Identification of Two Forms of TNF Tolerance in Human Monocytes: Differential Inhibition of NF-κB/AP-1– and PP1-Associated Signaling Free

Tumor necrosis factor is a master cytokine involved in inflammation and immunity (1, 2). The rapid induction of cytokines such as TNF, chemokines, and other antimicrobial effector molecules is fundamental for orchestrating a coordinated immune response. TNF tolerance means that pre-exposure to TNF reduces sensitivity to subsequent stimulation with this cytokine (3). This form of refractoriness is involved in the modulation of TNF signaling and may represent a protective mechanism preventing the cell and organism from excessive and/or prolonged cytokine stimulation (4). In contrast, TNF tolerance may be a paradigm for processes resulting in immune paralysis and shutdown of the immune response (4). TNF tolerance is presumably involved in inflammation, for example, sepsis (4) or chronic inflammatory disease (5), but also in malignant processes (6). The balance between protection against excessive immune response and immune paralysis determines the patients’ fate, for example, in severe sepsis.

Animal research reveals that TNF-mediated effects, such as fever, gastrointestinal toxicity, liver injury, and anorexia, are affected by TNF tolerance (7–11). Moreover, several forms of cross-tolerance between TNF and LPS have been described (7, 12, 13). Because TNF tolerance appears more slowly than that of LPS, different mechanisms seem to be responsible for the two phenomena (14). Only a few results from cell culture studies characterizing the molecular basis of TNF tolerance exist to date (9, 15, 16). At the beginning of this study, it was unclear whether the phenomenon of TNF tolerance exists in primary monocytes as major producers of TNF coordinating innate and adaptive immunity (17). An 18-h preincubation of monocytic THP-1 cells with a high TNF dose, IL-1β or LPS induces tolerance against stimulation with the same agonist and several forms of cross-tolerance, accompanied by reduced degradation of NF-κB inhibitor protein IκBα and attenuated phosphorylation of JNK and ERK (18). In contrast, when THP-1 cells were preincubated for 72 h with a low TNF dose, no inhibition of IκBα proteolysis and NF-κB DNA-binding activity was found (19). Under this condition the transcription factor C/EBPβ interacts with NF-κB-p65 and inhibits its phosphorylation, thereby blocking the expression of NF-κB–dependent target genes, for example, IL-8 (3). A recent report demonstrates that TNF induces glycogen synthesis kinase-3 (GSK3)–mediated cross-tolerance to endotoxin in macrophages (20). The hitherto available studies show that the basic mechanism and functional consequences of TNF tolerance have not yet been satisfactorily elucidated.

The present study uses human monocytes as the gold standard to investigate the phenomenon of TNF tolerance on a genome-wide level. We demonstrate that TNF tolerance is a prominent phenomenon in primary monocytes of healthy individuals. We established two forms of TNF refractoriness, as follows: absolute tolerance, mediated by low and high TNF doses, is a very specific mechanism inhibiting a small, albeit powerful group of effector molecules, whereas induction tolerance, predominantly induced by high doses, represents a more general phenomenon. TNF tolerance differentially modulates NF-κB/AP-1–associated signaling. Low-dose TNF-induced tolerance is regulated by GSK3, whereas high-dose TNF-mediated tolerance is controlled by A20/GSK3 and protein phosphatase 1 (PP1)–dependent mechanisms. Absolute and induction TNF tolerance dose dependently affect the kinase/phosphatase balance and may represent different cellular strategies to protect against excessive TNF stimulation and resolve inflammation.

Blood samples from healthy donors were provided by the Institute of Transfusion Medicine, Hannover Medical School. Informed patient consent was obtained, and the experiments were approved by the Hannover Medical School ethics committee in accordance with the Declaration of Helsinki. Monocytes were isolated by Biocoll (Biochrom) density gradient centrifugation using LeucoSeptubes (Greiner Bio-One), followed by negative selection with magnetic antibiotin microbeads (Monocyte Isolation Kit II; Miltenyi Biotec), according to manufacturer’s instructions. Purified cells were cultured in 12- or 24-well plates (Thermo Scientific) at a density of 4 × 10⁶ or 2 × 10⁶ cells in a final volume of 2 ml or 1 ml endotoxin-free RPMI 1640 supplemented with 7% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin (Biochrom), 1.8% OPI media supplement (Sigma-Aldrich), and 0.8% Life Technologies MEM nonessential amino acids solution (Invitrogen). Following an adherence step for 1 h, cells were washed three times with supplemented medium. Endotoxin contamination was excluded using the Limulus amebocyte lysate assay (Lonza) (<10 pg endotoxin/ml).

Purity of isolated monocytes was assessed by dual cell labeling (30 min, 4°C, dark) using Alexa Fluor405-CD45 (Invitrogen) and allophycocyanin-CD14 (BD Biosciences) human rAbs (rhAb). Contaminations with other cell types were excluded using allophycocyanin-CD3, PE-CD19, FITC-CD56 (BD Biosciences), and FITC-CD15 (Invitrogen) rhAbs. For detection of TNFR1 and 2, cells were labeled with PE-TNFR1 and FITC-TNFR2 rhAbs (Santa Cruz). Cells were then fixed in 1% paraformaldehyde (Sigma-Aldrich) and stored at 4°C until detection using the LSR II flow cytometer and FACSDiVa software (BD Biosciences).

