Bacterial meningitis is a severe inflammatory disease of the central nervous system that leads to long-term neurological disabilities or death in many patients, even when it is successfully treated with antibiotics (
29). The disease is characterized by massive infiltration of granulocytes into the cerebrospinal fluid (CSF) in response to local production of cytokines, chemokines, and other chemotactic stimuli (
17). The long-term neurological sequelae associated with the disease are due in part to ischemic neuronal damage to the cortex. Both focal and global cerebral blood flow deficits contribute to this damage (
22). While focal deficits appear to be a result of vasculitis and thrombosis, an increase in the intracranial pressure as a result of the vasogenic edema formed due to disruption of the blood-brain barrier (BBB) is, in concert with the loss of autoregulation, an important mechanism that causes global ischemia.
Reactive oxygen species (ROS) are involved in several aspects of the host response to bacterial infection. ROS production by granulocytes, for example, is an important antimicrobial mechanism. The initial step in the formation of granulocyte ROS is catalyzed by the multisubunit enzyme complex NADPH oxidase, which univalently reduces molecular oxygen to superoxide. Phagocyte NADPH oxidase (phox) consists of two membrane-anchored subunits, gp91
phox and p22
phox (which together form the cytochrome
b558 domain), and the cytosolic components p40
phox, p47
phox, p67
phox, as well as the GTP-binding regulatory protein Rac (
19). Normally, the oxidase is dormant, but it is activated upon stimulation through phosphorylation of the p47
phox subunit and subsequent translocation of the cytosolic components to cytochrome
b558 (
27). The importance of NADPH oxidase in the host response against invading microorganisms is manifested in an inherited disease, immunodeficiency syndrome chronic granulomatous disease (CGD). Patients affected by this disease have a mutation in one of the enzyme's subunits, resulting in defective phagocyte ROS production (
28). As a consequence, these patients suffer from recurrent life-threatening bacterial and fungal infections (
8). As implied by the name of the disease, CGD patients exhibit an enhanced inflammatory response which is characterized by excessive granuloma formation. The most common form of CGD, present in about two-thirds of the patients, is X linked and is due to mutations in the gp91 gene. Approximately 20% of affected patients have an aberrant form of the p47 subunit (
34). Defects in the genes encoding p22 and p67 are relatively rare (
5). Recently, transgenic mice with targeted disruption of either the gp91
phox subunit (gp91
−/−) (
23) or the p47
phox subunit (p47
−/−) (
11) were generated as models of CGD. Like patients with CGD, these mice exhibit enhanced susceptibility to common bacterial and fungal infections, as well as signs of an enhanced inflammatory response.
The aim of the present study was to assess the role of NADPH oxidase-derived ROS in pneumococcal meningitis. The results of the study revealed that the host response to pneumococcal CSF infection is different in p47−/− mice than in gp91−/− mice.
MATERIALS AND METHODS
Animals.
gp91
−/− and p47
−/− mice, generated as described previously (
11,
23), were bred under barrier isolation conditions at the animal care facility of Johann Wolfgang Goethe University and were sent to Switzerland for experiments. Animals were backcrossed onto the C57BL6 background for at least five generations. Deletion of the NADPH oxidase subunit was verified by the absence of blood leukocyte superoxide production, as measured by lucigenin-enhanced chemiluminescence (
3). C57BL6 mice from Harlan (Horst, The Netherlands) were used as wild-type (WT) controls. All of the mice were used when they were 12 to 16 weeks old and weighed between 22 and 32 g, and they were sex matched for each set of experiments. Animal experiments were approved by the Animal Care and Experimentation Committee of the Canton of Berne, Switzerland, and were carried out in accordance with the
Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health.
Induction of meningitis with live bacteria.
