Tumor Necrosis Factor α Stimulates Osteoclast Differentiation by a Mechanism Independent of the Odf/Rankl–Rank Interaction

Osteoclasts, the multinucleated giant cells that resorb bone, develop from hemopoietic cells of the monocyte/macrophage lineage 1,2. Osteoblasts or bone marrow stromal cells are involved in osteoclastogenesis through a mechanism involving cell-to-cell contact with osteoclast progenitor cells. We previously proposed a hypothesis that osteoblasts/stromal cells express osteoclast differentiation factor (ODF) in response to several osteotropic factors in supporting osteoclast differentiation from their precursors 1,3,4. This working hypothesis was proved by the recent discovery of a new member of the TNF ligand family, [ODF 5, also called osteoprotegerin ligand (OPGL) 6/TNF-related activation-induced cytokine (TRANCE) 7/receptor activator of nuclear factor-κB ligand (RANKL) 8] as a membrane-associated factor 3,4. Osteoclast precursors that express RANK 9,10,11, a TNF receptor family member, recognize ODF/RANKL through cell-to-cell interaction with osteoblasts/stromal cells, and differentiate into osteoclasts in the presence of macrophage CSF (M-CSF, also called CSF-1) 5,6. M-CSF has been shown to be an essential factor for osteoclast formation in vivo and in vitro 12,13. Osteoclastogenesis inhibitory factor [OCIF, also called osteoprotegerin (OPG)/TNF receptor-like molecule 1] is a soluble decoy receptor for ODF/RANKL 14,15,16,17,18. Nuclear factor κB (NF-κB) and c-Jun NH₂-terminal kinase (JNK) activated by the RANK-mediated signals in osteoclast precursors appear to be involved in their differentiation into osteoclasts 7,8.

We have also shown that osteoblasts/stromal cells play an integral role in inducing osteoclast function 19,20. When purified osteoclasts formed in cocultures were further cultured on dentine slices, they rapidly underwent apoptosis and failed to form resorption pits on the slices. Osteoblasts/stromal cells added to purified osteoclasts greatly stimulated the survival and resorption activity of osteoclasts through cell-to-cell contact. Both IL-1α and soluble ODF/soluble RANKL (sODF/sRANKL) prolonged the survival of purified osteoclasts and induced their pit-forming activity 9,21. OCIF/OPG inhibited both the survival and resorption activity of osteoclasts supported by not only sODF/sRANKL 9, but also osteoblasts/stromal cells 4. Fuller et al. 22 reported that TRANCE (ODF/RANKL) is involved in osteoclast activation induced by osteoblasts treated with parathyroid hormone. It was also demonstrated that OPGL (ODF/RANKL) activated mature osteoclasts in vitro, and that administration of OPGL induced rapid increases in blood-ionized calcium levels in mice 23. These results suggest that ODF/RANKL expressed by osteoblasts/stromal cells is responsible for inducing the differentiation, survival, and activation of osteoclasts 3,4. IL-1α can be substituted for ODF/RANKL in inducing the survival and activation of osteoclasts 21.

TNF-α, which is produced by many types of cells, including monocytes and macrophages, has been proposed to be involved in bone resorption, particularly in inflammatory bone diseases. Merkel et al. 24 demonstrated that TNF-α is involved in orthopedic osteolysis induced by implant-derived particles. It was also reported that TNF-α and IL-1α play important roles in bone loss associated with osteoporosis and periodontitis 25,26. Lipopolysaccharide-stimulated osteoclastogenesis has been shown to be mediated by TNF-α 27. In addition, Pfeilschifter et al. 28 reported that human TNF-α and IL-1α stimulated formation of multinucleated cells with the characteristics of osteoclasts in human bone marrow cultures. TNF-α induces a number of biological responses via two cell-surface receptors termed TNFR1 and TNFR2 (also called TNFR p55 and TNFR p75, respectively) 29,30,31. Both TNFR1 and TNFR2 can transduce intracellular signals that stimulate the proteolytic breakdown of IκB, a cytoplasmic inhibitor of NF-κB 32,33,34,35. The activated NF-κB is then translocated into the nucleus, where it induces the transcription of several TNF-α–responsive genes. In addition, the binding of TNF-α to TNFR1 triggers programmed cell death in many cells. This process depends on the presence of the “death domain” located in the cytoplasmic region of TNFR1, which is absent in TNFR2 36,37. It is well established that mouse TNF-α binds to both mouse TNFR1 and TNFR2 with high affinity, while human TNF-α binds to mouse TNFR1 with higher affinity than to mouse TNFR2 29,30.

