Deacetylation-mediated interaction of SIRT1-HMGB1 improves survival in a mouse model of endotoxemia

Inflammatory signal-mediated release of high-mobility group box 1 (HMGB1) is a damage-associated molecular pattern or alarmin. The inflammatory functions of HMGB1 have been extensively investigated; however, less is known about the mechanisms controlling HMGB1 release. We show that SIRT1, the human homolog of the Saccharomyces cerevisiae protein silent information regulator 2, which is involved in cellular senescence and possibly the response to inflammation, forms a stable complex with HMGB1 in murine macrophage RAW264.7 cells. SIRT1 directly interacted with HMGB1 via its N-terminal lysine residues (28–30), and thereby inhibited HMGB1 release to improve survival in an experimental model of sepsis. By contrast, inflammatory stimuli such as lipopolysaccharide (LPS) and tumor necrosis factor-α promoted HMGB1 release by provoking its dissociation from SIRT1 dependent on acetylation, thereby increasing the association between HMGB1 and chromosome region maintenance 1, leading to HMGB1 translocation. In vivo infection with wild-type SIRT1 and HMGB1K282930R, a hypo-acetylation mutant, improved survival (85.7%) during endotoxemia more than infection with wild-type SIRT1 and HMGB1-expressing adenovirus, indicating that the acetylation-dependent interaction between HMGB1 and SIRT1 is critical for LPS-induced lethality. Taken together, we propose that SIRT1 forms an anti-inflammatory complex with HMGB1, allowing cells to bypass the response to inflammation.

pathways 5,11 . HMGB1 has two non-classical nuclear export signals and, therefore, shuttles continually from the nucleus to the cytoplasm; however, the equilibrium is almost completely toward the nuclear accumulation of the protein in quiescent cells 12 . By contrast, HMGB1 translocates from the nucleus to the cytoplasm upon the activation of monocytes by inflammatory signals such as LPS or tumor necrosis factor (TNF)-α through the hyper-acetylation of two major clusters of lysine residues within two nuclear localization sequence (NLS) sites 12 . This acetylation-associated translocation is mediated by chromosome region maintenance 1 (CRM1), a nuclear exportin 13 . Serine phosphorylation by TNF-α is another requisite step for the nucleocytoplasmic translocation of HMGB1 in macrophages 14 . Although these findings suggest that post-translational modifications of HMGB1 are critical for its release, it is unclear how these specific modifications control HMGB1 release 12,14 .
SIRT1, a mammalian ortholog of yeast silent information regulator 2, is a NAD + -dependent class III protein deacetylase that governs a number of genetic programs acting on a wide range of histone and non-histone substrates [15][16][17] . SIRT1 emerged as a critical regulator of various metabolic and pathophysiological processes, such as mitochondrial biogenesis, cellular senescence, energy metabolism, stress resistance, and inflammation, by coordinating complex gene expression programs through the deacetylation of histones, transcription factors, and co-regulators [15][16][17] . In addition, SIRT1 was directly implicated in the modulation of inflammatory responses by deacetylating histones and critical transcription factors such as nuclear factor kappa B and activation protein 1, resulting in the transcriptional repression of various inflammation-related genes 18,19 . Furthermore, reduction in the level and activity of SIRT1 is closely correlated with chronic inflammatory conditions 20 . Knockout or knockdown of SIRT1 leads to increased cytokine release, whereas SIRT1 activators inhibit production of TNF-α , monocyte chemoattractant protein 1, and interleukin (IL)-8 21,22 , stressing the pivotal role of SIRT1 in cellular inflammatory control and the inflammatory response.
Recently, we and others demonstrated that upregulation and activation of SIRT1 inhibits LPS-primed or caloric restriction-mediated HMGB1 release in vitro and in vivo by unidentified mechanisms 23,24 .
Here, we report that HMGB1 release is modulated by SIRT1 in macrophages and an animal model of endotoxemia. SIRT1 physically interacts with and deacetylates HMGB1 at multiple lysine residues located at NLS sites, thereby increasing its association with HMGB1 and leading to retention of HMGB1 in the nucleus. These findings shed light on the regulation of HMGB1 release and have important implications in understanding the molecular mechanism underlying the inflammatory reaction, which may aid and encourage the development of new anti-inflammatory drugs.

HMGB1 physically interacts with SIRT1.
