High mobility group box chromosomal protein 1 plays a role in the pathogenesis of rheumatoid arthritis as a novel cytokine

Authors


Abstract

Objective

High mobility group box chromosomal protein 1 (HMGB-1), a nuclear DNA binding protein, was recently rediscovered as a new proinflammatory cytokine. The purpose of this study was to demonstrate HMGB-1 expression in vivo and to identify the role of HMGB-1 in the pathogenesis of rheumatoid arthritis (RA).

Methods

HMGB-1 concentrations in synovial fluid (SF) and serum from RA and osteoarthritis (OA) patients were measured by immunoblot analysis. The protein's specific receptor, receptor for advanced glycation end products (RAGE), was examined in SF macrophages (SFMs). We measured levels of proinflammatory cytokines released by SFMs treated with HMGB-1 via enzyme-linked immunosorbent assay and used soluble RAGE (sRAGE) to block the release of tumor necrosis factor α (TNFα). Immunohistochemical analysis and immunofluorescence assay were employed to examine localization of HMGB-1 in RA synovium and its translocation in SFMs after TNFα stimulation.

Results

HMGB-1 concentrations were significantly higher in SF of RA patients than in that of OA patients. SFMs expressed RAGE and released TNFα, interleukin-1β (IL-1β), and IL-6 upon stimulation with HMGB-1. HMGB-1 was found in CD68-positive cells of RA synovium, and TNFα stimulation translocated HMGB-1 from the nucleus to the cytosol in SFMs. Blockade by sRAGE inhibited the release of TNFα from SFMs.

Conclusion

HMGB-1 was more strongly expressed in SF of RA patients than in that of OA patients, inducing the release of proinflammatory cytokines from SFMs. HMGB-1 plays a pivotal role in the pathogenesis of RA and may be an original target of therapy as a novel cytokine.

High mobility group box chromosomal protein 1 (HMGB-1) has 219 residues in its primary amino acid sequence, and there is >98% sequence identity between the HMGB-1 of rodents and that of humans (1–6). In most cells, HMGB-1 is located in the nucleus. It is an abundant, highly conserved cellular protein and is widely known as a nuclear DNA binding protein that stabilizes nucleosome formation (7, 8), facilitates gene transcription, and regulates the activity of steroid hormone receptors (9, 10). However, it has been reported that HMGB-1 might be translocated from the nucleus to the cytosol and then released extracellularly.

A previous study demonstrated that extracellular HMGB-1 induces the production of proinflammatory cytokines in macrophages (11). When released by activated monocytes, it participates in the development of lethality and activates downstream cytokine release. Furthermore, like other cytokine mediators of endotoxemia, HMGB-1 activates proinflammatory cytokine release from human monocytes (12). Membrane HMGB-1 signals neurite outgrowth by binding to the receptor for advanced glycation end products (RAGE) (13–17). RAGE is a transmembrane protein that belongs to the immunoglobulin superfamily and is expressed in other tissues, including vascular smooth muscle cells, neurons, and monocyte/macrophages (18–21). Activation of RAGE mediates induction of cellular oxidant stress and triggers a cascade of intracellular signals, including p21 Ras and other downstream targets, such as mitogen-activated protein kinase kinase, mitogen-activated protein kinase, and nuclear factor κB (16, 20).

It has been postulated that tumor necrosis factor α (TNFα) plays a clinical role in a cytokine cascade that results in joint inflammation and destruction in rheumatoid arthritis (RA), and that inhibition of TNF ameliorates experimental arthritis (22–25). However, this treatment has not been effective in all patients, and the severity of its side effects, which include increased susceptibility to infection and impaired vaccination response, has yet to be elucidated (24, 26). The finding that severe arthritis could progress even in a TNF knockout mouse model of collagen-induced arthritis indicates that more complex interactions between TNF, interleukin-1 (IL-1), and other inflammatory mediators, including a new cytokine, occur in vivo (27).

We show here that HMGB-1 plays a role in the pathogenesis of RA as a novel cytokine. We measured HMGB-1 concentrations in vivo in the synovial fluid (SF) and serum of RA and osteoarthritis (OA) patients and compared them. Localization of HMGB-1 in RA synovial tissue was investigated, and the specific receptor for HMGB-1, RAGE, was examined in SF macrophages (SFMs) obtained from RA patients. SFMs were stimulated with HMGB-1, and proinflammatory cytokines were measured at protein and messenger RNA (mRNA) levels. Furthermore, with TNFα stimulation, we examined translocation of HMGB-1 from the nucleus to the cytosol in SFMs and observed its protein and mRNA expression. Soluble RAGE (sRAGE) was employed to inhibit TNFα release from SFMs by binding to HMGB-1.

