High-mobility group box 1 is an important mediator of microglial activation induced by cortical spreading depression

Abstract

Single episodes of cortical spreading depression (CSD) are believed to cause typical migraine aura, whereas clusters of spreading depolarizations have been observed in cerebral ischemia and subarachnoid hemorrhage. We recently demonstrated that the release of high-mobility group box 1 (HMGB1) from cortical neurons after CSD in a rodent model is dependent on the number of CSD episodes, such that only multiple CSD episodes can induce significant HMGB1 release. Here, we report that only multiple CSD inductions caused microglial hypertrophy (activation) accompanied by a greater impact on the transcription activity of the HMGB1 receptor genes, TLR2 and TLR4, while the total number of cortical microglia was not affected. Both an HMGB1-neurtalizing antibody and the HMGB1 inhibitor glycyrrhizin abrogated multiple CSD-induced microglial hypertrophy. Moreover, multiple CSD inductions failed to induce microglial hypertrophy in TLR2/4 double knockout mice. These results strongly implicate the HMGB1–TLR2/4 axis in the activation of microglia following multiple CSD inductions. Increased expression of the lysosomal acid hydrolase cathepsin D was detected in activated microglia by immunostaining, suggesting that lysosomal phagocytic activity may be enhanced in multiple CSD-activated microglia.

Introduction

Cortical spreading depression (CSD), defined as a slowly propagating wave (2–5 mm/min) of near-complete neuronal and glial cell depolarization followed by suppressed electrical activity of cortical neurons, was first discovered in rabbits by Leão. ¹ In rodents, CSD is known to cause a transient rise in cerebral blood flow followed by prolonged hypoperfusion.^1,2 Spreading oligemia and magnetic resonance imaging blood oxygenation level-dependent (BOLD) signal changes that spread at ∼3.5 mm/min have been observed during migraine aura, strongly suggesting CSD as the biological substrate.^3,4 Moreover, the aura percept was consistent with the observed BOLD changes in both space and time. ³ Recent observations using electrocorticography revealed that spreading depolarization occurs in stroke.^5,6 In contrast to CSD, which occurs in well-nourished brain tissue, spreading depolarization is always accompanied by depression of activity except when the brain tissue is already depressed like in ischemic tissue. ¹ Spreading depolarizations were observed in 13 of 18 subarachnoid hemorrhage patients ⁶ and 14 of 16 middle cerebral artery infarction patients. ⁵ In such cases, spreading depolarizations occur in clusters. Clustering of spreading depolarizations is thought to exacerbate tissue damage in stroke by cycling around the peri-infarct zone, thereby contributing to the enlargement of ischemic lesions. ⁷ Single-photon emission computed tomography (SPECT) data indicate that clustering of CSD may also develop in hemiplegic migraine, a special subtype of migraine characterized by prolonged aura with motor weakness.^8,9 Nevertheless, to the best of our knowledge, the pathophysiological difference between single and clustering spreading depression/depolarization has not been fully explored.

We recently reported that the release of high-mobility group box 1 (HMGB1) from cortical neurons increases with the number of CSD inductions in a rodent model. ¹⁰ Therefore, CSD clusters evoke far greater HMGB1 release compared with single events. Although HMGB1 was originally identified as a nonhistone nuclear protein, it also acts as a cytokine when released into the extracellular space. ¹¹ The pathophysiological events downstream of multiple CSD-induced HMGB1 release remain to be elucidated. Receptors for HMGB1 include Toll-like receptors 2 and 4 (TLR2 and TLR4, respectively) and the receptor for advanced glycation end products (RAGE), all of which are expressed by microglia. ¹² Microglia are the resident macrophages of the brain. ¹³ In the quiescent state, microglia extend fine motile processes and monitor alterations in the neural microenvironment. When activated under various pathological conditions, both microglial morphology and function are altered. Activated microglia exhibit enlarged soma and processes. ¹⁴ Functionally, activated microglia produce reactive oxygen species and proinflammatory cytokines, typically for the purpose of combating and eliminating invading pathogens. ¹⁴ However, excessive activation of microglia can culminate in tissue damage.

