high mobility group box 1 (HMGB1) protein, originally described as a DNA-binding protein that stabilizes nucleosomes and facilitates transcription, can also be released extracellularly by monocytes and macrophages stimulated by LPS, TNF-α, or IL-1 (2, 44). Extracellular HMGB1 has been demonstrated to participate in inflammatory processes, including delayed endotoxin lethality and acute lung injury (1, 44, 46), and also appears to be involved in pathophysiological processes associated with cellular necrosis, such as acetaminophen-induced liver injury (34).

Although HMGB1 and LPS appear to initiate similar intracellular events, including activation of kinases such as p38, ERK1/2, and Akt and transcriptional factors including NF-κB, that lead to production of proinflammatory cytokines, gene arrays demonstrated differences in expression profiles with each of these stimuli (12, 30). Unlike LPS, which primarily increased the activity of IKK-β, HMGB1 exposure resulted in activation of both IKK-α and IKK-β (31). In addition, culture of neutrophils lacking Toll-like receptor (TLR)4 with HMGB1, but not with LPS, still resulted in enhanced nuclear translocation of NF-κB (31). Such results suggest that the receptors interacting with HMGB1 and leading to cellular activation and gene transcription are likely to be distinct from TLR4, which is responsible for LPS-induced responses (40). Recent data indicate that HMGB1 interacts not only with TLR4 but also with TLR2 and the receptor for advanced glycation end products (RAGE) (31, 46). In particular, a decrease in NF-κB-dependent reporter gene expression after transfection with dominant-negative constructs to TLR2, TLR4, or both, demonstrated that TLR2 and TLR4 were both involved in HMGB1-induced activation of NF-κB (31). In contrast, RAGE played only a minor role in the activation of macrophages by HMGB1 (31).

In the present experiments, we used fluorescence resonance energy transfer (FRET), transient receptor expression, and immunoprecipitation to directly investigate cell surface interactions of HMGB1 with TLR2, TLR4, and RAGE. We show that interactions of HMGB1 with TLR2 and TLR4 are early events after macrophage exposure to HMGB1. In contrast, there was little evidence of binding between HMGB1 and RAGE. Because of the similarities between intracellular signaling pathways initiated by TLR2 and TLR4, these studies provide an explanation for the ability of HMGB1 to induce cellular activation and generate inflammatory responses that are similar to those initiated by LPS. They also provide the first direct evidence that HMGB1 can interact with both TLR2 and TLR4.

MATERIALS AND METHODS

Reagents.

DMEM, DMEM-F-12, and penicillin-streptomycin were obtained from Invitrogen (Carlsbad, CA). Defined fetal bovine serum was purchased from HyClone (Logan, Utah). LPS (from Escherichia coli O111:B4) was obtained from Sigma (St. Louis, MO). LPS was reextracted twice with phenol and then precipitated from the aqueous phase to ensure that signaling occurred only through TLR4 (15). Pam₃Cys-Ser-(Lys)₄ (Pam₃CSK₄) was obtained from Alexis Biochemicals (San Diego, CA). The micro-BCA (bicinchoninic acid) protein assay reagent was obtained from Pierce (Rockford, IL). Rabbit anti-human HMGB1 antibody was from S. Yamada. Rat anti-mouse TLR4 antibody and rabbit anti-mouse TLR4 were from Dr. Sachiko Akashi (University of Tokyo, Tokyo, Japan). Antibodies for rabbit anti-mouse TLR2 were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Mouse anti-RAGE monoclonal antibody was obtained from Chemicon International (Temecula, CA). Advanced glycation end product-BSA (AGE-BSA) was obtained from Calbiochem (La Jolla, CA). Mouse anti-β-actin antibody was from Sigma. Horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG, goat anti-mouse IgG, and rabbit anti-goat IgG were from Bio-Rad (Hercules, CA). Normal rabbit IgG, normal rat IgG, and normal mouse IgG (obtained from Santa Cruz Biotechnology) were a negative control in immunostaining.

