TNF activates calcium–nuclear factor of activated T cells (NFAT)c1 signaling pathways in human macrophages

Abstract

Acute activation of cells by tumor necrosis factor (TNF) has been well characterized, but little is known about later phases of TNF responses that are relevant for cells exposed to TNF for several days during inflammation. We found that prolonged exposure of human macrophages to TNF resulted in a wave of delayed but sustained activation of c-Jun and nuclear factor κB (NF-κB) proteins and of calcium oscillations that became apparent 1–3 d after TNF stimulation. These signaling events culminated in the induction and activation of the calcium-dependent transcription factor, nuclear factor of activated T cells (NFAT)c1, which mediated a gene expression program leading to cell fusion and osteoclast differentiation. TNF-induced NFATc1 activity primed macrophages for enhanced osteoclastogenesis in response to RANKL. High NFATc1 expression was apparent in synovial macrophages in a subset of patients with TNF-driven inflammatory arthritis. Thus, long-term exposure to TNF activates calcium-dependent signaling and an NFATc1-mediated gene activation program important for cell fusion and osteoclastogenesis. These findings identify a signaling pathway activated by TNF that is important for myeloid cell differentiation and suggest a role for TNF-induced calcium and NFAT signaling in chronic inflammation and associated bone resorption.

Tumor necrosis factor (TNF) is a pleiotropic cytokine that regulates a broad range of biological activities, including inflammation, innate and adaptive immune responses, and tissue development (1, 2). TNF is a major driver of acute inflammation and plays a key role in chronic inflammation in autoimmune diseases such as rheumatoid arthritis and Crohn disease (3, 4). Myeloid cells, including monocytes, dendritic cells (DC), and macrophages, are major producers of TNF, which activates multiple cell types, including immune, endothelial, stromal, and epithelial cells, at sites of inflammation (2, 5). TNF also regulates myeloid cell function by augmenting inflammatory cytokine production, synergizing with receptor activator of nuclear factor κB (NF-κB) ligand (RANKL) to induce osteoclast differentiation, and promoting DC maturation (2, 6, 7).

TNF binds to two distinct cell surface receptors, TNFR1 (p55/p60) and TNFR2 (p75/p80), and induces intracellular signaling cascades regulating cell survival, apoptosis, differentiation, proliferation, and activation of immune functions (1, 2). Most of the biological effects of TNF are mediated by TNFR1, which activates canonical NF-κB signaling leading to activation of NF-κB1 (p50), RelA (p65), and c-Rel; mitogen-activated protein kinase (MAPK) signaling leading to induction and activation of AP-1 family transcription factors such as Fos and Jun; and activation of caspase 8 and downstream proteolytic cascades. TNF-mediated NF-κB and AP-1 activation regulates expression of inflammatory genes and promotes cell survival, whereas activation of caspase 8-mediated pathways induces apoptosis (1, 8). The outcomes of cellular responses to TNF vary depending on the cell type and microenvironment and are controlled by the balance of activation of the different signaling pathways.

Members of the nuclear factor of activated T cells (NFAT) transcription factor family were originally described as important regulators of lymphocyte activation and differentiation (9). NFAT activation is calcium (Ca²⁺) dependent and requires dephosphorylation by the Ca²⁺-dependent phosphatase calcineurin (Cn), which allows nuclear translocation of NFAT and transcription of target genes (10). Thus, NFAT is most prominently activated by receptors that activate Ca²⁺ signaling, such as antigen receptors that signal via immunoreceptor tyrosine-based activation motif (ITAM)-containing adaptor proteins (11, 12). The function of NFATs is context dependent and modulated by interactions with other transcription factors, in particular, AP-1 proteins (9, 13). More recently, functions of NFAT in myeloid cells are emerging, including modulation of cytokine production (14, 15), DC survival (16), and induction of cell fusion and differentiation into osteoclasts (17). In myeloid cells, basal NFAT expression is low and the strong activation of NFAT necessary for driving cell fusion and osteoclast differentiation requires two signals: induction of NFAT expression via NF-κB and AP-1 and posttranslational activation by Ca²⁺–Cn signaling induced by ITAM-associated immunoreceptors (18, 19). NFATc1 is the major inducible protein in myeloid cells and is the only NFAT protein that undergoes autoamplification (20, 21). The only known strong inducer of NFATc1 expression in myeloid cells is the TNF family member RANKL, which activates AP-1 and both canonical and noncanonical NF-κB signaling that are required for NFATc1 expression (21, 22).