In the present study, we examined whether multiple CSD inductions activate microglia via the HMGB1–TLR2/4 pathway by employing pharmacological interventions and genetic ablation of the TLR2/4 genes. Following multiple CSD inductions, microglia were activated (hypertrophic) but did not show altered expression of the immune activation marker major histocompatibility antigen type II. Rather, multiple CSDs increased the microglial expression of cathepsin D, a major lysosomal acid hydrolase. Our results identify a possible molecular mechanism controlling the microglial activation and ensuing responses induced by clustering of spreading depolarization/depression in stroke and hemiplegic migraine.

Materials and methods

Experimental animals

Adult male C57BL/6 mice (CLEA Japan Inc., n = 96) were housed in cages with free access to water and food. Sixty mice were used for histological analysis, and the remaining animals were used for quantitative real-time polymerase chain reaction (qRT-PCR). All experimental procedures were approved by the Keio University Institutional Animal Care and Use Committee (authorization no. 15052-0). All experimental procedures were in accordance with the university’s guidelines and the ARRIVE (Animal Research: Reporting In Vivo Experiments) reporting guidelines for the care and use of laboratory animals.

Eighty C57BL/6 mice were randomly divided into eight groups. The control group did not undergo any surgical procedures. In the other groups, skin incision over the site of craniotomy and installation of a fixation bar on the head for a stereotaxic apparatus were performed 7 days prior to the CSD inductions to minimize the effects of surgical tissue injury. Sham-operated mice underwent the surgery only, while CSD was induced once or five times in the remaining seven groups as described below. The two groups subjected to a single CSD induction were sacrificed 3 or 24 h later (CSD1x-3 h and CSD1x-24 h groups, respectively). The four groups subjected to five CSD inductions were sacrificed at either 30 min or 3, 24, or 72 h after the final CSD induction (CSD5x-30 m, CSD5x-3 h, CSD5x-24 h, and CSD5x-72 h groups, respectively). Sham-operated mice were sacrificed 1 week later.

Sixteen C57BL/6 mice received pharmacological interventions for inhibiting HMGB1 activity before five CSD inductions. Six mice were given glycyrrhizin (GL, 600 mg/kg; Tokyo Chemical Industry, Tokyo, Japan) by intraperitoneal injection 15 min prior to the CSD inductions. ¹⁵ In 10 mice, mouse monoclonal anti-HMGB1 antibody was topically administered to the cerebral cortex through two different cranial windows (1 µg to each window; Medical & Biological Laboratories, Nagoya, Japan) 60 min prior to CSD inductions. These pretreated mice were sacrificed 24 h after CSD.

Adult male TLR2 and TLR4 double knockout mice (n = 12) on the C57BL/6 background were also examined. The details of this knockout mouse are described elsewhere.^16,17 The knockout mice were randomly divided into two groups: KO-control (mice without any surgical procedures) and KO-CSD5x-24 h (mice sacrificed 24 h after five CSD inductions). All knockout mice were used for immunohistochemistry.

CSD induction

Mice were anesthetized with isoflurane (1.0% in room air at a flow rate of 400 mL/min), which proved not to affect the occurrence of CSD. ¹⁸ The mice were fixed in a stereotaxic apparatus. During all procedures, rectal temperature of the animals was maintained at 37℃. Craniotomy was performed using a dental drill (Tas-35XL, Shofu, Kyoto, Japan). To measure direct current (DC) potentials, an Ag/AgCl electrode (tip diameter = 200 µm, EEG-5002Ag; Bioresearch Center Co., Nagoya, Japan) was inserted above the dura mater (2-mm posterior and 2-mm lateral to bregma). Ag/AgCl reference electrodes (EER-5004Ag; Bioresearch Center Co.) were placed in the subcutaneous tissue ipsilateral to the CSD induction side. DC potentials were amplified at 1–100 Hz and digitized at 1 kHz with a differential head stage and differential extracellular amplifier (Models 4002 and EX1; Dagan Co., Minneapolis, MN, USA). After craniotomy, we installed a small open cranial window with a diameter of 0.5 mm at a site distinct from that of CSD inductions (coordinate: 4-mm posterior and 2-mm lateral to bregma). After the cerebral cortex was exposed, a cup-shaped plastic tube was inserted into the cranial window and CSD was induced by applying 1-M KCl solution to the cortical surface. The cup-shaped plastic tube was used to prevent diffusion of the KCl solution to the dura mater from causing a nociceptive effect. The induction of CSD was verified by the appearance of typical deflections of DC potentials. We regulated the number of CSD inductions by washing out the KCl solution with physiological saline at an appropriate time point. Five CSD inductions could be achieved within 30 min. ¹⁰ We measured the amplitudes and durations of DC potential deflections in all the animals examined and confirmed that there were no differences in these parameters among the animal groups. In a subset of animals in each group undergoing five CSD inductions (n = 3 in each group), arterial blood analysis to measure PaO₂, PCO₂, and pH and blood pressure monitoring were carried out. Arterial blood was analyzed before and after the induction of CSD. ¹⁹ Besides, regional cerebral blood flow was continuously monitored employing a laser Doppler flowmeter (ALF 21; Advance Co., Ltd., Tokyo, Japan). We did not find any difference in these parameters in the animals examined.