In the present study, the role of TNF-α in osteoclast differentiation was studied in detail. Mouse TNF-α strongly stimulated the differentiation of M-CSF–dependent mouse bone marrow macrophages (M-BMMφ) into osteoclasts in the presence of M-CSF. Human TNF-α induced only a few osteoclasts in the M-BMMφ cultures. OCIF/OPG did not inhibit TNF-α–induced osteoclast formation. We report here that the ODF/RANKL–RANK interaction is not the sole pathway that leads osteoclast progenitors to differentiation into osteoclasts in vitro; the TNF-α–TNFR1/TNFR2 interaction can be substituted for the ODF/RANKL–RANK interaction in inducing osteoclast differentiation under some pathological conditions.

Recombinant mouse sODF/sRANKL and human OCIF/OPG were prepared as described previously 5,16. Recombinant proteins of mouse and human TNF-α, and human IL-1α were obtained from R & D Systems, Inc. Human M-CSF was obtained from Yoshitomi Pharmaceutical Co. Anti–mouse TNFR1 and TNFR2 antibodies were purchased from Genzyme Diagnostics. The Fab fragment of anti–RANK polyclonal antibody and nonimmunized rabbit immunoglobulin were prepared using the immunopure Fab preparation kit (Pierce) as described previously 10. Anti–Mac-1, Moma-2, and F4/80 rat monoclonal antibodies were obtained from Serotec. (3-[¹²⁵I] iodotyrosyl¹²) Human calcitonin (specific activity, 74 TBqmmol) was obtained from Amersham International.

5–8-wk-old male ddY and C57BL/6J mice and newborn ddY mice were obtained from Sankyo Labo Service Co. C57BL/6J mice, in which the TNFR1 or TNFR2 gene had been deleted, were obtained from Jackson ImmunoResearch Laboratories, Inc. Bone marrow cells prepared from the tibia of ddY mice, TNFR1-deficient mice [TNFR1(−/−)], or TNFR2-deficient mice [TNFR2(−/−)] were suspended in αMEM containing 10% fetal bovine serum (JRH Biosciences), and cultured in 48-well plates (1.5 × 10⁵ cells/0.25 ml per well) in the presence of M-CSF (100 ng/ml). After culturing for 3 d, nonadherent cells were completely removed from the culture by pipetting. Characteristics of adherent cells were examined by staining with antibodies against Mac-1, Moma-2, and F4/80 antigens using a Vectastain ABC AP kit and Vector Red (Vector Laboratories, Inc.). Positive cells were stained red (see Fig. 1). As almost all of the adherent cells were positive for these antibodies, we called the adherent cells M-BMMφ. M-BMMφ were further cultured for 3 d with either cytokine of sODF/sRANKL, mouse TNF-α, human TNF-α, or IL-1α in the presence or absence of OCIF/OPG and/or M-CSF. Some cultures were also treated with anti–mouse TNFR1 or TNFR2 antibody, or the Fab fragment of anti–RANK antibody. Cells were then fixed and stained for tartrate-resistant acid phosphatase (TRAP) as described previously 38. Cells were also stained for alkaline phosphatase (a marker enzyme of osteoblasts) as described previously 38. Positive cells appeared as blue cells. The number of TRAP-positive cells, including mononuclear and multinucleated cells, was scored under microscopic examination. In some experiments, TRAP-positive cells containing more than three nuclei were also counted as TRAP-positive multinucleated cells.