The direct early signaling events described above that occur minutes to several hours after TNF stimulation and are important for acute cell activation have been extensively characterized. However, the mechanisms of signal propagation over time are not clear. Such mechanisms that sustain and regulate temporal activation of signaling pathways and transcription factors after the initial transient response to TNF are important for understanding the role of TNF in cell differentiation and chronic inflammation, where cells are exposed to TNF for prolonged periods. We recently described a TNF-induced cascade of gene activation leading to the delayed induction of an IFN-STAT1 response that is apparent 1–2 d after TNF stimulation of primary macrophages (23). Here, we extended the time frame of analysis to 10–14 d of TNF stimulation and found that TNF-induced fusion of human macrophages into osteoclast-like multinucleated polykaryons is associated with induction and activation of NFATc1, which became apparent after 1–3 d of TNF stimulation. TNF-mediated induction of NFATc1 was associated with delayed and sustained activation of c-Jun/AP-1 and NF-κB pathways and with sustained Ca²⁺ oscillations. TNF-induced NFATc1 expression primed cells for enhanced fusion and increased osteoclastogenesis in response to RANKL, and NFATc1 expression was elevated in synovial macrophages from patients with TNF-driven inflammatory arthritis. These results identify a new signaling pathway activated by TNF that mediates delayed and sustained signaling responses important for macrophage differentiation and chronic inflammation and may play a role in pathological bone resorption during inflammatory arthritis.

Results

TNF Triggers a Differentiation Program in Human Macrophages via Activation of NFATc1.

Recent studies demonstrated that exposure to TNF for 24–48 h induces a complex cascade of gene expression in human and mouse macrophages (23, 24). We hypothesized that sustained exposure to TNF can further affect gene expression and reprogram human macrophages. We cultured human monocytes with TNF for 10–14 d and found that TNF-treated macrophages became tightly adherent with a spread morphology and started clustering after 3 d, and a subpopulation of cells (≈8–10%) fused and formed large multinuclear cells (Fig. 1 A and B). Although the extent of multinuclear cell formation in response to TNF varied among donors, cell fusion was detected in 45 of 52 experiments. In contrast, we did not observe TNF-induced cell fusion in cultures of mouse bone marrow-derived macrophages (BMDM). Next, we analyzed the expression of inflammatory genes and genes involved in cell fusion in TNF-treated macrophages. Interestingly, the previously described (23) expression of IFN-dependent genes and inflammatory chemokines declined after 2 d of culture (Fig. 1C and Fig. S1A, Left). Instead, we found that expression of a new group of genes known to be involved in cell fusion, such as DC-STAMP and ITGB3 (encoding β3 integrin), gradually increased over the 9-d culture period (Fig. 1C). TNF also induced delayed expression of genes associated with differentiation of osteoclasts, multinucleated cells that are specialized for bone resorption (Fig. S1A, Right). Moreover, we observed formation of resorption pits when cells were cultured on dentin slices in the presence of TNF (Fig. S1B). Our results suggest a transition of macrophages from an inflammatory to a tissue remodeling phenotype and imply that prolonged TNF exposure triggers a differentiation program in human macrophages that is distinct from previously described inflammatory activation.

Fig. 1.

TNF triggers a new differentiation program in human macrophages. Human macrophages were stimulated with TNF or left untreated for 8–12 d. (A) Cells were fixed and stained for TRAP activity (TRAP+ cells are stained purple). (B) Quantification of TRAP+ multinuclear cells. Each circle indicates an independent experiment (n = 13). Horizontal lines show the median; **P < 0.01. (C) Real-time PCR analysis of gene expression in human macrophages treated with TNF, normalized to expression of GAPDH and presented relative to expression in untreated cells at the same time point (set as 1). Data are representative of four independent experiments.