Our recent study showed that SIRT1 is a critical factor in the negative regulation of HMGB1 release 23 . To further investigate the detailed mechanism, we examined the interaction between HMGB1 and SIRT1 by co-immunoprecipitation. Lysates of HEK293T cells expressing epitope-tagged proteins were mixed with an anti-Flag antibody, and the resulting immune complexes were analyzed by immunoblotting with an anti-Myc antibody. Immunoprecipitation of HMGB1 from lysates of co-transfected cells resulted in the co-precipitation of SIRT1 (Fig. 1A). We also detected this interaction reciprocally by using an anti-Myc antibody for immunoprecipitation and an anti-Flag antibody for immunoblotting of the precipitate (Fig. 1B). Control mouse immunoglobulin G (IgG) did not precipitate any proteins. Similar conclusions were reached using in vitro protein-binding assays in which HMGB1-containing cell lysates were incubated with bacterially produced GST-fused SIRT1 protein (Fig. 1C). Consistent with these results, confocal microscopy showed the co-localization of ectopically expressed red fluorescent protein (RFP)-tagged HMGB1 and green fluorescent protein (GFP)-tagged SIRT1, mainly in the nuclei of HEK293T cells (Fig. 1D).
To identify the regions of HMGB1 that are responsible for its interaction with SIRT1, we generated a panel of HMGB1 deletion mutants (Fig. 1E). These mutants were individually transfected into HEK293T cells together with Myc-SIRT1, and co-immunoprecipitation was performed to evaluate the ability of these mutants to bind to SIRT1. HMGB1-full-length (FL) and HMGB1-A&B displayed a strong interaction with SIRT1, while HMGB1-B&C showed no interaction, suggesting that the N-terminal region of HMGB1 is indispensable for its interaction with SIRT1 (Fig. 1F). To dissect which part of the N-terminal region of HMGB1 interacts with SIRT1, we performed immunoprecipitation with mutants containing only the A-box or B-box. As expected, HMGB1-A interacted strongly with SIRT1, whereas HMGB1-B did not, indicating that the A-box of HMGB1 mediates its interaction with SIRT1 (Fig. 1G). Similar results were obtained by GST pull-down assays in which cell lysates containing HMGB1 deletion mutants were incubated with bacterially produced GST-fused SIRT1 protein (Fig. 1H). These observations provide the first indication that HMGB1 and SIRT1 can form physical complexes with each other.
LPS promotes the dissociation of HMGB1 and SIRT1 leading to HMGB1 release. HMGB1 released into the extracellular milieu acts as a proinflammatory cytokine in diverse pathological conditions 9 ; therefore, homeostatic regulation of this release appears to be essential. We investigated whether SIRT1 functions in this context via its direct interaction with HMGB1. The complex of HMGB1 and SIRT1 dramatically dissociated in the presence of LPS, as judged by co-immunoprecipitations with anti-Flag ( Fig. 2A) and anti-Myc (Fig. 2B) antibodies, in HEK293T cells ectopically expressing HMGB1 and SIRT1. Similar results were obtained using another anti-Flag antibody to verify the antibody specificity (Supplemental Fig. S1A). Among stimulators that cause HMGB1 release in monocytic cell lines, including HL-60, U-937, and RAW 264.7 (Supplemental Fig. S1B,C), LPS and TNF-α triggered dissociation of the HMGB1 and SIRT1 complex, while polyinosinic-polycytidylic acid (Poly (I:C)) and interferon (IFN)-γ did not (Fig. 2C). Similar results along with increased acetylation of HMGB1 were observed in a GST pull-down assay using bacterially produced GST-fused SIRT1 protein (Fig. 2D,E). In addition, the complex of HMGB1 and SIRT1 was observed in RAW 264.7 cells without ectopic expression of these proteins under quiescent conditions, and this complex was dissociated in the presence of LPS, leading to the release of HMGB1 into the extracellular milieu (Fig. 2F). HEK293T cells were co-transfected with the indicated plasmids for 48 h, and then whole-cell lysates were prepared and immunoprecipitated with IgG, anti-Flag, or anti-Myc antibody. The immunoprecipitates and total lysates (input) were subjected to immunoblot analysis with anti-Flag, anti-Myc, and anti-β -actin antibodies to detect HMGB1, SIRT1, and