High mobility group box 1 protein is released by neural cells upon different stresses and worsens ischemic neurodegeneration in vitro and in vivo


Address correspondence and reprint requests to Alberto Chiarugi MD PhD, Department of Pharmacology, University of Florence, Viale Pieraccini 6, 50139 Firenze, Italy. E-mail: suchende@unifi.it


High mobility group proteins are chromatin binding factors with key roles in maintenance of nuclear homeostasis. The evidence indicates that extracellularly released high mobility group box 1 (HMGB1) protein behaves as a cytokine, promoting inflammation and participating to the pathogenesis of several disorders in peripheral organs. In this study, we have investigated the expression levels and relocation dynamics of HMGB1 in neural cells, as well as its neuropathological potential. We report that HMGB1 is released in the culture media of neurons and astrocytes challenged with necrotic but not apoptotic stimuli. Recombinant HMGB1 prompts induction of pro-inflammatory mediators such as inducible nitric oxide synthase (iNOS), cyclooxygenase-2, interleukin-1β, and tumor necrosis factor α, and increases excitotoxic as well as ischemic neuronal death in vitro. Dexamethasone reduces HMGB1 dependent immune glia activation, having no effect on the protein’s neurotoxic effects. HMGB1 is expressed in the nucleus of neurons and astrocytes of the mouse brain, and promptly (1 h) translocates into the cytoplasm of neurons within the ischemic brain. Brain microinjection of HMGB1 increases the transcript levels of pro-inflammatory mediators and sensitizes the tissue to the ischemic injury. Together, data underscore the neuropathological role of nuclear HMGB1, and point to the protein as a mediator of post-ischemic brain damage.

Abbreviations used

bovine serum albumin




high mobility group box 1




inducible nitric oxide synthase


lactate dehydrogenase




oxygen glucose deprivation


phosphate-buffered saline


tumor necrosis factor

High mobility group box 1 (HMGB1) protein is a 215 amino acid nuclear protein recently emerged as a signaling factor with key roles in cell proliferation, differentiation as well as disease pathogenesis. HMGB1 is also known as amphoterin because of the presence of two basic DNA-binding domains called box A and B, and an acidic C-terminal tail (Landsman and Bustin 1993). The protein binds to the minor groove of linear DNA (Yu et al. 1977) and induces conformational changes of chromatin architecture, thereby facilitating assembly of supramolecular nucleoprotein complexes and transcription factor recruitment. Indeed, the evidence demonstrates that HMGB1 is a key regulator of enhanceosome organization and overall activation of the basal transcriptional machinery (Allain et al. 1999; Ellwood et al. 2000; Verrijdt et al. 2002).

In addition to its role in the regulation of nuclear homeostasis, the protein is also endowed with extracellular signaling functions by interacting with different receptors such as receptor for advanced glycation end products or toll-like receptor-2 and -4 on the plasma membrane of various cell types. Specifically, nuclear HMGB1 can be released into the extracellular space either passively from cells undergoing necrosis or actively from numerous cell types (Bianchi 2004). Conversely, HMGB1 is tightly associated with condensed chromatin during apoptosis (Scaffidi et al. 2002a). An explosion of interest in the pathophysiological implication of HMGB1 release occurred when immune cells were identified as a major source of extracellular HMGB1. Additional evidence shows that the cells of the immune system are highly sensitive to extracellular HMGB1 (Lotze and Tracey 2005) contributed to increase interest in HMGB1 signaling. It was indeed hypothesized that the capability of the protein of activating immune cells, and that of immune cells of releasing HMGB1 could establish a pro-inflammatory vicious circle sustaining the immune response and participating to tissue damage. Accordingly, extracellular HMGB1 contributes to the pathogenesis of disorders of the liver, lung, gut, joints as well as septic shock and cancer (Dumitriu et al. 2005; Lotze and Tracey 2005; Ulloa and Messmer 2006). Given its roles in the immune response, HMGB1 has been recently included into the ‘alarmin’ family, a group of endogenous factors released into the extracellular space and activating the inflammatory response through the engagement of membrane receptors (Oppenheim and Yang 2005; Bianchi 2007).

Several studies highlight a role of HMGB1 in CNS development (Guazzi et al. 2003; Chou et al. 2004). Also, the ability of HMGB1 to functionally link the immune and endocrine system at the hypothalamic level has been proposed. Specifically, pro-inflammatory cytokines trigger release of HMGB1 from pituicytes (Wang et al. 1999). HMGB1, in turn, increases brain levels of tumor necrosis factor (TNF) α and interleukin (IL)-1β, induces anorexia and the loss of body weight in the mouse, as well as fever and allodynia in the rat (Agnello et al. 2002; O’Connor et al. 2003). Interestingly, astrocytes challenged with different stimuli release HMGB1 (Passalacqua et al. 1998). These findings taken together point to HMGB1 as a key regulator of the neuroimmune system and suggest its involvement in neurological disorders. In this regard, a large body of information indicates that factors involved in delayed inflammatory reaction are of significance to progression of ischemic brain injury (Barone and Feuerstain 1999; Iadecola and Alexander 2001). Yet, triggers of the ischemic neuroinflammatory response still wait to be clearly identified at the molecular level (Dirnagl et al. 1999; Lo et al. 2003). Also, the neurotoxic potential of intracellular factors released form neurons undergoing ischemic cell death remains in large part to be understood. In this study, we have investigated whether HMGB1 is released by neural cells upon exposure to different stresses and contributes to neuroinflammation and post-ischemic brain damage.

