Expression of RANKL by stromal cells and of RANK and both NF-κB p50 and p52 by osteoclast precursors is essential for osteoclast formation. To examine further the role of RANKL, RANK, and NF-κB signaling in this process, we used NF-κB p50−/−;p52−/− double knockout (dKO) and wild-type (WT) mice. Osteoclasts formed in cocultures of WT osteoblasts with splenocytes from WT mice but not from dKO mice, a finding unchanged by addition of RANKL and macrophage colony-stimulating factor (M-CSF). NF-κB dKO splenocytes formed more colony-forming unit granulocyte macrophage (CFU-GM) colonies than WT cells, but no osteoclasts were formed from dKO CFU-GM colonies. RANKL increased the number of CFU-GM colonies twofold in WT cultures but not in dKO cultures. Fluorescence-activated cell sorting (FACS) analysis of splenocytes from NF-κB dKO mice revealed a two-to threefold increase in the percentage of CD11b (Mac-1) and RANK double-positive cells compared with WT controls. Treatment of NF-κB dKO splenocytes with interleukin (IL)-1, TNF-α, M-CSF, GM-CSF, and IL-6 plus soluble IL-6 receptor did not rescue the osteoclast defect. No increase in apoptosis was observed in cells of the osteoclast lineage in NF-κB dKO or p50−/−;p52+/− (3/4KO) mice. Thus, NF-κB p50 and p52 expression is not required for formation of RANK-expressing osteoclast progenitors but is essential for RANK-expressing osteoclast precursors to differentiate into TRAP+ osteoclasts in response to RANKL and other osteoclastogenic cytokines.
UNDERSTANDING OF the molecular mechanisms that regulate osteoclast formation has increased substantially after the recent identification and characterization of a number of new members of the TNF ligand and TNF receptor families. Expression of RANKL(1–5) by stromal cells is essential for osteoclast formation because RANKL knockout (KO) mice do not form osteoclasts.(6) Osteoprotegerin (OPG) is a member of the TNF receptor family and acts as a soluble decoy receptor for RANKL. It is produced by many cell types and regulates osteoclast formation by binding to and inactivating RANKL.(2,7) RANKL binds to its cognate receptor RANK, a member of the TNF receptor family that is expressed by osteoclast precursors in the spleen and bone marrow(2,8–11) and by dendritic cells and T cells.(1,4) RANKL/RANK interactions are not only essential for osteoclast formation, but also appear to regulate the activity of mature osteoclasts(12–14) as well as dendritic cell functions and T cell activation.(6) Anti-RANK antibody and a soluble form of the extracellular domain of RANK (RANK:Fc) completely inhibited RANKL-mediated osteoclast formation.(15,16) Furthermore, RANK KO mice do not form osteoclasts,(17,18) suggesting that RANK is the sole receptor for RANKL in osteoclast precursors. RANKL/RANK interactions lead to activation of NF-κB and of Jun kinase in purified osteoclasts,(19–21) effects that are similar to those of interleukin (IL)-1 treatment of mature osteoclasts.(22) However, it is not known if NF-κB activation is essential for the intracellular signaling in osteoclast precursors that leads to osteoclast formation after RANKL/RANK interaction or if some other signaling molecules such as Jun kinase can substitute for it.
NF-κB is a family of five transcription factors that are expressed in most cell types and was so named because it regulates κ-light-chain expression in murine B lymphocytes. It plays an essential role in immune and inflammatory responses by regulating the expression of a variety of proinflammatory cytokines and other inflammatory mediators. NF-κB transcription factors form homo- or heterodimers to activate target genes. These dimers are composed of various combinations of the structurally related proteins p50 (NF-κB1), p52 (NF-κB2), p65 (RelA), c-Rel, and RelB, the most common of which in most cell types is a p50/p65 heterodimer.(23–25) In unstimulated cells, NF-κB dimers generally are retained in the cytoplasm in an inactive form by binding to the inhibitory proteins IκBα and IκBβ.(26,27) In response to extracellular signals that can lead to cell growth, differentiation, inflammatory responses, apoptosis, and neoplastic transformation, IκBα and β are phosphorylated by homo- or heterodimers of IκB kinase α and β, ubiquitinated and then degraded by nonspecific proteases in the 26S proteasome.(28) The activated NF-κB factors translocate to the nucleus where they attach to κB binding sites in the promoter region of a large number of genes, including IL-1, IL-6, TNF-α, and granulocyte macrophage colony-stimulating factor (GM-CSF) to induce their transcription.
