Tumor Necrosis Factor α Induces Sustained Signaling and a Prolonged and Unremitting Inflammatory Response in Rheumatoid Arthritis Synovial Fibroblasts




The nonresolving character of synovial inflammation in rheumatoid arthritis (RA) is a conundrum. To identify the contribution of fibroblast-like synoviocytes (FLS) to the perpetuation of synovitis, we investigated the molecular mechanisms that govern the tumor necrosis factor α (TNFα)–driven inflammatory program in human FLS.


FLS obtained from the synovial tissues of patients with RA or osteoarthritis were stimulated with TNFα and assayed for gene expression and cytokine production by real-time quantitative reverse transcription–polymerase chain reaction analysis and enzyme-linked immunosorbent assay. NF-κB signaling was evaluated by Western blotting. Histone acetylation, chromatin accessibility, and NF-κB p65 and RNA polymerase II (Pol II) occupancy at the interleukin-6 (IL-6) promoter were measured by chromatin immunoprecipitation and restriction enzyme accessibility assays.


In FLS, TNFα induced prolonged transcription of messenger RNA (mRNA) for IL-6 and progressive accumulation of IL-6 protein over 4 days. Similarly, induction of mRNA for CXCL8/IL-8, CCL5/RANTES, matrix metalloproteinase 1 (MMP-1), and MMP-3 after TNFα stimulation was sustained for several days. This contrasted with the macrophage response to TNFα, which characteristically involved a transient increase in the expression of proinflammatory genes. In FLS, TNFα induced prolonged activation of NF-κB signaling and sustained transcriptional activity, as indicated by increased histone acetylation, chromatin accessibility, and p65 and Pol II occupancy at the IL-6 promoter. Furthermore, FLS expressed low levels of the feedback inhibitors A20-binding inhibitor of NF-κB activation 3 (ABIN-3), IL-1 receptor–associated kinase M (IRAK-M), suppressor of cytokine signaling 3 (SOCS-3), and activating transcription factor 3 (ATF-3), which terminate inflammatory responses in macrophages.


TNFα signaling is not effectively terminated in FLS, which leads to an uncontrolled inflammatory response. The results suggest that prolonged and sustained inflammatory responses by FLS in response to synovial TNFα contribute to the persistence of synovial inflammation in RA.

Rheumatoid arthritis (RA) is characterized by synovitis, systemic manifestations, and significant morbidity and mortality (1, 2). The pathogenesis of RA is a multistep process consisting of a preclinical phase involving the generation of autoantibodies, an initiation phase during which synovial inflammation emerges, and a perpetuation phase dominated by nonresolving synovial inflammation and joint destruction (3). During each step, there is interplay between environmental, genetic, hormonal, and stochastic factors (1, 4). The T cell–centric pathogenetic model of RA was challenged by the discovery that cytokines derived from macrophages and fibroblast-like synoviocytes (FLS), such as tumor necrosis factor α (TNFα) and interleukin-6 (IL-6), respectively, play a central role in its pathogenesis (5). The current pathogenetic model is the so-called integrated model, which implies that within the synovium, there is cross-talk between T and B lymphocytes, macrophages, and FLS, involving cell-to-cell interactions and soluble factors that drive the initiation and perpetuation of synovial inflammation (1).

Synovial inflammation in RA is chronic, and even with aggressive immunosuppression, long-term remission is rarely achieved (6). One potential explanation is that the continuous presence of arthritogenic antigens or the emergence of neoantigens sustains autoimmune responses against synovial elements (7). An alternative, but not mutually exclusive, scenario is that the anatomic components of the synovium, such as stromal cells, are hypersensitive to systemic or local inflammatory stimuli. This model posits that in contrast to epithelial surfaces, where homeostatic mechanisms effectively control responses to exogenous or endogenous inflammatory factors, the joint synovium mounts an exaggerated or unremitting response. Susceptibility of synovial tissue to inflammation triggered by diverse factors such as injury, infection, immune complexes, and cytokines is supported by data from animal models (8–13).

RA FLS have been described as transformed cells, sharing morphologic features with tumor cells, such as resistance to apoptosis, potentially due to somatic mutations in p53 (14). RA FLS display and retain an invasive capacity against articular cartilage (15). FLS express adhesion molecules and receptors for cytokines and Toll-like receptor (TLR) ligands that mediate their activation during synovial inflammation (16). In RA, FLS are the major synovial producers of IL-6, a key pathogenic cytokine (17, 18). Upon activation, FLS produce a constellation of cytokines, growth factors, chemokines, adhesion molecules, costimulatory molecules, and tissue destructive factors (16, 19, 20). FLS mediate synovial recruitment and the retention, organization, activation, and survival of inflammatory cells, and they enhance synovial neoangiogenesis and induce osteoclastogenesis and cartilage degradation (14, 21).