Material and methods

Neural cell cultures

Pure neuronal cultures were prepared by seeding cortical cells obtained from 16-day-old mouse embryos as previously described (Chiarugi 2002). Neurons were cultured in Neurobasal™ medium with B-27 supplement (GIBCO, Rockville, MD, USA) and 0.5 mmol/L glutamine onto poly-d-lysine-coated multiwell plates. Cells were used at 7 days in vitro from preparation. Cell damage was quantified by measuring the amount of lactate dehydrogenase (LDH) released from injured cells into the incubating media as previously reported (Koh and Choi 1987). Evaluation of nuclear morphology was performed by Hoechst-33258 staining as described (Chiarugi 2002). Primary cultures of mixed glial cells were prepared as described (Chiarugi and Moskowitz 2003) and grown in Dulbecco’s modified Eagle’s medium + 10% fetal bovine serum. Cells were subcultured in 24 well plates for 48 h before stimulation with 0.3 μg/mL lipopolysaccharides (LPS) or eukaryotic recombinant HMGB1 (Sparatore et al. 1996). PJ34 or trichostatin-A were dissolved in dimethylformamide and added to the culture medium 1 h before activation. Conditioned media were collected for TNFα measurement or evaluation of HMGB1 release, whereas cells were lysed for western blotting. For immunohistochemistry, pure neurons or glial cells were cultured on glass slides. Primary cultures of mixed cortical cells were prepared as previously described (Moroni et al. 2001).

Western blotting

For western blotting, cells were scraped, collected in Eppendorf tubes, centrifuged (1500 g for 5 min at 4°C), and resuspended in lysis buffer (50 mmol/L Tris, pH 7.4, 1 mmol/L EDTA, 1 mmol/L phenylmethylsulfonyl fluoride, 4 μg/mL aprotinin and leupeptin, 1% sodium dodecyl sulfate). Twenty-to-forty micrograms of protein per lane were loaded. After a 4–20% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and blotting, membranes (Hybond ECL; Amersham Biosciences, Buckinghamshire, UK) were blocked with phosphate-buffered saline (PBS) containing 0.1% Tween 20 and 5% skimmed milk (TPBS/5% milk) and then probed overnight with primary antibodies (1 : 1000 in TPBS/5% milk). The polyclonal anti-HMGB1 antibody was from Abcam (Cambridge, UK); the anti-iNOS, anti-cyclooxygenase (COX)-2 and anti-IL-1β were from Santa Cruz Biotechnology (Santa Cruz, Tamecula, CA, USA); the anti-phospho-p38 rabbit polyclonal antibody was from Cell Signaling Technology (Beverly, MA, USA). Membranes were then washed with TPBS and incubated for 1 h in TPBS/5% milk containing the corresponding peroxidase-conjugated secondary antibody (1 : 2000). After washing in TPBS, ECL (Amersham Biosciences) was used to visualize the peroxidase-coated bands.

Quantitative RT-PCR

One microgram of total RNA extracted with Trizol (GIBCO-BRL) from the mouse cortex 24 h after HMGB1 or LPS microinjection was reverse transcribed into DNA and subjected to real-time PCR. Expression levels of iNOS, COX-2, and IL-1β were investigated using real-time quantitative RT-PCR and TaqMan (Applied Biosystems, Foster City, CA, USA, http://www.appliedbiosystems.com). Primers for iNOS, COX-2, and β-actin were obtained from Applied Biosystems’ TaqMan Gene Expression Assay catalog (iNOS, Nos2: cat. # Mm00440485_m1; COX-2, Ptgs2: cat. # Mm00478374_m1; and β-actin: cat. # Mm00607939_s1). Primers and TaqMan Probe for IL-1β gene were designed as previously reported (Rioja et al. 2004) and obtained from Applied Biosystems’ TaqMan Gene Expression Assay. Real-time RT-PCR reactions were performed using TaqMan Universal PCR Master Mix (Applied Biosystems) in a 20-μL reaction volume containing 50 ng of cDNA. All reactions were performed in triplicate and included a negative control. PCR reactions were carried out using an ABI Prism 7500 Sequence Detection System (Applied Biosystems). Cycling conditions were 2 min at 50°C, 10 min at 95°C, and 40 cycles of 15 s at 95°C and 1 min at 60°C. Relative quantification of mRNA levels was determined by the 7500 system software, which only uses the comparative method (ΔΔCT).

TNFα measurement

The amount of TNFα in culture media was measured by ELISA (Bender MedSystems, Burlingame, CA, USA), according to the manufacturer’s instructions.

