NF-κB Signaling Participates in Both RANKL- and IL-4–Induced Macrophage Fusion: Receptor Cross-Talk Leads to Alterations in NF-κB Pathways

Macrophages reside in all tissues and undergo a homotypic fusion process under specific conditions. There are two major types of macrophage-derived multinucleated giant cells (MNG): osteoclasts in bone and MNG found in chronic inflammatory reactions (1). The osteoclast resorbs mineralized matrix, whereas the MNG engulfs large foreign bodies. Macrophage fusion is mediated through protein–protein interaction. Several proteins important for macrophage fusion have been identified, including macrophage fusion receptor, CD47, CD44, E-cadherin, CD36, and DC-STAMP (2–8); however, the molecular mechanism controlling macrophage fusion and the pathways that determine whether macrophages will form MNG or osteoclasts remain poorly understood.

Receptor activator for NF-κB ligand (RANKL) signals macrophages to differentiate into osteoclasts and undergo fusion, forming large, mutlinucleated cells expressing high levels of the enzyme tartrate-resistant acid phosphatase (TRAP) (9). This process is dependent on induction of NFATc1 and NF-κB. RANKL can activate NF-κB by either the canonical or alternative pathways (10). The canonical pathway is mediated by the NF-κB essential modulator (NEMO)-dependent activation of IκB kinase (IKK) 2, leading to the phosphorylation and degradation of the classic inhibitors of κB (IκB α, ε). Degradation of IκB leads to the translocation of the classic NF-κB dimer consisting of p50 and p65/RelA. The p50 subunit is derived from the nfkb1 gene that also encodes p105. The activation of this pathway is very rapid and independent of new protein synthesis. The alternative pathway is mediated by the NF-κB–inducing kinase (NIK)-dependent activation of IKK1 that leads to the phosphorylation and partial degradation of the p100 precursor (derived from the nfkb2 gene) to generate the p52 subunit. p100 also acts as an IκB, and thus, its degradation allows the generation of p52 and the release of NF-κB dimers including p52/RelB (11). Mice lacking expression of nfkb1 or nfkb2 have modest defects in osteoclastogenesis, whereas the double-deficient mice have severe osteoporosis caused by the absence of osteoclasts (12).

In contrast to RANKL, IL-4 signals macrophages to differentiate to the alternative type (13, 14) and to undergo fusion, forming MNG expressing high levels of arignase I and chitinases (15). This IL-4–induced macrophage differentiation and MNG formation is dependent on STAT6 (15, 16). The IL-4–induced expression of the fusion-related molecules E-cadherin and DC-STAMP were also STAT6 dependent. These fusion molecules are important for the formation of both osteoclasts and MNG (15, 17). In the presence of both RANKL and IL-4, the IL-4–induced differentiation was dominant. IL-4 suppressed the RANKL-induced osteoclast differentiation and promoted MNG formation in a STAT6-dependent manner (15, 16).

The effect of IL-4 on the activation of NF-κB in this setting is unclear. There are several studies reporting that IL-4 inhibited osteoclast differentiation through inhibition of the RANKL-induced activation of the classical NF-κB pathway (18–20). However, in our report, we did not observe a direct effect of IL-4 on RANK signaling (16). Others have shown that NF-κB is essential for macrophage fusion induced by culture at high cell density (21) and for the expression of fusion-related genes such as DC-STAMP (17).

Therefore, in this study, we carefully investigated the role of NF-κB in IL-4–induced MNG formation and whether IL-4 regulates RANKL-induced NF-κB activation. We found that NF-κB signaling is also required for IL-4–induced MNG formation. IL-4 treatment did not inhibit the RANKL-induced activation of the NF-κB canonical pathway but inhibited the activation of the alternative pathway. In addition, IL-4 induced the expression of p105/p50 and stimulated the accumulation of p50 in the nucleus in an NIK-dependent manner. Overexpression of NF-κB dimer proteins enhanced macrophage fusion, whereas knockdown of p100 enhanced macrophage fusion. These results indicate that NF-κB pathways regulate macrophage fusion induced by IL-4 treatment, in addition to RANKL-induced osteoclast formation, although the specific pathways are different.