To induce meningitis with live bacteria, a type 3
Streptococcus pneumoniae strain originally isolated from a patient with bacterial meningitis was used as the infecting organism (
16). A total of 22 WT mice, 17 gp91
−/−mice, and 20 p47
−/− mice anesthetized by intraperitoneal injection with 40 mg of ketamine (Ketalar; Parke-Davis) per kg and 30 mg of xylazinum (Xylasol; Graeub AG) per kg were intracisternally infected with 10 μl of a preparation containing ∼2 × 10
7 CFU of bacteria per ml. A total of 19 WT mice, 14 gp91
−/−mice, and 11 p47
−/− mice were mock infected with sterile saline and served as uninfected controls. Eighteen hours after infection, the animals were weighed, and CSF was obtained by puncture of the cisterna magna. Five-microliter portions were cultured quantitatively on blood agar plates. Animals were sacrificed with an overdose of pentobarbital (100 mg/kg given intraperitoneally) between 20 and 26 h after infection (i.e., when they became moribund), and they were perfused with ice-cold phosphate-buffered saline via the left cardiac ventricle. The rate of spontaneous death was <30% at this time, and no significant difference in the mortality rates of the different strains was observed.
Induction of sterile meningitis.
To induce sterile meningitis, the nonencapsulated
S. pneumoniae mutant R6 was used (
31). R6 grown at 37°C in brain heart infusion broth to the mid-logarithmic phase was harvested by centrifugation, washed with sterile saline, and resuspended in saline at a concentration of ∼5 × 10
10 CFU/ml. The bacteria were heat inactivated at 80°C for 20 min. Ten-microliter portions of heat-inactivated R6 pneumococci were intracisternally injected into 12 WT mice, seven gp91
−/− mice, and six p47
−/− mice. Four WT animals injected with sterile saline served as controls. Animals were sacrificed 12 h after injection, following collection of the CSF. Each CSF sample was centrifuged for 10 min at 10,000 ×
g and 4°C, and the resulting pellet was used to determine white blood cell (WBC) counts after dilution in 0.4% Türk's solution. The CSF supernatant was frozen at −80°C until cytokines were analyzed as described below.
Preparation of cortex homogenates.
Animals were perfused with ice-cold phosphate-buffered saline, and cortical hemispheres were removed, separated from meninges, immediately frozen on dry ice, and stored at −80°C until analysis. Tissue was homogenized (1:9, wt/vol) in 20 mM HEPES buffer (pH 7.4) containing 1 mM EGTA, 0.1 mM dithiothreitol, 0.1% Triton X-100, 0.1 mM phenylmethylsulfonyl fluoride, and a cocktail of additional protease inhibitors (complete mini; Roche Diagnostics, Rotkreuz, Switzerland). Homogenates were centrifuged at 9,000 × g for 15 min at 4°C, and aliquots of the supernatant were stored at −80°C. Protein concentrations were determined by the Pierce bicinchoninic acid assay (Pierce, Rockford, Ill.).
Determination of subarachnoid and ventricular inflammation.
Since the limited amount of CSF available was used to measure bacterial titers in experiments with live bacteria, inflammation was semiquantitatively determined by histological evaluation of tissue sections. Twenty-micrometer-thick coronal sections of formaldehyde-fixed, paraffin-embedded brain tissue at the level of the hippocampus and third ventricle were stained with hematoxylin and eosin and investigated by light microscopy for five animals per group. Images were captured with OpenLab (Improvision, Coventry, United Kingdom) by using a digital camera (Hamamatsu, Photonics, Hamamatsu, Japan) and were scored for the extent of inflammatory infiltrate (>90% of which was neutrophilic granulocytes [
30]) by using NIH Image. Subarachnoid inflammation was determined by measuring the area occupied by infiltrating cells in the subarachnoid space covering the area of (i) the retrosplenial motor cortex or (ii) the perirhinal cortex over a length of 850 μm by using the density slicing procedure (
4). The data were expressed in square micrometers of infiltrate per micrometer of subarachnoid space and were transformed into scores as follows: 0 (no inflammation), 1 (>0 to 10 μm
2/μm), 2 (>10 to 20 μm
2/μm), 3 (>20 to 30 μm
2/μm), etc. (i.e., 1 point per 10 μm
2/μm). Ventricular inflammation was determined by expressing the area of infiltrate as a percentage of the area of the ventricle by using the outline tool and then was given one of the following scores: 0 (no infiltrate), 1 (>0 to 2.5%), 2 (>2.5 to 10%), 3 (>10 to 25%), or 4 (>25%). Finally, the scores for the three separate areas were added to obtain the total inflammation score. Noninfected control animals (three for each strain) all had a score of 0.
