Release of High Mobility Group Box 1 by Dendritic Cells Controls T Cell Activation via the Receptor for Advanced Glycation End Products¹

Dendritic cells are the most important APCs and are crucial to activate immune responses (1). Dendritic cells challenged with primary proinflammatory signals undergo a well-characterized differentiation program, referred to as maturation, that endows them with the ability to migrate to secondary lymphoid organs and to activate naive T cells (2, 3).

The high mobility group box 1 (HMGB1)³ protein is expressed in the nuclei of vertebrate cells at a level that varies in vivo over 2 log scales and appears to correlate with the cellular differentiation state (4). HMGB1 is also a cytokine that mediates endotoxin lethality, inflammation, and macrophage activation (5, 6). It has been argued that the presence of a nuclear protein outside of the cell signals unprogrammed cell death and tissue damage (7). Indeed, HMGB1 is passively released from necrotic cells, but is avidly retained within the remnants of apoptotic cells (8). However, some cell types, including monocytes and macrophages, can secrete HMGB1 actively without dying. Active secretion involves extensive acetylation of HMGB1 and transfer to a secretory endolysosomal compartment (9, 10, 11). Extracellular HMGB1 activates membrane receptors, including the receptor for advanced glycation end products (RAGE), a member of the Ig superfamily (12, 13), and possibly TLR (14).

Exogenous HMGB1 is one of many signals that promote dendritic cell maturation (15, 16). In vivo, exogenous HMGB1 effectively controls dendritic cell function, with substantial adjuvant effects on the immunogenicity of soluble or corpusculate Ags (15). Under certain conditions, necrotic cells provide HMGB1 for dendritic cell activation (15); we have suggested that this reflects the need for increased immune surveillance after traumatic tissue death (17).

In this study we show that the role of HMGB1 and its receptor, RAGE, is much more important than previously suspected. Human maturing dendritic cells actively secrete HMGB1 in response to inflammatory stimuli, and secreted HMGB1 is necessary for proliferation, survival, and polarization of naive CD4⁺ T cells. The blockage of either HMGB1 or RAGE restrains T cell responses. Thus, the presence of an NF in the extracellular environment represents a bottleneck for the initiation of adaptive immune responses in humans. When HMGB1 is not passively leaked in the microenvironment of dendritic cells by necrotic cells, the dendritic cells themselves provide the required extracellular HMGB1 by an active secretion mechanism.

The medium used throughout was RPMI 1640 (Invitrogen Life Technologies) supplemented with 100 U/ml penicillin, 100 μg/ml streptomycin, 1.5 mM l-glutamine, and 10% heat-inactivated FCS. Rabbit polyclonal anti-HMGB1 Ab raised against peptide 166–181 was purchased from BD Pharmingen. Goat polyclonal Ab against human RAGE was purchased from Chemicon International. Expression and purification of recombinant HMGB1, box A, and GST protein (used as control) were performed as previously described (18). Endotoxin was removed by passage through Detoxy-Gel columns (Pierce).

Human PBMCs were isolated from healthy donor blood by Ficoll density gradient centrifugation. The method for the generation of dendritic cells from PBMCs has been previously described (19). Briefly, monocytes isolated by adherence to plastic were incubated with GM-CSF (800 U/ml) and IL-4 (800 U/ml; R&D Systems) for 6 days. The purity and maturation of dendritic cells were routinely assessed by staining with Abs to CD1a, CD14, CD83, CD86, HLA-DR, and CD40 (BD Pharmingen). Immature dendritic cells were CD14⁻, CD1a⁺, CD83⁻, CD80⁻, CD86⁺, and HLA-DR⁺. Murine dendritic cells were derived from RAGE^−/− mice or age- and sex-matched wild-type animals. For this purpose, precursors isolated from bone marrows were propagated for 7 days in complete medium containing 1000 U/ml murine rGM-CSF and 5 ng/ml murine rIL-4 (R&D Systems). Endotoxin (LPS; 1 μg/ml; Sigma Aldrich), rabbit anti-HMGB1 (1/200 dilution), goat anti-RAGE (1/200 dilution), or box A (10 μg/ml) were added as indicated to induce dendritic cell maturation. For detection of cytokines, supernatants were collected and either used immediately or stored frozen until analysis. When indicated, the up-regulation of surface markers was expressed as the average increase, calculated as: ((mean fluorescence intensity of endotoxin treated cells − mean florescence intensity of untreated cells)/mean florescence intensity of untreated cells) × 100.

