Negative regulation of osteoclastogenesis and bone resorption by cytokines and transcriptional repressors

Bone remodeling in physiological and pathological conditions represents a balance between bone resorption mediated by osteoclasts and bone formation by osteoblasts. Bone resorption is tightly and dynamically regulated by multiple mediators, including cytokines that act directly on osteoclasts and their precursors, or indirectly by modulating osteoblast lineage cells that in turn regulate osteoclast differentiation. The critical role of cytokines in inducing and promoting osteoclast differentiation, function and survival is covered by the accompanying review by Zwerina and colleagues. Recently, it has become clear that negative regulation of osteoclastogenesis and bone resorption by inflammatory factors and cytokines, downstream signaling pathways, and a newly described network of transcriptional repressors plays a key role in bone homeostasis by fine tuning bone remodeling and restraining excessive bone resorption in inflammatory settings. In this review we discuss negative regulators of osteoclastogenesis and mechanisms by which these factors suppress bone resorption.

and c-fos, to drive osteoclastogenesis [2] (Figure 2). More recently, transcriptional repressors that suppress RANKLinduced gene expression and diff erentiation have been described ( Figure 2). Th ese repressors can work as homeostatic factors in regulating osteoclasto genesis in physiological bone development and re modeling, and also as feedback inhibitors that limit bone resorption asso ciated with infl ammation. Th e extent of bone destruc tion in infl ammatory diseases is determined by the balance between osteoclastogenic and anti-osteoclastogenic factors.

IL-4/IL-13 and granulocyte-macrophage colonystimulating factor
IL-4 and IL-13 have pleiotropic immune functions and are produced by Th 2 lymphocytes, although IL-13 can also be produced by stromal cells. Since IL-4 and IL-13 utilize closely related receptor complexes, they have many overlapping features, including downstream signaling and some biological functions. IL-4, more eff ectively than IL-13, directly prevents osteoclast precursors from diff erentiating into osteoclasts in a signal transducer and activator of transcription (STAT)6-dependent manner [3,4]. IL-4 suppresses RANK expression, NF-κB, MAPK and calcium signaling, and expression of NFATc1 and c-Fos during osteoclasto genesis [3][4][5]. In addition, IL-4 inhibits bone resorption and actin ring formation in human mature osteoclasts by suppressing NF-κB and calcium signaling. On the other hand, IL-4 and IL-13 indirectly suppress osteoclasto genesis by inhibiting RANKL but enhancing OPG expres sion in osteoblastic cells [3,4]. Although IL-4 suppresses spontaneous or para thyroid hormone-related protein (1-34)-stimulated osteo clast formation in mice, IL-4 transgenic mice exhibit an osteoporotic phenotype that is attributed to a more dominant suppressive eff ect of IL-4 on osteoblast formation in vivo relative to its role in suppressing osteo clastogenesis. Th us, it is important to note that the net eff ect of IL-4 on bone turnover in vivo represents an integrated outcome of its infl uence on various cell populations.
Granulocyte-macrophage colony-stimulating factor (GM-CSF) inhibits osteoclastogenesis by diverting osteoclast precursors to a macrophage lineage [6]. Th e osteoclast suppressive mechanism was recently suggested to involve proteolytic cleavage of cell surface M-CSF receptor after treatment with GM-CSF and IL-4 [7]. Th e combination of GM-CSF and IL-4 enhances expression and activity of TACE (TNF-α converting enzyme)/ ADAM17 (a disintegrin and metalloproteinase 17) in human monocytes. Th is results in cleavage of cell surface M-CSF receptor, leading to disruption of M-CSF signaling and thereby suppressing osteoclastogenesis and divert ing the cells toward the dendritic cell lineage [7]. Osteoclasts are derived from myeloid precursors. Macrophage colony-stimulating factor (M-CSF) induces myeloid precursors to diff erentiate to osteoclast precursors that express RANK (Receptor activator of NF-κB) and TREM2 (Triggering receptor expressed by myeloid cells-2) receptors. Upon RANK ligand (RANKL) stimulation and ITAM (Immunoreceptor tyrosine-based activation motif ) activation, osteoclast precursors undergo further diff erentiation to mononuclear osteoclasts with NFATc1 (Nuclear factor of activated T cells, cytoplasmic 1) induction and express osteoclast-related genes such as those encoding TNF-receptor associated protein (TRAP), cathepsin K (CtsK) and αvβ3. Mononuclear osteoclasts then fuse to multinuclear osteoclasts and function as polarized bone resorbing cells. This process of osteoclast diff erentiation is regulated by various transcription factors and exogenous factors at diff erent stages. Infl ammatory factors that promote osteoclastogenesis are shown in red. Inhibitors of osteoclastogenesis are shown in blue. Calc, calcitonin; Calc R, calcitonin receptor; CSF-1R, colony stimulating factor 1 receptor; DC-STAMP, dendritic cell-specifi c transmembrane protein; ECM, extracellular matrix; GM-CSF, granulocytemacrophage colony-stimulating factor; M-CSF, macrophage colony-stimulating factor; MITF, microphthalmia-associated transcription factor; OPG, osteoprotegerin; TLR, Toll-like receptor.

