Tumor Necrosis Factor Inhibits Glucocorticoid Receptor Function in Mice

As glucocorticoid resistance (GCR) and the concomitant burden pose a worldwide problem, there is an urgent need for a more effective glucocorticoid therapy, for which insights into the molecular mechanisms of GCR are essential. In this study, we addressed the hypothesis that TNFα, a strong pro-inflammatory mediator in numerous inflammatory diseases, compromises the protective function of the glucocorticoid receptor (GR) against TNFα-induced lethal inflammation. Indeed, protection of mice by dexamethasone against TNFα lethality was completely abolished when it was administered after TNFα stimulation, indicating compromised GR function upon TNFα challenge. TNFα-induced GCR was further demonstrated by impaired GR-dependent gene expression in the liver. Furthermore, TNFα down-regulates the levels of both GR mRNA and protein. However, this down-regulation seems to occur independently of GC production, as TNFα also resulted in down-regulation of GR levels in adrenalectomized mice. These findings suggest that the decreased amount of GR determines the GR response and outcome of TNFα-induced shock, as supported by our studies with GR heterozygous mice. We propose that by inducing GCR, TNFα inhibits a major brake on inflammation and thereby amplifies the pro-inflammatory response. Our findings might prove helpful in understanding GCR in inflammatory diseases in which TNFα is intimately involved.

Tumor necrosis factor alpha (TNF␣) is a strong pleiotropic pro-inflammatory cytokine. It acts by binding to the membrane-bound TNF␣ receptor, which trigger a signaling cascade that results in the activation of two main transcription factors (TF), 3 NFB and AP1 (1). Consequently, TNF␣ causes the release of a secondary wave of various pro-inflammatory cytokines, such as IL-6, TNF␣, and IL-1␤. The strong pro-inflammatory nature of TNF␣ is also manifested in the pathogenesis of many inflammatory diseases, such as rheumatoid arthritis, inflammatory bowel diseases, asthma, and psoriasis (2)(3)(4). Therefore, several of these pathologies are currently treated with TNF␣ inhibitors (5,6). However, these anti-TNF␣ molecules are very expensive, they cause many side effects, e.g. increased sensitivity to bacterial infections (7), and many patients are resistant to this treatment.
Another effective treatment of inflammatory diseases is administration of synthetic glucocorticoids (GCs), such as dexamethasone. Glucocorticoids have impressive anti-inflammatory effects and are frequently used to treat a wide variety of inflammatory and autoimmune diseases. This treatment is based on the knowledge of the role of natural GCs as an endogenous brake on inflammation. Inflammatory cytokines activate the hypothalamus-pituitary-adrenal (HPA) axis, which results in the release of GCs from the adrenal glands (8,9). GCs can diffuse freely across the plasma membrane and bind to their intracellular receptor, the glucocorticoid receptor (GR). GR is ubiquitously expressed and exerts a wide range of functions in the body, including maintenance of homeostasis and regulation of central nervous system functions, metabolism, and growth (10,11). Upon ligand binding, GR translocates to the nucleus and regulates the activity of specific target genes. GR can homodimerize and bind to glucocorticoid response elements (GREs) in the promoter region of GC-responsive genes (12). GR homodimers subsequently recruit transcriptional coactivators and chromatin remodeling factors, and initiate transcription in a process termed transactivation (reviewed in Ref. 12). Transactivation of GC-induced genes can also occur via interaction of liganded GR monomers with other transcription factors. On the other hand, GR monomers also interfere with important TFs, such as NFB (13) and AP1 (14,15). By tethering transcription factors, GR transrepresses transcription of inflammatory genes. Another way GR can transrepress genes is by binding as a homodimer to a negative GRE in promoter regions of GCresponsive genes.
However, despite excellent efficacy, GC therapy is often hampered by two major drawbacks. First, long-term treatment with GCs can be accompanied by severe metabolic side effects, including diabetes, osteoporosis, hypertension, and skin and muscle atrophy (16,17). Second, the occurrence of insensitivity to the therapeutic effects of GCs, a condition referred to as glucocorticoid resistance (GCR), limits the success of many GC-based therapies, and is therefore associated with substantial health care costs. The percentage of patients suffering from GCR depends on the disease, with incidence rates ranging from a few percent in asthma to about 30% in rheumatoid arthritis, ulcerative colitis, and Crohn disease, to almost 100% in atherosclerosis, cystic fibrosis, and COPD (chronic obstructive pulmonary disease). Several distinct molecular mechanisms contribute to the decrease in the anti-inflammatory effects of GCs (reviewed in Ref. 18). However, molecular mechanisms of GCR are incompletely understood and remain the focus of intense research. For example, a specific disease can have a variety of mechanisms, but at the same time different inflammatory diseases can have similar mechanisms. The latter suggests that common therapeutic approaches for a variety of inflammatory diseases might be developed.
Several studies have suggested a role for inflammatory cytokines, such as TNF␣, in inhibition of GR activity and thus in steroid insensitivity (19). The effects of cytokines and their signaling pathways on GR function are therefore an important area of research, especially with respect to treatment of inflammatory diseases. In this study, we investigated the effect of one of the strongest pro-inflammatory cytokines, TNF␣, on one of the most powerful anti-inflammatory molecules, GR. We injected TNF␣ in mice to model acute inflammation, and we found that GR and GCs are essential in the protection against TNF␣. In addition, using specific knock-out mice, we show that liver GR is indispensable for protection against TNF␣ lethality. We also show that high doses of TNF␣ cause GCR in the liver: (i) when dexamethasone was administered after TNF␣ injection, it did not protect against a lethal dose of TNF␣; (ii) a high dose of TNF␣ impaired the transactivation and transrepression actions of GR. We propose that TNF␣-induced GCR is at least partly due to a drastic decrease in hepatic GR levels. The importance of GR levels in determining the outcome of a TNF␣ challenge was confirmed by using mice underexpressing GR. Further elucidation of the mechanism of TNF␣-induced GCR could lead to the development of better therapeutic strategies for GC-insensitive patients suffering from TNF␣-mediated diseases.