Human CD4⁺ naive T cells were purified by two-step immunomagnetic selection. Briefly, CD4⁺ T cells were first negatively selected for CD4⁺ using the CD4⁺ T cell isolation kit (Miltenyi Biotec) according to the manufacturer’s instructions. Then memory/effector T cells were depleted by incubation with anti-CD45RO-conjugated magnetic beads (Miltenyi Biotec). The purity of naive CD4⁺ T cells was typically >95%.

The levels of HMGB1 in culture medium were assessed by Western blotting as previously described (8). Cellular localization of HMGB1 was assessed by immunostaining of dendritic cells using affinity-purified rabbit anti-HMGB1 polyclonal Abs. Dendritic cells were transferred onto adhesion slides and incubated at 37°C for 1 h to allow adherence. The adherent cells were fixed with phosphate-buffered formaldehyde (4%; pH 7.4; for 15 min) and permeabilized with saponin. After blocking the slides with 10% BSA, cells were sequentially incubated with anti-HMGB1 Abs and FITC-conjugated anti-rabbit IgG. Nuclei were counterstained with Hoechst 33342 (Molecular Probes).

Cell lysates (30–40 μg) were electrophoresed on 12% SDS-PAGE gels and transferred onto nitrocellulose membranes (Amersham Biosciences). Membranes were blocked with 5% (w/v) nonfat dry milk/1% (v/v) Tween 20 in PBS for 1 h at room temperature and incubated overnight with primary Ab at 4°C. Abs to phospho- and total p38, phospho- and total ERK, and phospho- and total IκBα were all purchased from Cell Signaling Technology. Abs to β-actin were obtained from Sigma-Aldrich. Detection was performed by ECL (Amersham Biosciences). Relevant bands were quantified by densitometry using a computerized image analysis program (Amersham Biosciences).

The concentrations of IL-12 p70, TNF-α, and IL-10 in culture medium were determined using a two-site sandwich ELISA (R&D Systems), with reference to standard curves of purified recombinant cytokines. Samples were analyzed in serial 2-fold dilutions in duplicate.

Dendritic cells were incubated with endotoxin alone or in the presence of anti-HMGB1 Abs or box A for 8 h, then extensively washed before incubation with naive CD4⁺ T cells. T cells were mixed with allogeneic dendritic cells at a ratio of 10:1 and were plated at 1.5 × 10⁶ cells/ml. On day 3, the cultures were supplemented with 100 U/ml IL-2 (Roche). In some experiments, naive CD4⁺ T cells were stimulated with 1 μg/ml plate-bound anti-CD3 and 10 μg/ml soluble anti-CD28 mAbs as previously described (20). When indicated, 1 μg/ml rHMGB1 or supernatant of stimulated dendritic cells was filtered (0.2 μm), supplemented with 1 μg/ml anti-IL-12 (BD Pharmingen) blocking mAb, and added to naive T cells triggered by CD3/CD28 ligation. On day 8, cells were recovered, and Th1 and Th2 polarization was assessed as previously described (21). To assess T cell proliferation, cells were plated at 10⁵ cells/well in a round-bottom, 96-well plate at a T cell to dendritic cell ratio of 10:1 for 5 days. Microcultures were pulsed with [³H]thymidine (1 μCi/well) for 16 h. Cell cultures were harvested with an automated multiple sample harvester, and incorporated isotope was determined by liquid scintillation β emission.

