Main

Fatality and neurologic deficits still occur following bacterial meningitis in children despite newer and more potent antibiotics(1–5). Adverse neurodevelopmental sequelae are believed to arise both from direct bacterial damage and from the inflammatory host response(1–3,6–8). Neurotrophic growth factors are peptides that support cell populations and maintain homeostasis in the CNS(9–11). Their biosynthesis can be modulated by activation of neuronal pathways or through interaction with neurotransmitters and cytokines(9–11). Beyond their potent neurotrophic activity, these molecules can act as adaptive agents by rescuing or repairing brain cells after injury(10–12). Basic fibroblast growth factor (bFGF) may have such neurotrophic effects. It is expressed at high levels in the CNS and exerts trophic activity on neurons and astrocytes in different brain regions(9–11). Expression of bFGF in the CNS is elevated in response to neuronal activation and to several different kinds of damage to brain cells(11,13–15). However, little is known regarding the expression of bFGF during bacterial meningitis.

We tested the hypotheses that the expression of bFGF in CNS is enhanced in experimental bacterial meningitis and the level of bFGF in cerebrospinal fluid (CSF) is elevated in children with bacterial meningitis. In the experimental study, we used immunohistochemical methods to examine the expression of bFGF in the brains of mice with experimental Streptococcus pneumoniae meningitis. In the clinical study, we used an immunoassay to measure the level of bFGF in CSF from children with bacterial meningitis and related the levels to their disease severity.

SUBJECTS AND METHODS

Animal Model of Experimental Meningitis

Animals and bacterial strains. C3H/HeNCrj mice were obtained from Charles River Breeding Laboratories, Inc. (Portage, Japan). The strains of serotype 14 S. pneumoniae were clinical isolates obtained from National Cheng Kung University Medical Center. All cultures were grown in Todd Hewitt broth (Difco Laboratories, Detroit) at 37°C, and at midlogarithmic phase the organisms were harvested, washed in saline, and diluted to 103 colony forming units (CFU) as determined by OD at 600 nm (OD600 of 1 = 1 × 108 CFU/mL) (Beckman DU-640, Spectrophotometer, Beckman Instrument; Somerset NJ).

Experimental meningitis. The model of experimental meningitis was modified from Tsai et al.(16). Under ether anesthesia, 14 mice were infected intracranially by injecting 103 CFU S. pneumoniae in 50 µL of normal saline suspension through the right orbital surface of the zygomatic bone (posterior corner of the right eye). This was performed with 27-gauge 0.5-inch disposable hypodermic needles (Becton-Dickinson, Rutherford, NJ). The needle was introduced just beneath the zygomatic bone and directed toward the right side of the skull to a depth of about 5 mm. The control group (n = 10) received intracranial injection with 50 µL of normal saline. After the procedure, the mice were allowed to awaken and were returned to their cages. To confirm the development of bacterial meningitis, the brains of six mice injected with bacteria and four mice injected with saline were aseptically removed and homogenized in 2 mL of 3% PBS 24 h after the injections. After 500-fold dilution with PBS, 10 µL of the homogenized brains were cultured on blood agar plates and incubated for 24 h at 37°C in room air plus 5% CO2. All the homogenized brains from the S. pneumonia infected group were positive for S. pneumoniae (with an average count of 17 × 106 CFU/brain), and those from the control group were all negative.

Histologic and immunohistochemical staining. Twenty-four hours after the intracranial injections of bacteria or saline, the mice were killed and brains were removed for examination. The mice were deeply anesthetized with 60 mg/kg pentobarbital and perfused through the heart with 100 mL of warm PBS followed by 400 mL of ice-cold 3% paraformaldehyde in PBS for 1 h. The brains were removed and fixed in the same fixative overnight. Coronal sections were embedded in paraffin and cut into 6-µm-thick sections. Sections for histology were stained with hematoxylin and eosin. Immunohistochemistry was performed according to the standard streptavidin-biotin peroxidase technique using a Dako LSAB kit as described previously(13,14). The bFGF antibody, purchased from Oncogen Science, was a rabbit affinity-purified polyclonal antibody against peptide (AILFLPM), which corresponded to amino acids 147-153 of human bFGF. The specificity of the anti-bFGF was confirmed by abolishing positive immunoreactivity by prior incubation of the antibody (1 µg/mL) with 1 µg of recombinant bFGF (Oncogen Science). After the sections were blocked in 3% H2O2 for 15 min and in a blocking solution for 20 min to eliminate nonspecific binding, the sections were incubated overnight at 4°C in the primary antibody of polyclonal anti-bFGF 1:100 (Oncogen Science). After the sections were rinsed, they were sequentially incubated in biotinylated secondary antibody for 1 h and streptavidin-peroxidase solution for 1 h at room temperature and rinsed with PBS three times between incubations. The sections were finally reacted in 3-amino-9-ethylcarbazole in the presence of 0.003% H2O2 for 15 min and counterstained with hematoxylin. Sections were then washed and mounted on gelatin-coated slides.

