B, C, and E–H, filled symbols represent data from dead rabbits, and open symbols represent data from surviving rabbits that were killed 48 h postinfection. Unpaired Student's t test was used to compute two-sided P values for comparison of wild-type and Δpvl. (I) LukS-PV (in ng) per gram of lung tissue, as measured by ELISA, after each extraction (open symbols, dashed lines) and final cumulative LukS-PV (in μg) per lung (filled symbols, solid lines); data represent mean of three necrotic lungs harvested from rabbits 12–39 h postinfection with either the wild-type strain (blue) or compΔpvl strain (green). LukS-PV was not detected in lungs from rabbits infected with Δpvl mutant.

Lung histopathology showed that wild-type strains (12 lungs scored) and Δpvl strains (11 lungs scored) induced lung injury (Table S1), a finding consistent with the capacity of PVL-negative S. aureus to elaborate other virulence determinants, such as staphylococcal protein A and α-hemolysin, which also contribute to acute lung injury (13, 26, 27). However, compared with the Δpvl strain, the wild-type strain caused significantly more extensive necrosis with disruption of pulmonary architecture, hemorrhagic infiltration, exudate/fibrin deposition, alveolar and interstitial edema, and PMN infiltration and destruction (Table S1). Pulmonary edema was more severe in rabbits infected with the wild-type strain than in animals infected with the Δpvl strain, as evidenced by significantly greater lung wet weight to body weight (LW/BW) ratios (Fig. 2C). Severe pulmonary edema (i.e., LW/BW ratio >12), which was accompanied by profound hypoxemia [mean arterial partial pressure of oxygen (Po₂), 33 on room air], respiratory failure (mean arterial Pco₂, 78), and both respiratory and metabolic acidosis (mean arterial pH 7.14 and lactate 15.0 mmol/L) was observed more frequently in rabbits infected with the wild-type strain than in those infected with the Δpvl strain (Fig. 2C).

Restoration of PVL Production in the Δpvl Strain Restored Bacterial Virulence.

The Δpvl strain was significantly less virulent than either the wild-type or PVL-complemented Δpvl (compΔpvl) strains, as determined by animal mortality, cfu, and LW/BW ratios (Fig. 2 D–F). Levels of IL-8, but not monocyte chemotactic protein 1 (MCP-1), were significantly higher in animals infected with wild-type and compΔpvl strains than in animals infected with the Δpvl strain (Fig. 2 G and H). PVL-induced production of IL-8 in the lung reported here is consistent with previous observations that PVL stimulates PMNs to produce IL-8 in vitro (21, 28).

PVL Is Produced in Toxic Amounts in the Lung.

PVL expression in the rabbit lung was measured by ELISA as previously described (29, 30). LukS-PV was produced in 12–18 μg per lung in rabbits infected with 6 × 10⁹ cfu of either wild-type or compΔpvl strains; no LukS-PV was detected in the lungs of rabbits infected with the Δpvl strain (Fig. 2I). PVL was not detected in the lungs of rabbits infected with 1.5 × 10⁹ cfu of the wild-type strain, indicating that a high bacterial burden in the rabbit lung is required for PVL to be produced in amounts that are comparable to those achieved during human skin infections (31). It is of interest that inocula for the wild-type and compΔpvl strains contained 0.1–0.2 μg of LukS-PV. Therefore, de novo PVL production during infection, not preformed toxin, was responsible for the increased virulence of wild-type and compΔpvl strains.

Time-Course of PVL-Induced Acute Lung Injury.

Although PVL is strongly associated with necrotizing pneumonia and high mortality (Fig. 2), it is unclear whether PVL promotes bacterial survival, which in turn is responsible for the pathology, or whether PVL directly induces acute lung injury and lung inflammation. To address this issue, we conducted a time-course experiment in which rabbits were killed at 3, 6, and 9 h after endotracheal instillation with wild-type strain, Δpvl strain, or sterile vehicle control. Infection with the wild-type and Δpvl strains, but not instillation of sterile vehicle control, resulted in severe leukopenia by 9 h postinfection, evidenced by a decline in peripheral white-cell count to 8.6% and 16.3% of baseline values, respectively (P = 0.311).

