Identification of Pharmacological Modulators of HMGB1-Induced Inflammatory Response by Cell-Based Screening

High mobility group box 1 (HMGB1), a highly conserved, ubiquitous protein, is released into the circulation during sterile inflammation (e.g. arthritis, trauma) and circulatory shock. It participates in the pathogenesis of delayed inflammatory responses and organ dysfunction. While several molecules have been identified that modulate the release of HMGB1, less attention has been paid to identify pharmacological inhibitors of the downstream inflammatory processes elicited by HMGB1 (C23-C45 disulfide C106 thiol form). In the current study, a cell-based medium-throughput screening of a 5000+ compound focused library of clinical drugs and drug-like compounds was performed in murine RAW264.7 macrophages, in order to identify modulators of HMGB1-induced tumor-necrosis factor alpha (TNFα) production. Clinically used drugs that suppressed HMGB1-induced TNFα production included glucocorticoids, beta agonists, and the anti-HIV compound indinavir. A re-screen of the NIH clinical compound library identified beta-agonists and various intracellular cAMP enhancers as compounds that potentiate the inhibitory effect of glucocorticoids on HMGB1-induced TNFα production. The molecular pathways involved in this synergistic anti-inflammatory effect are related, at least in part, to inhibition of TNFα mRNA synthesis via a synergistic suppression of ERK/IκB activation. Inhibition of TNFα production by prednisolone+salbutamol pretreatment was also confirmed in vivo in mice subjected to HMGB1 injection; this effect was more pronounced than the effect of either of the agents administered separately. The current study unveils several drug-like modulators of HMGB1-mediated inflammatory responses and offers pharmacological directions for the therapeutic suppression of inflammatory responses in HMGB1-dependent diseases.


Introduction
High-mobility group box 1 protein (HMGB1) was initially considered a nuclear protein regulating gene transcription. However, data emerging over the last decade identified its separate role as a pro-inflammatory cytokine that is released actively and passively from cells during inflammation and injury [1][2][3]. According to a current classification, the immune response can be regulated by endogenous danger signals (damage-associated molecular patterns; DAMPs; alarmins) as well as exogenous pathogen-associated molecular patterns (PAMPs). In this context, HMGB1 has been identified as a bona fide DAMP (i.e. a mediator released during sterile inflammatory processes), as well as a mediator released during PAMP-associated inflammatory events (e.g. sepsis and septic shock), which participates in the pathogenesis of the delayed inflammatory response, organ injury and contributes to disease mortality [1][2][3].
Significant work has focused on the molecular mechanisms of HMGB1 release and on the therapeutic neutralization of HMGB1, either by antibodies, or by inhibiting its binding to its receptors RAGE and TLR4 [1][2][3][4][5][6][7]. Several compounds have been identified that attenuate the release of HMGB1, including glucocor-ticoids, chloroquine, gold salts, nicotinic receptor agonists, ethyl pyruvate and inhibitors of poly(ADP-ribose) polymerase [3,[8][9][10][11]. However, the inflammatory cellular responses downstream from HMGB1 are less understood, and no systematic survey has been conducted to characterize these pathways or to identify their pharmacological modulators. One determinant of the bioactivity of extracellular HMGB1 is based on the redox status of its three conserved thiol groups. The all thiol confirmation has been show to facilitate the binding of CXCL12 to CXCR4 and thus exhibit chemokine-like properties [12]. The C23-C45 disulfide C106 thiol conformation binds to the CD14/MD2/TLR4 receptor complex [13,14] and demonstrates cytokine-like properties. Using a cellbased medium-throughput screening approach, the goal of the current study was to identify drug-like compounds that downregulate the cytokine-like activity of HMGB1-induced inflammatory processes in murine macrophages in vitro.

