2.1. The Functions of NF-κB Signaling in Physiological Osteoclastic Bone Resorption
Osteoclasts differentiate from hematopoietic stem cells into osteoclasts via macrophage and monocyte pathways [
3,
4,
5,
17]. During differentiation, osteoclast progenitor cells proliferate and differentiate into mono- and binucleated osteoclasts that fuse to become multinucleated osteoclasts. Multinucleated osteoclasts recognize the bone matrix, form a sealing zone to separate the resorption surface from the outside, form a ruffled border, and secrete acid and proteolytic enzymes into the resorption lacunae [
3,
4,
5,
17].
Osteoclast differentiation is controlled by two cytokines: macrophage colony-stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL). M-CSF is essential for differentiation into osteoclast progenitors, and it induces the expression of the RANKL receptor RANK. Transcription factors PU.1 and MITF induce the expression of M-CSF receptor (
c-fms), and individuals lacking these transcription factors have impaired osteoclast differentiation and present with marble bone disease. In addition,
op/
op mice and
c-fms-deficient mice that cannot produce functional M-CSF exhibit marble bone disease and lack osteoclasts [
3,
4,
5,
17].
RANKL is produced by various cells, such as osteoblasts, osteocytes, T cells and B cells [
3,
4,
5,
17]. Mice that lack RANKL and its receptor RANK have severe osteopetrosis caused by a total lack of osteoclasts [
18,
19]. On the other hand, the number of osteoclasts increases in mice that lack the RANKL decoy receptor, osteoprotegerin (OPG), resulting in osteoporosis [
20,
21]. In human hereditary bone disease, mutations in RANKL, RANK, and OPG have been found, and these three molecules have been shown to be important for osteoclast formation, which maintains bone mass [
22].
RANK belongs to the TNF receptor family, and various adapter molecules can interact with the intracellular domain of RANK [
23]. Among TNF receptor-activating factor (TRAF) members, TRAF6-deficient mice exhibit an osteopetrosis phenotype that is similar to RANKL- or RANK-deficient mice [
24]. Of the downstream molecules of TRAF6, c-Fos and c-Jun regulate the transcription factor AP-1. c-Fos-deficient mice also exhibited osteopetrosis [
25]. Another downstream molecule, the transcription factor NF-κB, is composed of five family members. In mice with both NF-κB1 and NF-κB2 knocked out, there is also osteopetrosis due to the total lack of osteoclasts, but deletion of either NF-κB1 or NF-κB2 alone causes no detectable bone phenotype [
13,
14]. The molecular mechanism by which osteoclasts cannot form in NF-κB1 and NF-κB2 dKO mice is still unknown, but it is certain that NF-κB signaling is important for osteoclast formation.
Among the molecules involved in the signal transduction of NF-κB, p65 (RelA), IKKβ and NEMO could not be analyzed regarding a bone phenotype because these molecules are embryonic lethal [
26,
27,
28,
29,
30,
31]. Thus, to make IKKβ specifically deficient (IKKβcKO) in myeloid cells, IKKβ
flox/flox mice were crossed with Mx1 or CD11b-Cre transgenic mice to generate conditional knockout mice in which IKKβ is specifically deficient (IKKβcKO) in myeloid cells [
32,
33]. IKKβcKO mice showed an increase in the trabecular bone volume due to a decrease in the number of osteoclasts. Furthermore, the number of osteoclast precursor cells (F4/80 positive cells) was also significantly reduced. When IKKβcKO mice were crossed with tumor necrosis factor receptor 1(TNFR1)
–/– mice to generate IKKβcKO/TNFR1KO dKO mice, osteoclast precursor cells were resistant to apoptosis; further, IκBα was not degraded by RANKL stimulation, and osteoclast differentiation was still suppressed. By contrast, in IKKα knock-in (IKKα
A/A) mice in which the serine residue necessary for IKKα kinase activity was substituted with alanine, osteoclast formation by RANKL stimulation was suppressed in vitro but not in vivo. The trabecular bone volume in IKKα
A/A mice was comparable to that of wild-type (WT) mice [
30]. Furthermore, IKKβ-deficient osteoclasts resulted in RANKL-induced apoptosis by the activation of c-Jun N-terminal kinase (JNK), and the addition of JNK inhibitor restored RANKL-induced apoptosis derived from IKKβcKO mice in vitro [
33]. Thus, IKKβ, but not IKKα, is important as a RANK downstream signal in osteoclast differentiation. Consistent with these results, treatment with specific inhibitors of IKKβ activity suppressed RANKL-induced osteoclastogenesis in vitro and in vivo [
34,
35,
36,
37].