Oxygen glucose deprivation

For oxygen glucose deprivation (OGD) of primary cultures of mixed cortical cells culture medium was replaced by a glucose-free balanced salt solution saturated with 95% N2/5% CO2 and heated to 37°C. Multiwells were then sealed into an airtight incubation chamber equipped with inlet and outlet valves and 95% N2/5% CO2 was blown through the chamber for 10 min to ensure maximal removal of oxygen. The chamber was then sealed and placed into the incubator at 37°C for 30 min. OGD was terminated by removing the cultures from the chamber, replacing the exposure solution with oxygenated medium and returning the multiwells to the incubator under normoxic conditions. The extent of neuronal death was assessed 24 h later. In this system, 30 min OGD induced a neuronal damage that was approximately 40% of the maximal degree of neuronal injury achieved by exposing the cultures for 24 h to 1 mmol/L glutamate. OGD-induced cell injury was evaluated by propidium Iodide staining and quantified by optical density analysis of immunostainings using NIH Image software and expressed as the mean of four fields per slide of three different experiments. OGD neurotoxicity was also quantified by measuring the amount of LDH released from injured cells into culture media 24 h following exposure to OGD, as previously described (Koh and Choi 1987). The LDH level corresponding to complete neuronal death (with no glial death) was determined for each experiment by assaying sister cultures exposed to 1 mmol/L glutamate for 24 h. Background LDH release was determined in control cultures not exposed to OGD and subtracted from all experimental values. The resulting value correlated linearly with the degree of cell loss estimated by observation of cultures under phase-contrast microscopy.

Immunocytochemistry, immunohistochemistry, and image deconvolution

Neuronal or glial cultures grown on coverslips and challenged with different stressors were washed twice with PBS and fixed with cold PBS/4% paraformaldehyde for 10 min. Cells were incubated 1 h in PBS/0.1% Triton-X-100/20 mg/mL bovine serum albumin (BSA) at 20°C, rinsed in PBS and incubated 1 h with a polyclonal anti HMGB1 antibody (Santa Cruz Biotechnology) dissolved 1 : 500 in PBS/5 mg/mL BSA. After washing with PBS, slides were incubated 1 h with a Cy-3 conjugated anti-rabbit antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) dissolved 1 : 500 in PBS/5 mg/mL BSA. Fluorescence was examined using a rhodamine filter (Ex 520, Em >580). Specificity of immune-detection was established by the omission of the primary antibody.

For immunohistochemistry, mice were perfused with cold PBS and then PBS + 4% paraformaldehyde. Brains were collected and post-fixed overnight in PBS + 4% and cryoprotected with PBS + 20% sucrose for 24 h. Later on, brain sections of 20 μm were cut with a cryostat and used for free-floating immunohistochemistry or H&E staining for infarct volume determination. Free-floating immunohistochemistry was conducted in wells of a 24-well plate. Briefly, brain sections were pre-incubated 1 h with PBST/10% HS and then overnight with the anti-HMGB1 antibody (see above, 1 : 500 in PBST/2% HS). The binding was visualized by means of a Cy3-conjugated anti-rabbit secondary antibody. Imaging was performed with a Nikon microscope equipped with piezoelectric motorization and a CCD camera. For 3D imaging, stacks of images were acquired through the depth of the section using Metamorph/Metafluor software, and deconvolved using Image Autodeblur software as previously described (Cipriani et al. 2005). To better visualize the decrease of HMGB1 staining in the ischemic brain, the contrast has been increased to the same extent during image acquisition of contralateral and ischemic regions.

Animals and intracerebral microinjection of HMGB1

Male C57 mice (25–27 g, Harlan, Italy) were housed five per cage and kept at constant temperature (21 ± 1°C) and relative humidity (60%) with regular light/dark schedule (7 am–7 pm). Food and water were available ad libitum. A Procedure involving animals and their care were conducted in compliance with the Italian guidelines for animal care (DL 116/92) in application of the European Communities Council Directive (86/609/EEC) and was formally approved by the Animal Care Committee of the Department of Pharmacology of the University of Florence.

One microliter saline containing 0.5 μg LPS or different amounts of HMGB1 or vehicle (50 mmol/L sodium borate pH 8.5, 0.15 mol/L NaCl and 10 mmol/L dithiothreitol) was injected (0.2 μL/min) in the parietal cortex on the left of the middle cerebral artery of anesthetized mice. One group of mice was killed 24 later to evaluate the inflammatory reaction by western blotting. Samples were collected from the parietal cortex of mouse brain slices (20 μm) obtained with the cryostat. Another group of mice receiving HMGB1 was immediately subjected to distal middle cerebral artery occlusion (see below).

Surgical procedure and measurement of infarct volume

Permanent distal middle cerebral artery occlusion was induced in age-matched male C57 mice (n = 8, per group). Animals (25–30 g) were anesthetized with 4% isoflurane and maintained on 1.5% isoflurane in air. Rectal temperature was monitored and maintained between 36.5 and 37.5°C with a homeothermic blanket. A 1-cm vertical scalp incision was made between the right eye and ear. The temporalis muscle was bisected and a 2-mm burr hole was made at the junction of the zygomatic arch and squamous bone. The distal middle cerebral artery was exposed and permanently occluded by cauterization above the rhinal fissure. In randomly selected animals, the left femoral artery was cannulated with a PE-10 polyethylene tube for arterial blood pressure measurement and blood gas determination. Arterial blood samples (50 μL) were analyzed for pH, arterial oxygen pressure (PaO2) and partial pressure of carbon dioxide (PaCO2) using a Ciba-Cornig 248 PH/blood gas analyzer (Ciba-Corning Diagnostics, Corp. Medfield, Norwood, MA, USA). Physiological parameters such as rectal temperature, mean arterial blood pressure, pH, PaO2 as well as PaCO2 did not differ between groups before, during and 1 h after ischemia. Also, as revealed by a flexible skull probe connected to a Laser-Doppler (PF2B; Perimed, Stockholm, Sweden), drug treatment did not affect regional cerebral blood flow (rCBF) in control brain tissue, as well as regional cerebral blood flow drop upon artery cauterization. After surgery, mice were kept at 37°C for 1 h in an incubator and then placed in their cage until sacrifice. Mice were killed after 24 h and their brains snap frozen in N2 vapor for cryostat sectioning. For infarct determination, H&E stained coronal sections (20 μm) were imaged by using the Image 3.0 ProPlus analysis software (Faraco et al. 2006). Twelve sections per animal were analyzed and infarct areas were calculated subtracting the area of intact tissue in the ipsilateral hemisphere from the area of the contralateral hemisphere to minimize the error that is introduced by edema, which distorts and enlarges the infarcted tissue and surrounding white matter. Infarct volumes were calculated multiplying the infarct area by the distance among sections as previously described (Swanson and Sharp 1994).


