Glycyrrhizin Suppresses HMGB1 Inductions in the Hippocampus and Subsequent Accumulation in Serum of a Kainic Acid-Induced Seizure Mouse Model

Abstract

Glycyrrhizin (GL), a triterpene present in the roots and rhizomes of licorice (Glycyrrhiza glabra), has been shown to have anti-inflammatory and anti-viral effects. In our previous reports, we demonstrated the neuroprotective effects of GL in the postischemic brain and in kainic acid (KA)-induced seizure animal model. In this KA-induced seizure model, the systemic administration of GL 30 min before KA administration significantly suppressed neuronal cell death and markedly suppressed gliosis and proinflammatory marker inductions. In the present study, we showed that high-mobility group box 1 (HMGB1), an endogenous danger signal, was induced in hippocampal CA1 and CA3 regions of the same KA-induced model, and peaked at ~3 h and at 6 days post-KA. HMGB1 was transiently induced in neurons and astrocyte at 3 h post-KA, and it was released from dying neurons and accumulated in serum at 12 h post-KA. Furthermore, after ~4 days of almost undetectable levels in the hippocampus, delayed and marked HMGB1 induction was detected at 6 days post-KA, mainly in astrocytes and endothelial cells, in which HMGB1 was localized in nuclei, and not secreted into serum. Interestingly, GL suppressed HMGB1 inductions in hippocampus and also suppressed its release into serum in KA-treated mice. Since we established previously that GL has anti-inflammatory and anti-excitotoxic effects in this KA-induced seizure model, these results indicate that the neuroprotective effect of GL in the KA-injected mouse brain might be attributable to the inhibitions of HMGB1 induction and release, which in turn, mitigates the inflammatory process.

Introduction

High-mobility group box 1 (HMGB1) is an endogenous danger signal, and when released extracellularly, it evokes inflammatory reactions by activating various immune-related cells, which includes microglia in the case of the brain (Bianchi and Manfredi 2007; Kim et al. 2006; Qiu et al. 2008). In our previous studies, we showed that extracellular HMGB1 released from damaged neurons in the postischemic brain not only activates microglia (Kim et al. 2006), but also induces apoptotic death in neighboring neurons (Kim et al. 2011). In addition to cerebral ischemia, this mediator-like function of HMGB1 with respect to amplifying neuronal damage might be relevant under various neuropathological conditions, such as, epilepsy or spinal cord injury, because they feature delayed damage progresses after acute and massive neuronal death (Rosenzweig and McDonald 2004; Huttunen et al. 2002). In particular, HMGB1 has attracted considerable attention in epilepsy research field, because it is induced and released by dying or activated immune cells in epileptic tissues, and contributes to the etiopathogenesis of seizure by increasing neuronal excitability (Maroso et al. 2010).

Licorice is a natural product that is used to treat liver disease in traditional Chinese medicine, and glycyrrhizin (GL), which is extracted from licorice root, is used in food industry as a flavoring additive (Wang et al. 2011). GL has been reported to have a variety of pharmacological effects (Gu et al. 2002). Cherng et al. (2006) reported that GL has a neuroprotective effect against glutamate-induced excitotoxicity in primary neurons, whereas Kao et al. (2009) reported the neuroprotective effects of GL and 18β-glycyrrhetinic acid (18βGA), a bioactive compound of licorice, in PC12 cells. Recently, it has been shown that GL attenuates rat ischemic spinal cord injury by suppressing inflammatory cytokine induction and HMGB1 secretion (Gong et al. 2012). In a previous study, we also showed that GL efficiently suppressed infarct formation in the postischemic rat brain after middle cerebral artery occlusion (MCAO) and that the inhibition of HMGB1 phosphorylation and secretion is responsible, at least in part, for this neuroprotective effect (Kim et al. 2012). In addition, we more recently reported that GL attenuates kainic acid (KA)-induced neuronal death in the mouse hippocampus, and significantly suppressed the activations of microglia and astrocytes and the productions of proinflammatory markers in the hippocampus (Luo et al. 2013).