PATIENTS AND METHODS

Patients and samples.

Study patients with RA and control patients with OA were diagnosed according to the American College of Rheumatology classification criteria (28,29). SF and peripheral blood (PB) were obtained at the time of total knee arthroplasty or intraarticular injection. White blood cell (WBC) counts, C-reactive protein (CRP) levels, and erythrocyte sedimentation rates (ESRs) in PB were each measured upon collection. Samples were centrifuged at 1,500g for 10 minutes, and the supernatant was used as SF and serum. All RA patients were receiving nonsteroidal antiinflammatory drugs or various disease-modifying antirheumatic drugs, or both. Patients with OA exhibited no evidence of a systemic inflammatory response, as assessed by WBC counts, serum CRP levels, and ESRs (Table 1). All samples were obtained with informed consent.

Table 1. Clinical data for rheumatoid arthritis (RA) and osteoarthritis (OA) patients whose synovial fluid (SF) and peripheral blood (PB) were investigated in this study*
 RA patients (n = 30)OA patients (n = 30)
  • *

    Except where indicated otherwise, values are the mean ± SD. SF and PB were obtained at the time of total knee arthroplasty or intraarticular injection. White blood cell (WBC) counts, C-reactive protein (CRP) levels, and erythrocyte sedimentation rates (ESRs) in PB were each measured upon collection. OA patients exhibited no evidence of a systemic inflammatory response.

  • P < 0.01 versus OA patients.

Age, years59.2 ± 10.772.4 ± 7.6
No. of men/no. of women1/2911/19
Disease duration, years14.6 ± 10.66.5 ± 6.8
WBC count, ×103/liter7.7 ± 2.75.7 ± 1.5
CRP level, gm/dl3.3 ± 2.50.3 ± 0.4
ESR, mm/hour61.6 ± 25.914.7 ± 10.8

Cell preparation and culture.

Primary culture of SFMs was established from the SF obtained from RA patients, using the method of Hayes et al and Smith et al (30, 31). SF was diluted in RPMI 1640 (Nissui Pharmaceutical, Tokyo, Japan) containing 2% heat-inactivated fetal bovine serum (FBS), penicillin (100 units/ml), streptomycin (100 units/ml), and fungizone (0.25 μg/ml). Following 72 hours of culture in an atmosphere of humidified 5% CO2 at 37°C, nonadherent polymorphonuclear cells, lymphocytes, and monocytes were decanted, leaving adherent cells. These cells were used after 14 days in primary culture, during which time they developed from small adherent cells into larger macrophages. Adherent cells were collected and stained with fluorescein isothiocyanate (FITC)–labeled anti-CD14 antibody (Beckman Coulter, Fullerton, CA) to determine by fluorescence-activated cell sorting analysis whether the monocyte/macrophage population was predominant (>95%). We used von Willebrand factor as a marker for endothelial cells, which were not detected in SFMs by immunoblot analysis (results not shown).

Immunoblot analysis of HMGB-1.

SF and serum were fractionated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and HMGB-1 levels were determined by immunoblot analysis with reference to the standard curve for purified calf thymus HMGB-1 (a gift from Shino-Test Corporation, Kanagawa, Japan), obtained with serial dilutions. The concentrations of purified calf thymus HMGB-1 standard ranged from 0 ng/ml to 160 ng/ml. SF and serum were treated with lysis buffer and then loaded onto a 12.5% polyacrylamide gel and blotted onto a nitrocellulose membrane (Immobilon P; Millipore, Bedford, MA). The membrane was soaked in phosphate buffered saline containing 1% bovine serum albumin (PBS/BSA; Sigma-Aldrich, St. Louis, MO) at room temperature for 1 hour to effect blocking, followed by washing with PBS containing 0.05% Tween 20. The membrane was then incubated with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml; a gift from Shino-Test Corporation) and diluted with PBS/BSA overnight at 4°C. After washing, the membrane was incubated with horseradish peroxidase (HRP)–conjugated anti-rabbit IgG polyclonal antibody (Dako, Glostrup, Denmark) diluted 1:3,000 with PBS containing 3% BSA at room temperature for 2 hours.

