Osteoclast Differentiation Is Impaired in the Absence of Inhibitor of κB Kinase α*

Signaling through the receptor activator of nuclear factor κB (RANK) is required for both osteoclast differentiation and mammary gland development, yet the extent to which RANK utilizes similar signaling pathways in these tissues remains unclear. Mice expressing a kinase-inactive form of the inhibitor of κB kinase α (IKKα) have mammary gland defects similar to those of RANK-null mice yet have apparently normal osteoclast function. Because mice that completely lack IKKα have severe skin and skeletal defects that are not associated with IKKα-kinase activity, we wished to directly examine osteoclastogenesis in IKKα-/- mice. We found that unlike RANK-null mice, which completely lack osteoclasts, IKKα-/- mice did possess normal numbers of TRAP+ osteoclasts. However, only 32% of these cells were multinucleated compared with 57% in wild-type littermates. A more profound defect in osteoclastogenesis was observed in vitro using IKKα-/- hematopoietic cells treated with colony-stimulating factor 1 and RANK ligand (RANKL), as the cells failed to form large, multinucleated osteoclasts. Additionally, overall RANKL-induced global gene expression was significantly blunted in IKKα-/- cells, including osteoclast-specific genes such as TRAP, MMP-9, and c-Src. IKKα was not required for RANKL-mediated IκBα degradation or phosphorylation of mitogen-activated protein kinases but was required for RANKL-induced p100 processing. Treatment of IKKα-/- cells with tumor necrosis factor α (TNFα) in combination with RANKL led to partial rescue of osteoclastogenesis despite a lack of p100 processing. However, the ability of TNFα alone or in combination with transforming growth factor β to induce osteoclast differentiation was dependent on IKKα, suggesting that synergy between RANKL and TNFα can overcome p100 processing defects in IKKα-/- cells.

multinucleated cells that differentiate from monocyte/macrophage lineage hematopoietic precursors. The key factors regulating osteoclastogenesis are the TNF 1 receptor family member RANK, its ligand (RANKL), and the RANKL inhibitor, osteoprotegerin. Ablation of RANK or RANKL in mice results in a complete absence of osteoclasts and severe osteopetrosis, whereas lack of osteoprotegerin leads to excessive osteoclast activity and osteoporosis (2)(3)(4)(5).
Genetic studies in mice, as well as in vitro culture of osteoclasts using CSF-1 and RANKL, have elucidated many key components of RANK signaling during osteoclastogenesis. Binding of RANKL to RANK stimulates recruitment of TRAF2, -5, and -6 (6) followed by activation of mitogenactivated protein kinases and IB kinases, which ultimately lead to activation of the transcription factors AP-1 and NF-B. Osteopetrosis has been observed in c-fos Ϫ/Ϫ mice (7), Nfkb1/2 double knock-out mice (8), and TRAF6 Ϫ/Ϫ mice (9). The contribution of other signaling components to RANKmediated osteoclastogenesis remains unclear because ablation of many of these genes, such as the NF-B family member p65, TRAF2, and IKK␤, results in embryonic or neonatal lethality (10 -12). Additionally, although cytokines that contribute to osteoclast differentiation during inflammation, such as TNF␣ and interleukin 1, can activate NF-B and AP-1 (13,14) in osteoclast precursor cells, RANK and RANKL remain essential for osteoclastogenesis in vivo. For instance, administering osteoprotegerin to rats with adjuvant-induced arthritis abrogates osteoclastogenesis and subsequent bone loss without inhibiting inflammation (15). Furthermore, bone erosion induced by overexpression of TNF␣ in transgenic mice can be alleviated by administering RANK-Fc or by crossing TNF␣ transgenic mice into a RANK-null background (16). These studies indicate that RANK may activate unique signaling pathways in osteoclast progenitor cells.
In addition to osteoclast differentiation, RANK signaling is required for mammary gland development during pregnancy (17). Female RANK-or RANKL-null mice are unable support live birth because of an inability to lactate. This defect is because of insufficient mammary epithelial cell proliferation during pregnancy as well as an increase in mammary epithelial cell apoptosis, indicating that RANK signaling is required for both the proliferation and survival of mammary epithelial cells that are necessary for the full development of the mammary * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. lobuloalveolar structures during pregnancy. As with osteoclasts, genetic studies in mice have been useful for determining the specific signaling pathways utilized by RANK in the mammary gland. Specifically, mice expressing a kinase-inactive allele of the inhibitor of B kinase ␣ (IKK␣ AA ) had defects in mammary gland differentiation similar to those in RANK or RANKL-null mice (18). These defects were linked to a lack of NF-B activation and subsequent up-regulation of cyclin D1 in mammary epithelial cells. Interestingly, IKK␣ AA mice had no apparent defects in osteoclast differentiation, indicating that RANK may utilize tissue-specific signaling pathways to activate NF-B.