EXPERIMENTAL PROCEDURES
Mice-Normal and adrenalectomized (Adx) female C57BL/6 mice were purchased from Janvier (Le Genest-St. Isle, France). GR flox/flox mice were generated by Prof. Dr. Günther Schütz and we crossed these GR flox/flox mice with AlfpCre (20) transgenic mice to obtain GR AlfpCre//flox/flox mice. GR heterozygous mice (GR ϩ/Ϫ ) were generated by Prof. Dr. Günther Schütz and backcrossed into a C57BL/6 background by Dr. Jan Tuckermann. All offspring were genotyped by PCR on genomic DNA isolated from tail biopsies. Mice were housed in a temperaturecontrolled, air-conditioned animal house with 14 and 10-h light/dark cycles and received food and water ad libitum. The drinking water of Adx mice was supplemented with 0.9% NaCl. All mice were used at the age of 8 -10 weeks and all experiments were approved by the institutional ethics committee for animal welfare of the Faculty of Sciences, Ghent University, Belgium.
Reagents-Recombinant mouse TNF␣ was produced in Escherichia coli and purified to homogeneity in our laborato-ries. TNF␣ had a specific activity of 1.2 ϫ 10 8 IU/mg and no detectable endotoxin contamination. RU38486 (Mifepristone), a GR antagonist, and water-soluble dexamethasone (a GR synthetic ligand) were purchased from Sigma. Luciferin (D-luciferin firefly, potassium salt) was purchased from Caliper Life Sciences (Hopkinton, MA). Antibodies against GR (M20) and actin were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and MP Biomedicals Europe (Illkirch, France), respectively. Secondary antibodies goat anti-rabbit IRDye800 and goat anti-mouse IRDye680 were purchased from Li-Cor, Westburg (Leusden, The Netherlands). The p(GRE) 2 -50-luc plasmid (21) was a kind gift from Dr. Karolien de Bosscher (Department of Physiology, Ghent University). The pCAGGS-Luc plasmid was a kind gift from Dr. Xavier Saelens (Department for Molecular Biomedical Research, Ghent University, VIB).
Cell Line-The mouse hepatoma cell line BWTG3 was cultured in DMEM supplemented with 10% FCS, L-glutamine, and penicillin and streptomycin. Cells were seeded in 6-well cultures at 10 6 cells/well and stimulated with 1000 units/ml of TNF and/or 10 Ϫ6 M dexamethasone. Three hours after treatment cells were lysed with TRIzol and RNA was subsequently isolated using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions.
Injections and Sampling-TNF␣ and dexamethasone were diluted in pyrogen-free PBS. RU486 was prepared as a solution of 20 mg/ml in dimethyl sulfoxide and each mouse received 1 mg of RU486 15 min before the other injections. All injections were given intraperitoneally except for the hydrodynamic intravenous injection of the DNA plasmid. Injection volumes were always adapted to the weight of the mice. Blood was withdrawn with a glass capillary from the retro-orbital plexus and allowed to clot overnight at 4°C. The clot was then removed and serum was collected after centrifugation at 20,000 ϫ g for 10 min and stored at Ϫ20°C. For sampling liver tissue, mice were killed by cervical dislocation and the liver was perfused with sterile ice-cold PBS. A piece of liver was stored in RNA later (Qiagen) for RNA preparation and the rest was snap-frozen in liquid nitrogen and stored at Ϫ80°C for Western blot analysis.
Hydrodynamic Tail Vein Injection and in Vivo Imaging-Mice were injected in the tail vein over 5 s with a plasmid solution (10 g/ml in sterile, endotoxin-free PBS) in a volume equivalent to 10% of the body weight, as described (22)(23)(24). In vivo bioluminescent imaging was performed using a cooled charged-coupled device bioluminescence camera (The Night-OWL LB 981 UltraSens Frontlit, Berthold Technologies). Briefly, mice were anesthetized with ketamine/xylazine, shaved on the ventral side, and injected intraperitoneally with 200 l of a 15 mg/ml of potassium salt luciferin solution. The mice were placed in a dark chamber and a gray scale image was recorded. Photon emission was integrated over a period of 60 s and recorded as pseudo color images (pixel binning 7 ϫ 7). For co-localization of the bioluminescent photon emission on the liver, gray scale and pseudo color images were combined by using WinLight software (Berthold Technologies). Localization and measurement of luminescence emitted from the liver was performed by using the overlay of the real image and the luminescence scan. Data were acquired as photons/cm 2 /s and results are presented as relative expression values normalized to a control plasmid, pCAGGS-Luc. This plasmid is regulated under the control of an albumin promoter, which is stably expressed. Using this plasmid, we did not observe an effect on expression due to TNF challenges, which indicates that the TNF-induced reduction of GRE expression is a real transcriptional effect.
IL-6 and Corticosterone (CS) Measurements-Mice were housed individually for at least 3 days before blood sampling to obtain basal, non-stressed hormone levels. Serum was immediately separated by centrifugation and frozen at Ϫ20°C. Concentrations of CS were determined using commercially available radioimmunoassay kits with [ 125 I]corticosterone (COAT-A-COUNT, Diagnostic Products Corp., Los Angeles, CA), according to the manufacturer's instructions. Serum IL-6 was determined with a 7TD1 bioassay (25).
Q-PCR-Liver samples were collected in RNA Later (Qiagen), and RNA was isolated with the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. RNA concentration was measured with the Nanodrop 1000 (Thermo Scientific) and 1 g of RNA was used to prepare cDNA with Super-Script II (Invitrogen). Q-PCR was performed with the SYBR Green mastermix (Applied Biosystems) and Lightcycler 480 (Roche Applied Science). The best performing housekeeping genes were determined by Genorm (26). Every sample was run in triplicate and results are presented as relative expression values normalized to the geometric mean of the two best-performing housekeeping genes (Gapdh and Tbp).
Statistics-Data are expressed as the mean Ϯ S.E. Statistical significance of differences between groups was assessed using a Student's t test with 95% confidence intervals and with one-way or two-way analysis of variance. Survival curves (Kaplan-Meyer plots) were compared by a log-rank test and final survival was assessed by the 2 test. Error bars in the figures represent the mean Ϯ S.E. *, **, and *** represent p Ͻ 0.05, p Ͻ 0.01, and p Ͻ 0.001, respectively.