Several lines of in vitro evidence have shown that NF-κB may be involved in the regulation of osteoclast function. For example, the nonspecific NF-κB inhibitor pyrrolidine dithiocarbamate (PDTC) dose-dependently inhibited osteoclast activity in an in vitro bone resorption assay.(29) NF-κB DNA binding activity in osteoclasts can be induced by 1,25-dihydroxyvitamin D3 [1,25(OH)2D3], prostaglandin E2, and IL-1,(30) and antisense oligonucleotides to the p50 or the p65 subunits of NF-κB block IL-1-mediated survival of mature osteoclasts.(22) NF-κB inhibitors induce apoptosis of isolated mature osteoclasts and inhibit bone resorption.(31) Thus, studies to date indicate that NF-κB factors are expressed in osteoclasts and that they appear to be involved in the regulation of a variety of osteoclast functions.
An essential role of NF-κB for the formation of osteoclasts during embryonic development was discovered following the generation of double-KO (dKO) mice lacking NF-κB p50 and p52. These mice develop severe osteopetrosis because they form almost no osteoclasts or TRAP+ osteoclast precursors in vivo because of an autonomous defect in the osteoclast lineage.(32,33) Because NF-κB has diverse functions including regulation of cell proliferation and survival(34,35) and of the expression of the cytokines, IL-1, IL-6, and TNF-α, that also can up-regulate NF-κB expression,(2,36) the absence of osteoclasts in the NF-κB dKO mice could result from a number of possible defects in cells in the osteoclast lineage. Using both in vitro and in vivo approaches, we show that expression of NF-κB p50 and p52 is required for RANKL/RANK, IL-1, IL-6, and TNF-α-mediated osteoclast formation, but not for osteoclast precursor survival. In addition, we found that both colony-forming unit granulocyte macrophage (CFU-GM) colony formation and RANK surface expression by splenocytes are increased in NF-κB dKO mice, implying that blockage in osteoclastogenesis in NF-κB dKO mice is at or beyond the stage when osteoclast progenitors start to express RANK on their surface.
MATERIALS AND METHODS
Recombinant IL-6, IL-6 soluble receptor, TNF-α, macrophage colony-stimulating factor (M-CSF), and GM-CSF were purchased from R & D Systems Inc. (Minneapolis, MN, USA). PDTC and N-tosyl-phenylalanine chloromethyl ketone (TPCK) were obtained from Sigma Chemical Co. (St. Louis, MO, USA). MG-132 was purchased from BIOMOL Research Laboratories, Inc. (Plymouth Meeting, PA, USA). Antibodies phycoerythrin (PE)-conjugated CD11b (clone M1/70), biotin-conjugated CD11c (HL3), and CD16/32 were obtained from Pharmingen (San Diego, CA, USA); PE-conjugated F4/80 (clone CI:3-1) was from Serotec (Oxford, England). 1,25-(OH)2D3 was a generous gift from Dr. Milan Uskokovic (Hoffmann- LaRoche, Nutley, NJ, USA), Risedronate (2,[3-pyridinyl-]1-hydroxyethylidene-1,1-bisphosphonate) was from Dr. Rhone-Poulenc-Rorer (Collegeville, PA, USA), and RANKL was from Dr. Bill Dougall (Immunex Corp., Seattle, WA, USA).
Generation of osteoclasts
Bone marrow or spleen cells were used with or without wild-type (WT) osteoblasts to generate osteoclasts from WT mice, whereas only spleen cells from NF-κB KO mice with or without primary calvarial osteoblasts from WT mice were used to attempt to generate osteoclasts from dKO mice because their severe osteopetrosis makes it impossible to obtain sufficient osteoclast precursors from the bone marrow cavities of their long bones.