After washing again, immunostaining was performed with enhanced chemiluminescence (ECL; Amersham Biosciences, Amersham, UK) according to the manufacturer's instructions. ECL-labeled bands were visualized on X-Omat XAR5 film (Eastman Kodak, Rochester, NY) after 10 minutes of exposure. HMGB-1 concentrations were quantified from the blots by measurement of optical band intensity using image analysis software (NIH Image, version 159; National Institutes of Health, Bethesda, MD) (online at http://rsb.info.nih.gov/nih-image/). HMGB-1 concentrations in the supernatant of SFMs after stimulation with recombinant TNFα (R&D Systems, Minneapolis, MN) were also determined in triplicate using the methods described above.

Immunohistochemical analysis of HMGB-1.

Sections of synovial tissue obtained from RA patients at the time of synovectomy were fixed in 10% neutral buffered formalin. Slides were deparaffinized, and endogenous peroxidase activity was blocked by treatment with 0.3% H2O2 blocking solution for 30 minutes, followed by blocking with 2% goat serum and incubation with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) diluted with PBS/BSA overnight at 4°C. After washing, slides were incubated with anti-rabbit biotinylated antibody diluted 1:200 with PBS and were then incubated with avidin–biotinylated enzyme complex (Vector, Burlingame, CA) and washed. The slides were incubated with diaminobenzidine substrate (Vector) and counterstained with Carrazzi's hematoxylin solution (Wako, Osaka, Japan). Rabbit IgG (1 μg/ml) diluted with PBS/BSA was used as a nonimmune control. Hematoxylin and eosin staining was performed at the same time.

Immunofluorescence assay of HMGB-1.

Sections of synovial tissue from RA patients were pretreated as described above. The slides were incubated with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) and anti-CD68 mouse monoclonal antibody for 1 hour at 25°C. After washing with PBS, the slides were incubated with FITC-labeled anti-rabbit IgG (Cappel Products, Irvine, CA) diluted 1:200 with PBS and rhodamine-labeled anti-mouse IgG (Cappel Products) diluted 1:200 with PBS for 1 hour at 25°C. In some cases, 4′,6-diamidino-2-phenylindole (Nacalai Tesque, Kyoto, Japan) was used to label the nucleus. Finally, the sections were washed and then examined with an Axioskop microscope (Carl Zeiss, Oberkochen, Germany).

To investigate the translocation of HMGB-1 in SFMs, SFMs were cultured in a Lab-Tech chamber slide (Nalge Nunc International, Cambridge, MA) coated with gelatin and then stimulated with recombinant TNFα (100 ng/ml). After stimulation, cells were fixed with 4% paraformaldehyde solution containing 0.2% Nonidet P40 (Sigma-Aldrich) for 5 minutes. The chamber was then incubated with a blocking buffer containing 1% BSA and PBS–0.02% Tween (PBST) for 1 hour at 25°C and incubated for 30 minutes with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) and anti-CD68 mouse monoclonal antibody at 25°C. The chamber was then washed with PBST, followed by incubation with FITC-labeled anti-rabbit IgG diluted 1:50 with PBST and rhodamine-labeled anti-mouse IgG diluted 1:50 with PBST for 20 minutes. Finally, the chamber was washed and then examined as described above.

Immunoblot analysis of RAGE.

SFMs established from RA patients and human brain extract were lysed immediately by sonication in SDS sample buffer (62.5 mM Tris HCl (pH 6.8), 2% SDS, 10% glycerol, and 0.002% bromphenol blue) and boiled for 5 minutes. About 50 μg of the cell lysates was resolved by SDS-PAGE (10%) and then transferred onto a nitrocellulose membrane (Amersham Biosciences). After the transfer, the membrane was blocked by incubation with 5% nonfat dry milk in TBST (0.1% Tween 20 in Tris buffered saline, pH 7.4) for 1 hour at room temperature. The membrane was then incubated with anti-RAGE antibody (Santa Cruz Biotechnology, Santa Cruz, CA) diluted 1:500 with TBST containing 2.5% nonfat dry milk for 1 hour at room temperature. Next, the membrane was washed and incubated with HRP-conjugated anti-goat IgG polyclonal antibody (Santa Cruz Biotechnology) diluted 1:2,500 with TBST containing 1% nonfat dry milk for 1 hour at room temperature. After washing, immunoreactive bands were visualized with an ECL detection system.