Bone marrow cells of ddY mice (2 × 10⁵ cells/chamber) were plated on Lab-Tek 8-chamber slides (Nalge Nunc International). Cells were first cultured with M-CSF (100 ng/ml) for 3 d, and further cultured with or without cytokines for 3 d. Cultures were then treated with 0.2 nM [¹²⁵I]-human calcitonin for 1 h at room temperature. After washing twice with PBS, cells were fixed with 0.1 M cacodylate buffer, pH 7.4, containing 1% formaldehyde and 1% glutaraldehyde, stained for TRAP, and processed for autoradiography as described previously 38. Nonspecific binding of [¹²⁵I]-labeled calcitonin was assessed in the presence of an excess amount (200 nM) of unlabeled eel calcitonin (Asahi Chemical Industry).

To determine resorption activity of TRAP-positive cells, bone marrow cells of ddY mice (2 × 10⁵ cells/well) were plated on dentine slices (4 mm in diameter) that had been placed in 48-well culture plates. Bone marrow cells were first cultured with M-CSF (100 ng/ml) for 3 d. The slices, on which M-BMMφ were formed, were then well washed with αMEM to remove nonadherent cells, and further cultured with or without cytokines for an additional 3 d. Some dentine slices were subjected to TRAP staining. The remaining slices were cleaned by ultrasonication in 1 M NH₄OH to remove adherent cells and stained with Mayer's hematoxylin to visualize resorption pits as described 38. The number of TRAP-positive cells on the slices and that of resorption pits were counted under microscopic examination.

Osteoblasts obtained from the calvaria of newborn ddY mice and bone marrow cells from ddY mice were cocultured in αMEM containing 10% fetal bovine serum, 1α,25-dihydroxyvitamin D₃ [1α,25(OH)₂D₃] (10⁻⁸ M) (Wako Pure Chemical Co.), and prostaglandin E₂ (10⁻⁶ M) (Sigma Chemical Co.) in 100-mm–diameter dishes coated with Type I collagen gels (Nitta Gelatin Co.) as described previously 38. Osteoclasts formed within 6 d were recovered from the dishes by treating with 0.2% collagenase (crude osteoclasts). To purify osteoclasts, the crude osteoclast preparation was replated and cultured for 8 h. Osteoblasts were then removed by treating with PBS containing 0.001% pronase E and 0.02% EDTA 38. Some of the cultures were then stained for TRAP. The other cultures were further incubated for 36 h in the presence or absence of mouse TNF-α (20 ng/ml), human TNF-α (20 ng/ml), IL-1α (10 ng/ml), or sODF/sRANKL (100 ng/ml), and stained for TRAP. TRAP-positive multinucleated cells containing more than three nuclei were counted as living osteoclasts.

For semiquantitative reverse-transcriptase PCR (RT-PCR) analysis, total RNA was extracted using Trizol solution (GIBCO BRL) from freshly isolated bone marrow cells, M-BMMφ, and purified osteoclasts. Purified osteoclasts were obtained from cocultures of mouse bone marrow cells and calvarial osteoblasts as described previously 21. First strand cDNA was synthesized from total RNA with random primers and was subjected to PCR amplification with EX Taq polymerase (Takara Shuzo) using specific PCR primers: mouse TNFR1, 5′-GAGTGCTGACCTATAACACATTCCT-3′ (forward, nucleotides 3152–3176) and 5′-CATCTTGCCAGTTCAAAACTAACTT-3′ (reverse, nucleotides 3553–3577); mouse TNFR2, 5′-CATTCTAAGAACAATTCCATCTGCT-3′ (forward, nucleotides 246–270) and 5′-GGGTACTGGAGACAGGAGAACTAA-3′ (reverse, nucleotides 961–984); mouse RANK, 5′-ACACCTGGAATGAAGAAGATAAATG-3′ (forward, nucleotides 260–284) and 5′-AGCCACTACTACCACAGAGATGAAG-3′ (reverse, nucleotides 684–708); mouse c-Fms, 5′-AACAAGTTCTACAAACTGGTGAAGG-3′ (forward, nucleotides 2653–2677) and 5′-GAAGCCTGTAGTCTAAGCATCTGTC-3′ (reverse, nucleotides 3381–3405); mouse glyceraldehyde-3-phosphate dehydrogenase, 5′-ACCACAGTCCATGCCATCAC-3′ (forward, nucleotides 566–585) and 5′-TCCACCACCCTGTTGCTGTA-3′ (reverse, nucleotides 998–1017). The PCR products were separated by electrophoresis on a 2% agarose gel.