Surface Ag staining was performed using mouse mAbs against CD1a, CD14, CD40, CD80, CD83, CD86, and HLA-DR (BD Pharmingen). The intracellular cytokine content was determined after stimulation of cells with 50 ng/ml PMA and 1 μg/ml ionomycin (21) (both from Sigma-Aldrich). GolgiPlug (BD Pharmingen) was added during the last 2 h of culture to prevent cytokine secretion. Cells were fixed and permeabilized using Cytofix/Cytoperm (BD Pharmingen), then stained with FITC-conjugated anti-IFN-γ and PE-conjugated anti-IL-4 mAbs (BD Pharmingen). The samples were analyzed on a FACScan (BD Biosciences).

Apoptosis was detected by staining with annexin V (Bender MedSystems) and propidium iodide (Sigma-Aldrich) as previously described (22).

Statistical analysis was performed using two-tailed Student’s t test for unpaired samples with unequal variance. A value of p < 0.05 was considered statistically significant.

Human immature dendritic cells derived in vitro from circulating precursors and cultured in the presence of rGM-CSF and IL-4 express high levels of HMGB1 in their nuclei, as shown by immunocytochemistry (Fig. 1, A and C). TLR4 engagement elicits the maturation of dendritic cells (2, 23) and changes HMGB1 localization; HMGB1 leaves the nucleus and reaches cytoplasmic vesicles (Fig. 1, B and D). Other stimuli known to elicit the maturation of dendritic cells, such as TNF-α, induced the translocation of HMGB1 to cytoplasmic vesicles as well (not shown). Similar changes occur in human monocytes and murine macrophages after endotoxin challenge (10, 24). These changes possibly represent a common response of cells belonging to the innate immune system after recognition of pathogen-associated molecular patterns (25).

The translocation of nuclear HMGB1 to cytoplasmic vesicles precedes its release in the medium (Fig. 1,E). Lactate dehydrogenase is not released in the medium (Fig. 1,E), ruling out the possibility that cells dying in culture passively leak their contents. The level of intracellular HMGB1 does not change over the 48-h incubation period (Fig. 1 E, inset), indicating that sustained production of the molecule supports active secretion during dendritic cell maturation.

We then analyzed whether the active secretion of nuclear HMGB1 from maturing dendritic cells influences the cross-talk with naive CD4⁺ T cells. This event controls T cell activation, proliferation, and polarization (2, 3). We committed dendritic cells to maturation and verified their action on naive allogeneic CD4⁺ T cells in the absence or the presence of reagents that antagonize secreted HMGB1. We used anti-HMGB1 polyclonal Abs and an N-terminal fragment of HMGB1 (box A) that behaves as an HMGB1 antagonist (26). The Abs we used recognize peptide 166–181 of HMGB1, which is part of the domain that interacts with RAGE (27); they inhibit the release of TNF-α by activated murine macrophages in vitro and reduce sepsis-induced mortality in vivo after cecal ligation and puncture (26). Box A attenuates HMGB1-induced release of proinflammatory cytokines and reduces mortality after cecal ligation and puncture or endotoxin injection (26).

Maturing dendritic cells induce the expansion of CD4⁺ T cells (from 2 × 10⁶ to 19.6 × 10⁶± 0.6; mean ± SE; Fig. 2,A). HMGB1 blockade by either Abs or box A abrogated the T cell expansion (p < 0.005 and p < 0.01, respectively; Fig. 2,A). Irrelevant control Igs or a GST control protein were ineffective (Fig. 2 A).

The T cell expansion reflects a combination of at least two events, i.e., the proliferation of T cells and the acquired resistance to activation-induced cell death (28). Fig. 2,B shows that T cell proliferation abates in the presence of HMGB1 inhibitors, indicating that T cell proliferation requires the paracrine release of endogenous HMGB1. Moreover, HMGB1 blockade causes a substantial increase in T cell death; 18% of T cells primed with maturing dendritic cells alone died, whereas 39 and 46% of T cells died in the presence of anti-HMGB1 Abs or box A, respectively (Fig. 2 D).

