High mobility group 1 B-box mediates activation of human endothelium


Carl Johan Treutiger MD, PhD, Division of Infectious Diseases, Karolinska Institutet, Huddinge University Hospital, S-141 86 Stockholm, Sweden (fax: +46-8-746-62-80; e-mail: carl.johan.treutiger@medhs.ki.se).


Abstract Treutiger CJ, Mullins GE, Johansson A-SM, Rouhiainen A, Rauvala HME, Erlandsson-Harris H, Andersson U, Yang H, Tracey KJ, Andersson J, Palmblad JEW (Center for Infectious Medicine; Center for Inflammation and Haematology Research; Karolinska Institute at Huddinge University Hospital, Stockholm, Sweden, Finnish Red Cross Blood Transfusion Service; Institute of Biotechnology, University of Helsinki; Helsinki, Finland, Center for Molecular Medicine, Karolinska Institutet, Stockholm, Sweden and North Shore-Long Island Jewish Research Institute, NY, USA). High mobility group 1 B-box mediates activation of human endothelium. J Intern Med 2003; 254: 375–385.

Objectives. Severe sepsis and septic shock is a consequence of a generalized inflammatory systemic response because of an invasive infection that may result in acute organ dysfunction. Mortality is high despite access to modern intensive care units. The nuclear DNA binding protein high mobility group 1 (HMGB1) protein has recently been suggested to act as a late mediator of septic shock via its function as a macrophage-derived pro-inflammatory cytokine (J Exp Med 2000; 192: 565, Science1999; 285: 248). We investigated the pro-inflammatory activities of the A-box and the B-box of HMGB1 on human umbilical venular endothelial cells (HUVEC).

Design. The HUVEC obtained from healthy donors were used for experiments. Recombinant human full-length HMGB1, A-box and B-box were cloned by polymerase chain reaction (PCR) amplification from a human brain quick-clone cDNA. The activation of HUVEC was studied regarding (i) upregulation of adhesion molecules, (ii) the release of cytokines and chemokines, (iii) the adhesion of neutrophils to HUVEC, (iv) the activation of signalling transduction pathways and (v) the involvement of the receptor for advanced glycation end-products (RAGE).

Results. The full-length protein and the B-box of HMGB1 dose-dependently activate HUVEC to upregulate adhesion molecules such as ICAM-1, VCAM-1 and E-selectin and to release IL-8 and G-CSF. The activation of HUVEC could be inhibited to 50% by antibodies directed towards the RAGE. HMGB1-mediated HUVEC stimulation resulted in phosphorylation of the ELK-1 signal transduction protein and a nuclear translocation of p65 plus c-Rel, suggesting that HMGB1 signalling is regulated in endothelial cells through NF-κB.

Conclusions. The HMGB1 acts as a potent pro-inflammatory cytokine on HUVEC and the activity is mainly mediated through the B-box of the protein. HMGB1 may be a key factor mediating part of the pro-inflammatory response occurring in septic shock and severe inflammation.


Severe sepsis and septic shock caused by an invasive infection, generate a systemic inflammatory response, including hyperactivation of complement and cytokines. This may result in acute organ dysfunction with 40% mortality in western countries [1]. Mortality is still high despite advances in therapy. Bacterial toxins such as endotoxins from Gram-negative bacteria and exotoxins from Gram-positive bacteria, bind selectively to Toll-like receptors, which activate the innate immune system. This leads to an immediate pro-inflammatory response combined with a hypercoagulation state. Dysfunction of the vascular endothelium is a critical component in the development of severe sepsis and septic shock. In addition to bacterial toxins, pro-inflammatory cytokines bind to specific endothelial surface receptors causing further release of cytokines and tissue factors, upregulation of adhesion receptors, adhesion and transmigration of leucocytes. Taken together, this cascade of endothelial cell surface-associated events are linked to microvascular dysfunction, thrombosis and multiple organ failure.

