IκB kinase (IKK)β, but not IKKα, is a critical mediator of osteoclast survival and is required for inflammation-induced bone loss

To determine the roles of the IKK catalytic subunits in osteoclastogenesis, we used mice that carry mutant forms of either subunit—Ikkα^AA and Ikkβ^Δ mice. Ikkα^AA mice are homozygous for a knock-in mutant allele, in which the activation loop serines of IKKα—whose phosphorylation is required for its activation (28)—were replaced with alanines; this prevents IKKα activation by upstream stimuli, including RANK (29) and NIK (23). Ikkα^AA mice are viable, healthy, and fertile. These mice exhibit defective organization of secondary lymphoid organs and B cell maturation (30), a defect that is very similar to what was found in Nik^−/− (31) or Nik^aly/aly (26) mice. Ikkβ^Δ mice were generated by crossing Ikkβ^F/F mice, homozygotes for a “floxed” Ikkβ allele (32), with Mx1-Cre transgenic mice that express Cre recombinase from the IFN-inducible Mx-1 promoter (33). Injection of polyinosinic-polycytidylic acid (poly[IC]), which induces IFN production, into Ikkβ^F/F:Mx1-Cre mice results in efficient deletion of the floxed third exon of the Ikkβ gene and generation of IKKβ-deficiency in IFN-responsive cells, including myeloid cells (34). To determine whether defective IKKα activation or complete loss of IKKβ affect osteoclastogenesis, we first used an in vitro differentiation system. BM hematopoietic progenitors that were isolated from WT, Ikkα^AA, and Ikkβ^Δ mice were incubated with M-CSF and RANKL for 7 d, and stained for tartrate-resistant acid phosphatase (TRAP), a marker for osteoclasts. Wt cultures showed robust osteoclastogenesis and formed giant TRAP-positive cells, whereas osteoclast formation was absent in Ikkα^AA and Ikkβ^Δ cultures (Fig. 1). Expression of several other osteoclast markers, including CatK and matrix metalloproteinase-9, also was defective in RANKL-treated Ikkα^AA and Ikkβ^Δ BM cultures (unpublished data). These results suggest that IKKα and IKKβ are required for, and play nonredundant roles in, M-CSF and RANKL-induced osteoclastogenesis in vitro.

To gain insights into the possible mechanisms by which IKKα and IKKβ promote osteoclastogenesis in vitro, we examined RANKL-induced NF-κB activation in Ikkα^AA and Ikkβ^Δ BM cultures. BM-derived precursors from WT, Ikkα^AA, and Ikkβ^Δ mice were incubated with M-CSF and RANKL and analyzed for expression of IKKα and IKKβ by immunoblot analysis (Fig. 2 A). As expected, no IKKβ protein was detectable in Ikkβ^Δ BM cells. Interestingly, the deletion of IKKβ also resulted in slightly decreased levels of IKKα (Fig. 2 A); this suggests that IKKα:IKKγ complexes or IKKα homodimers are not as stable as the heterotrimeric IKKα:IKKβ:IKKγ complexes (22). Electrophoretic mobility shift assays revealed marked induction of NF-κB DNA binding activity in response to RANKL treatment in WT and Ikkα^AA cultures, but only a weak response in Ikkβ^Δ cultures (Fig. 2 B). Similarly, RANKL activated IKK in WT and Ikkα^AA cultures, but no IKK activity could be detected before or after RANKL treatment of Ikkβ^Δ BM cells (Fig. 2 C). In addition, RANKL induced the nuclear translocation of all three Rel proteins in WT cells; this response was slightly reduced for RelB and cRel, but not for RelA, in Ikkα^AA cells (Fig. 2 D). By contrast, the basal level of all three Rel proteins in the nucleus and their RANKL-induced nuclear translocation were diminished severely in Ikkβ^Δ cultures. Although IKKα plays a minor role in NF-κB activation, other experiments that were conducted with Ikkα^AA BM cells revealed its requirement for induction of NF-κB2/p100 processing to p52 in response to RANKL (Fig. 2 E), as previously shown for several other TNF family members (30).