For bone marrow culture experiments, nonadherent bone marrow cells were cultured in α-modified essential medium (α-MEM; Gibco BRL Co., Grand Island, NY, USA) supplemented with 10% FCS (Hyclone Laboratories, Logan, UT, USA) in the presence of 10−8 M of 1,25(OH)2D3 as described previously.(37) For coculture experiments, primary osteoblasts (0.25 × 104 cells/well), which were isolated from calvariae of 3- to 5-day-old WT mice using a sequential collagenase/protease digestion, were cocultured with spleen cells (0.5 × 105 cells/well in 96-well plates) from mutant mice or WT littermates under the same conditions for murine bone marrow culture as described previously.(38) Cultures were maintained for 7 days at 37°C in an atmosphere of 5% CO2/air and were fed every 2 days by replacing half of the spent medium with fresh medium and 1,25(OH)2D3. Cells were then fixed and stained for TRAP activity to identify osteoclasts and their TRAP+ precursors, as described previously.(39)
CFU-GM colony assay
The in vitro colony-forming assay was performed as described previously.(40) In brief, freshly isolated spleen cells from WT or dKO mice were plated at a density of 105 cells/ml in a 3-cm dish. Cells were cultured in a methylcellulose-based medium (StemCell Technologies, Inc., Vancouver, Canada) supplemented with 20 ng/ml of GM-CSF and 30 ng/ml of M-CSF for 10 days to form CFU-GM colonies. Colonies composed of >40 cells were counted under an inverted microscope. Colonies were picked individually from the agar with a pipette tip, washed extensively with α-MEM, and cultured in the presence of RANKL (40 ng/ml) and M-CSF (30 ng/ml) for various times depending on the purpose of experiments.
Reverse-transcription polymerase chain reaction
RNA from spleens and livers of dKO mice and WT littermates was extracted using RNAzol (Tel-Test, Inc. Friendswood, TX, USA). cDNA synthesis was performed in 20 μl from 1 μg of total RNA using the Moloney murine leukemia virus reverse transcriptase, random primers, and deoxynucleoside triphosphates (dNTPs; Perkin Elmer, Foster City, CA, USA). Polymerase chain reaction (PCR) amplification was performed with gene-specific primers under the conditions that have been determined experimentally for each pair of primers. The primer sequences were c-fms, forward 5′-GAGCCTCTTGCAGGAGGTG-3′ and reversed 5′-GGTCCAATGGGCAGCTGG-3′; OPG, forward 5′-CAAGAGCAAACCTTCCAGCTG-3′ and reversed 5′-GTACATTGTGAAGCTGTG-3; and GAPDH, forward 5′-CCACATGGAGAAGGCCGGGG-3′ and reversed 5′-GACGGACACATTGG-GGG-TAG-3′. The PCR products were separated by electrophoresis on 1.2% agarose gels and were visualized by ethidium bromide staining with UV light illumination.
Fluorescence-activated cell sorting analysis
Surface immunophenotyping was performed on single cell suspensions of freshly isolated spleen cells from dKO and WT mice. After red blood cell lysis, nonspecific interactions were blocked by incubation with a combination of 50% heat-inactivated normal rat serum (Zymed, San Francisco, CA, USA) and CD16/32 (Fc receptor block), according to the manufacturer's instructions. Staining was performed as described previously,(41) using fluorescein-conjugated RANKL alone or in combination with PE-conjugated CD11b, F4/80 or biotin-conjugated CD11c. Biotin-labeled reagents were revealed with streptavidin-conjugated PE. After staining, cells were sorted on an FACScalibur instrument (Beckton Dickinson, Bedford, MA, USA) and analyzed using Cellquest software (version 3.1). Fluorescent labeling of RANKL was performed by incubation of recombinant murine RANKL (from lys158 to end of protein) with N-hydroxysucinimide-activated fluorescein (FluoReporter fluorescein isothiocyanate [FITC] protein labeling kit; Molecular Probes, Eugene, OR, USA) according to manufacturer's protocol. The activity of the labeled RANKL was confirmed by ensuring that labeled cells isolated by cell sorting differentiated to osteoclasts.
