N-linked glycosylation plays a crucial role in the secretion of HMGB1

High mobility group box 1 (HMGB1) is a damage-associated molecular pattern (DAMP) molecule located in the nucleus. HMGB1 is secreted from activated monocytes and macrophages after post-translational modification (PTM) (Bonaldi et al., 2003; Youn and Shin, 2006), activation of an inflammasome pathway mediated by double-stranded RNA-dependent protein kinase (PKR, also known as EIF2AK2) (Lu et al., 2012), and through a pyruvate kinase M2 (PKM)-mediated pathway (Yang et al., 2014). HMGB1 is secreted through a non-classical vesicle-mediated secretory pathway (Gardella et al., 2002; Dupont et al., 2011). Additionally, HMGB1 can be released from dying cells as a result of autophagy (Thorburn et al., 2009), apoptosis (Bell et al., 2006), pyroptosis (Nyström et al., 2013) and necroptosis (programmed necrosis) mechanisms (Zou et al., 2013). HMGB1 can also initiate inflammation (Scaffidi et al., 2002), and it has a role as a late mediator of sepsis in both animal models and human patients (Ombrellino et al., 1999; Wang et al., 1999; Sundén-Cullberg et al., 2005). Extracellular HMGB1 binds to toll-like receptor (TLR) 2 and 4, as well as the receptor for advanced glycation end products (RAGE), and activates NF-κB and extracellular-signal-regulated protein kinases 1 and 2 (ERK1/2, also known as MAPK3 and MAPK1, respectively) (Salmivirta et al., 1992; Hori et al., 1995; Park et al., 2004) during sterile inflammation (Yang et al., 2010, 2015). HMGB1 interacts with host molecules of interleukin (IL)-1β (Sha et al., 2008), chemokine (C-X-C motif) ligand 12 (CXCL12) (Schiraldi et al., 2012) and nucleosomes (Tian et al., 2007), as well as with pathogen-associated molecular pattern (PAMP) molecules of lipopolysaccharide (LPS) and lipoteichoic acid (LTA), which are finally transferred to TLR4 and TLR2 receptors, respectively (Youn et al., 2008; Kwak et al., 2015), and augment or modify pro-inflammatory reactions. Intracellular HMGB1 is primarily located in the nucleus, where it functions as a nucleosome stabilizer and a regulator of transcription (Einck and Bustin, 1985; Lotze and Tracey, 2005), and also as a chaperone-like molecule that inhibits aggregation of polyglutamine (Min et al., 2013). In addition, HMGB1 binds to beclin-1 and regulates induction of autophagy (Tang et al., 2010).

Approximately 30 years ago, HMGB1 was reported to undergo extensive PTMs, including acetylation (Sterner et al., 1979; Dimov et al., 1990), phosphorylation (Sun et al., 1980; Kimura et al., 1985), methylation (Boffa et al., 1979) and adenosine diphosphate (ADP)-ribosylation (Tanuma et al., 1985). More recently, the molecular aspects of HMGB1 PTMs [acetylation (Bonaldi et al., 2003), phosphorylation (Youn and Shin, 2006), methylation (Ito et al., 2007), and cysteine oxidation (Hoppe et al., 2006)] have been investigated to determine the regulation of intracellular localization and extracellular secretion of HMGB1 owing to its role as an inflammatory mediator. The shuttling of HMGB1 between the nucleus and the cytoplasm is tightly regulated; the nuclear importin protein karyopherin-α1 and the nuclear exportin chromosomal maintenance 1 (CRM1, also known as XPO1) are involved in nucleocytoplasmic shuttling, which is influenced by modifications of HMGB1 (Bonaldi et al., 2003; Youn and Shin, 2006). HMGB1 has no signal peptide targeting it to the endoplasmic reticulum (ER), but a low degree of glycosyl modification has been identified in HMGB1 in calf thymus (Reeves and Chang, 1983; Elton and Reeves, 1986). The glycosylation residue and its role, however, have not been identified at a molecular level, although glycosylation is an important PTM.

