To clarify the function of osteoclast-like multinuclear cells differentiated from bone marrow–derived macrophages (BMMs) by a combination of tumor necrosis factor α (TNFα) and interleukin-6 (IL-6), and to investigate the molecular mechanisms underlying the differentiation.
BMMs were stimulated by TNFα and/or IL-6. The cells were then compared with conventional osteoclasts differentiated in vitro by RANKL. An in vitro pit formation assay on dentine slices and an in vivo resorption assay of calvarial bones were performed. We also evaluated the activities and expression levels of NF-κB, c-Fos, and NF-ATc1, which are essential to the differentiation of conventional osteoclasts. Small interfering RNA was used to knock down c-Fos. The effects of genetic ablation of STAT-3 and pharmacologic inhibitors of NF-AT, JAK, and ERK were also studied.
Osteoclast-like cell differentiation depended on TNFα and IL-6 and was not inhibited by osteoprotegerin. These differentiated cells were associated with both in vitro and in vivo bone resorption activity. TNFα and IL-6 had a synergistic effect on the activity and expression of c-Fos. Knockdown of c-Fos inhibited the expression of NF-ATc1 and the differentiation of osteoclast-like cells. All of these inhibitors blocked differentiation of the cells in vitro, but surprisingly, the conditional knockout of STAT-3 did not. Tofacitinib also inhibited the bone destruction caused by TNFα and IL-6 in vivo.
Our results demonstrate that a combination of the inflammatory cytokines TNFα and IL-6 can induce osteoclast-like cells that have in vitro and in vivo bone-resorptive activity.
Rheumatoid arthritis (RA) is a debilitating disease that causes chronic inflammation of the joints. The periarticular bone destruction resulting from this inflammation can seriously impair the quality of life of patients. Osteoclasts are multinucleated cells of monocyte/macrophage lineage and are believed to play a major role in arthritic bone destruction. The osteoclast differentiation factor (ODF) RANKL is a member of the tumor necrosis factor (TNF) family of cytokines. Although both RANKL and TNFα stimulate similar signaling pathways, the difficulty of inducing osteoclastogenesis by TNFα alone has been reported ([1, 2]). Nevertheless, TNFα must play an important role in the bone destruction observed in RA, because TNF blockers have been demonstrated to prevent arthritic bone destruction, particularly when administered in combination with methotrexate (MTX), which is the gold standard treatment of RA ().
Lam et al previously reported that TNFα can promote the differentiation of osteoclasts in the presence of a small amount (i.e., “a permissive level”) of RANKL (). The order in which osteoclast precursors encounter cytokines may also be important. For instance, Ochi and colleagues reported inhibition of osteoclastogenesis when TNFα was added to cultures of precursor cells before or at the same time as the addition of RANKL. In contrast, however, osteoclastogenesis was promoted when TNFα was added after the addition of RANKL (). Another possible role that TNFα plays is to differentiate osteoclasts in an indirect manner, via the effect of the induction of RANKL on osteoclastogenesis-supporting cells such as osteoblasts and synoviocytes. It is notable, however, that the role of MTX in this context remains unclear.
In this study, working on the hypothesis that TNFα in combination with one or more factors would induce the differentiation of bone-resorbing cells, we observed that TNFα in combination with interleukin-6 (IL-6) induced osteoclast-like multinucleated cells. Next, we investigated whether these cells had bone-resorbing activity in vitro and in vivo. We also analyzed the molecular mechanisms of multinucleated cell differentiation in comparison with RANKL-mediated osteoclastogenesis.
MATERIALS AND METHODS
C57BL/6 mice (ages 6–10 weeks) were purchased from Charles River Japan. LysM-cre/Stat3flox/flox mice with a C57BL/6 background have been described previously (). All mice were maintained under specific pathogen–free conditions, and all animal experiments were carried out with the approval of the Animal Study Committee of Saitama Medical University and conformed to relevant guidelines and laws.
In vitro assays for osteoclast differentiation and function
The assays used for in vitro osteoclast differentiation have been described in detail previously ([5, 6]). Briefly, bone marrow cells (BMCs) were cultured in α-minimum essential medium (Gibco Invitrogen) supplemented with 10% fetal bovine serum, 50 units/ml penicillin/streptomycin (Gibco Invitrogen), and 10 ng/ml macrophage colony-stimulating factor (M-CSF; R&D Systems). Cells were cultured at concentrations of 2 × 105 cells/well in 24-well plates and 5 × 106 cells/dish in 6-cm dishes for 2 days.