TNFα is a key driver of synovial inflammation in 50–70% of RA patients. Macrophages are likely the major source of synovial TNFα (5, 18). However, macrophages display a transient inflammatory response to TNFα because of homeostatic mechanisms that terminate inflammatory signaling and impose a chromatin-mediated barrier that suppresses inflammatory gene expression (22). Following initial exposure to TNFα, macrophages develop resistance to subsequent challenge with inflammatory stimuli, including TNFα, IL-1, and TLR ligands (23). This raises the question of which are the major cells that respond to synovial TNFα to sustain inflammation. Experiments with transgenic mice that express TNFα suggest that FLS, which are in close proximity to macrophages, are the major responders to TNFα (13).

In this study, we found that, in stark contrast to macrophages, TNFα stimulation of FLS resulted in a sustained inflammatory response, characterized by prolonged expression of cytokines, chemokines, and matrix metalloproteinases (MMPs). Prolonged gene expression in FLS was associated with sustained NF-κB signaling, prolonged transcriptional activity, as indicated by increased histone acetylation, chromatin accessibility, and NF-κB and RNA polymerase II (Pol II) occupancy at the IL-6 promoter, and ineffective induction of feedback mechanisms that restrain inflammatory signaling and gene expression in macrophages. These results suggest that ineffective termination of inflammatory signaling in FLS contributes to the persistence of synovial inflammation.



Synovial tissues were obtained from RA or osteoarthritis (OA) patients who were undergoing total knee replacement surgery (protocol approved by the Institutional Review Board at the Hospital for Special Surgery). The diagnoses of RA and OA were based on the American College of Rheumatology criteria (24, 25).

Cell purification.

Synovial tissue fragments were incubated with Dispase for 90 minutes at 37°C. Cells were allowed to adhere to tissue culture dishes and passaged every 3–5 days. A total of 3–4 passages yielded a relatively homogeneous population of FLS. CD14+ cells were purified with the use of anti-CD14 magnetic beads (Miltenyi Biotec) from peripheral blood mononuclear cells obtained from healthy volunteer donors.

Cell culture.

FLS and CD14+ cells were cultured in α-minimum essential medium plus 10% fetal bovine serum. The following reagents were used as indicated: TNFα (10 ng/ml) and macrophage colony-stimulating factor (20 ng/ml) (both from PeproTech), etanercept (10 μg/ml) and anakinra (10 μg/ml) (both from Amgen), infliximab (10 μg/ml; Janssen Biotech), human IgG (10 μg/ml; Sigma Aldrich), anti-gp130 (5 μg/ml) and mouse IgG2a (5 μg/ml) (both from R&D Systems), CP690550 (10 μM), BAY11-7082 (10 μM), IKK inhibitor II (50 μM), IKK inhibitor XII (10 μM), and MG132 (10 μM) (all from Calbiochem), and TNFα proteinase inhibitor 1 (TAPI-1) (10 μM; Peptides International). DMSO was used as a vehicle control.

Enzyme-linked immunosorbent assay (ELISA) and fluorescence-activated cell sorting.

We used a sandwich ELISA to measure levels of IL-6 and TNFα in culture supernatants (0.5 × 106 FLS in 3 ml of medium and 2 × 106 macrophages in 1 ml of medium). For flow cytometry, monoclonal antibodies to human TNF receptor superfamily 1A (TNFRSF1A; p55) and TNFRSF1B (p75), as well as isotype controls (R&D Systems), were used.

Real-time quantitative reverse transcription–polymerase chain reaction (qRT-PCR) analysis.

RNA was extracted from 0.5 ×106 FLS, 1 μg was reverse transcribed, and qRT-PCR analysis was performed.


Lysates from 105 FLS were fractioned on polyacrylamide gels, transferred to polyvinylidene difluoride membranes, and incubated with antibodies against IκBα, Akt, p65, lamin B1, and p38. Densitometry was performed using ImageJ software (National Institutes of Health).

Restriction enzyme accessibility (REA) assay.

Isolated nuclei were incubated for 30 minutes at 37°C with 50 units of Ssp I (New England Biosciences), and digested genomic DNA was purified. Equal amounts of purified DNA were digested to completion overnight at 37°C with 50 units of Bst XI, precipitated, and analyzed by Southern blotting using radiolabeled probe specific for the IL-6 gene (+51 to +614 relative to the transcription start site).

Chromatin immunoprecipitation analysis.