Human monocytic THP-1 and HeLa cells (DSMZ) were maintained in RPMI 1640 supplemented with 7.5% FCS, 100 U/ml penicillin, and 100 μg/ml streptomycin (Biochrom). THP-1 were cultured at 5 × 10⁵ cells/well in 12-well plates, and HeLa were plated at 2 × 10⁵ cells/well in 6-well plates (Sarstedt).

Unless otherwise indicated, cells were pretreated with 40 or 400 U TNF/ml (Sigma-Aldrich) for 48 h, followed by short exposure with 400 U TNF/ml for 2 h (mRNA expression), 4–6 h (protein secretion), 15 min (protein phosphorylation), or 30 min (EMSA).

Culture supernatants were harvested and subsequently analyzed using the Quantikine Human CXCL8/IL-8 Immunoassay (R&D Systems) and the ELx808 Absorbance Microplate Reader (BioTek Instruments), according to the manufacturers’ instructions. Protein levels were normalized to the respective DNA levels. Total DNA was isolated using the QIAamp DNA Mini Kit (Qiagen), and the concentration was assessed using a Nanodrop ND-1000 (PeqLab).

Cultured cells were lysed, and total RNA was isolated using the RNeasy Mini Kit or Micro Kit (Qiagen). To remove contaminating DNA, treatment with RNase-free DNase I (Qiagen) was performed. RNA concentrations were assessed using a Nanodrop ND-1000. Total RNA was reverse transcribed (SuperScript-II; Invitrogen), and quantitative PCR (qPCR) was performed using platinum SYBR-Green qPCR SuperMix UDG (Invitrogen) and a LightCycler 480 (Roche). The amplification protocol included enzymatic degradation of contaminating uracil-containing DNA (50°C, 2 min) and activation of the DNA polymerase (95°C, 2 min), followed by 45 amplification cycles (95°C, 10 s; 59°C, 15 s; 72°C, 20 s). The following primers were applied: IL8 (5′-TCCTGTTTCTGCAGCTCTGG-3′, 5′-GGCCACTCTCAATCACTCTC-3′), IL6 (5′-ACAGCCACTCACCTCTTCAG-3′, 5′-GTGCCTCTTTGCTGCTTTCAC-3′), IL1A (5′-TGACTGCCCAAGATGAAGAC-3′, 5′-CCAAGCACACCCAGTAGTC-3′), CCL20 (5′-GAAGGCTGTGACATCAATGC-3′, 5′-GGGCTATGTCCAATTCCATTC-3′), PTGS2 (5′-GGGCCAGCTTTCACCAAC-3′, 5′-ATCTTTGACTGTGGGAGGATAC-3′), and IκBα (5′-CGAGCAGATGGTCAAGGAGC-3′, 5′-CAGCCAAGTGGAGTGGAGTC-3′). IL-8 mRNA expression was alternatively assessed using universal probe library probe 72 (Roche) and IL-8–specific primers (5′-AGACAGCAGAGCACACAAGC-3′, 5′-CACAGTGAGATGGTTCCTTCC-3′), according to the standard protocol (21). Target gene expression levels were normalized to hGAPDH (5′-AGGTCGGAGTCAACGGAT-3′, 5′-TCCTGGAAGATGGTGATG-3′) or β₂-microglobulin (5′-TGTGCTCGCGCTACTCTCTCTT-3′, 5′-CGGATGGATGAAACCCAGACA-3′). Generation of external standard curves and normalization of cDNA amount were performed, as previously described (21). Relative expression values and fold changes were calculated, as published (22). Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Prism software).

The Whole Human Genome Oligo Microarray 4x44K (G4112F, design ID 014850; Agilent) was used in this study. The microarray contains 45,015 oligonucleotide probes covering ∼31,000 human transcripts. Total RNA was applied to prepare Cy3- or Cy5-labeled cRNA (Amino Allyl MessageAmp II Kit; Ambion), according to the manufacturer’s instructions. cRNA fragmentation, hybridization, and washing steps were carried out according to Agilent’s Two-Color Microarray-Based Gene Expression Analysis Protocol V5.7. A total of 300 ng of each labeled cRNA sample was used for cohybridization. Slides were scanned on the Agilent Micro Array Scanner G2505C (pixel resolution 5 μm, bit depth 20). Data extraction was performed with Feature Extraction Software V10.7.3.1 using the recommended default extraction protocol file (GE2_107_Sep09.xml). To correct for systematic bias at high-end fluorescence intensity measurements, the highest 2% of processed signals in the green channel were further normalized according to an intensity-dependent, nonlinear strategy, to optimize fitting to their respective counterparts in the red channel. Heat maps were created using the program Mayday (23).

The complete microarray data have been deposited in National Center for Biotechnology Information’s Gene Expression Omnibus and are accessible through GEO series accession number GSE45371 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE45371).