RESULTS
Hepatic GR Is Protective in TNF␣-induced Lethal Inflammation-TNF␣ is a well known strong pro-inflammatory protein and in high doses it induces lethal inflammation in mice. We first injected C57BL/6 mice intraperitoneally with a sublethal dose of TNF␣ (5 g) or with a lethal dose (25 g). Control mice were injected with PBS. As measurement of IL-6 can be used to assess sensitivity to TNF␣ (27,28), we measured IL-6 protein and mRNA levels in circulation and liver 3 and 12 h after TNF␣ stimulation, respectively. Both TNF␣ doses initially induced high levels of hepatic IL-6 mRNA expression and IL-6 serum levels 3 h after TNF␣ injection, but only the high dose induced IL-6 levels that were sustained 12 h after TNF␣ injection (Fig.  1A). This effect was due to prolonged expression of IL-6 mRNA rather than reduced clearance of IL-6, because high IL-6 transcription levels were still visible upon challenge with a high dose of TNF␣.
TNF␣ induction of IL-6 occurs predominantly via transcription factor NF-B, which is known to be negatively regulated by GR (29). GCs have strong anti-inflammatory properties. Therefore, we examined whether proper GR function is essential for protection against TNF␣ lethality. Control, RU-treated, and Adx C57BL/6 mice were injected intraperitoneally with 2 g of TNF␣, and survival and body temperature were monitored. We used this low dose of TNF␣ because it was the minimum LD 100 dose for Adx and RU-treated mice (experimentally determined, data not shown). Irreversible blocking of GR with RU486 or eliminating GCs by Adx led to a marked sensitization of C57BL/6 mice to a low dose of TNF␣, as both RU-treated and Adx mice showed a large and significant drop in body temperature ( Fig. 1B) and increased mortality (Fig. 1C). To further document this sensitization, we measured IL-6 levels in circulation 0, 1, 3, 6, and 12 h after injection of 2 g of TNF␣ in all groups. These serum IL-6 levels were compared with the IL-6 levels of C57BL/6 mice challenged with 25 g of TNF␣. In all groups, TNF␣ challenge caused a large increase in serum IL-6 levels at early time points (Fig. 1D). As shown in Fig. 1A, the high TNF␣ dose in C57BL/6 mice resulted in sustained IL-6 levels in contrast to the IL-6 levels of C57BL/6 mice treated with 2 g of TNF␣. Remarkably, after blocking GR by RU486 treatment or performing Adx, the IL-6 levels reached sustained levels similar to those in C57BL/6 mice treated with a high dose of TNF␣. These data suggest that injection of mice with a high TNF␣ dose compromises GR function, which might account for the sustained high levels of the pro-inflammatory cytokine IL-6.
As lipid and carbohydrate metabolism are tightly controlled by GR, this receptor is strongly expressed in the liver (20,30). The liver is also an important target of TNF␣ (31). Therefore, we investigated the effect of TNF␣ injection in mice in which GR is specifically ablated in hepatocytes. For this purpose, mice expressing Cre recombinase under control of both the mouse albumin regulatory elements and the ␣-fetoprotein enhancers (AlfpCre transgene, GR AlfpCre ) were used and crossed to mice, which are homozygous for GRloxP/loxP (GR flox/flox ), which show normal expression of GR (20,32). This way GR AlfpCre//flox/flox mice were generated. Upon intraperitoneal injection with a sublethal (10 g) dose of TNF␣, GR AlfpCre//flox/flox mice showed a large and significant drop in body temperature (Fig. 1E) and a significantly higher mortality (Fig. 1F) compared with control GR flox/flox mice. We measured IL-6 levels in circulation 2 h after injection with 10 g of TNF␣. In response to TNF, hepatic GR-deficient mice show significantly increased levels of proinflammatory cytokine IL-6 ( Fig. 1G). These findings indicate TNF Inhibits GR in Mice JULY 29, 2011 • VOLUME 286 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 26557
that hepatic GR plays a crucial role in controlling TNF␣-induced inflammation and justify our focus on the liver. Cultured mouse hepatoma cell BWTG3 (Fig. 1H) as well as freshly isolated hepatocytes (not shown), when stimulated with TNF indeed express IL-6 at the mRNA level, and activity that is completely repressed by co-culture of dexamethasone indicates that hepatocytes, at least partly, contribute to the TNF-induced IL-6 expression observed in livers of mice.
TNF␣ Induces Glucocorticoid Insensitivity in Mice-Mice injected with dexamethasone 30 min before a TNF␣ challenge are protected against TNF␣ lethality (33), because of the antiinflammatory effects of GCs. However, pro-and anti-inflammatory cytokines, such as IL-1␣, IL-10, and TNF␣, are known to modulate GC action (34,35). More recently, Cho et al. (36) reported that the TNF␣-dependent pathway could be behind the reduced response to GCs in asthma models. These data show that pro-inflammatory cytokines might contribute to GC insensitivity, which is observed in many inflammatory diseases. Therefore, we wondered whether mice challenged with TNF␣ develop GCR. Mice were injected intraperitoneally with 500 g of dexamethasone 30 min before challenge with a high dose of TNF␣: they were completely protected against TNF␣ lethality ( Fig. 2A), which agrees with previous findings (33). The same dose of dexamethasone was also administered to mice 1, 4, or 8 h after a high dose of TNF␣. We found that dexamethasone cannot protect mice against TNF␣ lethality when it is administered 4 or 8 h after challenge, probably because the TNF␣induced inflammation was overwhelming at that time. Remarkably, dexamethasone was also not protective when given only 1 h after TNF␣ challenge, which suggests that GR signaling is compromised or insufficient shortly after TNF␣ challenge.
GR can directly transactivate gene transcription by binding to a GRE in target gene promoters, and it can transrepress gene transcription by tethering other transcription factors (29,37). To further examine whether TNF␣ blocks GR function, we investigated in TNF␣-treated mice the hepatic expression of a typical GR-transactivated gene (Tsc22d3, which encodes GILZ) and a typical GR-transrepressed gene, il1␤. C57BL/6 mice injected intraperitoneally with 25 g of TNF␣ were compared with C57BL/6 mice and to Adx and RU-treated mice injected intraperitoneally with 2 g of TNF␣. Liver samples were isolated 0, 1, 3, 6, and 12 h after TNF␣ challenge. Total RNA was isolated and expression of Tsc22d3 and il1␤ mRNA was measured by q-PCR (Fig. 2B).
Tsc22d3 mRNA levels declined rapidly after TNF␣ challenge in all groups, including RU-treated and Adx mice. This shows that down-regulation of Tsc22d3 expression levels after TNF␣ challenge occurs independently of GCs. Interestingly, in C57BL/6 mice treated with 2 g of TNF␣, Tsc22d3 mRNA levels were restored to initial levels 6 -12 h after challenge, but the expression levels remained low in C57BL/6 mice treated with 25 g of TNF␣. In addition, Tsc22d3 levels remained low following TNF␣ challenge in RU-treated and Adx mice. These findings suggest that high doses of TNF␣ block transactivation of GR in the liver.