In the present study, we investigated when HMGB1 is induced in the mouse hippocampus using the same KA-induced seizure animal model and whether HMGB1 is released to serum. To identify the cell types responsible for the significant increase of HMGB1 in CA1 and CA3 and release into serum, double or triple immunofluorescent staining was conveyed. In addition, we investigated whether GL suppresses HMGB1 inductions in hippocampus and its release from hippocampal cells to extracellular milieu.

Materials and Methods

Animals

Male BALB/c mice (25–30 g) were housed under diurnal lighting conditions and allowed food and tap water ad libitum. This study was conducted in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. In addition, the animal protocol used in this study was reviewed and approved by the INHA University-Institutional Animal Care and Use Committee (INHA-IACUC) (Approval Number INHA-130607-209). Animals were randomly assigned to a KA-treated group (n = 45), a KA plus GL-treated group (n = 27), GL-treated group (n = 12), or a treatment naïve (Sham) control group (n = 19). Animals in treatment naïve control group were injected with the same volume of saline. At the start of the experiment, animals weighed 25–30 g and were 10 weeks old.

Kainic Acid Administration

Intracerebroventricular (i.c.v.) injection of KA into brain was performed as previously described (Cho et al. 2003). Briefly, male BALB/c mice (25–30 g) were anesthetized by intraperitoneal (i.p.) ketamine (0.12 g/kg) and xylazine hydrochloride (0.02 g/kg) injection, and then placed on a stereotaxic apparatus (Stoelting Co, Wood Dale, IL). Saline solution (4 μl) containing KA (0.2 μg) was then injected into the right lateral ventricle (stereotaxic coordinates in mm with reference to bregma were AP, −2.0; ML, −2.9; DV, −3.7; Franklin and Paxinos 1997) of mice at 0.5 μl/min. 5 min after completing the injection, the needle was removed over 3 min to minimize backflow. Mice were kept on a warm pad until awake.

Treatment with GL

GL (Sigma, St. Louis, MO) was dissolved in 0.89 % NaCl and administered (i.p.). In total, 300 μl of GL-containing solution was injected 30 min before KA injection (0.2 μg, i.c.v.).

Sampling of Protein

Mice were sacrificed by cervical dislocation and the hippocampal CA1 and CA3 regions were promptly removed and placed in ice-cold RIPA buffer (50 mM Tris–HCl (pH7.4), 1 % NP-40, 0.25 % sodium-deoxycholate, 150 mM NaCl, and one complete mini protease inhibitor cocktail tablet (Roche, Basel, Switzerland)). After homogenization, lysates were centrifuged at 19,000×g at 4 °C for 15 min, and supernatant liquors were frozen at −20 °C.

Sampling of Serum

Mice were anesthetized and the ventriculus sinister region was identified. About 500 μl of blood was quickly withdrawn from this region using a Vacutainer (BD, Franklin Lakes, NJ). Blood was centrifuged at 870×g for 15 min at 4 °C and supernatant liquors were frozen at −70 °C.

Immunohistochemistry

Mouse brains were fixed with 4 % paraformaldehyde by transcardiac perfusion, post-fixed in the same solution overnight at 4 °C, and incubated with 30 % sucrose overnight. Sections (30 μm) were prepared using a vibratome, and immunologic staining was performed using a standard procedure. Primary antibodies were diluted 1:200 for anti-NeuN (MAB377, Chemicon, Temecula, CA), anti-ionized calcium-binding adaptor molecule-1 (Iba-1) (Wako Pure Chemicals, Osaka, Japan), and anti-CD31 antibodies (Millipore, Billerica, MA), and 1:150 for anti-GFAP antibody (DB Bioscience, San Jose, CA). For double or triple fluorostaining, rhodamine-labeled anti-rabbit IgG (Jackson ImmunoRes Lab, West Grove, PA, USA) was used as a secondary antibody for anti-HMGB1. After washing stained sections with PBS containing 0.1 % Triton X-100, they were incubated with FITC-conjugated anti-mouse IgG (Jackson ImmunoRes Lab) for anti-Neu N, anti-GFAP, and anti-Iba-1 antibodies and with DyLight4TM405-conjugated AffiniPure Goat Anti-Armenian Hamster IgG (Jackson ImmunoResearch, West Grove, PA) for anti-CD31 antibody in PBS for 30 min at room temperature. After rinsing three times for 10 min in PBS, sections were observed using a confocal imaging system (Radiance 200, Bio-Rad, Hertfordshire, UK). The pictures presented are representative of three independent experiments.