To determine if IKKα and IKKβ play a role in osteoclastogenesis in vivo, bones from 4 mo-old Ikkα^AA and Ikkβ^Δ mice were analyzed. To study the role of IKKβ in bone development we induced deletion of the Ikkβ gene in Ikkβ^F/F:Mx1-Cre mice 9 d after birth. Histologic and histomorphometric analyses of long bone (tibias) sections from Ikkα^AA mice revealed no differences compared with WT bones, and no alterations were found in trabecular size and distribution or in the number of osteoclasts as determined by TRAP staining (Fig. 3 A–C). These results suggest that basal osteoclastogenesis is not affected by the loss of IKKα activation in vivo. By contrast, Ikkβ^Δ mice showed an osteopetrotic phenotype. Ikkβ^Δ mice are smaller and have a hunched back compared with their Ikkβ^F/F littermates (unpublished data). Histologic analysis of tibias from these mice showed increased trabecular size and distribution which resulted in obliteration of the bone marrow cavity compared with Ikkβ^F/F littermates or WT mice of the same age (Fig. 3 A). TRAP staining showed a greatly reduced number of osteoclasts in Ikkβ^Δ bones (Fig. 3, B and C). Furthermore, quantitative histomorphometric analyses of Ikkβ^Δ bones revealed a significant increase in bone volume due to increased trabecular number compared with bones of WT littermates (Fig. 3 C); this is diagnostic of reduced osteoclast-mediated bone resorption. The osteoclast parameters, number and surface, are reduced significantly in Ikkβ^Δ mice (Fig. 3 C). It also seems that osteoblast number is reduced at 7 mo in the mutant mice, a defect that was observed in other osteopetrotic mouse models (35). The osteopetrotic phenotype of Ikkβ^Δ mice becomes more dramatic with age (Fig. 3 C). These results suggest that Ikkβ^Δ mice develop osteopetrosis that is due to defective osteoclast formation, and furthermore, indicate a critical role for Ikkβ in bone development.

Osteoclasts also can be generated in vitro from BM hematopoietic precursors cultured in the presence of 1,25(OH)₂-vitamin D3 and dexamethasone together with osteoblastic/stromal cells from mouse calvarias (36). In such a system, activated osteoblasts play the major role in osteoclast differentiation and provide M-CSF and RANKL directly. To examine whether the osteoclastogenic defects that are observed in Ikkα^AA and Ikkβ^Δ BM cultures are caused by cell autonomous defects, we performed osteoclasts/osteoblast cocultures. Wt osteoblasts fully supported osteoclastogenesis of Ikkα^AA osteoclast precursors, but not Ikkβ^Δ osteoclast precursors; this indicates a cell–autonomous osteoclast differentiation defect in Ikkβ^Δ BM cells (Fig. 4 A, upper panels). Reciprocal cocultures of Ikkα^AA or WT BM cells with Ikkα^AA osteoblasts showed that Ikkα^AA osteoblasts had the same capacity as WT osteoblasts for supporting osteoclast differentiation (Fig. 4, A [lower panels] and B). Thus, Ikkα^AA osteoclasts can form normally as long as they receive osteoblast-derived signals; this suggests that these signals compensate for the defect in RANKL signaling.