Patients

Clinical characteristics. This study was approved by the Review Board of National Cheng Kung University Medical Center. Informed consent was obtained from all the parents. Between 1993 and 1996, we obtained CSF samples from 18 children with bacterial meningitis (10 boys and 8 girls aged 2.5 mo to 11 y; mean 22 mo), 12 children with aseptic meningitis (8 boys and 4 girls aged 65 d to 7 y; mean 29 mo), and 18 febrile children who required diagnostic lumbar puncture but who had no evidence of CNS infection when their CSF was examined (11 boys and 7 girls aged 72 d to 14 y; mean 25 mo) from our hospital.

The diagnosis of bacterial meningitis was based on clinical manifestations, CSF white blood cell (WBC) count >20 cells/mm3 with a predominance of polymorphonuclear leukocytes, and the presence of bacteria in CSF shown by at least one of the following: a positive Gram stain preparation, a positive bacterial latex particle agglutination (Wellcogen, Wellcome Diagnosticus, England), or a positive culture(4,5,12). The day of onset of fever was considered the first day of illness. Children who had received i.v. antibiotics before their initial lumbar puncture and those whose CSF remained sterile were excluded from analysis. All 18 patients with bacterial meningitis received dexamethasone (0.6 mg/kg/d i.v. for 4 d) before initiation of antibiotics and had proper antibiotic treatment according to the result of cultures or latex particle agglutination test (i.e. ampicillin or penicillin plus one of the following: cefotaxime, ceftriaxone, ceftazidime). Second lumbar punctures were performed in 17 of the 18 children 24-50 h after initiation of antibiotics. The disease severity at presentation was scored according to the system of Herson and Todd(17).

The diagnosis of aseptic meningitis was made based on the following criteria: 1) CSF WBC >20 cells/mm3 with a predominance of mononuclear cells; 2) negative bacteriologic studies, including CSF and blood cultures; and 3) self-limited clinical course without clinical evidence of encephalopathy, including seizures, disturbance of consciousness, or focal neurologic signs.

All CSF samples were collected and aliquoted at the time of diagnosis. They were immediately centrifuged and stored at -80°C for later assay. The CSF leukocyte count, glucose, and protein concentrations were determined by routine methods.

bFGF immunoassay. CSF samples were assayed for bFGF using Quantikine ELISA kit (Research and Diagnostics Systems, Inc., Minneapolis) based on a sandwich technique, as previously described(12,18). Briefly, the microtiter plate was precoated with murine MAb specific for bFGF. Both the recombinant standard and the sample were pipetted into the wells and left to react for 2 h at room temperature. After washing, horseradish peroxidase-linked polyclonal antibody specific for bFGF was added to the wells to "sandwich" the immobilized bFGF. Substrate solution was then added to develop the color after washing off the unbound antibody-enzyme reagent. The absorbance was read using a spectrophotometer (Molecular Device Corp., Sunnyvale, CA) set at 450 nm, and the concentration of bFGF was determined. Each case was analyzed in duplicate, and the mean values were used as the final concentration. No cross-reactivity to recombinant acidic FGF (Sigma) was demonstrated by testing up to 500 pg/mL, and the sensitivity (1 pg/mL) was determined by using recombinant bFGF (Sigma). For those samples undetectable by immunoassay, analysis was performed by assigning a value of 0.5 pg/mL (half of the detection limit).

Outcome. Patients were followed up for a mean duration of 13 months to determine any neurodevelopmental sequelae. The major sequelae included death, blindness, hydrocephalus, quadriplegia, microcephaly, severe retardation, and intractable seizures. The results of the bFGF assay were not known to those evaluating the patients (CCH and YCC) until after they had made their diagnoses.

Statistics. We compared the levels of bFGF in CSF between children with bacterial meningitis, aseptic meningitis, and control subjects with linear rank statistics based on the medians, which minimize the effect of any outliers. The differences between pretreatment and posttreatment CSF bFGF in children with bacterial meningitis were analyzed using Wilcoxon signed rank test. Values were expressed as median (range). Differences were considered significant at p < 0.05.