Measurement of cytokines.

We measured cytokine concentrations by enzyme-linked immunosorbent assay (ELISA) using commercially available kits. The TNFα, IL-1β, and IL-6 kits were purchased from R&D Systems. SF samples were diluted 1:5 in 20 mM Tris HCl (pH 7.4), 1% BSA, and 10 mM EDTA (32). HMGB-1 was added to the cultures of SFMs (established using the methods described above) in 24-well plates (Iwaki, Tokyo, Japan) with RPMI 1640 containing 2% FBS. The concentrations of TNFα, IL-1β, and IL-6 released into conditioned supernatant were measured in triplicate by ELISA. Soluble RAGE (a gift from Toray, Shiga, Japan) was incubated with HMGB-1 (500 ng/ml) at 25°C for 1 hour, and then the mixture was added to the cultured SFMs with the medium described above for 6 hours, followed by measurement of TNFα in the supernatant by ELISA.

Northern blot analysis of TNFα.

We determined expression of TNFα mRNA using Northern blot analysis. SFMs were stimulated with HMGB-1 (100 ng/ml), and total RNA was isolated from them with TRIzol reagent (Gibco BRL, Gaithersburg, MD). Total RNA (10 μg) was electrophoresed on a standard Northern gel and transferred to a nylon membrane (Bio-Rad, Hercules, CA). TNFα complementary DNA (cDNA) probes spanning the amino-terminal region to avoid cross-hybridization were radiolabeled to a specific activity using α32P-dCTP and a random primer labeling system (Takara, Tokyo, Japan). Blots were identified on Kodak film exposed at −20°C for 4 hours using intensifying screens.

Reverse transcription–polymerase chain reaction (RT-PCR) analysis of HMGB-1.

We determined expression of HMGB-1 mRNA by RT-PCR. SFMs were stimulated with TNFα (100 ng/ml), and total RNA was isolated from them with the methods described above, followed by synthesis of cDNA from 1 μg of total RNA with a cDNA synthesis system kit (Gibco BRL). The primers used in the PCR reaction were 5′-GAAGCACCCAGATGCTTCAG-3′ (sense) and 5′-ATAACGGGCCTTGTCCGCCT-3′ (antisense) for HMGB-1 (National Center for Biotechnology Information [NCBI]/GenBank accession no. NM002128), and 5′-GCCAAGTATGATGACATCAA-3′ (sense) and 5′-CCATATTCATTGTCATACCA-3′ (antisense) for GAPDH (33), which served as an internal control. The use of the above sequences from the NCBI as primers for HMGB-1 has not been reported previously. The predicted lengths of the HMGB-1 and GAPDH amplification products were 127 bp and 202 bp, respectively. The cDNA was amplified after determining the optimal number of cycles. The mixture was first incubated for 3 minutes at 95°C, then cycled 28 times at 95°C for 1 minute and 59°C for 1 minute, and finally elongated at 72°C for 2 minutes. This format allowed optimal amplification with little or no nonspecific amplification of contaminating DNA. The amplified products were separated on 3% agarose gels containing 1 μg/ml of ethidium bromide and were visualized and photographed using ultraviolet transillumination.

Statistical analysis.

We analyzed in vivo data using the Mann-Whitney U test for unpaired samples, since these variables demonstrated nonparametric distributions. Other data were analyzed with Student's t-test for unpaired values to compare differences between stimulated and control cultures. P values less than 0.05 were considered significant. Values are expressed as the mean ± SD.

RESULTS

HMGB-1 and TNFα levels in SF and HMGB-1 levels in serum of RA and OA patients.

Using immunoblot analysis, we measured the HMGB-1 concentrations in serum and SF obtained from RA and OA patients. We could detect expression of HMGB-1, a 30-kd protein, in the SF of RA patients, but not in that of OA patients. HMGB-1 was barely detected in the serum of patients from either group (Figure 1). We found that the mean ± SD HMGB-1 concentration in the SF of 30 RA patients was significantly higher than that in the SF of 30 OA patients (54.1 ± 73.0 ng/ml versus 12.0 ± 17.7 ng/ml; P < 0.01) (Figure 2A). The concentration of HMGB-1 was very low in the serum of patients in both groups, with no significant difference between the groups (RA 1.5 ± 4.2 ng/ml versus OA 1.8 ± 3.9 ng/ml; P = 0.54) (Figure 2B). TNFα levels in SF were analyzed by ELISA in the same patients and were also significantly higher in RA patients than in OA patients (29.2 ± 33.0 pg/ml versus 7.5 ± 20.5 pg/ml; P < 0.01) (Figure 2C).