The transcription factor NF-B is composed of homo-or heterodimers of the NF-B family members RelA/p65, RelB, cRel, p50, and p52 (reviewed in Ref. 12). The latter two proteins are synthesized as precursors p105 and p100, respectively. Processing of p105 into p50 is constitutively regulated, and p50 is most commonly associated with p65 and the inhibitor protein IB␣ in an inactive complex in the cytosol. Activation of the IKK complex containing the catalytic IKK␣ and -␤ subunits and the regulatory IKK␥ subunit leads to phosphorylation, ubiquitination, and subsequent proteasome-mediated degradation of IB␣, allowing nuclear translocation of p65/p50. In many cell types, such as fibroblasts or lymphocytes, IKK␣ kinase activity is dispensable for phosphorylation of IB␣ and activation of p65/p50 in response to TNF␣ or interleukin 1, but it is required for the processing of p100 into p52 in response to stimuli including lymphotoxin ␤, CD40 ligand, and B cell activating factor. In contrast, IKK␣ is required for the activation of p65/p50 by RANKL in mammary epithelial cells (18). The contribution of IKK␣ to RANKL-mediated osteoclastogenesis remains unclear, although both p100 processing and p65/p50 nuclear translocation have been implicated as important components of this process (19,20). To determine whether IKK␣ plays a role in osteoclast differentiation, we used mice completely lacking IKK␣ protein (IKK␣ Ϫ/Ϫ ) (21). Because these mice die at birth, we examined osteoclast differentiation in vivo in embryonic day 18.5 postcoitus (E18.5 dpc) IKK␣ Ϫ/Ϫ embryos and in vitro by culturing IKK␣ Ϫ/Ϫ hematopoietic cells with CSF-1 and RANKL. Using this system, we determined that the ability of RANKL to induce the formation of large multinucleated osteoclasts in vitro is impaired in the absence of IKK␣. This observation was correlated with a lack of p100 processing into p52. However, adding TNF␣ to RANKL-treated IKK␣ Ϫ/Ϫ cells partially restored osteoclastogenesis despite the presence of accumulated p100. Interestingly, we also observed a partial absence of multinucleated osteoclasts in vivo in IKK␣ Ϫ/Ϫ embryos, indicating that a lack of IKK␣ in osteoclast precursor cells in vivo may be overcome by cooperation between RANKL and other cytokines, such as TNF␣.

Precursor Cell Analysis and Osteoclast Differentiation Assays-
Timed matings were set up for IKK␣ hemizygous male and female mice as described previously (21) to obtain IKK␣ wild-type and knock-out embryos. Fetal livers were harvested from E18.5 dpc embryos and disrupted to obtain single-cell suspensions in ␣ minimum Eagle's medium (Invitrogen). Red blood cells were lysed in lysis solution (Sigma) followed by resuspension in ␣ minimum Eagle's medium containing 10% fetal bovine serum and antibiotics. To determine progenitor cell frequency, cells were plated in Methocult TM medium 3534 containing stem cell factor, interleukin 6, and interleukin 3 (StemCell Technologies, Inc., Vancouver, British Columbia, Canada) according to the manufacturer's directions, and morphologically distinct GM-CFU colonies were counted for duplicate wells of three separate wild-type and IKK␣ Ϫ/Ϫ embryos. For analysis of macrophage lineage markers, fetal liver cells were grown for 6 days in medium containing ␣ minimum Eagle's medium and 40 ng/ml CSF-1 (R&D Systems, Minneapolis, MN). Cells were trypsinized and incubated with fluorescently labeled anti-bodies to CD11b (BD Biosciences) and MOMA-2 (Serotec, Inc., Raleigh, NC) followed by flow cytometric analysis. For osteoclast differentiation assays, cells were plated at 5 ϫ 10 4 cells/well in 96-well plates in the presence of 40 ng/ml CSF-1 with or without 200 ng/ml recombinant murine RANKL (22). Where indicated, cells were also treated with recombinant murine TNF␣ at 20 ng/ml and/or TGF␤ (R&D Systems). Every 48 h, one-half of the cell medium was replaced with medium containing fresh cytokines. Cells were stained after 6 days in culture for TRAP in situ using a commercially available kit (Sigma). For direct analysis of osteoclasts in vivo, E18.5 dpc embryos were fixed in 10% neutral buffered formalin for 24 h and decalcified in 10% EDTA for 5 days. Embryos were step-sectioned into 40 sections and stained for TRAP activity.