IL-10
IL-10, produced by T and B lymphocytes and myeloid lineage cells, is predominantly an immunosuppressive and anti-infl ammatory cytokine that is best known as a potent deactivator of dendritic cells and macrophages. It plays a critical role in limiting tissue injury during infections and in preventing autoimmunity by limiting the duration and intensity of immune and infl ammatory reactions. A large body of work has established an important role for IL-10 in suppressing osteoclastogenesis in vitro and in vivo [8][9][10][11][12]. For example, IL-10 is expressed in periodontitis, and IL-10 polymorphisms have been linked to periodontitis in multiple studies. In periodontitis, IL-10 is a key negative regulator of bone resorption [8,9]. IL-10 directly inhibits osteoclast precursors by suppres sing RANKL-induced NFATc1, c-Fos and c-Jun expression [10,11]. Inhibition of RANKL expression and an increase in OPG expression due to IL-10 were found in dental follicle cells that support osteoclastogenesis, suggesting that IL-10 can also indirectly inhibit osteoclastogenesis via modulation of RANKL and OPG expression. A key biological activity of IL-10 is to attenuate infl ammation by suppressing TNF-α and IL-1 production and by antagonizing TNF-α and IL-1 function; thereby, IL-10 may suppress TNF-α-and IL-1-stimulated bone resorp tion. Recently, our lab, using human osteoclast pre cursors, showed that IL-10 inhibits calcium signaling by suppressing transcription of TREM-2, an important co-stimulatory receptor for osteo clastogenesis. Downregu lation of TREM-2 (Triggering receptor expressed by myeloid cells-2) expression leads to diminished calcium/calmodulin-dependent protein kinase (CaMK)-MEK-ERK activation induced by RANKL [12].

IL-27
IL-27 is produced by antigen-presenting cells and belongs to the IL-12 family of cytokines. IL-27 has pleiotropic immune functions with either activating or suppressive roles in various infectious and infl ammatory models. Th e IL-27 receptor is an IL-27Ra (WSX-1)/gp130 heterodimer. IL-27 mildly suppresses osteoclast diff erentiation in murine systems, potentially due to the low levels of WSX-1 expression on murine osteoclast precursors, limit ing the response of these cells to IL-27 [13][14][15]. Aggravated arthritic bone erosions and enhanced osteoclastogenesis were observed in Escherichia coli cell wall lysate-induced arthritis models in WSX-1 knockout mice compared to wild-type mice [14]. It should, however, be noted that the enhanced infl ammation and excessive Th 17 cells in WSX-1 knockout arthritis models could also explain the increase in osteoclastogenesis [14]. On the other hand, our lab and other groups [13,14] reported that IL-27 potently inhibits RANKL-induced human osteoclastogenesis and osteoclastic resorptive activity in vitro by downregulation of RANK and TREM-2 expression, inhibition of RANKL-activated ERK, p38 and NF-κB signaling, and by suppression of AP-1 (c-Fos and c-Jun) and NFATc1 expression in human osteoclast precursors. IL-27-induced STAT1 activation also partially contributes to its inhibitory function [14]. While expression of IL-27 is observed in human rheumatoid arthritis, synovial fl uid macrophages harvested from active rheumatoid arthritis patients are refractory to IL-27 [13]. Th is suggests that IL-27 has the capacity to protect bone tissue from resorption, but this homeostatic role of IL-27 might be compromised in an active infl ammatory microenvironment, such as occurs in RA.