HMGB1 has two N-glycosylation acceptor sites with a consensus motif of NxS/T (x can be any amino acid other than proline) (Welply et al., 1983 Marshall, 1974; Gavel and von Heijne, 1990). Here, we demonstrated that HMGB1 can be N-glycosylated at the consensus motifs at N37 and N134, and at a non-consensus motif at N135, by using liquid chromatography tandem mass spectrometry (LC-MS/MS) and gel shift analysis of the double mutant HMGB1^N37Q/N134Q. De-glycosylation of HMGB1 increases its binding to DNA and decreases its binding to CRM1, as demonstrated through gel shift and fluorescence recovery after photobleaching (FRAP) assays. We found that de-glycosylated HMGB1 localized to the nucleus and showed little extracellular secretion even when it had other PTMs. These results demonstrate, at a molecular level, that HMGB1 can be glycosylated at three asparagine residues, and that this modification is an important prerequisite for cytoplasmic transport and extracellular secretion resulting from pro-inflammatory stimuli.

The PTMs of HMGB1 are important for HMGB1 nucleocytoplasmic shuttling, extracellular secretion and transcriptional regulation. However, HMGB1 glycosylation and the functional role of this glycosylation have not been well characterized. The motif for N-linked glycosylation is NxS/T and there are four asparagine residues in HMGB1. Two motifs located at N37 and N134, in the HMGB1 A and B box, respectively, are part of this NxS/T motif. These motifs are conserved in rhesus macaques, cattle, rats, mice and chickens according to an N-linked glycosylation prediction program (NetNglyc, http://www.cbs.dtu.dk/services/NetNGlyc/) (Fig. 1A). Wild-type human HMGB1 (HMGB1^WT) does not contain an ER-targeting signal peptide to allow for glycosylation. Therefore, to observe HMGB1 glycosylation, which has been previously reported (Reeves, 1981), we subcloned HMGB1^WT DNA into a pFastBac HT-B plasmid and produced HMGB1 protein in insect SF9 cells (insHMGB1^WT) for de-glycosylation analysis. HMGB1^sig-pep, which has an artificially added signal peptide and was produced in HEK293F cells, was used as a positive control for N-glycosylated HMGB1. Both insHMGB1^WT and HMGB1^sig-pep were treated with PNGase F, a glycosidase enzyme that hydrolyzes N-linked oligosaccharides, and then a migration shift analysis was performed. We found that insHMGB1^WT was shifted down after enzyme treatment, similar to HMGB1^sig-pep, suggesting that HMGB1^WT is glycosylated (Fig. 1B). Next, we treated CHO-K1, HeLa and RAW264.7 cells with tunicamycin, an inhibitor of N-linked glycosylation, for 24 h and observed the migration shift of endogenous HMGB1 by western blotting. A band shift of endogenous HMGB1 was observed after tunicamycin treatment (Fig. 1C). Taken together, these data suggest that HMGB1 can be N-glycosylated, confirming a previous report (Reeves, 1981). When RAW264.7 cells were treated with brefeldin A and nocodazole, inhibitors of protein transport from the ER to the Golgi and of microtubule polymerization for vesicular transport, respectively, LPS-stimulated HMGB1 secretion did not decrease (Fig. 1D). These results suggest that the conventional ER-Golgi secretion pathway is not related to HMGB1 secretion.

Next, we performed LC-MS/MS analysis of insHMGB1^WT. Treatment with PNGase F results in an increase in molecular mass of 0.98 Da over the reference size of HMGB1 peptide fragments. This change is due to the conversion of an asparagine into an aspartic acid residue after a sugar molecule is deleted from the amine group of asparagine by PNGase F (Hao et al., 2010), confirming N-glycosylation in the site. HMGB1 was glycosylated at the N37 and N134 N-glycosylation consensus motifs. Interestingly, N135 is not a typical residue for N-glycosylation, but received a high score of glycosylation, and also displayed the molecular mass change associated with N-glycosylation (Fig. S1A). Next, we compared the molecular shift of insHMGB1^WT and insHMGB1^N37Q/N134Q by performing western blotting and found that insHMGB1^N37Q/N134Q was shifted down compared to insHMGB1^WT, suggesting N-glycosylation of HMGB1 (Fig. S1B). We repeated the LC-MS/MS analysis using purified Myc-tagged HMGB1, which was transiently overexpressed in HEK293T cells (Fig. 2A). Approximately 60% of the HMGB1 amino acid sequence was successfully detected and sequenced (Fig. S2A). A peptide including N37 differed between b6 (607.284) and b7 (722.310), and a peptide including N134 and N135 differed between b6 (761.365) and b7 (876.392), and between b7 (875.408) and b8 (990.435), respectively (Fig. 2B; Fig. S2B). We next prepared naturally expressed endogenous HMGB1, and repeated the LC-MS/MS analysis to prove that HMGB1 glycosylation was not limited to the artificial system. We purified endogenous HMGB1 protein from cultured HEK293T cells, using the immunoprecipitation method. The glycan modifications were identified at N37, N134 and N135 residues (Fig. S3). These data show that HMGB1 is N-glycosylated at N37, N134 and N135 as determined by LC-MS/MS analysis.