T cells differentiate after priming into effector type 1 (Th1) or type 2 (Th2) cells. The two subsets can be identified by their intracellular cytokine profiles (3). Priming with maturing dendritic cells results in polarization toward a Th1 cytokine pattern, as evidenced by the intracellular expression of IFN-γ (Fig. 2,C). The number of Th1 cells was dramatically reduced when the interaction between secreted HMGB1 and RAGE was blocked with either anti-HMGB1 Abs or box A (Fig. 2 C). In contrast, the percentage of Th2 cells, evaluated as IL-4⁺ T lymphocytes, was not influenced by the presence of HMGB1 inhibitors (2.02% when dendritic cells were triggered with endotoxin compared with 2.42% when dendritic cells were stimulated with endotoxin in the presence of anti-HMGB1 Abs). The ratio of IFN-γ⁺ to IL-4⁺ T cells was 103 when dendritic cells triggered with endotoxin alone were used for priming. This ratio decreased to 28 and 51 in the presence of anti-HMGB1 and box A, respectively.

Endotoxin-treated dendritic cells fail to sustain the proliferation, survival, and polarization of naive T cells when the HMGB1/RAGE interaction is blocked. This suggests that sustained HMGB1 secretion is necessary for the maturation of dendritic cells. We therefore evaluated by flow cytometry the membrane expression of CD80, CD83, CD86, and HLA-DR. In parallel, we measured the secretion of IL-10, IL-12, and TNF-α. Although mature dendritic cells up-regulate the expression of all markers (Fig. 3, A and B), anti-HMGB1 Abs and box A significantly hindered the up-regulation of CD80, CD83, and CD86 (p < 0.05) while leaving unaffected the expression of HLA-DR, the MHC molecules directly involved in Ag presentation to CD4⁺ T cells (Fig. 3, A and B). This is not surprising given the specific constraints regulating the intracellular trafficking of nascent HLA-DR to the plasma membrane (29, 30). No differences were detected in the presence of irrelevant Abs or a GST control protein (Fig. 3, A and B). HMGB1 blockade does not influence the expression of membrane molecules by immature dendritic cells (Fig. 3, A and B), which is in agreement with the observation that regulated HMGB1 secretion occurs only after induction of the maturation program (Fig. 1).

Maturing dendritic cells produce factors that sustain the maturation program, such as TNF-α (2, 3), or that contribute to limit it, such as IL-10 (31). A key factor for T cell priming is IL-12, which is produced upon dendritic cell maturation and behaves as a master regulator of the T cell polarization toward a Th1 phenotype (3). HMGB1 blockade does not influence the production of TNF-α and IL-10 (Fig. 3,C). However, in agreement with the requirement of HMGB1 for Th1 polarization of CD4⁺ T cells, HMGB1 blockade decreases the production of IL-12 (Fig. 3,C). Irrelevant Abs or the GST control do not influence the release of any of these cytokines (Fig. 3 C).

These results indicate that HMGB1 secretion by human dendritic cells enables them to prime naive T cells. Resting T cells express RAGE, the best-characterized receptor for HMGB1 (Fig. 4,A, left panel). We found that the endogenous HMGB1 protein secreted in the microenvironment by dendritic cells is not sufficient to influence the survival or the functional polarization of naive CD4⁺ T cells (Fig. 4,A, middle and right panels). T cell activation requires both stimulation via the T cell Ag receptor (signal 1) and additional costimulatory signals (signal 2) (32). In vitro T cell activation via signals 1 and 2 can be mimicked by CD3 and CD28 cross-linking (20). T cells activated by simultaneous triggering of CD3 and CD28 receptors become sensitive to secreted HMGB1, as revealed by proliferation and Th1 polarization (Fig. 4, B and C). The effect of secreted HMGB1 on CD3/CD28-triggered T cells was completely inhibited by HMGB1 blockade with either anti-HMGB1 Abs or box A (Fig. 4,B). As expected, T cell activation and polarization induced by the simultaneous cross-linking of CD3 and CD28 receptors in the absence of secreted HMGB1 were not influenced by HMGB1 blockade (Fig. 4,B). We observed that T cells triggered by CD3/CD28 cross-linking were more sensitive to secreted HMGB1 than to the recombinant protein (Fig. 4 C).