GR function can also terminate expression of the pro-inflammatory gene il1␤. In C57BL/6 mice stimulated with 2 g of TNF␣, il1␤ transcription levels returned to baseline 6 h after TNF␣ challenge. In RU-treated and Adx mice challenged with 2 g of TNF␣, as well as in C57BL/6 mice treated with 25 g of TNF␣, the il1␤ mRNA levels also returned to baseline, but not until 12 h after TNF␣. Because il1␤ mRNA levels after a high TNF␣ dose in C57BL/6 mice are comparable with those seen after low-dose TNF␣ treatment of RU-treated and Adx mice, these data indicate that a high dose of TNF␣ compromises hepatic GR function. Taken together, these results indicate that a lethal dose of TNF␣ blocks both GR transactivation and GR transrepression functions in the liver.
We next evaluated the response of TNF␣-treated mice to exogenous GCs. Again, we studied the induction of Tsc22d3 expression as well as repression of il1␤ expression by GR. C57BL/6 mice were treated with dexamethasone, and hepatic Tsc22d3 mRNA expression was examined. As expected, 2 h of dexamethasone stimulation resulted in strong induction of Tsc22d3 expression (data not shown). Furthermore, C57BL/6 mice were injected intraperitoneally with PBS, a low or high TNF␣ dose. Three or 9 h after TNF␣ challenge, all mice were treated with PBS or dexamethasone and sacrificed 2 h later. Expression levels of Tsc22d3 and il1␤ were measured in livers (Fig. 2C). These data confirmed the results shown in Fig. 2B: relative Tsc22d3 mRNA levels declined upon TNF␣ administration and remained low following the high TNF␣ dose, in contrast to the higher levels observed 11 h after low TNF␣ dose (Fig. 2C, left graph). Tsc22d3 was also clearly induced by dexamethasone in the PBS groups. Five hours after TNF␣ challenge, Tsc22d3 was still induced upon dexamethasone stimulation in all groups but less than in the PBS group. Eleven hours after challenge with a high dose of TNF␣, dexamethasone treatment could no longer induce Tsc22d3. Likewise, relative il1␤ mRNA levels returned to baseline at a slower rate after a high dose of TNF␣ compared with a low dose of TNF␣. Il1␤ expression FIGURE 1. Hepatic GR protects against TNF␣-induced lethal inflammation. A, serum IL-6 levels (left) and liver IL-6 mRNA levels (3) were measured in C57BL/6 mice 3 and 12 h after intraperitoneal injection with PBS (black bars, n ϭ 5), 5 g of TNF␣ (dashed bars, n ϭ 5), or 25 g of TNF␣ (white bars, n ϭ 5). Asterisks refer to significant differences compared with the PBS-treated group, unless otherwise indicated. Body temperature (°C, *** at all time points) (B) and survival (C) of C57BL/6 mice (f, n ϭ 9), C57BL/6 mice were pre-treated with the GR antagonist RU486 (Ⅺ, n ϭ 7), and Adx mice (E, n ϭ 8) were treated after intraperitoneal injection of 2 g of TNF␣. Asterisks refer to significant differences compared with control C57BL/6 mice. D, IL-6 concentration in serum 0, 1, 3, 6, and 12 h after injection of 2 g of TNF␣ in control mice (f, n ϭ 5), RU486-treated mice (Ⅺ, n ϭ 5), and Adx mice (E, n ϭ 5), or injection of 25 g of TNF␣ in control mice (F, n ϭ 5). Significance of differences from group treated with 2 g of TNF␣ (f) was least (*) p Ͻ 0.05 at the 3-h time point and (***) p Ͻ 0.001 at later times. E and F, body temperature and survival, respectively, of the control Cre line negative mice (GR flox/flox ) (f, n ϭ 6) and hepatocyte-specific GR knock-out mice (GR AlfpCre//flox/flox ) (Ⅺ, n ϭ 6) after TNF␣ (10 g) challenge. Significance was calculated for the difference between control and GR AlfpCre//flox/flox mice. G, IL-6 production in circulation of control (GR flox/flox ) (black bars, n ϭ 11) and GR AlfpCre//flox/flox mice (white bars, n ϭ 11) in response to 10 g of TNF. Significance was calculated for the difference between PBS-and TNF-treated mice in both genotypes. All mice were injected according to body weight. H, significant IL-6 mRNA expression, measured by q-PCR, by mouse hepatoma cells BWTG3 3 h after stimulation with TNF (1000 IU/ml, 3 h) is completely blocked by co-culture of dexamethasone (Dex) (10 Ϫ6 M). Cells were cultured in quadruplicate.