Immunoblotting

Initially, proteins (50 μg) were separated in 12 % sodium dodecyl sulfate-polyacrylamide gel, and after blocking membranes so obtained with 5 % non-fat milk for 1 h, they were incubated with primary antibodies diluted 1:1000 for anti-HMGB1 (Abcam, Cambridge, UK), anti-α-tubulin (Cell Signaling, Danvers, MA), and anti-albumin (Abcam, Cambridge, UK) overnight at 4 °C. The next day, membranes were detected using a chemiluminescence kit (Roche, Basel, Switzerland) using anti-rabbit HRP-conjugated secondary antibody (1:2000, Santa Cruz Biotechnology).

Statistical Analysis

Statistical analysis was performed by analysis of variance (ANOVA) followed by the Newman–Keuls test. All data are presented as means ± SEMs, and significance was accepted for p values of <0.05.

Results

HMGB1 Levels in Hippocampal CA1 and CA3 Regions and in Serum After KA Administration

To examine the levels of HMGB1 in KA-induced epileptic BALB/c mice, immunoblot analysis was carried out on whole cell extracts prepared from hippocampal CA1 or CA3 regions at 3 (Fig. 1a), 6, or 12 h or 1, 4, 6, or 8 days after KA injection. In CA1, HMGB1 levels transiently increased at 3 h post-KA (Fig. 1b, c), but then rapidly declined to the undetectable level at 12 h post-KA. This was maintained until 4 days post-KA (Fig. 1b, c), but at 6 days post-KA, HMGB1 levels in CA1 were markedly elevated and returned to the basal level at 8 days (Fig. 1b, c). Similar changes were observed in CA3, that is, a transient HMGB1 increase was observed at 3–6 h post-KA, and this was followed by significant decline for few days and delayed induction at 6 days post-KA (Fig. 1b, c). In contrast to that observed in CA1 and CA3 regions, in the sera of treatment naïve controls, HMGB1 was not detected (Fig. 1d). Interestingly, however, serum HMGB1 accumulation was detected from 6 h post-KA, peaked at 12 h, and then returned to basal level at 1 day post-KA (Fig. 1d). These results suggest that immediate induction and subsequent decline of HMGB1 in CA1 and CA3 after KA administration resulted in the marked accumulation of HMGB1 in serum. However, the delayed induction of HMGB1 at 6 days post-KA was not accompanied by an increase in serum HMGB1. Furthermore, α-tubulin levels in CA1 and CA and albumin levels in serum were unchanged at all time points (Fig. 1b, d).

Fig. 1.

HMGB1 levels in hippocampus and serum after KA administration. HMGB1 levels in the CA1 and CA3 hippocampal regions (a–c) and in serum (d) were determined at 3, 6, and 12 h and 1, 4, 6, and 8 days post-KA injection by immunoblotting. Representative images are presented (b, d). Results are presented as means ± SEMs (n = 3). * and ** or ^# and ^##, significantly different from Sham at p < 0.05 and p < 0.01, respectively

Suppression of KA-Induced HMGB1 Inductions in the Hippocampal CA1 and CA3 Regions and the Accumulation of HMGB1 in Serum by GL