Whole-cell extracts were prepared by incubating cells in hot lysis buffer (75 mM Tris[hydroxymethyl]-aminomethan [Tris]-HCl, 0.5% NaDodSO₄ [SDS], 50 mM DTT, 1 mg/100 ml bromphenol blue [Applichem], 15% glycerol [Merck]; 5 min, 96°C). Alternatively, sonication (10 s) in extraction buffer (150 mM NaCl [Merck], 25 mM MgCl₂ [Roth], 50 mM Tris-HCl, 10% glycerol, 1% Nonidet P-40, and three tablets protease inhibitor [Roche] per 50 ml) was performed. Cytosolic extracts were obtained using cytosolic extraction buffer (10 mM HEPES [Sigma-Aldrich], 10 mM KCl [Roth], 300 mM Saccharose [Roth], 1.5 mM MgCl₂, 0.1% Nonidet P-40, and three tablets protease inhibitor per 50 ml; for 5 min on ice). Protein concentrations were determined by Bradford assay (Bio-Rad). Electrophoresis was performed with 12% Tris/glycin SDS-polyacrylamide gels (Biostep), and proteins were transferred to nitrocellulose membranes (0.45 μm; Bio-Rad) using the Western blot technique. After blocking with 5% skim milk (Merck) or 5% BSA (Roche) in TBS (140 mM NaCl, 20 mM Tris-HCl)/0.1% Tween 20 (Sigma-Aldrich), membranes were incubated (4°C, overnight) with primary Abs against the following proteins: A20 (D13H3), AKT (67E7), β-catenin (6B3), c-jun (60A8), GSK3β (27C10), IκBα (44D4), p38 (D13E1), p65 (22B4), p-AKT (Ser⁴⁷³) (D9E), p-c-jun (Ser⁶³)-II, p-GSK3β (Ser⁹) (D85E12), p-IκBα (Ser³²) (14D4), p-IKKα/β (Ser^176/180) (16A6), p-JNK1/2 (Thr¹⁸³/Tyr¹⁸⁵) (81E11), p-p38 (Thr¹⁸⁰/Tyr¹⁸²) (D3F9), p-p65 (Ser⁵³⁶), p-p65 (Ser⁴⁶⁸) (Cell Signaling), actin (¹¹C; Sigma-Aldrich), or PPP1R14C (Novus Biologicals). Next, membranes were incubated (1 h, room temperature) with HRP-conjugated secondary Ab (Alexis Biochemicals). Proteins were visualized, as previously described (21). Densitometric analysis was performed with TotalLab TL100 software.

Nuclear extracts were prepared and analyzed by EMSA, as previously described (19).

Addition of GSK3-inhibitor SB216763 (Sigma-Aldrich) and TNF pretreatment were performed simultaneously. Calyculin A (Cell Signaling Technology) and okadaic acid (Merck) were added 30 min before TNF short exposure. Cycloheximide (Merck) was added for the indicated time intervals following TNF preincubation for 48 h.

Small interfering RNA (siRNA) transfection of HeLa cells was performed using A20 siRNA, Alexa Fluor 488–coupled AllStars negative control siRNA, and FlexiTube GeneSolution or HiPerFect Transfection Reagent (Qiagen), according to manufacturer’s instructions. For plasmid transfection of THP-1 cells, PPP1R14C human cDNA open reading frame clone (OriGene) and X-treme gene HP transfection reagent (Qiagen) were used. Freshly grown THP-1 were seeded at 3 × 10⁵ cells/ml in Opti-MEM (Life Technologies) and incubated at 37°C for 2 h. Following preincubation of 6 μg plasmid and 21 μl transfection reagent for 30 min in 600 μl Opti-MEM, 6 ml cell suspension was transfected and subsequently incubated for 4 h. Afterward, 6 ml RMPI/20% FCS were added for 24 h. For tolerance experiments, transfected cells were seeded in 12-well plates (1 ml/well) and pretreated with TNF.

Human monocytes were isolated from blood of healthy donors in a multiple step procedure (purity >97%) (Fig. 1A). Isolated monocytes were incubated with increasing doses of TNF, and IL-8 mRNA/protein was determined. These experiments showed a continuous increase in IL-8 mRNA/protein (TNF, 0.4–1000 U/ml). The expression of IL-8 mRNA was used as an indicator for the majority of our experiments. In initial preincubation time course experiments, monocytes were preincubated in medium or with 400 U/ml TNF up to 48 h and then stimulated with 400 U/ml TNF for 2 h. Following TNF preincubation, a high IL-8 mRNA concentration was observed at 6 h, whereas an intermediate level was found at 24 h and a low level after 48 h (Fig. 1B). When we restimulated the TNF-preincubated cells, we found a complete inhibition of IL-8 mRNA expression at 48 h, representing an optimal condition to further study tolerance. To evaluate the minimal dose required to induce TNF tolerance, monocytes were preincubated with decreasing doses of TNF for 48 h and then short exposed to high TNF doses. This showed that TNF preincubation as low as 25 U/ml was sufficient to induce tolerance (Fig. 1C). Following preincubation with TNF for 48 h, no downregulation of TNFR1/R2 was observed using flow cytometry (data not shown), showing that TNF tolerance is not due to receptor downregulation. Our incubation conditions are summarized, as follows: monocytes were preincubated with 400 (high dose) or 40 U/ml (low dose) to induce tolerance (Fig. 1D).