Tissue preparation

Under deep anesthesia by excess halothane (Fluothane; Takeda Pharmaceutical Company, Osaka, Japan), the animals were transcardially perfused with 4% paraformaldehyde in 0.1-M phosphate buffer, pH 7.0. Immediately after perfusion fixation, the brain was dissected out, immersed in the same fixative for 4 h at 4℃, and then kept in 0.01-M phosphate-buffered saline (PBS) solution containing 30% sucrose (w/v) for cryoprotection. The region of cerebral cortex located approximately 1.5-mm away from the site of CSD induction and 2.5-mm posterior to bregma was studied. Brains were embedded in Tissue TEK (Sakura Finetek, Torrance, CA, USA) and frozen in liquid nitrogen. Serial sections of 10 -µm thickness were prepared on a cryostat (Leica CM 3050S; Leica Biosystems, Nussloch, Germany) in the horizontal plane along the long axis.^10,20

Identification of cortical microglia and immunostaining

The sections were blocked by incubation in 10% normal donkey serum/0.1-M phosphate buffer for 30 min. To identify cortical microglia, sections were incubated with rabbit antisera against Iba1 (1:500; code 019-19741; Wako Pure Chemical Industries, Osaka, Japan), a well-established marker of microglia, for 24 h at room temperature. We also conducted immunostaining employing mouse antisera against major histocompatibility complex (MHC) class II antigen (1:500; MCA46GA; AbD Serotec, Raleigh, NC, USA) and goat antisera against cathepsin D (1:500; sc-6486; Santa Cruz Biotechnology, Santa Cruz, Dallas, TX, USA). After washing with 0.01-M PBS, the sections were incubated with species-specific fluorophore-labeled secondary antibodies for 2 h at room temperature. After rinsing with 0.01 -M PBS, the sections were cover-slipped in mounting medium (buffered glycerol, pH 8.6). Immunoreactivity was visualized using species-specific donkey secondary antibodies conjugated to fluorescein isothiocyanate (Jackson Immunoresearch Laboratories, West Grove, PA, USA). Nuclei were counterstained with 4′,6-diamidino-2-phenylinodole (DAPI). The immunolabeled specimens were assessed by an examiner blinded to the group allocations of the samples under a Keyence BIOREVO BZ-9000 microscope (Keyence, Osaka, Japan) and a TCS-SP5 confocal laser scanning microscope (Leica Microsystems, Mannheim, Germany). ¹⁰ Within the CSD-treated cerebral cortex, six regions of interest with a total area of 40,000 µm² were assigned, three each in the superficial and deep layers. Three animals were examined for each time point. The number of Iba1-positive cells per mm² in these regions of interest was designated as the representative density value for the animal. Additionally, the somal area of each Iba1-positive cell was calculated using Adobe Photoshop CS6 (Adobe systems, Inc., San Jose, CA, USA).

Terminal deoxynucleotidyl transferase dUTP nick end labeling assay

To detect apoptotic cells, we conducted terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL). Apoptotic cells are characterized by fragmentation of genomic DNA. Tissue sections were treated with the TUNEL reaction mixture containing terminal deoxynucleotidyl transferase (TdT) and fluorescein-labeled dUTP, which allows free fluorescein-labeled dUTP to attach with free 3′OH ends of DNA (Roche Diagnostics, Mannheim, Germany). Subsequently, the incorporated fluorescein was detected with an antifluorescein antibody conjugated to peroxidase. Peroxidase activity was visualized by a substrate reaction (Vector Laboratory, Burlingame, CA, USA). Stained tissue sections were observed under a light microscope. Apoptotic cells were identified by distinct peroxidase-positive nuclei.