The mounting medium for fluorescence with 4′,6′-diamidino-2-phenylindole (DAPI) was obtained from Vector Laboratories (Burlingame, CA). Alexa Fluor 488 goat anti-rat IgG (H+L) and Alexa Fluor 488 goat anti-mouse IgG (H+L) were purchased from Molecular Probes (Eugene, OR). Cy3-conjugated affinity-purified goat anti-mouse IgG (H+L) antibody was purchased from Jackson ImmunoResearch Laboratories (West Grove, PA). HMGB1 was purified from pig thymus using the method of Sanders (33) and contained <10 pg/ml LPS by chromogenic assay (35).

Cells.

Murine macrophage RAW264.7 and human embryonic kidney (HEK)-293 cell lines were obtained from the American Type Culture Collection (ATCC; Rockville, MD), and cells were grown according to ATCC guidelines.

Transient transfection and luciferase reporter assay.

The cDNA for human TLR2 and TLR4 were provided in the vector of pcDNA3.1 by Dr. Tatsushi Muta (Kyushu University, Fukuoka, Japan). The hMD-2 expression plasmid pEFBOS containing COOH-terminal FLAG and His epitopes was a gift of Dr. Kensuke Miyake (University of Tokyo, Tokyo, Japan). HEK-293 cells were plated on 12-well plates at 5 × 10⁴ cells/well on the day before transfection. Combinations of expression plasmid DNA (0.1 μg/ml) or empty vector (pcDNA3.1) were cotransfected with pMD-2 and pNF-κB-Luc (Clontech), using Lipofectamine 2000 Plus (Invitrogen) according to the manufacturer's instructions. All cells were also transfected with pRL-TK-plasmid, a Renilla luciferase control reporter vector (Promega), to normalize transfection efficiencies. At 24 h after transfection, cells were stimulated with HMGB1 (1 μg/ml), LPS (100 ng/ml), or Pam₃CSK₄ (10 μg/ml) for 5 h. Such concentrations of HMGB1 approximate those found in supernatants from LPS-stimulated macrophages and in serum from animals and humans with severe sepsis (44). The cells were then lysed, and luciferase activity was measured with a dual luciferase assay report system (Promega) according to the manufacturer's instructions. All of the luciferase assays were repeated at least three times.

Immunofluorescence and FRET analysis.

RAW264.7 cells cultured in BD Falcon culture slides (BD Bioscience) were directly fixed after HMGB1 stimulation in 4% paraformaldehyde in PBS for 15 min at room temperature. After fixation under nonpermeabilized conditions, cells were washed in PBS and the slides were blocked with 15% normal goat serum at 37°C for 1 h. Primary antibodies were diluted in PBS-1% BSA and incubated with the cells for 2 h at room temperature. After three rinses with PBS, cells were incubated with secondary antibodies for 1 h at room temperature and washed three times with PBS, and the slides were mounted immediately on Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA). Images were examined with a Leica DRMXA digital microscope. The polyclonal rabbit anti-HMGB1 and monoclonal mouse anti-HMGB1 antibodies were from S. Yamada and were used at 1:150 dilution. Monoclonal anti-TLR4 antibody was from Dr. Sachiko Akashi and was used at 1:50 dilution. The monoclonal anti-TLR2 antibody was purchased from eBioscience (San Diego, CA) and used at 1:50 dilution. Polyclonal anti-RAGE antibody was from Santa Cruz Biotechnology and was used at 1:50 dilution. Cy3-conjugated affinity-purified goat anti-mouse IgG (H+L) secondary antibody was from Jackson ImmunoResearch Laboratories; Alexa Fluor 488 goat anti-mouse IgG (H+L) and Alexa Fluor goat anti-rat IgG (H+L) were from Molecular Probes and were used at 1:200 dilution. Negative controls containing preimmune IgG from the same animal species as a primary antibody ruled out nonspecific binding.