Nonadherent bone marrow mononuclear cells were isolated from femurs and tibias of 4–8-wk-old female wild-type C57BL/6 (Taconic Laboratories) mice, nfκb1-deficient mice on the C57BL/6 background (provided by Dr. Ranjan Sen, National Institute of Aging, National Institutes of Health, Baltimore, MD), or NIK-deficient mice on the C57BL/6/129 background [developed by Dr. Robert Schreiber (22) and provided by Dr. Barry Sleckman, Washington University, St. Louis, MO]. Cells were cultured overnight in α-MEM media (BioWhittaker) to deplete adherent stromal cells. These nonadherent bone marrow mononuclear cells were cultured for 3 d in α-MEM media supplemented with penicillin, streptomycin, glutamine (BioWhittaker), 10% FBS, and 20 ng/ml recombinant mouse M-CSF (R&D Systems, Minneapolis, MN) to generate bone marrow-derived macrophages (BMM). Mature osteoclasts were obtained after 5 to 6 d culture in the presence of 20 ng/ml M-CSF and 150 ng/ml RANKL (provided by Dr. Mehrdad Tondravi, National Cancer Institute). BMM were also treated with 20 ng/ml recombinant mouse M-CSF and 10 ng/ml IL-4 (R&D Systems) to generate MNG. In the RANKL plus IL-4 group, BMM were treated with 20 ng/ml recombinant mouse M-CSF, 150 ng/ml RANKL, and 10 ng/ml IL-4 for various time points.

Osteoclasts were identified by high levels of TRAP staining. TRAP staining was performed as previously described (23). MNG were identified using crystal violet staining. Cells were fixed and stained with 1% crystal violet in 95% ethanol, washed with water, and air dried as described (15).

Total cell lysate preparation and immunoblotting were performed as described previously (16). Briefly, cell lysates were fractionated on 4–20% or 7.5% SDS-PAGE gels and electrophoretically transferred to polyvinylidene difluoride membranes (Millipore). Membranes were blocked and probed with rabbit anti-p105/50 (Abcam, Cambridge, MA), rabbit anti-p100/52 (Cell Signaling Technology, Danvers, MA), rabbit anti-p65 (RelA), -RelB, -STAT6, -lamin A, anti-IκBα, and mouse anti-actin, all from Santa Cruz Biotechnology (Santa Cruz, CA). The ratio of p100/p52 was calculated by densitometric analysis using Quantity One software (Bio-Rad).

Cells were lysed in lysis buffer A (10 mM HEPES [pH 7.9], 10 mM KCl, 10 mM EDTA, 1 mM DTT, 0.4% IGEPAL, 1 μg/ml pepstatin A, leupeptin, and aprotinin). The supernatant was collected by centrifugation and considered the cytoplasmic fraction. The precipitate was solubilized in nuclear lysis buffer B (20 mM HEPES [pH 7.9], 400 mM NaCl, 1 mM EDTA, 10% glycerol, 1 mM DTT, 1 μg/ml pepstatin A, leupeptin, and aprotinin), centrifuged, and the nuclear fraction was collected as supernatant lysates.

An NF-κB consensus oligonucleotide derived from the TNF promoter κB3 site (24) was labeled with biotin using the 3′ end DNA labeling kit (Pierce). The consensus probe sequence is 5′-AACAGGGGGCTTTCCCTCCT-3′. EMSA assay was performed using a chemiluminescent EMSA kit (Pierce). Briefly, 2 μg nuclear lysate was incubated with the probe in reaction buffer [1× binding buffer, 2.5% glycerol, 5 mM MgCl₂, 50 ng/μl poly(deoxyinosinic-deoxycytidylic) acid, and 0.05% Nonidet P-40] for 30 min. Reactants were loaded onto a 6% native polyacrylamide gel. The bands were transferred onto positively charged nylon membrane (Roche). The DNA was cross-linked by UV cross-linker. The membrane was incubated and exposed according to the kit instructions.

Total RNA was extracted from cultured cells using the RNeasy kit (Invitrogen). First-strand cDNA was transcribed from 2 μg RNA with the Superscript RT kit (Invitrogen). Quantitation of the target gene relative to that of β-actin was determined using the SYBR green dye method and detected on an ABI7700/SDS platform (Applied Biosystems). The primers used were the following: p100/52, 5′-GCCCGAGCGTTGCTGGACTA-3′ and 5′-CGCCGAGGTACTGAGCGTGAT-3′, 184 bp; RelB, 5′-ACACCGGGTACACCCACATAGC-3′ and 5′-TGGATGCCCAGGTTGTTAAAGC-3′, 114 bp; DC-STAMP, 5′-CGCTGTGGACTATCTGCTGT-3′ and 5′-CACTGAGACGTGGTTTAGGAAT-3′, 186 bp; E-cadherin, 5′-TCCAACAGGGACAAAGAAACAA-3′ and 5′-TGACACGGCATGAGAATAGAGG-3′, 171 bp; and CD44, 5′-CCCAGCCAGTGACAGGTTC-3′ and antisense 5′-CGGCAGGTTACATTCAAATCG-3′, 193 bp.