Preparation of tissue sections and TUNEL staining of apoptotic cells
NF-κB dKO mice were generated by breeding NF-κB p50+/− with NF-κB p52+/− mice. This also results in the generation of NF-κB p50−/−;p52+/− (3/4KO) mice. Although these mice express no p50 and only one allele of p52, they form osteoclasts in near normal numbers and do not have osteopetrosis. Spleens, livers, and long bones were removed from dKO, NF-κB p50−/−;p52+/− (3/4KO) and WT mice, (3-week-old) and fixed in 10% phosphate-buffered formalin. Bones were decalcified in 10% EDTA and embedded in paraffin wax, as were blocks of spleen and livers. Deparaffinized sections were digested for 15 minutes with 20 μg/ml of proteinase K (Sigma Chemical Co.) in distilled H2O, rinsed, and treated for 5 minutes with 3% H2O2 in methanol. After rinsing, sections were incubated for 45 minutes at 37°C in 19 U/ml of terminal deoxynucleotidyl transferase (Gibco BRL) and with 1.25 nmol/ml of biotin-16-deoxy-uridine triphosphate (UTP; Boehringer Mannheim, Indianapolis, IN, USA) in 30 mM of Trizma, 140 mM of sodium cacodylate, and 1 mM of CoCl2 buffer, pH 7.2. The biotin was detected using a standard avidin-biotin peroxidase technique (Dako, Carpinteria, CA, USA) and diaminobenzidine (Sigma Chemical Co.) as the chromogen. Sections were counterstained by methyl green. Small intestine of WT mice in which 0.05% of enterocytes typically are apoptotic and sections of TRAP-Tag transgenic mice in which osteoclast apoptosis is increased(42) were used as positive controls.
In vivo studies
Apoptotic osteoclasts and their precursors were identified in bone sections stained for TRAP activity using standard morphologic criteria: cell shrinkage, nuclear condensation and fragmentation, cytoplasmic condensation, and intense cytoplasmic TRAP staining as described previously.(42) We have confirmed previously that TRAP+ osteoclasts with these appearances in culture plates and in tissue sections are apoptotic using acridine orange staining and the TdT-mediated dUTP-biotin nick end-labeling (TUNEL) assay.(42,43)
Because NF-κB dKO mice do not form osteoclasts,(32) it is not possible to determine if the absence of p50 and p52 renders osteoclasts more sensitive to the effects of agents such as bisphosphonates, which induce osteoclast apoptosis in vivo.(41) However, because NF-κB p50−/−;p52+/− mice express no p50 and only one allele of p52, we used these mice instead in an attempt to address this question. Six-week-old NF-κB p50−/−;p52+/− and WT mice were given daily intraperitoneal injections of risedronate (0.5 mg/kg) for 4 days. The mice were killed by cervical dislocation 24 h after the last risedronate injection and the forelimbs were fixed and processed for histological analysis. Decalcified sections were prepared and stained for TRAP activity and morphologically viable and apoptotic osteoclasts were counted within the primary and secondary spongiosae and on the adjacent periosteal surfaces, as described previously.(43)
In vitro studies
To determine if inhibition of NF-κB activity promotes osteoclast apoptosis in vitro, osteoclasts were generated from WT bone marrow cells in the presence of 10−8 M of 1,25(OH)2D3, and mature osteoclasts were treated with various NF-κB inhibitors, including PDTC (1-50 μM), TPCK (1-50 μM), or MG-132 (0.5-4 μM) for 12-24 h. Then, the culture plates were fixed and stained for TRAP activity. The number of apoptotic and viable osteoclasts was counted and the former was expressed as a percentage of the total number of osteoclasts in each well, as described previously.(39) To attempt to determine if any observed proapoptotic effects of these agents were restricted to cells in the osteoclast lineage in these cultures, we examined their effects on stromal cell apoptosis, in addition to quantifying osteoclast apoptosis. Because most apoptotic stromal cells lose adhesion and disappear from the culture plates, we estimated the loss of these cells by this mechanism by measuring the area of culture wells occupied by the remaining adherent stromal cells using point counting and standard stereological methods. Values were expressed as a percentage of the total area of each culture well.