Cells were treated with 1% formaldehyde to crosslink chromatin. Fixed cells were incubated with lysis buffer and were sonicated using a Bioruptor UCD400 device (Diagenode). A total of 5% of sonicated cell lysates was saved as input. Chromatin was immunoprecipitated using antibodies to histone H3, histone H4, acetylated histone H4 (Millipore), NF-κB p65 subunit (Abcam), and Pol II (Santa Cruz Biotechnology). Crosslinks were reversed, DNA was purified, and enrichment of the target DNA was measured by real-time PCR.

Statistical analysis.

Results are expressed as the mean ± SEM. Analyses were performed using GraphPad Prism analytical software version 5.93 for Windows. When comparing between 3 or more groups, differences were tested using one-way or two-way analysis of variance followed by the Bonferroni test. When comparing between 2 groups, we used Student's paired 2-tailed t-test.


FLS display sustained inflammatory responses to TNFα.

We investigated the molecular mechanisms that govern the TNFα-driven production of proinflammatory and tissue-destructive mediators by FLS. Cells were cultured in the presence or absence of TNFα (10 ng/ml). TNFα was added on the first day and was not replenished. Upon TNFα stimulation, FLS secreted copious amounts of IL-6 (Figure 1A). IL-6 production was sustained and continuously increased over time, reaching >200 ng/ml at later time points (Figure 1A). We next addressed whether this sustained pattern of IL-6 production results from continuous transcription. We found a pattern of increasing IL-6 mRNA expression in TNFα-stimulated FLS during the first 48 hours, which remained at these high levels until day 4 (Figure 1B) (data were normalized relative to GAPDH mRNA [% GAPDH], which was not altered after TNFα stimulation). In addition, we measured levels of active transcription of IL-6 using primers specific for the fourth intron region of the IL-6 gene (amplifying primary transcripts). As shown in Figure 1C, there was a robust and prolonged induction of IL-6 primary transcripts by TNFα. Thus, continuous transcription of IL-6 correlates with the sustained pattern of IL-6 protein production. These results suggest that in FLS, TNFα induces prolonged transcription of IL-6.

Figure 1.

Sustained induction of interleukin-6 (IL-6), chemokines, and matrix metalloproteinases (MMPs) in fibroblast-like synoviocytes (FLS) by tumor necrosis factor α (TNFα). FLS were cultured in time-course experiments in the presence or absence of TNFα (10 ng/ml), which was added on the first day of culture and was not replenished. IL-6 protein (A) in culture supernatants was measured by enzyme-linked immunosorbent assay. IL-6 mRNA (B) and primary transcripts (PTs; primers specific for the fourth intron region of the IL-6 gene) (C) and CXCL8/IL-8, CCL5/RANTES, and MMP-1 mRNA (D) were measured by real-time quantitative reverse transcription–polymerase chain reaction analysis, and the values were normalized relative to mRNA for GAPDH. Values are the mean ± SEM. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001; ∗∗∗∗ = P < 0.0001 versus control, by two-way analysis of variance followed by the Bonferroni test.

Next, we investigated the kinetics of induction of additional TNFα targets in FLS. These included the chemokines CXCL8/IL-8 and CCL5/RANTES, which recruit effector cells to inflamed joints, and the metalloproteinases MMP-1 and MMP-3, which mediate cartilage degradation. There was a prolonged induction pattern by TNFα, similar to IL-6, for all the above genes (Figure 1D and data not shown for MMP-3 expression), suggesting that FLS display a sustained inflammatory and tissue destructive program in response to TNFα. We observed a similar pattern of sustained TNFα-induced inflammatory gene expression in RA and OA FLS, although, consistent with the literature (26), there was higher expression in a subset of RA FLS (data available upon request from the corresponding author). For the experiments described below, only RA FLS were used.

Macrophages display a transient inflammatory response to TNFα.

We wished to directly compare TNFα responses of FLS and macrophages in our system. Synovial macrophages spontaneously produce TNFα and induce an autocrine interferon (IFN)–STAT-1 loop that would confound this analysis (27), so instead, we analyzed TNFα responses in blood-derived macrophages. TNFα stimulation of macrophages resulted in a robust induction of TNFα, IL-1β, and CXCL8/IL-8 mRNA (Figures 2A–C). In stark contrast to FLS, the TNFα-induced inflammatory program in macrophages was transient, returning to baseline expression levels by 3–48 hours (Figures 2A–C). Interestingly, in macrophages the induction of IL-6 mRNA by TNFα was also transient (Figure 2D), with the maximum amounts of IL-6 mRNA being ∼50-fold lower than those induced in FLS (compare Figure 2D with Figure 1B). IL-6 protein levels in culture supernatants were <0.5 ng/ml by ELISA at all time points tested after TNFα stimulation of macrophages (data available upon request from the corresponding author), which was substantially lower than those in FLS culture supernatants (>200 ng/ml) (Figure 1A). Our observations suggest that there are kinetic, qualitative, and quantitative differences in the responses of macrophages and FLS to TNFα.