Figure 1.

High mobility group box chromosomal protein 1 (HMGB-1) concentrations in synovial fluid (SF) and serum obtained from patients with rheumatoid arthritis (RA) and osteoarthritis (OA). HMGB-1 concentrations were determined by immunoblot analysis. HMGB-1, a 30-kd protein, was expressed in the SF of RA patients but not in that of OA patients (A). In serum, there was little expression of HMGB-1 in either group (B). Shown are representative results comparing SF and serum levels of HMGB-1 in 3 RA and 2 OA patients. Lanes on the left indicate the range of standard HMGB-1 concentrations (0–160 ng/ml).

Figure 2.

Quantification of HMGB-1 and tumor necrosis factor α (TNFα) levels in SF and HMGB-1 levels in serum of RA and OA patients. Bars show the means. In SF, HMGB-1 concentrations in RA patients were significantly higher than those in OA patients (mean ± SD 54.1 ± 73.0 ng/ml versus 12.0 ± 17.7 ng/ml; ∗ = P < 0.01 versus OA patients) (A). In serum, both groups exhibited only very low concentrations, with no significant difference between them (RA 1.5 ± 4.2 ng/ml versus OA 1.8 ± 3.9 ng/ml; P = 0.54) (B). Enzyme-linked immunosorbent assay in the same patients revealed that TNFα levels in SF were significantly higher in RA patients than in OA patients (29.2 ± 33.0 pg/ml versus 7.5 ± 20.5 pg/ml; ∗ = P < 0.01 versus OA patients) (C). See Figure 1 for other definitions.

Localization of HMGB-1 in RA synovium.

We investigated HMGB-1 expression in RA synovial tissue by immunohistochemical analysis. As shown in Figure 3A, we found that HMGB-1 was localized in the cytoplasm of multinucleated cells situated mostly in the sublining layer. Rabbit IgG staining as a nonimmune control (Figure 3B) and hematoxylin and eosin staining (Figure 3C) were performed at the same time. The histologic characteristics and distribution of these cells expressing HMGB-1 were similar to those of synovial macrophages infiltrating the synovial membrane. To clarify this observation, we further employed double-labeling staining by immunofluorescence assay to compare the localization of HMGB-1–positive cells with that of CD68-positive cells. HMGB-1 was found in CD68-positive cells (Figures 4F–H), indicating that HMGB-1 is produced by synovial macrophages. Rabbit IgG and mouse IgG were used as nonimmune controls (Figures 4A–D). With an antigen-unmasking technique using autoclave heating, we could also detect HMGB-1 in the nuclei of these sections by both immunohistochemical and immunofluorescence assay (results not shown).

Figure 3.

Localization of HMGB-1 in RA synovium. Immunohistochemical analysis revealed that HMGB-1 was localized in the cytoplasm of multinucleated cells situated mostly in the sublining layer of RA synovium (A). The histologic characteristics and distribution of these cells resembled those of synovial macrophages. The sections were incubated with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) as a primary antibody overnight at 4°C. Rabbit IgG staining as a nonimmune control (B) and hematoxylin and eosin staining (C) were employed at the same time. See Figure 1 for definitions. (Original magnification × 250.)

Figure 4.

Comparison of localization of HMGB-1–positive cells with that of CD68-positive cells. To clarify the characteristics of HMGB-1–producing cells in RA synovial tissue, we employed double-labeling staining by immunofluorescence assay and verified that the same cells were immunoreactive for both HMGB-1 and CD68 (FH) (arrows in H indicate overlap of HMGB-1–positive cells and CD68-positive cells). This finding shows that HMGB-1 is produced by synovial macrophages infiltrating the RA synovium. The sections were incubated with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) and anti-CD68 mouse monoclonal antibody as primary antibodies, followed by incubation with fluorescein isothiocyanate–labeled anti-rabbit IgG and rhodamine-labeled anti-mouse IgG as secondary antibodies. Rabbit IgG and mouse IgG were employed as nonimmune controls (AD). The nucleus was labeled using 4′,6-diamidino-2-phenylindole (DAPI) (A and E). See Figure 1 for other definitions. (Original magnification × 400.)