Our findings raised the question of which mechanisms drive this newly described TNF-induced differentiation program. Macrophage fusion is induced by various cytokines (25); because NFATc1 plays a key role in fusion induced by TNF superfamily member RANKL, we hypothesized that NFATc1 may mediate TNF-induced macrophage fusion. First, we examined the effects of TNF on NFATc1 expression in human macrophages. TNF induced a transient and moderate increase in NFATc1 mRNA levels after 3 h of treatment with subsequent sustained up-regulation after 24 h, which persisted for several days (Fig. 2A). Consistent with this result, NFATc1 protein began to accumulate 24–48 h after TNF stimulation and was strikingly elevated after 3–7 d of culture of monocytes with TNF (Fig. 2B and Fig. S2A). In most human donors, TNF induced more prominent expression of the shortest NFATc1 isoform A (NFATc1/A) known to be the most abundant isoform in osteoclasts and effector T cells (20, 21), whereas induction of B and C isoforms was less robust. TNF-induced NFATc1 expression and macrophage fusion were not affected by osteoprotegrin (OPG), a decoy receptor for RANKL that blocked RANKL-induced multinuclear cell generation (Fig. S2B). These results suggest a role for TNF in cell fusion and NFATc1 induction independent of RANKL.

Fig. 2.

TNF induces NFATc1 expression and activation in human macrophages. Human macrophages were stimulated with TNF for the indicated times and NFATc1 expression was analyzed. (A) Real-time PCR analysis of NFATC1 expression in TNF-treated macrophages presented relative to expression in untreated cells at the same time point (set as 1). (B) Immunoblot of whole-cell extracts from macrophages cultured with (+) or without (−) human TNF for NFATc1. Arrows indicate three NFATc1 isoforms. (C) Immunoblot of whole-cell extracts from macrophages cultured for 72 h with (+) or without (−) TNF for NFATc1. Arrow indicates dephosphorylated NFATc1/A. (D) Microscopy of human macrophages cultured for 4 d with or without TNF and stained with anti-NFATc1 (green). Nuclei are counterstained with propidium iodide (PI; red). Original magnification, 40×. (E) Immunoblot of nuclear extracts from human macrophages cultured with (+) or without (−) TNF for 72 h. A total of 20 μg/mL of etanercept or human IgG (control) was added 0, 24, or 48 h after addition of TNF. Data are representative of 3 (A, B, and D), 10 (C), and 5 (E) independent experiments.

Activation of NFAT transcription factors requires dephosphorylation by the Ca²⁺-dependent phosphatase Cn, which allows nuclear translocation and transcription of target genes (10). TNF treatment induced a more rapidly migrating form of NFATc1/A (Fig. 2C), which is suggestive of dephosphorylation and thus posttranslational activation; shifts in mobility of the B and C isoforms were not apparent, likely secondary to limited resolution of bands under these electrophoresis conditions, but nuclear expression of these isoforms suggests dephosphorylation (Fig. 2E, and see below). Consistent with these results, TNF-induced nuclear localization of NFATc1 was detected by immunofluorescence staining (Fig. 2D). Importantly, blocking of TNF receptor signaling by addition of soluble TNF receptor etanercept at the beginning of the cultures as well as 24 or 48 h after TNF stimulation abrogated NFATc1 nuclear import (Fig. 2E) and, in parallel, inhibited cell fusion, albeit to a lesser extent when added 48 h after TNF (Fig. S2C). TNF also increased NFATc1 expression in mouse BMDM, but to a lesser extent than in human macrophages (Fig. S2D). It is not yet clear if differences in induction of macrophage fusion by TNF in mouse and human cells are related to quantitative differences in NFATc1 induction or other factors. Overall, our results indicate that continuous TNF signaling is required for NFATc1 activation and fusion of human macrophages.

Biphasic Activation of AP-1 and NF-κB Pathways in TNF-Treated Macrophages.