Interferons
IFN-γ, the sole type II IFN, is a product of innate immune cells and Th 1 cells. In bone marrow-derived macrophage culture systems, IFN-γ strongly inhibits osteoclastogenesis [16] by sup pres sing RANK signaling via rapid TNF receptor-asso ciated factor (TRAF)6 degradation in murine osteoclast precursors [16]. IFN-γ also inhibits human osteoclasto genesis, but TRAF6 expression is not together with calcium signaling drives expression of NFATc1 (Nuclear factor of activated T cells, cytoplasmic 1) and its targets, resulting in osteoclastogenesis. This process also requires releasing the 'brakes' on NFATc1 expression and osteoclastogenesis that are imposed by transcriptional repressors, including inhibitors of diff erentiation/ DNA binding (Ids), MafB (v-maf musculoaponeurotic fi brosarcoma oncogene family protein B), interferon regulatory factor (IRF)-8 and B cell lymphoma 6 (Bcl6). There is crosstalk between the activating and suppressive pathways, as Blimp1 (B lymphocyte-induced maturation protein-1) that is induced by NFATc1 suppresses expression of MafB, IRF-8 and Bcl6. ITAM, immunoreceptor tyrosine-based activation motif; MAPK, mitogen-activated protein kinase. signifi cantly aff ected [17], suggesting that IFN-γ acts through distinct mecha nisms in humans versus mice. Our lab recently found that IFN-γ, alone or in synergy with TLR stimulation, suppresses expression of the M-CSF receptor c-Fms, c-Fms's target RANK, and costimulatory receptor TREM2 in human osteoclast precursors [17]. In both collagen-induced arthritis and lipopolysaccharide-induced infl am ma tory bone resorption mouse models, loss of IFN-γ receptor leads to enhanced osteoclast formation and bone destruction [16,18]. IFN-γ also inhibits osteoclast formation to prevent tumor-associated bone loss [19]. Th ese data support an inhibitory role of IFN-γ in osteo clastogenesis in vivo. However, administration of re combi nant IFN-γ to rodents or osteopetrotic patients stimulates osteoclast formation and bone erosion [20,21]. Th ese contradictory observations of the in vivo role of IFN-γ may result from diff erences in the disease models and, more importantly, the impact of IFN-γ on various cell types. For example, recent data suggest that IFN-γ can not only directly inhibit diff erentiation of osteoclast precursors, but can also indirectly promote osteo clasto genesis by stimulating T-cell activation and secretion of the osteoclastogenic factors RANKL and TNF-α [22].
Type I IFNs, IFN-α and IFN-β, have also been implicated in suppression of bone resorption. During osteoclastogenesis, RANKL induces IFN-β expression in osteo clast precursors, and IFN-β, in turn, functions as a negative-feedback regulator to suppress osteoclast diff erentiation by decreasing c-Fos expression [23]. Mice defi cient in the type I IFN receptor component IFNAR1 spontaneously develop severe osteopenia with enhanced osteoclastogenesis due to interference of this feedback loop [23]. STAT3 and SOCS (Suppressor of cytokine signal ing) proteins downstream of Jak1 are also likely involved in the IFN-β-induced inhibition of osteoclastogenesis, and the ubiquitin-mediated degradation of Jak1 after RANKL stimulation may limit the suppressive eff ect of IFN-β on osteoclastogenesis [24][25][26]. IFN-α also blunts in vitro osteoclastogenesis, but exogenous IFN-α has no obvious eff ect on bone turnover in vivo. Interestingly, type I IFNs appear to protect from erosive arthritic lesions in the setting of an IFN-driven mouse model of systemic lupus erythematosus, potentially explaining the lack of erosive arthritis in human systemic lupus erythematosus [27].

Additional inhibitory cytokines: TRAIL, IL-12, IL-18, IL-6
TRAIL (TNF-related apoptosis inducing ligand), a TNF family member, impedes osteoclast diff erentiation [28] and induces apoptosis of osteoclasts [29]. IL-12 plays an inhibitory role in osteoclastogenesis, but it is still controversial whether IL-12 directly inhibits osteoclast precursors or targets other cell types such as stromal/ osteoblastic cells or T cells to indirectly suppress osteoclasto genesis [30]. Apoptosis induced by interactions between IL-12-induced FasL and TNF-α-induced Fas contributes to the inhibitory mechanisms of IL-12 in TNF-α-induced osteoclastogenesis [31]. IL-18 inhibits osteo clastogenesis by a variety of mechanisms, including stimulation of GM-CSF [32] and induction of IFN-γ and OPG. IL-18 alone or synergistically with IL-12 inhibits TNF-α-induced osteoclastogenesis through Fas-FasLinduced apoptosis. IL-18 is induced in rheumatoid arthritis, but contrarily it indirectly stimulates osteoclasto genesis via its induction of RANKL on synovial T cells. IL-6 has been regarded as a stimulator of osteoclastogenesis and bone resorption by stimulating osteoblastic/stromal cell-mediated osteoclast diff erentiation, but recent studies described an opposite eff ect of IL-6 that directly targets osteoclast precursors to suppress their diff erentiation [33,34].