Molecular proximity of cell surface proteins was investigated for FRET measurements by the acceptor photobleaching technique. FRET-positive pixels and efficiencies were calculated by directly subtracting images obtained before acceptor photobleaching from images acquired immediately after acceptor photobleaching to reveal the increased fluorescence of the donor (4, 13, 20). After various durations of exposure to HMGB1, cells were fixed and stained with antibodies labeled with spectrally compatible fluorophores (acceptor: Cy3) and Alexa Fluor 488 (donor) and imaged with a motorized Zeiss Axiovert microscope (I.I.I.) equipped with Chroma filter cubes appropriate for separating the fluorophores and a 12-bit charge-coupled device (Cooke Sensicam). After an initial image was acquired on all available channels (RGB), areas of cells were masked, and the acceptor fluorescence (Cy3) in the masked area was bleached with a pumped dye laser to <30%. After pixel alignment between the pre- and postbleach images was ensured, individual pixel FRET efficiencies were calculated by expressing the increased donor fluorescence as a fraction of the postbleach intensity (13, 20). The noise level in unbleached areas (typically <2%) was eliminated to identify FRET-positive areas and calculate image statistics.

Immunoprecipitation and Western blots.

RAW264.7 cells were lysed in lysis buffer [in mM: 20 Tris·HCl (pH 7.5), 150 NaCl, 1 Na₂EDTA, 1 EGTA, 2.5 sodium pyrophosphate, 1 β-glycerophosphate, and 1 Na₃VO₄ with 1% Triton X-100] containing protease inhibitor cocktail (Roche, Mannheim, Germany). Whole cell extracts were centrifuged at 14,000 rpm for 20 min to remove the debris. Immunoprecipitations were performed by incubating whole cell extracts with indicated antibody, preincubated with recombinant protein G agarose (Invitrogen) while rocking at 4°C overnight. Immunoprecipitates were washed three times with lysis buffer, resuspended in 50 μl of 1× Laemmli sample buffer, and then resolved by 4–15% Tris·HCl-PAGE.

For Western blot analysis, cells were lysed in lysis buffer supplemented with protease inhibitor cocktail. Whole cell lysates were resolved by SDS-PAGE and subsequently transferred onto nitrocellulose membrane and probed with primary antibody after blocking. Specific bands were visualized using an enhanced chemiluminescence detection system with subsequent exposure to X-ray film.

RESULTS

FRET analysis of association of HMGB1 with TLR2, TLR4, or RAGE.

The phenomenon of FRET describes the transfer of energy from a fluorophore (donor) in an excited state to a neighboring fluorophore (acceptor) through dipole-dipole interaction occurring in the range of 1–10 nm (14). It can be used for quantitative determination of molecular proximity and intermolecular interactions (36) including protein-protein (22, 39), protein-DNA (25), protein-membrane, and three-dimensional structure of molecules (6, 23).

To investigate the association of HMGB1 with TLR2, TLR4, and RAGE, we measured molecular interactions of HMGB1 with each of these receptors by FRET, using FRET images to reveal early protein-protein interactions after HMGB1 stimulation of RAW264.7 cells. As shown in Fig. 1, at the initial time point, the fluorescent signals generated by HMGB1 molecules appeared in the nucleus and cytoplasm, whereas TLR4 and TLR2 were found in the cytoplasm and plasma membrane of the macrophages. After 5-min incubation with HMGB1, FRET spots, showing interaction between HMGB1 and TLR2 or TLR4, appeared on the cell surface. The maximal density of association between HMGB1 with TLR2 or TLR4 was present after 15 min of culture with HMGB1, followed by decreased intensity of interaction at 60 min (Fig. 1, A and B). In contrast, no association of HMGB1 with RAGE was found in RAW264.7 cells stimulated with HMGB1 (Fig. 1C). The maximal percentage of FRET efficiencies for the association between HMGB1 and TLR4 or TLR2 molecules in RAW264.7 cells was present 5 min after stimulation with HMGB1 and then decreased during the 60-min incubation period (Fig. 1D).