Next, we investigated whether healthy individuals differ in their ability to develop TNF tolerance. Monocytes from 18 healthy donors were preincubated with 400 U/ml TNF for 48 h and then restimulated. This revealed that these individuals differ strongly in their sensitivity to TNF (Fig. 2A); IL-8 mRNA fold-induction values varied from 5-fold up to >110-fold. Initially, TNF tolerance was defined by the following formula: 400 + TNF (T) divided by medium (M) + T ≤ 0.25 (inhibition ≥ 75%). Fifteen of 18 individuals showed the picture of TNF tolerance, and a partial tolerance was observed in 3 individuals (Fig. 2B). Similar results were obtained when lower preincubation doses (6.25–200 U/ml, 20 donors) were used (Fig. 2C).

Microarrays were performed to characterize TNF tolerance on a genome-wide basis. Human monocytes from five different donors (Fig. 2, asterisks) were TNF preincubated for 48 h with 400 U/ml (n = 3) or 40 U/ml (n = 2). The cells were then short exposed to TNF (400 U/ml) for 2 h. To determine all the genes responding to TNF stimulation, we filtered the obtained data with regard to at least 2-fold induction following TNF short-term exposure of naive monocytes (M + T/M + M ≥ 2) identifying 360 genes (Supplemental Table I). First, TNF tolerance was defined by the following formula: 400 + T or 40 + T divided by M + T ≤ 0.25 (see Fig. 2B). Because under this condition the 400 + T or 40 + T value is <M + T, we defined this form of refractoriness as absolute tolerance. Following preincubation with 400 U/ml TNF, we detected 51 genes showing absolute tolerance (Fig. 3A, Supplemental Table II). Eleven of these genes belonged to a class of highly TNF-inducible genes (M + T/M + M ≥ 10) (Fig. 3B). Under 40 − T conditions, we identified 24 genes affected by absolute tolerance (Fig. 3C, Supplemental Table II). Applying less stringent criteria (40 + T/M + T ≤ 0.5), we identified 99 genes showing tolerance, including our indicator IL-8 (Supplemental Table III). Our data demonstrate that, under absolute tolerance, a selective number of TNF-inducible genes is inhibited, predominantly involved in inflammation, growth/differentiation, and chemotaxis/migration (Fig. 3D). Thirty-one (61%) of the 51 genes affected by high-dose absolute tolerance were regulated by NF-κB and/or AP-1 transcription factors, with similar results using low preincubation doses (Supplemental Table II) (24, 25). Microarrays were confirmed in primary monocytes by qPCR, for which several highly inducible genes were selected (IL-8, IL-6, IL-1A, CCL20, PTGS2) (Fig. 4A). mRNA data were confirmed on protein levels in primary monocytes using IL-8 as a readout (Fig. 4B). Absolute TNF tolerance was also demonstrated on both levels in monocytic THP-1 cells, another independent monocytic model, regardless of whether low (Fig. 4C) or high (data not shown) preincubation doses were used.

Second, TNF tolerance was defined by a different formula; the degree of induction was calculated within the same condition (naive, tolerant) for the 360 TNF-inducible genes: 400 + T/400 + M or 40 + T/40 + M ≤ 1.25. This means no induction following TNF short-term stimulation of tolerant cells leading to the name induction tolerance. When cells were preincubated with 400 U/ml TNF, we identified 227 tolerant genes (63%), also visualized by the heat map (white or pale red bands) (Fig. 5A, Supplemental Table IV), and 10 genes when the lower dose was used (data not shown). These 227 genes showing induction tolerance could be further subdivided, as follows: 72 genes displayed lower expression levels in 400 + M than in M + T cells (400 + M/M + T < 0.5) (Fig. 5B), and 155 genes showed comparable/higher 400 + M levels (400 + M/M + T ≥ 0.5) (Fig. 5C). These genes, again mostly controlled by NF-κB/AP-1, can be attributed to six prominent groups involved in signaling/transcription, inflammation, metabolism, growth/differentiation, chemotaxis/migration, and transport (Fig. 5D).