In our previous report, we showed that GL attenuates KA-induced neuronal death in the mouse hippocampus and significantly suppressed the activations of microglia and astrocytes and the productions of proinflammatory markers in the hippocampus (Luo et al. 2013). To examine whether GL suppresses KA-induced HMGB1 induction and secretion in our KA-induced model of epilepsy, 10 or 50 mg/kg of GL was administered (i.p.) 30 min prior to KA. HMGB1 levels were examined at 3, 12 h, and at 6 days post-KA. Transient inductions of HMGB1 in CA1 and CA3 at 3 h post-KA were suppressed by GL (10 or 50 mg/kg) (Fig. 2a), and reductions in HMGB1 levels in CA1 and CA3 at 12 h post-KA were not detected in GL-administered animals (Fig. 2b). As was expected, the accumulation of HMGB1 in serum at 12 h post-KA was almost completely prevented by GL (Fig. 2c). Although HMGB1 was increased higher than the basal level in KA plus GL (50 mg/kg)-administered animal (Fig. 2b), no HMGB1 induction was detected in GL (50 mg/kg)-administered animals in the absence of KA administration (Fig. 2d). These results indicate that GL suppressed acute neuronal damage in the CA1 and CA3 hippocampal regions and reduced subsequent HMGB1 release to serum.

Fig. 2.

Suppressions of KA-induced HMGB1 reductions in the hippocampus and HMGB1 accumulation in serum by GL. GL was administered (10 or 50 mg/kg, i.p.) 30 min prior to KA (0.2 μg, i.c.v.) and levels of HMGB1 in the CA1 and CA3 regions (Fig. 1a) (a, b) and in serum (c) were assessed by immunoblotting at 3 or 12 h after KA administration. d GL was administered (10 or 50 mg/kg, i.p.) and HMGB1 levels in the CA1 and CA3 regions were assessed by immunoblotting at 3 or 12 h after GL administration. Representative images are presented (a–d). Data are presented as the means ± SEMs (n = 3). * and ** or ^# and ^##, significantly different from Sham at p < 0.05 and p < 0.01, respectively

HMGB1 Expressions in the Hippocampal CA1 and CA3 Regions in the Treatment Naïve Brain

To identify the cell types responsible for the significant increase in HMGB1 in CA1 and CA3 at 3 h post-KA and for the decrease in HMGB1 at 12 h post-KA (Fig. 1), double-immunofluorescent staining was carried out with anti-HMGB1 and anti-neuN, -GFAP, or -Iba-1 antibodies. In treatment naïve brains, HMGB1 was present in almost all neurons in CA1 (arrows in Fig. 3b) and CA3 (arrows in Fig. 3f). In addition, HMGB1 immunoreactivity was also detected in NeuN-negative cells (arrowheads in Fig. 3b, f), which were probably astrocytes (long arrows in Fig. 3c, g) or microglia (double arrows in Fig. 3d, h).

Fig. 3.

HMGB1 localization in the hippocampus of the normal brain. Localizations of HMGB1 in neurons, astrocytes, and microglia were determined by double immunostaining with anti-HMGB1 antibody and anti-NeuN (a, b, e, f), anti-GFAP (c, g), or anti-Iba-1 (d, h) antibody, respectively. The images shown were obtained from the CA1 (a–d) and CA3 (e–h) regions. Arrows indicate HMGB1-positive/NeuN-positive cells and arrowheads indicate NeuN-negative/HMGB1-positive cells. Long arrows indicate GFAP-positive/HMGB1-positive cells and double arrows indicate Iba-1-positive/HMGB1-positive cells. b and f are high magnification images of the white boxes in a and e, respectively, and the insets in c, d, g, and h are high magnification pictures of the corresponding white boxes. Images are representative of three independent experiments. The scale bars represent 20 μm in b, f, and insets in c, d, g, and h and 50 μm in other images