β -actin, respectively. Two percent of whole-cell lysates were used as the input. (C) HEK293T cells were transfected with Flag-tagged HMGB1 for 48 h, and whole-cell lysates were incubated with recombinant GST or GST-SIRT1 fusion protein immobilized to glutathione-Sepharose 4B beads for 20 h. Bead-bound proteins were analyzed by Western blotting. GST and GST-fused proteins were stained with Ponceau S. HMGB1 contains reversibly acetylated lysine residues important for its release. Posttranslational modifications such as acetylation are critical for the release of HMGB1 into the extracellular milieu 12 ; therefore, we sought to determine whether acetylation affects complex formation between HMGB1 and SIRT1. When monocytic cells were stimulated with diverse signals to release HMGB1, the level of acetylated HMGB1 in immunoprecipitates increased (Supplemental Fig. S1D); the largest increase was over 3-fold in HEK293T cells treated with LPS for 3 h (Supplemental Fig. S1E). To evaluate whether this acetylation of HMGB1 correlated with its dissociation from SIRT1, we used p300/ CBP-associated factor (PCAF), which acetylated HMGB1 (Supplemental Fig. S2A). When HA-tagged PCAF was ectopically expressed in HEK293T cells, the association between HMGB1 and SIRT1 was markedly decreased, indicating that PCAF-mediated acetylation of HMGB1 hinders its interaction with SIRT1 (Fig. 3A). This acetylation-mediated dissociation of HMGB1 from SIRT1 was also demonstrated in LPS-stimulated RAW 264.7 cells with increased release of HMGB1 (Fig. 3B). These observations indicate that acetylation of HMGB1 causes it to dissociate from SIRT1, thereby promoting the release of HMGB1 into the extracellular milieu.
The A-box of HMGB1 was a requisite region for its interaction with SIRT1; therefore, we constructed A-box deletion mutants of HMGB1 to identify possible acetylation sites (Fig. 3C). These mutants were individually co-transfected into HEK293T cells together with Myc-SIRT1. A-box and ∆ 11 A-box interacted with SIRT1, and this was abolished by LPS, while ∆ 30 A-box displayed no interaction with SIRT1, indicating that N-terminal amino acids 12-30 of HMGB1 are essential for its interaction with SIRT1 (Fig. 3D). Similar results were obtained when cells were stimulated with TNF-α (Supplemental Fig. S2B).
To dissect the critical lysine residue(s) responsible for the interaction and dissociation of HMGB1 with and from SIRT1, we analyzed the amino acid sequences within N-terminal 12-30 residues of HMGB1. Within this region, HMGB1 has three lysine residues at positions 28, 29, and 30, which are evolutionarily well-conserved among diverse species (Fig. 3E). To confirm the importance of these lysine residues to the  interaction between HMGB1 and SIRT1, we introduced amino acid substitutions at residues 28, 29, and 30 of HMGB1 (HMGB1 K282930R ), replacing the normally present lysine with arginine, which mimics the hypo-acetylation state 25 . When the HMGB1 K282930R mutant was co-transfected together with Myc-SIRT1 into HEK293T cells, LPS-or TNF-α -mediated dissociation of HMGB1 and SIRT1 was not observed, in contrast to when wild-type HMGB1 was used (Fig. 3F, Supplemental Fig. S2C). The involvement of acetylation in the dissociation of HMGB1 from SIRT1 was further demonstrated using a mutant in which these three lysine residues were mutated to glutamate (HMGB1 K282930Q ), which mimics the hyper-acetylation state 26 . In this case, complex formation between HMGB1 and SIRT1 was dramatically decreased in quiescent cells in comparison to when wild-type HMGB1 was used (Fig. 3G). However, the individual substitution of any of these three lysine residues did not perturb the dissociation of HMGB1 and SIRT1 in response to LPS or TNF-α , suggesting that acetylation of all three residues underlies this dissociation (Supplemental Fig. S2D,E). These results suggest that lysine residues 28, 29, and 30 of HMGB1 are critical for its interaction with SIRT1 and that inflammatory signal-mediated acetylation of HMGB1 at these lysine residues promotes the relocation of HMGB1 to the cytoplasm, switching a nuclear protein into a cytokine in response to inflammatory stimuli.