Next, we monitored how TNF tolerance affects the NF-κB system (24). As expected, TNF (40 or 400 U/ml) led to initial IκBα proteolysis at 0.25 h and recovery within 1 h (Fig. 6A). Interestingly, following exposure up to 48 h, we observed a continuous decrease in IκBα predominantly in cells pretreated with the high TNF dose. When low-dose preincubated cells were re-exposed to TNF up to 120 min, we observed a regular, somewhat attenuated IκBα proteolysis/recovery (Fig. 6B, 6C). Remarkably, in high-dose pretreated cells, no TNF-inducible IκBα proteolysis was observed. In the following, cells were treated with a proteasome inhibitor briefly before short exposure because it is difficult to determine IκBα phosphorylation under the condition of stimulus-induced proteolysis. A modestly elevated level of IκBα phosphorylation was detected in unstimulated tolerant cells especially preincubated with high doses (Fig. 6D). As expected, we observed a dramatic increase in IκBα phosphorylation when naive cells were short-term TNF exposed, peaking at 5 min. A similar increase in IκBα phosphorylation was observed in 40 + T cells, albeit to a less degree and delayed. In contrast, in high-dose pretreated cells, no further TNF-induced phosphorylation of IκBα was observed. This was supported by gel shifts demonstrating a TNF-stimulated increase in NF-κB activity in 40 + T but not 400 + T cells (Fig. 6E). This demonstrates that TNF tolerance affects the NF-κB pathway in 400 + T cells by inhibiting restimulation-induced IκBα phosphorylation/proteolysis. In addition, mRNA expression and protein stability of IκBα were examined under tolerance-inducing conditions. Following preincubation with both the low and the high TNF dose, modestly increased IκBα mRNA levels were measured (Fig. 6F). As mentioned above, more prominent in cells preincubated with the high dose, reduced basal levels of IκBα protein were detected that were accompanied by increased IκBα protein degradation compared with naive cells (Fig. 6G). Taken together, our results suggest an increased IκBα turnover in tolerant cells.

The nuclear levels of NF-κB proteins were also assessed in tolerant cells. In response to TNF restimulation, a marked increase in nuclear p65 in nontolerant and low-dose preincubated cells, but not in high-dose preincubated cells, was observed (Fig. 7A). Preincubation with the low dose induced a modest increase in nuclear p50 (see TNF SE, 0 h; Fig. 7A), whereas a marked increase was observed following preincubation with the high dose, consistent with earlier results (18). When cells were restimulated, we found a marked increase in nuclear p50 in naive cells and a further increase in low-dose pretreated cells. No additional elevation of p50 in the nucleus was detected when high-dose preincubated cells were restimulated. This suggests a regular translocation of p65/p50 heterodimers into the nucleus following restimulation of naive and low-dose preincubated cells as well as a contribution of elevated p50 levels to the inhibition of NF-κB–dependent gene expression in tolerant cells (26). Short-term exposure of naive cells to TNF led to a marked p65 phosphorylation on Ser⁵³⁶ (Fig. 7B). In unstimulated 400 + M cells, a modest increase in p65 (Ser⁵³⁶) phosphorylation was found. Remarkably, TNF short-term–induced p65 (Ser⁵³⁶) phosphorylation was inhibited in all tolerant cells, regardless of whether low or high preincubation doses were applied. Furthermore, in unstimulated tolerant cells, an increased phosphorylation of p65 (Ser⁴⁶⁸), negatively regulating basal p65 activity (27), was found (Fig. 7B). We also investigated whether the MAPK/AP-1 pathway (25) is affected in tolerant cells. As expected, TNF short-term exposure lead to phosphorylation of p38, JNK, and c-jun (Fig. 7C, data not shown). Remarkably, phosphorylation of these proteins was strongly inhibited in tolerant cells under both low- and high-dose preincubation conditions. These data suggest that the NF-κB system is inhibited in both 40 + T and 400 + T tolerant cells by inhibition of p65 phosphorylation. The increase in p65 (Ser⁴⁶⁸) phosphorylation may contribute to tolerance-mediated inhibition of gene expression. In addition, the MAPK signaling pathway is inhibited in TNF-tolerant cells, presumably leading to inhibition of AP-1–dependent gene expression.

It has been shown that GSK3/A20 mediates TNF cross-tolerance to endotoxin (20). Most important, when cells were incubated with the GSK3 inhibitor SB216763, the effect of low- dose TNF-induced tolerance was completely reversed in a dose-dependent manner (Fig. 8A, left). Preincubation with SB216763 increased the level of β-catenin, an established substrate for GSK3 (28). Under high-dose TNF preincubation, the expression of IL-8 was modestly elevated in inhibitor-treated cells, regardless of whether they were restimulated or not (Fig. 8A, right). Remarkably, in low-dose pre-exposed cells, the inhibition of p65 (Ser⁵³⁶) phosphorylation was completely reversed in the presence of SB216763 (Fig. 8B). Under high-dose conditions, only a modest increase in p65 phosphorylation was detected in inhibitor-treated cells. In addition, we detected an upregulation of c-jun expression in the presence of SB216763 only in tolerant cells (Fig. 8C). Inhibition of GSK3 phosphorylation leads to the activation of this kinase (28, 29). Our experiments showed an inhibition of GSK3-β phosphorylation on Ser⁹ in TNF-tolerant cells (Fig. 8D). Our arrays also demonstrated a marked upregulation of A20 (30) in TNF-tolerant cells (400 + M: 8.9 ± 3.2-fold, n = 3, GEO series accession number GSE45371), and we showed that A20 levels were significantly increased following TNF exposure (Fig. 8E). Most important, using an A20 siRNA approach, we demonstrated a strong upregulation of IL-8 only in those cells in which TNF tolerance was induced by high-dose TNF (Fig. 8F). Finally, high TNF dose–induced inhibition of inhibitor of κβ kinase (IKK) phosphorylation was almost completely reversed by A20 siRNA (Fig. 8G). Our data suggest that GSK3 plays a key role in mediating low-dose–dependent TNF tolerance and also participates in high-dose–induced tolerance. Our results also demonstrate that A20 plays a major role in TNF tolerance forms that are induced by the high dose but is not involved in low-dose–dependent TNF tolerance.