qRT-PCR analysis

We dissected out the cerebral cortex subjected to CSD, and the collected tissue was quickly rinsed with 0.01-M PBS. Two aliquots of brain tissue lysate were prepared from the cerebral cortex subjected to CSD in each animal. Total RNA was prepared using RNeasy Plus (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. After quantification of total RNA by spectrophotometry, an aliquot of total RNA (100–500 ng) was used for cDNA synthesis employing the SuperScript III RT First-Strand Synthesis System (Invitrogen, Carlsbad, CA, USA). Quantitative fast real-time PCR was performed with the PikoReal and DyNAmo Flash SYBR Green qPCR Kit (Thermo Fisher Scientific, Waltham, MA, USA). Each reaction was run in duplicate. The resultant PCR products were sequenced for verification. The expression levels of TLR2, TLR4, RAGE, IRAK4, and MyD88 transcripts were quantified by the ΔΔCt method using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) transcript level (run in parallel) as the internal control. The primer pairs used for the analysis of mRNA expression and sequencing are listed in Supplementary Material Table 1.

Statistical analysis

Data are presented as mean ± standard deviation (SD). Group means were compared by one-way analysis of variance (ANOVA) with Bonferroni’s correction. We used IBM SPSS, ver. 22 (Chicago, IL, USA) for statistical analysis. Statistical significance was set at p < 0.05.

Results

Morphological changes in microglia after CSD

The control cerebral cortex exhibited a mesh-like distribution of Iba1-positive cells with small cytoplasmic volumes and fine processes, consistent with the morphological features of resident microglia. As depicted in Figure 1(a), there were no significant differences in the density of cortical Iba1-positive microglia among the eight treatment groups (n = 6 in each group, control: 101.5 ± 13.9/mm², sham: 99.7 ± 16.5/mm², CSD1x-3 h: 95.4 ± 27.5/mm², CSD1x-24 h: 97.8 ± 20.8/mm², CSD5x-30 m: 98.4 ± 6.3/mm², CSD5x-3 h: 100.9 ± 17.3/mm², CSD5x-24 h: 112.5 ± 10.5/mm², CSD5x-72 h: 105.8 ± 21.3/mm²). In general, the cell number is determined by the balance between proliferation and death. There is experimental and clinical evidence showing that apoptosis of microglia occurs in a variety of disease conditions.^21–24 However, the TUNEL assay revealed no obvious microglial apoptosis (data not shown). Although we did not observe any significant change in the microglial number, there were morphological changes in the microglia at 24 h after multiple CSD inductions (Figure 1(b)). Relative to the microglia in the control mice, those in the mice subjected to five CSD inductions had conspicuously larger cytoplasmic volume and thickened processes. The mean plus one SD somal size in the control animals (n = 6) was approximately 70 µm², so we defined Iba1-positive cells with somal size greater than 70 µm² as enlarged microglia.

Figure 1.

Multiple CSD-induced microglial activation in the cerebral cortex is dependent on HMGB1 activity. (a) The total number of Iba1-positive microglia in the cerebral cortex. n = 6 in each group. (b) Immunostaining of the cerebral cortex for Iba1 in the control, sham, CSD1x-24 h, and CSD5x-24 h groups (the upper row). Nuclear staining was performed with DAPI (the middle row). Merged images are shown in the lower row. Bar: 10 µm. (c) The number of enlarged Iba1-positive microglia. (d) The proportion of microglial somal area to the entire brain tissue area. Values are presented as mean ± SD. Statistical analysis was carried out by ANOVA and Bonferroni’s post hoc test.

The number of enlarged microglia significantly increased at 24 h after five CSD inductions (control: 17.7 ± 8.5/mm², sham: 14.1 ± 12.1/mm², CSD5x-24 h: 67.9 ± 17.3/mm²; p < 0.01, vs. control, Figure 1(c)). Significant changes in the number of enlarged microglia were not recognized earlier than 24 h after a single CSD induction (CSD1x-3 h: 21.4 ± 11.5/mm², CSD1x-24 h: 19.0 ± 14.9/mm²) or 30 min, 3 h, or 72 h after five CSD inductions (CSD5x-30 m: 25.1 ± 14.2/mm², CSD5x-3 h 28.1 ± 15.3/mm², CSD5x-72 h 35.5 ± 10.0/mm², Figure 1(c)). Moreover, the proportion of microglial somal area to the entire tissue area was significantly increased after 24 h after five CSD inductions (Figure 1(d)).