RNA Analysis-RNA was extracted from wild-type and IKK␣ Ϫ/Ϫ fetal liver cell cultures after 6 days treatment with CSF-1 and murine RANKL using the RNeasy TM kit (Qiagen). For quantitative reverse transcription-PCR, cDNA was prepared from 1 g of total RNA and used for PCR reactions using specific primers and SYBR Green dye (Applied Biosystems, Foster City, CA) as described previously (23). For microarray expression profiling, 5 g of total RNA from individual samples was labeled according to standard protocols (Affymetrix, Santa Clara, CA) and 10 g of biotinylated cRNA was hybridized to Mouse Affymetrix U430A (MOE430A) chips consisting of ϳ22,500 probe sets (Unigene Build 430). Hybridized chips were stained and washed on an Affymetrix FS400 fluidics station using the antibody amplification protocol and scanned using an Affymetrix GeneArray 2500 scanner. Scanned images were loaded into the Resolver TM system (Rosetta Biosoftware, Kirkland, WA) for analysis.
Western Blots-Cells were grown for 6 days in CSF-1 and treated with murine RANKL for indicated times prior to lysis in 1ϫ cell lysis buffer (Cell Signaling Technologies, Beverly, MA) supplemented with 1 mM phenylmethylsulfonyl fluoride and 1 mM NaF. Insoluble material was removed by centrifugation at 13,000 rpm in a microcentrifuge at 4°C for 10 min. Supernatants were stored in aliquots at Ϫ70°C. Protein concentration was determined using a BCA kit (Pierce), and 10 g of each sample were separated on 8 -16% gradient gels (Invitrogen) and then transferred to nitrocellulose membranes. Equivalent loading was confirmed by staining membranes with Ponceau S reagent followed by blocking in 5% milk in Tris-buffered saline, 0.1% Tween 20. Primary antibodies were applied in a solution containing 3% bovine serum albumin and 0.1% Tween 20 overnight at 4°C. Horse radish peroxidaselabeled secondary antibodies were applied after washing, which was followed by more washing and then detection using ECL reagent (Amersham Biosciences). IB␣ antibody (C-21) was obtained from Santa Cruz Biotechnology. Antibodies for phosphorylated proteins and corresponding total proteins were obtained from Cell Signaling Technologies. Antibody to mouse p100 was created by injecting rabbits with the NH 2 -terminal peptide DNCYDPGLDGIPEYDD, as described previously (24).

RESULTS
Defective Osteoclast Differentiation in the Absence of IKK␣-IKK␣-deficient mice die immediately after birth because of keratinocyte differentiation defects (21). Therefore, to determine whether IKK␣ was required for osteoclast differentiation, we analyzed osteoclasts in embryos of IKK␣ wild-type (ϩ/ϩ or ϩ/Ϫ) and IKK␣-null (Ϫ/Ϫ) littermates at E18.5 dpc. Although similar numbers of cells in both wild-type and knock-out embryos stained positive for the osteoclast marker TRAP (data not shown), the morphology and size of the TRAP ϩ cells observed in IKK␣ Ϫ/Ϫ embryos was dramatically different from those in wild-type littermates (Fig. 1A). In IKK␣ wild-type embryos, osteoclasts were predominantly multinucleated (57% multinucleated in ϩ/ϩ and 61% in ϩ/Ϫ, Fig. 1B) and contained a highly vacuolated cytoplasm. In contrast, significantly fewer multinucleated TRAP ϩ cells were identified in IKK␣ Ϫ/Ϫ embryos (32%, p Ͻ 0.001 compared with ϩ/ϩ, Fig. 1B), and these cells had an altered morphology compared with wild type, often containing a reduced dense granulated cytoplasm.