To verify that SIRT1 forms a complex with HMGB1 dependent on deacetylation to interfere with the release of HMGB1, adenoviruses expressing wild-type proteins (Ad-Flag-HMGB1 and Ad-Myc-SIRT1) or a deacetylated null mutant of HMGB1 (Ad-Flag-HMGB1 K282930R ) were constructed. Although complexes of HMGB1 and SIRT1 were detected in RAW 264.7 cells co-infected with Ad-Myc-SIRT1 and Ad-Flag-HMGB1 or Ad-Flag-HMGB1 K282930R , LPS-stimulated dissociation was markedly reduced for the complex of Ad-Myc-SIRT1 and Ad-Flag-HMGB1 K282930R . This was correlated with inhibition of the release and LPS-induced acetylation of HMGB1 (Fig. 3H). Following infection with these adenoviruses, the levels of exogenous HMGB1 and SIRT1 proteins were similar in the presence and absence of LPS (Fig. 3H, Supplemental Fig. S3A,B). Deacetylation-mediated inhibition of HMGB1 release was confirmed in RAW 264.7 cells treated with LPS or TNF-α (Supplemental Fig. S2F). These results support the hypothesis that HMGB1 and SIRT1 form a complex, maintaining the equilibrium toward the nuclear localization of HMGB1 in quiescent cells.
LPS stimulation rapidly induced the acetylation of HMGB1, which is required for its nuclear translocation and cytoplasmic accumulation 12 ; therefore, we determined whether these three lysine resides were acetylated in cells stimulated with LPS or TNF-α . In HEK293T cells transfected with tagged HMGB1 and SIRT1, lysine residues 28, 29, and 30 of HMGB1 were acetylated following stimulation with LPS or TNF-α , as determined by liquid chromatography-mass spectrometry (Fig. 3I, Supplemental Fig. S2G). However, acetylation of all three lysine residues was not detected in HMGB1 isolated from cells stimulated with IFN-γ or Poly (I:C). Although lysine residue 30 of HMGB1 was acetylated in cells stimulated with IFN-γ or Poly (I:C), this is unlikely to be sufficient to stimulate dissociation of the complex of HMGB1 and SIRT1, suggesting that acetylation of all three lysine residues is required for the dissociation of HMGB1 from SIRT1 and its cytoplasmic relocation (Supplemental Fig. S2H, I).

Nuclear export of HMGB1 via its interaction with CRM1 is negatively regulated by SIRT1.
CRM1 is an evolutionarily conserved protein that is an essential mediator of chromatin structure maintenance and nuclear protein export 27 . To investigate if CRM1 is involved in the export of HMGB1 following its acetylation-mediated dissociation from SIRT1, we examined the interaction between HMGB1 and CRM1 by co-immunoprecipitations. Upon stimulation with LPS or TNF-α , the amount of CRM1 immunoprecipitated with an anti-Flag antibody was increased (Fig. 4A), indicating a potential interaction with HMGB1. By contrast, the interaction between HMGB1 and SIRT1, as judged by co-immunoprecipitations, was substantially attenuated upon stimulation with LPS or TNF-α , suggesting the affinity for HMGB1 is inclined to the CRM1 from SIRT1 (Fig. 4B,C). However, the interaction between CRM1 and HMGB1 was not affected in HEK293T cells transfected with HMGB1 K282930R , even residues (lysine 28, 29, and 30) and alignment of the flanking regions of these residues of mouse HMGB1 with those of other species. The conserved lysine residues are highlighted in red. The following abbreviations are used: H, human; M, mouse; R, rat; B, cattle; S, pig; C, dog; O, rabbit; G, chicken; D, zebrafish; X, frog. (F,G) HEK293T cells co-transfected with Myc-SIRT1 and Flag-HMGB1 or Flag-HMGB1 mutants for 48 h were stimulated with LPS (100 ng/ml) for 3 h, and then whole-cell lysates were immunoprecipitated with an anti-Flag antibody. (H) RAW 264.7 cells infected with Ad-HMGB1, Ad-HMGB1 K282930R , and/or Ad-SIRT1 for 48 h were treated with LPS (100 ng/ml) for 6 h (for acetyl-HMGB1 and protein interaction) or 24 h (for HMGB1 release), and then whole-cell lysates were immunoprecipitated with an anti-Flag antibody. Release of HMGB1 was analyzed by immunoblotting the conditioned media. (I) HEK293T cells co-transfected with Myc-SIRT1 and Flag-HMGB1 for 48 h were stimulated with LPS (100 ng/ml) for 3 h, and then whole-cell lysates were immunoprecipitated with an anti-Flag antibody. The immunoprecipitates were digested with trypsin and subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. The fragmentation spectrums of 13 MSSYAFFVQTCREEHK (ac) K 29 , 29 K (ac) K (ac) HPDASVNFSEFSK 43 , and 30 K (ac) HPDASVNFSEFSK 43 revealed the presence of peptides with acetylation at lysine residues 28, 29, and 30, respectively. Procedure for LC-MS/MS analysis is schematically illustrated.