The enlarged microglia observed at 24 h after five CSD inductions exhibited greater immunostaining for cathepsin D than the microglia in the control cortex (Figure 2). Meanwhile, immunoexpression of MHC class II antigen was observed only in a small fraction of the enlarged microglia (<1%, data not shown).

Figure 2.

Upregulation of cathepsin D expression in hypertrophic microglia induced by multiple CSDs. Representative microglia immunostained for Iba1 (green) and cathepsin D (red) in the cerebral cortex of the control and CSD5x-24 h group mice. Granular immunostaining patterns of cathepsin D, consistent with intracellular lysosomal localization, are indicated by arrowheads. Bar: 10 µm.

Effects of HMGB1 inhibition on the morphological changes of microglia after CSD

We previously showed that CSD causes HMGB1 release from neurons in a manner dependent on the number of CSD inductions, ¹⁰ and this appeared to temporally correlate with the hypertrophic changes observed in the present study. To explore a possible causal link between HMGB1 release and microglial hypertrophy after CSD, we tested if the inhibition of HMGB1 activity using the HMGB1 inhibitor glycyrrhizin (GL) or a HMGB1-neurtalizing antibody could suppress these morphological changes. Both interventions significantly attenuated the hypertrophic changes of the microglia after CSD (n = 6 in each group, without treatment: 67.9 ± 17.3/mm², GL: 19.6 ± 11.0/mm²; p < 0.01 vs. without treatment; anti-HMGB1 antibody: 18.3 ± 11.1/mm²; p < 0.01 vs. without treatment) without significantly changing the numbers of total Iba1-positive microglia (without treatment: 112.5 ± 10.5/mm², GL: 100.3 ± 18.5/mm², anti-HMGB1 antibody: 99.1 ± 20.5/mm²). The proportion of microglial somal area to the entire tissue area was significantly attenuated by both interventions (Figure 3).

Figure 3.

The effects glycyrrhizin and HMGB1-neutralizing antibody pretreatment on the multiple CSD-induced increase in total number of microglia (a), enlarged Iba1-positive microglia (b), and proportion of microglial somal area to the entire brain tissue area (c). Values are expressed as mean ± SD. Statistical analysis was performed by ANOVA and Bonferroni’s post hoc test.

CSD-induced changes in the expression levels of HMGB1-associated pattern recognition receptors

We examined the expression levels of major HMGB1-associated pattern recognition receptors (PRRs) in the cerebral cortex subjected to CSD by performing qRT-PCR. While TLR2 expression was significantly enhanced irrespective of the number of CSD inductions, the expression level of TLR4 was significantly upregulated only after five CSD inductions. Treatment with the anti-HMGB1 antibody prior to five CSD inductions attenuated TLR4 upregulation, whereas there was no significant change in the expression of TLR2. Both TLR2 and TLR4 are known to associate with adaptor proteins, including MyD88 and IRAK4, during signal transduction. Consistent with upregulated HMGB1–TLR2/4 signaling, the expression levels of MyD88 and IRAK4 were significantly enhanced regardless of the number of CSD inductions (Figure 4, Supplementary Material Table 2). Meanwhile, the RAGE transcript was not detectable in any group (data not shown).

Figure 4.

The effects of single and multiple CSDs on the expression levels of TLR2 (a), TLR4 (b), MyD88 (c), and IRAK4 (d) as assessed by qRT-PCR (normalized to GAPDH expression). Values are presented as mean ± SD. Statistical analysis was carried out by ANOVA and Bonferroni’s post hoc test.

Effect of CSD on the number and morphology of microglia in TLR2/4-deficient mice

To examine the involvement of TLR2 and TLR4 in CSD-induced morphological microglial hypertrophy, we induced multiple CSDs in TLR2/4 double knockout mice. First, we confirmed that there was no difference in the basal density and morphological features of microglia between the wild-type and TLR2/4 double knockout mice. In the TLR2/4 knockout mice, there was no significant change in the total number of Iba1-positive microglia following CSD (KO-control: 96.6 ± 17.0/mm², KO-CSD5x-24 h: 96.0 ± 12.6/mm²), enlarged microglia after CSD (KO-control: 16.5 ± 15.0/mm², KO-CSD5x-24 h: 26.3 ± 11.7/mm²), or the proportion of microglial somal area to the entire tissue area (Figure 5).