To further characterize the role of IKK␣ in osteoclastogenesis, we next determined whether IKK␣-null fetal liver-derived hematopoietic cells could differentiate into osteoclasts in vitro. Fetal liver cells (FLC) from wild-type or IKK␣ Ϫ/Ϫ E18.5 dpc embryos were treated with CSF-1 and RANKL for 6 days and stained for TRAP activity. Treatment of wild-type FLC with CSF-1 and RANKL led to the formation of large multinucleated TRAP ϩ osteoclasts ( Fig. 2A). However, in IKK␣ Ϫ/Ϫ FLC treated with CSF-1 and RANKL, only small TRAP ϩ cells were observed. The number of multinucleated osteoclasts generated from IKK␣ Ϫ/Ϫ FLC was decreased by Ͼ15-fold compared with wild type (Fig. 2). To determine whether the lack of osteoclast differentiation observed in IKK␣ Ϫ/Ϫ FLC was because of a difference in osteoclast precursor cell numbers, cells were grown in methylcellulose medium containing monocyte/macrophage growth factors, and the frequency of GM-CFU, the earliest defined myeloid progenitor (25), was determined. No differences in GM-CFU frequency were observed between wildtype and IKK␣ Ϫ/Ϫ cells (Fig. 3A), indicating that the lack of osteoclast differentiation observed in IKK␣ Ϫ/Ϫ FLC is not because of a defect in the myeloid precursor cell compartment. The defect observed in IKK␣ Ϫ/Ϫ FLC is specific to osteoclasts but not macrophage, because similar numbers of cells expressing the markers CD11b and MOMA-2 were observed in wildtype and IKK␣ Ϫ/Ϫ FLC after 1 week of growth in CSF-1 (Fig.  3B). Taken together, these results indicate that IKK␣ is required for RANKL-mediated osteoclast differentiation from CSF-1-derived precursor cells in vitro.
Lack of Osteoclast-specific Gene Up-regulation in RANKLtreated IKK␣ Ϫ/Ϫ Cells-To further analyze the extent to which RANK-mediated osteoclast differentiation was inhibited in the absence of IKK␣, we assessed the expression of osteoclastspecific genes using quantitative reverse transcription-PCR (Fig. 4). In wild-type FLC grown in CSF-1 and RANKL for 6 days, expression of Mmp-9, c-Src, TRAP, and ␤ 3 integrin was greatly increased compared with cells grown in CSF-1 alone. In contrast, RANKL-induced expression of these genes in IKK␣ FLC was blunted, with a Ͼ16-fold decrease in expression compared with wild type. Importantly, however, expression of RANK was not different between wild-type or IKK␣ Ϫ/Ϫ FLC, indicating that loss of RANK expression was not responsible for the lack of response to RANKL in IKK␣ Ϫ/Ϫ FLC.
To determine whether IKK␣ was required for a specific subset of RANKL-induced genes or for overall RANKL-induced gene expression, we compared global gene expression in both knock-out and wild-type FLC grown for 6 days in CSF-1 with or without RANKL using an oligonucleotide array representing ϳ22,500 mouse genes (Affymetrix). A striking difference was observed in overall RANKL-induced gene expression in IKK␣ Ϫ/Ϫ cells compared with wild type. Although 209 genes were up-regulated Ն5-fold (p Ͻ 0.001) consistently in duplicate cultures of wild-type FLC treated with CSF-1 and RANKL compared with cells treated with CSF-1 alone (Fig. 5A and supplemental material), only 29 genes were up-regulated Ն5fold consistently in duplicate cultures of IKK␣ Ϫ/Ϫ FLC in response to RANKL. Conversely, 226 genes were down-regulated Ն5-fold in RANKL-treated wild-type FLC, but only three were down-regulated Ն5-fold in RANKL-treated IKK␣ Ϫ/Ϫ FLC. The expression patterns of the genes up-regulated Ͼ5-fold in wildtype FLC treated with CSF1 and RANKL fall into two broadly even categories in similarly treated IKK␣ Ϫ/Ϫ FLC. These categories are a group of genes that show no up-regulation and a group of genes that show "diminished" up-regulation. For example, of the 79 most up-regulated genes in wild-type FLC (Ͼ20-fold, p Ͼ 0.001), 32 show effectively no up-regulation (with a trend to down-regulation) in IKK␣ Ϫ/Ϫ FLC (Fig. 5B), whereas 47 show up-regulation but with diminished magnitude (Fig. 5C). Taken together, these results indicate that the overall changes in gene expression induced by RANKL during osteoclast differentiation in vitro are dependent on IKK␣.