Materials and Reagents
A comprehensive screening set of 5,546 compounds was gathered comprising the NIH Clinical Collection (446 phase I-III trial compounds) from BioFocus (South San Francisco, CA), the FDA Approved Library (640 FDA approved bioactive compounds) from Enzo Life Sciences (Farmingdale, NY), the Prestwick Chemical Library (1200 marketed drugs in Europe) from Prestwick Chemical (Washington, DC), the US Drug Collection (1040 clinical trial stage USP drugs), the International Drug Collection (240 compounds marketed in Europe or Asia but not in the US) and Killer Plates (160 toxic substances) from MicroSource Discovery Systems (Gaylordsville, CT), the LO-PAC1280 (1280 various biologically active compounds) from Sigma-Aldrich, (Saint Louis, MO) and the Natural Products (640 natural compounds and derivatives) from TimTec LLC (Newark, DE). The compounds were dissolved at 10 mM in dimethylsulfoxide (DMSO) and dilutions were made either in DMSO or in phosphate-buffered saline (PBS, pH 7.4) to obtain 0.5% final DMSO concentration. HMGB1 (C23-C45 disulfide C106 thiol form) was prepared as previously described [15] and diluted in OptiMEM I medium (Invitrogen, Carlsbad, CA). Unless specified otherwise, all other reagents were purchased from Sigma-Aldrich Co. (St. Louis, MO).

Screening Assay
RAW 264.7 cells (100 000/well) were plated into 96-well tissue culture plates and cultured overnight. Culture medium was replaced with OptiMEM prior to adding compounds. Test compounds were supplied at 10 mM in dimethyl sulfoxide (DMSO) and were diluted in DMSO and in phosphate buffer saline (PBS) to reach 3 mM final concentration (and 0.5% DMSO) in the culture medium. The Natural Products Library was screened at 1 mg/ml final concentration. Compounds were administered in 1/20 volume 1 hour prior to HMGB1 treatment. In the combined screen the cells received additional dexamethasone (3 mM) treatment. HMGB1 was added at 5 mg/ml final concentration in 1/10 volume and the cells were incubated for 18 hours at 37uC in 5% CO 2 atmosphere. Supernatant was collected to measure TNFa secretion and LDH release.

RNA Isolation, Gene Expression Measurements
Total RNA was isolated from RAW 264.7 cells exposed to HMGB1 or vehicle for 1.5 or 6 hours using a commercial RNA purification kit (SV total RNA isolation kit, Promega, Madison, WI). 2 mg RNA was reverse transcribed using the High Capacity cDNA Archive kit (Applied Biosystems, Foster City, CA) as previously described [17]. 1 mg RNA was used according to the manufacturer's protocol for gene expression measurements using the Toll-like receptor signaling pathway real-time PCR array (PAMM-0018ZD, SA Biosciences, Frederick, MD) on CFX96 thermocycler (Biorad, Hercules, CA) and analyzed with the tool provided by SA Biosciences. A full list of the genes investigated is deposited in Table S1. Taqman assay for TNFa was performed using a commercial assay (TNFa assay ID: Mm00443260_g1, Life Technologies, Carlsbad, CA) using GAPDH (VIC/MGB Probe, Applied Biosystems, Foster City, CA) control as normalizer.

Pharmacological Modulation of HMGB1-induced TNFa Production in vivo
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Committee on the Ethics of Animal Experiments of the University of Texas Medical Branch, Galveston (Permit Number: 1110054). The procedures were performed humanely with minimal suffering. 6-7 week-old Balb/c male mice (The Jackson Laboratory) were pretreated subcutaneously with 20 mg/kg prednisolone, 10 mg/kg salbutamol, the combination of prednisolone and salbutamol (doses as above), or the glucocorticoid receptor blocker mifepristone (30 mg/kg) or the b-receptor antagonist propranolol (10 mg/kg) or vehicle for 3 hours. Mice were injected intraperitoneally (i.p.) with 0.5 mg/mouse HMGB1 and animals were sacrificed 8 hours later. Serum levels of TNFa were measured by ELISA (as above).
Statistical analysis. Data are shown as means 6 SEM. One-way ANOVA was applied for statistical analysis and for the determination of significance, the Tukey's post-hoc test was used. A p value of ,0.05 was considered statistically significant. All statistical calculations were performed using Graphpad Prism 4 analysis software. Experiments were performed at least 3 times on different days.