was induced by TNF␣ challenge, and the levels were reduced when dexamethasone was administered 2 h after a 3-h TNF␣ treatment (Fig. 2C, right graph). Dexamethasone loses its activity 11 h after a high dose of TNF␣, consistent with the impaired induction of Tsc22d3. These data demonstrate that the ability of GR to induce or repress gene expression is completely lost 9 h after challenge with a lethal dose of TNF␣, again pointing to TNF␣-induced GR insensitivity. Mice also received 20 mg/kg of dexamethasone (Dex) at t ϭ Ϫ30 min (f), ϩ1 h (Ⅺ), ϩ4 h (E), or ϩ8 h (ϫ) relative to the TNF␣ challenge. Survival was monitored and compared with mice injected with 25 g of TNF␣ (F) but not with dexamethasone. Asterisk refers to significant difference in survival compared with TNF␣ control, unless otherwise indicated. B, relative Tsc22d3 (left) and Il1␤ (3) mRNA levels 0, 1, 3, 6, and 12 h after TNF␣ challenge in livers of C57BL/6 mice injected with 2 g of TNF␣ (f, n ϭ 5), 25 g of TNF␣ (F, n ϭ 5), RU, and 2 g of TNF␣ (Ⅺ, n ϭ 5), and in Adx mice injected with 2 g of TNF␣ (E, n ϭ 5). C, q-PCR analysis of levels of Tsc22d3 (left) and Il1␤ (3) in livers of C57BL/6 mice (n ϭ 3 for each group and time point). Mice were injected intraperitoneally with PBS (black bars), 5 g of TNF␣ (dashed bars), or 25 g of TNF␣ (white bars). Three or 9 h later, mice were injected intraperitoneally with PBS (Ϫ) or 20 mg/kg of dexamethasone (ϩ). Two hours after injection of PBS or dexamethasone (DEX) (5 h TNF␣ or 11 h TNF␣ in total), mice were euthanized and livers were harvested for q-PCR analysis. Asterisks refer to significant differences between dexamethasone and PBS groups; NS, no significant difference. D, C57BL/6 mice received a hydrodynamic injection with a p(GRE) 2 -50-luc plasmid solution in the tail vein, and the effect of TNF␣ on GRE-Luc activity was evaluated. A representative image of Luc activity measurements is shown. For the 3-h time point, mice were hydrodynamically injected with 10 g/10 g of plasmid followed 5 h later by intraperitoneal injection with PBS (black bars, n ϭ 12), 5 g of TNF␣ (dashed bars, n ϭ 5), or 25 g of TNF␣ (white bars, n ϭ 7). Luc activity was determined 3 h later. For the 9-h time point, mice were hydrodynamically injected with plasmid and at the same time received an intraperitoneal injection of PBS (black bars, n ϭ 12), 5 g of TNF␣ (dashed bars, n ϭ 6), or 25 g of TNF␣ (white bars, n ϭ 6). Nine hours later, Luc activity was determined. Mice injected hydrodynamically with empty plasmid as a negative control showed no Luc activity. PBS groups are set as 100%. Asterisks refer to significant differences compared with the PBS group. All mice were injected according to body weight.
Interpretation of gene expression data is often complicated by the involvement of multiple transcription factors. Therefore, to gain a clearer understanding of the previous results, we used a reporter plasmid containing typical GRE, because genes with a GRE in their promoter are induced by a dimerized GR and not by a GR monomer interacting with other transcription factors (29). We used a hydrodynamic transfection technique to deliver the GRE reporter plasmid p(GRE) 2 -50-luc (38) to the liver of mice. Several studies have reported that rapid injection of a large volume of plasmid DNA solution in the tail vein of mice results in high levels of transgene expression optimally 8 h later in the liver, which makes it relevant for our study (39,40). C57BL/6 mice were injected intravenously with a GRE-reporter plasmid to determine the GR response. Those mice were treated with TNF␣ for 3 and 9 h. Fig. 2D shows strong luciferase activity in livers of PBS-treated mice; this luciferase activity was inhibited by TNF␣ challenge. Inhibition of GR-induced GRE-Luc expression was most pronounced in mice that received the high TNF␣ dose, which is consistent with our gene expression data.
Taken together, our data demonstrate that GR function is compromised by a lethal dose of TNF␣. This TNF␣-induced GCR probably enhances the lethal effects of TNF␣ by enhancing inflammation.
GR Down-regulation Mediates GC Insensitivity Caused by TNF␣ Challenge-Different mechanisms have been proposed to explain GCR, including altered phosphorylation of GR, reduced GR-DNA binding, defective ligand binding, disrupted nuclear translocation, reduced GR levels, increased cofactor competition, and enhanced activity of pro-inflammatory transcription factors (41)(42)(43)(44)(45). To understand the fate of GR after TNF␣ challenge, we investigated GR mRNA and protein levels in mouse liver. C57BL/6 mice were injected intraperitoneally with PBS or a high dose of TNF␣, sacrificed 3 or 12 h later, and hepatic GR expression and protein levels were examined. Fig. 3, A and B, show a fast decrease in GR mRNA and protein levels, respectively, upon TNF␣ challenge. A high dose of TNF␣ reduced GR expression levels to 31.5% 3 h after TNF␣ and to 22.7% 12 h post-injection. GR protein levels were reduced to 60.8 and 28.0% 3 and 12 h after TNF␣ challenge, respectively. As dexamethasone administration 1 h after TNF␣ challenge cannot protect mice against TNF␣ lethality ( Fig. 2A), we performed an additional experiment to evaluate GR levels 1 h after TNF␣ challenge. Interestingly, both GR transcription and protein levels were already reduced upon TNF␣ after only 1 h (data not shown). Notably, GR levels were also reduced in other organs, such as spleen and lung (data not shown).
Several studies have shown the importance of GR concentration in determining the magnitude of the response to GCs and subsequently the response to inflammatory stimuli (46 -48). Therefore, it is conceivable that the large decrease in GR levels lies at the basis of the reduced GR response after TNF␣. To provide proof-of-principle that GR levels are associated with TNF␣ sensitivity, we used GR heterozygous mice (GR ϩ/Ϫ ), which lack one of the GR alleles. Injection of a low dose of TNF␣ in control and GR ϩ/Ϫ mice resulted in severe hypothermia (Fig.  3C, left graph) and a significantly higher mortality (Fig. 3C, right graph) in GR ϩ/Ϫ mice compared with control mice. To confirm that GR levels are indeed only half those in GR ϩ/Ϫ mice, hepatic GR expression and protein levels were examined. The results confirmed the presence of lower GR levels in GR ϩ/Ϫ mice (Fig.  3D). Next, we tested GR ϩ/Ϫ mice for their ability to induce GRE-Luc following injection of the plasmid in the tail vein. Fig.  3D shows that the GRE-Luc activity in GR ϩ/Ϫ mice is only half compared with control mice. These data indicate that GR concentration and activity are about half in GR ϩ/Ϫ mice. Together, these findings provide evidence that lower GR levels contribute to increased TNF-induced inflammation.
Hyperactivation of the HPA axis leads to a higher concentration of GCs and can result in down-regulation of GR, called homologous down-regulation (49,50). Pro-inflammatory stimuli, such as TNF␣, IL-1␤, and IL-6 can stimulate HPA axis activation, resulting in increased GC secretion (8,9,51). We hypothesized that TNF␣ could lead to activation of the HPA axis and subsequently to down-regulation of GR through increased levels of GCs, and so we evaluated endogenous GC production in mice. Mice were injected with PBS or a low or high TNF␣ dose and sacrificed 3 or 12 h after TNF␣ challenge to assess serum CS levels. Fig. 4A shows that CS was strongly induced by both TNF␣ doses and that CS production remained high after the high dose. To investigate whether this increase in GCs accounts for the observed decrease in GR levels after TNF␣, we injected C57BL/6 mice with either a low or high dose of TNF␣, and Adx mice with a low dose. Mice were sacrificed and livers were isolated 0, 1, 6, or 12 h after TNF␣ challenge and GR mRNA expression, and protein levels were determined (Fig.  4, B and C). The data show down-regulation of both GR expression and protein levels after both low and high TNF␣ treatment of normal mice. Surprisingly, GR levels are still reduced in Adx mice upon TNF. These results exclude the possibility that down-regulation of GR levels upon TNF␣ is due to higher GC levels.