IKK␣ Is Required for RANK-mediated p100 Processing but Not for Activation of the Classical NF-B Pathway in CSF-1derived Precursor Cells-Activation of the NF-B transcription factor pathway has been shown previously to occur immediately after binding of RANKL to RANK and has been implicated in osteoclast differentiation, activation, and survival (1). To determine whether IKK␣ was required for activation of NF-B by RANKL in osteoclast precursor cells, we analyzed the ability of RANKL to stimulate degradation of the NF-B inhibitor IB␣ in CSF-1-derived wild-type and IKK␣ Ϫ/Ϫ FLC.
In both wild-type and knock-out cells, RANKL treatment led to the complete degradation of IB␣ by 15 min, which was followed by IB␣ resynthesis at 30 min (Fig. 6A, top panel). Because IB␣ resynthesis is dependent on NF-B transcription factor activity, these data indicate that IKK␣ is not required for initial activation of NF-B by RANKL in osteoclast precursor cells. In addition to the classical NF-B pathway, RANKL activates a second, or non-canonical, NF-B activity involving processing of the inhibitory protein p100 into a smaller active p52 subunit. Previous studies have shown that RANKL-mediated p100 processing is dependent on NIK, an upstream activator of IKK␣ (19). To directly assess the role of IKK␣ in RANKL-mediated p100 processing, we analyzed the relative levels of both p100 and p52 in CSF-1-derived wild-type and IKK␣ Ϫ/Ϫ FLC treated with RANKL (Fig. 6A, bottom panel). In wild-type cells, p52 accumulation was apparent 8 h after RANKL treatment, whereas p100 levels remained unchanged. However, in IKK␣ Ϫ/Ϫ cells, RANKL treatment led to accumulation of p100 without any increase in p52 levels, indicating that IKK␣ is required for RANKL-mediated p100 processing in osteoclast progenitor cells.
Next we analyzed additional signaling pathways activated by RANKL, specifically mitogen-activated protein kinases p38, extracellular signal-regulated kinase, and c-Jun NH 2 -terminal kinase (Fig. 6B). Phosphorylated forms of these proteins were detected 10 min post-RANKL exposure in both wild-type and  IKK␣ Ϫ/Ϫ cells. Taken together, these data indicate that IKK␣ is not required for activation of RANKL-activated early signaling pathways involving classical NF-B and mitogen-activated protein kinases, but is necessary for later RANKL-induced p100 processing. Comparison of global RANKL-induced gene expression in wild-type and IKK␣ ؊/؊ fetal liver cells. RNA was prepared from WT and IKK␣ Ϫ/Ϫ (KO) fetal liver cells cultured in CSF-1 with or without RANKL for 6 days. A, 443 genes were identified that were upor down-regulated Ն5-fold (p Ͻ 0.001) in duplicate samples of either WT or KO cells treated with CSF-1 and RANKL compared with cells treated with CSF-1 alone. Red represents up-regulation of gene expression in RANKL-treated cells compared with cells grown in CSF-1 alone, green represents down-regulation of gene expression, and black indicates unchanged. A complete list of genes is available in a table in the supplemental material. B, 79 genes that were highly up-regulated (Ͼ20-fold, p Ͻ 0.001) in duplicate cultures of WT cells in response to RANKL but not in IKK␣ Ϫ/Ϫ . C, 47 genes that were highly up-regulated in WT cells in response to RANKL and were up-regulated to a lesser extent in IKK␣ Ϫ/Ϫ cells. compared with wild-type littermates. The discrepancy between in vitro and in vivo osteoclastogenesis in IKK␣ Ϫ/Ϫ cells may be because of the presence of additional factors such as TNF␣ or TGF␤ that can promote RANKL-mediated differentiation (13,26). Therefore, we wished to determine whether TNF␣ or TGF␤ could affect RANKL signaling (specifically p100 processing) or ultimately osteoclastogenesis in IKK␣ Ϫ/Ϫ cells. Neither TNF␣ nor TGF␤ stimulated