OUP user menu

High mobility group box 1 prolongs inflammation and worsens disease in pneumococcal meningitis

Christopher Höhne , Michael Wenzel , Barbara Angele , Sven Hammerschmidt , Hans Häcker , Matthias Klein , Angelika Bierhaus , Markus Sperandio , Hans-Walter Pfister , Uwe Koedel
DOI: http://dx.doi.org/10.1093/brain/awt064 1746-1759 First published online: 21 March 2013

Summary

Neutrophilic inflammation, which often persists over days despite appropriate antibiotic therapy, contributes substantially to brain damage in bacterial meningitis. We hypothesized that persistent inflammation is the consequence of a vicious cycle in which inflammation-induced cell injury leads to the release of endogenous danger molecules (e.g. high mobility group box 1) that drive the inflammatory response, causing further damage. The present study aimed to assess the mechanisms of high mobility group box 1 protein release and its functional relevance for the development and progression of pneumococcal meningitis. High mobility group box 1 was found in large quantities in cerebrospinal fluid samples of patients and mice with pneumococcal meningitis (predominantly in advanced stages of the disease). By using macrophages, we demonstrated that the release of high mobility group box 1 from macrophages following pneumococcal challenge is passive in nature and probably not connected with inflammasome- and oxidative stress-dependent inflammatory cell death forms. In a mouse meningitis model, treatment with the high mobility group box 1 antagonists ethyl pyruvate or Box A protein had no effect on the development of meningitis, but led to better resolution of inflammation during antibiotic therapy, which was accompanied by reduced brain pathology and better disease outcome. Additional experiments using gene-deficient mice and murine neutrophils provided evidence that high mobility group box 1 acts as a chemoattractant for neutrophils in a receptor for advanced glycosylation end products-dependent fashion. In conclusion, the present study implicated high mobility group box 1, likely released from dying cells, as a central propagator of inflammation in pneumococcal meningitis. Because persistent inflammation contributes to meningitis-associated brain damage, high mobility group box 1 may represent a promising target for adjunctive therapy of this disease.

  • meningitis
  • danger-associated molecular pattern
  • persistent inflammation

Introduction

Bacterial meningitis is among the top 10 causes of infection-related deaths worldwide and many survivors suffer from permanent neurological and otological sequelae (Brouwer et al., 2010). The most frequent aetiological agent in Europe and the United States is Streptococcus pneumoniae, being responsible for more than half of all cases (van de Beek et al., 2004; Thigpen et al., 2011). Pneumococcal infection of the meninges generates an excessive inflammatory reaction that contributes substantially to tissue injury in this disease (Koedel et al., 2010). The immune response is initiated by the recognition of pathogen-associated molecular patterns on invading pathogens by host pattern recognition receptors such as TLR2 and TLR4 (Klein et al., 2008). Toll-like receptor engagement induces MYD88-dependent production of pro-inflammatory cytokines of which interleukin 1 (IL1) family cytokines are of major importance (Koedel et al., 2002, 2004). IL1β and IL18, activated by the NLRP3/caspase 1 axis (the NLRP3 inflammasome), amplify and propagate the inflammatory response through a positive feedback loop involving IL1 receptor-mediated, MYD88-dependent signalling (Zwijnenburg et al., 2003; Hoegen et al., 2011; Witzenrath et al., 2011). As a consequence, large numbers of blood-borne leucocytes, predominantly neutrophils, are recruited into the leptomeninges. The invading neutrophils are not able to destroy their microbial targets (Ernst et al., 1983), instead the neutrophils damage bystander cells, presumably by secreting injurious factors such as free radicals or lysosomal proteases (Hoffmann et al., 2007; Koedel et al., 2009).

Stressed or injured cells can release alarm signals (so-called danger-associated molecular patterns) that elicit an immune response or modulate an ongoing immune response (Matzinger, 2007; Kono and Rock, 2008). Well-known danger-associated molecular patterns include but are not limited to heat shock proteins, IL1α, IL33, S100 proteins or the high mobility group box 1 protein (HMGB1) (Bianchi 2007). HMGB1 is an ubiquitously expressed, highly-conserved nuclear protein that stabilizes nucleosome formation and regulates transcription (Andersson and Tracey, 2011). The inside–outside translocation of HMGB1 can occur through two separate mechanisms. Inflammatory cells such as macrophages can secrete HMGB1 upon stimulation with pathogen-associated molecular patterns or pro-inflammatory cytokines through a non-conventional pathway that requires inflammasome assembly and caspase 1 activation (Lamkanfi et al., 2010; Andersson and Tracey, 2011). In addition, HMGB1 can be passively released from dying cells following nuclear and cell membrane disintegration (Scaffidi et al., 2002; Bell et al., 2006). Extracellular HMGB1 behaves much like a cytokine. HMGB1, by itself and/or by forming complexes with exogenous or endogenous pro-inflammatory molecules, induces and enhances cytokine synthesis (Bianchi, 2009; Andersson and Tracey, 2011). In addition, HMGB1 increases chemotaxis and accumulation of granulocytes at inflammatory sites (Orlova et al., 2007; Penzo et al., 2010; Berthelot et al., 2012), and inhibits phagocytosis and clearance of apoptotic leucocytes (Feng et al., 2008; Liu et al., 2008). The biological effects of extracellular HMGB1 are mediated by signalling pathways coupled to Toll-like receptors and the receptor for advanced glycation end products (RAGE) both of which are involved in inflammatory responses (Hori et al., 1995; Yu et al., 2006). The critical contribution of HMGB1 to disease pathogenesis was first described in an experimental mouse model of endotoxemia (Wang et al., 1999). In this setting, HMGB1 was not detectable in the circulation until hours after the onset of disease. The late occurrence preceded and paralleled with the onset of animal lethality (Andersson and Tracey, 2011). Treatment with HMGB1 antagonists conferred protection against lethal endotoxemia and sepsis (Wang et al., 1999; Yang et al., 2004). Moreover, systemic administration of purified HMGB1 to mice recapitulates many clinical signs of sepsis (Wang et al., 1999). In the brain, injection of recombinant HMGB1 was demonstrated to increase IL1β and tumour necrosis factor α expression, to mediate anorexia, fever and memory deficits, and to have proconvulsant activity (Agnello et al., 2002; Maroso et al., 2010; Mazarati et al., 2011). Moreover, endogenous HMGB1 was identified as an early mediator of post-ischaemic and post-traumatic inflammation (Liu et al., 2007; Su et al., 2011). Administration of HMGB1 inhibiting agents was also proven neuroprotective in animal models of cerebral ischaemia and trauma (Liu et al., 2007; Muhammad et al., 2008; Su et al., 2011). These reports suggest that HMGB1 is an important player in sterile neuroinflammation and acute neurodegeneration. The protein’s role in infections of the brain and its coverings, however, is unknown. Of note, two case studies have recently reported that HMGB1 is released into the CSF during acute bacterial meningitis in children (Tang et al., 2008; Asano et al., 2011). Our present in vitro and in vivo studies aimed to assess the mechanisms of HMGB1 release and its functional relevance for the development and progression of pneumococcal meningitis, which represents a common and very serious form of brain infection.