We next performed an LC-MS/MS analysis of the glycan structures of HMGB1. Myc–HMGB1 was purified from HEK293T cells transfected with myc–HMGB1 plasmid. A total ion chromatogram of Rapifluor-labeled glycans from 15 µg HMGB1 released by PNGase F was separated using an amide HILIC column. An analysis of glycan profiling of human Ig was used as a test control (data not shown). The chromatogram of the glycans eluted for 22 min was analyzed, where glycan composition and the proposed structures of glycans can be suggested according to the elution time (Fig. 3).

Next, we tested the functional role of HMGB1 N-glycosylation. HMGB1 is translocated to the cytoplasm when modified by acetylation, phosphorylation or oxidation. Therefore, we prepared expression plasmids for EGFP-tagged HMGB1^WT, HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q, and transiently overexpressed these proteins in HEK293T cells to observe their translocation and extracellular secretion. All HMGB1^WT was located in the nuclei in the resting state and over 80% was translocated to the cytoplasm upon treatment with trichostatin A (TSA; to induce acetylation), phorbol 12-myristate 13-acetate (PMA; to induce phosphorylation) or H₂O₂ (to induce oxidation) (Fig. 4A), confirming previous results (Bonaldi et al., 2003; Hoppe et al., 2006; Youn and Shin, 2006). However, the cytoplasmic translocation of EGFP–HMGB1^N37Q/N134Q and EGFP–HMGB1^N37Q/N135Q was severely decreased, to 20% and 40%, respectively, even after treatment with TSA, PMA or H₂O₂ (Fig. 4B–D).

We next observed the extracellular secretion of HMGB1 (Fig. 4E). When HEK293T cells were transfected with EGFP–HMGB1^WT plasmid, EGFP–HMGB1^WT secretion was profoundly increased upon treatment with PMA, TSA and H₂O₂. However, when transfected with the double mutants HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q, their secretion was barely detectable even after treatment with TSA, PMA or H₂O₂. When the cells were transfected with single mutants HMGB1^N37Q, HMGB1^N134Q or HMGB1^N135Q, HMGB1 secretion was partially decreased after stimulation. When we treated J774A.1 cells with LPS for 24 h or starved the cells (Fig. 4F), the secretion patterns of HMGB1^WT and mutant proteins were similar. These data suggest that HMGB1 N-glycosylation at N37, N134 and N135 is a prerequisite for nucleocytoplasmic translocation and secretion of HMGB1.

HMGB1 is located in the nucleus under resting conditions and binds to the minor groove of DNA; this binding ability is reduced by methylation (Ito et al., 2007). We next tested whether N-glycosylation of HMGB1 influences its movement in the nuclei of live HEK293T cells by using FRAP analysis. FRAP data indicate the rate at which fluorescent molecules exchange with photobleached molecules. The recovery percentage of fluorescence in the area of photobleaching will decrease if the protein mobility is decreased. The recovery percentages of EGFP-tagged HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q were lower than that of EGFP–HMGB1^WT within 4 s, suggesting slower mobility (Fig. 5A). Therefore, we hypothesized that HMGB1 N-glycosylation influences its DNA binding. In order to test this, plasmid DNA was incubated with insHMGB1^WT and insHMGB1^N37Q/N134Q, and complex formation was observed. As shown in Fig. 5B, the migration of complexes of insHMGB1^N37Q/N134Q and plasmid DNA were retarded compared to those of insHMGB1^WT, and some complexes did not migrate and remained in the wells. The effect of HMGB1 N-glycosylation on binding between the N-acetyl-D-glucosamine (GlcNAc)-tagged HMGB1 A or B box and double-stranded DNA (dsDNA) was modeled using the structures of the complexes of HMGB1 A box and B box bound to dsDNA as templates (PDB ID 1CKT and 2GZK, respectively) (Stott et al., 2006; Watson et al., 2007). As shown in Fig. 5C, HMGB1 GlcNAc molecules at N37 and N134 induce steric hindrance on binding between HMGB1 and dsDNA. Although the carbohydrate chains are flexible, they could interfere with the phosphate backbone or the nucleic acids of the dsDNA. As seen in Fig. 5C, only two GlcNAc molecules linked by a glycosidic bond were drawn instead of entire carbohydrate chains for clarity.