The high mobility group 1 (HMGB1) is a highly conserved protein with >99% amino acid homology between human beings and rodents [2]. It is known as a nonhistone DNA binding protein, involved in stabilization of nucleosome formation [3], facilitating gene transcription [4] and modulating steroid hormone receptors [5]. HMGB1 comprises three distinct domains. The two DNA-binding elements, termed A-box and B-box, each of which is made of approximately 80 amino acid residues. The third domain is the strongly negatively charged C-terminal. HMGB1 has recently been demonstrated to act as a late mediator of lethal endotoxaemia [6], a mediator of acute lung injury [7] and as a pro-inflammatory cytokine [8]. Serum levels of HMGB1 have been directly associated with mortality in patients with lethal sepsis [6], suggesting that HMGB1 may be a crucial member of the uncontrolled pro-inflammatory response associated with fatal outcome. HMGB1 can be released from cells of the macrophage/monocyte lineage following activation by pro-inflammatory stimuli [9]. The protein is also released by cells undergoing a necrotic cell death pathway [10] but not by cells undergoing apoptosis [11]. The receptor for advanced glycation end-products (RAGE) [12] has been suggested to be a high-affinity receptor for HMGB1. RAGE is a potent signalling receptor that activates parallel signalling pathways, affecting gene expression through NF-κB and cell migration through activation of the GTPases Rac and Cdc 42 [13]. As the endothelium plays a crucial pathophysiological role in the development of severe sepsis and septic shock, the role of HMGB1 in regulation of endothelial activation has been assessed. This study describes the effects of HMGB1 on human umbilical venular endothelial cells (HUVEC).

Materials and methods


Recombinant human HMGB1, 651 base pairs (bp), was cloned by polymerase chain reaction (PCR) amplification from a human brain quick-clone cDNA (Clontech, Palo Alto, CA, USA) using the following primers: forward primer: 5-GATGG GCAAA GGAGA TCCTA AG-3 and reverse primer: 5GCGGC CGCTT ATTCA TCATC ATCAT CTTC-3. Cloning and protein purification of human HMGB1 mutants: truncated form of human HMGB1 was cloned by PCR amplification from a human brain quick-clone cDNA (Clontech). The primers used were: carboxyl terminus-truncated mutant (557 bp): 5-GATGG GCAAA GGAGA TCCTA AG-3 and 5-GCGGC CGCTC ACTTG CTTTT TTCAG CCTTGAC-3; B-box (233 bp): 5-AAGTT CAAGG ATCCC AATGC AAAG-3 and 5-GCGGC CGCTC AATAT GCAGC TATATC CTTTTC-3; A-box (261 bp): 5-GATGG GCAAA GGAGA TCCTAAG-3 and 5-TCACT TTTTT GTCTC CCCTT TGGG-3. A stop codon was added to each mutant to ensure the accuracy of protein size. PCR products were sub-cloned into pCRII-TOPO vector EcoRI sites using TA cloning method as per manufacturer's instruction (Invitrogen, Carlsbad, CA, USA). After amplification, the PCR product was digested with EcoRI and sub-cloned into an expression vector (pGEX) with a glutathionine-S-transferase (GST) tag (Pharmacia, Piscataway, NJ, USA); correct orientation in positive clones were confirmed by DNA sequencing of both strands. The recombinant plasmids were transformed into protease-deficient Escherichia coli strains, BL21 (Novagen, Madison, WI, USA), and fusion protein expression was induced by 1 mm isopropyl-d-thiogalactopyranoside (IPTG) for 3 h. Recombinant proteins were obtained using affinity purification with the glutathione sepharose resin column (Pharmacia). A GST vector, lacking HMGB1 protein, was included as a control. The protein eluates were passed over a polymyxin B column (Pierce, Rockford, IL, USA) to remove any contaminating lipopolysaccharide (LPS) and dialysed extensively against phosphate-buffered saline (PBS) to remove reduced gluthatione, lyophilized and re-dissolved in sterile water before use. LPS content in the different HMGB1 preparations was 219 pg LPS μg−1 (HMGB1), 56 pg LPS μg−1 (carboxyl terminus truncated), 19 pg LPS μg−1 (B-box), 16 pg LPS μg−1 (A-box) and 4 pg LPS μg−1 (vector alone) as measured using the Limulus amoebocyte lysate assay (Bio Whittaker Inc., Walkersville, MD, USA). Integrity of the protein was verified by sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS–PAGE) with Coomassie blue staining.