Materials and methods

Ethics statement

This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, National Research Council, USA) and with the German Animal Protection Act. The study protocol was approved by the Committee on the Ethics of Animal Experiments of the Government of Upper Bavaria (Permit Numbers: 55.2-1-54-2531-31-09).

Animal model of pneumococcal meningitis

A well-characterized mouse model of pneumococcal meningitis was used (Koedel et al., 2009; Woehrl et al., 2011). Briefly, adult male C57BL/6n mice were weighed and their body temperature was taken. The mice were clinically examined and scored. Clinical scoring consisted of: a beam balancing test, a postural reflex test and the presence of piloerection, seizures or reduced vigilance. In healthy animals, the score was 0 points; mice with 11 or more score points were considered terminally ill and euthanized for ethical considerations. None of the investigated mice showed scores >10 within the observation period. After clinical evaluation, bacterial meningitis was induced by intracisternal injection of 20 µl of 107 colony forming units (cfu)/ml S. pneumoniae type 2 (D39 strain) under short-term anaesthesia with isoflurane. Controls were injected intracisternally with 20 µl PBS. Next, animals were allowed to wake up and food and water were supplied ad libitum. All animals that were studied longer than 24 h after infection received antibiotic therapy with ceftriaxone, starting 21 h after infection. At the end of each experiment, animals were weighed again, scored clinically, and their temperature was taken. Then, mice were anaesthetized with ketamine/xylazine and a catheter was placed into the cisterna magna. CSF samples were withdrawn for the determination of CSF leucocyte counts and CSF HMGB1. Thereupon, intracranial pressure was measured. Subsequently, blood samples were taken by transcardial puncture. After deep anaesthesia with ketamine/xylazine, mice were perfused with 15 ml ice-cold PBS containing 10 U/ml heparin. The brains were removed and frozen immediately.

Experimental groups

To assess the role of HMGB1 in the development of pneumococcal meningitis, C57BL/6n (Charles River GmbH) were infected and immediately thereafter treated intraperitoneally with 400 µg/mouse recombinant HMGB1 box A protein (n = 6; IBL International) and 50 mg/kg body weight Ringer’s ethyl pyruvate solution (REPS; n = 10 Sigma Aldrich), respectively. Infected mice that were either injected intraperitoneally with HEPES-buffered saline (HEPES-NaCl; vehBoxA; n = 6) or with Ringer’s lactate solution (RLS; n = 9) served as positive controls, whereas mice which were intracisternally injected with PBS were used as negative controls (n = 5). In order to obtain insights into the function of HMGB1 in the progression of meningitis, infected C57BL/6n mice were treated with 100 mg/kg body weight ceftriaxone either in combination with 400 µg/mouse recombinant HMGB1 box A protein (n = 8) or with 50 mg/kg body weight REPS (n = 9). Respective vehicles (RLS, n = 9; HEPES-buffered saline, n = 10) were injected to infected (positive) controls. In addition, mice injected intracisternally with PBS served as negative controls (n = 4). In order to check the impact of HMGB1 on neutrophil apoptosis in the CSF, treatment with ceftriaxone and REPS or ceftriaxone and RLS was started 18 h after infection and CSF samples were withdrawn 6 h later (n = 5, each group) and assessed for the presence of apoptotic leucocytes. For evaluation of the role of RAGE in mediating HMGB1 effects in meningitis, RAGE-deficient mice backcrossed to the C57BL6/n genetic background (RAGE−/−) were infected and compared with wild-type mice. These animals were followed until 24 h and 45 h (n = 7, each group) after infection, respectively. In an additional experimental series, RAGE−/− mice were infected, treated either with ceftriaxone and REPS or with ceftriaxone and RLS (n = 6, each group) at 21 h after infection, and investigated 24 h later. Finally, we assessed whether HMGB1 by itself can induce leucocyte migration into the CSF. Recombinant HMGB1 (5 µg; IBL International) was injected into the cisterna magna of wild-type, RAGE−/− and TLR2/TLR4 double-deficient mice, and CSF leucocyte counts were determined 6 h after injection (n = 4, each group). In an additional group, wild-type mice were pretreated intracisternally with 25 µg neutralizing anti-HMGB1 antibody 1 h before the application of the protein (n = 4). Mice which were injected intracisternally with PBS or with heat inactivated recombinant HMGB1 (5 µg) served as controls (n = 4, each group).

Determination of bacterial titres in blood and brain

Cerebella were dissected and homogenized in sterile saline. Blood samples and cerebellar homogenates were diluted serially in sterile saline, plated on blood agar plates, and cultured for 24 h at 37°C with 5% CO2.

Analysis of cerebral bleeding

Mice brains were cut in a frontal plane into 10-µm thick sections. Beginning from the anterior parts of the lateral ventricles, 10 serial sections were photographed with a digital camera in 0.3 mm intervals throughout the ventricle system. Haemorrhagic spots were counted and the bleeding area was measured (Image tool, UTHSCSA).

Immunohistochemical detection of RAGE and HMGB1

Ten micrometre-thick coronal brain sections containing the lateral ventricles and hippocampal tissue were stained with a goat anti-mouse polyclonal antibody directed against mouse RAGE (R&D Systems) or a rabbit anti-mouse polyclonal antibody directed against HMGB1 (Acris Antibodies). After quenching endogenous peroxidase activity with 0.3% methanolic hydrogen peroxide and blockage of non-specific binding by 10% normal rabbit serum, brain sections were incubated with the anti-RAGE antibody in a dilution of 1:200 or the anti-HMGB1 antibody in a dilution of 1:500 overnight at 4°C. Labelled cells were visualized using biotinylated rabbit anti-goat or goat-anti rabbit IgG at a 1:200 dilution, followed by horseradish peroxidase-conjugated streptavidin and then 3,3’-diaminobenzidine as a chromogen (all from Vector Laboratories). After counterstaining with Mayer’s haematoxylin solution, tissue sections were examined using an Olympus BX51 microscope and images captured with a cooled Moticam 5000 video camera connected to a PC.

Measurement of brain IL1β, TNFα, TGFβ, CXCL1 and CXCL2 concentrations

Mouse brain concentrations of IL1β, TNFα, TGFβ, CXCL1 and CXCL2 were assessed by ELISA (R&D Systems), according to the manufacturer’s instructions.