AP-1 and NF-κB signaling are important for RANKL-induced NFATc1 expression, which is further amplified by NFATc1 driving its own promoter in an autoamplification loop (21). AP-1 and canonical NF-κB signaling are strongly induced by TNF and orchestrate the early stages of TNF response in multiple cell types (1). However, regulation of AP-1 and NF-κB during chronic exposure to TNF has not been investigated. As expected, TNF rapidly induced nuclear expression of AP-1 proteins c-Fos and c-Jun after 1 h of TNF stimulation (Fig. 3A, lane 2). A more detailed time course showed that c-Fos was not detected in nuclear extracts at later time points. In contrast, c-Jun reemerged in nuclei after 48 h (24 h in some experiments) of TNF treatment, and its expression was sustained until the 72-h time point (Fig. 3A, lanes 10 and 12), which coincides with increased expression of NFATc1. Furthermore, blockade of TNF receptor signaling by adding etanercept either together with or after TNF down-regulated c-Jun nuclear expression (Fig. S3A, Left). Inhibitors of p38 and c-Jun amino-terminal kinase (JNK) MAPKs, known activators of AP-1, decreased TNF-induced NFATc1 expression (Fig. 3B, compare lanes 3 and 4), and inhibition of JNK also decreased TNF-induced polykaryon formation (Fig. S3C, Left) without significant effect on cell viability (Fig. S3C, Right, and Fig. S3B). To determine whether c-Jun directly binds to the NFATC1 promoter we used ChIP assays and found that c-Jun was recruited to the NFATC1 locus 1 h following TNF treatment but not at 40 h (Fig. 3D, Lower), suggesting that regulation of NFATC1 by c-Jun may be indirect at later time points. These results confirm the importance of MAPK-mediated pathways in NFATc1 induction but demonstrate differences in delayed AP-1 activation in response to TNF in human cells relative to previously described RANKL responses in mouse cells.

Fig. 3.

TNF induces delayed and sustained activation of AP-1 and NF-κB signaling in human macrophages. Human macrophages were stimulated with TNF (+) or left untreated (−) for the indicated times, and nuclear extracts were analyzed by immunoblotting for expression of (A) c-Fos and c-Jun or (C) NF-κB subunits. (B) Human macrophages were stimulated with TNF for 48 h. BAY 11-7082 (20 μM), a combination of SP600125 and SB220025 (10 μM each), and vehicle control, dimethyl sulfoxide (DMSO), were added for the last 3 h of culture, and whole-cell extracts were analyzed by immunoblot for NFATc1 expression. (D) NF-κB p65 (Upper) and c-Jun (Lower) recruitment to the NFATC1-P1 promoter was assessed by ChIP after stimulation with TNF for 1 and 40 h. Data are representative of four (A and B), three (C), and three (1 h) to seven (40 h) (D) independent experiments.

We next analyzed the effects of prolonged TNF stimulation on NF-κB activation in human monocytes. TNF acutely activates the canonical NF-κB pathway leading to IκB kinase β (IKKβ)-dependent activation of p50 and p65 (1, 2). TNF does not activate proximal components of the noncanonical NF-κB–inducing kinase (NIK)-IKKα–mediated NF-κB signaling pathway but has been shown to acutely activate RelB independently of IKKα (26) and suggested to suppress later phases of p52 and RelB activation by inducing expression of the p100 inhibitory protein in murine macrophages (22, 27). In human macrophages, TNF induced nuclear expression of components of both canonical and noncanonical NF-κB pathways with biphasic kinetics (Fig. 3C). Remarkably, despite differences in the levels and kinetics of expression at early time points, we detected sustained nuclear expression of p50, phospho(Ser536)-p65 (RelA), p52, and RelB 24–72 h after TNF stimulation (Fig. 3C, lanes 8, 10, and 12), which was dependent on continuous TNF receptor signaling (Fig. S3A). Similar to a previous report that studied RANKL signaling in murine macrophages (21), ChIP assays showed p65 (RelA) recruitment to the NFATC1 promoter 1 h after TNF stimulation, but in contrast to the RANKL study, we also observed p65 occupancy of the promoter after 40 h of treatment with TNF (Fig. 3D, Upper). Thus, TNF induced delayed and sustained activity of the canonical NF-κB pathway as well as nuclear accumulation of RelB and p52. As expected, NFATc1 induction by TNF was abrogated by an inhibitor of IκBα phosphorylation (BAY-11-7082) (Fig. 3B, lane 5). These results show that TNF induces sustained activation of multiple NF-κB proteins that are known to be important for NFATc1 expression and suggest a direct role for p65 in regulation of NFATC1.

In contrast to TNF, RANKL does not induce sustained canonical NF-κB activation (21, 28), and thus noncanonical NF-κB signaling typically mediated by NIK/IKKα has been implicated in RANKL-induced osteoclastogenesis (29, 30). Consistent with these reports that used murine cells, RANKL-induced multinuclear cell formation was decreased in human macrophages when the noncanonical NF-κB pathway was inhibited by RNA interference (RNAi)-mediated knockdown of IKKα expression (Fig. S3 D and E, Left). In contrast, RNAi-mediated knockdown of IKKα did not affect cell fusion in TNF-treated human macrophages despite effective IKKα down-regulation (Fig. S3E, Right); these results are consistent with a previous report showing IKKα-independent activation of RelB by TNF (26). Taken together, these results demonstrate that TNF induces delayed activation of AP-1 and NF-κB that is required for NFATc1 expression and identify differences between TNF and RANKL in the induction of these signaling pathways.