Mouse M-BMMφ were treated with sODF/sRANKL (100 ng/ml), mouse TNF-α (20 ng/ml), or human TNF-α (20 ng/ml) for 1 h. Nuclear extracts were then prepared from M-BMMφ as described previously 39. An NF-κB–binding oligonucleotide sequence 5′-AGCTTGGGGACTTTCCGAG-3′ was used as a radioactive DNA probe. DNA binding reaction was performed at room temperature in a volume of 30 μl, which contained the binding buffer (10 mM Tris/HCl, pH 7.5, 1 mM EDTA, 4% glycerol, 100 mM NaCl, 5 mM DTT, 100 mg/ml of BSA), 3 μg of poly(dI-dC), 10⁵ cpm of a ³²P-labeled probe, and 8 μg of nuclear proteins. After incubation for 15 min, the samples were electrophoresed on native 5% acrylamide/0.25 × Tris-borate-EDTA gels. The gels were dried and exposed to x-ray film.

Significance of differences was determined using Student's t test.

When mouse bone marrow cells were cultured with M-CSF for 3 d, adherent cells of uniform size and shape appeared on the culture plate. Almost all of the adherent cells strongly expressed antigens such as Mac-1, Moma-2, and F4/80, which are specific for macrophages (Fig. 1). In this study, these adherent cells were called M-BMMφ. Typically, 10⁴ M-BMMφ were obtained when 10⁵ bone marrow cells were cultured for 3 d in the presence of M-CSF. There were very few alkaline phosphatase–positive stromal cells detectable in the M-BMMφ preparation (data not shown).

When M-BMMφ were further cultured with sODF/sRANKL and M-CSF, TRAP-positive mononuclear and multinucleated cells were formed within 3 d (Fig. 2A and Fig. B). In the absence of M-CSF, most of the M-BMMφ died within 3 d. No TRAP-positive cells were formed when M-CSF was not added, even in the presence of sODF/sRANKL. Although varied in each experiment, 40–80% of the M-BMMφ usually differentiated into TRAP-positive cells in response to sODF/sRANKL and M-CSF. When nonadherent cells, which were recovered from bone marrow cultures treated for 3 d with M-CSF, were further cultured for 3 d with ODF/RANKL and M-CSF, only a few TRAP-positive cells were formed (data not shown). This suggests that osteoclast progenitors are present preferentially as adherent cells in M-CSF–treated bone marrow cells. Surprisingly, mouse TNF-α stimulated TRAP-positive cell formation from M-BMMφ in the presence of M-CSF in a dose-dependent manner (Fig. 2A and Fig. C). Morphology of TRAP-positive cells induced by mouse TNF-α was quite similar to that induced by sODF/sRANKL (Fig. 2 B). Repeated experiments showed that 3–5% of the TRAP-positive cells induced by either mouse TNF-α or ODF/RANKL were multinucleated cells, which contained more than three nuclei. In contrast, human TNF-α at 20 ng/ml induced a few TRAP-positive mononuclear cells in M-BMMφ cultures (Fig. 2 B). Neither human nor mouse TNF-α induced apoptosis of M-BMMφ in the presence or absence of M-CSF. No TRAP-positive cells were formed in M-BMMφ cultures treated with IL-1α at 10 ng/ml even in the presence of M-CSF (Fig. 2A and Fig. B). Addition of 1α,25(OH)₂D₃ at 10⁻⁸ M to the M-BMMφ cultures also failed to induce TRAP-positive cells, irrespective of the presence or absence of M-CSF (Fig. 2A and Fig. B). Alkaline phosphatase–positive cells were not detected in the M-BMMφ preparation after culture for 3 d with 1α,25(OH)₂D₃ and M-CSF (Fig. 2 B). Mouse TNF-α (20 ng/ml) stimulated TRAP-positive cell formation in cultures of mouse spleen cells or whole bone marrow cells in the presence of M-CSF (100 ng/ml) (data not shown). We used M-BMMφ instead of whole bone marrow cells in this experiment to avoid contamination of cells other than osteoclast progenitors.