HMGB1 is a specific ligand for RAGE (12), and immature dendritic cells express RAGE (Fig. 5,A) (16). To directly test whether RAGE is responsible for the effect of released nuclear HMGB1 on dendritic cells, we used anti-RAGE Abs. These Abs are directed against peptide 42–59 of human RAGE and hinder the interaction of RAGE with its ligand, HMGB1 (33). Blockade of RAGE prevents the up-regulation of CD80, CD83, and CD86 after TLR4 activation (Fig. 5,B; p < 0.05) while leaving unaffected the expression of HLA-DR. No effects were observed in the presence of control Abs (Fig. 5,B). RAGE blockade per se does not influence the expression of membrane molecules by immature dendritic cells (Fig. 5,B), nor does it influence the production of TNF-α and IL-10 (Fig. 5,C), whereas it significantly decreases the production of IL-12 p70 (Fig. 5,C; p < 0.005). Irrelevant control Abs do not influence the release of IL-12 p70 (Fig. 5 C).

Anti-RAGE Abs abrogate the expansion (Fig. 5,D; p < 0.005) and proliferation (Fig. 5,E; p < 0.005) of naive CD4⁺ T cells induced by maturing dendritic cells. The number of IFN-γ⁺ T cells was dramatically reduced by anti-RAGE Abs (Fig. 5,F; p < 0.05). Irrelevant control Abs did not alter the expansion or Th1 polarization of T cells (Fig. 5, D and F). Furthermore, anti-RAGE Abs caused substantial T cell death (from 18% for T cells primed with maturing dendritic cells alone to 35% in the presence of anti-RAGE Abs).

ERK1/2 and p38 stress-activated protein kinase (p38) together with the transcription factor NF-κB are involved in the maturation of dendritic cells (34) and IL-12 secretion (35). The activation of NF-κB can be assessed by analyzing the NF-κB binding inhibitory protein, IκB. Phosphorylation of IκB leads to its degradation; this allows the nuclear translocation of NF-κB and the subsequent activation of transcription. HMGB1 binding to RAGE in neuroblastoma cells or rat macrophages activates MAPK pathways and culminates in NF-κB activation and nuclear translocation (12, 13, 36).

We assessed the phosphorylation of p38, ERK1/2, and IκB in dendritic cells after TLR4 activation in the absence or the presence of HMGB1/RAGE blockade. Twenty-four hours after TLR4 activation, we observed HMGB1 secretion (as shown previously), which correlated with a significant increase in the phosphorylation of p38 and IκB (Fig. 6). Anti-HMGB1 Abs induced a significant decrease in the phosphorylation of p38, ERK1/2, and IκB (Fig. 6) at 24 h after TLR triggering; control Abs had no effect (Fig. 6,B). Anti-RAGE Abs had the same inhibitory effect as anti-HMGB1 Abs on p38, ERK1/2, and IκB phosphorylation (Fig. 6 A).

TLR4 activation per se induces an early and transient phosphorylation of p38, ERK, and IκB (37). Indeed, the activation of p38, ERK1/2, and IκB after TLR4 triggering in dendritic cells follows a fast kinetic and abates after 1 h (Fig. 7). Moreover, this early phosphorylation of p38, ERK1/2, and IκB is not influenced by anti-HMGB1 Abs (Fig. 7), an expected result because HMGB1 is not secreted until 8 h after TLR4 triggering.

Taken together, these data suggest that interaction of secreted HMGB1 with RAGE leads to late activation of p38, ERK1/2, and NF-κB in maturing dendritic cells. This late activation is necessary for dendritic cell maturation. The early and transient activation of the same MAPKs and NF-κB initiated by TLR4 triggering is not influenced by HMGB1 blockade.

We derived dendritic cells from the bone marrow of RAGE^−/− mice and wild-type counterparts. We stimulated these cells with endotoxin and assessed their maturation. Up-regulation of surface markers such as CD40 and CD86 was significantly decreased in dendritic cells derived from RAGE^−/− mice compared with wild-type animals, whereas no change was observed for MHC class II (Fig. 8,A). The production of IL-12 p70 and TNF-α upon endotoxin treatment was significantly decreased as well in RAGE^−/− dendritic cells (Fig. 8,B). TLR4 triggering led to decreased phosphorylation of p38 and IκB kinases in RAGE^−/− dendritic cells compared with wild-type counterparts (Fig. 8 C). These results support a role for secreted HMGB1 in the activation of dendritic cells by acting on its receptor, RAGE.