Figure 2.

Transient inflammatory response to tumor necrosis factor α (TNFα) in human macrophages. CD14+ cells from healthy blood donors were differentiated in vitro into macrophages by stimulation for 48 hours with macrophage colony-stimulating factor (20 ng/ml). Macrophages were cultured in time-course experiments in the presence or absence of TNFα (10 ng/ml), which was added the first day of culture and was not replenished. Levels of mRNA for TNFα (A), interleukin-1β (IL-1β) (B), CXCL8/IL-8 (C), and IL-6 (D) were measured by real-time quantitative reverse transcription–polymerase chain reaction analysis, and the values were normalized relative to mRNA for GAPDH. Values are the mean ± SEM. ∗ = P < 0.05; ∗∗∗∗ = P < 0.0001 versus control, by two-way analysis of variance followed by the Bonferroni test.

Continuous TNFα signaling is required for sustained inflammatory response in FLS.

The sustained pattern of TNFα-induced inflammatory gene expression described above for FLS suggests that TNFα may induce an autocrine cascade of factors that maintain gene expression after the initial TNFα signal resolves. Alternatively, ongoing direct signaling by TNFα may be required. We next performed experiments to distinguish between these two possible mechanisms that can sustain gene expression. As shown in Figure 3A, in FLS culture supernatants, concentrations of exogenous TNFα protein (added at 10 ng/ml) decayed over time, likely due to consumption by the cells and to protein degradation, reaching a concentration of <4 ng/ml after 2 days of culture; FLS did not produce significant amounts of endogenous TNFα, as expected (16). Addition of the MMP inhibitor TAPI-1 to prevent TNF receptor shedding did not change the levels of TNFα (Figure 3A), suggesting that the decrease in TNFα was not secondary to interference with the ELISA by soluble TNF receptors.

Figure 3.

Requirement of ongoing tumor necrosis factor α (TNFα) signaling for a sustained inflammatory response. Fibroblast-like synoviocytes (FLS) were stimulated with TNFα on day 0. A, TNFα protein in culture supernatants in the presence or absence of TNFα proteinase inhibitor 1 (TAPI-1) was measured at the indicated times by enzyme-linked immunosorbent assay (ELISA). B, Etanercept or human IgG (control) was added to cultures on days 0, 1, 2, or 3. Cells were harvested on day 4, and IL-6 mRNA was measured by real-time quantitative reverse transcription–polymerase chain reaction (qRT-PCR) analysis. C, Interleukin-6 (IL-6) protein in culture supernatants obtained at 48 hours and 72 hours in the presence (added at 48 hours) or absence of infliximab (IFX) was measured by ELISA. D, On day 3, etanercept, infliximab, anakinra, anti-gp130, or CP690550 was added, and IL-6 mRNA was measured on day 4 by qRT-PCR. Broken line represents the value for TNFα-stimulated control FLS (set at 100%). Values are the mean ± SEM. ∗ = P < 0.05; ∗∗∗ = P < 0.001; ∗∗∗∗ = P < 0.0001 versus control, by two-way analysis of variance followed by the Bonferroni test or by Student's paired 2-tailed t-test. NS = not significant.

Next, we investigated whether the late phase of gene expression requires the ongoing presence of the low residual amounts of TNFα, and we tested the role of other cytokines that may act in our system in an autocrine or paracrine manner. Strikingly, elevated expression of IL-6 mRNA was abrogated when the TNFα blocker etanercept was added 2 or 3 days after the initial stimulation with TNFα (Figure 3B). Consistent with this, we found that addition of the TNFα blocker infliximab on day 2 prevented further accumulation of IL-6 protein (Figure 3C). These results suggest that the late phase of inflammatory response in FLS was dependent on ongoing signaling by residual low amounts of TNFα. Next, we investigated whether other cytokines may cooperate with the low amounts of TNFα to sustain the late phase of the inflammatory response in our system. We added several inhibitors that successfully block inflammation and cytokine production in RA clinical trials, including the anti-TNFα monoclonal antibody infliximab, the IL-1 receptor antagonist anakinra, and the JAK/STAT signaling inhibitor CP690550. In addition, we tested a gp130-blocking antibody, which inhibits the effects of IL-6 and other gp130 signaling cytokines. All inhibitors and their controls were added on day 3 after the initial TNFα stimulation, and cells were harvested the next day.