Chemicals and antibodies

In this study, the following chemicals were used: LPS from E. coli serotype O55:B5 (Sigma Chemical, St. Louis, MO, USA); endothelial cell growth factor (Collaborative Research, Waltham, MA, USA); foetal bovine serum, HEPES, penicillin, streptomycin, RPMI-1640, PBS and Hanks’ balanced salt solution (HBSS; Life Technologies, San Diego, CA, USA); collagenase (type 3; Worthington, Lakewood, NJ, USA); polystyrene plates (96 and 24 well) and other tissue culture plastic materials (Techno Plastic Products (TPP)). The following antibodies were also used: monoclonal antibodies to E-selectin (33361 A; PharMingen, San Diego, CA, USA); VCAM-1 (33351 A; PharMingen); ICAM-1 (84H10; Serotec, Raleigh, NC, USA) and human leucocyte antigen (HLA; DAKO, Copenhagen, Denmark). Rabbit polyclonal antibodies to human NF-κB subunits c-Rel and p65 were purchased from Calbiochem (San Diego, CA, USA), horseradish peroxidase (HRP)-conjugated goat antibodies to mouse immunoglobulin (IgG) from Vector (Burlingame, CA, USA), HRP-conjugated antibodies to rabbit IgG from DAKO, dexamethasone from Apoteket AB (MSD Sweden, Sollentuna, Sweden) and trypsin from Sigma Chemical. HMGB1 and LPS were incubated with the trypsin solution [8]. The monoclonal mouse antibodies directed against phosphorylated ELK-1, MAPK and C-jun were purchased from Cell Signaling Technology (Beverly, MA, USA). Synthetic peptides based on RAGE sequences in Ig domains V1 and C1, P300: CNTGRTEAWKVLSPQG and P301: CRPLNTAPIQPRVRE were purchased from Peptide Laboratory of Department of Biochemistry, University of Helsinki, Helsinki, Finland. Antisera against peptides were produced in rabbits and antibodies were affinity purified using peptide column [14]. Neutralizing goat polyclonal antibodies were isolated from anti-RAGE antiserum (Chemicon, Temecula, CA, USA) using protein-G agarose according to the manufactures protocol (Upstate Biotechnology, Lake Placid, NY, USA). A goat polyclonal antibody directed against PECAM-1 was purchased from Chemicon. The tetravalent guanylhydrazone CNI-1493, CAS Reg. No. 154301-51-3, was synthesized and purified [15]. The purity was greater than 98% as estimated by melting point, nuclear magnetic resonance (NMR), elution from high-performance liquid chromatography and elemental analyses. A stock solution was prepared in sterile, deionized, LPS-free water. The macrophage deactivating agent CNI-1493, which is currently undergoing phase II clinical trial for the treatment of inflammatory bowel disease, was tested for its possible suppressive effect of HMGB1 pro-inflammatory activity. CNI-1493 inhibits the release of early cytokines, such as TNF-α and IL-1β, by preventing the phosphorylation of p38 MAP kinase [16]. Recombinant human TNF-α was purchased from PharMingen.

Endothelial cell culture and neutrophil isolation

The HUVEC were obtained [17] and used for experiments as confluent monolayers in the second or third passage. Neutrophils were isolated from healthy donors by one-step discontinuous gradient centrifugation on Percoll (Amersham Pharmacia Biotech, Piscataway, NJ, USA) [18]. Purified neutrophils (>95% purity and viability) were loaded with 2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester (BCECF-AM; Molecular Probes, Eugene, OR, USA) for assessment of adherence [19], with the modification that all assays were performed serum-free with the media supplemented with 1% human serum albumin (Amersham Pharmacia Biotech) instead of foetal calf serum.

Cytokine production by HUVEC

After incubation of HUVEC with LPS or HMGB1 for 4 or 16 h, culture supernatants and cells were harvested. The concentrations of cytokines (IL-1β, TNF-α, IL-8 and G-CSF) were determined using Quantikine assays (R&D Systems, Minneapolis, MN, USA).

