Paul - Ehrlich - Gesellschaft

Immunomodulation by Antibacterials in Meningitis

K. Stuertz, M. Mäder, R. Nau
Department of Neurology, University of Göttingen, Germany

Despite effective antimicrobial chemotherapy, bacterial meningitis is still associated with a high rate of mortality and neurological sequelae. Without antibacterial treatment the mortality rate is about 90 %, by use of bactericidal antibacterials it still is 20 to 25 % (1, 2). Surviving patients are often affected by neurological sequelae such as hearing loss, epileptic seizures, and paresis (2). Following the onset of antibacterial treatment a burst of meningeal inflammation is frequently observed (3, 4). This inflammatory reaction occurs despite rapid sterilisation of the cerebrospinal fluid (CSF). It is considered to be caused by a rapid release of bacterial components induced by the bacteriolytic action of the antibacterials (5). ß-lactam antibiotics are rapidly bactericidal. As the main aim of chemotherapy in bacterial meningitis is the sterilisation of the CSF (6), ß-lactam antibiotics are used for standard therapy. Their mechanism of antibacterial action is affecting bacterial cell wall synthesis. They are acting by lysing the bacterial cells. Thus, the therapy leads to a strong release of bacterial subcomponents as a result of cell lysis. Many of the subcomponents such as lipopolysaccharides in case of Gram-negative bacteria (7) or peptidoglycan fragments and teichoic and lipoteichoic acids in case of Gram-positive bacteria (8) are pro-inflammatory in animal experiments. The high amount of released bacterial subcomponents can induce the secondary inflammatory reaction and result in a deleterious outcome for the patient (8, 3).

In the brain, the released bacterial subcomponents induce the production of pro-inflammatory cytokines such as tumor necrosis factor alpha (TNF-alpha) and interleukin 1b (IL-1ß). These cytokines are triggering the expression of selectin molecules on the endothelial cell surfaces which enables neutrophils to bind to the endothelial cells. Continuous exposure to TNF-alpha and IL-1ß induces the endothelial cells to release IL-8. IL-8 causes a shift from the expression of selectin molecules to the expression of integrins. This enables the neutrophils to traverse the capillary endothelium. In the CSF, inflammatory cytokines stimulate the neutrophils to produce molecules such as radicals that contribute to the impairment of the blood brain barrier (5). The inflammatory reaction can result in brain oedema and increased intracranial pressure (5). In humans, 20 to 30 % of the meningitis patients that do not survive die from brain herniation (2). Furthermore, heat-inactivated rough pneumococci and pneumococcal cell walls produced direct toxicity on microglial cells and astrocytes cultured in vitro and neurons co-cultured with glial cells (9, 10). Several approaches were made to improve the therapy of bacterial meningitis. One approach is to mitigate the deleterious inflammatory reaction by administering immunosupressive drugs such as glucocorticoids prior to antibacterial treatment. In an animal model of E. coli-meningitis this therapy was shown to reduce several parameters of inflammation and the development of brain oedema in rabbits (6, 7, 3). Children suffering from bacterial meningitis - mainly H.-influenzae-meningitis -  were treated with dexamethasone adjunctively and showed significantly reduced frequency of hearing loss and overall neurological sequelae compared to a group of patients only treated with the antibacterial (11). In adults there is a lack of randomised studies documenting a beneficial effect from dexamethasone. Whether glucocorticoids are beneficial in adults and in children suffering from meningitis caused by other pathogens is still a matter of debate. Furthermore, animal experiments suggest that dexamethasone may increase neuronal damage in the dentate gyrus of the hippocampal formation (12). Another effect of an immunomodulating therapy is that the blood-brain-barrier tends to be less disturbed than without this treatment. However, in meningeal inflammation a disturbed blood brain barrier facilitates the entry of hydrophilic ß-lactam-antibiotics into the central nervous compartments and assure a high concentration in the CSF required for effective and rapid killing of the bacteria. Treatment with dexamethasone decreases the concentration of hydrophilic antibacterials in the CSF in meningitis  and affects CSF sterilisation in a negative way, particularly in the case of pathogens with a decreased sensitivity to antibacterials (13, 14).