HMGB1, a nuclear protein, has been recently described as an important extracellular mediator of inflammation (5, 17, 38). Its relocation outside the cell occurs as a consequence of either passive release after necrotic cell death (8) or active secretion by myeloid cells, involving extensive acetylation and transfer to secretory lysosomes (9, 10, 11). Exogenous HMGB1 is already known to play a role in dendritic cell activation (15, 16) and to behave as a potent adjuvant for immunization against soluble and cell-associated Ags in murine models (15).

In this study we have found that dendritic cells actively release their own HMGB1 into the microenvironment upon activation. This event is necessary for the maturation of dendritic cells themselves and for the cross-talk with naive T cells, because it is required for T cell survival, proliferation, and functional Th1 polarization. In particular, the active release of nuclear HMGB1 by dendritic cells is required for their ability to produce IL-12 and to up-regulate the expression of membrane molecules involved in T cell costimulation. We show that secreted, but not recombinant, HMGB1 stimulates proliferation and Th1 polarization of preactivated T cells. This highlights the importance of post-translational modifications for the biological activity of HMGB1 (39).

Furthermore, our data suggest a role for RAGE in the effects of HMGB1 on dendritic cell maturation via activation of MAPKs (p38 and ERK1/2) and NF-κB. Interestingly, the HMGB1/RAGE interaction determines the late activation of the MAPK and NF-κB pathways, contrary to the engagement of TLRs, which apparently only induces an early transient activation of the same pathways (Fig. 7). It may be that the late activation of the MAPK/NF-κB pathways, as opposed to an early transient activation, is key to the maturation of dendritic cells; alternatively, RAGE may activate an as yet unknown signaling pathway whose contribution is most relevant.

RAGE has been reported to be involved in innate immunity responses by recruiting inflammatory cells and contributing to the perpetuation of inflammation (40, 41). Our results suggest that RAGE plays a key role in regulating the responses of both human and murine dendritic cells to TLR triggering by microbial components. The human gene coding for RAGE, AGER, is located on chromosome 6 in the MHC class III region near the junction with class II (42). We note, however, that no null mutations in either the HMGB1 or AGER gene are known in humans; AGER variant alleles have been actively sought, because one missense mutation (G82S) and several silent point substitutions affect the predisposition to diabetic complications (43). In general, the rarity of loss-of-function alleles is correlated with a strongly selected phenotype, as immunodeficiency is sure to be. Additional receptors for HMGB1 are likely to be involved, because blockage of the HMGB1/RAGE pathway does not abrogate the maturation of dendritic cells upon TLR4 triggering.

In conclusion, our data indicate that HMGB1 secreted by maturing dendritic cells orchestrates the priming, activation, and Th1 polarization of T cells. This event may influence the initiation/maintenance of autoimmune diseases by regulating the outcome of Ag presentation under conditions involving tissue damage and infection. HMGB1 is normally a nuclear protein that is passively released in the microenvironment by necrotic cells, signaling cell damage. Cell necrosis is an event that reflects trauma or the presence of cytopathic pathogens, such as high virus loads (44), and can lead to inflammation. The ability of dendritic cells to recreate the environment associated with necrosis via the regulated secretion of HMGB1 represents a successful evolutionary strategy and places HMGB1 at the crossroads between innate and adaptive immunity.

We thank Julia Hering and Francesco Demarchis for recombinant HMGB1 and box A production and quality control, respectively, and Angela Raucci for critical reading of the manuscript. We thank Daniele D’Ambrosio for suggestions on dendritic cell-T cell cultures and critical reading of the manuscript.

M. E. Bianchi and A. A. Manfredi are named as investors in a patent application based partially on the data reported in this paper, with the assignee as Fondazione San Raffaele.