Following the initial 2-day culture period, BMCs were then used as bone marrow–derived macrophages (BMMs), and culture continued using culture medium supplemented with RANKL (PeproTech), TNFα (PeproTech), and/or IL-6 (R&D Systems). The medium was replenished with fresh medium every other day until various assays were performed. Tartrate-resistant acid phosphatase (TRAP) was assayed with a TRAP Staining Kit (Primary Cell) according to the manufacturer's instructions. The NF-AT inhibitor tacrolimus (FK506; Sigma-Aldrich), osteoprotegerin (OPG; R&D Systems), an anti-mouse IL-1β antibody (eBioscience), the pan JAK inhibitor tofacitinib (CP-690550; Selleckchem), and the MEK inhibitors PD98059 and U0126 (Sigma-Aldrich) were added at the same time as RANKL or the proinflammatory cytokines. For pit formation assay, BMMs cultured on dentine slices (Wako) were cultured for 14 days in the presence of the cytokines. After removal of the cells, the resorption pits were examined by electron microscopy (Hitachi).
Analysis of cytokine-induced bone destruction in vivo
In vivo bone resorption assays have been described in detail previously ([7, 8]). Briefly, phosphate buffered saline (PBS), IL-6, TNFα, or TNFα plus IL-6 (all at 0.375 μg/day) was injected into the supracalvariae of mice every day for 5 days (i.e., from day 0 to day 4). On day 5, the mice were killed, and decalcified paraffin sections of the calvarial bones were analyzed. Parameters such as the number of TRAP-positive cells per bone perimeter and the bone resorption area (eroded surface per bone surface) were determined. In one experiment, either tofacitinib (15 mg/kg body weight) or normal saline was administered intraperitoneally once daily beginning 2 days prior to the injection of PBS or TNFα plus IL-6 (i.e., from day −2 to day 4).
Immunofluorescence staining and Western blot analysis
Alexa Fluor 546–labeled phalloidin (Invitrogen), a mouse anti–NF-ATc1 monoclonal antibody (7A6; Santa Cruz Biotechnology), and an Alexa Fluor 488–labeled anti-mouse IgG antibody (Invitrogen) were used for immunofluorescence staining. Cells were observed by fluorescence microscopy (Olympus). For Western blot analysis, a rabbit anti–c-Fos monoclonal antibody, a rabbit anti–pSTAT-3 (Tyr705) monoclonal antibody (both from Cell Signaling), a mouse anti–NF-ATc1 monoclonal antibody (7A6), and a mouse anti–β-actin antibody (Sigma-Aldrich) were used. The protein level was determined using a Bio-Rad calibrated densitometer.
Assay for the activity of transcription factors
For the assessment of NF-κB p65/p52 activities, the BMMs were stimulated with the proinflammatory cytokines for 12 hours, restimulated for another 30 minutes, and then harvested. To determine c-Fos activity, the cells were stimulated for 24 hours and restimulated for 30 minutes. To assess NF-ATc1 activity, the cells were stimulated for 48 hours and restimulated for 24 hours. A Nuclear Extract Kit and TransAM Transcription Factor Assay Kits (Active Motif) were used according to the manufacturer's instructions.
BMMs were transfected using a Lipofectamine RNAiMAX Reagent (Invitrogen) according to the manufacturer's protocol. Briefly, cells were incubated with 100 nM scrambled small interfering RNA (siRNA; control) or c-Fos siRNA (Ambion), using 500 μl of a transfection solution. After 6 hours of transfection, the medium was changed, and the cells were stimulated with cytokines.
Values are presented as the mean ± SEM. The Mann-Whitney U test was used for comparisons between 2 groups. P values less than 0.05 were considered significant.
Differentiation of TRAP-positive multinucleated cells by the combination of TNFα and IL-6.
Osteoclasts can be differentiated in vitro by culturing mouse BMCs with M-CSF and subsequently adding RANKL and M-CSF (Figures 1A and B). TRAP expression is a characteristic of osteoclasts (). Indeed, TRAP staining revealed that these cells were positive for TRAP. When TNFα instead of RANKL was used in this culture system, multinucleated cells were barely detected, although TRAP-positive cells were abundantly present, indicating that TNFα cannot replace RANKL in the differentiation of osteoclasts (Figure 1B).
