Early release of HMGB-1 from neurons after theonset of brain ischemia

Jianhua Qiu, Masaki Nishimura, Yumei Wang, John R Sims, Sumei Qiu, Sean I Savitz,Salvatore Salomone and Michael A Moskowitz

Stroke and Neurovascular Regulation Laboratory, Department of Radiology, Massachusetts General Hospital,Charlestown, Massachusetts, USA

The nuclear protein high-mobility group box 1 (HMGB-1) promotes inflammation in sepsis, but littleis known about its role in brain ischemia-induced inflammation. We report that HMGB-1 and itsreceptors, receptor for advanced glycation end products (RAGE), Toll-like receptor 2 (TLR2), andTLR4, were expressed in normal brain and in cultured neurons, endothelia, and glial cells. Duringmiddle cerebral artery occlusion (MCAO), in mice, HMGB-1 immunostaining rapidly disappearedfrom all cells within the striatal ischemic core from 1 h after onset of occlusion. High-mobility groupbox 1 translocation from nucleus to cytoplasm was observed within the cortical periinfarct regions2 h after ischemic reperfusion (2 h MCAO). High-mobility group box 1 predominantly translocated tothe cytoplasm or disappeared in cells that colabeled with the neuronal marker NeuN. Furthermore,RAGE was robustly expressed in the periinfarct region after MCAO. Cellular release of HMGB-1 wasdetected by immunoblotting of cerebrospinal fluid as early as 2 h after ischemic reperfusion (2 hMCAO). High-mobility group box 1 released from neurons, in vitro, after glutamate excitotoxicity,maintained biologic activity and induced glial expression of tumor necrosis factor a (TNFa). Anti-HMGB-1 antibody suppressed TNFa upregulation in astrocytes exposed to conditioned media fromglutamate-treated neurons. Moreover, TNFa and the cytokine intercellular adhesion molecule-1increased in cultured glia and endothelial cells, respectively, after adding recombinant HMGB-1. Inconclusion, HMGB-1 is released early after ischemic injury from neurons and may contribute to theinitial stages of the inflammatory response.Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938; doi:10.1038/sj.jcbfm.9600582; published online14 November 2007

Keywords: HMGB-1; NeuN; brain ischemia; MCAO; neuroinflammation

Introduction

Ischemia induces inflammatory cell recruitment andmigration (neutrophils followed later by monocytes)and upregulates inflammatory mediators (cytokines,chemokines, and endothelial-leukocyte adhesionmolecules) in brain hours to days after the onset ofinjury (Huang et al, 1999). The significance of theinflammatory response to brain ischemia is com-plex, but there is evidence that in the early phase

after ischemia, inflammation contributes to tissueinjury, whereas in later stages, inflammation mightparticipate in brain repair (Friedlander et al, 1997;Garau et al, 2005; Hara et al, 1997; Kumai et al, 2004;Lovering and Zhang, 2005; Marchetti and Abbrac-chio, 2005; Wang et al, 1997). In this early phase,in vivo and in vitro evidence is consistent with theparadigm that the endothelium promotes inflamma-tion and recruits circulating leukocytes through theupregulation of adhesion molecules such as inter-cellular adhesion molecule-1 (ICAM-1), vascularcell adhesion molecule-1, E-selectin, and P-selectin(del Zoppo and Mabuchi, 2003). These recruitedleukocytes then release metalloproteinases, whichparticipate in the breakdown of the neurovascularmatrix with consequent blood–brain barrier disrup-tion, edema, and/or hemorrhage (Justicia et al, 2003;Kolev et al, 2003; Maier et al, 2004; Ponnampalamand Mayberg, 2004; Veldhuis et al, 2003; Wang andLo, 2003). The importance of the inflammatoryresponse in the pathophysiology of ischemic strokeis generally recognized. However, very early specific

Received 2 July 2007; revised 1 October 2007; accepted 10October 2007; published online 14 November 2007

Correspondence: Dr J Qiu, Department of Radiology, Massachu-setts General Hospital, 149, 13th Street, Room 6403, Charlestown,MA 02129, USA.E-mail: jqiu@partners.org

This work was supported by NIH R21NS056212-01A1,

R01NS048198, and Massachusetts General Hospital Neuroscience

Center Core Facility (NIH P30-NS045776). Jianhua Qiu was partly

supported by AHA N0335154. John R Sims was supported by

AHA 0535138N and NIH K08 NS049241. Sean I Savitz was

supported by AHA 0475008N.

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938& 2008 ISCBFM All rights reserved 0271-678X/08 $30.00

www.jcbfm.com

event(s) and mediator(s) have remained elusive and/or considered nonspecific (e.g., DNA and othercellular material released from damaged cells).Herein, we report the important participation ofhigh-mobility group box 1 (HMGB-1) as a potentialcandidate in a specific upstream pathway promotinginflammation after brain ischemia.

High-mobility group box 1 (also known as am-photerin or HMG-1) is a 216 amino acids (29 kDa)DNA-binding protein with a highly conservedstructure in several species (Thomas, 2001). High-mobility group box 1 participates in nucleosomeformation and regulation of gene transcription (Parket al, 2003; Stros et al, 2002). High-mobility groupbox 1 has recently been characterized as a keycytokine (Lotze and Tracey, 2005; Yang et al, 2005).It is secreted by activated macrophages (Bonaldi etal, 2003; Wang et al, 1999), natural killer cells(Semino et al, 2005), and myeloid dendritic cells(Lotze and Tracey, 2005) in response to inflamma-tory stimuli. Interestingly, administration of anti-HMGB-1 antibodies confers significant protectionagainst the lethal effects of endotoxin (Wang et al,2004; Yang et al, 2004). It is unknown whetherHMGB-1, released by neurons after brain ischemia,participates in neuroinflammation. The presentin vivo and in vitro studies were aimed to test thehypothesis that brain cells most susceptible toischemic injury (i.e., neurons) release HMGB-1 andthat early release from neurons or delayed releasefrom glia activates downstream inflammatory me-chanisms within astrocytes and endothelial cells.

Materials and methods

Cell Cultures

Mouse brain endothelial cell line (bEnd3) was purchasedfrom American Type Culture Collection (ATCC) andcultured in Dulbecco’s Modified Eagle’s Medium with10% fetal bovine serum. Primary cultured neurons wereisolated from embryonic day (E)-16 mouse fetuses andcultured in NeuroBasal medium with 2% B27 (Invitrogen,Carlsbad, CA, USA) on polyethylenimine-coated plates asdescribed previously (Qiu et al, 2002). Neurons weregrown for 10 to 12 days before use. Primary mixed glialcell cultures were prepared from postnatal day 1 pups andcultured in Dulbecco’s Modified Eagle’s Medium with10% fetal bovine serum.