Since p65-deficient (p65
–/–) mice are also embryonic lethal, p65
–/– fetal liver cells were studied; the cells were transplanted into irradiated mice to reconstitute bone marrow cells. Fewer osteoclasts were observed in p65
–/– chimera mice. When p65
–/– chimera mice were crossed with TNFR1
–/– mice, p65
–/– precursors were found to be sensitive to RANKL-induced apoptosis even on the TNFR1
–/– background. ZVAD, a caspase inhibitor, restored RANKL-induced osteoclastogenesis in p65
–/– precursors in vitro, suggesting that p65 induces proapoptotic gene expression in osteoclastogenesis [
38].
Several lines of evidence have shown that the alternative NF-κB pathway also involves RANKL-induced osteoclastogenesis. When RANKL is administered to NIK-deficient (NIK
–/–) mice, osteoclast formation is more inhibited than it is when RANKL is administered to wild-type mice. Osteoclast progenitor cells derived from NIK
–/– mice did not induce processing of p100 to p52 by RANKL stimulation due to IκB-like function of the C-terminus of p100 [
39,
40]. In addition, mice lacking IKKα, which is a molecule that is downstream of NIK, contain osteoclasts but are small in size and have reduced bone resorbing activity. As with NIK
–/– mice, processing of p100 to p52 by RANKL stimulation does not occur in IKKα-deficient mice [
32,
41]. The role of RelB in osteoclastic bone resorption is still unclear. Although the number of osteoclasts was normal, the bone mass was slightly increased. However, overexpression of RelB restored RANKL-induced osteoclastogenesis in NIK
–/– mice [
42]. Recently, NIK-deficient and RelB-deficient female mice, but not male mice, revealed a 2-fold increase in trabecular bone mass, suggesting that the alternative NF-κB pathway involves gender difference in bone metabolism [
43]. Alymphoplasia (
aly/aly) mice do not undergo p100 to p52 processing because NIK is inactive.
Aly/aly mice showed mild osteopetrosis and had a greatly reduced osteoclast count [
44,
45]. RANKL-induced osteoclast formation from the bone marrow cells of
aly/aly mice was also suppressed. RANKL still induced IκBα degradation and activated classical NF-κB, but p100 to p52 processing was abolished by the
aly/aly mutations. Overexpression of NFATc1 and constitutive activation of IKKα or p52 restored RANKL-induced osteoclastogenesis in
aly/aly cells. The overexpression of RelB in
aly/aly cells restored RANKL-induced osteoclastogenesis by inducing cancer Osaka thyroid (Cot) expression, which induces the processing of p52 from p100 in place of NIK [
46]. Taken together, the balance between p52 and p100 determines RANKL-induced osteoclastogenesis.
2.3. The Activation of NF-κB Suppresses Bone Formation
Bone is composed of hydroxyapatite crystals and various extracellular matrix proteins, including type I collagen, osteocalcin, osteopontin, bone sialoprotein, and proteoglycan. Most of these bone matrix proteins are secreted and deposited by mature osteoblasts that are aligned on the bone surface. The formation of hydroxyapatite crystals in osteoid is also regulated by osteoblasts. The expression of numerous bone-related extracellular matrix proteins and the activity of alkaline phosphatase (ALP) are key features of osteoblasts [
1,
2].