p50, p65, p52, RelB, and p100 were generated by PCR using cDNA plasmids containing p50, p65 [provided by Dr. Janet Stavnezer from the University of Massachusetts (25)], NF-κB2, and RelB (Open Biosystems) as templates. All constructs were cloned into the PLVX-puro lentiviral vector (developed by Clonetech, provided by Dr. Ricardo Feldman from the University of Maryland). p50 and p65 were also cloned into the pcDNA3.1 myc-HisB vector (Invitrogen). The primers for p52 (5′-ATAACTCGAGATGGACAATTGCTACGATCCAG-3′ and 5′-ATAAGGGCCCTCA TCGCTGCAGTGTGTCAAGGGC-3′) were designed to introduce a XhoI site at the 5′ end, a termination codon after the aa 442, and an ApaI site at 3′ end (26). PLVX-puro, PsPax2, and PMD.2G (developed by Dr. Didier Trono [Ecole Polytechnique Federale de Lausanne, Lausanne, Switzerland], provided by Dr. Ricardo Feldman from the University of Maryland) were cotransfected to 293T cells by using Fugene6 reagent (Roche). For the short hairpin RNA (shRNA) experiment, p100/52-PLKO.1 lentivirus vector (Sigma-Aldrich), PsPax2, and PMD.2G were cotransfected to 293T cells. After 24 h, transfection medium was replaced with fresh medium. Sodium pyruvate (1 mM) and HEPES (25 mM) were added into the fresh medium. Viral supernatant was collected at 48 and 72 h postinfection; the supernatant was filtered through a 0.45-μm filter and incubated with primary BMM for 72 h. Then the infection medium was replaced by induction medium composed of 150 ng/ml RANKL alone, 10 ng/ml IL-4 alone, or RANKL plus IL-4, respectively. The cells were kept in culture for 5 d and harvested for staining, real-time RT-PCR, cell fractionation, and immunoblot. p50, p65 pcDNA3.1 myc-HisB vectors were transfected to 293T cells using Fugene6 reagent (Roche). After 24 h, total cell lysate was harvested for EMSA assay as previously described (27).

BMM were treated with various concentrations of in-solution IKK2 inhibitor (SC-514; Calbiochem), IKKγ NEMO binding domain (NBD) inhibitory peptide, or control peptide (IMG-2000; Imgenex) for 5 d. Cells were stained for TRAP and crystal violet, respectively.

Samples were fixed and pictured at room temperature on a Nikon Eclipse E400 microscope using a 4× (plan APO) or 10× (plan fluor) objective lens (Nikon). Digital images were captured using a CCD camera and processed using CoolSNAP version 1.2 software (Roper Scientific). Three random fields (×40) per well were counted. The average area per giant cell and the total area of giant cells were measured as described (15) using ImagePro plus software (Media Cybernetics). The average size of giant cells was obtained by dividing the total giant cell size by the total giant cell number. The results are shown as the mean ± SD of triplicate fields.

Results were analyzed by using Microsoft Excel (Microsoft). The p values were calculated using a paired Student t test. Statistical significance was determined when a p value was <0.05.

Several groups reported that IL-4 was capable of inhibiting the rapid RANKL-induced activation of NF-κB; however, we did not observe any effect of IL-4 on RANK signaling (16, 19, 20). To examine whether IL-4 influenced the RANKL-induced canonical NF-κB pathway, BMM was treated with cytokine combinations for 30 min (Fig. 1). As expected, we found that RANKL induced the degradation of IκBα and the nuclear translocation of p50 (Fig. 1A); however, the presence of IL-4 had no effect on these rapid RANKL-induced responses. EMSA assay was also performed to analyze NF-κB DNA-binding activity. The sequence of the TNF κB3 site was used as the NF-κB probe, which has been shown to interact with high affinity with both p50–p65 heterodimers and p50 homodimers (24). Total lysate of 293T cells cotransfected with myc-tagged p50 and p65 cDNA was run simultaneously with BMM nuclear lysate to identify the NF-κB dimers (Supplemental Fig. 1). RANKL induced NF-κB dimer formation in primary BMM after 30 min; concurrent addition of IL-4 did not inhibit the RANKL-induced NF-κB dimer formation (Fig. 1B). Treatment of primary BMM with RANKL rapidly induced the translocation of p50 and p65 to the nucleus; IL-4 did not inhibit the RANKL-induced translocation of either p50 or p65. IL-4 treatment induced STAT6 translocation to the nucleus. We also performed IL-4 pretreatment experiments. We found that pretreatment of BMM with IL-4 for 1, 2, or 24 h did not suppress the translocation of p50 or p65 induced by RANKL (Fig. 1C). These results indicate that IL-4 does not block the rapid activation of the canonical NF-κB pathway by RANKL.