Fig. 3A shows there was no difference in bacterial counts at 3, 6, or 9 h postinfection for wild-type or Δpvl-infected lungs, indicating that PVL-induced acute lung injury is independent of bacterial survival/replication in the lung. In contrast, the LW/BW ratio increased in a time-dependent manner with the wild-type strain but not with the Δpvl strain (Fig. 3B). The gross appearance of wild-type–infected lungs was markedly different from that of Δpvl-infected lungs with extensive areas of necrosis and frothy edema brimming from the bronchi of wild-type–infected lung (Fig. 3H). The protein content of bronchoalveolar lavage (BAL) fluid of lungs of rabbits infected with the wild-type strain, but not those infected with Δpvl strain, increased over time (Fig. 3C), indicative of an influx of plasma proteins into the alveolar space caused by damage to the alveolar–endothelial barrier. Significantly greater levels of IL-8 and MCP-1 increased over time in lung and plasma of wild-type–infected rabbits compared with Δpvl-infected rabbits (Fig. 3 D–G), and this finding correlated with significantly more extensive PMN infiltration and destruction provoked by the wild-type strain as compared with the Δpvl strain (Fig. 3 I–K and Table S1).

Fig. 3.

Time-course of PVL-induced acute lung injury after endotracheal instillation with SF8300 wild-type (wt) strain, isogenic Δpvl mutant, or vehicle control. Twenty-four rabbits each were randomized to receive either the wild-type or Δpvl mutant strain, which were blinded, and then eight rabbits chosen at random from each group were killed at 3, 6, or 9 h postinfection (n = 57 total). Three vehicle control-instilled rabbits were included for each time point. (A) Bacterial cfu per lung. (B) LW/BW ratio. (C) total protein concentration in BAL fluid. (D) IL-8 (in ng) per lung. (E) IL-8 (pg/mL) in plasma. (F) MCP-1 (in ng) per lung. (G) MCP-1 (ng/mL) in plasma. Unpaired Student's t test was used to compute two-sided P values for between-group comparisons; statistical significant differences compared with wild type are indicated by an asterisk (P < 0.05); all other comparisons were nonsignificant. Photograph depicts lungs (H) and corresponding H&E-stained sections (magnification 200×) harvested from rabbits at 9 h after endotracheal instillation of wild-type strain (I), isogenic Δpvl mutant (J), or vehicle control (K).

PVL Causes Lung Inflammation and Injury by PMN-Dependent Mechanisms.

We further hypothesized that lung injury is the result of recruitment and subsequent lysis of PMNs, which damage the lung by releasing the contents of cytotoxic granules and/or reactive oxygen metabolites. To test this hypothesis, we instilled 12 μg each of LukS-PV and LukF-PV (i.e., 0.34 nmol of LukS-PV and 0.32 nmol of LukF-PV), amounts comparable to that produced during USA300 infection of rabbit lungs (Fig. 2I), endotracheally into lungs of normal rabbits and neutropenic (vinblastine-treated) rabbits (32). Because both LukS-PV and LukF-PV are required for biological activity (6, 7), each toxin subunit alone was used as a negative control. Purified LukS-PV+LukF-PV induced significant levels of IL-8 and MCP-1 in the lungs compared with no cytokine induction when either toxin subunit was administered alone (Fig. 4 C and D). Purified LukS-PV+LukF-PV, but neither subunit alone, was sufficient to induce acute lung injury, which was accompanied by a neutrophilic infiltrate, necrosis, diffuse alveolar hemorrhage, and pulmonary edema (i.e., an increase in the LW/BW ratio) in normal rabbits (Fig. 4 A–L). Purified LukS-PV+LukF-PV administered to neutropenic rabbits did not cause an increase in IL-8 and MCP-1 levels (Fig. 4 C and D), nor was there acute lung injury (Fig. 4 H and L). Immunohistochemistry for IL-8 showed staining of PMNs and to a lesser extent alveolar macrophages, but not epithelial or endothelial cells, in lungs of normal rabbits instilled with LukS-PV+LukF-PV (Fig. 4K). Lungs of normal rabbits instilled with either toxin subunit alone (Fig. 4 I and J) did not show staining for IL-8, whereas lungs of neutropenic rabbits instilled with both toxin subunits showed weak staining for IL-8 in only alveolar macrophages (Fig. 4L). This finding is consistent with the reported specificity of PVL for myeloid cells and not other cell types (6) and also indicates that PMNs can amplify or perpetuate the acute inflammatory response by IL-8–mediated recruitment of additional PMNs into the inflamed lung. Taken together, the data indicate that PMNs are critical for the development of PVL-induced lung inflammation and injury.