Nuclear HMGB1 is translocated to the cytoplasm after binding to the nuclear exportin CRM1 (Bonaldi et al., 2003); hence, we tested the binding of each mutant HMGB1 protein to CRM1. HEK293T cells were co-transfected with CRM1 plasmid and either Myc-tagged HMGB1^WT, HMGB1^N37Q/N134Q or HMGB1^N37Q/N135Q for 48 h, and then treated with 10 ng/ml TSA for 4 h to observe the effect of HMGB1 acetylation. The cells were lysed and immunoprecipitated with anti-Myc antibody and then immunoblotted with CRM1. As shown in Fig. 6, myc–HMGB1^WT binding to CRM1 increased after TSA treatment, as previously described (Bonaldi et al., 2003). HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q binding to CRM1 were profoundly decreased when compared with HMGB1^WT. These results suggest that HMGB1 N-glycosylation is important for binding to CRM1 and, thus, also for HMGB1 secretion.

Protein glycosylation is important in protein structure and stability. Therefore, we tested the effect of glycosylation on the intracellular stability of HMGB1 using HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q plasmids. HEK293T cells were transfected with HMGB1^WT, HMGB1^N37Q/N134Q or HMGB1^N37Q/N135Q plasmids and then treated with cycloheximide (CHX) to inhibit protein translation and observe protein stability. Cells were harvested and HMGB1 protein levels were analyzed by western blotting. HMGB1^WT protein levels were fairly consistent and decreased by only 10% after 48 h. However, HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q levels decreased by 60% and 30%, respectively, after 48 h (Fig. 7A). These data suggest that HMGB1 is degraded quickly when not properly glycosylated. HEK293T cells were then transfected with HMGB1^WT or HMGB1^N37Q/N134Q plasmids for 24 h and treated with a proteasome inhibitor, MG132, in the presence or absence of 20 mM NH₄Cl, a lysosomal inhibitor that prevents protease activity. No significant change was observed in HMGB1^WT protein levels, but a dramatic recovery of HMGB1^N37Q/N134Q was observed upon treatment with MG132 (Fig. 7B), suggesting that HMGB1^N37Q/N134Q is degraded in a proteasome-mediated pathway. In addition, HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q were ubiquitylated at K48, which is a degradation signal (Fig. 7C). These results suggest that HMGB1 glycosylation is important for HMGB1 stability.

HMGB1 secretion is modulated by the PTMs of acetylation, phosphorylation, methylation and oxidation, which drive HMGB1 to the cytoplasm and then to the extracellular space through an unconventional secretion pathway. Although it has no signal peptide, a low degree of glycosylation of HMGB1 has been reported previously (Reeves and Chang, 1983; Elton and Reeves, 1986; Chao et al., 1994), but the significance of glycosylation and the glycosylation residue had not been identified.

In this study, we demonstrated that HMGB1 could be glycosylated by performing a gel shift analysis after PNGaseF treatment and a tunicamycin-mediated N-glycosylation inhibition assay. LC-MS/MS analysis showed that there were two N-glycosylations at N37 and N134, at NxS/T consensus motifs, and one at the non-classical consensus residue N135, in recombinant HMGB1 proteins produced in both HEK293T and insect cells. Recently, the sequence requirements of the acceptor substrate for N-glycosylation have become less strict and it has been shown that atypical (non-consensus) N-linked glycosylation is possible (Valliere-Douglass et al., 2010; Schwarz and Aebi, 2011). Previously, it had been reported that HMGB1 contains a low degree of glycosyl modification in calf thymus (Reeves et al., 1981; Reeves and Chang, 1983; Elton and Reeves, 1986; Chao et al., 1994), and our data confirmed this glycosylation and identified the glycosylation residues. In addition, recombinant HMGB1 expressed in CHO cells treated with glycopeptidase F resulted in the removal of an N-linked oligosaccharide (Li et al., 2004), and the HMG family proteins HMGN1 and HMGN2 are also N-glycosylated (Reeves and Chang, 1983). Taken together, these results suggest that HMGB1 is N-glycosylated.