We previously reported that osteoclasts formed in the cocultures expressed mRNAs of TNFR1, TNFR2, and RANK 9. Using a semiquantitative RT-PCR, it was shown that not only purified osteoclasts but also M-BMMφ expressed these mRNAs (Fig. 3 A). Both M-BMMφ and osteoclasts expressed c-Fms mRNA as well. Freshly isolated bone marrow cells expressed mRNAs of TNFR1, TNFR2, RANK, and c-Fms only weakly, when compared with M-BMMφ and osteoclasts (Fig. 3 A). An electrophoretic mobility shift assay revealed that mouse TNF-α, human TNF-α, and sODF/sRANKL all activated NF-κB in M-BMMφ (Fig. 3 B).

TRAP-positive cell formation induced by sODF/sRANKL was inhibited completely by simultaneously adding OCIF/OPG (Fig. 4 A, top). In contrast, TRAP-positive cell formation was similarly induced by mouse TNF-α irrespective of the presence or absence of OCIF/OPG (Fig. 4 A, bottom). Although these results suggest that TNF-α induces osteoclast differentiation through TNFR1/TNFR2, there were still two possibilities that TNF-α directly binds to RANK or TNF-α increases endogenous production of some factor(s) that can interact with RANK. To exclude these possibilities, the Fab fragment of polyclonal antibodies against the extracellular domain of RANK was used in the next experiment. We previously reported that the anti–RANK antibody induced osteoclastogenesis in the presence of M-CSF, probably through clustering of RANK 10. In contrast, the Fab fragment of the antibody blocked the binding of ODF/RANKL to RANK, causing the inhibition of the sODF/sRANK-mediated osteoclastogenesis 10. The Fab fragment of the anti–RANK antibody inhibited TRAP-positive cell formation induced by sODF/sRANK (Fig. 4 B). In contrast, the Fab fragment of the anti–RANK antibody showed no inhibitory effect on mouse TNF-α–induced TRAP-positive cell formation in M-BMMφ cultures. These results indicate that TNF-α stimulates TRAP-positive cell formation by a mechanism independent of the ODF/RANKL–RANK interaction.

We next examined effects of blocking antibodies against TNFR1 or TNFR2 on TRAP-positive cell formation in M-BMMφ cultures treated with mouse TNF-α or sODF/sRANKL in the presence of M-CSF (Fig. 5 A). Both antibodies showed no inhibitory effect on the sODF/sRANKL-induced TRAP-positive cell formation, but each of the two antibodies strongly inhibited TRAP-positive cell formation induced by mouse TNF-α (Fig. 5 A). To further confirm the possibility that both TNFR1 and TNFR2 play important roles in TNF-α–induced TRAP-positive cell formation, M-BMMφ were prepared from TNFR1(−/−), TNFR2(−/−), and the control C57BL/6J (wild type) mice, and compared their capacity to differentiate into TRAP-positive cells in the presence of mouse TNF-α plus M-CSF or sODF/sRANKL plus M-CSF (Fig. 5 B). In the presence of sODF/sRANKL, the number of TRAP-positive cells formed from TNFR1(−/−) and TNFR2(−/−) M-BMMφ culture was similar to that from the control (wild type) M-BMMφ (Fig. 5 B). In contrast, no TRAP-positive cells were observed in TNFR1(−/−) M-BMMφ cultures treated with mouse TNF-α (Fig. 5 B). TRAP-positive cell formation induced by mouse TNF-α was markedly reduced in TNFR2(−/−) M-BMMφ cultures in comparison with that in the control (wild type) cultures. These results confirmed that both TNFR1- and TNFR2-mediated signals were important for TNF-α–induced osteoclast differentiation.