We also examined longer time periods of stimulation. After longer time periods (48 h), the RANKL-induced NF-κB binding was barely detectable (Fig. 2A). Surprisingly, IL-4 treatment in the presence or absence of RANKL induced a single NF-κB band that comigrated with the p50 homodimer position in the control 293T cell lysates (Fig. 2A). RANKL induced the accumulation of p50 in the nucleus at these longer times of stimulation (24 and 48 h); however, there was very little p65 remaining in the nuclear fraction. IL-4 enhanced the expression of p105/50 in the cytoplasm in the presence or absence of RANKL and increased the amount of p50 in the nucleus after 48 h (Fig. 2). These results indicate that IL-4 upregulates p105/50 expression and enhances the accumulation of p50 in the nucleus.

We next examined the effect of IL-4 on the alternative NF-κB pathway. We found that IL-4 reduced the RANKL-induced nuclear translocation of NF-κB p52 and RelB (Fig. 2A). We also observed changes in the cytoplasmic fraction (Fig. 2B). RANKL increased p100/52 and RelB expression in the cytoplasm after 24 h. In the presence of IL-4, the amount of cytoplasmic p100 and the p100/p52 ratio were altered. In the presence of RANKL, the ratio of p100/p52 was 0.7. However, in the presence of RANKL plus IL-4, the ratio was 1.7. These results suggest that IL-4 inhibits activation of the slower alternative NF-κB pathway by RANKL and alters the RANKL-induced processing of p100 to p52.

To determine the biological impact of the increase in p105/50, nfκb1^−/− BMM, which lack both p105 and p50, were treated with cytokines for 5 d to induce macrophage fusion. Interestingly, the formation of giant cells induced by RANKL (osteoclasts detected by TRAP staining) or by IL-4 or RANKL plus IL-4 (MNG detected by crystal violet staining) was impaired in nfκb1^−/− BMM (Fig. 3A). The average size of each multinucleated cell (nuclei >2) and the total area were significantly lower in the absence of nfκb1 (Fig. 3B). This suggests that p105/50 may be important for macrophage fusion in general. Although the RANKL-induced formation of TRAP⁺ osteoclasts was reduced in the absence of nfκb1, IL-4 was still able to suppress the TRAP⁺ osteoclast formation in these cultures (Supplemental Fig. 2A).

We also examined the alternative pathway in BMM derived from nfκb1^−/− mice. The RANKL-induced increase in p100/52 and RelB protein was still observed in the nfκb1^−/− cells (Supplemental Fig. 2B). Furthermore, RANKL was still able to induce the nuclear translocation of p52, p65, and RelB. IL-4 also suppressed the RANKL-induced nuclear translocation of p52 and RelB. Taken together, these results indicate that modulation of p105/50 by IL-4 was not essential for its inhibitory effect on osteoclast differentiation and TRAP expression. The effect of RANKL and IL-4 on the NF-κB alternative pathway was not affected by the lack of p105/50.

To confirm a role for NF-κB in macrophage fusion, we used several selective inhibitors of NF-κB (Fig. 4, Supplemental Fig. 3). The in-solution IKK2 inhibitor completely abrogated RANKL- and IL-4–induced macrophage fusion (Fig. 4A). Similar results were obtained using pyrrolidine dithiocarbamate (data not shown). NEMO forms an IκB kinase complex with IKK1 and IKK2 and is essential for both IκBα and p105 phosphorylation, degradation, and NF-κB pathway activation (28). A cell-permeable NBD inhibitory peptide blocked association of NEMO with the IKK complex and inhibited cytokine-induced NF-κB activation (29). To further identify the role of the IKK complex in IL-4–induced macrophage fusion, BMM were treated with IL-4 in the presence of NBD inhibitory peptide or control peptide. NEMO inhibitory peptide inhibited IL-4–induced MNG formation, but control peptide did not (Fig. 4B, Supplemental Fig. 3B). These results suggest that a NEMO-dependent pathway is involved in IL-4–induced MNG formation.