RESULTS

Experimental Meningitis

Clinical manifestations and histopathologic findings. By 24 h after intracranial bacteria inoculation, the mice were lethargic or obtunded, refused water, and had signs of CNS damage, such as ataxia, tremor of the head and extremities, myoclonic jerks, and arching of the neck and back. No such abnormalities were observed in the control mice. In the histopathologic sections from the bacteria-inoculated mice, the leptomeninges and the lateral ventricles were infiltrated with neutrophils, macrophages, fibrins (Fig. 1), as well as large numbers of pneumococci (Fig. 2). Neither inflammatory cells nor bacteria were present in animals injected with sterile saline.

Figure 1
figure 1

Photomicrographs of brain histopathology of S. pneumoniae meningitis by hematoxylin and eosin stain. (A) Normal leptomeninges from cerebral cortex of control mice; magnification ×40. (B and C) Acute leptomeningitis, 24 h after intracranial infection of S. pneumoniae, showing accumulation of fibrinopurulent materials, neutrophils, macrophages, and fibrins in the leptomeninges; magnification ×40. (D) Normal choroid plexus in the ventricle of control mice; magnification ×40. (E and F) Acute ventriculitis, 24 h after intracranial infection of S. pneumoniae, showing fibrinopurulent materials, neutrophils, macrophages, and fibrins accumulated in the lateral ventricle; magnification ×40.

Full size image
Figure 2
figure 2

Photomicrographs of brain 24 h following intracranial injection of S. pneumoniae demonstrating large amounts of leukocytes as well as pneumococci seen in the leptomeninges (A) and within the lateral ventricles (B). Hematoxylin and eosin stain, magnification ×1000.

Full size image

bFGF immunocytochemical findings. In the brains from the control mice, the leptomeninges, choroid plexus, blood vessel walls, and the ependymal lining of the lateral ventricles were stained for bFGF. Cerebral neurons showed very little bFGF immunoreactivity (Figs. 3, A and C, and 4A). In the brains from the mice with experimental meningitis, the infiltrating neutrophils, debris-laden macrophages, and fibrins in the leptomeninges, especially within the ventricles, showed strong bFGF immunoreactivity (Figs. 3B and 4, B-D). The neurons adjacent to the inflamed ventricles showed coarse cytoplasmic granules with bFGF immunoreactivity (Fig. 3D).

Figure 3
figure 3

Photomicrographs of bFGF immunohistochemical staining. (A) In control mice brain, bFGF immunostaining was seen in the leptomeninges; magnification ×40. (B) 24 h after intracranial infection of S. pneumoniae, strong expression of bFGF was seen in the infiltrating inflammatory cells and fibrins in the leptomeninges; magnification ×40. (C) In normal control, bFGF immunostaining was seen in the blood vessels, but very little bFGF immunoreactivity was demonstrated in the neurons; magnification ×100. (D) Enhanced bFGF immunoreactivity was demonstrated in the neurons adjacent to the inflamed lateral ventricle; magnification ×100.

Full size image
Figure 4
figure 4

Photomicrographs of bFGF immunohistochemical staining. (A) In normal control, bFGF immunostaining was seen in the choroid plexus and ependymal lining of lateral ventricles; magnification ×40. (B and C) Strong bFGF immunoreactivity demonstrated in the infiltrating neutrophils, debris-laden macrophages, and fibrins within the inflamed ventricle 24 h following intracranial infection of S. pneumonia; magnification ×40. (D) Close view of debris-laden macrophages and neutrophils within the lateral ventricle showed heavy bFGF staining; magnification ×1000.

Full size image

Patient Characteristics and CSF bFGF Concentrations

The CSF isolates from patients with bacterial meningitis included six Hemophilus influenzae type b, four S. pneumoniae, two Salmonella, two Pseudomonas aeruginosa, one Escherichia coli, one group B streptococcus, one citrobacter spp., and one Neisseria meningitidis. Three patients died during admission, and two patients had major neurologic sequelae. The pathogens in these five patients of major sequelae included one case of H. influenzae type b, two S. pneumoniae, one Salmonella, and one P. aeruginosa.