The inhibition of several kinase cascades in TNF-tolerant cells prompted us to look at the phosphatase system as counteracting principle (31, 32). Microarrays identified an increased expression of the PP1 phosphatase catalytic subunit PPP1CB and a downregulation of the PP1 phosphatase inhibitory subunit PPP1R14C (33) following TNF preincubation (Fig. 9A, 9B). Remarkably, preincubation with the PP1 phosphatase inhibitor calyculin A led to a strong upregulation of IL-8 expression predominantly in those tolerant cells that were pretreated with the high TNF dose (Fig. 9C). No effect on IL-8 was observed using the inhibitor okadaic acid in a dose selectively reducing PP2A activity (34). Furthermore, overexpression of the inhibitory subunit PPP1R14C increased the expression of IL-8 solely in tolerant cells, again more dramatic when the high preincubation dose was used (Fig. 9D). Finally, calyculin A treatment elevated the levels of phosphorylated IKK, IκBα, p65, and JNK (Fig. 9E), demonstrating a connection between NF-κB/MAPK and PP1 phosphatases. Our data show that, under TNF tolerance induced by the high dose, the PP1 phosphatase system is upregulated, which may be at least partially caused by a reduced expression of the inhibitory PPP1R14C subunit and finally contributes to the inhibition of NF-κB/AP-1–associated kinases.

Despite the fact that TNF is considered to be one of the master cytokines involved in regulation of inflammation and immune response (1, 24) the phenomenon of TNF tolerance is poorly understood to date. When this study was initiated, it was unclear whether this form of tolerance existed in monocytes at all. To our knowledge, this study represents the first genome-wide analysis of TNF tolerance using primary human monocytes. TNF tolerance was observed in monocytes preincubated for 48 h with a TNF dose as low as 25 U/ml. To systematically characterize TNF tolerance, we selected high and low TNF doses (400 and 40 U/ml) for preincubation to induce tolerance according to the literature (35) and our own dose-response experiments. Similar concentrations may be locally achieved in vivo for a certain time period in the presence of monocytic cells, for example, in an acute or chronic inflammatory environment (36–40). Monocytes of 83% of the healthy donors showed complete tolerance (400 U/ml preincubation), whereas 17% exhibited a partial tolerance (with similar results using lower preincubation doses), which demonstrates that TNF tolerance is a prominent phenomenon in this cell type.

Using microarrays, we detected two forms of TNF refractoriness, that is, absolute and induction tolerance. Absolute tolerance was defined by the equation 400 + T or 40 + T divided by M + T ≤ 0.25 (abbreviations: Fig. 1D) with the 400 + T or 40 + T values < M + T values (≥75% inhibition). Under high-dose preincubation, we detected 51 of 360 TNF-inducible genes affected by absolute tolerance, and, when we applied the 40 + T condition, we found 24 inhibited genes. The arrays were confirmed by RT-PCR and on protein levels in monocytes and monocytic THP-1 cells. This proves that the inhibition of mRNA expression observed under tolerance results in a functional protein output, and the experiments with pure cell lines demonstrate that TNF tolerance is not dependent on the potential presence of small amounts of contaminating other blood cells. Next, induction tolerance was defined by calculating the degree of induction within the same condition (naive, tolerant) for the identified TNF-inducible genes (400 + T/400 + M or 40 + T/40 + M ≤ 1.25), which means basically no induction following TNF short exposure. More than half (63%) of the TNF-inducible genes displayed the induction tolerance almost exclusively when the high preincubation dose was used.

TNF tolerance affected six major groups of genes predominantly regulated by NF-κB/AP-1 transcription factors (24, 25). Under absolute tolerance, an inhibition of genes playing a role in inflammation, growth/differentiation, and chemotaxis/migration was found, whereas induction tolerance inhibits genes involved in signaling/transcription, inflammation, metabolism, growth/differentiation, chemotaxis/migration, and transport. One interesting candidate not represented by these groups is tissue factor, a key procoagulatory molecule (41), which was also completely inhibited under absolute tolerance. Taken together, absolute tolerance appears to represent a selective negative-regulatory mechanism that dose dependently affects a relatively small, albeit extremely powerful group of molecules basically involved in each major monocytic function (17). This is different from LPS tolerance characterized by a more global nature of unresponsiveness of macrophages to restimulation (42, 43). TNF tolerance is not induced by direct contact with bacterial products such as LPS tolerance but is dependent on a specific cytokine. Therefore, TNF tolerance can be less manipulated by excessive production of certain bacterial products and may thus represent a more precise second protection line. A relatively large group of TNF-inducible genes is inhibited by induction tolerance, suggesting that this form of tolerance is a more general phenomenon to organize cellular functions in the presence of high TNF. It is striking to speculate that this form of tolerance represents a protective principle, leaving the basal mRNA supply open, but does not allow a further increase.