A therapeutic approach that circumvents  immunosuppression is to decrease the liberation of bacterial pro-inflammatory cell wall components into the CSF. Therefore antibacterials must be selected which kill the bacteria effectively but do not lyse them or at least delay lysis. Then the killed bacteria  remain normally shaped and can be phagocytosed without releasing great quantities of proinflammatory subcomponents.
Promising antibacterials that may release smaller quantities of proinflammatory components than ß-lactam antibiotics are those with an antibacterial mechanism not acting on the cell wall synthesis.
In vitro, the release of endotoxin during antibacterial therapy by Gram-negative bacteria is drug- and dose-dependent (15, 16, 17, 18). The overall amount of endotoxin released from E.coli cultures treated with ceftazidime and ciprofloxacin was similar but in serial investigation the ciprofloxacin-treated cells released only 12,7 % of the lipopolysaccharide within the first hour of exposure, whereas the ceftazidime-treated bacteria released 61,9 % in this time period (17). The TNF-alpha and IL-6 levels produced by stimulation of whole blood with antibacterially treated bacterial cultures were significantly higher with ceftazidime- and ciprofloxacin-treated cultures than after imipenem and gentamicin treatment. The levels corresponded to the amount of released endotoxin (16).

Treatment of Streptococcus-pneumoniae-cultures with ceftriaxone (ß-lactam) and rifampin (ansamycin) lead to effective killing of the bacteria. However, the scanning electron microscopy revealed obvious differences between both cultures (Fig. 1). The ceftriaxone-treated cells showed blebs and filaments whereas the rifampin-treated were normally shaped.
Recently, we were able to quantify the release of proinflammatory components from Streptococcus pneumoniae (19). The most  potent proinflammatory molecules of this species are the teichoic and lipoteichoic acids and the peptidoglycan fragments (8). The peptidoglycan fragments are not homogeneous in structure. On the other hand, in pneumococci there is the unique situation that the polysaccharide part of both, teichoic and lipoteichoic acid, is identical.
As these molecules are proinflammatory and ubiquitous in all pneumococcal serotypes they are ideal antigens for an enzyme linked immunosorbent assay (ELISA). Using this kind of assay we were able to detect pneumococcal teichoic and  lipoteichoic acid in bacterial culture supernatants down to 0.8 ng/mL.

Fig. 1. Representative scanning electron micrographs of S. pneumoniae type 3 after in vitro exposure to 10 µg of ceftriaxone (A) or rifampin (B) per mL for 6 h. The horizontal bar represents 500 nm. Note the numerous defects of the pneumococcal cell wall after exposure to ceftriaxone. Magnification, x 15,600.

With this ELISA we investigated the release of teichoic and lipoteichoic acids by drugs belonging to different classes of antibacterials. A pathogenic strain of  S. pneumoniae (type 3) was grown in vitro. After resuspension in fresh medium, drugs were added during the logarithmic phase of growth. In  the first set of experiments, ceftriaxone, rifampin, rifabutin, quinupristin-dalfopristin, trovafloxacin, or meropenem at concentrations of 10 µg/mL each were added to the bacterial suspension. The uniformly  high concentration of 10 µg/mL was applied to assure a maximum bactericidal effect for all drugs studied. Samples for the detection of free lipoteichoic acid and teichoic acid were collected every hour up to 12 h after the addition of drugs.
In these experiments (Fig. 2), the bactericidal rates of ceftriaxone, meropenem, quinupristin-dalfopristin, and trovafloxacin were slightly higher than those of rifabutin and rifampin. However, the differences did not reach statistical significance.
The amount of teichoic and lipoteichoic acids detected in the supernatants of the pneumococcal cultures revealed significant differences depending on the antibacterial. Compared to ceftriaxone which is used for standard therapy, only meropenem treatment did not significantly reduce the concentration of teichoic and lipoteichoic acid in the supernatant. Meropenem is a carbapenem belonging to the class of ß-lactam antibiotics and it also acts on cell wall synthesis. The other four antibacterials tested exhibit other modes of action. They are inhibitors of the DNA gyrase (trovafloxacin), of the bacterial RNA-polymerase (rifabutin, rifampin), and of  protein synthesis (quinupristin-dalfopristin), respectively. In the supernatants of the bacterial cultures treated with these four antibacterials, significantly lower concentrations of teichoic and lipoteichoic acids were found during the 12 h lasting experiment. The only exception represented the 10 and 12 h sample from trovafloxacin-treated cultures.

The in vitro experiments showed that despite similar bactericidal rate the release of teichoic and lipoteichoic acid to the culture supernatant was significantly lower in cultures treated with antibacterials not primarily acting on cell wall synthesis than in cultures treated with the standard therapy antibiotic, the ß-lactam ceftriaxone.