Middle Cerebral Artery Occlusion Model

C57 black/6j male mice (weighing 20 to 25 g, 8 to 10 weeksold) were used in all in vivo experiments, except fordetecting HMGB-1 in cerebrospinal fluid (CSF) (maleSprague–Dawley rat, weighing 250 to 300 g, 8 to 10 weeksold). All experiments were performed after institutionallyapproved protocols in accordance with The NationalInstitutes of Health Guide for the Care and Use ofLaboratory Animals under institutional guidelines. The

filament middle cerebral artery occlusion (MCAO) modelwas used, as described previously (Endres et al, 1998).Animals were anesthetized with 2% isoflurane and main-tained on 1.5% isoflurane in 70% nitrous oxide and 30%oxygen. A silicone-coated 8–0 monofilament was intro-duced in the internal carotid artery and advanced toocclude the middle cerebral artery for 0.5, 1, or 2 h. After2 h MCAO, the animals were briefly reanesthetized and thefilament was withdrawn for reperfusion studies. Regionalcerebral blood flow was measured by laser-Doppler (PF2B;Perimed, Stockholm, Sweden) using a flexible probe,placed over the temporal bone after removal of part of thetemporalis muscle, to confirm occlusion and reperfusion.Rectal temperature was maintained between 36.51C and37.51C with a homeothermic blanket (Frederick Haer andCo., Brunswick, ME, USA). To assess HMGB-1 in CSF, ratswere anesthetized and underwent filament occlusion withan 4–0 suture for 2 h followed by reperfusion. Two hoursafter reperfusion, the skin was incised, and the occipitalbone was cleared of muscle to expose the atlanto-occipitalmembrane. A 27-gauge needle was inserted into thecisterna magna and approximately 100mL of CSF waswithdrawn. Only clear CSF was used for analysis.

Immunostaining

Immunostaining was performed as previously described(Qiu et al, 2004; Teramoto et al, 2003). Briefly, mice werekilled at the indicated time points and transcardiallyperfused with cold saline. Brains were fixed with 4%phosphate-buffered paraformaldehyde. Coronal brain sec-tions (floating sections) were prepared (40mm thick). Thefollowing antibodies and dilution conditions were used:polyclonal anti-HMGB-1 antibody (1:200; Pharmingen, SanJose, CA, USA) and monoclonal anti-receptor for advancedglycation end products (RAGE) antibody (1:500; R&DSystem, Minneapolis, MN, USA). Neurons were stainedwith NeuN (1:1,000; Pharmingen) or microtubule-asso-ciated protein 2 (MAP2) (1:200; Sigma, St Louis, MO,USA); astrocytes were stained with glial fibrillary acidicprotein (GFAP) (1:1,000; Sigma). Fluorescent-stained sec-tions were analyzed by confocal microscopy as described(Qiu et al, 2004). For standardization of terminology, wedefine regions of periinfarct as corresponding to dorsalcortical regions 2±0.5 mm lateral to midline spanning1 mm rostral to the bregma and 1 mm caudal to the bregma.

Immunoblot

Immunoblot was performed using routine techniques (Qiuet al, 2002). The primary antibodies included the follow-ing: anti-HMGB-1 antibody (1:200; Pharmingen) andb-actin (1:1,000; Sigma).

Oxygen–Glucose Deprivation and GlutamateTreatment

Primary cultured neurons were subjected to 2 h oxygen–glucose deprivation (OGD) with balanced salt solution

Ischemic neurons release HMGB-1J Qiu et al

928

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

(Goldberg and Choi, 1993) and reoxygenation by replacingwith fresh Neurobasal medium immediately after termina-tion of ODG. For the conditioned medium, which containsneuronal HMGB-1, neurons (the number of initial seedingcells was one million per well in a six-well plate) werecultured in six-well plates with 0.6 mL of medium andtreated with 0.25 mmol/L of glutamate for 18 h and thenthe media were collected. A volume of 5 mg/mL (finalconcentration) of anti-HMGB-1 antibody was added intothe conditioned medium 2 to 3 mins before the mediumwas added into glial cultures.

Quantitative Real-Time Reverse Transcription-PCR

RNA was isolated from either cultured cells or braintissues and treated with DNase I to remove any contam-inating genomic DNA. Complementary DNA was gener-ated by using SuperScript III (Invitrogen). Samples wererun on an ABI Prism 7000 (Applied Biosystems, FosterCity, CA, USA), in triplicates, with negative controls (nocDNA). Primers and Taqman FAM-labeled probes (Ap-plied Biosystems) were mixed with AmpliTaq GoldUniversal PCR Master Mix (Applied Biosystems) and then

diluted with distilled deionized H2O up to 50 mL. Gene-specific products were normalized to an endogenouscontrol of 18S rRNA (Applied Biosystems).

Statistics

mRNA expression levels were quantified by real-timereverse transcription-PCR. mRNA amounts were thenaveraged by group/treatment and compared by analysisof variance, followed by Dunnet post hoc analysis. Pr0.05was considered statistically significant.

Results

High-Mobility Group Box 1 Expression in NormalBrain and Cultured Brain Cells

In mouse brain coronal sections, robust HMGB-1expression was observed in nuclei throughoutthe cortex and subcortical structures, in eitherNeuN or GFAP-positive cells (Figures 1A to 1F).High-mobility group box 1 was also observed in

Figure 1 Nuclear localization of HMGB-1 expression in neurons and astrocytes from normal mouse brain and cultured neurons andglial cells. Within the brain, immunofluorescence staining shows the nuclear distribution of HMGB-1 (green, B, C, E, F, and I) incortical neurons (A to C, NeuN = red), callosal astrocytes (D to F, GFAP = red), and cortical endothelial cells noted by arrows (I,GFAP = red). In culture, HMGB-1 (green, G and H) localizes to the nucleus in primary cultured neurons (G, MAP2 = red) and primarycultured glial cells (H, GFAP = red). Bars: 30 mm.

Ischemic neurons release HMGB-1J Qiu et al

929

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

endothelial cells (Figure 1I, endothelial cells notedby arrow). In primary cultures, both neurons andastrocytes expressed nuclear HMGB-1 (Figures 1G–H). Immunoblotting of normal brain extract alsoconfirmed expression of a 29 kDa protein that wasHMGB-1 (data not shown).