There are little and varying data as to how TNF tolerance regulates the NF-κB system. Under low dose–mediated tolerance, we found in THP-1 monocytic cells that TNF short-term–induced IκBα proteolysis and NF-κB activation (EMSA) were not affected (19). In contrast, another paper identified reduced IκBα degradation and NF-κB–binding activity in THP-1 cells preincubated for 18 h with high TNF doses (18). In addition, it was shown that TNF short-term pretreatment of fibroblasts negatively affects IκBα resynthesis (44). Our present results resolve these discrepancies, as follows. First, preincubation of THP-1 cells with TNF for 48 h led to reduced basal IκBα levels, which were more prominent when the high dose was used. This is in contrast to LPS tolerance characterized by elevated IκBα levels (43). In addition, we measured a modestly increased mRNA expression of IκBα in tolerant cells accompanied by increased protein degradation, suggesting an increased IκBα turnover. Second, under low-dose preincubation-dependent tolerance, restimulation with TNF induced a regular modestly impaired IκBα phosphorylation/proteolysis and translocation of NF-κB subunits p65 and p50 in monocytic cells. Third, in high-dose preincubated cells, no TNF-induced IκBα phosphorylation/proteolysis and increase in NF-κB activity could be detected.

The TNF-induced phosphorylation of p65 (Ser⁵³⁶) was inhibited in tolerant cells in our experiments, regardless of whether low or high preincubation doses were used. We have shown that low-dose TNF-induced tolerance is associated with an inhibition of p65 (Ser⁵³⁶) phosphorylation (3), leading to inhibited NF-κB–dependent transcription. Consistent with results reported earlier (18), we detected increased levels of p50 in the nucleus of tolerant cells that were more pronounced in high-dose preincubated cells and accompanied by a change of the cytosolic p50/p65 ratio toward p50 (data not shown). As a consequence, the p50 increase may lead to the formation of transcriptionally inactive p50/p50 homodimers (24), thereby contributing to the inhibition of NF-κB–dependent genes in the same way as found for LPS tolerance (26, 45). Taken together, we demonstrate that when cells were preincubated with high TNF doses, NF-κB signaling was inhibited at two regulatory levels, that is, IκBα phosphorylation/proteolysis and p65 phosphorylation, whereas in low-dose TNF-treated tolerant cells only an inhibition of p65 phosphorylation was observed. This is a likely explanation as to why high-dose TNF preincubation is more effective in inhibiting NF-κB–dependent genes. The increase in transcriptionally inactive p50 presumably contributes to the inhibition of the NF-κB system potentially to a greater extent in cells preincubated with the high dose. Our data also suggest that there are two different groups of NF-κB–regulated genes showing a differing sensitivity to the degree of inhibition.

Database analysis revealed that a major group of genes affected by absolute/induction tolerance is regulated by AP-1, which is activated by MAPK pathways (46). Our data showed that the TNF-induced phosphorylation of p38 and JNK was completely inhibited in tolerant cells regardless of whether high or low doses were used for preincubation. An impaired activation of JNK under TNF tolerance has been found earlier (18). Our data also demonstrated that the phosphorylation of c-jun was significantly inhibited in tolerant cells. It has been shown that JNK and p38 cooperate to regulate inflammatory gene expression, and a variety of genes is cooperatively controlled by AP-1 and NF-κB (24, 25, 46).

Innate and adapted immunity can be regulated by GSK3 (28, 29), and TNF-induced LPS cross-tolerance is reversed by GSK3 inhibitory strategies (20). Remarkably, our data showed that inhibition of GSK3 led to a complete reversal of low-dose TNF-induced tolerance. In addition, low-dose TNF-mediated inhibition of p65 (Ser⁵³⁶) phosphorylation could be reversed by this approach. This suggests that a GSK3-dependent mechanism plays a key role in mediating low-dose–induced TNF refractoriness. GSK3 has been shown to be involved in the inhibition of TNF-induced NF-κB–dependent gene expression on a transcriptional level (47), and it has been shown that GSK3 regulates chromatin accessibility (48), which may negatively affect p65 phosphorylation (49, 50). Previously, we have shown that C/EPBβ is involved in the inhibition of p65 (Ser⁵³⁶) phosphorylation mediated by low-dose TNF-dependent tolerance (3). Another group has found that GSK3-β–dependent phosphorylation of C/EBPβ is associated with inhibitory effects on gene expression (51), and our experiments demonstrate an inhibition of CEBPβ phosphorylation using a GSK3 inhibitor (data not shown). In unstimulated tolerant cells, we also observed an upregulation of p65 phosphorylation on Ser⁴⁶⁸, a site phosphorylated by GSK3 and involved in negative regulation of basal p65 transcriptional activity (27). GSK3 is inactivated by p38 and AKT-dependent phosphorylation (52, 53). Under TNF tolerance we observed an inhibition of p38, JNK, and AKT (data not shown) phosphorylation accompanied by reduced GSK3-β phosphorylation presumably associated with activation of this kinase. It has also been shown that GSK3 inhibits the activation of JNK and p38 (54) and is also involved in the inhibition of c-jun–mediated transcription (28). For example, it has been suggested that AP-1, CREB, and NF-κB form an integrated transcriptional network largely responsible for maintaining repression of genes downstream of GSK3 signaling (55). It should also be mentioned that GSK3 phosphorylates c-jun targeting this protein for degradation (56). In our experiments, we observed an upregulation of c-jun in the presence of a GSK3 inhibitor solely in tolerant cells, suggesting an increased c-jun turnover under tolerance.