All results are given as means ± SE. Comparisons were made by ANOVA and Student's t-test for unpaired data. Values of p < 0.05 were considered to represent statistical significance. The Institutional Animal Care and Use Committee approved all studies.
NF-κB is required for RANKL-induced osteoclast formation
To determine if NF-κB p50 and p52 signaling is essential to mediate the osteoclastogenic effects of RANKL, we cocultured spleen cells from NF-κB dKO mice with WT calvarial osteoblasts in the presence of RANKL and M-CSF, with and without 1,25(OH)2D3, for 7 days and counted TRAP+ cells. Addition of RANKL plus M-CSF to the WT splenocytes induced osteoclast formation. We have shown that these cells are osteoclasts by virtue of their ability to resorb bone on dentine slices and express osteoclast marker genes, including cathepsin K, c-fms, MMP9, TRAP, and the calcitonin receptor (data not shown). The effect of RANKL plus M-CSF was greater than that of 1,25(OH)2D3 alone, which had an additive effect along with RANKL and M-CSF (Fig. 1A). In contrast, no osteoclasts were formed with spleen cells from dKO mice as the source of osteoclast precursors in the presence of any of these agents.
Osteoclasts are derived from multipotent stem cells, and the earliest identifiable osteoclast precursors are present in CFU-GM colonies.(44) To determine if RANKL/NF-κB signaling influences CFU-GM formation, we cultured spleen cells from dKO or WT mice with GM-CSF and M-CSF in the presence and absence of RANKL. Numerous colonies were formed when WT spleen cells were cultured with GM-CSF and M-CSF, and addition of RANKL increased the number of colonies 1.7-fold (Fig. 1B). Spleen cells from dKO mice formed three times more colonies than WT cells when they were cultured under the same conditions, but the number of colonies did not increase in response to RANKL (Fig. 1B). Similar to the findings in coculture experiments, no TRAP+ mononucleated or multinucleated cells were formed when dKO CFU-GM cells were cultured with RANKL and M-CSF, while WT colony cells formed many osteoclasts (Fig. 1C). Taken together, these data indicate that RANKL enhances the formation of osteoclast precursors in CFU-GM colonies in WT mice and that NF-κB p50/p52 signaling is required for osteoclasts to form from these cells.
Proinflammatory cytokines do not rescue the osteoclast defect of NF-κB dKO mice
Expression of the cytokines, IL-1, IL-6, and TNF-α is regulated by NF-κB in hematopoietic cells. Thus, the lack of osteoclasts in NF-κB dKO mice could be caused by decreased production of these cytokines. Initially, we examined the effects of adding IL-6 ± IL-6 soluble receptor, IL-1, TNF-α, and GM-CSF alone and in combination to cocultures of spleen cells from dKO mice and WT osteoblasts in the presence of 1,25(OH)2D3 to determine if they could rescue the osteoclast defect in vitro. We also treated dKO splenocytes with M-CSF, RANKL with or without the above cytokines, or in combination. No TRAP+ osteoclasts were formed under any of these conditions (Fig. 2). To exclude the possibility that reduced expression of c-fms (the receptor tyrosine kinase for M-CSF) and OPG in spleen and liver cells from NF-κB dKO mice is responsible for the defect in osteoclastogenesis, we used reverse-transcription (RT) PCR with sequence-specific primers. The mRNA expression levels of these cytokines and their receptors were similar in cells from dKO mice and their WT littermates (Fig. 3).