Cerebral Ischemia Induces High-Mobility Group Box 1Translocation and Release in Brain Cells

When cells become injured or undergo necrosis,HMGB-1, normally residing in nuclei, translocatesto the cytoplasm and/or extracellular space (Scaffidiet al, 2002). Here, we examined several time pointsin a mouse MCAO model (Figure 2A) to determinewhether HMGB-1 translocation and release occurafter cerebral ischemia. There was no change inHMGB-1 staining in either the contralateral striatum(Figure 2D) or the contralateral cerebral cortex(Figure 2H) at any time point examined duringischemia or after reperfusion. During ischemia,HMGB-1 staining was slightly diminished in theischemic core of the striatum as early as 30 mins(Figure 2B) with substantial and diffuse loss in thestriatum 1 h after onset of MCAO (Figure 2C). Unlikethe striatum with complete loss of HMGB-1 staining1 h after onset of ischemia, the ipsilateral cortex hadno observed change in HMGB-1 staining at 30 mins

or 1 h during ischemia (Figure 2E). However, HMGB-1loss/translocation was observed at 2 h after MCAOin cortex adjacent to the striatal ischemic core (datanot shown).

To assess later time points, we analyzed tissueafter reperfusion after 2 h of occlusion. High-mobi-lity group box 1 staining was lost in ischemicstriatum and cortex, but appeared localized to thecell cytoplasm at the margins of ischemic injurywithin the cortex (Figures 2F and 2G) as early as 2 hafter reperfusion. The area of HMGB-1 translocationseemed to be maximum 2 h after reperfusion, as nosignificant change in staining pattern was noted atlater time points (6 and 24 h) (data not shown).These results suggest rapid and region-specifictemporal changes in HMGB-1 immunostaining afterischemia. Within the ischemic core, HMGB-1 ap-pears to be released from the nucleus and cytoplasminto the extracellular space, whereas periinfarctregions appear to maintain translocation of HMGB-1 from the nucleus into the cytoplasm.

To assess the extracellular release of HMGB-1 afterMCAO, we examined CSF for the presence ofHMGB-1. Rats were used in this experiment becausethe volume of CSF was more readily available. High-mobility group box 1 was observed 2 h afterreperfusion after 2 h MCAO (Figure 3A). NoHMGB-1 was detectable in CSF from sham-operatedanimals. We ran immunoblots with lysates from

Figure 2 Cerebral ischemia induces HMGB-1 translocation and release. (A) A representative coronal brain section stained by TTCshowing viable (red) and dead tissue (white) 24 h after a 2 h MCAO. (B to H) The lettered squares show the areas analyzed byconfocal microscopy at earlier time points and correspond to micrographs. High-mobility group box 1 staining was slightlydiminished in the striatum during 30 mins of ischemia (B) or lost during 1 h of ischemia (C) with preservation of the normal nuclearlocalization on the nonischemic contralateral side at the 1-h time point (D). (E) Within the cortex, during 1 h of ischemia, HMGB-1remains predominantly nuclear. A low magnification view of HMGB-1 immunostaining (F) shows a distinct border corresponding tothe cortical periinfarct region between anterior and middle cerebral artery territory 2 h after reperfusion (2 h MCAO). The dashed lineseparates the areas where HMGB-1 is predominantly nuclear (left side) or predominantly cytoplasmic (right side). A highermagnification view (G) of the region to the right of the border shows periinfarct cortical cells with a predominant cytosolic distributionof HMGB-1. As a control, the contralateral nonischemic hemisphere-shows normal nuclear staining of HMGB-1, 2 h after reperfusion(2 h MCAO). Bars: 30mm for (B, C, D, E, G, and H); Bar: 100 mm for (F). Cont, contralateral side; Ipsi, ischemic ipsilateral side.

Ischemic neurons release HMGB-1J Qiu et al

930

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

contralateral nonischemic and ispsilateral ischemicmouse brain tissues using the same antibody as inimmunostaining. As shown in Figures 3B and 3C,total HMGB-1 did not significantly differ betweenischemic and nonischemic tissues from homologousbrain regions after MCAO, despite the loss ofnuclear and cytoplasmic staining. These data sug-gest that the loss of both nuclear and cytosolicimmunostaining in the ischemic core was not due todegradation or modification of the HMGB-1 epitope,but was rather due to its release into the extra-cellular space. Furthermore, they suggest that theloss of HMGB-1 into the CSF is small relative to thetotal pool of HMGB-1. In contrast to CSF, HMGB-1was not detected in plasma samples from eithersham-operated mice or mice subjected to MCAO, forup to 24 h by Western blot.

Early High-Mobility Group Box 1 Translocation andRelease Induced by Middle Cerebral Artery OcclusionOccur Mostly in Neurons

To identify the cell types in which HMGB-1translocation/release occurred after focal ischemia,coronal sections were costained with anti-HMGB-1and anti-NeuN antibodies. Within the ischemicstriatal core, HMGB-1 staining was negative ordiminished in most NeuN-positive cells 1 h afterMCAO (Figures 4A to 4C), suggesting that HMGB-1

had already been released from neurons by 1 h afterMCAO. Two hours after reperfusion (2 h MCAO),HMGB-1 was observed in the cytoplasm of NeuN-positive cells, mainly within the outer margins ofthe cortical ischemic territory (Figures 4D to 4F),indicative of translocation from nucleus to cyto-plasm. However, nuclear HMGB-1 staining wasfound in some of GFAP-positive astrocytes withinthe regions where HMGB-1 cytoplasmic transloca-tion occurred in almost all neurons (Figures 4G to4I). This finding may in part reflect the selectivevulnerability of neurons to ischemic injury andsuggests that neurons could be one of the mainsources of released HMGB-1, at least at this earlytime point.

To confirm the susceptibility of neurons to earlyHMGB-1 translocation/release, we subjected pri-mary cultures to glutamate or 2 h of OGD andassayed time-dependent changes in HMGB-1immunoreactivity in conditioned media. As shownin Figure 5A, HMGB-1 was undetectable in mediumfrom untreated cells, but it was detected byimmunoblot starting as early as 1 to 2 h afterglutamate (0.25 mmol/L) or reoxygenation fromOGD; moreover, there was no further substantialincrease in the release of HMGB-1 out to 24 h afterreoxygenation. Assuming no protein degradation inconditioned medium, this time course suggests thatmost of HMGB-1 release occurred in the first 2 hunder these conditions. Despite HMGB-1 release atthis time point, the gross morphology of neuronswas still preserved and very few cells were lysed(Figure 5B). Thus, HMGB-1 release in vitro occurs asan early event after injury in cultured neurons,consistent with in vivo results. Moreover, HMGB-1translocation was observed in some neurons ex-posed to OGD or glutamate by immunofluorescence(data not shown). A prolonged exposure to OGD(18 h, followed by 6 h reoxygenation) led to a loss ofintracellular staining for HMGB-1 in astrocytes also,but such a loss did not occur after 2 h OGD and 6 hreoxygenation (data not shown).