Induction of Sustained Ca²⁺ Signaling and NFATc1 Activation by TNF in Human Macrophages.

Our findings that TNF treatment results in NFATc1 dephosphorylation and nuclear translocation imply induction of Ca²⁺ signaling, which is indispensable for NFATc1 activation. To our knowledge, activation of Ca²⁺ signaling by TNF has not been described. We measured intracellular Ca²⁺ and detected sustained oscillations in Ca²⁺ levels 3 d after TNF stimulation (Fig. 4A). Although irregular oscillations were observed both in control and in TNF-treated cells, especially early in the culture, at late time points there was a striking and highly significant (P < 0.0001 by Fisher's exact test) increase in regular oscillations (similar wavelength and amplitude) in TNF conditions compared with control (Fig. 4A). This finding is in line with our observation that NFATc1 dephosphorylation and nuclear translocation were predominantly detected at 72 h of TNF treatment (Fig. 2 C and D). Interestingly, the percentage of cells with Ca²⁺ oscillations and NFATc1 nuclear translocation was similar (∼10–12%) and reflected the fusion rate in TNF-treated cultures (Fig. 4A).

Fig. 4.

TNF triggers Ca²⁺ signaling in human macrophages and activates NFATc1 by a Ca²⁺–Cn-dependent mechanism. (A) Human macrophages were cultured with (Left) or without (Right) TNF for 72 h and intracellular Ca²⁺ levels were monitored for 180 s. Data are shown as representative oscillatory patterns for single cells. (Right) The total number of cells with a regular (similar wavelength and amplitude) oscillatory pattern in TNF-treated and control cultures obtained in three independent experiments is shown. Differences between TNF and control conditions were analyzed by Fisher's exact test (P < 0.0001). (B) Human macrophages were stimulated with TNF for 72 h in the presence of 10 μM of Cn-binding peptide 11R-VIVIT or control 11R-VEET peptide. Nuclear extracts were assayed for NFATc1 by immunoblotting. (C) Quantification of TRAP+ multinuclear cells after 6–8 d of culture with TNF in the presence of 11R-VIVIT (solid circles) or control peptide (open circles). Each circle indicates an independent experiment (n = 7). Horizontal lines show the median; *P < 0.05. (D) BMDMs from DAP12/FcRγ-deficient (DKO) or C57BL/6J wild-type (WT) control mice were stimulated with mouse TNF for the indicated times and expression of NFATc1 was assessed by immunoblotting. Data are representative of two (B) and three (D) independent experiments.

Sustained low-amplitude Ca²⁺ oscillations are sufficient to activate Cn and selectively activate Cn-NFAT signaling relative to other Ca²⁺-mediated signaling pathways (9, 31, 32). To investigate the role of Cn in TNF-induced NFATc1 activation, we used a high-affinity Cn-binding peptide, 11R-VIVIT, that blocks NFATc1 binding to Cn (33). 11R-VIVIT, but not control 11R-VEET peptide, blocked NFATc1 nuclear translocation and suppressed TNF-induced cell fusion (Fig. 4 B and C). These findings were corroborated using additional Cn inhibitors, cyclosporine A (CsA) and FK506 (Fig. S4 A–E). Consistent with inhibition of Cn, CsA treatment significantly decreased induction of the NFATc1-dependent integrin β3 gene (34, 35) by TNF (Fig. S4B). Taken together, our findings implicate TNF-induced Cn-NFATc1 signaling in human macrophage fusion.