GSK3 inhibition only partially affected high-dose–induced TNF tolerance and inhibition of p65 phosphorylation, suggesting that additional mechanism contributes to this form of refractoriness. For example, the upregulation of PP1 phosphatase systems under high-dose–dependent tolerance conditions also found in this study may prevent a complete reversal of TNF tolerance by GSK inhibitory strategies. GSK3 is also involved in the upregulation of A20 expression (30). Remarkably, when we used a siRNA approach against A20, a significant upregulation of gene expression was observed only in high-dose but not in low-dose TNF-induced tolerant cells. We also detected a dose-dependent upregulation of A20 in tolerant cells. In this study, it should also be mentioned that in those cells, in which tolerance was induced by the high dose, our array data showed an upregulation of TNIP1-3 (data not shown) involved in the regulation of A20 (30). These data suggest that A20 plays a major role in mediating TNF tolerance induced by high TNF doses and also indicate that GSK3-β–independent mechanisms contribute to A20 upregulation. A20 has been shown to negatively affect NF-κB signaling upstream of the IKK complex (57). Our data show an almost complete reversal of the high-dose TNF-induced inhibition of IKK phosphorylation in the presence of A20 siRNA. Therefore, the increased level of A20 presumably contributes to the inhibition of IκBα phosphorylation/proteolysis observed only in high-dose preincubated tolerant cells.

Because various kinases are differentially modulated under TNF tolerance in monocytic cells, we examined whether opposing/balancing phosphatase systems are also affected (32). Interestingly, the presented array data show an increased expression of the catalytic PP1 phosphatase subunit PPP1CB and a downregulation of the inhibitory subunit PPP1R14C in TNF-preincubated cells. Remarkably, treatment with calyculin A, a strong inhibitor of PP1 phosphatases (34), led to markedly elevated IL-8 expression in cells affected by high-dose–dependent tolerance, whereas the PP2A inhibitor okadaic acid had no effect. In addition, overexpression of PPP1R14C increased IL-8 expression predominantly under high-dose–induced tolerance. PPP1R14C, also named KEPI, has been proposed to bind to the catalytic side as a nondephosphorylable pseudosubstrate (33). We also showed that preincubation with calyculin significantly upregulated the phosphorylation of IKKβ, IκBα, p65, and JNK, which demonstrated a connection between the PP1 phosphatase and the NF-κB/MAPK systems (24, 32). These data suggest that the PP1 phosphatase system is upregulated following long-term TNF exposure, which may be at least partly caused by a reduced PPP1R14C expression and contributes to the differential inhibition of kinase systems particularly under high-dose TNF-induced tolerance. It has also been shown that PP1 inactivates specific transcription factors and promotes the recycling of splicing factors (31). Looking at the overall picture of inflammation, the upregulation of phosphatase systems may be an important means to reset pathways to basal and economical states of activity and terminate the maintenance of high activity conditions functioning as kind of a reset button to resolve inflammation (31).

Taken together, to our knowledge, we present the first genome-wide analysis of TNF tolerance in human monocytes that identifies two forms of TNF refractoriness representing prominent regulatory mechanisms, as follows: absolute tolerance, mediated by low and high TNF doses, appears to specifically inhibit a small group of monocytic effectors, whereas induction tolerance, induced by the high dose, represents a more general phenomenon. TNF tolerance dose dependently modulates NF-κB/AP-1–associated signaling. Low-dose–induced TNF tolerance is mediated by GSK3, whereas high-dose–induced tolerance is regulated by A20/GSK3- and PP1 phosphatase–dependent mechanisms. Absolute and induction TNF tolerance affects the kinase/phosphatase balance representing different cellular means to protect against excessive TNF stimulation and orchestrate/resolve inflammation. A pathophysiological scenario may arise from a situation in which one has not enough or vice versa too much tolerance, resulting in either excessive inflammation or immune paralysis.

We are very thankful to Prof. Dr. H. W. L. Ziegler-Heitbrock, Dr. C. Cappello, Dr. F. Bollig, J. Mages, and R. Lang for numerous critical discussions throughout the development of this study. We thank I. Rudnick for technical assistance. In addition, we are very thankful to Cornélia La Fougère-Brand for typing and proofreading the manuscript as well as for critical discussions.

The authors have no financial conflicts of interest.