Fig. 2. Release of lipoteichoic and teichoic acids by S. pneumoniae type 3 during treatment with 10 µg of ceftriaxone (A), rifabutin (B),
rifampin (C), quinupristin-dalfopristin (D), trovafloxacin  (E), and meropenem (F) per mL and by untreated cultures (G) (closed circles).
The means ± standard errors of the means of five experiments are shown. The asterisks denote significant differences versus the
results for ceftriaxone-treated bacteria. The open sqares represent the bacterial titers (means ± standard error of means).

Lowering the antibacterial concentration in the in vitro experiments led to increasing release of teichoic and lipoteichoic acids (Tab. 1). This effect was observed with all six antibacterials tested. Only rifabutin and rifampin caused, even at the MIC, relatively low release of bacterial subcomponents compared to the untreated control. These in vitro data do not support the concept of using a low first antibiotic dose to prevent the release of proinflammatory cell wall components (19).

In the rabbit model of pneumococcal meningitis several antibacterials tested in the in vitro experiments were used in vivo. Teichoic  and lipoteichoic acids as well as several parameters of inflammation and neurological sequelae were determined for each treatment group compared to the standard therapy with ceftriaxone.

Table 1. Release of LTA and TA by S. pneumoniae type 3 during treatment with antibacterial agents at different concentrations

In a first set of experiments the quinolone trovafloxacin was investigated as an antibacterial for pneumococcal meningitis (20). Animals were injected intracisternally with pneumococci that were precedingly treated with antibacterials in vitro. Following injection of ceftriaxone-treated bacterial cultures consistently higher CSF leukocyte counts (median 2568/µL versus 543/µL at 6 h; p = 0.03; 4560/µL versus 2207/µL at 18 h, p = 0.03) were observed than with trovafloxacin-treated bacterial cultures.
After intracisternal injection of live pneumococci and subsequent antibacterial treatment, the amount of teichoic and lipoteichoic acids in the CSF was lower in the trovafloxacin-treated group than in the ceftriaxone-treated group. At two hours after initiation of antibacterial therapy the difference was significant (21). The IL-1ß concentration in the CSF was lower at two and five hours after initiation of antibacterial therapy in the trovafloxacin-treated group of animals than in the ceftriaxone-treated group (455 pg/mL and 2921 pg/mL versus 1399 pg/mL and 4302 pg/mL; P = 0.02 at 2 hours, difference not significant at 5 h). The TNF concentration in the CSF was also significantly lower than in the standard group two hours after initiation of the antibacterial treatment (26 U/mL versus 141 U/mL;  P = 0.02)). However, the maximum TNF and IL-1ß concentrations were only delayed, but not decreased in comparison to ceftriaxone. The lactate increase in CSF during therapy was significantly lower in the trovafloxacin-treated group. Yet, the parameters of neuronal destruction (CSF neuron specific enolase
[NSE] and density of neuronal apoptoses in the dentate gyrus; 12) were nearly identical. This indicated that trovafloxacin delayed but did not inhibit the inflammatory burst induced by the initiation of antibiotic therapy and did not influence parameters of neuronal damage.
Schmidt et al. (22) tested the new quinolone moxifloxacin for antibacterial treatment of pneumococcal meningitis in the experimental animal model. At a concentration of 10 mg/kg/h it was found to be as bactericidal as the ceftriaxone standard (10 mg/kg/h) whereas it was less effective at 2.5 mg/kg/h. Following antibacterial treatment the TNF release was delayed versus treatment with ceftriaxone reaching its maximum at 17 h versus 14 h. The CSF leukocyte density, lactate and protein were similar in both groups. No significant difference in CSF NSE concentration was observed. Despite a reduced release of proinflammatory bacterial subcomponents in vitro and a delayed production of TNF in vivo the inflammatory reaction following antibacterial treatment of pneumococcal meningitis could not be avoided.

In the CSF of animals treated with rifabutin (5 mg/kg/h) in the rabbit model of pneumococcal meningitis, significantly lower concentrations of teichoic  and lipoteichoic acid than in the CSF of ceftriaxone (10 mg/kg/h)  treated animals were found at two, five and eight hours after initiation of antibacterial therapy while the bactericidal rates of both antibacterials were comparable (21). Consistently, the TNF concentration in the CSF of rifabutin-treated rabbits was significantly lower than in the CSF of ceftriaxone-treated animals at two and five hours after initiation of the antibacterial therapy (23). No difference was found after 12 hours (Fig. 3). Other parameters of inflammation such as CSF lactate, protein, and IL-1ß were found to be similar to those obtained with ceftriaxone-treated rabbits. The median leukocyte density was lower 2 h after initiation of therapy, the difference not reaching statistical significance. The CSF concentration of NSE and the density of apoptotic neurones in the dentate gyrus did not differ significantly (23).