HMGB1 Induces Inflammatory Mediator Production and Cytotoxicity in RAW 264.7 Macrophages
HMGB1 (1-10 mg/ml) induced concentration-dependent tumor necrosis factor a (TNFa) secretion by RAW 264.7 cells, an effect, which was potentiated by IFN-c (Fig. 1A). HMGB1 also reduced cell viability (Fig. 1B); this cytotoxic response became Non-toxic compounds that reduced the HMGB1-induced TNFa production by 2 standard deviation values are listed in order of potency, according to their inhibitory potency for TNFa secretion. The source library of the compounds, their known biological activity and the respective viability values are shown. Viability was measured by the MTT assay.   2) and was associated with a suppression of mitochondrial function (Fig. 3) and caspase activation (Fig. 4). In addition to TNFa, HMGB1 also upregulated multiple pro-inflammatory cytokine (IL1a, IL1b, IL6, TNFb) and chemokine (Ccl2, MCP-1, Cxcl10) genes, as well as the anti-inflammatory cytokine IL10 (Fig. 1C). The HMGB1mediated responses were also associated with an upregulation of nuclear factor kB (NF-kB) (Fig. 1C). Moreover, HMGB1 induced a down-regulation of TLR4 and MD2 and upregulation of TLR2, TLR9 and TLR adaptor molecule 1 (Ticam1) (Fig. 1C). Thus, the form of HMGB1 used for the screen exhibited the expected cytokine like properties of C23-C45 disulfide C106 thiol HMGB1.

Identification of Inhibitors of HMGB1-induced TNFa Production by Cell-based Screening
Cell-based screening of a focused library of over 5,000 clinical drugs, natural products and pharmacologically active compounds identified ,2% of the compounds, which suppressed TNFa production, without adversely affecting cell viability (Table 1; Fig. 5). Conversely, a limited number of compounds induced a significant enhancement of HMGB1-mediated TNFa response (Table 2). A full list of the primary screen data is deposited in Table S2.
More than 50% of the hit compounds that inhibited TNFa production were glucocorticoids ( Fig. 5; Table 1). Beta-adrenergic agonists represented the second-most common class. The activity of the hit compounds was next confirmed at 3 and 10 mM. Since glucocorticoids and beta agonists showed a clear class action, only a subset of these compounds was retested. Apart from glucocorticoids and beta agonists, the highest inhibitory activity was detected for the NF-kB inhibitors Bay 11-7085 and parthenolide, and the antioxidant piperlongumine. Increasing the concentration of the compounds to 10 mM did not produce more pronounced inhibitory responses, but approximately 15% of the hit compounds became slightly cytotoxic at this concentration ( Table 3).

Identification of Pharmacological Potentiators of Glucocorticoids by Cell-based Screening
We hypothesized that synergistic drug combinations may be more effective than single agents in controlling HMGB1-induced inflammatory responses. To identify compounds that potentiate the effect of glucocorticoids, a follow-up screen of the NIH Clinical Collection compound library was conducted in the presence of dexamethasone (3 mM). The screen identified beta2 agonists (salbutamol, salmeterol), the phosphodiesterase (PDE) inhibitor rolipram and as prostaglandin E1 as synergistic enhancers of the glucocorticoid's effect (Fig. 6, Table 4). In addition, the dopamine receptor antagonist SCH 23390 (R)-(+)-7-chloro-8-hydroxy-3methyl-1-phenyl-2,3,4,5-tetrahydro-1H-3-benzazepine HCl), the structurally related benzodiazepine lorazepam, the antioxidant ebselen and the d1 opioid receptor agonist SB 205607 decreased TNFa production in the presence of the glucocorticoid. As expected, the glucocorticoid receptor antagonist mifepristone attenuated the effect of dexamethasone (Table 4). A few drugs (e.g. cerivastatin, vindesine, vinorelbine) increased TNFa production in the presence dexamethasone (Table 4); this effect was related to the fact that, according to the results of the primary screen, these compounds, on their own, increase HMGB1-induced TNFa secretion (Table 1).