RANK-expressing splenocytes are increased in NF-κB dKO mice
To determine if NF-κB dKO mice produce RANK-expressing cells, we used FACS analysis to examine RANK expression on osteoclast precursors by labeling splenocytes from dKO and WT mice with FITC-labeled RANKL. We found using this method that the percentage of RANK+ cells was increased threefold in the dKO splenocytes compared with WT controls (Fig. 4A). Splenocytes from RANK KO mice were used as negative controls in these experiments to determine the specificity of FITC-labeled RANKL binding to RANK. Binding was detected in 3-5% of splenocytes from these animals (Fig. 4A), suggesting either nonspecific background staining or specific interaction of the synthesized protein with a receptor other than RANK. Previous studies have indicated that RANK is expressed only on the surface of osteoclast precursors, dendritic cells, and activated T cells(1,2) and that NF-κB dKO mice have almost no T cells in their spleens.(32) To examine the types of cells that express RANK in NF-κB dKO mice, splenocytes from dKO and WT mice were double-stained with FITC-labeled RANKL and antibodies against CD11b (Mac-1, an osteoclast and macrophage precursor marker), CD11c (a monocyte/granulocyte marker), or F4/80 (a mature macrophage marker). More CD11b+ osteoclast/macrophage precursors and CD11c+ monocytes expressed RANK in splenocytes from dKO mice than from WT mice, but no difference was observed in the percentage of F4/80+ mature macrophages expressing RANK (Fig. 4B). These data indicate that RANK expression was increased on the surface of cells in the monocyte/macrophage lineage in dKO mice, consistent with a blockage in differentiation, rather than a failure of cells in this lineage to survive because of lack of NF-κB p50 and p52 expression.
Apoptosis is not increased in bone marrow of NF-κB KO mice
Because NF-κB has an antiapoptotic role in some cell types, for example, in lymphocytes after TNF-α/Fas ligand binding,(34,35) the absence of osteoclasts in the NF-κB dKO mice could be caused by increased apoptosis of osteoclasts or their TRAP+ or TRAP− mononuclear precursors. Therefore, we examined bone, spleen, and liver sections from these mice and used hematoxylin and eosin (H&E) and TUNEL staining to detect apoptotic cells. We found no increase in the number of apoptotic cells in sections of bone marrow or liver (where there is extramedullary hematopoiesis as a result of the osteopetrosis in these animals) from the dKO mice. However, the number of TUNEL+ mononuclear cells was significantly increased in spleen sections of the mutant mice (Fig. 5). We do not yet know the histogenesis of these TUNEL+ cells. We also examined CFU-GM colony cells using the 3[4,5-dimethyl thiazol-2-yl]2-5-diphenyl tetrazolium bromide; thiazolyl blue (MTT) assay and found no difference in cell survival between dKO and WT mice (data not shown), providing further evidence that there is no massive osteoclast precursor death in dKO mice.
The bone phenotype of the NF-κB dKO mice also could result from increased sensitivity of osteoclasts or their precursors to proapoptotic signals because of the absence of NF-κB p50 and p52 expression. Because very few osteoclasts form in NF-κB dKO mice,(32) it is not possible to study the effects of proapoptotic agents on osteoclasts in them to determine if they have increased susceptibility to undergo apoptosis. Thus, in an attempt to examine this issue, we used NF-κB p50−/−;p52+/− (3/4 KO) mice that do not express NF-κB p50 and express only one allele of the NF-κB p52 gene but, nonetheless, form near normal numbers of osteoclasts. We treated NF-κB 3/4KO and WT mice for 4 days with the bisphosphonate risedronate, which promotes osteoclast apoptosis in vitro and in vivo,(43) and counted apoptotic and viable osteoclasts in TRAP-stained bone sections. A 10-fold increase in the percent apoptotic osteoclasts was observed in both WT and KO mice (Fig. 6). Furthermore, similar levels of increased osteoclast apoptosis were observed in bone marrow cultures from NF-κB p50−/− and WT mice in response to risedronate (48 ± 3.7 vs. 43 ± 1.2) and transforming growth factor (TGF) β (50 ± 3 vs. 45 ± 2.7) compared with PBS-treated cultures (30 ± 6.3 vs. 32 ± 2.4).