Cell culture experiments

The murine macrophage cell line J774.2 (Health Protection Culture Collections) was cultured in Dulbecco’s modified Eagle medium (Sigma Aldrich) supplemented with very-low endotoxin 10% foetal bovine serum (Biochrom) and penicillin/streptomycin (Sigma Aldrich). For experiments, the culture medium was replaced by Dulbecco’s modified Eagle medium supplemented with 1% Nutridoma-SP (Roche Diagnostics). Cells were seeded at a concentration of 100 000 cells per well in a 96 well plate. After 24 h of incubation, cells were exposed to different concentrations of a heat-inactivated (at 60°C for 60 min), non-encapsulated isogenic mutant of S. pneumoniae (D39Δcps strain) for different durations. In separate experiments, the following compounds were added to the culture medium: z-YVAD-fmk (50 µM; Biocat), potassium chloride (65 mM), oxidized ATP (0.01, 0.1 and 1 mM), A438079 (1, 10 and 100 µM), diphenylene iodonium (10 µM, Merck Chemicals), N-acetyl-l-cysteine (2.5 mM), Mn(III)tetrakis (4-benzoic acid) porphyrin (50 µM), and ethyl pyruvate (1 mM, 10 mM).

Bone marrow-derived macrophages [from wild-type and apoptosis-associated Speck-like protein containing a CARD (ASC)-deficient mice] were prepared from bone marrow cells isolated from the femur as described previously (Hoegen et al., 2011). Bone marrow-derived neutrophils from RAGE-deficient and wild-type mice were isolated from femurs and tibias and subsequently loaded on top of a discontinuous Percoll gradient (52%/64%/72%) and centrifuged at 1000g for 30 min (Frommhold et al., 2010). Bone marrow-derived neutrophils were harvested from the 64%/72% interface, washed in PBS, and suspended at a concentration of 2 × 106 cells/ml in chemotaxis medium.

Hoxb8 neutrophil progenitors derived from C57BL/6n mice (Wang et al., 2006; Koedel et al., 2009) were cultured in Opti-MEM® medium (Life Technologies) supplemented with 10% foetal calf serum, 30 mM β-mercaptoethanol (Sigma Aldrich), penicillin/streptomycin, 4% supernatant from stem cell factor-producing Chinese Hamster ovarian cells, and 1 mM oestrogen (Sigma Aldrich). Neutrophil differentiation was induced by removal of oestrogen. After culture for 5 days in differentiation medium, Hoxb8 neutrophils were used for chemotaxis experiments.

Chemotaxis assay

Neutrophil migration was assayed using a 48-well microchemotaxis chamber (Neuroprobe) as described previously (Woehrl et al., 2010). Briefly, recombinant HMGB1, diluted in RPMI 1640 with 1% foetal calf serum at concentrations of 0.05, 0.5 and 5.0 µg/ml, given alone or in combination with a neutralizing chicken anti-HMGB1 polyclonal antibody (50 µg/ml; IBL International), was placed in the lower well (25 µl). HoxB8 neutrophils (2 × 106 cells/ml), untreated or pretreated for 1 h with a RAGE-blocking antibody or its isotype control antibody (25 µg/ml each; both from R&D Systems), were added to the upper well, which was separated from the lower well by a polycarbonate, polyvinylpyrrolidone-free micropore filter. The filter pore sizes were 5 µm. In selected experiments, bone marrow-derived neutrophils from RAGE-deficient or wild-type mice were used instead of HoxB8 neutrophils. After 120 min, the filters were removed, fixed with methanol and stained with DiffQuik (Baxter Diagnostics AG). Chemotaxis was quantified by microscopic counting of cells that migrated completely through the filter pores in 10 randomly chosen high power fields. Cell migration was expressed as the mean number of leucocytes that migrated per field. Recombinant CXCL2 (R&D Systems) was used as a positive control at concentrations between 0.05 ng/ml and 5 ng/ml.

Determination of lactate dehydrogenase

The lactate dehydrogenase activity was determined in centrifuged cell culture supernatants, centrifuged supernatants of control cells after lysis with Triton™ X-100 (positive controls), and in control medium (negative controls) using a colorimetric assay kit (BioVision, Biocat). Cytotoxicity was calculated as percentage lactate dehydrogenase release by the ratio of (supernatants − negative control/positive control − negative control) × 100.

Western blot analysis of HMGB1

HMGB1 release was analysed in (i) CSF samples from four adult patients with culture-proven pneumococcal meningitis and four age- and sex-matched patients with non-inflammatory neurological disorders (two with migraine, two with idiopathic facial palsy); (ii) CSF samples obtained from mice of different experimental groups; and (iii) cell culture supernatants by western blot analysis. In all patients, lumbar puncture was performed for diagnostic purposes after informed consent. The use of the samples was approved by the local ethical committee. In order to concentrate HMGB1 in cell culture supernatants, 15 µl StrataClean resin (Agilent Technologies) was added to 200 µl cell culture supernatant. The resin-bound proteins were recovered by centrifugation, washed with PBS and redissolved in HEPES buffer containing 10 mM HEPES, 10 mM KCl 1.5 mM MgCl2, and a mixture of protease inhibitors. Ten microlitres of this protein solution or 3 µl of CSF were diluted 1:1 or 1:6 in sample buffer (125 mM Tris–HCl, 4% SDS, 0.05% bromophenol blue, 20% glycerol and 5% β-mercaptoethanol), and heated to 70°C for 10 min. Then, 20 μl were loaded on 4% to 12% NuPAGE® Tris-Bis gel (Life Technologies) and subjected to electrophoresis. After transfer of proteins on polyvinylidene fluoride membranes (Merck Millipore), membranes were incubated for 18 h at 4°C with a rabbit anti-HMGB1 polyclonal antibody (1:2000 dilution, Novus Biologicals Europe). Bound primary antibodies were detected using a peroxidase-conjugated antibody against rabbit IgG (1:2000 dilution; Sigma Aldrich) and the FemtoMax™ Super Sensitive chemiluminescence substrate kit (Rockland Inc.). Blots were visualized and digitalized using a Doc-It®LS Image analysis system (UVP Inc.).

Statistical analysis

The principal statistical test was one-way ANOVA, followed by Bonferroni post hoc testing. Differences were considered significant at P < 0.05. Data are displayed as means ± standard deviation (SD).

Results

HMGB1 release into cerebrospinal fluid during pneumococcal meningitis

Recent case studies reported significantly higher levels of HMGB1 in CSF samples obtained from children with bacterial meningitis than in controls (Tang et al., 2008; Asano et al., 2011). Accordingly, our western blot analysis demonstrated the release of large quantities of HMGB1 (with concentrations >4 µg/ml) into the CSF of adults with pneumococcal meningitis (Fig. 1A). Similar to the human situation, HMGB1 was detectable by immunoblotting in CSF samples obtained from mice with pneumococcal meningitis, but not from uninfected controls (Fig. 1B and Supplementary Fig. 1). Of note, the HMGB1 immunoreactivity was higher at 45 h than at 24 h after infection, pointing at a possible role of this protein in advanced rather than in early stages of the disease.