Putative High-Mobility Group Box 1 Receptors (RAGE,TLR2, and TLR4) in Brain and Cultured Brain Cells

High-mobility group box 1 acts as a proinflammatorycytokine after interaction with RAGE (Hori et al,1995), Toll-like receptor 2 (TLR2), and TLR4 (Parket al, 2004). We detected RAGE, TLR2, and TLR4mRNA expression in bEnd3 cell line, culturedprimary neurons, and astrocytes as well as in braintissue (Figure 6A). This suggested that cells in theneurovascular unit might be susceptible to regula-tion by HMGB-1 signaling. Very low levels of RAGEexpression were detected by immunohistochemistryin the normal brain (Figure 6B), but there was robustRAGE expression only in the cortical ischemicperiinfarct region—the area between translocationand nontranslocation of HMGB-1 corresponding to

Figure 3 High-mobility group box 1 expression in CSF andbrain parenchyma after MCAO. (A) Immunoblotting detectsHMGB-1 in the CSF 2 h after reperfusion. Rats were subjectedto a 2 h MCAO, and CSF was withdrawn from cisterna magna2 h after reperfusion in two ischemic rats. All lanes are loadedwith 20mL of CSF. No HMGB-1 is detected in two sham rats.(B) The immunoblot shows no significant qualitative change intotal HMGB-1 protein isolated from middle cerebral arteryterritory in control (sham) or contralateral (cont) or ipsilateral(ipsi) brain tissue from mice subjected to 2 h MCAO followed by2 h reperfusion. (C) Quantitation of densitometric measurementof immunoblots shows no significant change between contral-ateral and ipsilateral regions. The data are expressed as relativeexpression for each side after normalizing for b-actin expression(n = 4).

Ischemic neurons release HMGB-1J Qiu et al

931

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

Figure 2E 6 h after reperfusion (2 h MCAO) (Figure6C). Robust RAGE immunoreactivity was mainlyseen in cells exhibiting a neuronal morphology inlayers 4 to 5. These results suggest that RAGE, aputative HMGB-1 receptor, may be upregulated afterischemic injury.

Extracellular High-Mobility Group Box 1 InducesInflammatory Cytokines in Cultured Neurons,Astrocytes, and Endothelial Cells

Tumor necrosis factor a (TNFa), inducible nitricoxide synthase, and ICAM-1 expression are knownto increase in ischemic brain and play a significantrole in neuroinflammation. To determine whetherextracellular HMGB-1 could induce these cytokines,we exposed cultured neurons to different concen-trations of recombinant HMGB-1. In a concentration-dependent manner, HMGB-1 increased TNFa mRNAexpression in astroglia (Figure 7A). Tumor necrosisfactor a upregulation was noted as early as 1 h afterHMGB-1 exposure and peaked at 3 to 6 h (Figure7B). Tumor necrosis factor a expression was alsofound in the other cell types of the neurovascular

Figure 4 Early HMGB-1 translocation and release is induced by MCAO in neurons and some astrocytes. Immunofluorescencestaining shows loss of HMGB-1 (green) in striatal neurons (NeuN, red, A to C) and some striatal astrocytes (GFAP, red, G to I), during1 h of ischemia. The arrows point to the nuclear HMGB-1 staining preserved in some astrocytes. After 2 h of reperfusion (2 h MCAO),HMGB-1 (green) is predominantly cytoplasmic in cortical periinfarct neurons (NeuN, red, D to F). Bars: 30mm.

Figure 5 High-mobility group box 1 is released from injuredcultured neurons. Primary cultured neurons were subjected to0.25 mmol/L glutamate or oxygen–glucose deprivation (OGD)and assayed for time-dependent changes in HMGB-1 (A) Animmunoblot for HMGB-1 in culture medium (15ml of media/lane) shows no HMGB-1 in media from control neurons,whereas a rapid monophasic release of HMGB-1 into the mediaoccurs by 2 h after OGD (left lanes) and glutamate-induced(right lanes) excitotoxicity. (B) A phase contrast micrograph ofneurons, untreated (left) or after 6 h OGD (right) showspreservation of cellular morphology. Bar: 30 mm.

Ischemic neurons release HMGB-1J Qiu et al

932

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

unit such as in neurons and endothelial cells (datanot shown). However, in neurons, TNFa upregula-tion was not as robust as in glia and endothelium.The expression of other inflammatory moleculeswas assessed after HMGB-1 treatment. Induciblenitric oxide synthase was increased in glia, andICAM-1 was increased in endothelium at 3 h afterHMGB-1 treatment (Figure 7C to 7D).

Conditioned Medium from Glutamate-TreatedNeurons Induces Tumor Necrosis Factor a inAstrocytes, an Effect Inhibited by Anti-HMGB-1Blocking Antibody

To determine whether HMGB-1 released frominjured cultured neurons was biologically active asa pro-inflammatory mediator, we treated primaryastrocytic cultures with conditioned mediumfrom glutamate-treated neurons. As shown in Figure7E, conditioned medium robustly induced TNFamRNA expression in astrocytes. Glutamate didnot modify TNFa expression when added directlyto the astrocyte culture medium. Tumor necrosisfactor a expression was inhibited by 90% afterpretreatment with anti-HMGB-1 blocking antibody,

indicating that this response was mediated bysoluble HMGB-1.

Discussion

HMBG-1 is a key mediator of inflammation that weand others hypothesized might be released fromdamaged cells after focal cerebral ischemia. In thisstudy, we showed that (1) HMGB-1 was translocatedfrom nucleus to cytoplasm and release occurred inischemic neurons as early as 1 to 2 h after MCAO; (2)HMGB-1 release also occurred from cultured neu-rons as early as 1 to 2 h after glutamate treatment orOGD followed by reoxygenation (astrocytes ap-peared more resistant); (3) putative HMGB-1 recep-tors were expressed in neurons, glia, andendothelial cells, and one of these receptors, RAGE,appeared upregulated in the cortical periinfarctregion; and (4) recombinant HMGB-1 or HMGB-1released by injured neurons in culture inducedproinflammatory cytokine expression in neurons,astrocytes, and endothelial cells.

Previous studies have shown that HMGB-1 can bereleased into the extracellular space either actively,after HMGB-1 is hyperacetylated in the nucleus(Bonaldi et al, 2003), or passively, when HMGB-1diffuses out of cells that are undergoing necrosis andcell membrane is damaged (Scaffidi et al, 2002).Active secretion typically occurs not only in macro-phages and monocytes (Chen et al, 2004; Kalinina etal, 2004), but also in nonimmune cells such ashepatocytes where it is overexpressed after hypoxia(Tsung et al, 2005). Here, we suggest that HMGB-1 isreleased from the ischemic brain parenchyma with-in 1 h after MCAO, based on early loss of HMGB-1immunoreactivity in the striatum. Kim et al (2006)showed that HMGB-1 was not hyperacetylated,suggesting that its release is passive. Our demon-stration that HMGB-1 is present in rat CSF is also inagreement with Kim et al (2006), who detected thisprotein 3 h after transient MCAO. We further showthat neurons are one of the principal sources ofHMGB-1 release in the early stages of ischemicinjury. In fact, HMGB-1 staining was lost in NeuN-positive cells in the ischemic core and cytosolicHMGB-1 staining colocalized with NeuN in theischemic cortex. The presence of HMGB-1 in ratcisterna magna fluid probably reflects its release anddiffusion from ischemic tissue. We did not detectany HMGB-1 in mouse plasma in contrast to a studyin the rat (Kim et al, 2006).