TNFα-induced IL-6 mRNA was significantly suppressed by etanercept or infliximab, but the other inhibitors had minor (CP690550) or no effect (Figure 3D). Similar results were found for mRNA for CXCL8/IL-8, CCL5/RANTES, and MMP-1 (data not shown). The findings that FLS retain a prolonged responsiveness even to low amounts of TNFα suggest that the late phase inflammatory response in FLS requires the continued presence of, and signaling by, TNFα.

In FLS, TNFα induces prolonged NF-κB signaling that is required for IL-6 transcription.

Activation of the canonical NF-κB signaling pathway typically follows cell stimulation with TNFα. In addition, activation of the NF-κB pathway has been observed in RA synovium, and the detrimental role of this pathway in synovitis has been suggested by studies in animal models (28). After TNFα stimulation, IκBα protein, which inhibits NF-κB signaling by retaining NF-κB proteins in the cytoplasm, was rapidly degraded and then returned to baseline levels within 2–3 hours (Figure 4A, top). These results are consistent with rapid activation of NF-κB signaling, followed by termination through the resynthesis of IκBα, as has been previously established in many cell types (29). Strikingly, in FLS, there was a second phase of IκBα protein degradation beginning at 6 hours after TNFα stimulation (Figure 4A, top) that was maintained throughout the timeframe of our experiments (Figure 4A, bottom). We used p38, whose expression did not change, as a loading control, as previously described (23).

Figure 4.

Dependence on sustained NF-κB signaling for prolonged induction of interleukin-6 (IL-6) by tumor necrosis factor α (TNFα). Fibroblast-like synoviocytes (FLS) were cultured in the presence or absence of TNFα (10 ng/ml). A and B, Levels of IκBα (A) and nuclear p65 (B) proteins were measured at the indicated times by Western blotting (left), and the bands were quantified by densitometry (right). Akt was included as quality control of the nuclear extract, and lamin B1 was used as a loading control. IFX = infliximab; iIKK II = IKK inhibitor II. C, On day 3 of culture, DMSO, BAY11-7082, IKK inhibitors II and XII, or MG132 was added for 3 hours, and IL-6 primary transcripts (PTs) were measured by real-time quantitative reverse transcription–polymerase chain reaction analysis. Values are the mean ± SEM. ∗∗ = P < 0.01 versus DMSO plus TNFα, by Student's paired 2-tailed t-test. D, Levels of p55 or p75 protein expression on the cell surface were measured by fluorescence-activated cell sorting. Results are representative of 3 independent experiments.

In addition, TNFα stimulation resulted in an expected rapid (within 1–3 hours) increase in NF-κB p65 nuclear localization (Figure 4B, top). Consistent with the sustained pattern of IκBα degradation, p65 nuclear localization in TNFα-stimulated FLS was sustained throughout the timeframe of our experiments (96 hours) (Figure 4B). Sustained p65 nuclear localization was dependent on continuous TNFα signaling, as it was abrogated by the addition of infliximab (on day 3) or an IKK inhibitor (on day 4, for 3 hours) (Figure 4B, bottom, third through sixth lanes). Next, several inhibitors of IKKs (BAY11-7082; IKK inhibitors II and XII) or the proteasome inhibitor MG132 was added 3 days after TNFα stimulation to acutely terminate NF-κB signaling. Interruption of TNFα-mediated NF-κB signaling for 3 hours resulted in a substantial decrease in IL-6 primary transcripts (Figure 4C).

The above results suggest that the late phase of TNFα-induced IL-6 expression in FLS is dependent on sustained NF-κB signaling. This could be explained by sensitization of FLS to TNFα by up-regulation of TNF receptors or proximal signaling components, or by ineffective termination of signaling by feedback inhibitory homeostatic mechanisms. The potential sensitization of FLS to TNFα in our culture system was investigated using flow cytometry to measure cell surface expression of the two TNF Receptors, p55 and p75. There was a significant down-regulation of p55 receptor at 72 hours of TNFα-stimulation in FLS, and no up-regulation of p75 was observed upon early or prolonged TNFα stimulation (Figure 4D and data available upon request from the corresponding author). There was also no increase in the TNF signaling components TRADD, TNF receptor–associated factor 2, or RIP-1, as measured by immunoblotting (data available upon request from the corresponding author). These results suggest that sensitization of TNF receptors and upstream signaling components did not occur, and led us to investigate feedback regulatory mechanisms instead.

Feedback regulatory mechanisms that control inflammatory responses in macrophages are not effectively induced in FLS.