In myeloid cells, receptors associated with ITAM-containing signaling adaptors FcRγ and DAP12 play an important role in NFATc1 activation by mediating Ca²⁺ signaling via activation of spleen tyrosine kinase (Syk) and phospholipase Cγ 2 (PLCγ2) (18, 19, 36). We took advantage of our findings that TNF can activate NFATc1 in mouse BMDM to examine if deficiency in FcRγ and DAP12 affects TNF-mediated NFATc1 induction and activation. TNF failed to induce Ca²⁺ oscillations and up-regulate NFATc1 in BMDM from Dap12^−/−/FcRγ^−/− mice (DKO mice) (Fig. 4D and Fig. S4H). Deficiency in FcRγ alone had no discernable effect (Fig. S4F, lanes 3 and 4 compared with lanes 6 and 7), whereas DAP12 deficiency resulted in a modest but consistent decrease in NFATc1 expression only at the 72-h time point (Fig. S4G, lanes 3 and 4 compared with lanes 6 and 7) that correlated with the absence of Ca²⁺ oscillations at this late time point (Fig. S4H). These results suggest a role for DAP12 in Ca²⁺ signaling-dependent autoamplification of NFATc1 expression and are consistent with previous reports of redundant function of these adaptors in other systems (18, 19). We next tested whether NFATc1 induction by TNF depends on Syk and found that the Syk inhibitor piceatannol blocked NFATc1 up-regulation in human macrophages (Fig. S4I), but 80% knockdown of Syk using RNAi was not sufficient to diminish NFATc1 expression. Thus, we turned to the murine system to further address the role of Syk. As Syk deficiency is embryonic lethal, we used fetal liver-derived macrophages rather than BMDM. Induction of NFATc1 in fetal liver-derived macrophages was weak, and consistent differences in NFATc1 expression between control and Syk-deficient cells were not observed (Fig. S4J, Lower); however, we observed decreased dephosphorylation of NFATc1 in Syk-deficient macrophages at earlier time points (Fig. S4J, Upper, lanes 3 and 4 compared with lanes 6 and 7). Collectively, these results support a partial role for ITAM-coupled receptors and Syk in TNF-induced NFATc1 activation that is dependent on context, timing, and the source of myeloid cells; and is consistent with variable results obtained concerning the role of ITAM-coupled receptors and Syk in RANKL-induced NFATc1 induction in murine systems (18, 19, 36, 37). Taken together, our findings demonstrate that TNF triggers Ca²⁺ signaling that activates NFATc1 and thereby initiates a new differentiation program in human macrophages.

TNF Primes Macrophages for Increased and More Rapid Responses to RANKL Stimulation.

Next, we wished to test if TNF-induced NFATc1 activation in human macrophages has potential biological significance. We hypothesized that accumulation of transcriptionally active NFATc1 can prime macrophages for more robust responses to RANKL. To test that possibility, we cultured human monocytes for 3 d with or without TNF before addition of RANKL for 1 d (Fig. 5A). As expected, RANKL stimulation of control monocytes for only 24 h had minimal effects on NFATc1 activation and generation of TRAP+ multinuclear cells (Fig. 5A, lane 2, and 5B, column 3). In contrast, in TNF-primed monocytes, subsequent RANKL stimulation for 24 h rapidly activated NFATc1 as assessed by nuclear localization (Fig. 5A, lane 4) and strongly increased the number of TRAP+ polykaryons (Fig. 5B, column 5). When we added etanercept 24 h after TNF stimulation to prevent TNF-mediated induction of NFATc1 expression, nuclear NFATc1 was no longer detected after a subsequent 24-h stimulation with RANKL (Fig. 5A, lane 8), and RANKL-induced rapid osteoclastogenesis was significantly diminished (Fig. 5B, column 6 vs. column 5). These results demonstrate that TNF increases osteoclastogenesis in response to RANKL and can be potentially implicated in increased osteolysis at sites of chronic inflammation.

Fig. 5.

Continuous TNF signaling primes macrophages for more robust responses to RANKL. Human macrophages were treated with TNF for 72 h and RANKL was added where indicated for an additional 24 h. To block TNF signaling, etanercept was added to cultures as indicated 24 h after TNF stimulation; hIgG served as the negative control. After 96 h of culture, nuclear extracts were assayed for NFATc1 by immunoblotting (A), and TRAP+ multinuclear cells were quantitated (B). Each circle indicates an independent experiment (n = 6). Horizontal lines show the median; *P < 0.05. Data in A are representative of three independent experiments.

We further tested this idea by determining if NFATc1 expression is increased in synovial macrophages from patients with inflammatory arthritis obtained during an active phase of joint inflammation when they are exposed to high amounts of TNF. NFATc1 expression was increased in half of the inflammatory synovial macrophage samples (Fig. S5, lanes 4, 7, and 11–14), and expression levels were similar to those in TNF-stimulated macrophages from healthy donors described in this study (Fig. S5, lane 2). These results suggest that TNF may contribute to increased NFATc1 expression and thereby promote osteoclastogenesis in inflammatory arthritis that is associated with pathological bone resorption.