High-mobility group box 1 may act as a proin-flammatory signal and contribute to the delayeddeath of brain cells in the ischemic periinfarctregion. Extracellular HMGB-1 binds to its receptorssuch as RAGE, TLR2, and TLR4. We detectedmRNAs for these receptors in cell types of theneurovascular unit in vitro and in brain extracts.Toll-like receptor 2 and TLR4 expression hasalready been documented by others in endothelium

Figure 6 Putative HMGB-1 receptors (RAGE, TLR2, and TLR4)in brain and cultured brain cells. (A) mRNA expression forRAGE, TLR2 and, TLR4 is detected by end-point reversetranscription-PCR in normal mouse brain, primary culturesof glia and neurons, and a brain endothelial cell line (bEnd3).An increase in RAGE immunostaining (C) is seen 6 hafter reperfusion (2 h MCAO) relative to normal cortex (B).Bar: 50mm.

Ischemic neurons release HMGB-1J Qiu et al

933

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

(Faure et al, 2000) and in astrocytes (Bowman et al,2003). We found upregulation of RAGE in theperiinfarct region 6 h after reperfusion in the 2 hMCAO model. Because HMGB-1 is an endogenousligand with high affinity for RAGE (Hori et al, 1995)and RAGE expression is reportedly induced byHMGB-1 (Li et al, 1998), the upregulation of RAGEin ischemic brain further supports the centralhypothesis of a key proinflammatory role for

HMGB-1 in ischemic stroke. Conceivably, increasedexpression of HMGB-1 receptors after ischemia mayenhance the sensitivity of brain cells to HMGB-1. Toour knowledge, RAGE overexpression in ischemicbrain has not yet been reported, although its role inb-amyloid metabolism and Alzheimer’s disease hasbeen studied (Deane et al, 2003; Mackic et al, 1998).Receptor for advanced glycation end product signalsthrough pathways that involve ERK1 (extracellular

Figure 7 High-mobility group box 1 upregulates the expression of inflammatory mediators in cultured glia and endothelial cells.Tumor necrosis factor a, inducible nitric oxide synthase, and ICAM-1 mRNA expression was detected by real-time reversetranscription-PCR. (A) Recombinant HMGB-1 dose dependently upregulates TNFa mRNA in primary cultured glia treated for 6 h.(B) Recombinant HMGB-1 (100 ng/mL) rapidly increases TNFa expression as early as 1 h after treatment of primary cultured glia.(C) Recombinant HMGB-1 (100 ng/mL) upregulates inducible nitric oxide synthase mRNA expression in primary cultured glia 3 hafter treatment. (D) Recombinant HMGB-1 (100 ng/mL) upregulates ICAM-1 mRNA expression in brain endothelial cell line (bEnd3)3 h after treatment. (E) Upregulation of TNFa mRNA expression in primary cultured glia treated with conditioned media from neuronsexposed to excitotoxic levels of glutamate (0.25 mol/L) is blocked by the addition of HMGB-1 blocking antibody (5 mg/mL). Exposureof glia to same levels of glutamate present in the conditioned media has no effect on TNFa mRNA expression. Ordinate: mRNAexpression relative to untreated control. Error bars indicate s.d. *P < 0.01 versus control, one-way analysis of variance followed byDunnet post hoc analysis.

Ischemic neurons release HMGB-1J Qiu et al

934

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

signal-regulated kinase) and/or ERK2 as well as themitogen-activated protein kinase p38, and it pro-motes the activation of nuclear factor-kB (Taguchi etal, 2000), which itself activates the transcription ofproinflammatory genes (such as interleukin (IL)-1b,IL-6, and TNFa). Interaction of HMGB-1 with TLR2and TLR4 promotes inflammatory responses thatultimately also lead to nuclear factor-kB activation.The expression of these other putative HMGB-1receptors has been shown to increase duringinflammation (Cipollone et al, 2003; Koedel et al,2003). Both TLR2 and TLR4 are robustly upregu-lated in brain microvasculature of young rats injuredby neonatal hypoxia–ischemia (Maslinska et al,2004). Whether they are upregulated in experimen-tal brain ischemia, however, remains to be deter-mined.

If HMGB-1 acts as an early upstream inflammatorysignal in cerebral ischemia, we expect its transloca-tion/release to precede the upregulation of othercytokines and endothelial adhesion molecules andthe subsequent leukocyte recruitment into theischemic territory. We and others have shown thatERK activity increases as early as 0.5 to 1 h afterischemia in the middle cerebral artery territory(Alessandrini et al, 1999; Wang et al, 2003; Wu etal, 2000). This signaling is potentially linked toTLR2, TLR4, and/or RAGE activation (Ishihara et al,2003; Park et al, 2003; Taguchi et al, 2000) and isknown to promote transcription of cytokines such asTNFa (Fiuza et al, 2003). Tumor necrosis factor amRNA reportedly increases within 1 h after induc-tion of ischemia in the periinfarct and infarct area,peaks at 12 h, and remains elevated for up to 24 h(Gregersen et al, 2000; Lambertsen et al, 2002, 2005;Schroeter et al, 2003; Yang et al, 1999).

The data reported herein suggest that HMGB-1may participate as an early upstream initiator ofinflammation. In vitro, we observed a 3- to 4-foldTNFa mRNA induction in glial cultures already after1 h incubation with exogenous HMGB-1; TNFamRNA levels increased more than 10-fold after 3 h.Noteworthy, TNFa mRNA induction in cultured gliawas induced with 100 ng/mL HMGB-1; this concen-tration seems consistent with values that we canroughly estimate in CSF after MCAO, and thiscorresponds to approximately 3 nmol/L, which isclose to the kDa value of HMGB-1 for RAGE (Hori etal, 1995). High-mobility group box 1 proinflamma-tory signaling might begin early, possibly in the firsthour after focal ischemia, but may continue forhours thereafter (Faraco et al, 2007). Consistent witha more protracted HMGB-1 signaling, we havepreviously reported neuroprotection through ERKinhibition, even when the treatment had beeninitiated 3 h after MCAO (Namura et al, 2001).