Figure 1

HMGB1 is released into the CSF during pneumococcal meningitis. HMGB1 levels were assessed by western blot analysis of CSF samples either obtained from (A) patients with pneumococcal meningitis (PM) and non-inflammatory neurological disorders (NIND; n = 4 for each group) or (B) from mice subjected to pneumococcal meningitis. Mice were untreated, treated with REPS or its respective vehicle RLS. CSF samples were withdrawn from infected mice by intracisternal puncture 24 or 45 h after pneumococcal infection. Control CSF was obtained from mice injected intracisternally with PBS. rHMGB1 = recombinant HMGB1.

Extracellular HMGB1 potently drives inflammation during antimicrobial therapy

To test the functional significance of extracellular HMGB1 in the development and/or progression of pneumococcal meningitis, mice were treated either at the time point of or 21 h after infection with the HMGB1 antagonists box A or REPS and examined 24 h later (at 24 h or 45 h after infection). Box A is a competitive antagonist of extracellular HMGB1 that displaces HMGB1 binding to cells (Yang et al., 2004), whereas ethyl pyruvate (REPS) is a stable and lipophilic derivative of pyruvate known to inhibit HMGB1 release (Ulloa et al., 2002). According to the latter, application of REPS resulted in the near absence of extracellular HMGB1 in CSF samples obtained from mice 24 h (not shown) and 45 h after induction of pneumococcal meningitis (Fig. 1B).

Within 24 h after pneumococcal inoculation, all mice—irrespective of the treatment regimen—developed clinical signs of infection that manifested in significantly increased clinical score values, compared with uninfected control mice. Clinical symptoms were accompanied by significant increases in CSF leucocyte counts and brain cytokine concentrations, a significant rise in intracranial pressure, and the occurrence of cerebral haemorrhages (Fig. 2A–H). Neither pretreatment with box A nor with REPS had any impact on the development of meningitis. The number of leucocytes in the CSF was comparable between mice that received the HMGB1 antagonists and those that were injected with the respective vehicles HEPES-NaCl and RLS (Fig. 2A). In addition, brain concentrations of major inflammatory mediators of pneumococcal meningitis, namely IL1β, TNFα, CXCL1, CXCL2 (Fig. 2B–E), and TGFβ (3.34 ± 0.34 pg/mg protein in RLS-treated mice, compared with 3.11 ± 0.85 pg/mg protein in REPS-treated mice; 3.35 ± 0.76 pg/mg in HEPES-NaCl-treated mice, compared with 3.79 ± 0.88 pg/mg in box A-treated mice), were not altered by pretreatment with box A or REPS. Compatible with the lack of effect on brain inflammation, there were no between-group differences in the degree of intracranial complications like the rise in intracranial pressure or the magnitude of intracerebral haemorrhages. Correspondingly, the clinical status assessed 24 h after infection was similar between mice that were treated with box A or REPS and those that received the respective vehicles (Fig. 2F–H). In addition, pretreatment with box A or REPS did not result in significant alterations of pneumococcal titres in the brain (6.36 ± 0.31 log10 cfu/organ in RLS-treated mice, compared with 6.32 ± 0.34 log10 cfu/organ in REPS-treated mice; 6.29 ± 0.55 log10 cfu/organ in HEPES-NaCl-treated mice, compared with 6.07 ± 0.47 log10 cfu/organ in box A-treated mice) and the blood (data not shown).

Figure 2

HMGB1 antagonism has no affect on the development of meningitis. Mice were treated intraperitoneally with recombinant Box A protein (Box A; n = 6), ethyl pyruvate (REPS; n = 9), or their respective vehicles (vehBoxA; n = 6, or RLS; n = 9). Then, pneumococcal meningitis was induced by intracisternal injection of S. pneumoniae (strain D39). Twenty-four hours later, animals were evaluated. (A) The number of white blood cells (WBC) in the CSF was comparable between mice that received HMGB1 antagonists and those that were injected with their respective vehicles. (B–E) Brain concentrations of IL1β, TNFα, CXCL1, and CXCL2 were also not altered by pretreatment with HMGB1 antagonists. (F and G) There were also no between-group differences in the rise of intracranial pressure (ICP) and the number of cerebral haemorrhages. (H) Moreover, the clinical status was not affected by pretreatment with HMGB1 antagonists. Data are given as means ± SD.

Different results were obtained, however, when HMGB1 antagonists were given to mice with established meningeal inflammation. To rescue mice from overwhelming infection and associated death, mice have to be treated with an antibiotic in this series of experiments. Antibiotic treatment (with ceftriaxone) results in rapid elimination of the bacteria as no or only a few bacteria can be cultured from the blood, CSF or even the brain tissue 24 h after the start of antibiotic therapy (Koedel et al., 2009; Grandgirard et al., 2010). At this disease stage (45 h after infection), vehicle-treated mice still differed significantly from uninfected control mice with regard to CSF leucocyte counts, intracranial pressure values, numbers of cerebral haemorrhages, and clinical score values (Fig. 3A and F–I). Treatment with box A or REPS did not modulate the killing of S. pneumoniae by ceftriaxone (2.00 ± 1.80 log10 cfu/organ in RLS-treated mice, compared with 2.43 ± 0.77 log10 cfu/organ in REPS-treated mice; 1.78 ± 1.24 log10 cfu/organ in HEPES-NaCl-treated mice, compared with 0.87 ± 1.24 log10 cfu/organ in box A-treated mice). However, mice that received ceftriaxone in combination with box A or REPS showed significantly lower leucocyte counts in the CSF sampled at 45 h after infection than those that were injected with ceftriaxone and the respective vehicles (Fig. 3A; for more information on statistics see Supplementary Table 2). The reduction in CSF pleocytosis was not related to lower brain levels of IL1β, TNFα, TGFβ (data not shown), CXCL1 or CXCL2, as the expression of these factors had returned to nearly normal levels under antibiotic therapy, irrespective of whether or not the mice were treated with HMGB1 antagonists (Fig. 3B–E). Notwithstanding the above, the reduction in CSF leucocyte numbers was associated with a significant amelioration of both brain pathology and disease severity, as evidenced by lower intracranial pressure values, fewer haemorrhagic spots, and lower clinical scores (Fig. 3F–I and Supplementary Table 1). These data suggest a major role of HMGB1 in the persistence of massive leucocyte infiltrates in antibiotic-treated meningitis, which in turn results in an aggravation of meningitis-associated brain damage.