No significant effect of age on the concentrations of CSF bFGF was found in control subjects. The median levels of CSF bFGF at the time of diagnosis in patients with bacterial meningitis, aseptic meningitis, and control subjects are shown in Table 1. The CSF levels of bFGF were twice as high in children with bacterial meningitis (median 7.21 pg/mL), compared with those who had aseptic meningitis (2.9 pg/mL), or in control subjects (2.65 pg/mL, p < 0.0001, respectively). If we exclude the three children with bacterial meningitis who had CSF bFGF levels higher than 20 pg/mL who might be outliers, the median in the cases decreases to 6.8 pg/mL, but the differences still remain highly significant. The small difference between the values for the children with aseptic meningitis and controls was not significant.

Table 1 Levels of CSF bFGF at the time of diagnosis in three groups of children
Full size table

There was no correlation between CSF bFGF concentrations and CSF leukocyte count, glucose concentration, or protein concentration. Although the child with N. meningitidis meningitis who had the highest CSF WBC (27 750/mm3) and CSF bFGF level of 14.96 pg/mL had a complete recovery, the child with salmonella meningitis who had the highest CSF bFGF of 358 pg/mL and CSF WBC of 126/mm3 died. In general, the CSF bFGF levels in patients with bacterial meningitis decreased significantly 24-50 h after antibiotic treatment (p < 0.0005) (Fig. 5) to the point where there were no longer any differences between the levels in children who had had bacterial meningitis, aseptic meningitis, or controls. The exceptions were two children whose illnesses were fatal and whose bFGF actually rose after antibiotic therapy was begun.

Figure 5
figure 5

The levels of CSF bFGF at the time of diagnosis and 24-50 h after initiation of antibiotic therapy in children with bacterial meningitis. •, patients with fatality; ▪, patients with major neurologic sequelae; ▴, patients who survived without major sequelae.

Full size image

Five of the patients with bacterial meningitis died or survived with major neurologic sequelae and 13 survived without major sequelae. At admission, the CSF WBC count was lower (780 versus 3469 WBC/mm3) and the protein concentration higher (514 versus 236 mg/dl) in the more seriously ill children, but these measurements were variable, and the differences were not significant. The bFGF concentration, however, was 18-fold higher in the more severely affected children, and the difference was significant (p < 0.05) (Table 2). All three patients who had CSF bFGF above 20 pg/mL died of severe brain edema.

Table 2 Clinical characteristics, CSF bFGF, and outcome in patients with bacterial meningitis
Full size table

DISCUSSION

In a murine model of bacterial meningitis, our study demonstrated enhanced bFGF immunoreactivity in the neutrophils and macrophages infiltrating the leptomeninges and ventricles and also in the neurons adjacent to the inflamed lateral ventricles 24 h after bacterial inoculation. In brains from control mice, the choroid plexus epithelial cells and blood vessel walls had high concentrations of bFGF, but the amount of bFGF present in neurons was barely detectable with immunohistochemistry. Clinically, we found that CSF bFGF is significantly higher in children with acute bacterial meningitis as compared with those with aseptic meningitis or children without CNS infections. In addition, patients with bacterial meningitis who had markedly elevated levels of CSF bFGF on admission had much poorer outcomes. These findings of enhanced expression of bFGF during experimental bacterial meningitis and the association of CSF levels of bFGF with the disease severity in childhood bacterial meningitis suggest a biologic role for this neurotrophic factor in the pathophysiology of bacterial meningitis.

The FGFs are a family of closely related molecules showing affinity to heparin and heparin sulfate proteoglycan. The fact that FGFs have neurotrophic, angiogenic, and gliogenic capacities and that brain is a ready source of FGFs suggest that these molecules may play vital pathophysiologic roles in the nervous system. Several studies have shown that neurons and glial cells are capable of producing bFGF. The amount of FGFs present in normal brains is so low that FGFs are barely detectable with immunohistochemistry. Our study in normal mice demonstrated that high concentrations of bFGF were found in the choroid plexus, the ependymal lining of the ventricles and blood vessel walls. Subsequent to bacterial meningitis, the neurons and the proliferating nonneural cells (leukocytes and macrophages) were induced to express an increased amount of FGFs(11,14,15).