HMGB1 is present in cultured neural cells and released upon different stresses

We first evaluated the expression levels and cellular distribution of HMGB1 in cultures of pure cortical neurons or astrocytes. As shown in Fig. 1a and c, HMGB1 immunoreactivity was present in the nucleus and absent in the cytoplasm of cortical neurons. When the anti-HMGB1 antibody was pre-adsorbed with pure recombinant HMGB1, nuclear fluorescence was almost undetectable, indicating specificity of immunoreactivity. Conversely, HMGB1 was localized in the cytoplasm of neurons exposed to different necrotic stimuli such as chemical ischemia (6 mmol/L deoxy-glucose + 10 mmol/L sodium azide), oxidative stress (1 mmol/L hydrogen peroxide) and excitotoxicity (1 mmol/L glutamate). The cytoplasm accumulation of HMGB1 was time dependent and occurred as early as 1 h after exposure to the different stresses. In striking contrast with this, neurons challenged with 1 μmol/L staurosporine, a prototypical inducer of apoptosis, retained HMGB1 strictly confined into the nucleus up to 8 h after exposure to the chemical. Accordingly, neurons exposed to apoptotic concentrations (100 μmol/L) of the alkylating agent methyl-nitrosoguanidine (Yu et al. 2002; Cipriani et al. 2005) retained HMGB1 in the nucleus, whereas when the concentration of methyl-nitrosoguanidine was raised up to 1 mmol/L [which reportedly induces necrosis (Ha and Snyder 1999)] HMGB1 was released in the cytoplasm of neurons (Fig. 1a). At variance with neurons, HMGB1 immunoreactivity in control astrocytes, although highly represented in the nucleus (Fig. 1c), was also present in the cytoplasm as small vesicle-like bodies (Fig. 1b). This is in line with the endolysosome pathway extruding HMGB1 in immune cells (Gardella et al. 2002). The number of HMGB1-positive cytoplasmic granules was highly increased in astrocytes exposed to deoxy-glucose + sodium azide (Fig. 1b) or hydrogen peroxide (not shown). Again, staurosporine did not affect cellular distribution of HMGB1 in astrocytes (Fig. 1b). Taken together, these findings suggested that HMGB1 is released from the nucleus to the cytoplasm in both neurons and astrocytes undergoing necrosis but not apoptosis, and that nuclear release of HMGB1 in astrocytes occurs through a route which might involve endolysosomes. We also investigated whether under these conditions the protein was released extracellularly. Western blotting shows that HMGB1 was present as a double band (30 and around 25 kDa) in the electrophoresed media of cultures of pure cortical neurons exposed for 4 h to the necrotic stimulus deoxy-glucose + sodium azide, but absent in conditioned media of neurons challenged with staurosporine (Fig. 1c). In keeping with the constitutive presence of HMGB1-positive granules in cultured astrocytes, the protein was detectable as a single 30 kDa band in media of these cells and increased in media of astrocytes exposed 4 h to deoxy-glucose + sodium azide. Staurosporine did not alter the levels of HMGB1 in media of astrocytes (Fig. 1c). Extracellular release of HMGB1 during necrosis is passive and not regulated by acetylation (Scaffidi et al. 2002b; Bonaldi et al. 2003). Accordingly, the histone deacetylase inhibitor trichostatin-A did not affect the amount of HMGB1 present in media of neurons or astrocytes exposed to 6 mmol/L deoxy-glucose + 10 mmol/L sodium azide (Fig. 1c). Given that HMG proteins are poly(ADP-ribosyl)ated (Giancotti et al. 1996), and formation of poly(ADP-ribose) massively occurs during necrosis (Ha and Snyder 1999), we investigated the effect of PJ34, a potent inhibitor of poly(ADP-ribose) formation (Abdelkarim et al. 2001), on necrotic extracellular release of HMGB1. As shown in Fig. 1c, PJ34 had no effect on the protein contents in the media of both neurons and astrocytes, indicating that poly(ADP-ribosyl)ation does not regulate HMGB1 release during necrosis.

Figure 1.