Glucocorticoid/beta-adrenergic Agonist Synergy: Mechanism of Action
Using prednisolone (a prototypical glucocorticoid) and salbutamol (a prototypical beta 2 adrenergic agonist), follow-up experiments were designed to further characterize the pharmacological properties and underlying mechanisms of the glucocorticoid/betaadrenergic synergy. Both prednisolone and salbutamol, on their own, decreased the HMGB1-induced TNFa production in the low nanomolar concentration range: they reached their maximum effect at around 100-300 nM, exhibiting a 50% inhibition of TNFa production (Fig. 7). Salbutamol (1 mM), in combination   with prednisolone, significantly reduced TNFa production already at 10 nM (compared to salbutamol alone); at 100 nM prednisolone the combination reached its full potential (approximately 70% inhibition). Likewise, the combination of prednisolone (1 mM) with 30 nM salbutamol significantly reduced the HMGB1induced TNFa response (compared to prednisolone alone) and with 300 nM salbutamol the combination reached its full potential (approximately 70% inhibition) (Fig. 7). HMGB1-induced TNFaa secretion was associated with a rapidonset and marked increase in TNFa mRNA (Fig. 8). Prednisolone and salbutamol each decreased the TNFa mRNA level by 50%; combination of the two compounds synergistically inhibited the transcription of TNFa mRNA (Fig. 8). We next tested whether the early inhibition of TNFa production involves upstream signaling events such as mitogen-activated protein kinase (MAPK) activation and IkB phosphorylation. HMGB1 induced an early and sustained activation of the extracellular signal-regulated kinases 1/ 2 (ERK1/2, p44 and p42) and p38 and of IkB phosphorylation (Fig. 9). The combination of salbutamol and prednisolone resulted in a partial, but statistically significant inhibition of ERK1 phosphorylation and IkB phosphorylation (Fig. 9). These data indicate the regulation of HMGB1-mediated cellular signaling by the glucocorticoid/beta-agonist combination has an upstream regulatory component.
To further characterize the effect of the glucocorticoid/betaagonist combination on HMGB1-induced gene transcription, a TLR signaling pathway array was next employed. The responses could be characterized by four distinct expression patterns: a) prednisolone, but not salbutamol inhibiting gene expression, b) salbutamol, but not prednisolone inhibiting gene expression, c) the two compounds synergistically blocking gene expression and d) the two compounds synergistically enhancing gene expression (Fig. 10). The genes which were mostly inhibited by steroids included the interleukins (IL1a, Il1b, IL6, IL10) and Ptgs2 (COX-2); the inhibition exerted by the beta-2 agonist was dominant in case of the chemokines Ccl2 and Cxcl10 and TLR2 and TLR9; synergistic inhibition by the glucocorticoid and the beta-agonist was confirmed for TNFa, as well as demonstrated for lymphotoxin (Lta) and the TLR adaptor Ticam1 (Fig. 10). Unexpectedly, in a few instances, the steroid and the beta2 agonist led to a synergistic enhancement, as seen with Csf3 (GCSF), CD14, CCAAT/enhancerbinding protein beta (Cebpb), interleukin 1 receptor alpha (IL1R1) and TLR8 (Fig. 10).
Given the fact that both glucocorticoids and beta-receptor agonists represent endogenous hormones of the sympatheticadrenal-medullary axis, we have next evaluated whether cortisol and/or adrenaline/noradrenaline, at concentrations that are comparable to their endogenous plasma levels, affect HMGB1induced TNFa production. Cortisol, and, more markedly, the combination of adrenaline and noradrenaline, suppressed the HMGB1-induced TNFa response (Fig. 11).