Given its release from neurons, HMGB-1 may playmultiple, proinflammatory roles in the neurovascu-lar unit by upregulating cytokines and adhesionmolecules in astrocytes and endothelium. Astro-cytes directly contact and translate signals to

neurons and blood vessels. The demonstration thatcultured primary astrocytes upregulate TNFa assoon as 1 h after exposure to low concentrations ofHMGB-1 seems particularly relevant. After brainischemia, the astrocyte might represent an earlytarget for released HMGB-1, as might endothelialcells. In this study, we found that HMGB-1 upregu-lated mRNA ICAM-1 in bEnd3 cell line. Similarstudies showing an increase in protein levels ofICAM-1 at 4, 6, and 16 h in two other human celllines are consistent with our data (Fiuza et al, 2003;Treutiger et al, 2003). Although bEnd3 cell linesconstitutively express ICAM-1 and have beenwidely used for studying adhesion molecules, theymay have a different response to inflammationcompared to primary brain endothelial cells. Ourunpublished data suggest that the baseline of ICAM-1 expression in bEnd3 cell line is lower than that inrat primary brain endothelial cells. Upregulation ofICAM-1 expression in endothelial cells wouldrecruit immune cells into the ischemic area, leadingto a more severe inflammatory reaction. Further-more, it remains to be determined whether suchresponses by the endothelium are beneficial ordeleterious, as their timing may be critical and havecounterbalancing effects. Although the results ofthis study suggest that HMGB-1 may contribute tothe early stages of inflammation in brain ischemia,Kim et al (2006) showed that HMGB-1 mayalso contribute to later stages of inflammation byactivating microglia. Therefore, HMGB-1 may exertmultiple proinflammatory effects in the ischemicbrain.

The importance of neuroinflammation after focalbrain ischemia has led to the development of drugsthat target the mediators responsible for leukocyterecruitment. Unfortunately, clinical trials usingeither anti-ICAM-1 antibodies or a recombinantglycoprotein with selective binding to the CD11bintegrin did not show any significant beneficialeffect (Krams et al, 2003; Enlimomab Acute StrokeTrial Investigators, 2001). The complexity of inflam-matory signaling raises the question whether target-ing a single molecule or receptor would be aneffective therapeutic strategy. A more important taskis to identify central molecular mechanisms in theischemic brain that may generate key, early media-tors of the inflammatory response in the vasculature(endothelium) and in the circulating cell compart-ments. To date, very early event(s) and mediator(s)have not yet been identified. Our study points to apotential novel and pivotal role for HMGB-1 in areceptor-specific signaling pathway linking veryearly stages of cerebral ischemic injury with theactivation of local neuroinflammatory responseswithin the neurovascular unit. Additional researchwill be required to clarify this point.

In conclusion, HMGB-1 is released by injuredneurons early after ischemia and may represent amajor upstream paracrine inflammatory mediatorwithin the neurovascular unit in ischemic stroke.

Ischemic neurons release HMGB-1J Qiu et al

935

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

Targeting HMGB-1 signaling may be an importantnovel therapeutic strategy in acute ischemic stroke.

Acknowledgements

We thank Christian Waeber for preparation of figuresand Igor Bagayev for his excellent technical assis-tance.

References

Alessandrini A, Namura S, Moskowitz MA, Bonventre JV(1999) MEK1 protein kinase inhibition protects againstdamage resulting from focal cerebral ischemia. ProcNatl Acad Sci USA 96:12866–9

Bonaldi T, Talamo F, Scaffidi P, Ferrera D, Porto A, BachiA, Rubartelli A, Agresti A, Bianchi ME (2003) Mono-cytic cells hyperacetylate chromatin protein HMGB1 toredirect it towards secretion. EMBO J 22:5551–60

Bowman CC, Rasley A, Tranguch SL, Marriott I (2003)Cultured astrocytes express Toll-like receptors forbacterial products. Glia 43:281–91

Chen G, Li J, Ochani M, Rendon-Mitchell B, Qiang X,Susarla S, Ulloa L, Yang H, Fan S, Goyert SM, Wang P,Tracey KJ, Sama AE, Wang H (2004) Bacterial endotoxinstimulates macrophages to release HMGB1 partlythrough CD14- and TNF-dependent mechanisms.J Leukoc Biol 76:994–1001

Cipollone F, Iezzi A, Fazia M, Zucchelli M, Pini B,Cuccurullo C, De Cesare D, De Blasis G, Muraro R,Bei R, Chiarelli F, Schmidt AM, Cuccurullo F, MezzettiA (2003) The receptor RAGE as a progression factoramplifying arachidonate-dependent inflammatory andproteolytic response in human atherosclerotic plaques:role of glycemic control. Circulation 108:1070–7

Deane R, Du Yan S, Submamaryan RK, LaRue B, JovanovicS, Hogg E, Welch D, Manness L, Lin C, Yu J, Zhu H,Ghiso J, Frangione B, Stern A, Schmidt AM, ArmstrongDL, Arnold B, Liliensiek B, Nawroth P, Hofman F,Kindy M, Stern D, Zlokovic B (2003) RAGE mediatesamyloid-beta peptide transport across the blood-brainbarrier and accumulation in brain. Nat Med 9:907–13

del Zoppo GJ, Mabuchi T (2003) Cerebral microvesselresponses to focal ischemia. J Cereb Blood Flow Metab23:879–94

Endres M, Laufs U, Huang Z, Nakamura T, Huang P,Moskowitz MA, Liao JK (1998) Stroke protection by3-hydroxy-3-methylglutaryl (HMG)-CoA reductaseinhibitors mediated by endothelial nitric oxidesynthase. Proc Natl Acad Sci USA 95:8880–5

Enlimomab Acute Stroke Trial Investigators (2001) Use ofanti-ICAM-1 therapy in ischemic stroke: results of theEnlimomab Acute Stroke Trial. Neurology 57:1428–34

Faraco G, Fossati S, Bianchi ME, Patrone M, Pedrazzi M,Sparatore B, Moroni F, Chiarugi A (2007) High mobilitygroup box 1 protein is released by neural cells upondifferent stresses and worsens ischemic neurodegen-eration in vitro and in vivo. J Neurochem 103:590–603

Faure E, Equils O, Sieling PA, Thomas L, Zhang FX,Kirschning CJ, Polentarutti N, Muzio M, Arditi M(2000) Bacterial lipopolysaccharide activates NF-kappaB through Toll-like receptor 4 (TLR-4) incultured human dermal endothelial cells. Differential

expression of TLR-4 and TLR-2 in endothelial cells.J Biol Chem 275:11058–63

Fiuza C, Bustin M, Talwar S, Tropea M, Gerstenberger E,Shelhamer JH, Suffredini AF (2003) Inflammation-promoting activity of HMGB1 on human microvascularendothelial cells. Blood 101:2652–60