Figure 3

HMGB1 antagonism accelerates resolution of inflammation. Pneumococcal meningitis was induced by intracisternal injection of S. pneumoniae (strain D39). Twenty-one hours later, mice were treated intraperitoneally with a combination of ceftriaxone and recombinant Box A protein (Box A; n = 8), ethyl pyruvate (REPS; n = 9), or their respective vehicles (vehBoxA; n = 10, or RLS; n = 9). Twenty-four hours after the start of treatment (45 h after infection), animals were evaluated. (A) Mice that received ceftriaxone in combination with box A or REPS showed significantly lower white blood cell (WBC) counts in the CSF than those that were injected with ceftriaxone and the respective vehicles. (C–E) There were no between-group differences in the brain levels of IL1β, TNFα, CXCL1 or CXCL2. (F–I) The reduction in CSF pleocytosis was associated with a significant amelioration of brain pathology and disease severity, as evidenced by lower intracranial pressure values (F), less macroscopically visible haemorrhages (G and H) and lower clinical scores (I). Data are given as means ± SD. *P < 0.05, compared with vehBoxA-treated, infected mice using ANOVA and Bonferroni’s test for post hoc analysis. #P < 0.05, compared with RLS-treated, infected mice using ANOVA and Bonferroni’s test for post hoc analysis. Con = control.

HMGB1 functions as a chemokine in cerebrospinal fluid

HMGB1 can orchestrate an acute inflammatory process by modulating cytokine/chemokine production by immune cells, inhibiting neutrophil apoptosis and phagocytosis of apoptotic neutrophils, and mediating chemotaxis of neutrophils (and monocytes) (Andersson and Tracey, 2011). Our finding that HMGB1 antagonism did not affect the production of key cytokines and chemokines during meningitis argues against the possibility that HMGB1 mediates the persistence of meningeal infiltrates after the start of antibiotic therapy through acting as a pro-inflammatory cytokine-like factor. To test the hypothesis that the persistent leucocyte infiltration is due to HMGB1-mediated inhibition of leucocyte apoptosis, we analysed the cell composition and morphology in CSF samples obtained from mice 6 h after the start of therapy either with ceftriaxone and REPS or ceftriaxone and RLS. At this time point, total CSF leucocyte counts tended to be lower in REPS-treated than in RLS-treated mice (14 213 ± 1609 cells/µl versus 19 113 ± 4792 cells/µl). Differential leucocyte counts did not reveal differences in the cellular composition of the CSF between both groups. Likewise, the ratio of apoptotic leucocytes, identified by their nuclear pyknosis and karyrhexis, was similar between mice that received ceftriaxone and REPS and those that were injected with ceftriaxone and RLS (8.0 ± 3.7% versus 9.3 ± 3.2% of 200 counted leucocytes). Thus, the HMGB1-associated persistence of leucocyte infiltration seems not to be attributable to the inhibition of leucocyte apoptosis. We next tried to clarify whether HMGB1 propagates leucocyte recruitment in established meningitis through its direct chemotactic activity for neutrophils and monocytes (Rouhiainen et al., 2004; Orlova et al., 2007; Penzo et al., 2010; Berthelot et al., 2012). First, we investigated the effect of HMGB1 (0.05–5 µg/ml) on murine, in vitro-differentiated HoxB8 neutrophils in a Transwell migration assay. CXCL2 (0.05–500 ng/ml) was used as a positive chemoattractant control. The highest concentration of HMGB1 increased neutrophil migration to a comparable extent as CXCL2 (Fig. 4A). Co-application of HMGB1 and neutralizing anti-HMGB1 antibodies resulted in a significant reduction of the neutrophil migration across the porous membrane. Moreover, pretreatment of HoxB8 neutrophils with a RAGE-blocking antibody (but not with its isotype control antibody) significantly attenuated cellular migration towards HMGB1. Correspondingly, recombinant HMGB1 exerted chemoattractant activity in bone marrow-derived neutrophils from wild-type mice, but not from RAGE-deficient mice (wild-type neutrophils exposed to medium and HMGB1: 4.8 ± 0.9 cells/microscopic field and 17.8 ± 0.9 cells/microscopic field, compared with RAGE-deficient neutrophils exposed to medium and HMGB1: 5.9 ± 0.7 cells/microscopic field and 9.9 ± 0.8 cells/microscopic field, respectively). Next, we injected HMGB1 into the CSF of mice and determined CSF leucocyte counts 6 h later. This led to a significant increase in CSF leucocyte counts, compared with negative controls that received PBS or heat-inactivated HMGB1 (Fig. 4B). According to our in vitro finding, intrathecal pretreatment with neutralizing anti-HMGB1 antibodies resulted in a marked reduction of HMGB1-induced CSF pleocytosis. To gain further insight into the receptors responsible for HMGB1-induced leucocyte recruitment, we administered HMGB1 to RAGE-deficient and TLR2/TLR4-double-deficient mice, respectively. RAGE- (but not TLR2/TLR4-double-deficient) mice displayed lower leucocyte numbers than wild-type mice (Fig. 4B). Combined, these data suggest that, when present in the CSF, HMGB1 can act as a chemoattractant and this activity seems to depend on RAGE.

Figure 4

HMGB1 acts as a chemokine in the CSF. (A) Chemotactic activity of recombinant HMGB1 (rHMGB1) on murine neutrophils (differentiated, conditionally Hoxb8-immortalized cells) was assessed in vitro using a microchamber chemotaxis assay. Recombinant CXCL2 (n = 8 for each concentration) was used as a positive control. Data are given as means ± SD. *P < 0.05, compared with medium controls (Con; n = 16); #P < 0.05, compared with the group that received the highest CXCL2 concentration; $P < 0.05, compared with 5 µg /ml rHMGB1; +P < 0.05, compared with 5 µg/ml rHMGB1 plus 25 µg/ml RAGE isotype control antibody, using ANOVA and Bonferroni’s post hoc test. (B) To investigate the chemotactic activity of rHMGB1 in vivo, 5 µg of the recombinant protein was injected into the CSF through the cisterna magna. Six hours later, CSF samples were withdrawn and analysed for the presence of white blood cells (WBC). WT = wild-type mice (n = 19 in total). AkHMGB1 = neutralizing antibody directed against HMGB1 (n = 4). RAGE−/− = RAGE-deficient mice (n = 4). TLR2/4−/− = TLR2 and TLR4 double-deficient mice (n = 4). Data are given as means ± SD. *P < 0.05, compared to PBS-injected controls (Con; n = 4); #P < 0.05, compared with mice that received 5 µg of heat-inactivated rHMGB1 (HI-HMGB1; n = 4); $P < 0.05 compared with 5 µg rHMGB1 (n = 7), using ANOVA and Bonferroni’s post hoc test. MF = microscopic field; PMN = polymorphonuclear.