Scientific RepoRts | 5:15971 | DOi: 10.1038/srep15971 upon LPS or TNF-α stimulation (Fig. 4D). Furthermore, the interaction between HMGB1 and CRM1 was dramatically increased in HEK293T cells transfected with HMGB1 K282930Q , even in the absence of stimuli, indicating that acetylation-mediated dissociation of HMGB1 from SIRT1 is critical for the interaction of HMGB1 with CRM1 (Fig. 4E). Next, we examined if this acetylation-mediated interaction of HMGB1 and CRM1 is linked to the release of HMGB1 into the extracellular milieu upon LPS or TNF-α stimulation in RAW 264.7 cells ectopically expressing epitope-tagged proteins. In cells expressing wild-type HMGB1, the extracellular level of HMGB1 was increased upon LPS or TNF-α stimulation, and this was further increased in cells transfected with CRM1, indicating that CRM1 is critical for the shuttling of HMGB1. However, this increased release of HMGB1 was almost completely abolished in cells expressing HMGB1 K282930R , even in cells transfected with CRM1, suggesting that the deacetylation-mediated interaction between HMGB1 and SIRT1 is important in the regulation of HMGB1 release (Fig. 4F).
Acetylation is a critical determinant of HMGB1 relocation to the cytoplasm. To determine the importance of lysine residues 28, 29, and 30 of HMGB1 in its intracellular localization, we further examined the cellular localizations of HMGB1 and SIRT1 using fluorescent fusion proteins of wild-type HMGB1, HMGB1 K282930R , and SIRT1 by confocal fluorescence microscopy. In Chinese hamster ovary (CHO) cells, wild-type RFP-HMGB1 localized in the nuclear region and co-localized with GFP-SIRT1. Upon stimulation with LPS or TNF-α , although the majority of RFP-HMGB1 protein remained in the nuclear region, numerous signals were detected in the cytoplasm with a diffuse staining pattern. By contrast, this LPS-or TNF-α -induced cytoplasmic localization was almost completely abolished in cells expressing RFP-HMGB1 K282930R (Fig. 5A,B). However, this abolishment was not observed in cells stimulated with Poly (I:C) or IFN-γ (Fig. 6A,B). Consistent with these findings, Poly (I:C) and IFN-γ promoted the dissociation of both HMGB1 and HMGB1 K282930R from SIRT1, whereas LPS and TNF-α only stimulated the dissociation of the complex between SIRT1 and HMGB1, not between SIRT1 and HMGB1 K282930R (Fig. 6C,D). These results indicate that although Poly (I:C) and IFN-γ stimulate HMGB1 release, similar to LPS and TNF-α , in monocytic cells (Supplemental Fig. S1B,C), Poly (I:C)-or IFN-γ -mediated dissociation of HMGB1 from SIRT1 is independent of acetylation of lysine residues 28, 29, and 30 of HMGB1, which is facilitated by LPS and TNF-α .
Hyper-acetylation is a critical signal for the relocation of HMGB1 11 ; therefore, we examined a fusion protein with hyper-acetylation mutations (RFP-HMGB1 K282930Q ). The localization of RFP-HMGB1 K282930Q was shifted toward the cytoplasm even in the absence of stimuli, similar to the localization of wild-type HMGB1 in the presence of stimuli (Fig. 5A,B). These data suggest that HMGB1 K282930R with more nuclear localization is still capable of interacting with SIRT1, while HMGB1 K282930Q lose the ability to interact with SIRT1. Therefore, it is most likely that deacetylation is inevitable event for the interaction of HMGB1 and SIRT1. This fits well with the established notion that post-translational modifications of HMGB1, such as acetylation, regulate its release 12 .