Osteoblasts differentiate from mesenchymal stem cells, and their differentiation stage is cooperatively and dynamically controlled by specific signal transduction pathways, directly or indirectly. Osteoblasts differentiate from mesenchymal stem cells through various intracellular signaling mechanisms by various cytokines and hormones, such as bone morphogenetic proteins (BMPs), transforming growth factor (TGF)-β, Wnt, hedgehog, fibroblast growth factor, and estrogen. This intracellular signal transduction is activated by phosphorylation, ubiquitination, protein–protein interactions and structural changes following the binding of ligands to receptors. Since mice with either Runx2 or Osterix transcription factors knocked out exhibited impaired bone formation, these two transcription factors have been reported to be important for osteoblast differentiation [
1,
2,
80].
It is known that bone formation is suppressed in an inflammatory state, and, in particular, TNF-α is known to suppress osteoblast differentiation in various culture systems [
76,
81,
82,
83,
84]. TNF-α activates various signals in the cell, but a specific IKK inhibitor, BAY11-7082, restores the suppression of osteoblast differentiation induced by TNF-α [
85]. Recently, it has been reported that the inhibition of NF-κB by the dominant negative form of IKKβ enhances bone formation [
13]. The administration of an inhibitor of IKK, S1627, promoted bone formation in ovariectomized (OVX) mice [
14]. Mice expressing the dominant negative form of IKKβ in mature osteoblasts showed increased bone mass, bone mineral density, and osteoblast activity without exhibiting any changes to osteoclast activity. Furthermore, expressing the dominant negative form of IKKβ maintained the bone mass of OVX mice by increasing the expression of Fos-related antigen-1 (Fra1), which is an essential transcription factor involved in bone matrix formation [
13]. There are also reports supporting these findings that show that estrogen receptors inhibit the activation of the classical NF-κB pathway by interacting with NF-κB [
86]. As another possibility, TNF-α, IL-1β, IL-6, and IL-17 produced by T cells and other cells during osteoporosis have been reported to activate the classical NF-κB pathway [
87].
Bone morphogenetic proteins (BMPs) belong to the TGF-β superfamily and were originally identified by their ability to induce ectopic bone formation when implanted into muscle tissue [
88,
89]. Since BMP signaling and the classical NF-κB pathway have opposing biological activities, crosstalk between the two is possible. A cell-permeable inhibitor of the classical NF-κB pathway restored the inhibitory effects of TNF-α on BMP2-induced Runx2 expression and osteoblast differentiation [
90]. Zinc inhibits the classical NF-κB pathway by TNF-α and promotes BMP2-induced osteoblast differentiation [
91]. Pyrrolidine dithiocarbamate (PDTC), which inhibits the classical NF-κB pathway, partially blocked the TNF-α-induced suppression of osteoblast differentiation. These results indicate that inhibition of the classical NF-κB pathway by BMP [
92] reverses osteoblast differentiation in a mechanism dependent on TNF-α. Thus, the implantation of collagen sponges containing BAY11-7082, a selective inhibitor of the classical NF-κB pathway, with BMP2 under the fascia resulted in the formation of larger amounts of ectopic bone than what was seen following treatment with only BMP2 [
93]. These results suggest that selective inhibitors of the classical NF-κB pathway have the effect of promoting bone formation by BMP. However, the side effects of its administration must be considered, since inhibition of the classical NF-κB pathway activity might induce cell death [
26,
27,
28,
29,
30,
31]. Therefore, to enhance the effect of BMP without impairing the function of the classical NF-κB pathway, the suppression mechanism of the BMP/Smad signal of the classical NF-κB pathway was examined. There are various stages of BMP/Smad signaling, but the classical NF-κB pathway does not affect the phosphorylation of Smad1/5 or the formation of the Smad1–Smad4 complex; however, the classical NF-κB pathway does interfere with the DNA binding complex. Furthermore, we found that the p65 subunit of the classical NF-κB pathway associates with Smad4 but not Smad1 [
64]. Therefore, when the association sites of p65 and Smad4 were examined, the transactivation domain 2 (TA2) of p65 and the mad homology (MH) 1 region of Smad1 were directly associated. We further narrowed the association site to the amino acid level and found that the 16 amino acid sites on the N-terminal side of the TA domain of p65 were critical for binding to Smad4; we named the site the Smad binding domain (SBD) [
94]. We synthesized the SBD peptide to compete with the interaction of p65 with Smad4. The SBD peptide promoted ALP activity and calcification induced by BMP2 in vitro. Furthermore, administration of SBD peptide together with BMP2 induced ectopic thick cortical bone formation in vivo. The SBD peptide did not affect the activation of the classical NF-κB pathway by TNF-α stimulation [
94]. Based on these results, it is possible that peptides targeting the association site of NF-κB, p65 and Smad4 may be useful for promoting bone formation by BMP with few side effects (
Figure 3).