Taken together, these results indicate that TNF␣ challenge leads to down-regulation of GR levels independently of GC production. This finding could explain the higher sensitivity to TNF␣ lethality.

DISCUSSION
TNF␣ is a strong pro-inflammatory cytokine that can induce an inflammatory response by activating several major TFs, such as NFB, AP1, and IRF1, which leads to transcriptional induction of various pro-inflammatory mediators. This is illustrated by the rapid increase in serum IL-6 levels after TNF␣ injection in mice. Injection of a lethal dose of TNF␣ induced a sustained elevation of IL-6 in circulation, whereas a sublethal dose induced a transient elevation. This sustained IL-6 elevation could be indicative of impaired GR function, as GR exerts a strong anti-inflammatory action, mainly by repression of NFBand AP1-dependent gene expression (29).
GCs can protect against TNF␣ toxicity: activation of GR with exogenous dexamethasone can protect mice against TNF␣ toxicity (33). Furthermore, endogenous GC were also found to be essential for protection against the lethal effects of TNF␣. First, Bertini et al. (52) showed that surgical removal of the adrenal glands, the major source of GCs, causes significant sensitization of mice to TNF␣ lethality. Second, the use of the synthetic GR TNF Inhibits GR in Mice JULY 29, 2011 • VOLUME 286 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 26561
antagonist, RU486, was reported to sensitize mice to TNF␣induced inflammation (53). Using adrenalectomy and RU486, we confirmed that GCs and GR are protective against TNF␣induced gene expression, as reflected in the increased IL-6 lev-els upon impairment of GR signaling. The crucial role of GR, particularly in the liver, in the protection against TNF␣ was further illustrated using hepatocyte-specific GR-deficient mice, which exhibited higher IL-6 levels and subsequent increased  (3) in WT mice (black bars, n ϭ 3 for GR protein and mRNA levels and n ϭ 7 for GRE-Luc activity) and GR ϩ/Ϫ mice (white bars, n ϭ 4). Luc activity was measured 8 h after the hydrodynamic injection of 10 g/10 g of p(GRE) 2 -50-luc plasmid solution in the tail vein. Asterisks refer to significant difference between WT and GR ϩ/Ϫ mice. All mice were injected according to body weight. sensitivity to TNF␣. These data indicate that TNF␣ toxicity is controlled by GCs and GR, and that hepatic GR plays a determining role.
The use of GCs is the most effective anti-inflammatory treatment available for many inflammatory and immune diseases, including asthma, rheumatoid arthritis, and inflammatory bowel disease. However, many of these patients show a poor or no response to GCs, known as glucorticoid resistance or GCR (18). Unfortunately, the etiology of GCR is largely unknown and likely involves multiple factors. Nevertheless, recent data suggest that inflammation itself might contribute to reduced GC sensitivity, for example, inflammatory cytokines, such as TNF␣, have been associated with GCR, which points to an intricate interplay between TNF␣ and GR signaling (19). Several studies have suggested a role for TNF␣ in GCR disease (reviewed in Ref. 19). For example, Cho et al. (36) demonstrated that the TNF␣ inhibitor infliximab partially restores the reduced GC response in chlamydia-induced asthma. A strong correlation between TNF␣ and GCR was also observed in patients with Crohn disease, where GCR of areas in the gut epithelium was strongly linked with MAPK and NFB activation (54). In addition, polymorphisms in the TNF␣ promoter strongly correlated with GCR in Crohn disease, ulcerative colitis, and idiopathic nephritic syndrome (55,56). Furthermore, TNF␣ can disrupt the action of other hormone receptors, including vitamin D receptor (57) and insulin receptor (58), and insulin-like growth factor signaling (59), which might have implications in the development of osteoporosis, diabetes, and muscle wasting, respectively. Thus, the effects of cytokines and their signaling pathways on hormone signaling, particularly GR signaling, is an important area of research in relationship to the treatment of inflammatory diseases. Here, we investigated the effects of TNF␣ on GR functions. To address the question of the extent to which GCR can be mediated by TNF␣ and how this occurs, we examined the in vivo functions of GR after administration of TNF␣ in mice. Here, we demonstrate that injection of TNF␣ in FIGURE 4. TNF␣-induced down-regulation of GR mRNA and protein is not dependent on increased corticosterone production. A, serum CS levels in C56BL/6 mice 3 and 9 h after intraperitoneal injection with PBS (black bars), 5 g of TNF␣ (dashed bars), or 25 g of TNF␣ (white bars). Asterisks refer to significant differences compared with the PBS-treated group unless otherwise indicated. B, q-PCR analysis of GR mRNA expression (n ϭ 4 for each group and time point) in the liver 0, 1, 6, and 12 h after TNF␣ challenge. C56BL/6 mice were injected intraperitoneally with 2 g of TNF␣ (f) or 25 g of TNF␣ (F) and Adx mice (E) were injected intraperitoneally with 2 g of TNF␣. C, detailed kinetic study (0, 1, 3, 6, and 12 h after TNF␣ challenge) showing total liver GR protein levels measured by Western blot analysis, in C56BL/6 mice treated with 2 g of TNF␣ (black bars), C56BL/6 mice treated with 25 g of TNF␣ (dashed bars), and Adx mice treated with 2 g of TNF␣ (white bars). For all groups, n ϭ 4. C1 and C2 refer to control samples included in all blots to exclude variability of loading between blots. Asterisks refer to significant differences compared with controls (0 h time point) unless otherwise indicated. All mice were injected according to body weight. mice causes a severe and acute form of GCR: the protective effect of dexamethasone against TNF␣ lethality was abolished when dexamethasone was given after TNF␣ stimulation. Further indications that TNF␣ induces GCR are derived from gene expression analysis. These results show that injection of a high dose of TNF␣ results in a gene expression profile resembling that induced by a low dose of TNF␣ in Adx mice and RU486treated mice. GR signaling was required to restore the expression of the GR-inducible gene Tsc22d3 and to repress the expression of inflammatory genes such as Il1␤. The proof of TNF␣-induced GCR came from experiments designed to evaluate hepatic gene expression in response to exogenous GCs in TNF␣-injected mice. When dexamethasone was administered after TNF␣ challenge, the GR-dependent gene Tsc22d3 was not induced and the Il1␤ gene was not repressed at later time points. Following hydrodynamic injection of a GRE-reporter plasmid in the tail vein of mice, we observed that a high dose of TNF␣ compromised GR-dependent transcription already early after TNF injection. Thus, we conclude that high doses of TNF␣ impair the transactivation actions of GR. Note that the transactivation of Tsc22d3 was not completely abolished upon a high dose of TNF and was to the same extent as the induction after low dose of TNF. This might be explained by the equal reduction of GR levels at the early time point. Furthermore, the induction of other GRE-dependent genes by dexamethasone, such as Tat and Pepck, was completely inhibited 3 h after TNF (data not shown). This discordance between GR-responsive genes may be caused by the dynamic and promoter-specific regulation of the GR transactivation complex.