To further clarify the functional significance of SIRT1 in HMGB1 release, we assessed the impact of SIRT1 deacetylase activity on the interaction between HMGB1 and SIRT1. Activation of SIRT1 by resveratrol almost completely reversed LPS-induced dissociation of HMGB1 from SIRT1 (Fig. 7C). Regulation of SIRT1 activity by resveratrol or sirtinol, an inhibitor of SIRT1 29 , was also correlated to the acetylation level and release of HMGB1 in RAW 264.7 cells expressing epitope-tagged proteins (Fig. 7D). Furthermore, small interfering RNA (siRNA)-mediated knockdown of SIRT1 reduced the interaction between HMGB1 and SIRT1, thereby increasing the release of HMGB1 from RAW 264.7 cells (Fig. 7E), suggesting that SIRT1 has an anti-inflammatory function by inhibiting HMGB1 release.

HMGB1 release is correlated with its acetylation status in endotoxemia model mice. SIRT1
inhibited LPS-or TNF-α -induced HMGB1 release from macrophages by directly interacting with HMGB1 in an acetylation-dependent manner; therefore, we next analyzed whether SIRT1 affected the circulating HMGB1 level during endotoxemia, a standard model of systemic inflammation. BALB/c mice infected with Ad-Flag-HMGB1, Ad-Flag-HMGB1 K282930R , and/or Ad-Myc-SIRT1 via the tail vein were challenged with LPS to induce lethal endotoxemia. Expression of Flag-HMGB1, Flag-HMGB1 K282930R , and Myc-SIRT1 in heart, kidney, liver, and lung was observed in mice infected with adenoviruses in the presence or absence of LPS (Supplemental Fig. S5A,B). Complexes of Flag-HMGB1 and Myc-SIRT1 were detected in co-immunoprecipitated tissue lysates, and this was markedly reduced by LPS treatment. However, LPS-induced suppression of this co-immunoprecipitation was almost completely reversed in the tissues of mice infected with Ad-Flag-HMGB1 K282930R and Ad-Myc-SIRT1 (Fig. 8A). In line with these results, the serum level of Flag-HMGB1 was significantly increased in mice infected with Ad-Flag-HMGB1 and Ad-Myc-SIRT1 following endotoxin challenge, whereas it was significantly reduced in mice infected with Ad-Flag-HMGB1 K282930R and Ad-Myc-SIRT1 (Fig. 8B). These results support the hypothesis that SIRT1 forms a complex with and deacetylates HMGB1 in vivo, thereby inhibiting LPS-induced release of HMGB1.
Particular attention was paid to HMGB1 in the context of LPS-induced endotoxemia, wherein HMGB1 can reportedly exacerbate pathogenic inflammatory responses 4,30 . We therefore examined whether the acetylation status of HMGB1 is related to the lethality and survival rate of endotoxemia model mice. When mice were infected with Ad-Flag-HMGB1, their sensitivity to endotoxins was increased (data not shown). Infection of Ad-Flag-HMGB1 K282930R and Ad-Myc-SIRT1 conferred significant protection against lethality and improved survival during endotoxemia (survival rate, 85.7%) compared to infection of Ad-Flag-HMGB1 K282930R alone (survival rate, 15.3%). This protective effect was not observed in mice infected with Ad-Flag-HMGB1 and/or Ad-Myc-SIRT1, indicating that the acetylation-dependent interaction of HMGB1 and SIRT1 is critical in LPS-induced lethality (Fig. 8C). There were no late deaths of adenovirus-infected animals during the 2 weeks after LPS injection, indicating that SIRT1-mediated inhibition of HMGB1 release conferred lasting protection and did not merely delay the onset of death.