Figure 5.

(a) Representative immunostaining for Iba1 and nuclear counterstaining in the TLR2/4 KO mouse cerebral cortex. Bar: 10 µm. The numbers of total (b) and enlarged (c) Iba1-positive microglia in the cerebral cortex of the wild-type and TLR2/4 KO mice in the control and CSD5x-24 h groups. (d) The proportion of microglial somal area to the entire brain tissue area. Values are expressed as mean ± SD. Statistical analysis was carried out by ANOVA and Bonferroni’s post hoc test.

Discussion

Multiple CSD episodes induced microglial hypertrophy, a sign of activation, which was most prominent 24 h after CSD. Concurrently, the expression levels of TLR2 and TLR4 transcripts were elevated. The importance of TLR2 and TLR4 for CSD-induced microglial activation was substantiated by its attenuation in TLR2/4-deficient mice. Moreover, our pharmacological studies indicate that HMGB1, an important ligand of both TLR2 and TLR4, plays a crucial role in inducing these morphological alterations. Collectively, this is the first demonstration that the HMGB1-TLR2/4 axis mediates microglial activation by multiple CSD episodes.

CSD is widely believed to be the neurobiological correlate of migraine aura. ³ In addition, CSD-like spreading depolarizations are observed in brain tissue exposed to ischemia, hypoxia, or subarachnoid hemorrhage, and the occurrence of spreading depolarizations in the peri-infarct area likely contributes to the expansion of infarct size by imposing a bioenergetic burden on vulnerable tissue.^7,25 Migraine aura is clinically characterized by a short-lasting neurological symptom, often a gradually expanding visual field defect accompanied by the appearance of the fortification spectrum. ¹ The duration and pattern of such visual symptoms imply that migraine aura is caused by a single CSD episode, and this notion is supported by the pattern of the BOLD signal changes recorded during the aura phase. ³ In contrast, electrocorticographic data have demonstrated that stroke-associated spreading depolarizations usually occur in clusters.^5,6,26 Hence, an important implication of our data is that CSD-induced microglial hypertrophy is likely to be a pathological event more relevant to stroke than migraine aura.

Currently, little is known about the microglial reaction to CSD. Gehrmann et al. ²⁷ reported that the number of MHC class II antigen-positive microglia significantly increased in the rat cerebral cortex 24 h after multiple CSD inductions. MHC class II antigen is a marker of immunological activity. ²⁸ We did not observe a significant CSD-induced MHC class II antigen upregulation in microglia. This discrepancy may be attributed to the difference in the CSD induction methods. Gehrmann et al. ²⁷ used a higher concentration of KCl solution (4 M) and applied it to the cerebral cortex for a longer period (more than 60 min) than we did to induce CSD. Another source of the discrepancy may be the different species studied. As we utilized a universal microglial marker, Iba1, which is expressed in microglia irrespective of their immunological activation status, ²⁹ we were able to reveal the temporal profile of the changes in the entire cortical microglia population after CSD, thus providing a novel insight into the mode of microglial reaction to CSD. It is generally believed that cytoplasmic hypertrophy is indicative of the microglial activation state. ²⁸ We observed increased expression of cathepsin D, a lysosomal acid hydrolase, in the enlarged microglia, suggesting that multiple CSDs enhance microglial phagocytic activity.