It is not sure whether the transrepression actions of GR are also hampered by TNF␣ challenge, because GR-inducible genes, such as Tsc22d3 and Mkp1, also have strong anti-inflammatory activities (reviewed in Ref. 60). Therefore, defective induction of these GR-dependent genes might account for the reduced repression of inflammatory genes after TNF␣ stimulation. An important role for GR-responsive genes in controlling TNF-induced lethality was illustrated by the increased TNF␣ sensitivity of GR dim mice, which have a dimerization deficient GR and are thus impaired in GRE-driven gene transcription. 4 These GR-responsive genes can exert their anti-inflammatory actions by inhibiting MAPK, destabilizing target mRNAs, and inhibiting TF activity, for example, Tsc22d3 targets both NFB and AP1 (33,52), as reviewed in Ref. 61. The prominent role of Mkp1 in the control of excessive inflammation has been confirmed in Mkp1-deficient mice, which were more susceptible than normal mice to LPS-induced lethal inflammation (62)(63)(64). Furthermore, NFB inhibition is also mediated by the GR-induced transcription of IB␣, the inhibitory subunit of NFB (65,66). These data indicate that GR-transactivated genes are indispensable for terminating inflammatory responses, and that the defective repression of inflammatory genes upon lethal TNF␣ injection might be indirectly due to the constrained GRdependent induction of anti-inflammatory genes. We conclude that TNF␣ compromises GR actions, at least its transactivation potential. By inducing GCR and thus blocking the endogenous brake on inflammation mediated by GR, TNF␣ seems to amplify its own pro-inflammatory potency.
Numerous factors have been demonstrated to affect the responsiveness to GCs by regulating GR activity, such as GR cofactors, GR splice variants, GR isoforms, and GR post-translational modifications (reviewed in Ref. 67). By influencing these changes in GR structure, inflammatory kinases and TFs might inhibit GR activity. For example, it has been reported that activation of both JNK and p38 upon TNF␣ stimulation can suppress GR function in different cell lines (68,69). Although MAP kinases and TFs, as well as various other mechanisms, have been described to be involved in the unresponsiveness to GCs (18), detailed mechanistic insights are still lacking, especially of the in vivo TNF␣-induced GCR. Therefore, we wanted to elucidate the mechanism involved in the hepatic TNF␣-induced GCR, because a detailed understanding of this mechanism could have implications for the treatment of many inflammatory diseases in which GCR occurs. As GC responsiveness of target cells is consistent with receptor levels, we measured liver GR transcription and protein levels upon TNF␣ administration in mice. We found that GR transcriptional expression and protein levels were reduced in livers of TNF␣-injected mice starting from 1 h after TNF␣ challenge. This is consistent with the finding that administration of dexamethasone 1 h after TNF␣ injection cannot protect mice from TNF␣ lethality. These data indicate that GCR develops within 1 h of TNF␣ challenge, likely due to reduced GR levels. Previously, reduced expression of GR has already been linked to GCR in multiple sclerosis patients (70) and sepsis patients (71,72). Moreover, the importance of GR levels in the response to inflammatory stimuli such as TNF␣ has been demonstrated using different transgenic mice. For example, transgenic mice overexpressing GR are more resistant to LPS-and TNF␣-induced lethal inflammation (47). In addition, we demonstrated that GR heterozygous mice, which have approximately half of the GR mRNA and protein expression compared with control mice, are very sensitive to TNF␣-induced lethality. Furthermore, recent data from our lab showed that SPRET/Ei mice are strongly resistant to endotoxemia because of strongly increased GR levels (46). These reports indicate that TNF␣ enhances its pro-inflammatory effects by reducing the GR levels and thus by inhibiting the endogenous blockade on inflammation.
In the literature, data describing the effects of cytokines on the levels of GR expression are contradictory, as both cytokineinduced increases and decreases have been reported (35,44). In most cases, GR has been found to be up-regulated in vitro following treatment with inflammatory cytokines (73), which might be explained by the finding that the GR promoter regions contain several transcription factor sites, including sites for NFB and AP1 (67,74). Only one study showed a decrease in the number of GR molecules after stimulation of whole blood cell cultures with TNF␣ (34). In vivo, the situation is more complex because TNF␣ administration directly activates the HPA axis, which leads to the release of GCs. However, increased GC levels can result in reduced GR levels, a phenomenon known as homologous down-regulation (49,(75)(76)(77)(78)(79). This has been proposed as one of the mechanisms of GCR (80). Accordingly, we observed increased GC production upon TNF␣ challenge. Therefore, the observed effects on GR expression might have been due to cytokine-induced increases in GCs rather than to direct effects of TNF␣. For example, TNF␣-induced increase in GC secretion has been reported to be responsible for TNF␣mediated GR down-regulation in rat liver after immersion scald (81). However, we excluded a role for GCs in the reduced GR levels upon TNF␣ challenge, because TNF␣ also caused a decrease in GR mRNA and protein levels in Adx mice, which lack endogenous GC production. Importantly, it has been reported that the synthesis of bioactive GCs is not restricted to the adrenal glands, but extends to other extra-adrenal sources such as the skin, intestine, thymus, brain, and vascular endothelium (82,83). In addition, it has been postulated that TNF␣ promotes local steroidogenesis by directly inducing steroidogenic enzymes in intestinal epithelial cells (84,85). Therefore we investigated whether TNF␣ stimulates steroidogenesis in other organs besides the adrenal glands and whether this results in the observed TNF␣-mediated GR reduction in Adx mice. However, we found that local GC synthesis in intestine and skin is negligible and hence does not contribute to the systemic GC increase upon TNF␣ and the GR down-regulation in Adx mice (data not shown). Altogether, these findings suggest that TNF␣ can lead to a decrease in GR levels in a ligand-independent manner.