Expression of cell adhesion molecules

The cell surface expression of E-selectin, VCAM-1 and ICAM-1 were examined using a modified cellular enzyme-linked immunosorbent assay (ELISA) [20]. After incubation with the various stimuli for the indicated time points, HUVEC were fixed with 0.1% para-formaldehyde and incubated for 1 h at room temperature with 5% nonfat dried milk in PBS. Adhesion molecules were labelled with specific monoclonal antibodies (1 μg mL−1) for 2 h and then 1 h with HRP-conjugated secondary antibodies. The cells were washed and immune complexes were then detected by incubation with 5-thio-2-nitrobenzoic acid (TNB) peroxidase substrate (Bio-Rad, Hercules, CA, USA) for 30 min, after which the reaction was terminated with sulphuric acid, and absorbance was measured at 450 nm. Data are expressed in arbitrary units and as a percentage of the level of adhesion molecule expression in relation to nonstimulated HUVEC.

Neutrophil adherence

Adherence of neutrophils to stimulated HUVEC monolayers was assessed [21]. After stimulation and washing, HUVEC were exposed to BCECF-loaded neutrophils for 10 min. Nonadherent neutrophils were then removed, and the fluorescence of the adherent cells was measured. The number of adherent cells as the percentage of the total number of added neutrophils has been discussed in this study. The contribution of possible interfering factors, such as neutrophil aggregation, to the adherence assay has been evaluated [21].

Immunostaining of NF-κB and c-Rel

The staining of NF-κB and c-Rel was based on a previously described method [22]. HUVEC grown on glass cover-slips to confluence, were stimulated with either LPS or HMGB1 for indicated time points, fixed in methanol, permabilized with acetone and exposed to 3% H2O2. The cells were incubated with 1.5% normal goat serum for 20 min and then with either antibodies to NF-κB subunit p65 or against c-Rel. Immunocomplexes were detected with a Vectastain avidin–biotin complex (ABC) kit and 4-dimethylamino-azobenzene (DAB) substrate for peroxidase (Vector). Micrographs were scanned with a Jandel SigmaScan Pro instrument (Jandel Scientific Software, San Rafael, CA, USA) for densitometric assessment of the ratio of staining between the nucleus and cytoplasm. More than 200 cells were analysed on every micrograph.

Western blotting of intracellular signalling proteins

The HUVEC (5 × 106) were harvested and cytosolic and nuclear extracts were prepared from HUVEC using a modified mini-extraction protocol [23]. Portions of cytoplasmic and nuclear extracts were fractionated by 12% SDS–PAGE. The separated proteins were transferred to a nitrocellulose membrane (Bio-Rad) and were consecutively stained with antibodies directed against the different indicated proteins, after which HRP-conjugated secondary antibodies were applied and immune complexes were detected with an enhanced chemiluminescence (ECL) kit (Amersham Pharmacia Biotech), according to the protocol provided by the manufacturer. Phosphorylated ELK-1 was recorded using a 1 : 1000 dilution of antibodies directed against phosphorylated ELK-1 (Cell Signaling Technology). Staining of β-actin (Cell Signaling Technology) was used as a loading control.

Assessment of cell viability

The viability of HUVEC was assessed before experiments or after incubation with the various stimuli. Less than 5% of cells incubated with LPS (100 ng mL−1) or HMGB1 (1–10 μg mL−1) for 16 h exhibited an altered morphology or uptake of trypan blue, indicating that membrane integrity was not disrupted. Consistent with this conclusion, the amount of lactate dehydrogenase released into culture supernatants was not increased by experimental treatments (data not included).

Statistical analysis

Data are presented as mean ± SEM for the indicated number of separate experiments. All analyses are based on more than three separate experiments made in at least triplicate performed with HUVEC or neutrophils from different donors. The data were analysed with use of the Statistica software package (Stat Soft, Tulsa, OK, USA) using paired t-test, and differences were determined to be statistically significant at P < 0.05.