BBB integrity.
To visualize differences in BBB integrity, animals received 0.5 ml of 2% (wt/vol) Evans blue intraperitoneally 45 min before sacrifice. Ten-micrometer-thick cryosections were examined for Evans blue extravasation by fluorescence microscopy essentially as described previously (
12). To quantitatively assess changes in BBB permeability, extravasation of serum proteins into the cortex was assessed by measuring serum albumin levels with a mouse-specific enzyme-linked immunosorbent assay (ELISA) kit (Exocell Inc., Philadelphia, Pa.).
MMP-9.
Matrix metalloproteinase 9 (MMP-9) levels were determined by gelatin zymography with minor modifications (
16), essentially as described previously (
14). MMP-9 levels were determined by densitometric analysis of the gelatin lysis zone at ∼95 kDa by using NIH Image. Only the latent form of MMP-9 was detected. Purified human neutrophil MMP-9 (Calbiochem, Darmstadt, Germany) was used as the standard. Under these conditions, samples were within the linear range of the standard curve.
Cytokines.
In animals infected with live bacteria, interleukin-1β (IL-1β) and tumor necrosis factor alpha (TNF-α) were quantitatively determined by using mouse-specific ELISA kits (Quantikine M; R&D Systems Inc., Minneapolis, Minn.). Cortex homogenate supernatants were diluted 1:1 in assay buffer and analyzed in duplicate. The detection limit of the assay was <3.0 pg/ml for IL-1β and <5.1 pg/ml for TNF-α. In animals challenged with heat-inactivated R6 pneumococci, IL-12p70, TNF-α, gamma interferon, monocyte chemoattractant protein 1, IL-10, and IL-6 were simultaneously measured by flow cytometry by using a mouse inflammation cytometric bead array kit (BD Biosciences Pharmingen, San Diego, Calif.). CSF samples were prediluted 1:10 in assay diluent, while cortex homogenates were used undiluted. The detection limits ranged from ∼2.5 pg/ml for gamma interferon to ∼50 pg/ml for monocyte chemotactic protein 1.
Statistical analysis.
Data were analyzed by nonparametric statistics for multiple comparisons (Kruskal-Wallis test followed by Dunn's test) by using Prism 3.0 (GraphPad, San Diego, Calif.). Data were considered to be statistically different at a P value of <0.05. Data in the figures are expressed as medians with upper and lower quartiles and minimum and maximum values.
DISCUSSION
Using transgenic mice, we found that a deficiency in the NADPH oxidase subunit gp91 does not have a significant effect on CSF bacterial titers, BBB disruption, or the overall inflammatory response in animals infected with live pneumococci. In contrast to these results, infection was more severe in mice deficient in the p47 subunit. p47−/− animals had significantly higher CSF bacterial titers and exhibited more pronounced inflammation and increased BBB disruption than both WT and gp91−/− animals.
The results obtained with the gp91
−/− animals are in line with the results of previous studies on the role of granulocytes in the pathogenesis of pneumococcal meningitis. Thus, bacterial growth and final titers in CSF were only marginally higher in rabbits rendered leukopenic than in normal rabbits (
6,
30). Although pneumococci are sensitive to the bactericidal effects of ROS (
24) and granulocytes from animals with pneumococcal meningitis are known to generate ROS (
2), several factors can explain why these ROS are not able to effectively control pneumococcal growth in the CSF. Effective oxygen-dependent killing requires close proximity between bacteria and the location where ROS are generated (e.g., the phagolysosome). However, in CSF the levels of opsonizing complement and immunoglobulins necessary for efficient phagocytosis are low, and encapsulation of bacteria largely prevents phagocytosis by granulocytes. Furthermore, virulent pneumococci express high levels of manganese superoxide dismutase, which also contributes to protection from host-derived ROS (
35). Overall, these findings, including the ones presented here, document that gp91
−/−-dependent ROS formation does not contribute significantly to defense against pneumococcal CSF infection.