Fig. 4.

PVL-induced acute lung injury is mediated by PMNs. Normal rabbits were treated with endotracheal instillation of 12 μg of LukS-PV alone (n = 3), 12 μg of LukF-PV alone (n = 3), or 12 μg each of LukS-PV and LukF-PV (n = 9); neutropenic rabbits were treated with 12 μg each of LukS-PV and LukF-PV (n = 6). All rabbits were killed 3 h after instillation to determine (A) LW/BW ratio, (B) total protein concentration in BAL fluid, (C) IL-8 (in ng) per lung, and (D) MCP-1 (in ng) per lung. Unpaired Student's t test was used to compute two-sided P values for between-group comparisons; differences that are statistically significant compared with wild type are indicated by an asterisk (P < 0.05); all other comparisons were nonsignificant. H&E-stained sections (E–H; magnification 200×) or immunohistochemistry for IL-8 (I–L; magnification 400×) of lungs from normal rabbits instilled with LukS-PV only (E and I), LukF-PV only (F and J), LukS-PV and LukF-PV (G and K), or neutropenic rabbits instilled with LukS-PV and LukF-PV (H and L). Results in K showed deposition of brown reaction product (IL-8) in PMNs and to a lesser extent in alveolar macrophages.

Concluding Comments.

The striking epidemiological association of PVL and severe necrotizing staphylococcal infections (1–5), including pneumonia, has prompted numerous studies to evaluate whether PVL contributes to the disease process (10–18). The infection model for most of these studies was the mouse or the rat, which are relatively insensitive to the cytotoxic effects of PVL (6–8). To the extent that PVL target cell susceptibility is important in pathogenesis, experiments conducted in rodent models could obscure the contribution of PVL and might fail to detect a key role of PMNs, the major host cell target of PVL. Using a rabbit model of necrotizing pneumonia, we found a clear role for PVL in pathogenesis and established a mechanistic basis for PVL-induced acute lung injury and inflammation. Together with previously published results, we postulate the following model of how PVL induces acute lung inflammation and injury (Fig. 5). In this model, PVL-producing S. aureus gain access to the alveoli. At low bacterial burdens, the invading organisms eventually are cleared, suggesting that, if PVL has a role in establishing infection, other predisposing condition(s) such as antecedent or coincident influenza (2–5) are likely to be required also. At high bacterial burdens, PVL is produced in sufficient quantities to activate PMNs and macrophages to release proinflammatory mediators, including IL-8 (21, 28, 33, 34), thereby promoting recruitment of PMNs into the inflamed lung. PVL then lyse PMNs, possibly resulting in the release of granule contents, such as proteases and reactive oxygen metabolites (20, 21, 33). In turn, the toxic products derived from activated or lysed PMNs could damage the alveolar epithelial and endothelial barriers, resulting in influx of fluid and protein from the vascular space into the airspace. Noncardiogenic pulmonary edema and accompanying tissue injury and hemorrhagic lung necrosis ensue, ultimately resulting in death. Knowledge of mechanisms of lung injury gained from this model may lead to new approaches to improve outcome in patients with severe staphylococcal pneumonia.

Fig. 5.

Mechanisms of PVL-induced acute lung injury and lung inflammation. Black arrows indicate observed events; gray arrows indicate postulated events.

Methods

Bacterial Strains.

SF8300, a minimal-passaged USA300 clinical strain representative of the epidemic clone USA300-0114, and its isogenic Δpvl mutant, have been described previously (35). lukS-PV and lukF-PV were reintroduced into their original chromosome sites in the Δpvl mutant by allelic replacement mutagenesis with plasmid pKOR1 (36) using the strategy outlined in Fig. S1. PVL complementation was confirmed by PCR, DNA sequencing, and immunoblotting (Fig. S1).

Other experimental details are provided in SI Text.