The significance of HMGB1 glycosylation was not known, even though HMGB1 glycosylation was reported over 20 years ago. Here, we show that HMGB1 N-glycosylation is required for nucleocytoplasmic translocation and extracellular secretion. When HEK293T cells transfected with HMGB1^N37Q/N134Q or HMGB1^N37Q/N135Q plasmid were treated with PMA, TSA or H₂O₂, the mutant HMGB1 primarily showed nuclear localization with a profound reduction in extracellular secretion. These results demonstrate that the PTMs of acetylation, phosphorylation and oxidation are not sufficient for HMGB1 secretion, and glycosylation is a prerequisite for HMGB1 translocation. Based on the results showing that HMGB1^N37Q/N134Q and HMGB1^N37Q/N135Q had stronger binding to nuclear DNA and weaker binding to CRM1 compared to HMGB1^WT, we suggest that HMGB1 N-glycosylation at N37, N134 and N135 is important for HMGB1 intracellular trafficking and secretion.

N-glycans are synthesized and transferred to polypeptides containing a signal peptide, by glycosyltransferase, at aspargine residues within the NxS/T sequon on the luminal side of the ER and Golgi (Molinari, 2007; Hebert et al., 2014). However, the mechanism by which HMGB1 becomes glycosylated despite being devoid of signal peptide is still is unknown, as N-glycosylation rarely exists in the cytosol. Based on a molecular mass shift upon PNGase F treatment, as analyzed by western blot and profiling of N-glycans, some amount of glycosylation is linked to HMGB1.

The secretion of protein devoid of signal peptide is not fully understood as the protein is unable to access the conventional secretion pathway through the ER. IL-1β is a model protein without a signal peptide and with an unconventional secretion pathway based on the autophagy mechanism, an important pathway for extracellular delivery (Dupont et al., 2011). Autophagy has recently been shown to regulate selective HMGB1 secretion in dying tumor cells (Thorburn et al.l, 2009). An investigation of the interaction of autophagy-based unconventional secretion of HMGB1 and HMGB1 glycosylation is needed to further identify and understand the pathophysiologic mechanism of HMGB1-mediated inflammation. In summary, HMGB1 can be N-glycosylated at N37 and either at residue N134 or N135, and this determines the nucleocytoplasmic transport, extracellular secretion and protein stability of HMGB1.

Human embryonic kidney (HEK) 293T, Chinese hamster ovary (CHO)-K1 cells and HeLa cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Life Technologies), 100 U/ml penicillin, 100 µg/ml streptomycin and 2 mM L-glutamine at 37°C under 5% CO₂. Mouse macrophage cell lines J774A.1 and RAW264.7 were cultured in RPMI1640 medium supplemented with the reagents described above. Plasmid transfection was carried out using Fugene HD reagent (Promega) as recommended by the manufacturer. HEK293T cells were treated for 18 h with 15 ng/ml histone deacetylase inhibitor trichostatin A (TSA, Sigma-Aldrich), 250 nM phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich) and 20 μM H₂O₂ (Sigma-Aldrich), to induce acetylation, phosphorylation and oxidation, respectively. Proteasome inhibitor MG132 (0.5 μM, Sigma-Aldrich) and protein biosynthesis inhibitor cycloheximide (CHX) (70 μg/ml, Sigma-Aldrich), were also used. J774A.1 cells were treated with 100 ng/ml LPS (E. coli 0111:B4, Sigma-Aldrich) for HMGB1 secretion. Tunicamycin (10 µg/ml, Sigma-Aldrich) was used for inhibition of N-glycosylation in HeLa, RAW264.7 and CHO-K1 cells.