Another proinflammatory cytokine, IL-6, is also implicated in the pathogenesis of RA (). IL-6 has been shown to trigger osteoclast formation () and induce bone resorption (). However, because RANKL has been identified as the ODF, the role of IL-6 is considered to be indirect via the induction of RANKL on osteoblasts/stromal cells (). As expected, IL-6 alone induced only a scarce number of TRAP-positive cells and almost no multinucleated cells. Interestingly, however, the combination of TNFα and IL-6 induced the formation of TRAP-positive multinucleated cells (Figure 1B). The number of TRAP-positive multinucleated cells increased with increasing concentrations of either cytokine, revealing the dose-dependent nature of the response (Figure 1C).
We next added OPG, a decoy receptor for RANKL (), to the culture system in order to determine whether RANKL induced by IL-6 and/or by TNFα on stromal cells (which might be contained in BMCs) was involved in the differentiation of the TRAP-positive multinucleated cells. As shown in Figure 1D, OPG did not inhibit the differentiation of the TRAP-positive multinucleated cells, whereas OPG did inhibit osteoclastogenesis induced by RANKL. The combination of TNFα and IL-1β has been reported to induce osteoclastogenesis (). In our system, we demonstrated that IL-1β was not responsible for cell fusion, because the addition of an anti–IL-1β antibody did not affect the differentiation of TRAP-positive multinucleated cells (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html).
In vitro and in vivo bone-resorbing activity of TRAP-positive multinucleated cells induced by TNFα and IL-6.
The pit formation assay revealed that TRAP-positive multinucleated cells had the capacity to carry out bone matrix resorption in vitro (Figure 2A). Moreover, the formation of actin rings is critically important for bone resorption (), and indeed, these multinucleated cells formed actin rings similar to those observed in conventional osteoclasts induced by RANKL (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html).
The in vivo effects of the cytokines were assessed by application into the subcutaneous tissue overlying the calvariae of mice. Administration of TNFα alone induced some TRAP-positive cells and slight bone destruction, but when IL-6 was also administered, the degree of bone destruction was substantially increased (Figure 2B). The administration of IL-6 alone did not induce bone destruction or TRAP positivity. Interestingly, compared with administration of TNFα plus IL-6, administration of RANKL resulted in fewer numbers of TRAP-positive cells and a smaller bone resorption area (eroded surface/bone surface) (Figure 2C). It was therefore clear that this cytokine combination induced TRAP-positive cells with both in vitro and in vivo bone-resorption activity.
Analysis of activated intracellular signaling molecules in osteoclast-like cells
We next attempted to elucidate the mechanisms underlying differentiation of osteoclast-like cells induced by TNFα and IL-6. The master regulatory transcription factor for osteoclast differentiation is believed to be NF-ATc1 (). During the course of the differentiation of osteoclast-like cells, the NF-ATc1 protein level increased, and NF-ATc1 was translocated into the nucleus (Figures 3A and B). Accordingly, the transcriptional activity of NF-ATc1, as quantified by an enzyme-linked immunosorbent assay–based method, was elevated (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html). Translocation of NF-ATc1 into the nucleus depends on its dephosphorylation by calcineurin. As expected, the calcineurin inhibitor tacrolimus (FK506) inhibited the differentiation of osteoclast-like cells (Figure 3C). Thus, we demonstrated that NF-AT activity is necessary for the differentiation of osteoclast-like cells, although the elevation in the NF-ATc1 level was not as marked as that observed during the differentiation of conventional osteoclasts ().
The transcriptional activities of NF-κB and activator protein 1 (AP-1) are also considered to be important for the induction of NF-ATc1 during osteoclast differentiation. Adaptor proteins known as TNF receptor (TNFR)–associated factors (TRAFs) are activated downstream of RANK or TNFRs. TRAFs in turn activate NF-κB and MAPKs, and MAPKs then activate AP-1. Mice doubly deficient in NF-κB1 and NF-κB2 were rendered osteopetrotic due to the lack of osteoclasts ([18, 19]). Mice deficient in c-Fos, which is a critical component of AP-1, had a similarly osteopetrotic phenotype ([20, 21]).