HMGB1-induced cytokine production

Quiescent HUVEC released only restricted amounts of IL-8 and G-CSF after 4 and 16 h (Fig. 1a–d) (4 h: 1770 ± 410 and 29 ± 20 pg mL−1; 16 h: 3316 ± 696 and 93 ± 79 pg mL−1, respectively). HMGB1 (1 μg mL−1) exposure of HUVEC led to a release of both G-CSF and IL-8 (Fig. 1a–d). However, levels of the cytokines were consistently lower when compared with those following LPS stimulation. In the case of IL-8, the values for 4 h rose 6.6-fold for LPS compared with a 2.3-fold rise for HMGB1. The corresponding values for 16 h were a 27.8-fold rise for LPS versus a ninefold rise for HMGB1. The HMGB1-dependent release of IL-8 was dose-dependent (Fig. 1e).

Figure 1.

G-CSF or IL-8 release from human umbilical venular endothelial cells (HUVEC). The graphs (a) to (b) illustrate HUVEC incubated with lipopolysaccharide (LPS; 100 ng mL−1), high mobility group 1 (HMGB1; 1 μg mL−1) or media only for either 4 or 16 h. The supernatants were collected and analysed for indicated cytokines. (a) G-CSF at 4 h; (b) G-CSF at 16 h, *P < 0.05; (c) IL-8 at 4 h, *P < 0.05; (d) IL-8 at 16 h, *P < 0.01. P-values are for comparison with cells not exposed to HMGB1. The graph (e) illustrates HUVEC incubated with increasing concentrations of HMGB1 (0.5–10 μg mL−1). The supernatants were collected and analysed for IL-8. Data are mean ± SEM of three separate experiments (in triplicate).

Expression of adhesion molecules

Neutrophil adhesion to HUVEC is mediated by several adhesion molecules including ICAM-1 and E-selectin, whereas VCAM-1 is of significance for other leucocytes. The surface expression of these receptors was analysed when HUVEC had been stimulated with HMGB1 or LPS for 4 or 16 h. Nonstimulated HUVEC expressed low levels of adhesion molecules at the surface at 4 and 16 h for ICAM-1 was 1.2 ± 0.04 and 0.8 ± 0.07; VCAM-1 was 0.05 ± 0.01 and 0.13 ± 0.01 and E-selectin was 0.08 ± 0.02 and 0.04 ± 0.01, respectively, all values being expressed as arbitrary units. Stimulation of HUVEC with HMGB1 for 4 h induced 1.7-,3- and 11-fold increases in the abundance of ICAM-1, VCAM-1 and E-selectin, respectively (Fig. 2a–f). Incubation with HMGB1 for 16 h also induced an increase in the surface expression of the investigated adhesion molecules, 1.9-, 4.2- and 2.7-fold increase of ICAM-1, VCAM-1 and E-selectin abundance, respectively (Fig. 2a–f). In concordance with the cytokine results, HMGB1-mediated upregulation of adhesion molecules were consistently lower than the upregulation induced by LPS (Fig. 2a–f).

Figure 2.

Surface adhesion molecule expression on human umbilical venular endothelial cells (HUVEC): the bars depict results for HUVEC incubated with either lipopolysaccharide (LPS; 100 ng mL−1), high mobility group 1 (HMGB1; 1 μg mL−1) or media only for either 4 or 16 h. HUVEC were then analysed for surface expression of indicated adhesion molecule. (a) ICAM-1 at 4 h, *P < 0.0001; (b) ICAM-1 at 16 h, *P < 0.001; (c) VCAM-1 at 4 h, *P < 0.01; (d) VCAM-1 at 16 h, *P < 0.001; (e) E-selectin at 4 h, *P < 0.01; (f) E-selectin at 16 h, *P < 0.05. P-values are for comparison with cells incubated with media alone. Data are mean ± SEM of three separate experiments (in triplicate).