Recently, a heterozygous de novo missense mutation (c.1534_1535delinsAG, p.Asp512Ser) in exon 11 of RELA encoding Rela/p65 was found in a neonate who had died suddenly and unexpectedly with high bone mass (HBM) that was judged radiographically and by skeletal histopathology [
95]. Numerous morphologically normal osteoclasts in the neonate were observed in bone histology, suggesting that the missense change was associated with neonatal osteosclerosis from increased osteoblastic bone formation in utero rather than failed osteoclastic bone resorption. Moreover, LPS stimulation failed to activate the classical NF-κB pathway in fibroblasts derived from the neonate. This is the first report that demonstrates the importance of the Rela/p65 subunit within the classical NF-κB pathway for human skeletal homeostasis and represents a new genetic cause of HBM [
95].
Bone histomorphometric data from
aly/aly mice show an increase in trabecular bone volume caused by both the suppression of bone resorption and increased bone formation, suggesting that the alternative NF-κB pathway also regulates osteoblastic bone formation [
96]. ALP activity and the expression of osteoblastic markers (including osteocalcin, Id1, Osterix, and Runx2) induced by either β-glycerophosphate and ascorbic acid or BMPs were increased in primary osteoblasts (POB) derived from
aly/aly mice compared with WT mice. The ectopic bone formation in vivo induced by BMP2 was enhanced in
aly/aly mice compared with WT mice, due to enhancement of BMP2 signaling [
96]. Thus, the alternative NF-κB pathway via the processing of p52 from p100 negatively regulates osteoblastic differentiation and bone formation by modifying BMP activity. Mice that have RelB, a main subunit of the alternative NF-κB pathway, knocked out develop age-related increased trabecular bone mass associated with increased bone formation [
97]. RelB
–/– bone marrow stromal cells enhanced osteoblastic differentiation by increasing Runx2 expression. RelB directly bound to the Runx2 promoter to inhibit its activation. Moreover, RelB
–/– bone-derived mesenchymal progenitor cells (MPCs) formed bone more rapidly than WT cells after they were injected into a murine bone defect model [
97].
Notch is a family of evolutionarily conserved receptors that regulate cell fate, and its signaling plays various important roles in bone metabolism [
98]. Notch signaling and the alternative NF-κB pathway were identified as signaling pathways responsible for the inhibitory effects of TNF-α on osteoblastic differentiation. This was done by RNA sequencing and pathway analysis of mesenchymal stem cells using WT and TNF-α transgenic (Tg) mice, a model of RA [
99]. Notch inhibitors restored bone loss and osteoblast inhibition in TNF-α Tg mice. The transplantation of fibroblasts from TNF-α Tg mice treated with Notch inhibitors formed more new bone in recipient mice with bone defects. The activation of the alternative NF-κB pathway in a murine pluripotent stem cell line induced RPBjκ and HES1 in a Notch intracellular domain dependent manner (NICD-dependent). TNF-α enhanced the binding of p52/RelB heterodimer to NICD, which induced binding at the RBPjκ site within the Hes1 promoter. Elevated levels of HES1, p52, and RelB were observed in mesenchymal stem cells from RA patients [
99]. These results indicate that the inhibition of the alternative NF-κB pathway could reduce age-related bone loss and enhance bone repair as well as inflammation-mediated bone loss.