Friedlander RM, Gagliardini V, Hara H, Fink KB, Li W,MacDonald G, Fishman MC, Greenberg AH, MoskowitzMA, Yuan J (1997) Expression of a dominant negativemutant of interleukin-1 beta converting enzyme intransgenic mice prevents neuronal cell death inducedby trophic factor withdrawal and ischemic brain injury.J Exp Med 185:933–40

Garau A, Bertini R, Colotta F, Casilli F, Bigini P, CagnottoA, Mennini T, Ghezzi P, Villa P (2005) Neuroprotectionwith the CXCL8 inhibitor repertaxin in transient brainischemia. Cytokine 30:125–31

Goldberg MP, Choi DW (1993) Combined oxygen andglucose deprivation in cortical cell culture: calcium-dependent and calcium-independent mechanisms ofneuronal injury. J Neurosci 13:3510–24

Gregersen R, Lambertsen K, Finsen B (2000) Microglia andmacrophages are the major source of tumor necrosisfactor in permanent middle cerebral artery occlusion inmice. J Cereb Blood Flow Metab 20:53–65

Hara H, Friedlander RM, Gagliardini V, Ayata C, Fink K,Huang Z, Shimizu-Sasamata M, Yuan J, Moskowitz MA(1997) Inhibition of interleukin 1beta converting en-zyme family proteases reduces ischemic and excito-toxic neuronal damage. Proc Natl Acad Sci USA94:2007–12

Hori O, Brett J, Slattery T, Cao R, Zhang J, Chen JX,Nagashima M, Lundh ER, Vijay S, Nitecki D et al (1995)The receptor for advanced glycation end products(RAGE) is a cellular binding site for amphoterin.Mediation of neurite outgrowth and co-expression ofrage and amphoterin in the developing nervous system.J Biol Chem 270:25752–61

Huang J, Kim LJ, Mealey R, Marsh Jr HC, Zhang Y, TennerAJ, Connolly Jr ES, Pinsky DJ (1999) Neuronal protec-tion in stroke by an sLex-glycosylated complementinhibitory protein. Science 285:595–9

Ishihara K, Tsutsumi K, Kawane S, Nakajima M, Kasaoka T(2003) The receptor for advanced glycationend-products (RAGE) directly binds to ERK by aD-domain-like docking site. FEBS Lett 550:107–13

Justicia C, Panes J, Sole S, Cervera A, Deulofeu R,Chamorro A, Planas AM (2003) Neutrophil infiltrationincreases matrix metalloproteinase-9 in the ischemicbrain after occlusion/reperfusion of the middle cerebralartery in rats. J Cereb Blood Flow Metab 23:1430–40

Kalinina N, Agrotis A, Antropova Y, DiVitto G, KanellakisP, Kostolias G, Ilyinskaya O, Tararak E, Bobik A (2004)Increased expression of the DNA-binding cytokineHMGB1 in human atherosclerotic lesions: role ofactivated macrophages and cytokines. ArteriosclerThromb Vasc Biol 24:2320–5

Kim JB, Sig Choi J, Yu YM, Nam K, Piao CS, Kim SW, LeeMH, Han PL, Park JS, Lee JK (2006) HMGB1, a novelcytokine-like mediator linking acute neuronal deathand delayed neuroinflammation in the postischemicbrain. J Neurosci 26:6413–21

Koedel U, Angele B, Rupprecht T, Wagner H, RoggenkampA, Pfister HW, Kirschning CJ (2003) Toll-like receptor 2participates in mediation of immune response inexperimental pneumococcal meningitis. J Immunol170:438–44

Ischemic neurons release HMGB-1J Qiu et al

936

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

Kolev K, Skopal J, Simon L, Csonka E, Machovich R, NagyZ (2003) Matrix metalloproteinase-9 expression in post-hypoxic human brain capillary endothelial cells: H2O2

as a trigger and NF-kappaB as a signal transducer.Thromb Haemost 90:528–37

Krams M, Lees KR, Hacke W, Grieve AP, Orgogozo JM,Ford GA (2003) Acute Stroke Therapy by Inhibition ofNeutrophils (ASTIN): an adaptive dose–responsestudy of UK-279276 in acute ischemic stroke. Stroke34:2543–8

Kumai Y, Ooboshi H, Takada J, Kamouchi M, Kitazono T,Egashira K, Ibayashi S, Iida M (2004) Anti-monocytechemoattractant protein-1 gene therapy protects againstfocal brain ischemia in hypertensive rats. J Cereb BloodFlow Metab 24:1359–68

Lambertsen KL, Gregersen R, Finsen B (2002) Microglial–macrophage synthesis of tumor necrosis factor afterfocal cerebral ischemia in mice is strain dependent.J Cereb Blood Flow Metab 22:785–97

Lambertsen KL, Meldgaard M, Ladeby R, Finsen B (2005)A quantitative study of microglial–macrophage synth-esis of tumor necrosis factor during acute and late focalcerebral ischemia in mice. J Cereb Blood Flow Metab25:119–35

Li J, Qu X, Schmidt AM (1998) Sp1-binding elements inthe promoter of RAGE are essential for amphoterin-mediated gene expression in cultured neuroblastomacells. J Biol Chem 273:30870–8

Lotze MT, Tracey KJ (2005) High-mobility group box 1protein (HMGB1): nuclear weapon in the immunearsenal. Nat Rev Immunol 5:331–42

Lovering F, Zhang Y (2005) Therapeutic potential of TACEinhibitors in stroke. Curr Drug Targets CNS NeurolDisord 4:161–8

Mackic JB, Stins M, McComb JG, Calero M, Ghiso J, KimKS, Yan SD, Stern D, Schmidt AM, Frangione B,Zlokovic BV (1998) Human blood-brain barrier recep-tors for Alzheimer’s amyloid-beta 1-40. Asymmetricalbinding, endocytosis, and transcytosis at the apicalside of brain microvascular endothelial cell monolayer.J Clin Invest 102:734–43

Maier CM, Hsieh L, Yu F, Bracci P, Chan PH (2004) Matrixmetalloproteinase-9 and myeloperoxidase expression:quantitative analysis by antigen immunohistochemis-try in a model of transient focal cerebral ischemia.Stroke 35:1169–74

Marchetti B, Abbracchio MP (2005) To be or not to be(inflamed)—is that the question in anti-inflammatorydrug therapy of neurodegenerative disorders? TrendsPharmacol Sci 26:517–25

Maslinska D, Laure-Kamionowska M, Maslinski S (2004)Toll-like receptors in rat brains injured by hypoxic-ischaemia or exposed to staphylococcal alpha-toxin.Folia Neuropathol 42:125–32