HMGB1 induced neutrophil adhesion to HUVEC

Treatment of HUVEC with HMGB1 led to activation of the endothelial cells and a dose-dependent increase of neutrophil adhesion (Fig. 3a). The two boxes of HMGB1 (A-box and B-box) were tested for their ability to stimulate HUVEC. The B-box protein of HMGB1 efficiently upregulated HUVEC at 1 μg mL−1 (22.5-fold increase compared with vector alone), whereas the A-box protein only gave a very weak nonsignificant upregulation of neutrophil adhesion at this concentration (1.3-fold increase compared with vector alone) (Fig. 3a). A dose-dependent increase was evident with both the boxes, but still at 10 μg mL−1 the B-box was a stronger inducer of neutrophil adhesion (34.6-fold increase compared with vector alone), whereas the A-box (ninefold increase compared with vector alone) (Fig. 3a). Pretreatment of HMGB1 with trypsin [8] eliminated bioactivity, but activity of HMGB1 was unaffected by addition of polymyxin B (1 μg mL−1; data not included). The kinetics of HMGB1-induced neutrophil adhesion to HUVEC was studied: low-level neutrophil adhesion occurred at 3 h (3.75-fold increase compared with vector alone) and remained for at least 12 h, then neutrophil adhesion increased further (15-fold increase at 14 h). Peak of adhesion was observed at 16 h (17.5-fold increase); subsequently, it declined slowly (11.6-fold increase at 24 h) (Fig. 3b). Neutrophil adhesion persisted for 48 h (data not included). In contrast, when HUVEC were stimulated with LPS (100 ng mL−1), an initial peak of neutrophil adhesion at 3 h (42.5-fold increase compared with media alone), which then declined until 14 h (16.2-fold increase), was observed. A second peak of neutrophil adhesion was apparent at 16 h (27.5-fold increase), and later declined slowly (Fig. 3b). In order to determine whether the increase of neutrophil adhesion was mediated by a HMGB1-induced IL-8 release, a neutralizing monoclonal anti-IL-8 antibody was added to HUVEC prior to stimulation with HMGB1. However, this antibody had no effect on the upregulation of neutrophil adhesion (data not included).

Figure 3.

Neutrophil adhesion to human umbilical venular endothelial cells (HUVEC). (a) HUVEC were incubated with increasing concentrations (0.1–10 μg mL−1) of high mobility group 1 (HMGB1) ▮, B-box □ or A-box bsl00047 for 16 h, whereafter neutrophils were added to the activated HUVEC. Adhesion of neutrophils to the HUVEC were analysed and expressed as the percentage of total number of added neutrophils adhering to the HUVEC. (b) A kinetic assessment of neutrophil adhesion after either lipopolysaccharide (LPS; 100 ng mL−1) ▴ or HMGB1 (1 μg mL−1) ▮ stimulation of HUVEC. Neutrophils were added to the HUVEC at the indicated time points. Adhesion of neutrophils to the HUVEC were analysed and expressed as the percentage of total number of added neutrophils adhering to the HUVEC. (c) HUVEC incubated with HMGB1 (1 μg mL−1) either together with purified antibodies from a goat anti-RAGE antiserum ▮ or antibodies from a goat anti-PECAM-1 antiserum □ in increasing concentrations for 16 h. Adhesion of neutrophils to the activated HUVEC were analysed and expressed as relative adhesion compared to incubation with nonimmune goat IgG (20 μg mL−1) bsl00047. #P < 0.005, *P < 0.0001, for the comparison with nonimmune goat IgG. Data are mean ± SEM of three separate experiments (in triplicate).

Inhibition of HMGB1 stimulation by antibodies directed towards RAGE

In order to investigate whether HMGB1 activation of endothelial cells was mediated by RAGE, HUVEC were stimulated with 1 μg mL−1 of HMGB1 together with increasing concentrations of antibodies directed against RAGE (either affinity-purified neutralizing antibodies from a RAGE antiserum or the antibodies P300 and P301) for 16 h. All three antibodies inhibited activation of HUVEC. Neutrophil adhesion to the activated HUVEC was inhibited in a dose-dependent manner (46% of control at 20 μg mL−1; P < 0.0001) (Fig. 3c; data not shown for P300 and P301). Goat antisera directed towards PECAM-1 (Fig. 3c) and a mouse monoclonal antibody directed against HLA were used as control antibodies (data not included) and neither affected neutrophil adhesion. The anti-RAGE antibodies had no effect on LPS or TNF-α stimulation (data not included). The HMGB1 induced release of the cytokine IL-8 from HUVEC could also be inhibited by the anti-RAGE in a dose-dependent manner (data not included). This demonstrated that the pro-inflammatory activities of HMGB1 were partly initiated via RAGE.