Transient Induction of HMGB1 in Neurons and Astrocytes at 3 h Post-KA

At 3 h post-KA, HMGB1 immunoreactivity in neurons in the KA-treated group was slightly increased as compared with treatment naïve control and HMGB1 was localized in cytoplasm as well as in nucleus in CA1 (Fig. 4a, arrows) and CA3 (Fig. 4g, arrows) regions. The cytoplasmic translocation of HMGB1 was more evident in neurons of CA3 regions, in which the total number of NeuN-positive cells appeared to be lower than in the treatment naïve controls. However, marked cytoplasmic accumulation of HMGB1 was obvious in these NeuN-positive cells and the total HMGB1 immunoreactivity seemed to be higher than in treatment naïve control (Fig. 4g, arrows). However, in the GL plus KA-treated group (10 mg/kg, i.p., 30 min prior to KA administration), HMGB1 was detected in nuclei of neurons in CA1 (Fig. 4b) and CA3 (Fig. 4h). At 3 h post-KA, total HMGB1 immunoreactivity was also higher in astrocytes (Fig. 4c, d, i, j) and microglia (Fig. 4e, f, k, l), and in particular, in CA3, it was increased in the cytoplasm of astrocytes (Fig. 4j, long arrow) and microglia (Fig. 4l, double arrow). Results indicate that HMGB1 was significantly accumulated in cytoplasm of neuron and in nucleus and cytoplasm of astrocytes and microglia.

Fig. 4.

HMGB1 localization in hippocampus 3 h after KA administration. GL was administered (50 mg/kg, i.p.) 30 min prior to KA treatment (0.2 μg, i.c.v.). Hippocampal sections were prepared at 3 h post-KA treatment and the localizations of HMGB1 in neurons, astrocytes, and microglia were determined by double immunostaining with anti-HMGB1 antibody and anti-NeuN (a, b, g, h), anti-GFAP (c, d, i, j), or anti-Iba-1 (e, f, k, l) antibody, respectively. The images shown were obtained from the CA1 (a–f) and CA3 (g–l) hippocampal regions (Fig. 1a). Arrows indicate HMGB1-positive/NeuN-positive cells, long arrows indicate GFAP-positive/HMGB1-positive cells, and double arrows indicate Iba-1-positive/HMGB1-positive cells. Images in d, f, j and l are high magnification images of the white boxes in c, e, i, and k, respectively, and insets in a and g are high magnification images of corresponding white boxes. The images shown are representative of three independent experiments. The scale bars present 10 μm in d, 20 μm in f, j, and l, and 50 μm in the other images

Massive HMGB1 Release from Damaged Neurons and Glia

At 12 h after KA administration in the KA-treated group, numbers of NeuN-positive cells were remarkably decreased, and the morphologies of these cells were abnormal in CA1 (Fig. 5a, b) and CA3 (Fig. 5h, i). Furthermore, HMGB1 was rarely detected in NeuN-positive cells in CA1 (Fig. 5a, b, arrows) or CA3 (arrows, Fig. 5h, i). Interestingly, however, a few small HMGB1-positive/NeuN-negative cells were still detected at 12 h after KA administration (Fig. 5b, i, arrowheads), which were probably astrocytes (Fig. 5d, e, k, l, long arrow) and microglia (Fig. 5f, g, m, n, double arrow), and their morphologies were also abnormal. On the other hand, in the KA plus GL group (10 mg/kg, i.p., 30 min prior to KA administration), the numbers and morphologies of NeuN-positive cells were maintained and HMGB1 was detected in almost all neurons in CA1 (Fig. 5c) and CA3 (Fig. 5j) at 12 h post-KA. Together these results indicate that HMGB1 was rapidly released from neurons due to KA-induced acute and massive neuronal death, and suggest that this release contributed to the marked accumulation of HMGB1 in serum at 12 h post-KA observed in the KA-administered group (Fig. 1b).

Fig. 5.