HMGB1 induces cellular activation through TLR4 and TLR2, but not RAGE, in HEK-293 cells.

HEK-293 cells do not express TLR2 or TLR4 and are normally unresponsive to activation through these receptors (24). Although HEK-293 cells become responsive to TLR2, TLR4, and RAGE ligands when these receptors are expressed after transfection, optimal signaling through TLR2 and TLR4 requires the presence of MD-2, a cell surface protein that associates with these receptors (5, 19, 27, 37).

To determine whether TLR4, TLR2, and/or RAGE mediate HMGB1-induced activation of NF-κB, HEK-293 cells were transiently cotransfected with TLR4, TLR2, RAGE, or empty vector control (pcDNA3.1) and MD-2 cDNA, as well as with a luciferase reporter construct under the control of the transcription factor NF-κB. At 24 h after transfection, the cells were either left untreated or incubated with HMGB1 (1 μg/ml), the TLR4 stimulus LPS (100 ng/ml), the TLR2-specific stimulus Pam₃CSK₄ (10 μg/ml), or the RAGE-specific stimulus AGE-BSA (200 μg/ml) for an additional 5 h and then lysed and assayed for luciferase activity.

Addition of HMGB1 to HEK-293 cells transfected with either TLR2 or TLR4 induced activation of the NF-κB reporter gene to levels approximately threefold those found in controls that were transfected with the empty vector (Fig. 2A). The degree of NF-κB-dependent transcription produced by stimulation with HMGB1 was similar in magnitude to that produced by cellular exposure to the specific TLR2 or TLR4 stimuli, Pam₃CSK₄ or LPS, respectively. In contrast, HMGB1 did not produce any apparent activation of cells transfected with RAGE, in contrast to that found with the RAGE-specific ligand AGE-BSA.

To further examine whether TLR2 and TLR4 are involved in HMGB1 signaling, HEK-293 cells were cotransfected with combinations of TLR4, TLR2, RAGE, or empty vectors. As shown in Fig. 2B, after stimulation with HMGB1, NF-κB reporter luciferase activity in cells coexpressing TLR2 and TLR4 was greater than in cells in which TLR2 plus RAGE or TLR4 plus RAGE had been transfected.

Coprecipitation of HMGB1 with TLR2 and TLR4.

To confirm that HMGB1 interacts with TLR2 and TLR4, we carried out immunoprecipitation experiments using anti-HMGB1 antibodies and then detection of the components of the HMGB1-associated complexes with antibodies to TLR2 or TLR4. As shown in Fig. 3, both TLR2 and TLR4 coprecipitated with HMGB1. There were similarities in the kinetics of association between TLR2 and TLR4 with HMGB1. In particular, the greatest amounts of coprecipitated TLR2 and TLR4 were found at the earlier time points, i.e., 5 and 15 min, after addition of HMGB1 to the RAW264.7 cells. Of note, exposure of RAW264.7 cells to HMGB1 resulted in rapid decreases in expression of both TLR2 and TLR4 that became greatest after 60 min (Fig. 3). No change in the amounts of HMGB1 in the cell lysates or associated with immunoprecipitated TLR2 or TLR4 was found at any time point after culture initiation (Fig. 3). There was no detectable association of RAGE with immunoprecipitated HMGB1 in these experiments (data not shown).

