A critical cysteine is required for HMGB1 binding to Toll-like receptor 4 and activation of macrophage cytokine release

Abstract

During infection, vertebrates develop "sickness syndrome," characterized by fever, anorexia, behavioral withdrawal, acute-phase protein responses, and inflammation. These pathophysiological responses are mediated by cytokines, including TNF and IL-1, released during the innate immune response to invasion. Even in the absence of infection, qualitatively similar physiological syndromes occur following sterile injury, ischemia reperfusion, crush injury, and autoimmune-mediated tissue damage. Recent advances implicate high-mobility group box 1 (HMGB1), a nuclear protein with inflammatory cytokine activities, in stimulating cytokine release. HMGB1 is passively released during cell injury and necrosis, or actively secreted during immune cell activation, positioning it at the intersection of sterile and infection-associated inflammation. To date, eight candidate receptors have been implicated in mediating the biological responses to HMGB1, but the mechanism of HMGB1-dependent cytokine release is unknown. Here we show that Toll-like receptor 4 (TLR4), a pivotal receptor for activation of innate immunity and cytokine release, is required for HMGB1-dependent activation of macrophage TNF release. Surface plasmon resonance studies indicate that HMGB1 binds specifically to TLR4, and that this binding requires a cysteine in position 106. A wholly synthetic 20-mer peptide containing cysteine 106 from within the cytokine-stimulating B box mediates TLR4-dependent activation of macrophage TNF release. Inhibition of TLR4 binding with neutralizing anti-HMGB1 mAb or by mutating cysteine 106 prevents HMGB1 activation of cytokine release. These results have implications for rationale, design, and development of experimental therapeutics for use in sterile and infectious inflammation.

Fig. 1.

Mammalian HMGB1 secreted by genetically engineered CHO cells activates macrophage TNF release and is significantly inhibited by neutralizing anti-HMGB1 mAb. (A) A eukaryotic expression plasmid was engineered to secrete an N-terminal 3 X FLAG-tagged rat HMGB1 recombinant protein (24). Stably transfected CHO cells were cultured in CHO-S-SFM II media supplemented with 300 μg/mL geneticin. As control, non-HMGB1-secreting CHO cells were maintained in the same culture medium without geneticin. (B) Conditioned medium from CHO- or CHO-HMGB1-secreting cells was harvested and concentrated about 10-fold using a 10-KDa cut off membrane. TNF-stimulating activity of the conditioned medium was assayed on human primary macrophages in 96-well plates. Supernatants were collected after 16 h and TNF levels were quantified using commercial ELISA. Data are presented as mean ± SEM (n = 3). *P < 0.05 vs. unstimulated control. (C) Primary human macrophages were stimulated with conditioned medium from CHO-HMGB1 cells in the presence or absence of anti-HMGB1 mAb. Data are presented as means ± SEM (n = 3). *P < 0.05 vs. CHO-HMGB1 alone. (D) Human primary macrophages were stimulated with various amounts of conditioned medium from CHO-HMGB1 cells (0, 2, 8, 18 and 20 μl/mL, corresponding to 0, 0.1, 0.4, 0.9 and 1 μg HMGB1/mL, respectively, as measured by Western blot). TNF levels in the supernatants were measured by ELISA. Data are presented as mean ± SEM (n = 4–5). (E) RAW 264.7 cells were treated with necrotic HMGB1 KO or wild-type fibroblasts (10⁶ cells/mL) in the presence or absence of anti-HMGB1 mAb for 16 h. TNF levels in the supernatant were analyzed by ELISA. Data are presented as mean ± SEM (n = 5–6). *P < 0.05 vs. WT necrotic cells alone. (F) Western blot autoradiogram of HMGB1 from wild-type and HMGB1 KO fibroblasts. Approximately 20 μg of total proteins from necrotic HMGB1 KO or wild-type fibroblasts were subjected to Western blot analysis using anti-HMGB1 antibodies. Data shown is representative from two separate measurements.

Fig. 2.

TLR4 is required for HMGB1 signaling on macrophages. Thioglycollate-elicited peritoneal macrophages were isolated from wide-type (WT) or RAGE, TLR2, and TLR4 KO, or RAGE/TLR2 and RAGE/TLR4 double KO mice and stimulated with 1 μg/mL HMGB1 for 16 h. Supernatants were analyzed for levels of TNF (A), IL-1β, IL-6, IL-10, and MIP-2 (B) by ELISA. Data shown are mean ± SEM (n = 4–5). *, P < 0.05 vs. WT.

Fig. 3.

HMGB1 and B box bind to TLR4/MD2 as revealed by surface plasmon resonance (BIAcore) (A–C) and in situ proximity ligation assay (for HMGB1 only, D). (A) Different concentrations of HMGB1 (0, 0.0625, 1.25, 2.5, 5, and 10 μM) were flowed over immobilized TLR4/MD2 on the sensor chip. Data are presented as response units over time (seconds). HMGB1 bound to TLR4/MD2 in a concentration-dependent manner; with an apparent K_d of 1.5 μM. Data shown are representative of three experiments. (B) Compared with buffer alone, HMGB1 (10 μM) bound to TLR4/MD2 complex. Addition of neutralizing anti-HMGB1 mAb (10 μM) completely inhibited HMGB1 binding to TLR4/MD2. Data are representative of three experiments. (C) Increasing concentrations of B box (0, 0.0625, 1.25, 2.5, 5, and 10 μM) were flowed over immobilized TLR4/MD2 on the sensor chip. HMGB1 B box showed a concentration-dependent binding to TLR4/MD2 complex with a K_d of ≈22 μM. Data shown are representative of three experiments. (D) Synovial fibroblasts from human rheumatoid arthritis patients were subjected to in situ proximity ligation assay (Materials and Methods), using primary antibody pairs of anti-HMGB1 or control IgG2b antibody together with anti-TLR4 antibody. Dual binding by a pair of corresponding proximity probes, secondary antibodies with attached oligonucleotides, generates a spot (blob) if the two antibodies are in close proximity. Thus, each individual blob represents HMGB1 in close proximity with TLR4. The cells were counterstained with anti-actin (green) and Hoechst dye (blue) to visualize the cytoplasm and nucleus, respectively. Data shown are representative of three experiments.

Fig. 4.

Cysteine 106 is required for HMGB1-TLR4 binding and activation of TNF release from macrophages. (A) HMGB1 (10 μM) bound to TLR4/MD2 as compared with buffer alone in surface plasmon resonance analysis (BIAcore). In contrast, C106A HMGB1 did not bind to TLR4/MD2. Data shown are representative of three repeats. (B) Mouse peritoneal macrophages were stimulated with increasing concentrations of HMGB1 or C106A HMGB1 for 16 h. TNF release was measured by ELISA. Data shown are means ± SEM (n = 3). (C) Peritoneal macrophages from WT, TLR4, or RAGE KO mice were cultured with C106 or S106 peptides at concentrations indicated for 16 h, and TNF released was measured. Data shown are mean ± SEM (n = 4–5). *, P < 0.05 vs. unstimulated controls. (D) Primary human macrophages and RAW 264.7 cells were treated with C106 peptide for 16 h and TNF released was measured. Data shown are mean ± SEM (n = 3–4). *, P < 0.05 vs. unstimulated controls.