Fig. 3. Time course of CSF TNF-alpha activity during S. pneumoniae meningitis with rifabutin 5 mg/kg/h (n = 8) and ceftriaxone (CRO) 10 mg/kg/h (n = 10).

In conclusion, rifabutin was rapidly bactericidal in vivo and caused a lower release of proinflammatory bacterial subcomponents and a lower level of production of TNF. The probable reason is the dissociation between bacterial killing and lysis. As trovafloxacin, rifabutin did not diminish parameters of neuronal damage in comparison to ceftriaxone.

In the CSF of rabbits treated with a bolus dose  or a continuous infusion of the streptogramin quinupristin-dalfopristin, respectively, the teichoic and lipoteichoic acid concentrations were lower than in the CSF of animals treated with ceftriaxone (21, 24). The difference was significant at five and eight hours after initiation of therapy. The TNF concentrations were lower, too. The differences were significant at two and five hours after initiation of antibacterial therapy in both groups and at eight hours after initiation in the group treated with a bolus of quinupristin-dalfopristin. The CSF lactate and protein concentrations were similar in all groups. After 12 h of treatment with quinupristin-dalfopristin, however, the NSE concentrations in the CSF as a parameter of neuronal damage were significantly lower with both applications than in the ceftriaxone-treated group (Fig. 4). The density of apoptotic neurones was also lower in quinupristin-dalfopristin-treated rabbits, yet, the difference did not reach statistical significance (24).

Fig. 4. Neuron-specific enolase concentrations in the CSF 12 h after initiation of antibiotic therapy were significantly lower in quinupristin-dalfopristin-treated animals (with continuous (a) or short duration (b) infusion) than in ceftriaxone-treated animals. *, P = 0.03; **P = 0.009.

The bactericidal rate of quinupristin-dalfopristin was lower than that of ceftriaxone, probably due to low CSF concentrations, in particular, of dalfopristin. In vitro quinupristin-dalfopristin was as active as ceftriaxone. The very low release of proinflammatory bacterial components, the low TNF concentrations produced, and the delayed increase of leukocyte density as well as the significantly lower CSF NSE values observed, altogether  suggest an attenuated inflammatory response and a reduction of the extent of neurological sequelae. Therefore, streptogramins should be taken into account as an alternative antibacterial option, provided that modified dosing regimens or agents reaching higher CSF concentration and a rapid bactericidal activity in the CSF will be available in the future.

Antibiotics not primarily affecting cell wall synthesis reduce the concentrations of proinflammatory pneumococcal sub-components  in vitro and in vivo in comparison with ß-lactam antibiotics following initiation of therapy. In vivo, the levels of several
parameters of inflammation and with quinupristin-dalfopristin even the CSF NSE concentration as a parameter of  neurological sequelae were reduced following treatment with less bacteriolytic antibacterial.

In the mouse model of pneumococcal meningitis we recently demonstrated that treatment with the ansamycin rifampin reduces the early mortality rate significantly compared to treatment with ceftriaxone (26 % versus 49 %, P = 0.04; 25). Correspondingly, the teichoic and lipoteichoic acid levels were lower in serum samples and pooled CSF of the animals treated with rifampin than in the specimens of the animals treated with ceftriaxone 8 h after the initiation of therapy.

In conclusion, the present experiments demonstrate that the use of antibacterials that act bactericidally without lysing bacteria  is a promising approach toward reducing the secondary inflammatory reaction following antibacterial treatment and early mortality in bacterial meningitis.


1.Wenger, J.D., Hightower, A., Facklam, R.R., Baventa, S. & Broome, C.V. 1990. J. Infect. Dis. 162:1316.

2.Durand, M.L., Calderwood, S.B., Weber, D.J., Miller, S.I., Southwick, F.S., Caviness, V.S. & Swartz, M.N. 1993. N. Engl. J. Med. 328: 21.

3.Mustafa, M.M., Ramilo, O., Mertsola, J, Risser, R.C., Beutler, B., Hansen & E.J., McCracken, G.H. JR. 1989. J. Inf. Dis. 160: 818.