DISCUSSION

HMGB1 is a ubiquitous, abundant nuclear protein highly conserved among all mammals (11, 45). Incubation of macrophages with LPS or TNF-α results in increased release of HMGB1, and HMGB1 itself can stimulate macrophages, monocytes, and neutrophils to release proinflammatory cytokines, including IL-1β, TNF-α, and IL-8 (2, 30, 44). Increased circulating levels of HMGB1, attributed to release from activated macrophages, have been measured in the serum of mice treated with endotoxin as well as from septic patients (44). Additionally, HMGB1 appears to function as a late mediator of endotoxin lethality, as demonstrated by improved survival after treatment with specific anti-HMGB1 antibodies (44). When added to target cell cultures, HMGB1 can potently induce multiple responses, including differentiation, cytoskeletal reorganization, migration, and release of proinflammatory cytokines (1, 2, 9, 32). Recent information also suggests that HMGB1 may be an indicator of cell death by necrosis because its extracellular release is increased in this setting, whereas it remains confined to the nuclei of apoptotic cells (34).

Several studies have identified RAGE as a receptor for HMGB1 in neurites and malignant cell populations (16–18). However, other receptors also appear to be involved in HMGB1 signaling. Of note, incubation of HMGB1 with soluble RAGE or RAGE-blocking antibodies decreases, but does not eliminate, the stimulatory effects of HMGB1 on cellular activation (16, 26, 41, 42). The likelihood that there are alternative HMGB1-binding receptors was also suggested by recent studies demonstrating that HMGB1-induced differentiation of erythroleukemia cells is a RAGE-independent event (38) and that extracellular HMGB1 induces mesangioblast migration and proliferation even when the RAGE receptor has been disabled (29). Such findings are consistent with our previous experiments (31) in which RAGE was found to play only a modest role in HMGB1 signaling.

In the present experiments, FRET directly revealed early protein-protein interactions of HMGB1 with TLR2 and TLR4 in HMGB1-stimulated RAW264.7 cells. Initially, the fluorescent signals generated by HMGB1 molecules were identified only in the nucleus and cytoplasm, consistent with a purely intracellular localization for this molecule. At these early time points, TLR4 and TLR2 are found in the cytoplasm and plasma membrane of the macrophages. After 5-min incubation with HMGB1, FRET spots, showing interaction between HMGB1 and TLR2 or TLR4, appeared on the cell surface, consistent with extracellular HMGB1, either from that added to the cultures or from intracellular transport, being directly associated with these receptors. The maximal density of association between HMGB1 with TLR2 or TLR4 was present at 5 and 15 min of culture with HMGB1, followed by decreased intensity of interaction at the later time points examined. These results suggest that early interaction between HMGB1 and TLR2 or TLR4 may be followed by internalization and downregulation of TLR2 and TLR4. In contrast, there was no association of HMGB1 with RAGE on RAW264.7 cells stimulated with HMGB1. Such findings are consistent with our previous experiments (31) using transfections of dominant-negative RAGE constructs, which showed only minimal involvement of RAGE in HMGB1-induced macrophage activation.

Although the experiments using FRET showed direct association of HMGB1 with TLR2 and TLR4, this methodology does not demonstrate the significance of such interactions in terms of cellular activation. To examine this issue directly, TLR2 and TLR4 were transiently expressed in HEK-293 cells, which normally do not bear these receptors (24). Expression of either TLR2 or TLR4 in HEK-293 cells rendered them sensitive to stimulation by HMGB1. In contrast, transfection with RAGE was without effect. This lack of participation of RAGE in HMGB1-induced cellular activation is consistent with our FRET results, in which no association between HMGB1 and RAGE was found, and our earlier experiments (31) in which transfection of dominant-negative RAGE only minimally affected HMGB1-induced activation of RAW264.7 cells. However, the present findings do not eliminate a role for RAGE in HMGB1 signaling in other cell populations.