NF-κB and c-Rel translocation in HUVEC

The HUVEC were stimulated with HMGB1 for 5 min to 4 h before the nuclear translocation of the p65 subunit of NF-κB or c-Rel was assessed by immunostaining. In nonstimulated HUVEC, p65 was detected in the cytoplasm (Fig. 4a). LPS stimulation led to a translocation of p65 from the cytoplasm to the nucleus after approximately 15 min (data not included), whereas that induced by HMGB1 could be detected at 30 min. HMGB1 induced an 8.3-fold enhancement of nuclear intensity compared to HUVEC stimulated with vector alone (Fig. 4a). A similar delay of translocation was evident in c-Rel, with a start at 1 h with LPS, compared with 2 h using HMGB1 (data not included).

Figure 4.

NF-κB translocation and ELK-1 phosphorylation in human umbilical venular endothelial cells (HUVEC). (a) HUVEC were incubated with high mobility group 1 (HMGB1; 1 μg mL−1) for 0.5 h or vector alone and for cellular location of the P65 subunit of NF-κB was analysed. Micrographs were scanned with a Jandel SigmaScan Pro instrument for densitometric assessment of the ratio of staining between the nucleus and cytoplasm. The graph depict the increase of nuclear intensity, *P < 0.05. (b) Immunoblot analysis of phosphorylation ELK-1. HUVEC were incubated with HMGB1. Extracts were prepared as described in Materials and Methods and subjected to immunoblot analysis with antibodies to phosphorylated ELK-1. The graph demonstrates the increase of intensity, *P < 0.05. Data are mean ± SEM of three separate experiments.

Signal transduction pathways

In order to explore possible involved signal transduction pathways, induced by HMGB1 stimulation of HUVEC, the intracellular signalling proteins C-jun, MAPK and ELK-1 by Western blotting was analysed. No phosphorylation of C-jun or MAPK intracellular signalling protein could be detected (data not included). However, a clear phosphorylation of ELK-1 was recorded (8.2-fold induction of intensity compared to HUVEC stimulated with vector alone), indicating a role for this signalling protein in the HMGB1-induced activation of HUVEC (Fig. 4b). CNI-1493 was then used to inhibit the p38 MAP kinase. However, CNI-1493 had no effect on HMGB1-mediated neutrophil adhesion to HUVEC and no effect on expression of ICAM-1 or VCAM-1 at 4 or 16 h, although a slight decrease of E-selectin expression at 16 h was observed (data not included).


The HMGB1 is a newly described cytokine with remarkable properties. It consists of two boxes (A and B), of which the A-box comprises the first 98 amino acids and the B-box the following 87. There is no leading sequence typical for intracellular transportation via the golgi apparatus and the endoplasmic reticulum. HMGB1 was initially described as a nuclear protein of significance for DNA stabilization as well as having transcriptional activity. It was subsequently demonstrated to act as a late mediator of septic shock and as a pro-inflammatory cytokine [6, 8]. Macrophages (and pituitary cells) release HMGB1 in response to LPS [9] and it has also been demonstrated that HMGB1 can be released from cells when membrane integrity is disturbed by increased permeabilization or necrosis [10, 11, 24, 25], whereas apoptosis led to covalent binding of HMGB1 to chromatin and, subsequently, a lack of inflammatory signalling [11]. This may be coupled to inflammatory reactions associated with tissue breakdown. The present study focuses on interactions of HMGB1 with the endothelium, as these cells are crucial for acute (and chronic) inflammatory responses including sepsis with organ failure. Exogenous HMGB1 causes release of cytokines from HUVEC and enhances neutrophil adhesion to HUVEC. This is partly mediated via RAGE and NF-κB-dependent steps and surface expression of adhesion molecules.