Autoradiography using [¹²⁵I]-labeled calcitonin showed that numerous grains, as a result of the binding of labeled calcitonin, accumulated on TRAP-positive cells induced by mouse TNF-α together with M-CSF even in the presence of OCIF/OPG (Fig. 6 A). In agreement with previous findings 5, numerous grains of the binding of labeled calcitonin were observed on TRAP-positive cells induced by sODF/sRANKL in the presence of M-CSF (Fig. 6 B). However, no calcitonin receptor-positive cells appeared in M-BMMφ cultures treated with sODF/sRANKL plus M-CSF in the presence of OCIF/OPG (Fig. 6 C). The binding of [¹²⁵I]-labeled calcitonin to TRAP-positive cells was inhibited by an excess amount of unlabeled calcitonin simultaneously added (data not shown), suggesting that TRAP-positive cells induced by mouse TNF-α as well as sODF/RANKL expressed calcitonin receptors.

We previously reported that IL-1α, M-CSF, and sODF/sRANKL similarly stimulated the survival of purified osteoclasts via the respective receptors 9,21. Using purified osteoclasts, we next examined whether TNF-α supports the survival of osteoclasts (Fig. 7). Most of the osteoclasts died within 36 h after removal of osteoblasts in the presence and absence of OCIF/OPG. Treatment of purified osteoclast preparations with mouse TNF-α stimulated the survival of osteoclasts. OCIF/OPG showed no inhibitory effects on the survival of osteoclasts supported by mouse TNF-α (Fig. 7). In agreement with previous finding 9, sODF/sRANKL and IL-1α stimulated the survival of purified osteoclasts. OCIF/ OPG strongly inhibited the survival of osteoclasts supported by sODF/sRANKL but not by IL-1α (Fig. 7).

We previously reported that purified osteoclasts rapidly died and did not form resorption pits on dentine slices 19,20. The pit-forming activity of purified osteoclasts was greatly recovered by adding either osteoblasts/stromal cells, IL-1α, or sODF/sRANKL 9,19,21. When M-BMMφ were prepared on dentine slices, and cultured with mouse TNF-α, IL-1α, or mouse TNF-α plus IL-1α in the presence of M-CSF and OCIF/OPG, a similar number of TRAP-positive cells appeared on dentine slices in response to mouse TNF-α irrespective of the presence and absence of IL-1α (Fig. 8A and Fig. B). However, resorption pits on the slices were detected only in the presence of both mouse TNF-α and IL-1α (Fig. 8A and Fig. B). These results suggest that TNF-α stimulates the differentiation and survival of osteoclasts, but not the function of osteoclasts.

Since the discovery of the ODF/RANKL–RANK signal transduction, it has been believed that ODF/RANKL is the sole factor responsible for inducing osteoclast differentiation. This notion was supported by the finding that the targeted disruption of the gene encoding OPGL (ODF/RANKL) in mice developed severe osteopetrosis with complete absence of TRAP-positive cells in bone tissues 40. However, the present study clearly shows that TNF-α together with M-CSF induces TRAP-positive cell formation in M-BMMφ cultures without any help of osteoblasts/stromal cells. Alkaline phosphatase–positive cells were seldom detected in the M-BMMφ preparation even after culture for an additional 3 d. No TRAP-positive cells were formed in the cultures of the M-BMMφ preparation treated with 1α,25(OH)₂D₃ in the presence or absence of M-CSF. A recent study on vitamin D receptor knockout mice proved that the target cells of 1α,25(OH)₂D₃ in inducing osteoclast formation are osteoblasts/stromal cells, but not osteoclast progenitors 41. These results suggest that even if a small number of stromal cells were present in the M-BMMφ preparation, they could not support osteoclast formation.