Compared to the results obtained with gp91
−/− animals, infection of p47
−/− mice with live pneumococci was significantly more severe. A striking observation was the markedly more intense inflammatory response in p47
−/− mice than in the other two strains of mice. The most likely explanation for this more intense inflammatory response is the increased bacterial titer in p47
−/−mice compared to the bacterial titers in the two other strains. Our experiments with sterile inflammation provide a likely explanation for the higher CSF titers observed after 18 h of infection in p47
−/− mice. In these animals, the inflammatory response (in particular, the WBC infiltration in the CSF) was greatly diminished, while the response in gp91
−/− mice was similar to that in WT animals (Table
1).
Extravasation of leukocytes into the CSF during bacterial meningitis is dependent on expression of intercellular adhesion molecule 1 (ICAM-1) on endothelial cells of the cerebral vasculature (
1,
32). A recent study by Fan et al. showed that TNF-α-induced pulmonary ICAM-1 expression is greatly reduced in p47
−/− mice compared to expression in WT mice (
7). This effect was attributed to a defect in activation of nuclear factor κB, which is important for induction of ICAM-1 expression. Interestingly, inhibition of ICAM-1 expression appeared to be less pronounced in gp91
−/− mice than in p47
−/− mice. A possible explanation for this observation is that oxidants produced by cells that possess a p47-dependent but gp91-independent NADPH oxidase(s), such as vascular smooth muscle cells, are involved in control of endothelial ICAM-1 expression.
In addition to the differential effect of gp91 and p47 on the course of pneumococcal meningitis, our data are of interest with regard to the role of MMPs in the disruption of the BBB in bacterial meningitis. In our model, cortical MMP-9 levels correlated significantly with accumulation of albumin in the cortex, a marker for BBB disruption. A correlation between the level of MMP-9 in CSF and BBB disruption has also been observed for patients suffering from bacterial meningitis (
18). In a rat model of meningococcal meningitis, treatment with batimastat, which inhibits MMP-9 and other MMPs, effectively diminished BBB disruption and the increase in intracranial pressure (
21). Since MMPs such as MMP-9 can be activated by ROS (
20,
26), one could anticipate that granulocyte-derived ROS contribute to BBB disruption via activation of latent MMPs. However, our data clearly show that NADPH oxidase-dependent formation of superoxide is not required for BBB disruption in pneumococcal meningitis.
The actual source of MMP-9 implicated in BBB disruption also remains unresolved. Since we measured MMP-9 levels in cortex homogenates that did not contain contaminating neutrophilic granulocytes (as judged by the absence of increased cortical myeloperoxidase activity [unpublished results]), our data suggest that at least some MMP-9 originates within the parenchyma. This hypothesis is supported by the fact that MMP-9 appears in the CSF before granulocytes can be detected in the CSF (
15). It is known that astrocytes (
33), microglia (
9), and endothelial cells (
10) all produce MMP-9 after stimulation with cytokines in vitro and may therefore be potential sources of MMP-9 in bacterial meningitis, in addition to infiltrating granulocytes. Indeed, exposure of rat brain slices to heat-inactivated R6 pneumococci resulted in increased secretion of MMP-9 in the absence of granulocytes (20).
In summary, in the present study we demonstrated that ROS formation by granulocytes cannot properly control CSF bacterial growth in pneumococcal meningitis. However, our results obtained with p47-deficient animals demonstrated that ROS formation by a gp91-independent NADPH oxidase(s) is important for establishing an adequate inflammatory response to pneumococcal CSF infection.