To identify potential signals that may compensate for the defect in RANKL signaling of Ikkα^AA osteoclast precursors, we examined the effect of the proinflammatory cytokines, IL-1 and TNFα. IL-1 and TNFα are potent osteoclastogenic factors and are likely to be involved with RANKL in inflammation-induced bone loss (10, 37). We cultured osteoclast progenitors from WT, Ikkα^AA, and Ikkβ^Δ mice in the presence of IL-1 or TNFα, alone or together with RANKL. IL-1 and TNFα strongly augmented the osteoclastogenic response of WT BM to RANKL and led to the formation of numerous TRAP-positive giant cells (Fig. 5 A). Either IL-1 or TNFα, when combined with RANKL, completely rescued the osteoclastogenic defect of Ikkα^AA BM, although they did not induce differentiation on their own (Fig. 5 A and not depicted). However, the defect in osteoclast differentiation of Ikkβ^Δ BM cultures could not be rescued by IL-1 or TNFα (Fig. 5, A and B). The bone-resorbing activity of these cells was assayed by an in vitro resorption pit assay and further quantified. Ikkα^AA, but not Ikkβ^Δ, BM cells were able to resorb bone, once their differentiation defect was rescued by IL-1 (not depicted) or TNFα (Fig. 5 C). Furthermore, we found that when incubated with TNFα, either in the absence or presence of RANKL, most Ikkβ^Δ osteoclast precursors were dead within 48 h (Fig. 5 A). This prompted us to examine whether Ikkβ^Δ BM cells undergo apoptosis in response to TNFα. Terminal deoxynucleotidyl transferase–mediated dUTP nick-end labeling (TUNEL) assay revealed the appearance of cells with distinct apoptotic morphology and fragmented DNA within 24 h of TNFα addition to Ikkβ^Δ osteoclast precursors (Fig. 5, D and E). Very few such cells were detected in cultures that were not exposed to TNFα or in WT osteoclast precursors that were incubated with TNFα.

TNFR1 contains a death domain and is capable of engaging the apoptotic machinery (38). To further examine the mechanism that underlies TNFα-induced death of Ikkβ-deficient preosteoclasts, we crossed Ikkβ^Δ mice with Tnfr1^−/− mice to generate Ikkβ^F/F:Mx1-Cre:Tnfr1^−/− double mutants. After poly(IC) injection, BM cultures that were isolated from these mice were stimulated with RANKL alone or with RANKL plus TNFα; osteoclast differentiation was analyzed by TRAP staining. Loss of TNFR1 prevented TNFα-induced death of Ikkβ^Δ osteoclast progenitors (Fig. 5 A, bottom). Furthermore, the absence of TNFR1 rescued the inability of Ikkβ^Δ cells to become TRAP-positive, but did not allow them to differentiate fully into multinucleated giant osteoclasts (Fig. 5 A) with the ability to resorb bone (Fig. 5 C). Thus, IKKβ is required for the prevention of TNFα-induced death and is needed for formation of fully functional bone-resorbing osteoclasts.

Next we analyzed femurs of 4-mo-old WT, Ikkβ^Δ, Tnfr1^−/−, and Ikkβ^Δ:Tnfr1^−/−mice. Ikkβ deletion in both strains was induced as early as 9 d after birth. Histochemical staining revealed osteopetrosis in Ikkβ^Δ:Tnfr1^−/− bones (Fig. 6 A). However, TRAP-positive cells could be detected easily in Ikkβ^Δ:Tnfr1^−/− mice, whereas they were scarce in Ikkβ^Δ mice (Fig. 6 B). Analysis of deoxypiridinoline cross-links in urine samples—which reflects osteoclast activity in vivo—revealed that in Ikkβ^Δ:Tnfr1^−/− mice, osteoclast resorption activity remained severely attenuated, as in Ikkβ^Δ mice. Normal levels of deoxypiridinoline cross-links were found in urine samples from Ikkα^AA or Tnfr1^−/− mice.