Glucocorticoid/beta-adrenergic Agonist Synergy in vivo
The combination of prednisolone and salbutamol effectively suppressed HMGB1-induced TNFa production in Balb/c male mice in vivo, and this effect was more pronounced than the effect of either agent alone (Fig. 12). In contrast, the glucocorticoid receptor blocker mifepristone or the b-receptor antagonist propranolol did not enhance HMGB1-induced TNFa production (Fig. 12), suggesting that the response is not under significant control by endogenous glucocorticoids acting on the mifepristone- Figure 6. Combined screening to identify pharmacological potentiators of dexamethasone-mediated inhibition of the HMGB1induced pro-inflammatory response. RAW 264.7 cells were pre-treated with dexamethasone (3 mM) in combination with test compounds and exposed to HMGB1 for 18 hours. TNFa production was measured from the supernatant and the viability of the cells was measured by the MTT assay. A: TNFa responses measured in the combination screen are plotted versus the TNFa production values measured in the single compound screen. TNFa production values higher than MEAN+2SD are shown in red (''steroid inhibitors'') and values lower than MEAN+2SD in green boxes (''potentiators of steroids) for the combination screen. Red dots denote the toxic compounds, green the steroid potentiators and purple those that increase the TNFa production. Compounds that inhibited the HMGB-induced TNFa production in the single compound screen, but failed to potentiate the action of steroids are shown in yellow.