We then examined the serum levels of proinflammatory cytokines that are thought to participate in the pathogenic responses to endotoxemia. The serum levels of TNF-α and IL-6 in mice infected with Ad-Flag-HMGB1 were significantly increased by LPS treatment, while these increases were reduced in the presence of Ad-Myc-SIRT1 (Fig. 8D). Furthermore, infection of Ad-Flag-HMGB1 K282930R and Ad-Myc-SIRT1 almost completely abolished LPS-induced secretion of these cytokines, yielding levels similar to those in the control group. These results suggest that SIRT1-mediated hypo-acetylation of HMGB1 attenuates the secretion of proinflammatory cytokines such as TNF-α and IL-6 in endotoxemia,

Discussion
HMGB1, a cytokine as well as a nuclear architectural protein, elicits particular functions depending on its localization 2,4,[8][9][10] . Although the functions of extracellular HMGB1 in conjunction with its nuclear actions have been well-documented, few of the regulatory mechanisms that determine its cellular localization have been elucidated. In a recent study, we showed that the NAD + -dependent deacetylase SIRT1 is an epigenetically regulated anti-inflammatory gene that can functionally cooperate with HMGB1 in cellular inflammation 23 . Here, we present a key mechanism via which post-translational modification of HMGB1 determines its cellular localization, and this process occurs through the interaction of HMGB1 with SIRT1 under inflammatory stimuli. SIRT1 directly interacts with HMGB1, and this protein-protein interaction is favored in quiescent cells. However, this complex dissociates in response to inflammatory signals in an acetylation-dependent manner, leading to the release of HMGB1, a late mediator of endotoxic shock lethality 4 . Loss or gain of SIRT1 function clearly showed that the acetylation level of HMGB1 is intimately related with cellular inflammatory responses 23,24,28,31 . This may indicate a role for the interaction between SIRT1 and HMGB1 in the anti-inflammatory response, i.e., SIRT1-mediated deacetylation inactivates HMGB1 to assist the anti-inflammatory response. In line with this notion, the deacetylation-mediated interaction of HMGB1 and SIRT1 in mice was sufficiently potent to robustly protect against endotoxemia in response to LPS challenge by inhibiting the secretion of HMGB1 and cytokines such as TNF-α and IL-6.
Lysine residues 28, 29, and 30 of HMGB1 were identified as being part of a putative region that mediates the interaction with SIRT1 in an acetylation-dependent manner. Posttranslational modification of HMGB1 reportedly modulates its subcellular localization, either positively or negatively 12,32,33 . In line with previous studies, inflammatory stimuli induced acetylation of lysine residues 28, 29, and 30 in the N-terminal region of HMGB1, which includes the NLS domain 12 . This stimuli-mediated acetylation promoted the dissociation of HMGB1 and SIRT1, leading to alteration of the subcellular localization of HMGB1. This effect of acetylation on HMGB1 localization correlated with the deacetylase activity of SIRT1, indicating that SIRT1 interacts with and deacetylates HMGB1, thereby preventing its release. Accordingly, acetylation of these sites appears to induce a conformational change in the binding domain of HMGB1 and, hence, alter its interaction with SIRT1. HMGB1 K282930Q , a hyper-acetylation mutant, exhibited a significantly reduced interaction with SIRT1, while HMGB1 K282930R , a hypo-acetylation mutant, exhibited an increased interaction with SIRT1 in comparison to wild-type HMGB1, even in the presence of inflammatory stimuli. These findings are consistent with previous studies demonstrating that inflammation-and cellular stress-mediated acetylation of HMGB1 prevents its nuclear reentry and leads to the accumulation of HMGB1 in the cytoplasm 12,32,33 . Similarly, JAK/STAT-or interferon regulatory factor 1-mediated hyper-acetylation of HMGB1 stimulates its release 11,34 . Thus, epigenetic modification of HMGB1 by acetylation has emerged as a critical regulator that can determine the localization of HMGB1. Such findings provide insight into the key role of SIRT1 as a binding partner that maintains HMGB1 in a hypo-acetylated state to inhibit its cytoplasmic accumulation and extracellular release. Accordingly, understanding the mechanisms by which inflammatory cells regulate HMGB1 release may enable the targeting of therapeutics to attenuate HMGB1-related inflammation by the selective activation or expression of the SIRT1 Although HMGB1 is released into the extracellular milieu in response to cellular stimuli such as LPS, TNF-α , IFN-γ , and Poly (I:C), these stimuli did not equally affect the interaction between HMGB1 and SIRT1. In this study, we identified acetylation of lysine residues 28, 29, and 30 of HMGB1 as a key aspect of the regulation of its active release from cells stimulated with inflammatory signals. In line with these findings, LPS and TNF-α induced the acetylation of lysine residues 28, 29, and 30 of HMGB1, whereas IFN-γ and Poly (I:C) only induced the acetylation of lysine residue 30. This difference in the acetylated residues of HMGB1 might be attributed to the induction of different signaling cascades by each stimulus: specifically, LPS transduces signals via Toll-like receptor 4-mediated pathways, while IFN-γ activates JAK-STAT signaling pathways 12,35 . Accordingly, it seems most feasible that the location of acetylation determines the signaling cascades that mediate the dissociation of HMGB1 and SIRT1, which, depending on the nature of the stimulus, can lead to the release of HMGB1. During release of HMGB1 following stimulation, HMGB1 is heavily acetylated and relocates to the cytoplasm through an association with CRM1, a nuclear export receptor 5,12 . Formation of a complex between HMGB1 and CRM1 accompanies LPS-or TNF-α -induced release of HMGB1 13,36 . In addition, leptomycin B, a CRM1 inhibitor, significantly blocks LPS-induced nuclear export of HMGB1 12 . To our knowledge, this is the first study to demonstrate that the specific sites of acetylation modulate HMGB1 release in response to different stimuli via a protein-protein interaction. These novel findings have important implications regarding our understanding of the molecular mechanisms underlying the anti-inflammatory effect of SIRT, as well as the regulation of HMGB1 release.