Like authentic osteoclasts, TRAP-positive cells formed in response to TNF-α and M-CSF expressed calcitonin receptors. TNF-α–induced TRAP-positive cells formed resorption pits only in the presence of IL-1α. This property of TRAP-positive cells induced by TNF-α is recognized in ODF/RANKL-induced osteoclasts as well 19,21. M-CSF was essentially required for osteoclast formation induced by TNF-α as in sODF/sRANKL-induced osteoclast formation. These results indicate that TNF-α–induced TRAP-positive cells in M-BMMφ cultures satisfy major criteria for osteoclasts. OCIF/OPG and the Fab fragment of anti–RANK antibody did not inhibit the TNF-α–induced TRAP-positive and calcitonin receptor-positive osteoclast formation. Using RT-PCR, we found that M-BMMφ expressed ODF/RANKL mRNA, but its level was very low and was not upregulated by adding mouse TNF-α (data not shown). These results further confirm that ODF/RANKL is not involved in the TNF-α–induced osteoclast formation in M-BMMφ cultures. In our preliminary experiments, human TNF-α stimulated the formation of TRAP- and vitronectin receptor (23C6)–positive osteoclast-like cells in cultures of human peripheral blood mononuclear cells even in the presence of OCIF/OPG (data not shown). These results suggest that the ODF/RANKL–RANK signaling system is not the sole pathway for inducing osteoclast differentiation in vitro (Fig. 9).

M-BMMφ expressed both TNFR1 and TNFR2. Activation of NF-κB in M-BMMφ was induced not only by mouse TNF-α, but also human TNF-α. However, the potency of human TNF-α to induce TRAP-positive osteoclast formation in M-BMMφ cultures was much weaker than that of mouse TNF-α. Anti–TNFR1 antibody completely inhibited osteoclast formation induced by mouse TNF-α, but not osteoclast formation by sODF/sRANKL. M-BMMφ prepared from TNFR1(−/−) mice did not differentiate into TRAP-positive cells in response to mouse TNF-α. These results suggest that TNFR1-mediated signals are essential for TRAP-positive cell formation induced by TNF-α. Furthermore, addition of anti–TNFR2 antibody to M-BMMφ cultures significantly inhibited TRAP-positive osteoclast formation induced by mouse TNF-α. In addition, TRAP-positive osteoclast formation induced by mouse TNF-α was markedly reduced in TNFR2(−/−) M-BMMφ cultures. These results suggest that TNFR2-mediated signals play important roles in osteoclast differentiation. Collectively, both signals mediated by TNFR1 and TNFR2 appear to be required for inducing osteoclast formation at the level similar to ODF/RANKL-induced osteoclast formation (Fig. 9). Using the anti–TNFR1 and –TNFR2 antibody, it was clearly shown that only TNFR1-mediated signals led to the endotoxin-induced shock and protective response against Listeria monocytogenes infection, and both TNFR1- and TNFR2-mediated signals were required for the development of TNF-α–induced skin necrosis in mice 42. Abu-Amer et al. 27 reported that osteoclastogenesis induced by lipopolysaccharides is mediated by TNF-α via TNFR1. They also showed that human as well as mouse TNF-α stimulated the expression of c-Src, a marker protein of osteoclasts, in mouse bone marrow macrophages 27. These findings support the conclusion that TNFR1-mediated signals are essential for osteoclast differentiation.

We previously reported that purified osteoclasts rapidly underwent apoptosis and did not form resorption pits on dentine slices 19. IL-1α as well as sODF/sRANKL induced the survival and pit-forming activity of purified osteoclasts 9,21. The present study showed that the survival of purified osteoclasts was stimulated by human as well as mouse TNF-α. However, resorption pits were not formed on dentine slices in M-BMMφ cultures treated with mouse TNF-α and M-CSF. Resorption pits were observed on dentine slices in the M-BMMφ culture further treated with IL-1α. These results suggest that TNF-α is capable of stimulating the differentiation and survival of osteoclasts, but not the pit-forming activity of osteoclasts. On the other hand, ODF/RANKL supports all the processes of differentiation, survival, and activation of osteoclasts (Fig. 9).