HMGB1 localization in hippocampus 12 h after KA administration. GL was administered (50 mg/kg, i.p.) 30 min prior to KA treatment (0.2 μg, i.c.v.). Hippocampal sections were prepared at 12 h post-KA treatment and the localizations of HMGB1 in neurons, astrocytes, and microglia were determined by double immunostaining with anti-HMGB1 antibody and anti-NeuN (a, b, c, h, i, j), anti-GFAP (d, e, k, l), or anti-Iba-1 (f, g, m, n) antibody, respectively. The images shown were obtained from the CA1 (a–g) and CA3 (h–n) hippocampal regions (Fig. 1a). Arrows indicate NeuN-positive/HMGB1-negative cells, arrowheads indicate NeuN-negative/HMGB1-positive cells, long arrows indicate GFAP-positive/HMGB1-positive cells, and double arrows indicate Iba-1-positive/HMGB1-positive cells. Images in b, e, g, i, l and n are high magnification images of the white boxes in a, d, f, h, k, and m, respectively. The images shown are representative of three independent experiments. The scale bars represent 20 μm in b, e, g, i, l and n and 50 μm in the other images

Delayed Induction of HMGB1 in Astrocytes and Microglia in CA1 and CA3 Regions of the Hippocampus at 6 Days Post-KA and Suppression of These Inductions by GL

At 6 days post-KA in the KA-administered group, NeuN-immunoreactivities had not recovered in CA1 (Fig. 6a, b) or CA3 (Fig. 6i, j), and HMGB1-immunoreactivity was detected mostly in NeuN-negative cells (Fig. 6b, j, arrows). Double-immunofluorescent staining with anti-GFAP and anti-HMGB1 antibodies revealed that HMGB1 was markedly induced in nuclei of activated astrocytes (GFAP-positive) in CA1 and CA3 (Fig. 6d, l). Interestingly, HMGB1 immunoreactivity was found to be closely associated with blood vessels (Fig. 6e, m, arrowheads). Cytoplasmic localization of HMGB1 in astrocytes was also detected, especially in white matter (Fig. 6f, g, n, o, long arrows). In addition to astrocytes, HMGB1 was detected in activated microglia, although levels of activation, and numbers of activated cells were relatively lower (Fig. 6h, p, double arrow) than those of astrocytes. In the KA plus GL-administered group (10 mg/kg, i.p., 30 min prior to KA administration), numbers of NeuN-positive cells were maintained. HMGB1 was detected in almost all neurons and numbers of activated astrocytes were significantly lower in CA1 (Fig. 6c) and CA3 (Fig. 6k) than in the KA-administered group. Therefore, the delayed induction of HMGB1 in CA1 and CA3 observed at 6 days post-KA administration in the KA group was found to occur mainly in activated astrocytes and GL suppressed these inductions.

Fig. 6.

HMGB1 localization in the hippocampus 6 days after KA administration. GL was administered (50 mg/kg, i.p.) 30 min prior to KA treatment (0.2 μg, i.c.v.). Hippocampal sections were prepared at 6 days post-KA treatment and the localizations of HMGB1 in neurons, astrocytes, and microglia were determined by double immunostaining with anti-HMGB1 antibody and anti-NeuN (a, b, c, i, j, k), anti-GFAP (d–g, l–o), or anti-Iba-1 (h, p) antibody, respectively. The images shown were obtained from the CA1 (a–h) and CA3 (i–p) hippocampal regions (Fig. 1a). Arrows indicate NeuN-negative/HMGB1-positive cells, arrowheads indicate GFAP-positive/HMGB1-positive cells (nucleus), long arrows indicate GFAP-positive/HMGB1-positive cells (cytoplasm), and double arrows indicate Iba-1-positive/HMGB1-positive cells. b, g, j, and o are high magnification images of the white boxes in a, f, i, and n, respectively. The images shown are representative of three independent experiments. The scale bars represent 20 μm in b, g, h, j, and o and 50 μm in the other images

Induction of HMGB1 in Endothelial Cells in the CA1 and CA3 Hippocampal Regions at 6 and 8 Days Post-KA

Triple immunofluorescent staining revealed that HMGB1 was induced in nuclei (asterisks) and cytoplasm (double asterisks) of endothelial cells of CA1 and CA3 at 6 days post-KA (Fig. 7a–d). As is shown in Fig. 6e and m, HMGB1 induction in the nuclei of astrocytes associated with blood vessels was evident (Fig. 7a–d, long arrows). HMGB1 induction in endothelial cells was also detected at 8 days post-KA in a manner similar to that observed at 6 days (Fig. 7e–h). These results indicate that the delayed induction of HMGB1 in CA1 and CA3 at 6–8 days post-KA mainly occurred in astrocytes and endothelial cells and only weakly in microglia.