Namura S, Iihara K, Takami S, Nagata I, Kikuchi H,Matsushita K, Moskowitz MA, Bonventre JV, Alessan-drini A (2001) Intravenous administration of MEKinhibitor U0126 affords brain protection against fore-brain ischemia and focal cerebral ischemia. Proc NatlAcad Sci USA 98:11569–74

Park JS, Arcaroli J, Yum HK, Yang H, Wang H, Yang KY,Choe KH, Strassheim D, Pitts TM, Tracey KJ, AbrahamE (2003) Activation of gene expression in humanneutrophils by high mobility group box 1 protein. AmJ Physiol Cell Physiol 284:C870–9

Park JS, Svetkauskaite D, He Q, Kim JY, Strassheim D,Ishizaka A, Abraham E (2004) Involvement of Toll-like

receptors 2 and 4 in cellular activation by high mobilitygroup box 1 protein. J Biol Chem 279:7370–7

Ponnampalam SN, Mayberg MR (2004) Mediators ofblood–brain barrier disruption and potential therapeu-tic interventions for protection of the barrier followingfocal ischemia. Clin Neurosurg 51:112–9

Qiu J, Takagi Y, Harada J, Rodrigues N, Moskowitz MA,Scadden DT, Cheng T (2004) Regenerative response inischemic brain restricted by p21cip1/waf1. J Exp Med199:937–45

Qiu J, Whalen MJ, Lowenstein P, Fiskum G, Fahy B,Darwish R, Aarabi B, Yuan J, Moskowitz MA (2002)Upregulation of the Fas receptor death-inducing signal-ing complex after traumatic brain injury in mice andhumans. J Neurosci 22:3504–11

Scaffidi P, Misteli T, Bianchi ME (2002) Release ofchromatin protein HMGB1 by necrotic cells triggersinflammation. Nature 418:191–5

Schroeter M, Kury P, Jander S (2003) Inflammatorygene expression in focal cortical brain ischemia:differences between rats and mice. Brain Res Mol BrainRes 117:1–7

Semino C, Angelini G, Poggi A, Rubartelli A (2005) NK/iDC interaction results in IL-18 secretion by DCs at thesynaptic cleft followed by NK cell activation andrelease of the DC maturation factor HMGB1. Blood106:609–16

Stros M, Ozaki T, Bacikova A, Kageyama H, Nakagawara A(2002) HMGB1 and HMGB2 cell-specifically down-regulate the p53- and p73-dependent sequence-specifictransactivation from the human Bax gene promoter.J Biol Chem 277:7157–64

Taguchi A, Blood DC, del Toro G, Canet A, Lee DC, Qu W,Tanji N, Lu Y, Lalla E, Fu C, Hofmann MA, Kislinger T,Ingram M, Lu A, Tanaka H, Hori O, Ogawa S, Stern DM,Schmidt AM (2000) Blockade of RAGE-amphoterinsignalling suppresses tumour growth and metastases.Nature 405:354–60

Teramoto T, Qiu J, Plumier JC, Moskowitz MA (2003) EGFamplifies the replacement of parvalbumin-expressingstriatal interneurons after ischemia. J Clin Invest111:1125–32

Thomas JO (2001) HMG1 and 2: architectural DNA-binding proteins. Biochem Soc Trans 29:395–401

Treutiger CJ, Mullins GE, Johansson AS, Rouhiainen A,Rauvala HM, Erlandsson-Harris H, Andersson U, YangH, Tracey KJ, Andersson J, Palmblad JE (2003) Highmobility group 1 B-box mediates activation of humanendothelium. J Intern Med 254:375–85

Tsung A, Sahai R, Tanaka H, Nakao A, Fink MP, Lotze MT,Yang H, Li J, Tracey KJ, Geller DA, Billiar TR (2005)The nuclear factor HMGB1 mediates hepatic injuryafter murine liver ischemia–reperfusion. J Exp Med201:1135–43

Veldhuis WB, Floris S, van der Meide PH, Vos IM, de VriesHE, Dijkstra CD, Bar PR, Nicolay K (2003) Interferon-beta prevents cytokine-induced neutrophil infiltrationand attenuates blood–brain barrier disruption. J CerebBlood Flow Metab 23:1060–9

Wang H, Bloom O, Zhang M, Vishnubhakat JM, Ombrelli-no M, Che J, Frazier A, Yang H, Ivanova S, BorovikovaL, Manogue KR, Faist E, Abraham E, Andersson J,Andersson U, Molina PE, Abumrad NN, Sama A,Tracey KJ (1999) HMG-1 as a late mediator of endotoxinlethality in mice. Science 285:248–51

Wang H, Czura CJ, Tracey KJ (2004) Lipid unites disparatesyndromes of sepsis. Nat Med 10:124–5

Ischemic neurons release HMGB-1J Qiu et al

937

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938

Wang X, Barone FC, Aiyar NV, Feuerstein GZ (1997)Interleukin-1 receptor and receptor antagonist geneexpression after focal stroke in rats. Stroke 28:155–61;discussion 161–162

Wang X, Lo EH (2003) Triggers and mediators of hemor-rhagic transformation in cerebral ischemia. Mol Neu-robiol 28:229–44

Wang X, Wang H, Xu L, Rozanski DJ, Sugawara T, ChanPH, Trzaskos JM, Feuerstein GZ (2003) Significantneuroprotection against ischemic brain injury byinhibition of the MEK1 protein kinase in mice:exploration of potential mechanism associated withapoptosis. J Pharmacol Exp Ther 304:172–8

Wu DC, Ye W, Che XM, Yang GY (2000) Activation ofmitogen-activated protein kinases after permanent

cerebral artery occlusion in mouse brain. J Cereb BloodFlow Metab 20:1320–30

Yang GY, Schielke GP, Gong C, Mao Y, Ge HL, Liu XH, BetzAL (1999) Expression of tumor necrosis factor-alpha andintercellular adhesion molecule-1 after focal cerebralischemia in interleukin-1beta converting enzyme defi-cient mice. J Cereb Blood Flow Metab 19:1109–17

Yang H, Ochani M, Li J, Qiang X, Tanovic M, Harris HE,Susarla SM, Ulloa L, Wang H, DiRaimo R, Czura CJ,Roth J, Warren HS, Fink MP, Fenton MJ, Andersson U,Tracey KJ (2004) Reversing established sepsis withantagonists of endogenous high-mobility group box 1.Proc Natl Acad Sci USA 101:296–301

Yang H, Wang H, Czura CJ, Tracey KJ (2005) The cytokineactivity of HMGB1. J Leukoc Biol 78:1–8

Ischemic neurons release HMGB-1J Qiu et al

938

Journal of Cerebral Blood Flow & Metabolism (2008) 28, 927–938