 Nuclear release of high mobility group box 1 (HMGB1) from cultured mouse neurons or astrocytes exposed to different stimuli. (a) HMGB1 is confined in the nucleus of pure cultured neurons. The protein relocates into the cytoplasm upon 1 h exposure to necrotic stimuli such as 6 mmol/L deoxy-glucose + 10 mmol/L sodium azide (NaAz + DeGluc), 1 mmol/L H2O2 or 1 mmol/L glutamate (Glu), but not after a challenge with the apoptotic inducer staurosporine (STP, 1 μmol/L/8 h). Higher magnification (NaAz + DeGluc 60×) of necrotic neurons demonstrates that cytoplasmic HMGB1 staining is almost homogeneous. (b) HMGB1 is expressed into the nucleus of cultured resting astrocytes but also present into their cytoplasm as vesicle-like structures. Exposure to 6 mmol/L deoxy-glucose + 10 mmol/L sodium azide (NaAz + DeGluc) but not to staurosporine (STP) triggers massive HMGB1 relocation into the cytoplasm. Note that, at variance with neurons, cytoplasmatic HMGB1 in necrotic astrocytes maintains a granular distribution. (c) Colocalization of HMGB1 immunoreactivity with Hoechst in control neurons or astrocytes. (d-i) Western blotting analysis of HMGB1 in the medium of neurons or astrocytes in culture. The protein is constitutively present in the media of astrocytes but not neurons. Four hours after exposure to 6 mmol/L deoxy-glucose + 10 mmol/L sodium azide (NaAz + DeGluc) but not staurosporine (STP, 1 μmol/L) the protein is present in the culture media of both cell types. The histone deacetylase inhibitor TSA and the poly(ADP-ribose) polymerase inhibitor PJ34 do not affect extracellular release of HMGB1 form both neurons and astrocytes. Note that the protein appears as a double or single band when electrophoresed from media of necrotic neurons or astrocytes, respectively. (d-ii) Densitometric analysis of results shown in (d-i). One experiment representative of four (a, b) and three (c) is shown. Bar = 30 μm (a) and 10 μm (b and c).

HMGB1 is induced during glia activation and triggers expression of pro-inflammatory mediators

To understand whether inflammatory glia activation alters the expression levels and localization of HMGB1, cultures of glial cells were exposed to bacterial LPS or recombinant HMGB1 for different times, and HMGB1 expression evaluated by western blotting. The data demonstrates that both stimuli increased HMGB1 expression after 24 h exposure (Fig. 2a), but did not cause relocation of the protein in the cytoplasm or culture media (not shown). We then evaluated the effect of recombinant HMGB1 on inflammatory glia activation. In line with the cytokine functions of HMGB1, the recombinant protein prompted phosphorylation of p38 (Fig. 2b-i), a kinase with pivotal role in the inflammatory response. Also, HMGB1 concentration dependently triggered expression of pro-inflammatory mediators such as iNOS, COX2, and IL-1β in cultured glial cells (Fig. 2b-ii). Recombinant HMGB1 also prompted the release of TNFα in the culture media. Oddly, TNFα concentrations did not increase over time after HMGB1 challenge (Fig. 2b-iii). We speculate that this is due to early inactivation of HMGB1 because of its well-known instability in solution. In addition, this finding indicates that TNF release by recombinant HMGB1 was not due to LPS contamination (2.5 pg/μg of protein; Bianca Sparatore and Marco Pedrazzi, personal communication). Accordingly, when tested at this concentration, cytokine induction by LPS was below detection limit (not shown). Of note, HMGB1-dependent pro-inflammatory activation of glia was reduced by the presence of dexamethasone in the incubating media (Fig. 2c).

Figure 2.

 Immune activation of mouse glia by high mobility group box 1 (HMGB1). (a) Exposure to lipopolysaccharides (LPS) or recombinant HMGB1 time dependently induces HMGB1 expression in cultured glia. (b-i) Both LPS (0.3 μg/mL/1 h) and recombinant HMGB1 (10 μg/mL/1 h) prompt phosphorylation of p-38 kinase. (b-ii and c) Exposure of glial cells to HMGB1 induces expression of iNOS, cyclooxygenase (COX)-2, interleukin (IL)-1β, and tumor necrosis factor (TNF)α in a concentration dependent manner. *< 0.05 versus 6 and 24 h, anova and Dunnett’s post hoc test. LPS is used as a positive control. (c) Dexamethasone (DXE) inhibits the pro-inflammatory effects of HMGB1 on cultured glial cells. One blot/experiment representative of three (a, b) and two (c) is shown. In (b-iii) bars represent the mean ± SEM of three experiments conducted in duplicate.

HMGB1 worsens excitotoxic and ischemic neuronal death in vitro

Given the ability of HMGB1 to trigger immune glia activation and the well-established role of the latter in neurodegeneration, we then investigated the neurotoxic potential of recombinant HMGB1 in vitro. Mixed cortical cells containing both glia and neurons were exposed to different concentrations of HMGB1 and neuronal death evaluated by both propidium Iodide staining and measurement of LDH release. The challenge of mixed cortical cells to a wide range of HMGB1 concentrations (0.01–10 μg/mL) did not induce apparent neurotoxicity up to 48 h exposure (Fig. 3a and not shown). Conversely, HMGB1 increased neuronal death in cultures undergoing 30 μmol/L glutamate-dependent excitotoxicity or subjected to OGD (Fig. 3a–c); 100 μmol/L glutamate triggered massive excitotoxic cell death which was not potentiated by HMGB1 (not shown). Of note, dexamethasone had no effects on the sensitization to glutamate- and OGD-induced neurotoxicity caused by HMGB1 (Fig. 3b and c).

Figure 3.