First, we measured the activation levels of both the canonical and noncanonical NF-κB pathways and observed little difference between stimulation with TNFα alone and stimulation with TNFα plus IL-6 (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html). Next, we investigated the expression and activity of c-Fos at the protein level. Previous studies () have detected c-Fos as several distinct bands, suggesting extensive posttranslational modification. In our study, we observed that administration of TNFα plus IL-6 induced a higher level of c-Fos than the level induced by TNFα alone, and that the slower migrating band was clearly dominant (Figure 3D). The activity of c-Fos also showed a similar pattern (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html). We attempted to knock down Fos by introducing siRNA into BMCs and observed that siRNA against Fos significantly down-regulated the expression level of not only Fos but also Nfatc1 relative to the scrambled siRNA control (Figure 3E). Consistent with this result, the differentiation of TRAP-positive multinucleated cells by TNFα plus IL-6 was also significantly inhibited, demonstrating the importance of c-Fos in the differentiation of these cells.
Dependence of osteoclast-like cells, but not conventional osteoclasts, on JAK
Intracellular signaling by IL-6 is largely transmitted via the JAK/STAT pathway, in which STAT-3 plays an important role (). Thus, we sought to determine whether the addition of the pan JAK inhibitor tofacitinib () would inhibit the differentiation of osteoclast-like cells. Tofacitinib was recently shown to be effective in the treatment of RA (). In fact, we demonstrated that the in vitro addition of 10–100 nM tofacitinib inhibited the multinucleation of precursor cells (Figure 4A). In this instance, TRAP positivity was not affected. Consistent with a previous report (), the same concentrations of tofacitinib did not inhibit RANKL-induced osteoclastogenesis (data not shown).
Our next step was to use LysM-cre/Stat3flox/flox mice in which Stat3 is knocked out in macrophages (). Consistent with the very high deletion efficiency (∼90%) achieved in the mature macrophages of the LysM-cre mice () and the dominant-negative effect of the truncated STAT-3 protein (), phosphorylation of STAT-3 (Tyr705) after IL-6 stimulation was substantially reduced in the BMMs derived from the LysM-cre/Stat3flox/flox mice compared with those derived from the control Stat3flox/flox mice (additional information is available at http://www.saitama-med.ac.jp/uinfo/riumachi/supplementary_figure.html). As expected, BMMs from the conditional knockout mice and those from control mice were differentiated into conventional osteoclasts by RANKL at similar levels. Unexpectedly, we observed little difference in the ability of the 2 types of BMMs to differentiate into TRAP-positive multinucleated cells following stimulation with TNFα plus IL-6 (Figure 4B), indicating that their differentiation is not dependent on STAT-3.
This result focused our attention on another important IL-6 signaling pathway. Activation of the MAPK pathway has also been reported to depend on JAK (). In particular, ERKs, one subgroup of MAPKs, have been implicated in the activation and stabilization of c-Fos in IL-6–mediated signaling (). We applied the MEK-specific inhibitors PD98059 and U0126 at concentrations that did not significantly affect the number of TRAP-positive multinucleated cells induced by RANKL. Overall, we observed that these inhibitors, especially U0126, significantly blocked differentiation of TRAP-positive multinucleated cells by TNFα plus IL-6 (Figure 4C). Finally, we questioned whether tofacitinib was also effective in the treatment of the bone destruction induced by TNFα plus IL-6 in vivo. This was indeed the case. Systemic administration of tofacitinib significantly ameliorated bone erosion at the outer periosteal surface (mean ± SEM 17.0 ± 1.1% in mice treated with tofacitinib versus 28.6 ± 6.6% in untreated mice), whereas it did not decrease the number of TRAP-positive cells (mean ± SEM 518 ± 55 and 266 ± 179, respectively) (n = 3 independent experiments) (Figure 4D and results not shown).
In this study, we demonstrated that the combination of TNFα and IL-6 differentiates TRAP-positive multinucleated cells, which resemble osteoclasts. Indeed, these osteoclast-like cells have the capacity to absorb bone matrix in vitro, and the same combination of proinflammatory cytokines induced in vivo erosion of the calvarial bones. This finding is surprising, because Duplomb et al reported that IL-6 inhibited osteoclast differentiation in vitro (). Although proinflammatory cytokines are believed to induce RANKL on stromal cells, the fact that the addition of OPG did not inhibit the differentiation of osteoclast-like cells indicates that TNFα and IL-6 do not differentiate the cells via the induction of RANKL. We cannot rule out the possibility, however, that a trace amount of RANKL that escaped blocking by OPG was necessary for their differentiation. The apparent discrepancy between our results and those of the previous study () may arise because 1) the effect of IL-6 in the study by Duplomb et al was evaluated in a culture system with exogenous addition of RANKL, and 2) TNFα was not added to the system used in that study. Similarly, the combination of IL-1β and TNFα was reported to induce the differentiation of osteoclasts in a RANKL-independent manner (). However, in our system, the addition of anti–IL-1β antibody revealed that IL-1β was not involved in the differentiation of osteoclast-like cells.