Immunohistochemical analyses by hematoxylin-eosin (H&E) staining revealed a small number of osteoclast precursors in the vicinity of the growth plate that were positive for F4/80 staining in Ikkβ^Δ mice, whereas many more such cells were seen in WT mice (Fig. 6 D). Remarkably, the loss of TNFR1 in Ikkβ^Δ:Tnfr1^−/− double mutant mice restored the presence of F4/80 positive cells next to the growth plate (Fig. 6 D). Double staining experiments revealed that in Ikkβ^Δ mice, most of the few F4/80-positive cells that were present were apoptotic, based on TUNEL staining (Fig. 6 E). This further supports the hypothesis that the decreased osteoclast number in Ikkβ^Δ mice is due to increased apoptosis of their precursors. Control experiments showed that in other tissues, such as the liver, the number of F4/80-positive cells did not change in the absence of Ikkβ (Fig. 6 F). Despite the decreased apoptosis of osteoclast precursors and restoration of TRAP-positive cells, the Ikkβ^Δ:Tnfr1^−/− double mutant mice remain osteopetrotic. This suggests that osteoclasts that lack IKKβ and TNFR1 are not fully functional under basal conditions; this interpretation is consistent with the severely reduced bone-resorbing activity in these mice (Fig. 6 C).

IKKβ plays a crucial role in osteoclast differentiation under physiologic conditions. We next analyzed whether IKKβ or IKKα is involved in inflammation-induced bone loss. We used an established model of endotoxin-induced bone resorption (39, 40). Ikkα^AA, Ikkβ^Δ, Tnfr1^−/−, Ikkβ^Δ:Tnfr1^−/−, and WT mice were injected with 500 μg LPS in saline into the synovial space of the hind limb knee joint, whereas the contralateral knee was injected with saline alone; mice were killed 5 d after injection. Ikkβ deletion was induced 10 d before knee injection. Before sacrificing the mice, their ability to move and flex their hind limb was assessed by video cinematography (Videos 1 and 2). The staining of the joint bones (femurs and tibia) with H&E, TRAP, and F4/80 showed a considerably lower number of osteoclasts and osteoclast precursors in Ikkβ^Δ mutant mice compared with WT controls or Ikkα^AA mice (Fig. 7, A–D; not depicted for Ikkα^AA). Although LPS-injected WT, Ikkα^AA, and Tnfr1^−/− mice completely lost the ability to flex the hind limb, no such aberrations were evident in Ikkβ^Δ mice, which retained normal flexibility and movement of the LPS-treated hind limb (Videos 1 and 2). The loss of TNFR1 restored inflammation-induced bone loss in mice that lack Ikkβ^Δ (Fig. 7), although the bone damage is not as extensive as in WT mice (compare resorption sites indicated by arrows in Fig. 7 A). The degree of inflammation-induced bone loss was quite similar in Ikkβ^Δ:Tnfr1^−/− and Tnfr1^−/− mice. These results suggest that under strong inflammatory conditions, which are likely to result in massive cytokine production, IKKβ is no longer needed for osteoclastogenesis once its survival function has been rendered unnecessary by ablation of TNFR1.

Previous studies that were based on ablation of the p50 and p52 NF-κB members outlined an important role for these transcription factors in osteoclast differentiation (15). NF-κB is activated by RANKL, a cytokine whose expression and binding to the receptor, RANK, are essential for osteoclastogenesis. NF-κB also is activated by the proinflammatory cytokines, TNFα and IL-1 (21), which enhance inflammation-induced bone loss, although they are not essential for developmental osteoclastogenesis (41, 42). The activation of NF-κB by all of these cytokines depends on integrity of the IKK complex (21); recent results show that a small peptide that can prevent binding of the IKKα and IKKβ catalytic subunits to the IKKγ regulatory subunit can inhibit inflammation-induced bone loss (27). Although these results provide further support for the role of NF-κB in osteoclastogenesis, it has not been established which of the two IKK catalytic subunits plays a more critical role in basal osteoclastogenesis and inflammation-induced bone loss. For instance, it recently was described that the NIK- and IKKα-dependent alternative pathway is required for RANKL-induced osteoclast differentiation in vitro (25). Our results demonstrate a critical role for IKKβ, but not IKKα, in basic osteoclastogenesis in vivo and in inflammation-induced bone loss.