Fig. 7.

HMGB1 localization in endothelial cells 6 and 8 days after KA administration Hippocampal sections were prepared at 6 (a–d) or 8 (e–h) days post-KA treatment and the localizations of HMGB1 in endothelial cells were determined by triple immunostaining with anti-HMGB1, anti-GFAP, and CD31 antibodies. The images shown were obtained from CA1 (a, b, e, f) and CA3 (c, d, g, h) hippocampal regions (Fig. 1a). Long arrows indicate GFAP-positive/HMGB1-positive cells, asterisks indicate CD31-positive/HMGB1-positive cells (nucleus), and double asterisks indicate CD31-positive/HMGB1-positive cells (cytoplasm). b, d, f, and h are high magnification images of the white boxes in a, c, e, and g, respectively. The images shown are representative of three independent experiments. The scale bars present 20 μm in b, d, f, h and 50 μm in the other images

Discussion

The present study shows that HMGB1 was induced in the hippocampus of KA-induced seizure mouse model, and that it peaked twice at 3 h and 6 days post-KA. Furthermore, it shows that a substantial amount of HMGB1 accumulated in serum at 12 h post-KA probably due to massive HMGB1 release due to KA-induced neuronal death. We also found that GL suppressed KA-induced HMGB1 induction and release from neurons and its subsequent accumulation in serum. In a previous study, we found that GL suppresses neuronal death in the hippocampus of a KA-administered mouse model and significantly suppressed NMDA-, KA-, and glutamate-induced neuronal death in primary cortical cultures (Luo et al. 2013). These anti-excitotoxic effects of GL might underlie its suppressions of HMGB1 induction and release into serum, as demonstrated by the present study. Since, GL has also been reported to have anti-inflammatory effects in the postischemic brain, which, in part, derived from the direct suppression of HMGB1 release by inhibiting its phosphorylation (Kim et al. 2012), the results of our previous and present studies suggest that the neuroprotective effect of GL is due to the combined effects of anti-inflammatory and anti-excitotoxic functions.

The transient induction of HMGB1 after proconvulsant injuries has been reported in experimental animals, in which HMGB1 was induced in neurons and astrocytes (Maroso et al. 2010). In the present study, HMGB1 was transiently induced mainly in neurons, released after acute neuronal death, and markedly accumulated in serum at 12 h post-KA (Fig. 1). It was interesting to find that HMGB1 accumulation in serum generated a sharp peak at 12 h post-KA (Fig. 5b), which is probably due to the rapid induction and release of HMGB1 caused by neuronal damage. In contrast to neurons, in which HMGB1 was translocated from nucleus to cytoplasm, HMGB1 was maintained in the nucleus in astrocytes at 3 h post-KA, indicating that neuronal HMGB1, and not astrocyte HMGB1, was responsible for its accumulation in serum at 12 h post-KA. Therefore, it appears that the neuroprotective effect of GL is responsible for its suppression of massive HMGB1 release from neurons immediately after KA administration. Regarding the molecular mechanism underlying the neuroprotective effect of GL, GL-mediated suppressions NF-κB activation, calcium influx, and glutamate-induced neurotoxicity have been reported in primary neuron cultures (Cherng et al. 2006). In addition, carbenoxolon, a synthetic derivative of GL, has been reported to block NMDA receptors and to impair long-term potentiation (Chepkova et al. 2008). In addition to this NMDA receptor-mediated direct anti-excitotoxic effect of GL (Kim et al. 2011), Maroso et al. (2010) reported that HMGB1 has proconvulsant effects mediated by an ifenprodil-sensitive NMDA receptor. Therefore, it is conceivable that the neuroprotective effect of GL is due to a combination of two effects, those are to the direct neuroprotection afforded by the anti-excitotoxic effect of GL and to the suppression of acute HMGB1 release.