Discussion
It is well established that HMGB1 plays a central role in sterile inflammation [1][2][3]. This screen was undertaken to identify inhibitors of HMGB1-induced, TLR4 dependent TNFa production. The hit compounds emerging from the primary screen included several signal transduction pathway modifiers, such as the IkB phosphorylation inhibitor Bay 11-7085 and the Src/Syk kinase inhibitor MNS. These findings are consistent with a role of NF-kB and kinase activation in HMGB1-mediated cellular signaling. Glucocorticoids and beta-receptor agonist activators of intracellular cAMP (such as such as salbutamol, clenbuterol, metaproterenol, ethylnorepinephrine and ritodrine) were two most prominent drug classes emerging from the screen. Because of their therapeutic potential, and because of the endogenous physiological regulatory implications, these two classes of compounds were subject of follow-up studies (see below). Additional classes of hit compounds included the natural compounds, piperlongumine and parthenolide (the latter compound is known pharmacological actions as a NF-kB and MAP kinase inhibitor). The mechanism of action and potential anti-inflammatory utility of miscellaneous additional compounds that showed inhibitory effects -such as the HIV protease inhibitor indinavir, the local anesthetic lidocaine, the surfactant tyloxapol, the calcium antagonist prenylamine and the diuretic ethacrynic acid -requires further characterization. It is interesting to note that indinavir [18], tyloxapol [19] and lidocaine [20] have previously been demonstrated to suppress TNFa production in various experimental models in vitro, although the underlying molecular pathways have not been fully characterized. It is intriguing to speculate that these compounds may have additional, hitherto unrecognized, secondary modes of pharmacological action (as well as potential therapeutic utility) due to inhibition of HMGB1-mediated inflammatory responses.
We identified several different activators of intracellular cAMP signaling as part of the screen for enhancers of the inhibitory effect of glucocorticoids. The enhancers exerted their effects their effects either through beta-adrenergic receptor agonism (such as salbu- tamol and salmeterol), through prolongation of the intracellular half-life of cAMP (such as the phosphodiesterase inhibitor rolipram) or by activating the cAMP-dependent protein kinase (PKA) (such as prostaglandin E1). While neither the glucocorticoids nor the cAMP-stimulating agents, on their own, produced a complete inhibition of HMGB1-mediated TNFa response, the combination of these two agents yielded a robust inhibition, and did so at low micromolar/nanomolar concentrations. Previous studies have demonstrated synergistic interactions between steroids and beta-agonists in various experimental systems in vitro and suggested that cAMP and glucocorticoids act via distinct upstream pathways, which activate transcription though separate hormone response elements, the glucocorticoid receptor (GR) element (GRE) and the cAMP-response element (CRE), respectively. The site of the synergistic convergence was identified at the level of inhibition of the promoter activation of various pro-inflammatory genes [21][22][23]. Based on our findings, at least some of the synergistic inhibition of HMGB1-induced signaling by the glucocorticoid/beta-agonist combination occurs upstream from NF-kB activation, upstream from GRE and CRE and upstream from the promoter region of the inflammatory genes studied.
Our analysis of the gene expression profiles using a TLR signaling pathway array demonstrated that the synergistic inhibition of HMGB1-induced TNFa production by the glucocorticoid and the beta agonist does not represent a generalized phenomenon. In the case of several mediators, neither the glucocorticoid tested (e.g. Ccl2, Tlr2, Tlr9, Cd14, Cebpb, Csf3, Tlr8), nor the beta agonist tested (e.g. Il-1a, IL1b, IL6, Csf3, IL1r1) showed any inhibition. In some cases an enhancement was seen (IL1ra, Ptgs2, IL-10). These findings clearly demonstrate that HMGB1-mediated pro-inflammatory mediator production is regulated by glucocorticoids and by cAMP in a fashion that is specific to each gene product, and may be, at least in part, related to individual differences in the steroid and cAMP-responsive elements in individual promoters. Nevertheless, the combination of the beta agonist and the glucocorticoid resulted in a partial suppression for the majority of the genes studied, yielding a shift towards an overall anti-inflammatory phenotype (without suppressing the expression of the anti-inflammatory cytokine IL-10).
HMGB1 signals through numerous receptors, depending upon the molecular conformation of the three cysteines [2,[12][13][14][24][25][26]. For example, extracellular HMGB1 is post-translationally regulated via redox mechanisms, and the C23-C45 disulfide C106 (CTL: vehicle treated control, HMGB: cells exposed to HMGB1, Pred: cells pretreated with prednisolone and exposed to HMGB1, Salb: cells pretreated with salbutamol and exposed to HMGB1, Pred+Salb: cells pretreated with both prednisolone and salbutamol and exposed to HMGB1. 1 p,0.05 HMGB1-treated group compared to vehicle treated control, *p,0.05 compared to HMGB1 group, # p,0.05 compared to single compound treatment). doi:10.1371/journal.pone.0065994.g008 thiol conformation binds to and activates the TLR4/MD2 receptor complex in the absence of LPS [12][13][14]. Here we utilized this recombinant conformation of HMGB1 (purified and characterized as previously described), which primarily signals through TLR4 to induce TNFa. Because RAGE and TLR2 are dispensable for this effect, our studies would not be expected to address signaling mediated through these receptors.