Although IKKα and IKKβ are important transducers of osteoclastogenic signals which emanate from RANKL in vitro, the mutation that prevents IKKα activation by its upstream kinase, NIK (23), had no effect on osteoclast formation in vivo and did not increase bone density. Furthermore, Ikkα^AA mice were fully sensitive to inflammation-induced bone loss (unpublished data). These results are consistent with the small effect of the Ikkα^AA mutation on RANKL-induced translocation of NF-κB proteins to the nucleus, as well as the absence of an osteopetrotic phenotype in mice defective in NIK, the upstream activator of IKKα (25). Like BM progenitors from Nik^−/− mice, Ikkα^AA BM progenitors do not differentiate in response to RANKL when cultured in the absence of osteoblasts. However, either osteoblasts or the proinflammatory cytokines, IL-1β and TNFα, together with RANKL induce osteoclastogenesis of Ikkα^AA BM progenitors. Thus, although IKKα contributes to RANKL-induced differentiation in vitro, its function in vivo is dispensable because of the action of other factors that activate the IKKβ-driven classic NF-κB pathway. During normal bone development, these factors are likely to be derived from osteoblasts, whereas during inflammation these factors could be TNFα and IL-1β.

In contrast with IKKα, our findings illustrate a critical function for IKKβ. IKKβ-deficient BM progenitors do not form osteoclasts in vitro in response to RANKL or when cocultured with osteoblasts. Furthermore, their inability to respond to RANKL cannot be complemented by IL-1 or TNFα, and Ikkβ^Δ mice are osteopetrotic. IKKβ-deficient BM progenitors are extremely sensitive to TNFα and undergo extensive apoptosis, despite the presence of the myeloid survival factor, M-CSF. Because TNFα-induced apoptosis of IKKβ-deficient preosteoclasts is prevented by the loss of TNFR1, we propose that one of the mechanisms by which IKKβ-dependent NF-κB activation contributes to osteoclastogenesis in vivo, especially during inflammation, is through prevention of TNFα-induced apoptosis of osteoclast progenitors. In the absence of IKKβ, such cells become very sensitive to TNFα and are eliminated when TNFα is produced in sufficiently large amounts. Nonetheless, although the death of Ikkβ-deficient osteoclast progenitors is prevented by loss of TNFR1 and Ikkβ^Δ:Tnfr1^−/− double mutants display close to normal numbers of TRAP-positive cells in their bones (Fig. 6 B), these mice become osteopetrotic (Fig. 6 A) if Ikkβ deletion is induced early during bone development. The explanation for these results is that IKKβ-deficient osteoclasts remain defective in bone resorption, even when their TNFα-induced elimination does not take place. Our in vitro results (Fig. 5 A) suggest that Ikkβ^Δ:Tnfr1^−/− progenitors cannot give rise to fully differentiated osteoclasts, although they do become TRAP-positive in response to RANKL. Morphologically, such cells look like monocytes, rather than multinucleated giant cells; this suggests that their ability to undergo cell fusion is eliminated. Similar defects in basal osteoclast functions were observed in Traf6^−/− and Src^−/− mice (16, 43). These mice are osteopetrotic as a result of the presence of osteoclasts that are unable to form ruffled borders, and therefore, are defective in bone resorption.

Thus, in addition to the prevention of TNFα-induced apoptosis, IKKβ is required for terminal osteoclast differentiation. Although IKKβ-dependent NF-κB activation is essential for this process, it is not sufficient; potent NF-κB activating cytokines, such as TNFα, cannot substitute for RANKL (Fig. 7 E). Most likely, another pathway or factor, designated X in Fig. 7 E, needs to be switched on—along with IKKβ and NF-κB—for terminal osteoclast differentiation to take place. Nonetheless, our results illustrate the potential ability of IKKβ inhibition to prevent inflammation-driven bone destruction. Again, the mechanism through which IKKβ inhibition prevents inflammation-induced bone loss involves sensitization of osteoclast progenitors to TNFα-induced apoptosis, because Ikkβ^Δ:Tnfr1^−/− mice are fully susceptible to inflammation-induced bone loss. The bone-resorbing ability of osteoclasts in Ikkβ^Δ:Tnfr1^−/− mice seems to be restored after LPS injection. Because these cells do not respond to TNFα (as a result of the loss of TNFR1), the inflammation-induced factor that may stimulate their bone-resorbing activity could be IL-1β, which is known to be induced by LPS administration (44). However, in vitro IL-1β is unable to induce the formation of multinucleated bone-resorbing cells when given together with RANKL once IKKβ is absent (Fig. 5 A). Alternatively, Ikkβ^Δ:Tnfr1^−/− osteoclast progenitors may be only partially defective in their ability to respond to RANKL; this defect may be eliminated when high levels of RANKL are present along with other proinflammatory cytokines. LPS-induced inflammation results in induction of RANKL along with other cytokines (45).