RAGE expression in SFMs.

We examined expression of the specific receptor for HMGB-1, RAGE, in SFMs. SFMs were established from the SF of RA patients as described in Patients and Methods. Using these culture cells, we could identify RAGE in SFMs by immunoblot analysis (Figure 5). Human brain extract was used as a positive control, with a band observed at ∼50 kd (34).

Figure 5.

Expression of receptor for advanced glycation end products (RAGE) in SF macrophages (SFMs). RAGE, the specific receptor for HMGB-1, was identified in SFMs by immunoblot analysis. Human brain extract was used as a positive control, with a band observed at ∼50 kd (arrow). SFMs were established from the SF obtained from RA patients as described in Patients and Methods. See Figure 1 for other definitions.

Proinflammatory cytokine release by SFMs with stimulation by HMGB-1.

To investigate the functional role of HMGB-1 in the progression of RA, SFMs were stimulated with HMGB-1 and the concentrations of TNFα, IL-1β, and IL-6 released into conditioned supernatant were measured by ELISA. The peak TNFα levels after stimulation with HMGB-1 (500 ng/ml) occurred at 6 hours and tended to decrease after 12 hours (Figure 6A). TNFα was released in a dose-dependent manner after stimulation for 12 hours (Figure 6B). IL-1β levels increased in a time-dependent manner, except as shown at 12 hours, and were all significantly higher than those at 3 hours (Figure 6C). IL-1β was released in a dose-dependent manner after stimulation for 12 hours (Figure 6D). IL-6 was released by SFMs in both time- and dose-dependent manners (Figures 6E and F). Release of TNFα, IL-1β, and IL-6 had significantly increased at 12 hours with stimulation with HMGB-1 at doses ≥100 ng/ml (Figures 6B, D, and F).

Figure 6.

HMGB-1–induced production of proinflammatory cytokines by SF macrophages (SFMs). SFMs were stimulated with HMGB-1, and enzyme-linked immunosorbent assay was used to measure levels of tumor necrosis factor α (TNFα), interleukin-1β (IL-1β), and IL-6 released into conditioned supernatant. Values are the mean ± SD. After stimulation with HMGB-1 (500 ng/ml), TNFα levels peaked at 6 hours (646.1 ± 21.2 pg/ml) and then tended to decrease (∗ = P < 0.01 versus 3 hours) (A). IL-1β levels increased in a time-dependent manner, except for those at 12 hours, and were all significantly higher than those at 3 hours (∗ = P < 0.01) (C). IL-6 was released in a time-dependent manner (∗ = P < 0.01 versus 3 hours) (E). All cytokines were released in a dose-dependent manner for 12 hours after stimulation with HMGB-1, and the rate of release increased significantly at doses of ≥100 ng/ml (∗ = P < 0.01; † = P < 0.05 versus 0 ng/ml) (B, D, and F) See Figure 1 for other definitions.

Expression of TNFα mRNA in SFMs following stimulation with HMGB-1.

We found that HMGB-1 evokes TNFα release by SFMs via gene transcription. SFMs were stimulated with HMGB-1 (100 ng/ml), and expression of TNFα mRNA was determined by Northern blot analysis. As presented in Figure 7, the mRNA level began to increase at 2 hours and reached a maximum at 4 hours after the addition of HMGB-1.

Figure 7.

Expression of TNFα mRNA in SFMs following stimulation with high mobility group box chromosomal protein 1 (HMGB-1). SFMs were stimulated with HMGB-1 (100 ng/ml), and expression of TNFα mRNA was determined by Northern blot analysis. The expression of TNFα mRNA peaked at 4 hours, then decreased and returned to the baseline level by 12 hours. Note that TNFα levels following stimulation with HMGB-1 peaked at 6 hours at the protein level (see Figure 6A). See Figure 6 for other definitions.

HMGB-1 expression at protein and mRNA levels in SFMs with TNFα stimulation.

We examined the release of HMGB-1 by SFMs treated with TNFα. SFMs were stimulated with TNFα (100 ng/ml), and the concentrations of HMGB-1 released into conditioned supernatant were determined by immunoblot analysis and quantified. The mean ± SD level of HMGB-1 released by SFMs peaked at 48 hours (72.3 ± 12.5 ng/ml; P < 0.01 versus 12 hours) (Figure 8A). RT-PCR analysis revealed that HMGB-1 was up-regulated at the mRNA level after stimulation with TNFα for 8 hours (Figure 8B).