Discussion

TNF signaling has been extensively studied for several decades, and the proximal signaling events leading to acute cell activation or apoptosis via NF-κB, MAPK, and caspase-mediated pathways have been well characterized. In contrast, little is known about later phases of TNF response that are relevant for cells that are exposed to TNF for several days under inflammatory conditions. In this study, we identified delayed TNF-induced signaling responses in primary human macrophages and found a sustained wave of signaling that begins after 1–2 d of TNF stimulation and is characterized by nuclear accumulation of multiple NF-κB proteins and c-Jun and by Ca²⁺ oscillations. These signaling events culminate in the induction and activation of NFATc1, which mediates a gene expression program linked to cell fusion, osteoclast differentiation, and potentially additional NFAT-mediated functions. High NFATc1 expression was apparent in synovial macrophages in a subset of patients with inflammatory arthritis and primed human macrophages for enhanced osteoclastogenesis in response to RANKL. These findings identify TNF-induced signaling responses important for myeloid cell differentiation and suggest a role for TNF-induced NFAT-mediated functions in chronic inflammation and associated bone resorption.

A link between TNF and Ca²⁺ signaling was not previously appreciated, and TNF is not known to activate proximal signaling pathways that are important for an acute Ca²⁺ flux. Indeed, we detected TNF-induced calcium oscillations only after several days of culture. Such delayed low-amplitude Ca²⁺ oscillations have been previously linked to Cn-NFAT activation and occur after tonic stimulation of antigen receptors in anergic B cells and after RANKL stimulation of mouse osteoclast precursors (18, 19, 32). In both cases, Ca²⁺ signaling is mediated by ITAM-containing adaptors; RANKL is not directly associated with ITAM-containing adaptors and how RANKL activates ITAM signaling has not been clarified despite intensive efforts. Similarly, TNF-induced activation of NFATc1 was mediated at least in part by ITAM-containing adaptors DAP12 and FcRγ that likely induced NFATc1 dephosphorylation and autoamplification of its own expression. A plausible explanation for TNF-mediated activation of ITAM signaling is the induction of endogenous ligands for ITAM-coupled receptors; many of these receptors are basally or inducibly expressed on myeloid cells. The identity of most myeloid cell-expressed ligands for ITAM-associated receptors remains unknown. Overall, we demonstrate that initiation of a differentiation program by TNF required costimulatory signals from ITAM-coupled receptors to activate tonic Ca²⁺ signaling and sustain NFATc1 activity.

In myeloid cells, the best established functions of NFATc1 are to induce expression of genes important for cell fusion and osteoclast differentiation. Thus, induction and activation of NFATc1 by TNF provide a mechanistic explanation by which TNF can induce giant cell formation in infectious granulomas such as in Mycobacterium tuberculosis infection (38) and in TNF-driven chronic granulomatous diseases such as sarcoidosis (39). In addition to fusion and osteoclastogenesis, NFAT regulates cytokine production, including TNF production, in myeloid cells (14, 15, 35) and thus can play an important role in myeloid cell biology independently of osteoclast differentiation. Context-dependent function of NFAT is determined by its interactions with NF-κB and AP-1 proteins that direct NFAT to composite binding sites present in different target genes (9). Thus, NFATc1 in TNF-primed macrophages can activate different target genes and potentially subsume different functions depending on additional environmental cues that regulate AP-1 and NF-κB signaling. In our system, NFATc1 may interact with canonical and noncanonical pathway NF-κB proteins and with c-Jun, which would promote differentiation into osteoclast-like cells and prime cells for enhanced and rapid RANKL-mediated osteoclastogenesis. The synergy between TNF and RANKL that we have described is distinct from but complementary to the previously reported synergy where prior exposure to RANKL initiates a differentiation program and subsequent stimulation with TNF then enhances osteoclastogenesis (discussed in ref. 22 and references therein). In contrast, we show that prior exposure to TNF of sufficient duration to induce expression of NFATc1 enhances subsequent RANKL responses. During chronic inflammation both mechanisms of TNF-RANKL synergy may be engaged.