The potential role of enhanced bFGF expression in the inflammatory cells 24 h after experimental S. pneumoniae meningitis remains to be elucidated. In vivo, bFGF was shown to dilate pial arterioles and increase cortical blood flow, which may be related to the microvascular changes and early hyperemia during the initial phase of experimental pneumococcal meningitis(19–21). In vitro, bFGF has potent trophic properties for neurons, glia and endothelia, and supports the growth and survival of CNS neurons(22). Moreover, increased expression of bFGF immunoreactivity can be found after brain injury or stroke, suggesting bFGF plays a role in the neural sprouting and in the glial and vascular proliferation that occurs during these processes. Brain injury can induce repair responses characterized by proliferation of glial cells (gliosis), blood vessels (angiogenesis), and macrophages. Twenty-four hours following experimental S. pneumoniae meningitis, strong expression of bFGF immunoreactivity was observed in the infiltrating neutrophils and macrophages within the leptomeninges and ventricles and in the periventricular neurons. However, gliosis and angiogenesis were not yet evident. These findings suggest that the early expression of bFGF may be related to its trophic properties in response to cell damage following bacterial meningitis. Whether there is a later phase or continued expression of bFGF related to its role in promoting cellular proliferation and tissue repair days after bacterial meningitis needs further investigation(14).

It is still not clear why bFGF is elevated during the acute phase of bacterial meningitis, but several possibilities are consistent with our results. First, the elevated bFGF may result from up-regulation of bFGF production in response to bacterial invasion(11,15). During experimental bacterial meningitis, there is a continued influx of neutrophils and macrophages that express strong bFGF immunoreactivity. And, as a part of the inflammatory response, CSF and tissue levels of tumor necrosis factor-α and IL-1β are elevated(1–3,6–8). Both cytokines may be the initiators of a cascade of brain responses, among which is the stimulation of production of bFGF(23,24). Excitatory amino acids, especially glutamate, released by brain cells or accumulated within CSF during bacterial meningitis may turn on the expression of target genes including bFGF(23,25–27). Brain structures adjacent to the ventricles are freely exposed to high levels of CSF glutamate, as the damaged ependymal lining of the ventricle is highly permeable to diffusion particularly during bacterial meningitis(25–27). It is conceivable that high glutamate concentrations in the ventricular CSF contribute to the enhanced expression of bFGF in the periventricular neurons, which in turn may also contribute to the elevated level of CSF bFGF during bacterial meningitis. Taken together, enhanced expression of bFGF in the recruited inflammatory cells and in the periventricular neurons, as demonstrated in our experimental meningitis model, may be responsible for the elevation of CSF bFGF during bacterial meningitis in children.

Alternatively, tissue damage and leakage of bFGF into CSF may be brought about by bacterial meningitis. In our data, higher levels of CSF bFGF meant a higher case fatality and more frequent neurologic sequelae among survivors. Support of this hypothesis derives from observations that bFGF is bound to heparan sulfate in the extracellular matrix and could be released in an active form when the extracellular matrix-heparan sulfate is degraded by heparanase expressed by platelets and neutrophils(28). Hence, the extracellular matrix could be a reservoir for bFGF in the brain. The unusually high CSF levels of bFGF occur with poor outcome because the high levels are a consequence of the destruction of brain tissue. In addition, ventriculitis, which generally has a poor outcome, seemed to be produced in our animal model and might be associated with higher CSF bFGF clinically.

It has been recently reported that FGF binding protein (FGF-BP) is detectable in human CSF. FGF-BPs are soluble truncated forms of the high affinity FGF receptor and function in vivo to modulate the biologic activity of FGF(29,30). Because FGF-BPs avidly bind bFGF, the FGF/FGF-BP dimer may function to deliver FGF to sites of neuronal injury or to sequester FGF and complete with the high affinity cell surface receptor. There is not yet a method for quantitating these binding proteins, so we did not measure them in our studies.

Although statistically nonsignificant, our data showed that the patient with major sequelae have lower CSF WBC than those without. The finding that patients with a low CSF WBC count (<1000/mm3) tend to have a poor outcome, including fatality, had been reported in our previous study and by others as well(5,17,31,32).

In our patients, dexamethasone was routinely administered to children with bacterial meningitis before initiation of antibiotic therapy. The advantage of this adjunctive therapy was suggested to be down-regulating the production of tumor necrosis factor-α and IL-1, which are the fundamental events in the initiation and maintenance of inflammatory responses during bacterial meningitis(1–3). In addition, dexamethasone can up-regulate the synthesis of bFGF in the cerebral cortex and hippocampus(33,34). Because bFGF is mitogenic for endothelial cells, dexamethasone may help stimulate the repairing process of angiogenesis in the injured brain, which can be one of the beneficial effects of dexamethasone adjunctive therapy in reducing neurologic complications in children with bacterial meningitis.

bFGF may play a role during bacterial meningitis. Additional experimental investigations of the time course of this response and whether it is modulated by anti-inflammatory drugs might provide information useful in understanding this severe and often fatal childhood disease.