Figure 8.

High mobility group box chromosomal protein 1 (HMGB-1) production and release by SFMs evoked by TNFα. SFMs were stimulated with TNFα (100 ng/ml), and the concentrations of HMGB-1 released into conditioned supernatant were determined by immunoblot analysis and quantified as described in Patients and Methods. Values are the mean ± SD. The levels of HMGB-1 released by SFMs peaked at 48 hours (72.3 ± 12.5 ng/ml) (∗ = P < 0.01 versus 12 hours) (A). Up-regulation of HMGB-1 at the mRNA level was verified by reverse transcription–polymerase chain reaction analysis after stimulation with TNFα (100 ng/ml) for 8 hours (B). See Figure 6 for other definitions.

Translocation of HMGB-1 in SFMs after TNFα stimulation.

To examine the translocation of HMGB-1 in SFMs treated with TNFα, SFMs were cultured in a Lab-Tech chamber slide with medium as described in Patients and Methods and stimulated with TNFα (100 ng/ml) for 24 hours. As a result of observation by immunofluorescence assay, we found that HMGB-1 was translocated from the nucleus to the cytosol in CD68-positive cells in the stimulated group (Figures 9F–H). In the control group without stimulation, HMGB-1 staining was observed only in the nucleus (Figures 9A–D).

Figure 9.

Translocation of high mobility group box chromosomal protein 1 (HMGB-1) in SFMs after TNFα stimulation. Immunofluorescence assay indicated that HMGB-1 was translocated from the nucleus to the cytosol following stimulation with TNFα. SFMs were stimulated with TNFα (100 ng/ml) for 24 hours (EH) and then compared with control SFMs (AD). CD68-positive SFMs stimulated with TNFα exhibited HMGB-1 staining in the cytosol (FH). Note that control SFMs exhibiting HMGB-1 stained only in the nucleus (AD). SFMs were incubated with anti–HMGB-1 rabbit polyclonal antibody (1 μg/ml) and anti-CD68 mouse monoclonal antibody for 30 minutes after stimulation. Fluorescein isothiocyanate–labeled anti-rabbit IgG and rhodamine-labeled anti-mouse IgG were then used as secondary antibodies, and 4′,6-diamidino-2-phenylindole (DAPI) was used to label the nuclei. See Figure 6 for other definitions. (Original magnification × 400.)

Blockade of TNFα release using sRAGE.

We investigated whether neutralization of HMGB-1 could suppress TNFα release by SFMs. Blockade of TNFα release from SFMs was performed using sRAGE, which binds to HMGB-1, as described in Patients and Methods. As shown in Figure 10, sRAGE inhibited TNFα release in a dose-dependent manner after stimulation for 6 hours and significantly inhibited release at a dose of 250 ng/ml.

Figure 10.

Blockade of tumor necrosis factor α (TNFα) release using soluble RAGE (sRAGE). Soluble RAGE inhibited TNFα release from SFMs by binding to high mobility group box chromosomal protein 1 (HMGB-1). Soluble RAGE was incubated with HMGB-1 (500 ng/ml) at 25°C for 1 hour, and the mixture was then added to the cultured SFMs for 6 hours, followed by measurement of TNFα levels in the supernatant by enzyme-linked immunosorbent assay. Release of TNFα was significantly inhibited by sRAGE at a dose of 250 ng/ml (∗ = P < 0.01 versus control [cont] [HMGB-1 stimulation alone]). Values are the mean and SD. See Figure 5 for other definitions.

DISCUSSION

This study's findings suggest that HMGB-1 plays a pivotal role as a novel cytokine in the pathogenesis of RA. HMGB-1, a 30-kd member of the high mobility group nonhistone chromosomal protein family (1, 35, 36), is considered a mediator of delayed endotoxin lethality and systemic inflammation (11). We measured HMGB-1 in the SF and serum of both RA and OA patients using immunoblot analysis and found that the HMGB-1 concentrations in SF were significantly higher in RA patients than in OA patients (Figure 2A). Similarly, TNFα levels in SF were also significantly higher in the same RA patients than in the same OA patients (Figure 2C). Although anti–HMGB-1 antibody was frequently detected in sera from patients with systemic rheumatic diseases, including RA (37), HMGB-1 concentrations in serum were very low and did not differ significantly between RA and OA patients (Figure 2B). In RA synovial tissue, HMGB-1 was found in the cytoplasm of CD68-positive cells infiltrating the sublining layer (Figures 3 and 4), indicating that HMGB-1 is secreted by synovial macrophages and suggesting that HMGB-1 concentrations in synovial membrane may be higher than those in the SF we examined.