To confirm that NF-κB proteins participated in macrophage fusion, cDNA encoding p50, p65, p52, and RelB were transduced by lentiviral vector in nfκb1^−/− or nfκb1^+/+ BMM (Supplemental Fig. 4). Overexpression of p50, p52, p65, or RelB in nfκb1^−/− BMM restored macrophage fusion induced by RANKL or IL-4 (Supplemental Fig. 4A). Overexpression of these proteins also enhanced IL-4–induced fusion in nfκb^+/+ BMM (Supplemental Fig. 4B). Overexpression of these proteins was confirmed by Western blotting analysis (Supplemental Fig. 4C).

To examine the role of the alternative NF-κB pathway in RANKL- and IL-4–induced macrophage fusion, nfκb2 shRNA or scrambled control was transduced by lentivirus to wild-type BMM (Fig. 5, Supplemental Fig. 5). Transduction of BMM with nfκb2 shRNA resulted in substantial knockdown of p100 and p52 (Fig. 5A). Nfκb2 knockdown enhanced RANKL-induced macrophage fusion, which is in accordance with the study that nfκb2-deficient BMM showed enhanced sensitivity to RANKL (30). Surprisingly, nfκb2 knockdown also significantly enhanced the IL-4– and RANKL plus IL-4–induced macrophage fusion (Fig. 5, Supplemental Fig. 5). These results suggest that p100/52 (Nfκb2) acts as an inhibitor of macrophage fusion. Because p52 transduction enhanced macrophage fusion, the inhibitory effect is likely mediated by p100 itself. Indeed, dimers of p100 have been identified as IκBδ (11). Interestingly, nfκb2 shRNA knockdown significantly enhanced the basal expression of CD44, DC-STAMP, and E-cadherin in wild-type BMM (M-CSF treatment, Fig. 5C), whereas the RANKL-induced or IL-4–induced expression was unchanged.

To determine whether BMM lacking both NF-κB1 (p105/50) and NF-κB2 (p100/52) would have a defect in macrophage fusion, nfκb2 shRNA was transduced by lentivirus to nfκb1-deficient BMM (Fig. 6, Supplemental Fig. 5B). The RANKL-induced formation of osteoclasts was completely abrogated by the knockdown of nfκb2 in the nfκb1^−/− background, whereas overexpression of p50 or p65 enhanced osteoclast formation. This result agrees with previous reports showing that NF-κB1/NF-κB2 double-knockout mice have a defect in osteoclastogenesis and confirms the efficiency of nfκb2 knockdown (12). In contrast, knockdown of nfκb2 enhanced the IL-4–induced formation of MNG in the nfκb1^−/− background in the absence and presence of RANKL (Fig. 6A, Supplemental Fig. 5B). However, the MNG observed with nfκb2 knockdown in the nfκb1^−/− background were smaller than MNG formed in nfκb1^+/+ BMM (Supplemental Fig. 5), revealing a contribution for p105/p50 in IL-4–induced fusion in a cell devoid of p100/p52.

The mRNA level of osteoclastogenic and fusion-related genes were examined by real-time PCR (Fig. 6B, 6C). Analysis of mRNA for nfκb2 demonstrated the efficiency of knockdown in these cells. In the absence of nfκb1, RANKL was still able to induce expression of NFATc1, TRAP, and the fusion-related genes, demonstrating redundancy among NF-κB family members for responsiveness to RANKL. As expected, nfκb2 knockdown completely abrogated the RANKL-induced expression of NFATc1, TRAP, and the fusion-related genes DC-STAMP, E-cadherin, and CD44 in nfκb1^−/− BMM. The effect of nfκb2 knockdown on IL-4–induced gene expression was different. DC-STAMP, E-cadherin, and CD44 were significantly upregulated in the presence of IL-4. Nfκb2 knockdown significantly enhanced TRAP expression in IL-4 and RANKL plus IL-4–treated BMM without affecting NFATc1 expression (Fig. 6B). This is consistent with our previous findings that IL-4–induced TRAP expression is mediated by STAT6 and unrelated to NFATc1 (23, 31). These results suggest that p100/p52 inhibits the expression of STAT6-dependent genes induced by IL-4.