Recently, a number of authors have suggested inflammation might be both a consequence and cause of epilepsy (Vezzani et al. 2011). It was reported that inflammation contributes substantially to delayed brain damage after acute injury and detrimentally affects neurological outcome following KA-induced epileptic seizures (Kim et al. 2004; Weise et al. 2005; Penkowa et al. 2005). In particular, astrocytes and microglia are activated first and they upregulate proinflammatory cytokines (IL-1β, TNFα, and IL-6) and cytokine receptor (Vezzani and Granata 2005), and subsequently, an inflammatory process is induced not only in astrocytes and microglia, but also in BBB endothelial cells and by peripheral immune cells, which migrate into brain parenchyma. Furthermore, the massive HMGB1 release observed at ~12 h post-KA (Fig. 1d) might play a critical role as a proinflammatory mediator inducing delayed inflammation and as an aggravator of neuronal damage. In this regard, Choi et al. (2011) reported that serum HMGB1 levels were significantly higher in febrile seizure patients than in fever only controls, and that IL-1β levels in serum were significantly correlated with serum HMGB1 levels, suggesting that HMGB1 might contribute to the generation of febrile seizures. In addition, recently, it was reported that blocking HMGB1 using anti-HMGB1 antibody exerts neuroprotective effects during the early phase of KA-induced status epilepticus in juvenile rats (Li et al. 2013).

In the present study, we observed a second wave of HMGB1 induction at 6 days post-KA (Fig. 1) and coincident HMGB1 inductions mainly in astrocytes and endothelial cells (Figs. 6, 7). In particular, HMGB1 was markedly induced in activated astrocytes (mainly in the nucleus), and this delayed induction did not generate a detectable peak in serum. It is of note that this is completely different from that observed in the postischemic rat brain, in which HMGB1 released into CSF and serum, generating an obvious 2nd peak at 6–7 days post-MCAO (Kim et al. 2012; Shin et al. 2013). We confirmed no other delayed peak occurred later, for example, at 10 or 12 days post-KA (data not shown), indicating that in our KA-induced seizure model, HMGB1 remains in the nucleus of astrocytes or is secreted at an undetectable level. Therefore, we speculate that delayed HMGB1 induction in the postischemic brain aggravates inflammation by modulating inflammatory mediators localized in the extracellular space, whereas delayed HMGB1 induction in KA-induced epileptic animals might be involved in repair and recovery-related processes possibly by modulating the gene expressions of astrocytes and endothelial cells. However, HMGB1 translocation from nucleus to cytoplasm of neurons and astrocytes has been reported in other studies on experimental animals (Maroso et al. 2010), which is different from the observations made in the present study. Furthermore, the increased nuclear expression of HMGB1 and its translocation from the nucleus to cytoplasm has been detected in the epileptic tissues of human patients (Maroso et al. 2010). This difference could have been caused by the use of different strains or different seizure activity intensities and requires further study. In addition, the functional significances of the delayed inductions of HMGB1 in CA1 and CA3 also warrant additional study.

The present study shows dynamic changes in HMGB1 expression in the hippocampus of the KA-administered mouse brain, involving two peaks at 3 h and 6 days and a massive accumulation of HMGB1 in serum immediately after 3 h-peak in our KA-induced seizure animal model. The suppression of both acute and delayed HMGB1 inductions in CA1 and CA3 and its accumulation in serum by GL might be attributable to its neuroprotective effect, which suggests that GL has potential therapeutic value for the alleviation of neuronal damage elicited by glutamate-induced excitotoxicity.