Our results suggest that TNFα-induced inflammatory responses are not effectively terminated in FLS by homeostatic mechanisms that deactivate inflammatory signaling in other cell types, such as macrophages. The inflammatory response in macrophages is rapidly down-regulated by inhibitory molecules, including IL-1 receptor–associated kinase M (IRAK-M), A20-binding inhibitor of NF-κB activation 3 (ABIN-3), A20, suppressor of cytokine signaling 3 (SOCS-3), SHIP-1, and activating transcription factor 3 (ATF-3), which inhibit inflammatory signals or repress transcription of inflammatory genes, such as IL-6 (22). Interestingly, we observed that the TNFα-induced levels of ABIN-3, IRAK-M, SOCS-3, and ATF-3 expression were considerably higher in macrophages compared to FLS (Figure 5). In addition, baseline IRAK-M and ATF-3 mRNA levels were substantially higher in macrophages than in FLS (Figures 5B and D). These data suggest that relatively low expression of ABIN-3, IRAK-M, suppressor of cytokine signaling 3 (SOCS-3), and ATF-3 in FLS may enable the sustained inflammatory response observed in TNFα-stimulated FLS.

Figure 5.

Ineffective induction in fibroblast-like synoviocytes (FLS) of regulatory mechanisms that control tumor necrosis factor α (TNFα)–induced signaling and gene transcription in macrophages (Mϕ). FLS (n = 5) and macrophages (n = 6) were cultured in the presence or absence of TNFα (10 ng/ml) in time-course experiments. Levels of mRNA for A20-binding inhibitor of NF-κB activation 3 (ABIN-3) (A), interleukin-1 receptor–associated kinase M (IRAK-M) (B), suppressor of cytokine signaling 3 (SOCS-3) (C), and activating transcription factor 3 (ATF-3) (D) were measured by quantitative polymerase chain reaction analysis, and the values were normalized relative to mRNA for GAPDH, which was comparable in FLS and macrophages. Values are the mean ± SEM. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001; ∗∗∗∗ = P < 0.0001 versus control, by two-way analysis of variance followed by the Bonferroni test. ND = not done.

TNFα induces a sustained increase in chromatin accessibility, histone acetylation, and NF-κB p65 and Pol II recruitment at the IL-6 promoter in human FLS.

In macrophages, closing of chromatin accessibility at the IL-6 promoter after prolonged stimulation with TLR ligands or TNFα is an important mechanism for abrogating the expression of IL-6 (23, 30, 31). Such epigenetic mechanisms work in tandem with feedback inhibition of signaling to ensure postinduction suppression of IL-6 production (22). We used REAs to measure changes in chromatin accessibility at the IL-6 promoter after TNFα stimulation of FLS. An accessible, or “open,” chromatin state is vulnerable to nuclease cleavage, which results in DNA cleavage products of smaller sizes, while inaccessible, or “closed,” chromatin is more resistant to cleavage. We measured accessibility at the Ssp I restriction endonuclease site in the IL-6 promoter.

Cleavage at the Ssp I site upstream of the transcription start site was substantially increased in FLS by TNFα stimulation during the early phase (3 hours), as measured by decreased amounts of uncut chromatin (Figure 6A, left top panel, first and second lanes), with a concomitant increase in amounts of Ssp I cut products (Figure 6A, left bottom panel, first and second lanes). Remarkably, increased cleavage by Ssp I was detected during the late phase (days 1, 2, and 4 after TNFα stimulation) (Figure 6A, left bottom panel, third through sixth lanes and center bottom panel, first and second lanes). In stark contrast, macrophages stimulated with TNFα for 4 days displayed no detectable cleavage by Ssp I (Figure 6A, center panels, third and fourth lanes); as a positive control, a 3-hour stimulation with lipopolysaccharide increased Ssp I cleavage, as expected (Figure 6A, right). These results correlate with the sustained pattern of IL-6 mRNA expression in TNFα-stimulated FLS and the transient pattern observed in macrophages, and suggest that a sustained increase in chromatin accessibility contributes to sustained IL-6 transcription.

Figure 6.