Myc- and enhanced green fluorescent protein (EGFP)-tagged HMGB1^WT and HMGB1 glycosylation mutant genes were inserted into the pCMV plasmid for mammalian cell expression. Site-directed mutations of HMGB1 were generated utilizing pFastBac HT-B-HMGB1 as a template using the QuikChange site-directed mutagenesis kit (Stratagene). Three asparagine residues, N37, N134 and N135 were single or doubly mutated to glutamine: HMGB1 N37Q, N134Q, N135Q, N37Q/N134Q, or N37Q/N135Q. The sequences of primers used for construction of human HMGB1 mutants are as follows: HMGB1 N37Q, 5′-CAGATGCTTCAGTCAATTCTCAGAGTTTTC-3′ (forward) and 5′-GAAAACTCTGAGAATTGGACTGAAGCATCTG-3′ (reverse); HMGB1 N134Q, 5′-AGAGATGTGGCAAAACACTGCTGCA-3′ (forward) and 5′-TGCAGCAGTGTTTTGCCACATCTCT-3′ (reverse); and HMGB1 N135Q, 5′-AGATGTGGAATCAAACTGCTGCAGATG-3′ (forward) and 5′-CATCTGCAGCAGTTTGATTCCACATCT-3′ (reverse).

An EcoRI–XhoI fragment containing full-length HMGB1^WT and HMGB1^N37Q/N134Q (insHMGB1^WT and insHMGB1^N37Q/N134Q) were cloned into pFastBac HT-B, and recombinant baculovirus was produced according to the manufacturer's manual (Invitrogen). The N-terminus six histidine (6× His)-containing HMGB1 WT and N37Q/N134Q plasmids were expressed by SF9 cells for 36 h. The insect cells were then purified on a Ni-NTA column (GE Healthcare) with lysis buffer (25 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% glycerol, 30 mM imidazole, 5 mM β-mercaptoethanol). HMGB1 was then eluted with elution buffer (25 mM Tris-HCl pH 7.5, 300 mM NaCl, 10% glycerol, 300 mM imidazole, 5 mM β-mercaptoethanol). After tobacco etch virus (TEV) protease treatment for 16 h at 4°C, size exclusion chromatography (Superdex-200 column, GE Healthcare) was carried out under buffer containing 25 mM Tris-HCl pH 7.5 and 150 mM NaCl to verify protein homogeneity.

HEK293T cells in six-well plates were transfected with various combinations of plasmids using Fugene HD. For immunoblotting, cells were washed with phosphate-buffered saline (PBS) and lysed in 1× radioimmunoprecipitation assay (RIPA) buffer (GenDEPOT) containing 150 mM NaCl, 1% Triton X-100, 1% deoxycholic acid sodium salt, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris-HCl pH 7.5, 2 mM EDTA and protease inhibitor cocktail. Whole-cell lysates were centrifuged at 20,000 g for 10 min at 4°C. Protein sample buffer was added to the whole-cell lysates (100 mM Tris-HCl pH 6.8, 2% SDS, 25% glycerol, 0.1% bromophenol blue, and 5% β-mercaptoethanol) followed by heating at 94°C for 5 min. In all experiments, 20 μg protein were separated by molecular mass through SDS-PAGE and transferred onto nitrocellulose membranes (GE Healthcare). Non-specific binding sites were blocked by incubating the membranes in Tris-buffered saline (TBS) supplemented with 0.1% Tween 20 and 5% (w/v) low-fat milk for 1 h. Anti-HMGB1 (Abcam; ab18256, rabbit antibody against the region from amino acid 150 to the C-terminus of HMGB1), anti-CRM1 and anti-HA (Santa Cruz Biotechnology), anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH, AbFrontier), anti-ubiquitin-K63 and -K48 (Millipore), and anti-c-Myc (Invitrogen Technology) antibodies were used. Membranes were washed three times for 10 min each with TBS with 0.1% Tween 20 (TBST), probed with the appropriate horseradish peroxidase (HRP)-conjugated secondary antibody for 1 h, and washed three times for 10 min each with TBST. Enhanced chemiluminescent (ECL) substrate (GenDEPOT) was then used for visualization. Membranes were then stripped by submerging in stripping buffer (50 mM glycine, pH 2.2) at room temperature for 1 h under constant shaking.

Immunoprecipitation was performed with 200 μg protein, which was pre-cleared with 20 μl of protein G beads (Sigma-Aldrich) for 1 h at 4°C, followed by overnight incubation with 1 µg of the specific antibody. The next day, samples were added to 20 μl of protein G beads and rotated at 4°C for 2 h. The beads were then washed three times with ice-cold PBS, and then mixed with protein sample buffer followed by heating at 94°C for 5 min. The proteins were separated using SDS-PAGE.