Of particular interest is the possibility that the acetylation-dependent interaction of SIRT1 and HMGB1 participates in the pathophysiology of sepsis. Pharmacological or genetic manipulation of SIRT1 markedly attenuated LPS-and TNF-α -induced release of HMGB1 in a process mediated by acetylation. Ectopic expression of the hypo-acetylated mutant HMGB1 K282930R inhibited LPS-induced increases in the level of circulating HMGB1, indicating that HMGB1 release is tightly regulated by the acetylation status of these residues. Furthermore, expression of HMGB1 K282930R reduced endotoxin-induced lethality of LPS in mice. These effects are intimately correlated with the interaction between HMGB1 and SIRT1 as well as the secretion of secondary cytokines such as TNF-α and IL-6 in endotoxemic mouse tissues. These findings are in line with previous studies reporting that HMGB1 is a novel deacetylation target of SIRT1, and that its release and nuclear translocation are intimately linked to SIRT1 deacetylase activity, emphasizing the critical role of SIRT1 in inflammatory responses 23,24,31,37 . In fact, HMGB1 was deacetylated by SIRT1 at four lysine residues (55, 88, 90 and 177) in quiescent endothelial cells 31 . However, these lysine residues were not involved in SIRT1-mediated control of HMGB1 release in the LPS-stimulated murine macrophages. Thus, it may be possible to target SIRT1 to selectively inhibit HMGB1 release without significantly compromising innate immune responses. Modulation of SIRT1 deacetylase activity by pharmacological or genetic manipulation altered the acetylation-dependent release of HMGB1 upon inflammatory stimulation. Consistent with the present findings, pharmacological activation of SIRT1 by resveratrol significantly inhibits HMGB1 release and reduces septic liver injury 24,31,37 . Accordingly, targeting of SIRT1 in inflammation-related diseases may elicit therapeutic effects by decreasing the extracellular level of HMGB1.
In the current study, we demonstrated that SIRT1 regulates the release of the proinflammatory cytokine HMGB1 via a direct interaction mediated by deacetylation (Fig. 8E). Consequepgntly, the physical interaction between SIRT1 and HMGB1 is associated with a blunted inflammatory response to endotoxin stimuli, leading to a significant increase in the survival of endotoxemic animals.
Ethics statement. All animal studies were performed in accordance to the Korean College of Laboratory Animal Medicine's guidelines for Animal Care and Use. These regulations were approved by the Institutional Animal Care and Use Committee of Konkuk University (approval number: KU14118).
Serum cytokine analysis. Serum levels of HMGB1, TNF-α , and interleukin (IL)-6 were analyzed in circulating blood samples obtained from BALB/c mice infected with 1 × 10 10 particles of purified recombinant adenoviruses expressing Ad-LacZ, Ad-Flag-HMGB1, Ad-Flag-HMGB1 K282930R , and/or Ad-Myc-SIRT1 in 100 μ l of saline with or without LPS for 16 h. Blood was collected, allowed to clot for 2 h at room temperature, and centrifuged for 20 min at 4,000 rpm as described previously 45 . Circulating Flag-HMGB1 in serum was determined by Western blot analysis. Serum levels of TNF-α and IL-6 were measured using a mouse TNF-α ELISA Ready-SET-Go! Kit (eBioscience, San Diego, CA, USA) and IL-6 ELISA Max Set Standard (BioLegend, San Diego, CA, USA), respectively. Statistical analysis. Data are expressed as means ± standard error. Statistical significance was determined using a one-way ANOVA, followed by the Tukey-Kramer test. A value of p < 0.05 was considered statistically significant.