Besides homologous down-regulation, GR expression is controlled by a variety of mechanisms, including chromatin condensation, transcription initiation, epigenetic modifications, mRNA stability, and others. mRNA stability is in turn regulated by various mechanisms, such as RNA destabilizing agents and microRNAs (miRs). A well known mRNA-destabilizing protein is tristetraprolin, which binds to AU-rich elements (4) in the 3Ј-untranslated region (UTR) of target mRNAs and promotes deadenylation and subsequent degradation of the mRNA (86,87). It has been postulated that the presence of AREs in the 3Ј UTR of GR mRNA could mediate mRNA decay of GR by recruiting ARE-binding proteins (51, 88 -90). However, it has not been shown whether tristetraprolin binds to these GR AREs and leads to reduced GR levels upon TNF␣ stimulation. Interestingly, we observed fast up-regulation of tristetraprolin levels in liver of TNF␣-stimulated mice (data not shown). This might contribute to the TNF␣-mediated decrease in GR levels, although a direct role of tristetraprolin has to be confirmed. Another level of fine-tuning GR levels can be provided by miRs, by directing mRNA destabilization and translational repression (91). Recently, it was reported that miR18 and miR124a reduce GR expression, which leads to impairment of the transcriptional response of GR (92). Therefore, TNF␣ might induce the expression of GR down-regulating miRs. However, further investigation will be required to elucidate the mechanism of TNF␣-induced GR down-regulation. Next, strategies should be developed to prevent this decrease in GR expression, as it probably contributes to TNF␣-induced GCR.
On the other hand, GR protein levels also reduce rapidly after TNF stimulation. However, in the unliganded form, GR is normally degraded slowly over time, with a half-life of more than 30 h. As we observed a faster degradation of GR protein upon TNF, TNF probably triggers an additional post-translational mechanism resulting in enhanced GR degradation. For exam-ple, TNF induces the activity of the MAPK JNK, which phosphorylates GR on serine 226 (68). This in turn facilitates GR sumoylation, which affects GR protein stability. In addition, protein phosphorylation also enhances the recognition of target proteins by ubiquitin enzymes and ligases (93), which leads to degradation of GR through the ubiquitin-proteasome pathway (50). However, further studies will be required to elucidate the mechanisms by which TNF reduces GR amounts so quickly.
Based on our results, we propose that the decreased amount of GR at least partly determines the GR response and outcome of TNF␣-induced shock. However, it is likely that other mechanisms contribute to TNF␣-induced GCR. TNF␣ and its downstream signaling molecules have been shown to compromise the function of GR by a variety of mechanisms, including transrepression, post-translational modifications, and competition for cofactors by pro-inflammatory TFs. First, the repressive effect of GR on NFB and AP1 works both ways, in other words, these TFs can interfere with the transactivation function of GR (13-15, 94 -99). Therefore, high doses of TNF␣, which lead to prolonged activation of these inflammatory TFs, can suppress GC action by this mutual repression mechanism. Recently, it was suggested that the TF IRF1 plays a role in the suppression of GR signaling (100) by competing for the cofactor GRIP1 (101). Cellular accumulation of IRF1 may therefore constitute another mechanism for TNF␣-induced GC insensitivity. Preliminary data from our lab, using IRF1-deficient mice, however, exclude a role for IRF1 in the TNF␣-induced GCR (data not shown). The involvement of the coactivator GRIP1 in the previous mechanism indicates that inhibition of GR function can also occur at the level of nuclear cofactors. Inflammatory TFs and GR compete for the same coactivators, and so it is possible that the decreased GR levels result in recruitment of higher amounts of coactivators to TFs, in which would further decrease GR transactivation activity. In addition, the involvement of coactivators in TNF␣-induced inhibition of GR can occur independently of TFs. For example, the FLICE-associated huge protein, which is induced by TNF␣, can inhibit the physical interaction between GR and the p160 family of coactivators (102,103). Furthermore, TNF␣ has also been shown to decrease the expression levels of coactivators in several different cell lines (104 -106). However, as most of these studies relied on transient overexpression (61), the in vivo role of mutual GR and NFB/AP1 inhibition and cofactor competition remains to be addressed. To conclude, it is conceivable that the severe GCR observed after a lethal dose of TNF␣ is the result of several additive mechanisms, including a large decrease in GR levels. However, further investigation will be necessary to elucidate the exact mechanism of TNF␣-induced GCR. Also, as the inhibitory effect of TNF␣ on GR function is probably not direct and other downstream mediators might be involved, other pro-inflammatory assaults may trigger a similar response.
In conclusion, we report that TNF␣ is a potent and effective inhibitor of the anti-inflammatory properties of the GR, and thereby it enhances TNF␣-induced inflammation. Using a GRE-reporter plasmid in vivo and measuring the endogenous levels of transactivated genes, we demonstrate that GR transactivation is diminished in mice injected with a high dose of TNF␣. Therefore, inhibition of the production of anti-inflam-matory proteins after TNF␣ administration might be responsible for the prolonged expression of inflammatory genes. We further demonstrate that a strong decrease in GR levels after TNF␣ challenge likely contributes to the reduced GR response after TNF␣. Hence, the progression of lethal responses specifically in the liver in response to TNF comprise a TNF-induced impairment of GR anti-inflammatory function leading to increased levels of pro-inflammatory cytokines and subsequent lethality. We believe that more detailed insights in the molecular events contributing to TNF␣-induced GCR could lead to a more effective GC therapy of patients suffering from chronic inflammatory diseases in which GCR occurs. In several of these diseases, such as Crohn disease and rheumatoid arthritis, TNF␣ plays an important role, as illustrated by the success of anti-TNF␣ molecules (107,108). Therefore, TNF␣ might contribute to GCR in many of these patients, but further research is required to confirm this. If we can target the mechanism that contributes to TNF␣-induced GCR, we might succeed in a more broad application of GCs in patients suffering from TNF␣-mediated diseases. Thus, combination therapy might lead to a more effective, safer and cheaper inhibition of inflammation.