 High mobility group box 1 (HMGB1) sensitizes to excitotoxic and ischemic neuronal death in vitro. (a) Propidium Iodide (PI) staining of mixed cortical cell cultures exposed to HMGB1 and/or glutamate (Glu) for 24 h. Exposure to a low excitotoxic insult (30 μmol/L glutamate) but not to HMGB1 (10 μg/mL) triggers mild neuronal death. The latter is significantly increased when cultures are exposed to glutamate together with HMGB1. (b) Measurement of lactate dehydrogenase (LDH) released in the media of cultures shown in (a) confirms that HMGB1 sensitizes neurons to excitotoxic cell death. (c) Similarly, LDH quantification in media of cultures exposed to 30 min oxygen glucose deprivation (OGD) indicates that HMGB1 sensitizes to ischemic neuronal demise in vitro. (b and c) Note that dexamethasone does not affect the effect of HMGB1 on glutamate- or OGD-dependent neurotoxicity. Of three experiments one is shown in (a). In (b) and (c) bars represent the mean ± SEM of three experiments conducted in duplicate. *< 0.05 versus control, §< 0.01 versus control and < 0.05 versus Glu (b) or OGD (c), anova and Dunnett’s post hoc test.

HMGB1 is abundantly expressed in the mouse brain and promptly released during cerebral ischemia

As the previous findings indicated that HMGB1 was of relevance to neurodegeneration in vitro, we investigated whether the neurotoxic potential of the protein is also maintained in vivo. Constitutive expression of HMGB1 in the adult rodent brain is controversial (Guazzi et al. 2003; Takata et al. 2004). As shown in Fig. 4a, HMGB1 immunoreactivity was well represented in sections of the mouse brain cortex and colocalized with Hoechst-33258-positive nuclei (Fig. 4a–c). Interestingly, fluorescence deblurring by image deconvolution and 3D reconstruction revealed that HMGB1 immunoreactivity is not homogeneously distributed within the cell nucleus but clustered in numerous foci scattered throughout the nucleoplasm. HMGB1-positive foci did not colocalize with nuclear regions highly stained by Hoechst-33258 typically representing heterochromatic domains (Fig. 4d). Similar expression levels and cellular distribution of HMGB1 were detected in the striatum, hippocampus, brainstem, and cerebellum of the mouse brain. HMGB1 appeared highly expressed in the ependima and subventricular zone (not shown).

Figure 4.

 Immunohistochemical localization of high mobility group box 1 (HMGB1) in the mouse brain. HMGB1 immunofluorescence (a) colocalizes (c) with Hoechst-33258-positive nuclei (b) of neural cells of the mouse brain cortex. (d and inset) Image deconvolution and 3D reconstruction reveals that nuclear HMGB1 is not evenly distributed in the nucleoplasm but concentrated in regions distinct from heterochromatin. (e) HMGB1 (red) is expressed in nuclei of Neu-N (green) positive neurons (arrowheads). The anti-HMGB1 antibody also stains small bodies (arrows) probably representing nuclei of glial cells. (f) Image deconvolution and 3D reconstruction reveals that GFAP-positive (red) astrocytic processes converge to HMGB1 (green) expressing nuclei (arrowhead) smaller than the round ones typical of neurons (arrow). Bar = 40 μm (a–c, e) or 20 μm (d, f).

To understand whether neurons and/or astrocytes express HMGB1, colocalization experiments with specific neuronal and astrocytic markers were performed. We found that numerous, but not all, HMGB1-positive cells also expressed the neuronal marker Neu-N (Fig. 4e). As for the expression of HMGB1 in astrocytes, unambiguous evidence for HMGB1-positive nuclei in glial fibrillary acidic protein-expressing astrocytes was hard to obtain. Image deconvolution and 3D reconstruction, however, allowed to identify glial fibrillary acidic protein-positive astrocytic processes converging to HMGB1-positive nuclei. This finding, along with the evidence that these nuclei were elongated and smaller than the prototypical round-shaped neuronal nuclei (Fig. 4f), suggested that HMGB1 is also expressed in mouse brain astrocytes.

We then investigated the localization of HMGB1 within the ischemic mouse brain. Strikingly, HMGB1 immunoreactivity was significantly reduced in the ischemic area compared with the contralateral one as early as 1 h after the occlusion of the middle cerebral artery. Hoechst-33258 staining revealed that reduction of HMGB1 immunoreactivity was not due to tissue loss or edema (Fig. 5a). Three hours after ischemia, decrease of HMGB1 signal was even more pronounced (not shown). At this time point, HMGB1 was present in the cytoplasm of numerous cells of the ischemic cortex (Fig. 5b-i ). Higher magnification revealed that HMGB1 localizes in structures resembling soma and axons of cortical neurons (Fig. 5b-ii). Conversely, HMGB1 appeared strictly confined into the nucleus of cells in the contralateral cortex (Fig. 5b-iii). Twenty-four hours after ischemia, the immunoreactivity of HMGB1 was further reduced in the injured cortex (not shown) and localized in structures resembling remnants of degenerating neurons (Fig. 5c-i). It is worth noting that at this time point the background of the ischemic cortex was often higher than that of the ipsilateral one (compare Fig. 5c-i with Fig. 5c-ii), suggesting the extracellular presence of HMGB1. A striking difference between HMGB1 immunoreactivity in the ischemic tissue and that in the penumbra was evident 24 h after the insult (Fig. 5c-iii ).

Figure 5.