HMGB1 activity is mediated through RAGE signalling

Because HMGB1 provoked CSF pleocytosis through RAGE activation, and box A was suggested to preferentially antagonize HMGB1 binding to RAGE (Muhammad et al., 2008), we reasoned that HMGB1-mediated effects in meningitis are triggered through its interaction with RAGE. First, we investigated the in vivo expression pattern of RAGE in mouse brains by immunohistochemistry, finding mild staining in the wall of large meningeal vessels and in some leptomeningeal cells in control mice. In infected mice, large vessel walls stained strongly positive for RAGE (Fig. 5A–D). The staining intensity was alike in the different stages of the disease (24 h and 45 h after infection). In addition, a positive immunostaining for RAGE was seen in the inflammatory infiltrate. Next, we subjected RAGE-deficient mice to pneumococcal meningitis. RAGE deficiency had no impact on the development of meningitis, as evidenced by comparable CSF leucocyte counts, brain bacterial titres, and clinical scores in the RAGE-deficient and wild-type mice at 24 h after infection (Fig. 5E–G). However, infected RAGE-deficient mice displayed significantly lower CSF leucocyte counts than wild-type mice when CSF samples were obtained from antibiotic-treated animals at 45 h after infection (Fig 5E–G). The reduction in CSF pleocytosis was also accompanied by an amelioration of disease severity. Moreover, treatment of RAGE-deficient mice with REPS did not result in further significant decreases of CSF leucocyte counts and clinical scores (Fig. 5E–G). Together, these data support the concept that the HMGB1–RAGE axis is a crucial pathway mediating the persistence of leucocytic inflammation in meningitis.

Figure 5

HMGB1 activity requires interaction with RAGE. (A–D) Immunohistochemistry of brain sections with antibodies to RAGE: (A) uninfected control mouse, (B and C) wild-type mouse 24 h and 45 h after induction of meningitis, (D) RAGE-deficient mouse 45 h after induction of meningitis. Black arrows and white asterisks indicate positive immunostaining for RAGE in walls of large meningeal vessels and the inflammatory infiltrate, respectively. Scale bar = 50 µm. (E–G) To assess the functional role of RAGE, we subjected RAGE-deficient mice to pneumococcal meningitis. RAGE-deficient mice had lower CSF white blood cell (WBC) counts clinical score values than wild-type mice at 45 h (but not at 24 h) after infection (n = 7 for each group). Treatment of RAGE-deficient mice with ethyl pyruvate (REPS; n = 6) did not result in further decreases of CSF WBC counts and clinical scores, as compared with RAGE-deficient mice and RAGE-deficient mice that received the vehicle RLS (n = 6). Data are given as means ± SD. #P < 0.05, compared to infected wild-type (WT) mice that were evaluated 45 h after infection, using ANOVA and Bonferroni’s test for post hoc analysis.

HMGB1 is released from dying macrophages independent of inflammasome activation and oxidative stress

To characterize HMGB1 release during pneumococcal infection, we next challenged mouse J774.2 macrophage-like cells with heat-killed pneumococci. Immunoblot analyses revealed that large amounts of HMGB1 were liberated from J774.2 cells in a time- and concentration-dependent manner, beginning 4 h after stimulation at concentrations above 2 × 107 cfu/ml (Fig. 6A). The release of HMGB1 was strictly accompanied by a loss of cell viability, as judged by an increase of lactate dehydrogenase in the cell supernatants (Fig. 6B).

Figure 6

HMGB1 release from macrophages is passive in nature and independent from inflammasome activation and oxidative stress. HMGB1 levels were assessed by western blot analysis of cell culture supernatants from J774.2 macrophages and bone marrow-derived macrophages (BMDM). Cell death was determined by measurement of lactate dehydrogenase (LDH) release. If not otherwise indicated, cells were exposed to 5 × 107 cfu/ml heat-inactivated pneumococci (HKP). (A) Pneumococci induced a concentration- and time-dependent release of HMGB1 from J774.2 cells. (B) HMGB1 release was paralleled by an increase of lactate dehydrogenase in the cell supernatants. To gain insight into the role of inflammasome assembly and caspase 1 activation in HMGB1 release, we treated J774.2 macrophages with inflammasome antagonists and additionally used bone marrow-derived macrophages from wild-type and ASC-deficient mice. (C and D) Neither addition of z-YVAD-fmk nor potassium chloride (KCl) substitution did alter HMGB1 and lactate dehydrogenase release into the cell culture supernatant. (E) Correspondingly, equal HMGB1 levels were detectable in supernatants from J774.2 cells, wild-type and ASC-deficient bone marrow-derived macrophages after exposure to live S. pneumoniae (D39 strain). Supplementation of the culture medium with oxidized-ATP (oxATP) abolished the liberation of HMGB1 and lactate dehydrogenase into the supernatant following pneumococcal challenge. (F) To clarify the role of oxidative stress in pneumococci-induced HMGB1 release, we treated J774.2 cells with the antioxidants diphenylene iodonium (DPI), N-acetyl-l-cysteine (NAC), and Mn(III)tetrakis (4-benzoic acid) porphyrin (MnTBAP). None of these substances affected HMGB1 release provoked by pneumococcal challenge. In contrast, HMGB1 levels were substantially lower in ethyl pyruvate (EP)-treated cells than in control cells. Data are given as means ± SD. *P < 0.05, compared with unstimulated cells (medium; n = 6); #P < 0.05, compared to heat-killed pneumococci-stimulated cells, using ANOVA and Bonferroni post hoc test.

HMGB1 release was recently shown to occur downstream of inflammasome assembly and caspase 1 activation through unconventional protein secretion or as a consequence of cell lysis during pyroptosis (Lamkanfi et al., 2010; Lu et al., 2012). Inflammasome activation, in turn, was found to be triggered by pneumococcal infection (Hoegen et al., 2011; Witzenrath et al., 2011). We therefore comparatively analysed the caspase 1 antagonist z-YVAD-fmk, the P2X7 purinoceptor antagonists oxidized-ATP and A438079, as well as purposeful increase of extracellular potassium concentration for their potential to interfere with HMGB1 release. Their effectiveness as inflammasome inhibitors was proven by IL1β ELISA measurements. The heat-killed pneumococci-induced rise in extracellular IL1β concentrations was absent in supernatants from cells treated with z-YVAD-fmk, P2 purinoceptor antagonists or high extracellular potassium (data not shown). Neither addition of z-YVAD-fmk nor potassium chloride substitution altered HMGB1 and lactate dehydrogenase release into the cell culture supernatant of J774.2 cells (Fig. 6C and D). Correspondingly equal levels of HMGB1 (and lactate dehydrogenase; data not shown) were found in supernatants from pneumococci-stimulated wild-type and ASC-deficient bone marrow-derived macrophages (Fig. 6E). Moreover, treatment with the selective P2X7 purinoceptor antagonist A438079 failed to modulate pneumococci-induced HMGB1 and lactate dehydrogenase release (data not shown). Remarkably, however, supplementation of the culture medium with oxidized-ATP completely abolished the liberation of HMGB1 and lactate dehydrogenase into the supernatant following pneumococcal challenge (Fig. 6C and D).