Our results support a role for IKKβ as an important regulator of bone homeostasis and a mediator of inflammation-induced bone loss. Our results also suggest that the major mechanism through which deletion or inhibition of IKKβ exerts its therapeutic effect in inflammation-induced bone loss is by predisposing osteoclast precursors to TNFα-induced apoptosis. A schematic model that summarizes our findings is presented in Fig. 7 E. Binding of RANKL to its receptor, RANK, induces a cascade of events that leads to activation of IKKβ and at least one more factor—that together with IKKβ—is required for induction of terminal osteoclast differentiation. During inflammation, proinflammatory cytokines, such as TNFα and IL-1β, are induced and strongly potentiate RANKL-induced osteoclastogenesis, although such factors cannot induce osteoclast differentiation on their own. TNFα signaling through TNFR1 has the potential to induce apoptosis through caspase 8, a process that is prevented by IKKβ-dependent NF-κB activation (46). Once IKKβ is inhibited, TNFα-induced apoptosis results in elimination of Ikkβ-deficient osteoclast progenitors, and thereby, prevents inflammation-induced bone destruction. Thus, IKKβ inhibition presents a logical strategy for prevention of numerous bone-resorbing disorders that are triggered by inflammation, such as rheumatoid arthritis.

Ikkα^AA and Ikkβ^F/F mice were generated as described (29, 32). To delete IKKβ in hematopoietic cells, Ikkβ^F/F mice were crossed with Mx1-Cre transgenic mice (33), and Ikkβ^F/F:Mx1-Cre progeny were injected three times with poly(IC) every 2 d. Injections started at day 9 after birth for in vivo analysis, whereas injections were performed 10 d before killing mice or collecting BM cells, respectively. Deletion of Ikkβ was confirmed by PCR, whereas the absence of IKKβ protein was examined by immunoblotting. These mice are referred to as Ikkβ^Δ mice. All experimental procedures were approved by the Animal Subjects Committee at the University of California San Diego, according to U.S. National Institutes of Health guidelines.

Tissues were fixed in PBS-buffered 4% formaldehyde, embedded in paraffin, sectioned at 5 μm, and stained as indicated using standard techniques. Calcified tissues were decalcified in EDTA (0.5 M, pH 8) for 12 d before embedding. TRAP staining was performed using a leukocyte acid phosphatase kit (Sigma-Aldrich). For histomorphometry, tibiae were embedded in methacrylate (Echnovit; Heraeus Kulzer) without previous decalcification and 3–4-μm sections were stained with Goldner trichrome. Histomorphometry of metaphyses was performed using an Axioskop 2 microscope (Carl Zeiss MicroImaging, Inc.) and OsteoMeasure Analysis System (OsteoMetrics) according to international standards (47). TUNEL assay and immunohistochemistry were as described previously (48, 49).