 High mobility group box 1 (HMGB1) relocation in the ischemic mouse brain. (a) HMGB1-positive cells are present throughout the layers of the mouse brain cortex (contralateral). In the cortex of mice subjected to 1 h middle cerebral artery occlusion (MCAO), HMGB1 immunofluorescence appears significantly reduced in the ischemic tissue. Hoechst-33258 staining reveals integrity of the ischemic tissue and cell nuclei. The dotted line demarcates the healthy from the ischemic tissue. (b-i) In the ischemic brain subjected to 3 h MCAO, numerous cells (arrowheads) show cytoplasmic HMGB1 localization (red). (b-ii) Higher magnification reveals that these cells show the typical morphology of neurons (arrowheads). (b-iii) Nuclear localization of HMGB1 in the nuclei of neural cells in the contralateral hemisphere. (c-i) Twenty-four hours after MCAO, the anti-HMGB1 antibody stains structures in the ischemic core resembling remnants of dead neurons, whereas the nuclear localization of HMGB1 is conserved in the contralateral hemisphere (c-ii). Note the higher intensity of the background in (c-i) compared with (c-ii). (c-iii) The dotted line demarcates the border between the healthy (right) and ischemic (left) brain 24 h after MCAO. Note the presence of HMGB1 in the nucleus of cells of the penumbra. Bar = 200 μm (a), 40 μm (b-ii, b-iii, and c), and 20 μm (b-ii).

Microinjection of HMGB1 increases transcripts for neuroinflammatory mediators and sensitivity of the brain tissue to ischemic injury

In light of the pro-inflammatory activity of HMGB1 in glial cell cultures and the causal role of inflammation in post-ischemic brain damage (Dirnagl et al. 1999; Iadecola and Alexander 2001), we investigated the effect of iontophoretic injection of recombinant HMGB1 on the brain’s inflammatory response and sensitivity to ischemic injury. Transcript levels of iNOS and IL-1β increased in the mouse cortex after HMGB1 injection (7.5 μg) to an extent lower than that induced by LPS. COX-2 mRNA was increased by LPS injection only (Fig. 6a). However, the expression levels of the related proteins remained below detection limits 24 (not shown) and 48 h (Fig. 6b) after cortical HMGB1 microinjection. On the contrary, LPS induced the two inflammatory proteins for up to 48 h (Fig. 6b). In spite of the lack of immune activation, the brain cortex receiving recombinant HMGB1 displayed an increased sensitivity to the ischemic challenge. Specifically, injection of 2.5 μg had no effect on the size of the ischemic injury after 24 h, whereas at the same time point the ischemic volume increased from 17 ± 2 to 25 ± 4 mm3 (< 0.05, Student t-test; = 5 per group) when the dose of HMGB1 was raised up to 7.5 μg (Fig. 6c and d).

Figure 6.

 Effects of recombinant high mobility group box 1 (HMGB1) on the brain immune response and ischemic damage. (a) Real-time RT-PCR reveals that the transcript levels of iNOS and interleukin (IL)-1β but not those of cyclooxygenase (COX) 2 are increased in the mouse cortex 24 h after HMGB1 microinjection (7.5 μg). Lipopolysaccharides (LPS) (0.5 μg) more efficiently increase mRNAs of the three inflammatory mediators. Values are the mean ± SEM of three experiments. *< 0.05, **< 0.01, ***< 0.001, anova and Dunnett’s post hoc test. (b) At variance with LPS (0.5 μg), iontophoretic microinjection of different doses of HMGB1 does not induce expression of iNOS, IL-1β, and COX2 evaluated 24 h after microinjection. (c) Visualization of hematoxylin–eosin stained brain sections (15 μm) of mice subjected to 24 h after middle cerebral artery occlusion (MCAO) and injected with vehicle or different doses of HMGB1. (d) HMGB1 microinjection dose dependently increases the ischemic volumes of mice subjected 24 h to distal MCAO (= 5 per group, *< 0.05, Student’s t-test).


The present study reports that HMGB1 is released from neural cells exposed to different necrotic stimuli, triggers immune glia activation and worsens excitotoxic as well as ischemic in vitro neuronal death. It also demonstrates that HMGB1 is highly expressed in mouse brain neurons and astrocytes, and released during brain ischemia. Notably, cerebral microinjection of exogenous HMGB1 increases post-ischemic brain injury. Overall, data are in keeping with prior work by Kim et al. (2006).