NIK activates IKK1, which phosphorylates p100. After phosphorylation, p100 undergoes partial degradation to form p52 (32). NIK^−/− mice demonstrate reduced p100 processing and p52 generation (22). The nfκb2 shRNA knockdown strategy depletes both p100 and p52 expression and therefore cannot distinguish their respective roles in macrophage fusion. To clarify the roles of p100 and p52 in RANKL- and IL-4–induced macrophage fusion, NIK^−/− BMM were prepared. Both RANKL- and IL-4–induced fusion were significantly reduced in NIK^−/− BMM as compared with controls (Fig. 7). Cytoplasmic and nuclear extracts of these cultures were evaluated by Western blotting. In wild-type BMM, RANKL induced the expression of p100 and enhanced the formation of p52 in the cytoplasm. The RANKL-stimulated increase in p52 in the cytoplasm was greatly reduced in the absence of NIK, as expected. The ratio of p100/p52 was significantly increased from 0.7 in wild-type cells to 5.2 in NIK^−/− BMM. In contrast, treatment of wild-type BMM with IL-4 induced expression of p105 and p50 in the cytoplasm. This IL-4–induced response was intact in NIK^−/− BMM. As shown in Fig. 2, the combination of IL-4 with RANKL resulted in an increase in the p100/p52 ratio in wild-type cells. However, in the absence of NIK, IL-4 plus RANKL still induced the production of p52; in this case, the amount of p100 was in excess of p52 (p100/p52 ratio 1.9) (Fig. 8A).

In wild-type BMM, RANKL stimulated the nuclear translocation of both p52 and p50 (Fig. 8B). The levels of p65 in the nucleus were almost negligible. IL-4 alone induced the accumulation of p50 in the nucleus. Addition of IL-4 inhibited the RANKL-induced translocation of p52. In NIK^−/− BMM, the pattern of nuclear translocation was quite different. In the absence of NIK, RANKL did not induce translocation of p52, and the amount of p50 in the nucleus was reduced. Furthermore, the IL-4–induced accumulation of p50 in the nucleus was greatly reduced in the absence of NIK even though IL-4 induced expression of p105 and p50 in the cytoplasmic fraction. The IL-4–induced expression of the fusion-related genes DC-STAMP, CD44, and E-cadherin were reduced in NIK^−/− BMM as compared with wild-type (Fig. 8C). Taken together, these results demonstrate the important role for NIK in both RANKL- and IL-4–induced macrophage fusion. They also support the model that excess p100 in the cytoplasm acts as an inhibitor of NF-κB dimers and suppresses macrophage fusion (Fig. 9). In support of this model, we found that overexpression of p100 in BMM suppressed RANKL- and IL-4–induced macrophage fusion (Supplemental Fig. 5A).

IL-4 is known to induce macrophage fusion to form MNG by a STAT6-dependent process requiring protein–protein interaction (33). Although directly driving MNG formation, IL-4 suppressed the RANKL-induced formation of osteoclasts, cells that also form by macrophage fusion. Several studies showed that inhibition of IκBα degradation and prevention of p65 translocation were responsible for the IL-4 inhibitory effect on osteoclast formation (18–20, 34). However, the regulation of NF-κB signaling by IL-4 and the potential consequences of this regulation are highly controversial. Some studies showed that IL-4 suppressed NF-κB activity and inhibited proinflammatory cytokine production (35, 36). The inhibition of IκBα degradation and suppression of p65 translocation were proposed as the inhibitory mechanism by which IL-4 modulates cellular responses to stress (37). Other mechanisms of regulation were also proposed. For example, IL-4 inhibited TNF-α–stimulated E-selectin expression through STAT6 competition for an NF-κB DNA binding site in the promoter (38). IL-4/STAT6 also can suppress NF-κB activity through competing for coactivators such as CREB-binding protein (39). However, there are also reports showing that activation of the H chain germline ε promoter by IL-4 required both STAT6 and NF-κB binding (40, 41). STAT6 and NF-κB were found to directly associate and mediate the synergistic activation of IL-4–induced transcription (25). In B cells, the ability of IL-4–stimulated STAT6 to bind to DNA required NF-κB activity (42). Our previous study showed that NF-κB activation plays an important role in the IL-4–induced protection from apoptosis (43). In addition, another report showed that NF-κB activation is required for CD44-mediated macrophage fusion (21). Taken together, these studies argue that IL-4 activates signaling pathways, including STAT6, that can cross-talk with constitutive or inducible NF-κB pathways. However, the characterization of this cross-talk has remained elusive.