The RAGE is a member of the Ig superfamily of receptors and expressed on endothelial cells and various other cell types [26, 27]. RAGE has been demonstrated to act as a receptor for HMGB1 [12]. The adhesion of neutrophils is inhibited to HUVEC by 50% using either neutralizing antibodies from RAGE antiserum or either of two antibodies (P300 and P301) directed to Ig-domains V1 or C1 of the RAGE receptor, suggesting that RAGE is of significance for HMGB1-dependent endothelial responses, but that additional recognition mechanisms are also present. The pathways downstream of the putative receptors were determined to involve phosphorylation of ELK-1 by HMGB1 exposure without contribution of C-jun and MAPK. In addition, CNI-1493 an inhibitor of the p38 MAP kinase failed to block the ability of HUVEC to adhere neutrophils or to upregulate adhesion molecules. Furthermore, HMGB1 signalling involved translocation of the NF-κB subunits p65 and c-Rel from the cytoplasm to the nucleus. This phenomenon might be important for transcription of genes coding for pro-inflammatory cytokines, including IL-8 and G-CSF, and for induction of E-selectin, ICAM-1 and VCAM-1. Inhibition of NF-κB by dexamethasone blocked the HMGB1 induced IL-8 release (G. E. Mullins, J. Sundén-Cullberg, A.-S. M. Johansson, H. Erlandsson-Harris, U. Andersson, H. Yang, K. J. Tracey, J. Andersson, J. E. W. Palmblad and C. J. Treutiger, unpublished data). Although LPS was a stronger inducer of most of the adhesion molecules, the effect of HMGB1 was more pronounced on VCAM-1 at the early time point (i.e. 4 h). Effects of the two mediators varied with time but were rather comparable. Although the same effector molecules seems to be upregulated by LPS or HMGB1 activates target cells to part through different signalling systems, i.e. HMGB1 induces septic shock in LPS-resistant mice [6]. The anti-RAGE antibodies did not affect the LPS-stimulated increase of neutrophil adhesion or release of IL-8 (data not included).

Neutrophil adhesion to HMGB1-activated HUVEC was dose-dependently enhanced at a late time point (16 h), whereas the response to LPS was biphasic, with peaks at 4 and 16 h. However, when either the translocation of NF-κB or the release of cytokines were studied, and such a pronounced delay was not seen, suggesting that other mechanisms may explain the observed delay in neutrophil adhesion. Virtually all HMGB1 activity resided in the B-box. This points to different biological properties for the different boxes in some situations, but not all [24, 28]. When concentrations of the A-box were increased, some pro-inflammatory activity was evident, but it remained much lower compared with that associated with the B-box. The secretion of G-CSF and IL-8 from HMGB1-stimulated HUVEC implicates a role for enhanced myelopoeisis emigration of neutrophils from the bone marrow and subsequently neutrophilia. The generation of IL-8 may also be a factor for neutrophil adhesion to HUVEC and traffic into the tissues. A recent report by Fiuza et al. [29] demonstrates that HMGB1 activates human microvascular endothelial cells which further strengthens the evidence for the importance of HMGB1 as a cytokine.

In conclusion, this report illustrates that HMGB1 activates a number of pro-inflammatory properties of endothelial cell. Taken together with HMGB1 generation from and effects on macrophages, this cascade of rather late occurring events probably perpetuates the septic process and contributes to organ dysfunction in sepsis patients. Such a series of events may be abrogated by blocking the activities of HMGB1, opening up a possibility for new therapeutic approaches.

Conflict of Interest Statement

No conflict of interest declared.


This work was supported by grants from the Swedish Medical Research Council (JA: 14089-01A, JP: 05991-71X and CJT: 10850-08A), The Swedish Foundation for Strategic Research, NIH (A141536-04), Swedish League against Rheumatism, the foundations of Uggla, Karolinska Institutet, the Swedish Society for Medicine and King Gustav V: 80-year Fund. H. Rauvala and A. Rouhiainen were supported by grants from the Finnish Academy and the Sigrid Juselius Foundation. K. J. Tracey were supported by the National Institute of General Medicine. We thank Professor R. A. Harris for linguistic advice.