Oxidative stress (e.g. peroxynitrite) was also assumed to cause passive release of HMGB1 into the extracellular space through the induction of necrosis (Loukili et al., 2011). In an attempt to clarify its role in pneumococci-induced HMGB1 release, we treated J774.2 cells with the well-known antioxidants diphenylene iodonium, N-acetyl-l-cysteine, and Mn(III)tetrakis (4-benzoic acid) porphyrin. None of these substances was found to be protective against cell injury (data not shown) and subsequent HMGB1 liberation provoked by exposure to heat-killed pneumococci (Fig. 6F). In contrast, supplementation with ethyl pyruvate led to results similar to those obtained in vivo. The levels of HMGB1 in the cell culture supernatant were substantially lower in ethyl pyruvate-treated cells than in control cells. This effect was paralleled by an inhibition (of 41.4 ± 2.9%; 10 mM ethyl pyruvate) but not prevention of pneumococci-induced cell lysis. Taken together, our data suggest that, following pneumococcal challenge, HMGB1 leaks out of macrophages that are presumably dying through inflammasome- and oxidative stress-independent processes.

Discussion

During the past decades, experimental investigations provided strong evidence that meningitis-related brain injury is largely caused by the massive neutrophilic inflammatory reaction in the brain (Gerber and Nau, 2010; Koedel et al., 2010). Neutrophilic inflammation is only little modified by appropriate antibiotic therapy over days even though complete CSF sterilization occurs within hours (Blazer et al., 1983; Viallon et al., 2005). We hypothesized that the delay in resolution of inflammation is—at least partly—the consequence of a cycle in which inflammation-induced cell injury leads to the release of endogenous danger-associated molecular pattern that drives the inflammatory response, causing further damage. Here, we showed that (i) HMGB1 is released in large quantities into the CSF of patients and mice with pneumococcal meningitis; (ii) HMGB1 antagonisms interfered with the progression (but not the development) of meningitis, resulting in less brain pathology and better short-term outcome; and (iii) HMGB1 acts primarily as a chemoattractant through binding to RAGE.

In a case study, Tang et al. (2008) reported that HMGB1 levels were significantly elevated in CSF samples from children with bacterial meningitis (3 of 13 children suffered from pneumococcal meningitis) as compared with those from children with no or aseptic meningitis. Correspondingly, we detected large quantities of HMGB1 in the CSF of adult patients with pneumococcal meningitis as well as in mice subjected to pneumococcal meningitis. In the mouse model, we observed a substantial rise in CSF HMGB1 between early and advanced meningitis (between 24 h and 45 h after infection). This pattern fits in with the currently favoured concept of HMGB1 release during invasive thread, mainly deduced from sepsis studies (Andersson and Tracey, 2011). HMGB1 is considered as a later mediator in infection that is actively released into the extracellular milieu by macrophages in response to bacterial products and/or pro-inflammatory cytokines (Wang et al., 1999; Jiang and Pisetsky, 2006). In an attempt to gain further insight into the mechanism of HMGB1 release, we exposed J774.2 macrophage-like cells to pneumococci. HMGB1 was liberated from J774.2 cells as early as 4 h after pneumococcal challenge. The appearance of HMGB1 in the culture medium closely paralleled that of lactate dehydrogenase, a measure of cell death and plasma membrane damage. This finding is in line with recent studies that demonstrated that the release of HMGB1 from macrophages (and other cells) following challenge with bacterial products like peptidoglycan (Rose et al., 2011) or live bacteria such as Streptococcus suis, Shigella flexneri and Klebsiella pneumoniae (Tenenbaum et al., 2006; Willingham et al., 2007, 2009) is passive in nature and probably connected with the inflammatory cell death forms pyroptosis and pyronecrosis (Fernandes-Alnemri et al., 2007; Willingham et al., 2007). Both cell death programmes require the inflammasome adapter ASC, whereas caspase 1 activation is important only in pyroptosis, but not in pyronecrosis. Our experiments using inhibitors or gene-deficient cells showed that pneumococci-induced macrophage cell death is independent of inflammasome activation, caspase 1 activity and ASC. Of note, we observed that oxidized-ATP, but not A438079 (both substances are widely used as P2X7 receptor antagonists) was capable of preventing lactate dehydrogenase and HMGB1 liberation from stimulated macrophages. This difference may be explained by P2X7-independent effects of oxidized-ATP, like direct interference with Toll-like receptor activation or blockade of downstream signalling pathways after cellular entry as proposed by Beigi et al. (2003). In addition, oxidized-ATP was reported to block endosomal acidification as well as to reduce intracellular superoxide concentrations (Chen et al., 2006; Moayeri et al., 2006). Reactive oxygen species are generated rapidly and in large quantities by macrophages upon exposure to pneumococci (Koedel and Pfister, 1999) and are known inducers of cell death as well as HMGB1 release (Tang et al., 2007; Loukili et al., 2011). In our study, neither the superoxide dismutase mimetic and peroxynitrite scavenger Mn(III)tetrakis (4-benzoic acid) porphyrin nor the NADPH oxidase inhibitor diphenylene iodonium and the hydrogen peroxide scavenger N-acetyl-l-cysteine modulate the appearance of lactate dehydrogenase and HMGB1 in the cell culture medium after pneumococcal challenge. Taken together, our data suggest that S. pneumoniae-induced HMGB1 leakage may require cell death that is blocked by oxidized-ATP, but independent from P2X7 and inflammasome activation as well as reactive oxygen species production. The mechanisms underlying oxidized-ATP-inhibitable cell death need to be clarified in future studies.

Once released into the extracellular space, HMGB1 can act as a danger-associated molecular pattern; HMGB1 can (i) alert the innate immune system for the initiation of host defence; or (ii) sustain inflammation when danger is still present (Andersson and Tracey, 2011). In this study, we found that HMGB1 antagonism did not modulate the initiation, but the propagation of inflammation during antibiotic therapy. Together with recent findings on strong pro-inflammatory activities of subcapsular pneumococcal fragments (Paterson and Mitchell, 2006; Kadioglu et al., 2008), our observation suggests that the induction of meningitis may primarily be dependent on the presence of exogenous alarm signals (pathogen-associated molecular patterns). When having entered the CSF space, S. pneumoniae can multiply easily as a consequence of a local immunodeficiency, which includes the absence of soluble pattern recognition receptors and the presence of immunosuppressive mediators (Koedel et al., 2010). Bacterial outgrowth results in brain/CSF acidosis (Denziot, 1991), which may trigger some bacteria to undergo autolysis (Pinas et al., 2008). Autolysis releases peptidoglycan, lipoteichoic acid and other pneumococcal constituents into the CSF. Their recognition by host pattern recognition receptors including TLR2 and TLR4 (Koedel et al., 2010) elicits the inflammatory reaction which, however, is inefficient in limiting bacterial outgrowth (Ernst et al., 1983). As a result, pneumococcal growth reaches potentially dangerous levels (>107 cfu/ml) shown by us and others to be cytotoxic in cell cultures (Kim et al., 1995). Even more important, the increasing amounts of pneumococcal pathogen-associated molecular patterns amplify the inflammatory reaction, finally leading to ‘collateral’ damage of host cells and passive HMGB1 release. Extracellular HMGB1 can fuel inflammation, either by acting in complex with or by substituting for pathogen-associated molecular patterns (Bianchi, 2009). In this study, mice were treated with antibiotics to rescue them from death due to overwhelming infection. Antibiotic therapy sterilizes the CSF quickly and is also paralleled by a fast reduction in CSF pneumococcal fragments (within 6–10 h) (Kanegaye et al., 2001; Gerber et al., 2003; Stucki et al., 2007). The elimination of pathogens and pathogen fragments, however, does not bring neutrophilic inflammation to a quick end. This may be, at least in part, explained by our observation that the endogenous danger-associated molecular pattern HMGB1 is released into the CSF during established meningitis and contributes to the persistence of inflammation.