Tumor necrosis factor α (TNFα)–induced sustained chromatin remodeling, histone acetylation, and p65 and RNA polymerase II (Pol II) recruitment on the interleukin-6 (IL-6) gene promoter in fibroblast-like synoviocytes (FLS). FLS and macrophages (Mϕ) were cultured in the presence or absence of TNFα (10 ng/ml) in time-course experiments. A, Nuclei from control and TNFα- or lipopolysaccharide (LPS)–treated cells were digested with Ssp I, and IL-6 promoter was determined by restriction enzyme accessibility assay. Purified genomic DNA was analyzed by Southern blotting. Results are representative of 2 independent experiments. B–D, At 72 hours after TNFα stimulation, FLS were harvested, and chromatin immunoprecipitation analysis was performed with anti–histone H4, anti–acetylated histone H4 (B), anti-p65 (C), or anti–Pol II (D) antibodies. The enrichment of immunoprecipitated proteins is shown as the percentage of input, and levels of acetylated histone H4 were normalized to those of total histone H4 at the IL-6 and hemoglobin B (HBB; negative control) levels. Values are the mean ± SEM. Results are representative of 3 independent experiments.

Chromatin accessibility can be increased by depletion of histones/nucleosomes, which exposes DNA, or by histone acetylation, which weakens DNA–histone interactions. We thus examined the effects of TNFα stimulation on histone occupancy and acetylation at the IL-6 locus. Stimulation with TNFα for 72 hours resulted in decreased histone H4 levels at the IL-6 promoter (Figure 6B, left), and the remaining histones showed increased acetylation, as measured by the ratio of acetylated histone H4 to total histone H4 (Figure 6B, right). Histone H4 acetylation was not detected at the hemoglobin B promoter, which served as negative control. Induction of a more accessible chromatin state would facilitate binding of transcription factors and Pol II; indeed, we observed a striking increase in NF-κB p65 (Figure 6C) and Pol II (Figure 6D) occupancy at the IL-6 promoter 72 hours after TNFα stimulation. These data further corroborate that TNFα induces sustained NF-κB signaling and IL-6 transcription. Overall, our results suggest that TNFα-induced sustained signaling by the NF-κB pathway, coupled with prolonged chromatin accessibility at the IL-6 promoter, result in prolonged IL-6 transcription.


Our study reveals fundamental differences between FLS and macrophages in terms of the kinetics, quality, and quantity of their TNFα-induced inflammatory program. Whereas macrophages display a transient inflammatory response, TNFα-stimulated FLS exhibit a prolonged inflammatory response. Strikingly, we found in FLS very low expression of the negative regulators of inflammatory responses ABIN-3, IRAK-M, SOCS-3, and ATF-3, in combination with sustained activation of the canonical NF-κB pathway. In addition, TNFα induced in FLS sustained histone modifications and increased chromatin accessibility at the IL-6 promoter. This removal of a chromatin barrier augments cell responsiveness to ongoing TNFα signaling by facilitating recruitment of NF-κB and Pol II to the IL-6 promoter. These findings provide a molecular explanation for the well-known capacity of FLS to produce, upon stimulation, large quantities of IL-6 and suggest that sustained expression of inflammatory genes by FLS contributes to unremitting synovial inflammation in RA.

From a teleological point of view, the differences we detected between FLS and macrophages probably reflect their distinct functions. The main function of FLS is to create the structural scaffold of synovium and to produce extracellular matrix, lubricin, and cartilage nutrients (16). FLS reside in a sterile internal environment and are not normally exposed to environmental antigens or microbes and, thus, under physiologic conditions, do not require homeostatic mechanisms used by immune cells, such as macrophages, that are often exposed to microbes and their products. On the other hand, macrophages need to limit their responses to inflammatory factors in order to prevent local and systemic toxicity caused by high levels of mediators of inflammation. They possess remarkable plasticity and a capacity for a wide spectrum of tasks, varying from rapid inflammatory responses to homeostatic functions (32).

During inflammation, macrophages normally exhibit a biphasic response, initially promoting acute inflammation, while later dampening inflammation and triggering tissue repair (22). Thus, it is important for the proper function of macrophages to rapidly adapt to the distinct requirements of defense and homeostasis. From a mechanistic perspective, one path that macrophages use to accomplish these divergent functions is to become tolerant to further inflammatory stimulation via regulatory mechanisms that either terminate the input of inflammatory signaling or repress inflammatory genes by epigenetic modifications (22). On the other hand, our data show that FLS do not become tolerant to continuous inflammatory stimulation, but instead, display sustained responsiveness, even to low amounts of TNFα. The low expression of ABIN-3, IRAK-M, SOCS-3, and ATF-3 in FLS support the hypothesis that in FLS there are insufficient “brakes” for turning off the response to TNFα, whereas in macrophages, the inflammatory response is tightly controlled via cooperation of a series of signaling and epigenetic “brakes.”