We investigated the molecular mechanisms driving differentiation of the cells and observed that the addition of IL-6 has little impact on the NF-κB pathway but augments c-Fos at the protein level. The expression level and activity of NF-ATc1 are critical for the differentiation of osteoclast-like cells, and we observed both to be elevated in response to the combination of TNFα and IL-6. In addition, the conditional knockout of Stat3 did not significantly reduce the efficiency of the differentiation of TRAP-positive multinucleated cells. This result was unexpected given the important role of STAT-3 in IL-6 signal transduction. Thus, another important IL-6 signaling pathway downstream of JAK, namely the MEK/ERK pathway, is likely to play a crucial role in the induction of c-Fos (). Indeed, the osteoclast-like cells were more susceptible to MEK inhibitors than were conventional osteoclasts (see Figure 5).
MTX has been reported to inhibit IL-6 production under a variety of in vitro and in vivo conditions ([30, 31]). Thus, we speculate that the inhibition of production and/or the activity of IL-6 by MTX may be one of the mechanisms by which the combination of a TNFα blocker and MTX almost completely blocks the periarticular bone destruction that occurs in RA (). Considering the wide range of pharmacologic effects of MTX, however, it cannot be ruled out that it may also contribute to bone protection through other mechanisms.
The contribution of proinflammatory cytokines to bone resorption has been established both clinically and experimentally (). In terms of the mechanism underlying the bone damage, most of the studies have focused on the capacity of the proinflammatory cytokines to induce RANKL ([33, 34]). The aforementioned report by Kim et al () is noteworthy in that it claimed that the combination of IL-1 and TNFα induced the differentiation of osteoclasts in a RANKL-independent manner. There are also reports that TNFα alone is able to induce osteoclastogenesis independent of RANKL ([35, 36]). Interestingly, Kobayashi and colleagues reported that such osteoclasts have bone-resorbing activity only in the presence of IL-1 (). In a study by Sabokbar et al, human macrophages derived during surgery performed for aseptic loosening of hip implants were used as the osteoclast precursors. In that study, too, the addition of IL-1 further promoted osteoclast differentiation and function compared with TNFα alone (). Moreover, a substantial number of TRAP-positive multinucleated cells was observed, even in the absence of exogenous TNFα. Thus, it is likely that the macrophages derived from periprosthetic sites were already activated and thus released various proinflammatory cytokines involved in osteoclastogenesis.
Taken together, the results of these studies imply that several combinations of proinflammatory cytokines, which are likely to coexist in the sites affected by arthritis, can induce the differentiation of bone-resorbing cells. These cells may have characteristics different from those of conventional osteoclasts that are induced by RANKL. For example, conventional osteoclasts are susceptible to an anti-RANKL antibody () or OPG. However, compared with differentiation of conventional osteoclasts, the differentiation of osteoclast-like cells induced by TNFα and IL-6 is more easily inhibited by ERK inhibitors and tofacitinib, although tofacitinib did not affect TRAP positivity. In this sense, such cells might well be called “inflammatory osteoclast-like cells.”
We expect that a greater understanding of the relative contribution of inflammatory osteoclast-like cells and conventional osteoclasts to RA can be gained through close scrutiny of the effects of novel antirheumatic drugs, including tofacitinib, on bone destruction. It is hoped that as a result, new combinations of drugs that are more efficient in preventing bone damage and have fewer side effects will become available in the near future.
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. Sato 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. Yokota, Sato, Mimura.
Acquisition of data. Yokota, Sato, Miyazaki, Kitaura, Kayama, Takeda.
Analysis and interpretation of data. Yokota, Sato, Miyoshi, Araki, Akiyama, Mimura.
We are grateful to N. Murai, N. Shiraishi, N. Koga, Y. Yamada, N. Kurosawa, Y. Aizaki, and T. Ishibashi (Saitama Medical University) for technical assistance. We also thank S. Hida (Shinshu University), M. Asagiri (Kyoto University), and U. Sato (Tokyo Hitachi Hospital) for helpful discussions.