Using coimmunoprecipitation, we confirmed that HMGB1 associates with TLR2 and TLR4. In these experiments, the kinetics for the interaction between HMGB1 and TLR2 or TLR4 was largely similar. In particular, the greatest amounts of TLR4 coprecipitating with HMGB1 were found at 5–15 min after addition of HMGB1 to the RAW264.7 cells. These results suggest that interactive structural relationships are involved in the association of HMGB1 with either TLR2 or TLR4. Of note, expression of both TLR2 and TLR4 progressively decreased in cell lysates over the 60-min culture period with HMGB1, whereas no change of HMGB1 in the cell lysates was found at any time point. Therefore, our experiments indicate that the relative percentage of TLR2 or TLR4 associated with HMGB1, particularly at the later culture time points, was greater than earlier in the incubation period. Finally, in the immunoprecipitation experiments, there was no detectable association of RAGE with HMGB1, consistent with experiments using FRET as well as transfections of dominant-negative RAGE constructs (31) and confirming the minimal role that interactions between HMGB1 and RAGE appear to play in macrophage activation.

The present experiments, as well as our previous studies (31), indicate that both TLR2 and TLR4 are involved in cellular activation by HMGB1. Although the participation of multiple receptors in HMGB1 signaling is somewhat surprising, similar results have been reported by other investigators, who also found that HMGB1 could activate cells through TLR-dependent and -independent pathways (43, 46). Because distinct sets of ligands have been shown to interact with TLR2 and TLR4, it is likely that different regions of HMGB1 may be responsible for the associations observed in the present experiments with each of these receptors. HMGB1 has been shown to interact with a wide range of proteins, including transcriptional factors (21, 47), steroid receptors (3, 28), and viral components (7), through its A and B box domains, and possibly through the highly acidic COOH-terminal domain (10). With a phage display approach, HMGB1 was shown to recognize more than 12 different peptide sequences, indicating that HMGB1 is capable of associating with multiple proteins (10). In those studies, different peptide sequences were found to associate with either the A or B box of HMGB1, suggesting that HMGB1 does not have a restricted interaction sequence. Such findings demonstrate that HMGB1 has the potential to interact with more than one receptor and are consistent with the findings in our studies. Future experiments are necessary to identify the specific sequences in HMGB1 that associate with TLR2 or TLR4.

Interactions of HMGB1 with TLR4 may provide an explanation for the similarities in intracellular signaling pathways initiated by HMGB1 and LPS. In the case of cellular activation by both HMGB1 and LPS, the p38 MAP kinase appears to play a primary role in inducing proinflammatory cytokine expression, with kinases associated with phosphatidylinositol 3-kinase and ERK1/2 also being involved (30). Gene arrays revealed differences in expression profiles among neutrophils cultured with LPS and HMGB1, consistent with HMGB1 using receptors that were distinct from those associated with LPS (30). The present studies, directly demonstrating interactions between HMGB1 and TLR2 as well as TLR4 provide an explanation for such differences in gene expression.

In models of severe infection, circulating levels of HMGB1 increase with delayed kinetics, rising only hours after the initial infectious insult (44). The late appearance of HMGB1 appears to contribute to the increased levels of circulating and tissue cytokines that are present hours to days after exposure to LPS or bacteria and that contribute to mortality in these settings (2, 8, 44). Interaction of HMGB1 with TLR2 and TLR4, even after the initial infectious insult results in downregulation of membrane exposure of these receptors, would be expected to result in additional proinflammatory responses, including cytokine release, that contribute to organ dysfunction and death. The finding that HMGB1 associates with both TLR2 and TLR4 indicates that interventions directed against either of these receptors alone would not be able to prevent cellular activation and inflammatory injury induced by HMGB1. Of note, in our previous experiments (31), dominant-negative TLR2 or TLR4 constructs alone were insufficient to block HMGB1- induced signaling completely, whereas transfections with both constructs provided much greater inhibition of NF-κB-dependent transcription. However, because of redundancy in signaling pathways induced by TLR2 and TLR4, it is likely that interventions able to inhibit downstream intracellular events, such as activation of p38 or NF-κB, may be effective in reducing the proinflammatory actions of HMGB1.

GRANTS

This work was supported by National Institutes of Health Grants 1-P01-HL-068743 and P50-GM-049222.

FOOTNOTES

• The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.