BM cells from 6-wk-old mice were plated in the presence of 5 ng/ml recombinant M-CSF for 24 h Nonadherent cells were replated in the presence of recombinant M-CSF (10 ng/ml) and recombinant-RANKL (50 ng/ml) for 7 d and then fixed and stained for TRAP activity. For biochemical analysis, BM cells were plated in the presence of M-CSF (10 ng/ml) for 6 d, collected, and counted. Osteoclast activity was assayed in vitro after differentiation on a calcium phosphate film (BioCoat Bone Cell Culture System, Osteologic) and the resorpted area was quantified using AxioVision 4.3 software. Osteoclast activity in vitro was measured in urine samples using RatLaps ELISA (Nordic Bioscience Diagnostics).

Primary osteoblasts were isolated from calvarias of neonatal (2–4-d-old) WT and Ikkα^AA mice and were digested for 10 min in modified Eagle's medium (α-MEM) which contained 0.1% collagenase and 0.2% dispase. Cells from two to five mice were combined as an osteoblastic cell population and plated at a density of 5 × 10⁵ cells/ml in α-MEM with 10% FCS for 24 h. Cocultures were performed as described (36, 50). Briefly, BM cells (10⁶ per well) were added to primary osteoblasts (5 × 10⁵ cells per well) and cultured in α-MEM which contained 10% FCS, 10⁻⁸ M 1,25(OH)₂-vitamin D3, and 10⁻⁷ M dexamethasone in 24-well plates. Intensity of TRAP staining was measured by Image Pro Plus 5.1.

Cells were resuspended in buffer L1 (50 mM Tris-Cl, pH 8.0, 2 mM EDTA, 0.1% NP-40, 10% glycerol) that contained protease inhibitors, incubated for 5 min at 4°C, and centrifuged for 5 min at 4,000 revolutions/min in a microcentrifuge. Cytoplasmic supernatants were stored and nuclear pellets were extracted further in buffer L2 (20 mM Hepes-KOH, pH 7.6, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 10% glycerol, 25 mM β-glycerophosphate) which contained protease inhibitors. After lysis, nuclei were centrifuged at 15,000 revolutions/min and the supernatant was collected for further analysis. Nuclear (5 μg) and cytoplasmic (20 μg) extracts were electrophoresed, transferred to nitrocellulose membranes, and immunoblotted with anti-IKKα (Imgenex), anti-IKKβ (UBI), anti-RelA, anti-RelB, and anti-c-Rel (Santa Cruz Biotechnology, Inc.) antibodies. p100 processing in total extracts was assayed using anti–NF-κB/p52 K-27 antibody (Santa Cruz Biotechnology, Inc.).

IKK immunocomplex kinase assay was as described (51), except that an IKKγ antibody (BD Biosciences) was used for immunoprecipitation. Electrophoretic mobility shift assay for NF-κB was described previously (23).

2- to 3-mo-old mice were given an intrajoint injection of Escherichia coli LPS (Sigma-Aldrich), 500 μg in saline, and a vehicle control into the contralateral joint, 5 d after the last poly(IC) treatment. 5 d later, mice were killed and joint histology was examined in the Histologic and histomorphometric analyses section.

Data are expressed as mean ± SEM. Differences were analyzed by Student's t test.

Ikkβ^F/F (Video 1) and Ikkβ^Δ (Video 2) mice were given an intrajoint injection of E. coli LPS (Sigma-Aldrich), 500 μg in saline, (left hind limb) and a vehicle control (right hind limb). The two videos show the ability of the mice to move and stretch their hind limbs.

The authors would like to thank Drs. A. Hoebertz and L. Kenner for helpful discussion and A. Chang for LPS knee joint injections.

S. Maeda, J.M. Park, L.-C. Hsu, and Y. Cao were supported by postdoctoral fellowships from the Japan Society for the Promotion of Science, Bristol-Myers Squibb Foundation Research Fellowship at the Irvington Institute, the Cancer Research Institute, and an American Association for Cancer Research–Genentech BioOncology Career Development Award for Cancer Research. Support was provided by National Institutes of Health grants nos. ES06376 and AI43477 (to M. Karin). In addition, M. Karin is an American Cancer Society Research Professor.

The authors have no conflicting financial interests.