processing of p100 into p52, either alone or in combination with RANKL in IKK␣ Ϫ/Ϫ cells (Fig. 7A). However, TNF␣ treatment alone, but not TGF␤, led to accumulation of p100 protein (without p52 formation) in IKK␣ Ϫ/Ϫ cells, similar to RANKL treatment. In wild-type cells, TNF␣ treatment increased both p100 and p52 protein levels, in contrast to RANKL treatment, which only increased p52. Taken together, these data indicated a potential role for IKK␣ during TNF␣ signaling in osteoclast precursor cells. Consistent with this, we observed that fewer TRAP ϩ multinucleated cells formed in TNF␣-treated IKK␣ Ϫ/Ϫ cells compared with those observed in wild-type cells. Furthermore, treatment with TNF␣ and TGF␤ led to the formation of large, multinucleated TRAP ϩ osteoclasts in wild-type but not IKK␣ Ϫ/Ϫ cells. Interestingly, we found that TNF␣ in combination with RANKL was able to induce the formation of large, multinucleated TRAP ϩ osteoclasts in IKK␣ Ϫ/Ϫ FLC, despite the accumulation of p100 and lack of p52 formation following this treatment. The overall number of multinucleated TRAP ϩ cells observed in IKK␣ Ϫ/Ϫ cells treated with RANKL and TNF␣ was similar to that observed in wild-type cells; however the majority of these cells were smaller in IKK␣ Ϫ/Ϫ cultures than in wild type (Fig. 7, B and  C). The combination of RANKL, TNF␣, and TGF␤ also resulted in the formation of large, multinucleated TRAP ϩ os-teoclasts in IKK␣ Ϫ/Ϫ FLC. This effect was completely inhibited by the RANKL inhibitor osteoprotegerin (data not shown), indicating that signaling via RANK was essential for osteoclast formation. DISCUSSION Here we have demonstrated a clear role for IKK␣ during osteoclastogenesis in vitro. IKK␣ Ϫ/Ϫ fetal liver cell cultures treated with CSF-1 and RANKL failed to form large, multinucleated TRAP ϩ osteoclasts, even after extended culture periods or with increased amounts of RANKL (not shown). However, unlike RANK or RANKL knock-out mice, which completely lack TRAP ϩ osteoclasts in vivo, IKK␣ Ϫ/Ϫ E18.5 dpc embryos did contain multinucleated TRAP ϩ osteoclasts, although the frequency of these cells was less than that observed in wild-type embryos. Because IKK␣ Ϫ/Ϫ mice die at birth, it is unclear whether the decreased number of multinucleated osteoclasts observed in these mice would result in a functional deficit in adult animals, such as a defect in tooth eruption or proper bone remodeling. Interestingly, mice lacking the upstream activator of IKK␣, NIK, have no defects in basal osteoclastogenesis during development and thus have normal tooth eruption and skeletal morphology but fail to respond to osteoclastogenic stimuli, including RANKL and vitamin D (19). Additionally, bone marrow cells from NIK Ϫ/Ϫ mice are unable to form osteoclasts when cultured in vitro with CSF-1 and RANKL. Taken together, these studies indicate that the NIK/IKK␣ pathway is important during stimulated but not basal osteoclastogenesis. However, NIK Ϫ/Ϫ mice are not phenotypically identical to either IKK␣ Ϫ/Ϫ or IKK␣ AA mice because they have no defects in keratinocyte differentiation, as observed in IKK␣ Ϫ/Ϫ mice, or in mammary gland development, as observed in IKK␣ AA mice (in addition to RANK-and RANKL-null mice) (17,18). Mice expressing a mutant form of NIK (aly/aly) do have defects in mammary gland development that are similar to, albeit less severe than, defects in IKK␣ AA mice (27), yet it is unknown whether either aly/aly mice or IKK␣ AA mice have defects in stimulated osteoclastogenesis. Given the differences between NIK Ϫ/Ϫ , aly/aly, IKK␣ Ϫ/Ϫ , and IKK␣ AA mice, it remains to be determined which functions of IKK␣ overlap with NIK during osteoclastogenesis (such as p100 processing) and which functions, if any, are independent of NIK.