A deglycosidase assay was performed to determine HMGB1 de-glycosylation. Recombinant HMGB1 proteins HMGB1^sig-pep and insHMGB1^WT were resuspended in 20 mM ammonium bicarbonate buffer (pH 8.0) and boiled for 10 min at 98°C. HMGB1^sig-pep (A&RT) was produced in HEK293F cells by adding an ER-targeting Ig signal peptide to HMGB1 (Leahy et al., 1988). The proteins were then treated with PNGase F and incubated overnight at 37°C. Proteins were separated using SDS-PAGE for western blotting.

RAW264.7 cells (1×10⁶ cells/well) were treated with 1 µg/ml of LPS in the presence of 10 µg/ml of brefeldin A and 10 µM nocodazole (Sigma-Adrich) for 24 h to see whether HMGB1 was secreted by the conventional secretion pathway. Culture supernatants and whole-cell lysates were harvested, and HMGB1 was immunoblotted for HMGB1 secretion.

HMGB1 protein was purified from HEK293T cells transfected with a Myc–HMGB1 plasmid for 48 h. Cells were lysed with 1× RIPA buffer and then protein G beads conjugated with anti-Myc antibody were added to the lysates. Beads were rotated at 4°C overnight and then washed three times with ice-cold PBS. Samples were mixed with protein sample buffer, boiled at 94°C for 5 min, and then separated using SDS-PAGE followed by Coomassie Blue staining. The HMGB1 band was extracted from the gel for analysis. Endogenous HMGB1 protein was prepared from whole-cell lysate of HEK293T cells to detect the presence of N-glycosylated residues in naturally expressed HMGB1. Whole-cell lysate was immunoprecipitated with anti-HMGB1 antibody conjugated to protein G beads, which were pre-cleared.

HMGB1 proteins were treated with PNGase F for LC-MS/MS analysis. LC-MS/MS analysis was performed with a nano high performance liquid chromatography (HPLC) system (Agilent) using the nano chip column (150 mm×0.075 mm) for peptide separation. Mobile phase A for liquid chromatography separation was 0.1% formic acid in deionized water and mobile phase B was 0.1% formic acid in acetonitrile. The chromatography gradient was designed for a linear increase from 5% B to 30% B in 25 min, 40% B to 60% B in 5 min, 90% B in 10 min, and 5% B in 15 min. The flow rate was maintained at 300 nl/min. Product ion spectra were collected in the information-dependent acquisition (IDA) mode and were analyzed by Agilent 6530 Accurate-Mass quadrupole time-of-flight (Q-TOF) using continuous cycles of one full scan TOF mass spectrometry reading from 200–1500 m/z (1.0 s) plus three product ion scans from 50–1800 m/z (1.5 s each). Precursor m/z values were selected starting with the most intense ion and using a selection quadrupole resolution of 3 Da. The rolling collision energy feature was used, which determines collision energy based on the precursor value and charge state. The dynamic exclusion time for precursor ion m/z values was 60 s. The mascot algorithm (Matrixscience) was used to identify peptide sequences present in a protein sequence database. Mascot searches against an in-house database of HMGB1 (Genbank accession no. CAG33144.1) were performed with the following criteria: fixed modification, carboxyamidomethylated at cysteine residues, variable modification, oxidized at methionine residues, deamidation at aspartic acid and asparagine, maximum allowed missed cleavage of 2, mass spectrometry tolerance of 100 ppm, and MS/MS tolerance of 0.1 Da. Only peptides resulting from trypsin digests were considered.

For the HMGB1 N-glycan analysis, HEK293T cells were transfected with myc–HMGB1 plasmid for transient expression. Whole-cell lysate was prepared and the protein supernatant was applied to the affinity column of a c-Myc tag immunoprecipitation or co-immunoprecipitation kit (Pierce). The purity of myc–HMGB1 was confirmed to be sufficient by analyzing for N-glycans. A total of 15 μg of myc–HMGB1 protein was treated with PNGase F and then labeled with RapiFlour reagent (Waters). The final solution was diluted with acetonitrile for HILIC enrichment. The glycan enrichment was performed with a Waters GlycoWorks™ HIULIV µElution Plate, based on the manufacturer's protocol. Briefly, the well was washed with 200 µl of deionized water and primed with 200 µl of 85% acetonitrile. Then, the labeled glycan sample was loaded to the well and washed with 2×600 µl of 1:9:90 (v/v/v) formic acid:water:acetonitrile solution. Finally, the glycans were eluted with 3×30 µl of 200 mM ammonium acetate in 5% acetonitrile. The enriched glycan was injected onto the ACQUITY I UPLC Glycan BEH amide column (2.1 mm×150 mm, 1.7 µm particle size, Waters). Separation was performed with eluent A, consisting of 50 mM ammonium formate, and eluent B, consisting of 100% acetonitrile, with a 35-min liner gradient from 25% to 46% eluent A at a flow rate of 0.4 ml/min. The effluent was electro-sprayed into the LTQ Elite (Thermo). Blanks were run prior to the sample run to ensure that there were no significant signals from the solvents or the column. A database search was carried out using a SimGlycan algorithm (SimGlycan v5.0).