Considerable progress has been made in understanding the early phase of NF-κB and AP-1 activation 1–3 h after TNF stimulation (1, 2, 8), but little is known about how TNF activation of these pathways is propagated over time. In the early phase, TNF activates primarily canonical NF-κB signaling via IKKβ to induce activation and nuclear translocation of p50 and p65/RelA. p52 and RelB are typically activated by other ligands such as RANKL and CD40 via a noncanonical NIK-IKKα–mediated pathway, although TNF can also activate RelB by an IKKα-independent pathway in certain cell types (26). We found that in human macrophages, TNF induced biphasic nuclear accumulation of c-Jun and both “canonical” (p50/p65) and “noncanonical” (p52/RelB) NF-κB proteins. Jun, p50/p65, and p52/RelB are all required for osteoclast differentiation, and activation of these proteins explains how TNF can induce osteoclast differentiation in human macrophages. In striking contrast, in murine macrophages, TNF strongly induces p100 expression that inhibits generation and activation of p52 and RelB (22). These differences in TNF-mediated p52 and RelB activation help explain why TNF-mediated induction of NFATc1 and osteoclast differentiation are much less efficient in murine cells. Our study also revealed considerable differences between TNF- and RANKL-induced differentiation programs in macrophages: (i) TNF preferentially induced strong and sustained nuclear accumulation of c-Jun, in contrast to c-Fos induction by RANKL; (ii) TNF induced sustained NFATC1 promoter occupancy by canonical NF-κB factor RelA; (iii) TNF-induced osteoclastogenesis was substantially less dependent on IKKα. Our findings demonstrate that TNF induces delayed activation of AP-1 and NF-κB pathways in human macrophages that contribute to NFATc1 expression and highlight differences between TNF and RANKL in the induction of NFATc1 and downstream human macrophage differentiation.

In summary, our data show that long-term exposure to TNF reprograms human macrophages by increasing NFATc1 expression and function through Ca²⁺-dependent mechanisms and that high NFATc1 expression primes cells for enhanced fusion leading to increased osteoclastogenesis in response to RANKL in chronic inflammatory conditions, such as arthritis. Inhibition of TNF-induced signaling pathways leading to Cn-NFAT activation may offer new opportunities for therapeutic intervention in TNF-driven chronic inflammatory diseases.

Materials and Methods

Cell Culture and Mice.

Peripheral blood mononuclear cells (PBMCs) were obtained from the blood of healthy donors by density gradient centrifugation using Ficoll (Invitrogen), using a protocol approved by the Hospital for Special Surgery Institutional Review Board. CD14⁺ human monocytes were purified from PBMCs by positive selection using anti-CD14 magnetic beads, as recommended by the manufacturer (Miltenyi Biotec). Monocytes were cultured in α-MEM medium (Invitrogen) supplemented with 10% FBS (Defined; HyClone), penicillin/streptomycin (Invitrogen), l-glutamine (Invitrogen), and 20 ng/mL of human macrophage colony-stimulating factor (M-CSF; Peprotech) in the presence or absence of 40 ng/mL human TNF (Peprotech). Inflammatory arthritis macrophages were purified using anti-CD14 magnetic beads from preexisting synovial fluids as described in SI Materials and Methods. Murine monocyte progenitor cells were isolated from bone marrow, differentiated into macrophages, and treated as described in SI Materials and Methods.

Multinuclear Cell/Osteoclast Differentiation.

To generate osteoclast precursors, human monocytes were incubated in 96-well plates with 20 ng/mL of M-CSF and 40 ng/mL of human TNF or soluble RANKL for various times. Cytokines were replenished every 3 d. At the end of the culture period, cells were stained for tartrate-resistant acid phosphatase (TRAP) activity, according to the manufacturer's instructions (Sigma). Multinucleated (more than three nuclei), TRAP-positive cells were counted in triplicate wells of 96-well plates.

Intracellular Ca²⁺ Measurements.

For Ca²⁺ measurement, cells were loaded with 5 μM of fluo-4 AM and 10 μM of Fura Red AM (Molecular Probes and Invitrogen) and analyzed using confocal microscopy (Leica). To estimate intracellular Ca²⁺ concentrations in single cells, the ratio of the fluorescence intensity of fluo-4 to that of Fura Red was calculated.

See SI Materials and Methods for more detailed information.