4.Tuomanen, E., Hengstler, Rich, R., Bray, M.A., Zak, O. & Tomasz, A. 1987. J. Infect. Dis. 155: 985.

5.Quagliarello, V.J. & Scheld, W.M. 1997. N. Engl. J. Med. 336: 707.

6.Täuber, M.G., Khayam-Bashi, H. & Sande, M.A. 1985. J. Infect. Dis. 151: 528.

7. Syrogiannopoulos, G.A., Olsen, K.D.,  Reisch, J.S. & McCracken JR., G.H. 1987. J. Infect. Dis. 155: 213. [Erratum: J. Infect. Dis. 1987, 155: 1359]

8.Tuomanen, E., Liu, H., Hengstler, B., Zak, O. & Tomasz, A. 1985. J. Infect. Dis. 151: 859.

9.Kim, Y. S., Kennedy, S. & Täuber, M. G. 1995. J. Infect. Dis. 171: 1363.

10.Kim, Y. S. & Täuber, M. G. 1996. Infect. Immun. 64: 3148.

11.Odio, C.M., Faingezicht I., Paris, M., Nassar, M., Baltodano, A., Rogers, J., Sáez-Llorenz, X., Olsen, K.D. & McCracken JR., G.H. 1991. N. Engl. J. Med. 323: 1525.

12.Zysk, G., Brück, W., Gerber, J., Brück, Y., Prange, H.W. & Nau, R. 1996. J. Neuropath. Exp. Neurol. 55: 722.

13.Paris, M.M., Hickey, S. M., Uscher, M.I., Shelton, S., Olsen, K.D. & McCracken JR., G.H. 1994. Antimicrob. Agents Chemother. 38: 1320.

14.Cabellos, C., Martinez-Lacasa, J., Martos, A., Tubau, F., Fernández, A., Viladrich, P.F. & Gudiol, F. 1995. Antimicrob. Agents Chemother. 39: 2158.

15.Dofferhoff, A.S.M. Nijland, J.H., DeVries.Hospers, H.G., Mulder, P.O.M., Weits, J. & Bom, V.J.J. 1991. Scand. J. Infect. Dis. 23:745.

16.Prins, J.M., Speelman, P., Kuijper, E.J., Dankert, J. & van Deventer, J.H. 1997. J. Antimicrob. Chemother. 39, 13.

17.Evans, M.E. & Pollack, M. (1993). J. Infect. Dis. 167: 1336.

18.Prins, J.M.,  Kuijper, E.J., Mevissen, S.L.C.M., Spellman N.P. & Van Deventer, S.J.H. (1997). Infect. Immun. 63: 2236.

19.Stuertz, K., Schmidt, H., Eiffert, H, Schwartz, P., Mäder, M. & Nau, R. 1998. Antimicrob. Agents Chemother. 42: 277.

20.Nau, R., Zysk, G., Schmidt, H., Fischer, F.R., Stringaris, A., Stuertz, K. & Brück, W. 1997. J. Antimicrob. Chemother. 39: 781.

21.Stuertz, K., Schmidt, H., Trostdorf, F., Eiffert H., Mäder, M., Nau, R. 1999. Lower lipoteichoic and teichoic acid CSF concentrations during treatment of pneumococcal meningitis with non-bacteriolytic antibiotics than with Ceftriaxone. Scand. J. Infect. Dis. 31: 367.

22.Schmidt, H., Dalhoff, A., Stuertz, K., Trostdorf, F., Chen, V., Schneider, O., Kohlsdorfer, C., Brück, W. & Nau, R. 1998. Antimicrob. Agents Chemother. 42: 1397.

23.Schmidt, H., Zysk, G., Reinert, R.R., Brück, W., Stringaris, A., Chen, V., Stuertz, K., Fischer, F., Bartels, R., Schaper, K.-J., Weinig, S. & Nau, R. 1997. Chemotherapy 43: 264.

24.Trostdorf, F., Reinert, R.R., Schmidt, H., Nichterlein, T.,  Stuertz, K., Schmitz-Salue, M., Sadowski, I. Brück, W. & Nau, R. 1999. J. Antimicrob. Chemother. 43: 87.

25.Nau, R., Wellmer, A., Soto, A., Koch, C., Schneider, O., Schmidt, H., Gerber, J., Michel, U. & Brück, W. Rifampin reduces early mortality in experimental Streptococcus pneumoniae meninigitis. J. Infect. Dis., in press.