NF-κB inhibitors induce both osteoclast and stromal cell death in vitro
Several classes of compounds have been shown to inhibit NF-κB activation in a variety of cell types. These include antioxidants such as PDTC, the protease inhibitor TPCK and the proteasome inhibitor MG-132 that induce apoptosis of isolated mature osteoclasts in vitro.(22,31) Because bone formation appears to be normal in NF-κB dKO mice, and calvarial osteoblasts from these mice support osteoclast formation from WT spleen cells,(32) inhibition of NF-κB p50 and p52 activity in vitro should not have a detrimental effect on WT osteoblasts but could affect osteoclast survival. In contrast, inhibition of all members of the NF-κB family could promote apoptosis of all cell types in bone. To determine if the above compounds had proapoptotic effects on osteoclasts specifically among bone cells, we tested them in our in vitro osteoclast apoptosis assay that includes osteoclasts and stromal cells from murine marrow.(43) At concentrations that have been reported to inhibit NF-κB nuclear translocation in several cell types,(45) these compounds promoted apoptosis of both osteoclasts and stromal cells in a dose-dependent manner. PDTC (Fig. 7) caused both osteoclast and stromal cell apoptosis at a concentration of 5 μM, an effect also observed with TPCK at 5 μM and with MG-132 at 1 μM concentration (data not shown). In addition, these compounds blocked osteoclast formation when they were added at the beginning of the bone marrow culture after removal of cells from WT mouse long bones, but this also was associated with extensive death of stromal cells (data not shown).
RANKL/RANK signaling is essential for osteoclast differentiation from hematopoietic precursors and plays a critical role in normal bone remodeling.(6) Studies of the mechanisms of RANK signal transduction have shown that RANKL stimulation leads to activation of NF-κB in RANK-expressing T cells(1) and of JNK in mouse thymocytes,(4) indicating that NF-κB and AP-1 are downstream of RANKL/RANK signaling. Our present studies indicate that RANKL-induced osteoclastogenesis requires expression of both NF-κB p50 and p52 proteins because NF-κB p50;p52 dKO mice do not form osteoclasts in vivo, and treatment of splenocytes from these mice with RANKL, M-CSF, and 1,25(OH)2D3 in vitro fails to produce osteoclasts, unlike treatment of WT splenocytes.
The lack of osteoclast formation in NF-κB dKO mice is not caused by failure of RANKL expression by osteoblast/stromal cells or to the absence of receptors for RANKL and/or M-CSF on osteoclast progenitors because expression of RANKL by stromal cells and of RANK and c-fms mRNA by splenic and hepatic hematopoietic cells from the mutant mice was similar to that of their WT littermates. Indeed, the percentage of splenocytes with surface expression of RANK protein was increased in the dKO mice. RANK mRNA is expressed in a wide variety of tissues, but its surface expression has been detected only in dendritic cells, activated T cells, osteoclasts, and their precursors.(1,16) Because of restricted RANK protein expression and because only RANK+ cells (sorted by FITC-labeled RANKL) form osteoclasts(2) (our unpublished observations), it is reasonable to conclude that RANK expression may be used as an early marker for osteoclast precursors. Our finding that the percentage of splenocytes expressing RANK is increased in NF-κB dKO mice and that most of them are in the monocyte/macrophage lineage support this conclusion and indicate that expression of NF-κB p50 and p52 is required to permit further RANKL-induced differentiation of these cells along the osteoclast lineage. Thus, our studies provide primary evidence that (1) RANK is expressed not only on osteoclasts but also on macrophages because RANK+ cells also express CD11b or F4/80, both markers for macrophages (Fig. 4), and, therefore, RANK cannot be used as a marker to differentiate osteoclast precursors from macrophage precursors; (2) blockage of osteoclastogenesis in dKO mice occurs at the stage at which osteoclast precursors express RANK on their surface; and (3) the absence of NF-κB p50 and p52 expression by these cells does not adversely affect their survival, which if reduced sufficiently could in theory decrease the pool of precursors to a point where few or no osteoclasts would form.
Our findings also indicate that the lack of osteoclasts in NF-κB dKO mice is not caused by overexpression of OPG, the secreted TNF receptor family member that inhibits osteoclastogenesis and promotes osteoclast apoptosis by interfering with RANKL/RANK interaction,(7,46) because OPG expression is normal in the dKO mice. Our findings strongly suggest that NF-κB p50 and p52 are essential downstream molecules in the RANKL/RANK signaling pathway during osteoclast progenitor proliferation and differentiation and that other DNA-binding NF-κB complexes present in these dKO mice, the majority of which are p65 and c-Rel,(32) cannot compensate for the absence of NF-κB p50 and p52. Our conclusion is supported by the close similarities in the phenotypes of NF-κB, RANKL, and RANK KO mice, all of which fail to form not only osteoclasts but also mature B cells, and all have small spleens and thymuses and absent or small lymph nodes.