Using a variety of pharmacologic, knockout, knockdown, and overexpression approaches, we investigated the contribution of the NF-κB pathways to IL-4– and RANKL-induced macrophage fusion. Although focused on macrophage fusion and cross-talk with RANKL, these studies have implications for the mechanisms by which IL-4 cooperates with other NF-κB–activating receptors including TLR4, B cell Ag receptor, TCR, and CD40. We found that IL-4 treatment and pretreatment did not inhibit the rapid RANKL-induced activation of the NF-κB canonical pathway. However, we found that IL-4 inhibited the RANKL-induced p52 and RelB translocation, which suggests that IL-4 inhibits the RANKL-activated alternative pathway (Fig. 9). We do not know the mechanism by which IL-4 suppresses p52/RelB translocation, but it may be due, in part, to an alteration in processing of p100. We found a moderate increase in the p100/p52 ratio in cells cultured in RANKL plus IL-4 as compared with cells cultured in RANKL alone. Because p100 can act as an IκB (IκBδ) and retain p52/RelB dimers in the cytoplasm (11), an increase in the relative amount of p100 may prevent NF-κB dimer translocation.

IL-4 did not induce rapid NF-κB translocation by itself. IL-4 enhanced p105/50 expression and p50/50 homodimer formation after 48 h treatment. The nfκb1 gene, which encodes p105 and p50, was required for IL-4–induced macrophage fusion, whereas ectopic overexpression of p50 in nfκb1^−/− BMM significantly enhanced IL-4–induced MNG formation. Interestingly, NIK, which regulates the inducible proteolysis of p100, was required for IL-4–induced MNG formation, accumulation of p50 in the nucleus, and induction of fusion-related genes in response to IL-4. However, NIK was not required for the induction of cytoplasmic p105/p50 by IL-4. These results suggest that IL-4–induced upregulation of p105 and accumulation of p50 in the nucleus play a critical role in MNG formation and that NIK participates in the IL-4–driven accumulation of nuclear p50 (Fig. 9). The mechanism by which NIK regulates this process is unknown. It is possible the inability to process p100 leads to excess cytoplasmic p100, which could act to sequester p50/p50 homodimers (Fig. 9). The importance of p105/p50 in these IL-4–driven responses are consistent with several other studies showing a crucial role for p50 in the orchestration of alternative activated macrophage-driven inflammatory reactions and inhibition of the classical activation (44, 45).

The NF-κB inhibitor IKK2 inhibitor suppressed both osteoclast and MNG formation induced by RANKL or IL-4 treatment, respectively. The NEMO inhibitory peptide also inhibited IL-4–induced MNG formation. The classical IKK complex, composed of IKK1, IKK2, and the regulatory subunit NEMO, triggers IκBα degradation and also regulates p105 proteolysis to form p50 (46, 47). However, only IKK1, but not IKK2, triggers p100 phosphorylation and proteolysis to form p52 (48, 49). The finding that the IKK2 inhibitor and NEMO peptide blocked IL-4–induced fusion further suggests that the classic IKK complex plays an important role in RANKL- and IL-4–induced macrophage fusion. Because IL-4 did not induce translocation of p65 at any time points studied, the mechanism by which this complex controls IL-4–induced events is unclear. It is possible that constitutive activity of this complex is required to generate the p50 needed for IL-4–induced fusion.

Overexpression of p52 or RelB and nfκb2 knockdown enhanced RANKL- and IL-4–induced macrophage fusion, whereas overexpression of p100 reduced IL-4–induced macrophage fusion. Because p52 was shown to enhance macrophage fusion, these results suggest that p100 plays a negative role in MNG formation. NIK-deficient BMM have an increased p100/p52 ratio and reduced fusion, suggesting that NIK acts to generate p52 in an inducible fashion, as observed during RANKL signaling, and to decrease the amount of p100. Several studies have also shown that p100 is a negative regulator of RANKL-induced osteoclast formation (31, 50, 51). These results indicate that the alternative pathway also participates in RANKL- and IL-4–induced macrophage fusion. However, IL-4 alone did not induce p52 and RelB translocation and inhibited the RANKL-induced nuclear translocation of p52 and RelB. These results suggest that p52/RelB are not necessary for IL-4–induced MNG formation.