In order to observe nuclear–cytoplasmic translocation of HMGB1 in HEK293T cells under various stimulation conditions, HEK293T cells were transfected with EGFP-tagged HMGB1^WT and mutant plasmids and cultured on cover glass in a six-well plate. After 48 h, cells were treated with 50 μM H₂O₂ for 2 h, 250 µM PMA for 4 h and 10 ng/ml TSA for 2 h. Cells were fixed with 4% paraformaldehyde-PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA and 4 mM MgSO₄, pH 7.0) for 20 min at room temperature and washed with cold PBS. After mounting with 4′,6′-diamidino-2-phenylindole (DAPI), cells were observed with a confocal FV1000 microscope (Olympus).

Nuclear and cytosolic fractionation was performed using a nuclear and cytosolic fractionation kit (BioVision). HEK293T cells in six-well plates were transfected with HMGB1 WT, N37Q/N134Q or N37Q/N135Q plasmids and incubated for 48 h. Cells were incubated with Earle's balanced salt solution (EBSS) for 4 h to induce starvation and then harvested by centrifugation at 600 g for 5 min at 4°C. Samples were added to 0.2 ml cytosol extraction buffer (CEB)-A containing dithiothreitol (DTT) and protease inhibitors. Samples were vigorously vortexed for 15 s to fully resuspend the cell pellet, and then incubated on ice for 10 min followed by addition of 11 μl of ice-cold CEB-B. Samples were vortexed for 5 s, incubated on ice for 10 min and then centrifuged for 5 min at 16,000 g. The supernatant fraction (cytoplasmic extraction) was then transferred to a clean pre-chilled tube. The pellets containing the nuclei were resuspended in 100 μl of ice-cold nuclear extraction buffer (NEB) mixture, vortexed for 15 s, returned to ice and vortexed for 10 min, and then centrifuged at 16,000 g for 10 min. The supernatant (nuclear extract) was immediately transferred to a clean pre-chilled tube.

In order to observe the DNA binding of wild-type and mutant HMGB1 proteins, a gel retardation assay was performed. Plasmid DNA was prepared by EcoRI digestion and then purified. Recombinant HMGB1^WT and HMGB1^N37Q/N134Q were then incubated with plasmid DNA at 37°C for 2 h and migration was analyzed on a 1% agarose gel.

A FRAP assay was performed to analyze the binding of HMGB1 to nuclear DNA within the nucleus by measuring the mobility of HMGB1. HEK293T cells were seeded on covered glass-bottom dishes and transfected with GFP-tagged HMGB1 WT, N37Q/N134Q or N37Q/N135Q plasmids for 48 h. Transfected cells were viewed under a confocal FV1000 fluorescent microscope and bleached using laser diode at 405 nm at 90% for 3 s. The nuclei were then scanned to measure the recovery rate of GFP fluorescence by region optical intensity (ROI) value. FRAP graphs were created using Liveplot.

In order to draw a model of the N-glycosylated HMGB1 structure, we used the complex structures of HMGB1 A box and B box bound to dsDNA as templates (PDB ID 1CKT and 2GZK, respectively). Two N-acetyl-D-glucosamine (GlcNAc) molecules, which are the first two sugars of glycosylation, were manually attached to HMGB1 N37 and N134 using the WinCoot program (Emsley and Cowtan, 2004). The cartoon or stick structures were produced by the PyMOL program (PyMOL, http://www.pymol.org).

Analysis of experiments was performed with Student's t-test using GraphPad Prism. Data represent the mean±s.e.m., as indicated in the individual figure legends. The difference was considered statistically significant when P<0.05.

We thank Nahee Kim and Young Joo Sun for providing a protocol of HMGB1 protein purification.