Previously, we demonstrated that CSD induced early HMGB1 release from neurons in a manner dependent on the number of inductions, ¹⁰ which prompted us to explore the involvement of HMGB1 in the development of microglial hypertrophy. Glycyrrhizin, a triterpene glycoside found in licorice root, is commonly used as an anti-inflammatory and antiallergic agent in China and Japan, and is known to inhibit HMGB1 function.^15,30–32 We found that glycyrrhizin and HMGB1-neutralizing antibody abrogated CSD-induced microglial hypertrophy, which strongly suggests that HMGB1 is a crucial mediator in this pathological process. HMGB1 is now recognized as a major member of the damage-associated molecular pattern molecules (DAMPs). ¹² Upon injurious stimuli, DAMPs are released into the extracellular space and initiate innate immune responses by binding to PRRs. HMGB1 is a representative endogenous ligand for several PRRs, including TLR2, TLR4, and RAGE. Of these, only TLR2 and TLR4 exhibited significant changes in expression in our experimental CSD model. TLR2 and TLR4 are expressed primarily in microglia, although the expression of both receptors in neurons and astrocytes has been reported. ³³ Our qRT-PCR data revealed distinct temporal expression profiles of these PRRs (Figure 4). TLR2 appeared to be more sensitive to CSD than TLR4, because a single CSD sufficed to upregulate its expression. Meanwhile, TLR4 expression peaked at 24 h after multiple CSD inductions, the timing of which closely correlated with the emergence of microglial hypertrophy. Multiple CSD-induced microglial activation was suppressed in TLR2/4-deficient mice. Only from our data, it is hard to judge whether the microglia-activating action of HMGB1 requires TLR4 alone or both TLR4 and TLR2. In this regard, it was an intriguing observation that only the upregulation of TLR4 was inhibited by the HMGB1-neutralizing antibody: This seems to raise the possibility that TLR4 expression is induced by HMGB1, with TLR2 upregulation merely concurring with HMGB1 release.

What is the relevance of the HMGB1–TLR2/4 pathway to the pathophysiology of CSD? A large number of studies have demonstrated that HMGB1 is released after cerebral ischemia.^34–36 HMGB1 appears to cause ischemic neural damage by enhancing excitotoxicity, neuroinflammation, and blood–brain barrier permeability, and the importance of the microglia/macrophage HMGB1–TLR4 pathway in the development of ischemia/reperfusion injury has been demonstrated. ³⁷ Notably, Shichita et al. ¹⁷ showed that the HMGB1-mediated inflammatory response is activated in the early stage of ischemic injury (within 6 h after middle cerebral artery occlusion), which overlaps with the early interaction between HMGB1 and TLR2/4 in our experimental CSD condition. Extensive HMGB1 release after a single CSD, which led to dural neurogenic inflammation and the generation of headache, was recently reported. ³⁸ However, our data do not lend support to the idea that HMGB1–TLR2/4 pathway is fully operative in such a paradigm. Activation of the HMGB1–TLR2/4 axis has been reported in other disease conditions.^15,39,40 Microglial activation caused by HMGB1–TLR2/4 signaling in these disease conditions is considered pathogenic. Microglia continuously monitor the extracellular milieu and synaptic activity using their motile processes. ⁴¹ Under pathological conditions, they can exert beneficial actions by secreting neurotrophic factors and restoring neuronal circuits by pruning synapses and axon terminals. ⁴¹ A recent in vivo study demonstrated morphological alterations of axons and dendritic spines after CSD. ⁴² Furthermore, it has been shown that microglial TLR4 regulates the clearance of axon debris, a critical step for the outgrowth of regenerating axons. ⁴³ In this regard, the increased cathepsin D expression in hypertrophic microglia may be relevant to the role of microglia in restoring the integrity of axons and dendritic spines after CSD.

There are several limitations to the present study. There are currently no immunostaining-compatible antibodies specific for TLR2 and TLR4, so the precise cellular localization of these PRRs is unknown. Hence, it remains unsettled whether microglial activation by the HMGB1–TLR2/4 pathway is driven by a cell-autonomous or intercellular signal-requiring process. Second, we induced CSD by applying a high concentration of potassium chloride solution to the cortical surface, which might have directly activated the cortical microglia. However, given the absence of significant morphological changes after single CSD, it appears unlikely that the concentration of potassium chloride used in our experiments alone triggered microglial activation.

In summary, we have clearly shown that CSD-induced microglial activation is dependent on the number of episodes. We also provided evidence that the HMGB1–TLR2/4 signaling pathway is an early mediator of multiple CSD-induced microglial activation. Further work is required to determine whether microglial activation resulting from multiple CSD episodes is pathogenic, neuroprotective, or both depending on context. If deleterious, inhibition of multiple CSD-induced microglial activation by glycyrrhizin, a BBB-permeable agent already in clinical use, could be an effective therapeutic option to prevent CSD-associated neural damage.