A major feedback loop that limits inflammatory cytokine production in macrophages is induction of IL-10, which in turn, activates STAT-3 (33). The STAT-3–mediated feedback inhibition is not effectively induced in FLS, as these cells produce minimal IL-10, and STAT-3 appears to play an activating, rather than an inhibitory, role in FLS (34). Our results show modest suppression of IL-6 by a JAK inhibitor, further supporting the notion that STAT-3 does not suppress inflammatory cytokine production in FLS. Thus, the absence of an effective IL-10/STAT-3–mediated inhibitory axis also likely contributes to the sustained pattern of inflammatory gene expression we observed in TNFα-stimulated FLS. In contrast to IL-10, TNFα induces IFNβ in FLS, resulting in JAK/STAT signaling and expression of CCL5/RANTES and CXCL10/IFNγ-inducible protein 10 that are suppressed by the JAK inhibitor CP690550 (35).

Several signaling pathways, activated directly or indirectly, downstream of TNF receptors mediate the pleiotropic effects of TNFα on FLS. The classic NF-κB pathway has been linked primarily with the production of mediators of inflammation and is critical for FLS survival (28). Activation of the JNK pathway contributes to the induction of tissue-degrading enzymes (36), whereas the induction of phosphatidylinositol 3-kinase δ by TNFα results in the activation of Akt, which in turn, leads to increased growth and survival of FLS (37). Notably, a plethora of regulatory molecules that affect the balance and activity of these pathways in FLS have been described, modulating (amplifying, attenuating, or reprogramming) the effects of TNFα during homeostasis or synovial inflammation. In this context, it has been suggested that the low expression in RA FLS of clusterin, which inhibits NF-κB by stabilizing IκB proteins, may amplify the induction of NF-κB and IL-6 by TNFα (38). In contrast, TWEAK, a member of the TNF superfamily, may attenuate TNFα-mediated IL-6 production via the activation of RelB (39). In addition, there is induction of the autotaxin/lysophosphatidic acid axis during synovitis, which enhances the positive effects of TNFα on survival, growth, migration, and production of inflammatory cytokines, chemokines, and MMP-9 in FLS (40). Interestingly, ablation of this axis switches the effects of TNFα on FLS from prosurvival to proapoptotic (41). Another potential modulator of the effects and signaling of TNFα on FLS is the p21-activated kinase 1, which amplifies the production of MMPs via the activation of JNK (42). These various TNF-induced signaling events cooperate to induce the full pathogenic phenotype of RA FLS.

The role of IL-6 in RA, as pathogenic and treatment target, is now well established (43). Tocilizumab, which blocks the IL-6 receptor, a monoclonal antibody which neutralizes circulating IL-6, and JAK inhibitors, which block cytokine signaling, all target the IL-6 pathway (44–47). FLS are the major source of IL-6 in RA (18), and our study provides a model on how the cross-talk of synovial macrophages and FLS leads to continuous production of inflammatory mediators, such as IL-6, and contributes to chronic synovitis. The model is that, in the course of synovial inflammation, there is a continuous recruitment of activated monocytes and macrophages that produce TNFα. Our results suggest that TNFα, even in low amounts, acts on neighboring FLS in a paracrine manner, inducing prolonged NF-κB signaling and sustained chromatin remodeling at the IL-6 promoter, due to ineffective homeostatic regulatory mechanisms in FLS. The combination of these signaling and epigenetic events results in continuous transcription of IL-6, overall contributing to the maintenance of synovial inflammation.

Complete remission of synovitis is still an unmet need for most RA patients, and even in patients fulfilling the criteria for clinical response, joint destruction may proceed, which suggests that there is residual subclinical inflammation (6, 48, 49). Notably, the approved therapies for RA target primarily immune cells, and only tocilizumab targets a product mainly derived from FLS. Thus, direct targeting of FLS emerges as an attractive alternative strategy for breaking the vicious circle that leads to unresolved synovitis in RA. Therapeutic targeting of FLS by blocking cadherin 11, a molecule primarily expressed on FLS within the synovium, has been proven effective in animal models of arthritis (50). Our observations suggest that blocking sustained inflammatory signaling or altering the chromatin state of inflammatory and tissue-destructive genes in FLS represent additional strategies to suppress the detrimental functions of FLS in RA.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Kalliolias had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Lee, Qiao, Grigoriev, Chen, Park-Min, Park, Ivashkiv, Kalliolias.

Acquisition of data. Lee, Qiao, Grigoriev, Chen, Park-Min, Park, Ivashkiv, Kalliolias.

Analysis and interpretation of data. Lee, Qiao, Grigoriev, Chen, Park-Min, Park, Ivashkiv, Kalliolias.


We thank the patients and Dr. Mark Figgie (Hospital for Special Surgery) for providing the synovial tissues. We thank Laura Donlin (Hospital for Special Surgery) for critical review of the manuscript.