Hormones of the hypothalamic-pituitary-adrenal axis, the sympathetic-adrenal-medullary axis, and the sympathetic and parasympathetic arms of the autonomic nervous system have powerful roles in the control of inflammation [27][28][29][30][31]. Adrenalectomy or pharmacological blockade of endogenous glucocorticoid receptors exacerbates [30], while beta-receptor activation suppresses systemic inflammatory responses [31]. Considering the fact that the biologically active concentrations of glucocorticoids and catecholamines in the current study are in the physiological range, we have also explored whether the HMGB1-mediated inflammatory responses are under the tonic control of these hormones. While the combination of exogenous glucocorticoid and beta agonist inhibited HMGB1-induced TNFa production (thereby Figure 9. MAPK activation and IkB phosphorylation in response to HMGB1 are ameliorated in synergy by prednisolone and salbutamol. A: RAW 264.7 cells were exposed to HMGB1 (5 mg/ml) for the indicated length and the phosphorylation of ERK1/2, p38 and IkB was detected. B: RAW 264.7 cells pretreated with prednisolone (1 mM) and salbutamol (1 mM) were exposed to HMGB1 (5 mg/ml) for 30 min (ERK1/2, p38) or 1 hour (IkB) and the activation was detected as phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204), phospho-p38 (Thr 180) or phospho-IkB-a (Ser 32/ 36). C: Bar graph shows the phosphorylation signal normalized to the total amount of the respective protein. (CTL: vehicle treated control, HMGB: cells exposed to HMGB1, Pred: cells pretreated with prednisolone and exposed to HMGB1, Salb: cells pretreated with salbutamol and exposed to HMGB1, Pred+Salb: cells pretreated with both prednisolone and salbutamol and exposed to HMGB1. 1 p,0.05 HMGB1-treated group compared to vehicle treated control, *p,0.05 compared to HMGB1 group). doi:10.1371/journal.pone.0065994.g009 Figure 10. The interaction of prednisolone and salbutamol in the inhibition of HMGB-induced gene expression. RAW 264.7 cells pretreated with prednisolone (1 mM) and salbutamol (1 mM) were exposed to HMGB1 (5 mg/ml) for 1.5 hours and the expression of TLR-associated genes was analyzed with TLR signaling pathways array. Gene expression normalized to control genes (GAPDH, actin, B2m, Gusb, Hsp90ab1) is shown as fold expression values of vehicle treated cells. (CTL: vehicle treated control, HMGB: cells exposed to HMGB1, Pred: cells pretreated with prednisolone and exposed to HMGB1, Salb: cells pretreated with salbutamol and exposed to HMGB1, Pred+Salb: cells pretreated with both prednisolone and salbutamol and exposed to HMGB1. 1 p,0.05 HMGB1-treated group compared to vehicle treated control, *p,0.05 compared to HMGB1 group, # p,0.05 compared to single compound treatment). doi:10.1371/journal.pone.0065994.g010 Figure 11. Inhibition of the HMGB-induced inflammatory response by endogenous catecholamines and glucocorticoids at physiological concentrations. RAW 264.7 cells were pretreated with cortisol (0.7 mM), noradrenaline (0.5 ng/ml), adrenaline (0.5 ng/ml), dexamethasone (1 mM) and salbutamol (1 mM) and exposed to HMGB1 (5 mg/ml) for 18 hours. TNFa secretion was measured in the supernatant.
( 1 p,0.05 HMGB1-treated group compared to vehicle treated control, *p,0.05 compared to HMGB1 group, # p,0.05 cells treated with all compounds in combination versus treated with a combination of two.). doi:10.1371/journal.pone.0065994.g011 extending the in vitro findings to an in vivo system), blockade of the endogenous glucocorticoid receptors with mifepristone or inhibition of the beta receptors with propranolol failed to potentiate the HMGB1-induced TNFa responses in vivo. Thus, circulating HMGB1 does not result in a more severe inflammatory response in subjects with impairment of their endogenous sympatheticadrenal-medullary homeostasis, at least in the current in vivo experimental system.
In summary, the current study unveils several drug-like modulators of HMGB1-mediated inflammatory responses and offers pharmacological directions for the therapeutic suppression of inflammatory responses in diseases driven by the HMGB1-TLR4 axis. Glucocorticoids remain a mainstay of therapy for rheumatoid arthritis, as well as many other inflammatory diseases. In rheumatoid arthritis HMGB1 has been shown to play a significant pathogenetic role [2,3]. We hypothesize that the mode of the therapeutic action of glucocorticoids, in addition to inhibiting HMGB1 release [11], also involves an inhibition of HMGB1's downstream signaling action. Furthermore, we conclude that the synergistic administration of a glucocorticoid and a beta-receptor agonist or (another cAMP-elevating agent) is an effective approach to suppress HMGB1-mediated inflammatory responses in vitro and in vivo.

Supporting Information
Table S1 A full list of the genes investigated in the realtime PCR array experiments.

(XLS)
Table S2 A full list of the primary data produced by the primary cell-based screens. (XLS) Figure 12. Inhibition of the HMGB-induced TNFa production by catecholamines and glucocorticoids in vivo. Balb/c male mice (Charles River Laboratories) were injected with 0.5 mg/kg HMGB1 in the presence of 60 min pretreatment of either vehicle, or 20 mg/kg prednisolone, 10 mg/kg salbutamol, the combination of prednisolone and salbutamol (doses as above), or the glucocorticoid receptor blocker mifepristone (30 mg/kg) or the b-receptor antagonist propranolol (10 mg/kg). At 8 hours after HMGB1 injection, animals were sacrificed and serum levels of TNFa were measured. # p,0.05 represents a significant increase in TNFa serum levels in response to HMGB1; *p,0.05 represents significant inhibition of HMGB1-induced TNFa production by the various pharmacological agents indicated. n = 7 animals per group. doi:10.1371/journal.pone.0065994.g012