Like other cytokine mediators of endotoxemia, HMGB-1 activates proinflammatory cytokine release from human monocytes (12). We stimulated SFMs obtained from RA patients with HMGB-1 and confirmed that SFMs released proinflammatory cytokines, such as TNFα, IL-1β, and IL-6, in a dose-dependent manner (Figures 6B, D, and F). It has been reported that cytokine release after HMGB-1 stimulation was delayed and biphasic compared with that following lipopolysaccharide stimulation, as determined by detection of intracellular TNF synthesis in human PB mononuclear cultures (12). In our study, peak levels of TNFα in the supernatant released by SFMs occurred at 6 hours (Figure 6A), and TNFα mRNA peaked at 4 hours after stimulation with HMGB-1 (Figure 7), but levels of other cytokines, such as IL-1β and IL-6, were still increasing after 24 hours (Figures 6C and E). These results indicate that since TNFα is an “upstream” cytokine, IL-1β and IL-6 were released by TNFα during the late phase in addition to responding to the initial HMGB-1 stimulation (38, 39). It is thus possible that HMGB-1 induces SFMs to release increasingly not only TNFα, but also IL-1β and IL-6, and that this contributes to the pathogenesis of severe arthritis in the TNF knockout mouse model of collagen-induced arthritis (27) and explains why inhibition of TNF has not been effective in all patients (24).

On the other hand, TNFα also induced HMGB-1 release from SFMs, with HMGB-1 levels in supernatant detected by immunoblot analysis (Figure 8A). Interestingly, release from SFMs occurred much later than that of TNFα by HMGB-1, and peak release was found at 48 hours after stimulation. Immunofluorescence assay revealed that HMGB-1 was translocated from the nucleus to the cytosol in CD68-positive cells after 24 hours (Figure 9), implying that release of HMGB-1 extracellularly occurs after this. This shows that HMGB-1 acts as a late mediator, as reported by Wang et al (11,40), and contributes to prolonging and sustaining systemic inflammation (11, 40–42). HMGB-1 can also be released by damage or necrosis of a variety of cell types, including endothelial cells (43). However, after stimulation with TNFα, cell death of SFMs was not observed (data not shown), and augmented expression of HMGB-1 mRNA ensured that this release was not due to necrosis of SFMs but to stimulation with TNFα (Figure 8B). These findings suggest the possibility that a proinflammatory loop exists between HMGB-1 and TNFα via SFMs in RA.

Schmidt et al identified RAGE in human mononuclear phagocytes (18), and its molecular weight is ≤55 kd (18, 44). We also found RAGE expression in SFMs obtained from RA patients (Figure 5). It has recently been reported that RAGE is not required for process outgrowth or differentiation of neuroblastoma cells involving a dominant negative and antisense strategy (45), and that stimulation of erythroleukemia cell differentiation by extracellular HMGB-1 is independent of RAGE (46). RAGE may thus not be only a specific receptor for HMGB-1, but this remains to be established. However, we found that sRAGE significantly inhibited TNFα release by binding to HMGB-1 (Figure 10), which may be useful in suppressing the release of proinflammatory cytokines involved in this unique HMGB-1/RAGE system.

To our knowledge, this is the first report of a role for HMGB-1 in the pathogenesis of RA. Our findings improve understanding of the multiple overlapping activities of TNF and IL-1β in patients with RA and clarify the role of HMGB-1 as a newly identified inflammatory mediator. Blocking proinflammatory cytokine release with HMGB-1 stimulation using sRAGE might prove useful for the treatment of RA.

Acknowledgements

We appreciate the provision of samples by Kanehisa Hashiguchi, MD, Minoru Kijima, MD, Tsutomu Sonoda, MD, Akihiko Sonoda, MD, Nobuhiko Sunahara, MD, Koichiro Sugiyasu, MD, and Yoshihiro Ryoki, MD.

Ancillary