Extracellular HMGB1 can orchestrate an inflammatory response in multiple ways. First, a substantial body of evidence suggests that, once released from cells, HMGB1 can function as a cytokine-like pro-inflammatory mediator (Wang et al., 1999; Abraham et al., 2000). It activates immune and endothelial cells to produce other pro-inflammatory factors like IL1β, IL6 and TNFα (Wang et al., 1999; Andersson et al., 2000; Fiuza et al., 2003; Park et al., 2004). The cytokine-inducing ability of HMGB1 seems to depend on its redox status and its interaction with TLR2 and TLR4 (Park et al., 2004; Yu et al., 2006; Yang et al., 2010). Moreover, HMGB1 can form complexes with exogenous and endogenous pro-inflammatory molecules (like lipopolysaccharide or IL1β), greatly potentiating their cytokine-inducing properties by signalling through the partner molecule receptor (Bianchi, 2009). In this study, the kinetics of HMGB1 release into the CSF distinctively differed from that of inflammatory factors including IL1β, TNFα, TGFβ, CXCL1 and CXCL2. In addition, brain cytokine and chemokine production was neither altered by pre- nor by post-treatment with HMGB1 antagonists. In accordance to the in vivo data, we found that recombinant HMGB1 did not act in synergy with heat killed pneumococci to promote IL1β and IL6 generation by J774.2 macrophages (data not shown). Combined, these data argue against a cytokine-like function of HMGB1 in pneumococcal meningitis as it was demonstrated in other disease models like sepsis (Wang et al., 1999; Yang et al., 2004).

Additional mechanisms by which HMGB1 can promote inflammation is by interfering with granulocyte apoptosis and efferocytosis (Feng et al., 2008; Liu et al., 2008) and by diminishing bacterial killing through RAGE-dependent mechanisms (Tadie et al., 2012). In line with the latter finding is the observation that RAGE-deficient mice exhibited lower bacterial loads in the lung as well as a decreased dissemination of the bacteria to blood and spleen after intranasal pneumococcal inoculation. In addition, RAGE-deficient macrophages were reported to have an improved killing capacity of S. pneumoniae in vitro (van Zoelen et al., 2009a). In this study, we did not detect any differences in brain bacterial titres irrespective of whether or not wild-type mice were treated with HMGB1 antagonists. Bacterial concentrations in the brain were also similar in wild-type and RAGE-deficient mice subjected to pneumococcal meningitis. The lacking effect of HMGB1 inhibitors and RAGE deficiency on pneumococcal spread may be related to the local immunodeficiency within the CSF space that renders neutrophils (nearly) incapable of killing pathogens (Ernst et al., 1983). In additional analyses, we did not observe any differences in the cell composition and the ratio of apoptotic leucocytes in the CSF between mice that received the HMGB1 inhibitor REPS or its vehicle. Taken together, these findings suggest that the faster resolution of inflammation observed in mice treated with HMGB1 antagonists or lacking RAGE is unlikely to be due to increased granulocyte apoptosis and/or pneumococcal killing.

Another conceivable mechanism by which HMGB1 can contribute to the persistence of inflammation is by facilitating recruitment of leucocytes to the site of infection and/or injury. Accumulating evidence indicates that HMGB1, by itself and by acting in complex with chemokines like CXCL12 (Schiraldi et al., 2012), elicits neutrophil and monocyte recruitment in vitro and in vivo (Rouhiainen et al., 2004; Orlova et al., 2007; Schiraldi et al., 2012). The HMGB1-mediated chemotactic response seems to require its interaction with RAGE (Rouhiainen et al., 2004; Orlova et al., 2007; Penzo et al., 2010). HMGB1 can also indirectly contribute to leucocyte recruitment through the induction of chemokine production; this effect requires the engagement of TLR2 and TLR4 (van Zoelen et al., 2009b; Berthelot et al., 2012). In this study, HMGB1 antagonism attenuated CSF pleocytosis (at 48 h after infection) without affecting brain CXCL1 and CXCL2 expression, suggesting a direct effect of HMGB1 on leucocyte recruitment. This hypothesis is strengthened by the following observations: (i) in a Transwell migration assay, murine neutrophils migrated towards recombinant HMGB1 in a RAGE-dependent fashion; (ii) intracisternal injection of recombinant HMGB1 by itself induced CSF pleocytosis in a RAGE- (but not TLR2 and TLR4-) dependent manner; (iii) RAGE deficiency led to similar reduction in CSF pleocytosis as seen with HMGB1 inhibitors; and (iv) HMGB1 antagonism did not further decrease CSF pleocytosis in RAGE-deficient mice. Combined, these findings suggest that HMGB1 may retard the resolution of inflammation in antibiotic-treated pneumococcal meningitis by acting as a chemoattractant for blood leucocytes like neutrophils.

In conclusion, the present study demonstrated that delayed resolution of inflammation, often seen in pneumococcal meningitis, is—at least partly—due to the release of the endogenous alarm signal HMGB1 from stressed or dying cells. In this context, HMGB1 exerts its pro-inflammatory effects through its interaction with RAGE. Interference with HMGB1 release or HMGB1 blockade may be promising targets for adjunctive therapy of pneumococcal disease, since the persistence of inflammation is a major cause of further damage to the brain and thus of an unfavourable disease outcome.

Funding

This work was supported by the German Research Foundation (to HWP/UK: Pf 246/7-1), the Else-Kröner-Fresenius Stiftung (to UK: A84/08 and MK: 2011-A105), and by the American Lebanese Syrian Associated Charities (ALSAC) (to H.H.).

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

We would like to thank Susanne Bierschenk for technical assistance and Monika Pruenster for preparation of bone marrow-derived neutrophils.

Abbreviations
ASC
apoptosis-associated Speck-like protein containing a CARD
RAGE
receptor for advanced glycation end products
REPS
Ringer’s ethyl pyruvate solution
RLS
Ringer’s lactate solution

References

View Abstract