Determination of signals involved in osteoclast differentiation is an important issue in the research on osteoclast biology. At present, two signals, NF-κB and JNK, have been proposed to be involved in osteoclast differentiation. Hsu et al. 11 reported that ODF/RANKL stimulated JNK activity of murine myeloid RAW-264.7 cells and induced differentiation of these cells into osteoclasts. Activation of NF-κB in RAW-264.7 cells was not induced by ODF/RANKL. This indicates that the signals mediated by JNK rather than NF-κB are important for ODF/RANKL-induced osteoclast differentiation in RAW-264.7 cells. In our experiments, IL-1 activated NF-κB in M-BMMφ (data not shown), but failed to induce osteoclast differentiation in M-BMMφ cultures. These results suggest that some other signals than NF-κB may play an important role in osteoclast differentiation. RAW-264.7 cells differentiated into TRAP-positive cells in response to mouse TNF-α as well as ODF/RANKL (data not shown). Possible involvement of JNK and NF-κB in osteoclastogenesis is currently under investigation in our laboratories.

Mounting evidence has demonstrated that TNF receptor family members interact with TNF receptor–associated factors (TRAFs) to modulate JNK and NF-κB activity as well as apoptosis in the target cells 43,44. TRAF2 has been shown to be an adaptor protein for TNFR1- and TNFR2-mediated signals in the target cells 43,44. Recent studies have demonstrated that the cytoplasmic tail of RANK interacts with TRAF1, TRAF2, TRAF3, TRAF5, and TRAF6 45,46,47,48. Recently, it was reported that TRAF6 knockout mice developed severe osteopetrosis 49,50. Similar numbers of TRAP-positive osteoclasts were observed in TRAF6 knockout mice and normal mice, but osteoclasts in TRAF6-deficient mice did not form ruffled borders, structures necessary for bone resorption 49. IL-1α receptors have been shown to use TRAF6 as a signal-transducing molecule in the target cells 51. We previously showed that IL-1α activated NF-κB in purified osteoclasts 39. IL-1α also induced pit-forming activity of osteoclasts placed on dentine slices 21. However, IL-1α failed to induce osteoclast differentiation in M-BMMφ cultures. These results suggest that TRAF6 is a common signaling molecule responsible for osteoclast activation induced by ODF/RANKL and IL-1α. TRAF2 and/or its related TRAFs may be involved in the differentiation of osteoclasts. Thus, ODF/RANKL–RANK interaction induces signals necessary for both differentiation and activation of osteoclasts (Fig. 9).

TNF-α stimulates osteoclastogenesis in the absence of ODF/RANKL in vitro. However, this is not consistent with the finding that the disruption of the ODF/RANKL gene in mice showed a complete lack of osteoclasts in bone 40. Recently, it was reported that TNF-α– and IL-1α–induced bone resorption in mice was inhibited by concomitant treatment with OCIF/OPG 52. Our preliminary experiments showed that TNF-α as well as IL-1α upregulated expression of ODF/RANKL mRNA in osteoblasts/stromal cells (data not shown), suggesting that a part of bone resorption induced by TNF-α and IL-1α could be inhibited by OCIF/OPG. In fact, we reported that bone resorption induced by parathyroid hormone and 1α,25(OH)₂D₃ was completely inhibited by adding OCIF/OPG in an organ culture system, but the inhibition by OCIF/OPG of IL-1α–induced bone resorption was only partial 53. Ammann et al. 54 also reported that transgenic mice insensitive to TNF due to the overexpression of soluble TNF receptors are protected against bone loss due to estrogen deficiency. These results suggest that TNF-α and IL-1α are directly involved in pathologically occurring bone resorption observed in inflammatory and metabolic bone diseases such as rheumatoid arthritis, periodontitis, and/or postmenopausal osteoporosis 24,25,26,27. Further studies will elucidate the biological significance of osteoclast differentiation and function stimulated by a combination of TNF-α and IL-1α in pathological bone resorption.

This work was supported in part by grants-in-aid (09877355, 08672088, and 07557118) and the High-Technology Research Center Project from the Ministry of Education, Science, Sport and Culture of Japan.