Rescue of RANKL-mediated Osteoclast Differentiation in IKK␣
RANKL and TNF␣ share many common downstream signaling activities. In particular, the ability of both cytokines to activate the transcription factors AP-1 and NF-B, which are required for osteoclastogenesis (7,28,29), indicates the potential for redundant functions of RANKL and TNF␣ in osteoclast precursor cells. One distinction between TNF␣-and RANKLsignaling pathways is the utilization of IKK␣. We and others (19) have demonstrated that RANKL activates p100 processing, which is dependent on IKK␣ and its upstream activator NIK (30). In addition to p100 processing in osteoclast precursor cells, RANKL requires IKK␣ for activation of the p65/p50 NF-B transcription factor in mammary epithelial cells via phosphorylation and degradation of the inhibitor IB␣ (18). In contrast, TNF␣ has no such requirement for IKK␣ during activation of p65/p50 but instead relies solely upon IKK␤ and IKK␥ (31,32). Furthermore, TNF␣ has not been reported previously to induce p100 processing but instead causes accumulation of p100 mRNA and protein via p65/p50-mediated upregulation of the nfkb1 gene (33). Here we have shown that treatment of wild-type but not IKK␣ Ϫ/Ϫ osteoclast precursor cells with TNF␣ does result in a small amount of p52 processing, although the overall net effect is to increase p100 levels compared with p52. RANKL treatment, on the other hand, results in complete processing of p100 into p52. In IKK␣ Ϫ/Ϫ cells, both RANKL and TNF␣ increased p100 levels without FIG. 6. IKK␣ is required for activation of p100 processing by RANKL but not early signaling events. A, wild-type and IKK␣ Ϫ/Ϫ FLC were grown in the presence of 40 ng/ml CSF-1, and RANKL was added at 200 ng/ml at indicated times prior to cell harvest on day 6. Whole cell lysates were used for immunoblot analysis with antibodies to IB␣ (upper panel) and p100/p52 (lower panel). Arrows indicate specific bands detected for each protein. B, cells were treated as described in A. Lysates were immunoblotted with antibodies for phosphoextracellular signal-regulated kinase (p-ERK), phospho-c-Jun NH 2 -terminal kinase (p-JNK), and phospho-p38 (p-p38), and then blots were stripped and reprobed with corresponding total protein antibodies.
formation of p52, effects that have also been observed in NIK Ϫ/Ϫ cells. However, although Novack et al. (19) reported that the result of p100 accumulation was to inhibit osteoclast differentiation in NIK Ϫ/Ϫ cells (19), we found that RANKL and TNF␣ in combination could stimulate osteoclast formation despite the presence of high levels of p100 protein. As previous studies have shown that TNF␣ can synergize with RANKL to activate NF-B and c-Jun NH 2 -terminal kinase (13), it remains possible that these types of signals are sufficient to overcome inhibition of osteoclast differentiation by p100.
TGF␤ has been reported to have both stimulatory and inhibitory effects on osteoclastogenesis (26). In contrast to a recent report indicating that TGF␤ could directly stimulate osteoclast differentiation in the presence of CSF-1 (34), we did not observe any effect of TGF␤ and CSF-1 on TRAP ϩ cell formation in wild-type fetal liver cells after 6 days in culture. However, TGF␤ did augment TNF␣-and RANKL-mediated osteoclast differentiation in wild-type fetal liver cells, which increased the size of the TRAP ϩ cells but not the overall number. Although TGF␤ has been linked to IKK␣ and NF-B p65/p50 activation via the upstream kinase TAK1 (35), we did not observe p100 processing in response to TGF␤ treatment alone in either wildtype or IKK␣ Ϫ/Ϫ cells. Despite a lack of p100 processing, the ability of TNF␣ and TGF␤ to stimulate osteoclast differentiation was IKK␣-dependent. It remains to be determined whether these effects are mediated via TNF␣ or TGF␤ signaling to IKK␣ via TAK1 or some other mediator.
In addition to its multiple roles in diverse tissues, including mammary gland, skin, and lymphocytes, we have now shown a role for IKK␣ during osteoclast differentiation. Ohazama et al. (36) have recently reported that IKK␣ Ϫ/Ϫ mice have defects in tooth development, specifically abnormal cusp formation, a defect that has been linked previously to the TNF receptor family member ectodysplasin A receptor (37). Although osteoclast activity is required for eruption of teeth, it is unlikely that these defects are linked to deficient osteoclast activity because cusp formation is regulated through epithelial cell signaling interactions. IKK␣ Ϫ/Ϫ mice have also been reported to have multiple skeletal abnormalities, including syndactyl and shortened limbs (21,38,39). Taken together, these findings indicate that IKK␣ plays multiple roles in the regulation of skeletal development and may contribute to pathological conditions associated with increased osteoclast activity.