NF-κB regulates the expression of the proinflammatory cytokines, IL-1, IL-6, and TNF-α, and IL-1 and TNF-α in turn also regulate NF-κB expression. These cytokines also have been linked to the up-regulation of osteoclastogenesis and activity that leads to the bone loss associated with estrogen deficiency after menopause(47,48) and with inflammatory disorders of bone.(49,50) Therefore, it is possible that NF-κB p50 and p52 are required for expression of these cytokines in cells involved in osteoclast formation such as stromal cells. To test this hypothesis, we added IL-1, IL-6 plus its soluble receptor, TNF-α, or GM-CSF alone or in combination [with or without 1,25(OH)2D3] to spleen cells isolated from NF-κB dKO mice in coculture with WT calvarial osteoblasts. No TRAP+ osteoclasts were formed under these conditions. Although IL-6 production from dKO macrophages(33) and stromal cells (data not shown) is reduced, the lack of IL-6 is not the essential factor for halted osteoclastogenesis in dKO mice because (1) dKO stromal cells can support osteoclast formation in vitro(32) and (2) IL-6 KO mice form near normal numbers of osteoclasts in vivo and in vitro. More recently, Kudo et al. reported that IL-6 induces human osteoclast formation and bone resorption independent of RANK/RANKL interaction.(51) However, we did not find IL-6 able to induce osteoclast formation in either NF-κB dKO mice or in RANK KO mice (data not shown). Because none of these cytokines that are regulated by NF-κB could rescue the osteoclast defect, we conclude that p50 and p52 regulate the expression other genes in TRAP− osteoclast precursors that are essential for osteoclast formation.
Because NF-κB regulates the survival of a variety of cell types,(34,35) the lack of osteoclasts in NF-κB dKO mice also could be caused by increased apoptosis of osteoclasts or their precursors. We examined bone sections from the NF-κB dKO mice using morphologic analysis of H&E-, TRAP-, and TUNEL-stained sections, but we found no evidence of increased apoptosis of bone marrow cells. Furthermore, we found no evidence of increased apoptosis in osteoclasts cultured from the NF-κB 3/4 and single KO mice compared with those from WT mice in the presence or absence of inducers of osteoclast apoptosis, such as risedronate and TGF-β, or in bone sections of 3/4 KO mice when they were treated with risedronate. Two previous studies have described induction of osteoclast apoptosis by pharmacologic inhibitors of NF-κB,(22,31) in cultures containing a relatively pure population of osteoclasts. To determine if osteoclasts have increased sensitivity to these NF-κB inhibitors in comparison to stromal cells, we examined their effects on both cell types in our in vitro bone marrow apoptosis assay. We found that these agents caused apoptosis of both osteoclasts and stromal cells when they were added to formed osteoclasts (Fig. 7) and halted stromal cell-induced osteoclastogenesis when they are added during the first 3 days of culture (data not shown). Thus, the proapoptotic effects of these NF-κB inhibitors show no specificity for the cells in the osteoclast lineage for which p50 and p52 expression is essential. We conclude that the major physiological role of NF-κB in osteoclasts is to regulate their formation rather than their survival.
In summary, a molecular mechanism by which NF-κB may control osteoclast numbers is proposed whereby NF-κB is a key downstream mediator of RANKL- and cytokine-induced osteoclastogenesis. The previously identified NF-κB target gene products, IL-1, TNF, and IL-6, cannot rescue the NF-κB dKO osteoclast defect in vitro. Thus, the downstream target(s) of NF-κB that is essential for osteoclast differentiation remains to be identified. Although the development of B lymphocytes also is halted in NF-κB dKO mice, the B cell defect largely is one of increased apoptosis,(52) suggesting different essential NF-κB targets in these cells from those in osteoclasts.
We thank Dr. Bill Dougall (Immunex Corp., Seattle, WA, USA) for the gift of RANKL and Beryl Story and Arlene Farias for technical assistance. Funding for this work was provided by the National Institutes of Health (NIH; grant PHS AR 43510; to B.F.B.).