Several lines of evidence point to HMGB1 as an active mediator of liver, lung, and gut as well as joint disorders (Lotze and Tracey 2005; Ulloa and Messmer 2006), but also highlight its beneficial effects in a model of myocardial infarction (Limana et al. 2005). Although HMGB1 has been originally identified as a neurite-outgrowth inducing protein (Rauvala and Pihlaskari 1987), and its membrane receptors are functionally expressed in the brain (Ding and Keller 2005; Konat et al. 2006), the protein’s role in diseases of the CNS is less appreciated. Our data underscore the neuropathologic role of HMGB1, and build upon previous findings showing that the protein promotes β-amyloid neurotoxicity (Takata et al. 2003, 2004), and contributes to ischemic brain injury (Kim et al. 2006). Given the relevance of the glia inflammatory response to neurodegeneration, the present finding that HMGB1 prompts immune glia activation might be of pathogenetic significance. Similarly, the data showing that dexamethasone reduces HMGB1-dependent inflammatory glia activation suggest that corticosteroids can be used to counteract at least some of the neurotoxic effects of the protein. We also report that the extracellular presence of recombinant HMGB1 increases excitotoxic and ischemic neuronal death in vitro. The fact that dexamethasone does not reduce HMGB1 neurotoxicity hints that in these two models, the protein prompts neurotoxicity through mechanisms different from immune glia activation. In this regard, the recent finding that HMGB1 promotes glutamate release from gliosomes (Pedrazzi et al. 2006) suggests that its neurotoxicity is mediated, at least in part, by increased astrocytic release of glutamate and enhanced excitotoxic neuronal death. This assumption is in keeping with the emerging role of glutamate released from astrocytes in neuropathology (Volterra and Meldolesi 2005). Furthermore, the finding that HMGB1 is constitutively present in both cytoplasm and conditioned media of cultured astrocytes hints that the protein represents a novel autocrine/paracrine regulator of astrocytic glutamate release and astrocyte-neuron crosstalk. It is also worth mentioning that cytoplasmic staining of HMGB1 is almost homogeneous in neurons undergoing necrosis whereas it appears granular in necrotic astrocytes. These findings, together with evidence that extracellular HMGB1 respectively runs as a double or single band in electrophoresis of media from neurons or astrocytes suggest that HMGB1 undergoes different post-translational modification(s) and excretory pathways in neurons and astrocytes during necrosis.

We report here that HMGB1 is highly expressed in the adult mouse brain. In apparent contrast with this finding, prior work demonstrated that protein is only present in areas of continuing neurogenesis in the adult animal (Guazzi et al. 2003). It this study, the authors performed HMGB1 immunohistochemistry on brain sections mounted on slides, whereas we adopted the free-floating technique. Surprisingly, when we adopted the method reported by Guazzi et al. (2003). we also found no HMGB1-positive cells. Although we do not know the molecular mechanism underlying the requirement of the free-floating method for HMGB1 immunostaining, alterations of HMGB1 immunoreactivity has been reported between living and fixed cells (Pallier et al. 2003). Regardless, our data indicate that both neurons and astrocytes do express HMGB1 in their nuclei. Image deblurring and 3D reconstruction also allowed to demonstrate that HMGB1 is not evenly distributed within the cell nucleus, being concentrated in numerous nuclear subregions not colocalizing with heterochromatin. Given the key role of HMGB1 in transcriptional activation, we assume that these domains represent transcriptionally active chromatin foci.

The present study shows that HMGB1 immunostaining is rapidly (1–3 h) lost in the ischemic brain tissue, in keeping with the kinetic of release we obtained in vitro. It also reports that HMGB1 localizes in soma and processes of ischemic neurons, indicating nucleus-to-cytoplasm translocation of the protein during the first stages of neuronal ischemia. These findings, together with the evidence that HMGB1 increases in the cerebrospinal fluid and serum after brain ischemia in the rat (Kim et al. 2006), suggest that the protein translocates from the nucleus of ischemic neurons to the extracellular space and then to the liquor and bloodstream. Importantly, down-regulation of HMGB1 brain levels by shRNA correlates with diminished infarct volumes in the rat (Kim et al. 2006). Again, this finding is in keeping with the present study showing that brain microinjection of HMGB1 increases ischemic volumes in the mouse. Evidence that inflammation plays an active role of in the pathogenesis of ischemic brain injury (Dirnagl et al. 1999), along with the cytokine functions of HMGB1 suggest that the detrimental effect of the protein during brain ischemia is due, at least in part, to its capability of promoting neuroinflammation. Data showing that iontophoretic injection of HMGB1 does not trigger expression of pro-inflammatory mediators (Fig. 6b); however, is in apparent contrast with this assumption and with data obtained with glia exposed to HMGB1 in vitro (Fig. 2). Yet, microinjection of HMGB1 increased transcript levels of iNOS and IL-1β (Fig. 6a). These apparently conflicting data might be reconciled considering that expression levels of pro-inflammatory proteins were below detection limits in our experimental setting. Alternatively, given the key role of mRNA stabilization during immune cell activation (Clark 2000; Seko et al. 2006), the possibility that transcript levels of inflammatory genes induced by HMGB1 decay rapidly and necessitate stabilization by additional immune stimuli can be advanced. Extracellular HMGB1 per se may not be sufficient to prompt neuroinflammation in vivo, necessitating additional glia-activating factors such as those released by necrotic and/or immune cells during cerebral ischemia. Also, the evidence that COX2 mRNA is not increased by HMGB1 in vivo indicates that the protein is a weak immune activator. Relevance of HMGB1 to neuroinflammation, however, is evidenced by prior work showing that suppressing HMGB1 expression reduces ischemia-dependent microglia activation and induction of inflammatory cytokines/enzymes in a model of ischemia-reperfusion in the rat brain (Kim et al. 2006). Taken together, our findings point to HMGB1 as a nuclear protein contributing to neurodegeneration when released extracellularly. The recent report by Goldstein et al. (2006) showing that plasma levels of HMGB1 are elevated up to 10-fold in stroke patients strengthens the possibility that the protein is also of pathogenetic relevance to stroke in humans. The improvement of the pharmacokinetic and pharmacodynamic profile of current anti-HMGB1 compounds (Sakamoto et al. 2001; Ulloa et al. 2002; Lotze and Tracey 2005; Killeen et al. 2006) might help elucidating the protein’s significance to neurological disorders and possibly provide innovative tools to be harnessed to therapeutic interventions.