Knockdown of nfκb2 in nfκb1-deficient BMM completely blocked osteoclast differentiation. In contrast, nfκb2 knockdown in nfκb1-deficient BMM enhanced IL-4–induced MNG formation. This suggests that IL-4–induced MNG formation may not require both NF-κB1 and NF-κB2. However, IKK2 and NEMO inhibitory peptide completely blocked MNG formation. There are two possible explanations for these somewhat paradoxical findings. One is that NF-κB transcription factor dimers are not absolutely required for IL-4– induced macrophage fusion and act only to modulate the activity of STAT6. Another possibility is that other NF-κB transcription factor dimers may participate in this process in the absence of NF-κB1 and NF-κB2. Overexpression of either p65 or RelB enhanced IL-4–induced MNG formation. In addition, Rel proteins can form homodimeric and heterodimeric complexes in the absence of NF-κB1 and NF-κB2. For example, p65 can form a relatively stable homodimer complex and is rapidly translocated to the nucleus after stimulation (52). p65 and c-Rel also can form heterodimer complexes and induce the expression of NF-κB–responsive genes (53). Further investigation will be required to determine the particular NF-κB pathways and which transcription factor dimers are involved in the IL-4–induced fusion process.

Knockdown of nfκb2 in wild-type BMM significantly enhanced the basal expression of CD44, DC-STAMP, and E-cadherin. This may partially explain why nfκb2 knockdown enhanced both RANKL-induced osteoclast and IL-4–induced MNG formation. The fusion-associated genes would already be upregulated, and the fusion machinery would be ready for rapid response to fusion stimuli. Compared with responses observed in nfκb2 knockdown in nfκb1-deficient BMM, the basal levels of CD44, DC-STAMP, and E-cadherin were significantly higher in nfκb2 knockdown of wild-type BMM. This further suggests that p105/50 plays a critical role in the expression of these fusion-related genes. Expression of these downstream genes induced by IL-4 does not require RANKL-activated NF-κB translocation. It is still not clear if these fusion-related genes are regulated directly by NF-κB binding to their promoters.

Our previous study indicated that IL-4 modestly enhanced TRAP expression in BMM through STAT6 binding to the TRAP promoter (23). Interestingly, we found that nfκb2 knockdown significantly enhanced IL-4–induced TRAP expression. One possibility for this result is that NF-κB molecules coordinate and enhance STAT6 function in the absence of p100. Another possibility is that p100 has a direct inhibitory effect on STAT6 function. The possible regulation of STAT6 by p100 will require further investigation. The physiological function of TRAP expression in the IL-4–induced MNG is still unknown.

The mechanism of macrophage fusion shares several characteristics with the process of phagocytosis (54, 55). Raveh et al. (56) showed that IL-4 and IFN-γ alone or together enhanced mannose receptor-mediated phagocytosis. However, fusion and phagocytosis can be uncoupled based on their differential susceptibility to Rac1 inhibition. In this model, inhibition of fusion did not inhibit phagocytosis (57). In another study, increasing the stimulus of phagocytosis did not inhibit cell fusion (58). Varin et al. (59) indicated that nonopsonic phagocytic uptake of Neisseria meningitidis and several other microbial particles were markedly reduced in IL-4–treated peritoneal macrophages, although IL-4 induced macrophage fusion. Papadimitriou et al. (60) showed that plastic film-induced MNG (more than seven nuclei) lost phagocytic ability. These studies show that fusion and phagocytosis are different processes. Phagocytosis may not be enhanced through cell fusion. It will be interesting to further investigate whether macrophage phagocytosis will be affected by manipulating NF-κB signaling.

Implanted biomaterials, some infected parasites, or products of infected parasites are subject to a foreign body reaction (61, 62). During that process, implants or parasites are surrounded by macrophages that fuse to form MNG, as well as a collagenous and avascular fibrotic capsule, and finally form granuloma (63, 64). This damaging effect can lead to reduced functionality, failure of implants, or organ fibrosis (65–67). This foreign body reaction is highly dependent on IL-4 and IL-13 (63, 68). Our new findings indicate that NF-κB signaling plays an essential role in both RANKL- and IL-4–induced macrophage fusion. Because NF-κB inhibitors can block both osteoclast and MNG formation, it may be possible to develop NF-κB inhibitors as therapeutic agents to prevent Th2-type giant cell and granuloma formation. This concept provides a new explanation for the finding that glucocorticoids, potent NF-κB inhibitors (69), effectively reduced both Schistosoma mansoni egg-induced granuloma formation and maturation (70) and strongly inhibited granuloma formation induced by implanted biomaterial (61).

We thank Dr. Ranjan Sen for providing nfκb1^−/− mice, Dr. Robert Schreiber and Dr. Barry Sleckman for providing NIK^−/− mice, Dr. Janet Stavnezer, Dr. Ricardo Feldman, and Dr. Didier Trono for generously supplying plasmid cDNA for these studies, and Mary Kaileh and Ryan Irwin for excellent technical assistance.

The authors have no financial conflicts of interest.