Toll-like receptors and interplay with interferons
TLRs are the best characterized 'pattern recognition receptors' that recognize conserved microbial molecules and mediate immune and infl ammatory cellular responses to infection and microbial products and in some cases responses to endogenous factors generated during cell death, infl ammation, and tissue damage. Activation of various TLRs directly inhibits the early stages of RANKL-induced osteoclastogenesis [35,36]. Th e underlying molecular mechanisms include TLR-induced production of IFN-β that suppresses RANKL-induced c-Fos, and inhibition of NFATc1 by decreased JNK activation in response to TLR ligands [37]. However, in a human osteoclast culture system, TLRs can inhibit human osteoclastogenesis independently of type I IFNs [17]. TLR ligands can suppress human osteoclastogenesis by inhibiting expression of c-Fms, RANK and TREM2, thereby rendering osteoclast precursors refractory to M-CSF and RANKL stimulation [17]. Inhibition of RANK expression by TLRs was also observed in murine osteoclast precursors but to a lesser extent [17], suggesting that TLR-induced inhibition of osteoclastogenesis can be mediated by distinct IFN-dependent and IFN-independent mechanisms that can act in parallel. Moreover, TLRs cooperate with IFN-γ to inhibit osteoclastogenesis by synergistically suppressing expression of RANK and c-Fms [17]. Th ese data revealed a complex interplay between TLRs and IFN-γ in the inhibition of osteoclastogenesis, and new mechanisms by which TLRs and IFN-γ prevent osteoclast precursors from diff erentiating to osteoclasts, while directing them toward becoming infl ammatory macrophages. Interferon regulatory factor (IRF)-8, induced by IFN-γ, is a critical negative regulator for osteo clastogenesis in humans and mice, and its downregulation by RANKL is essential for osteoclastogenesis [38]. We found that RANKL-induced downregulation of IRF-8 is abrogated by TLR activation (Zhao B et al., unpublished data). Th e inhibitory eff ect of TLRs on osteo clastogenesis is compromised by IRF-8 defi ciency [38], suggesting that regulation of IRF-8 is involved in the mechanisms by which TLRs and IFN-γ inhibit osteoclastogenesis.
TLRs are activated during acute infection, during chronic microbial colonization and invasion as occur in periodontitis, and during chronic sterile infl ammation as occurs in rheumatoid arthritis, most likely by tissue degradation products. TLRs are highly expressed on hematopoietic cells and are also expressed on various other cell types, including epithelial cells, fi broblasts, and osteoblasts. Th erefore, it is not surprising that, in contrast to their direct inhibitory eff ect on osteoclast precursors, TLRs can stimulate infl ammatory osteolysis in vivo by aff ecting various cell populations and by distinct mechanisms. TLRs have been implicated in the induction of RANKL and TNF-α expression on osteoblastic/ stromal cells and thus are involved in stimulating osteoblast/stromal cell-mediated osteoclastogenesis and bone resorption [39]. In addition, TLRs are among the most potent inducers of infl ammatory cytokines such as TNF-α and IL-1, which then act to increase RANKL expression on stromal cells and also synergize with RANK signals to drive osteoclastogenesis. Furthermore, TLR activation accelerates diff erentiation of committed osteoclasts, and promotes mature osteoclast survival [39][40][41]. Th us, the net eff ect of TLRs on osteoclastogenesis in vivo is mediated by various cell types and is determined by the potency of pro-osteoclastogenic versus antiosteoclastogenic mechanisms.

Cytotoxic T-lymphocyte antigen 4 and regulatory T cells
Recent exciting work has identifi ed a role for regulatory T cells (Tregs) in restraining osteoclastogenesis and limit ing bone resorption [42,43]. Tregs suppress osteoclast precursors directly by a mechanism predominantly dependent on cytotoxic T-lymphocyte antigen 4 (CTLA-4). CTLA-4 is expressed on the surface of activated T cells and Tregs and transmits an inhibitory signal to T cells after binding to its cognate ligands, CD80 and CD86 (also known as B7.1 and B7.2), on antigen-presenting cells. Recent work showed that CTLA-4, which is constitutively expressed by Tregs, directly inhibits osteoclast formation by binding to CD80 and CD86 expressed by osteoclast precursors. Th is suggests that CTLA-4-mediated ligation of its counter-receptors CD80 and CD86 delivers a negative signal to osteoclast precursors, and provides a potential new explanation for the anti-erosive eff ect of abatacept, a CTLA-4 immunoglobulin fusion protein used for the treatment of rheumatoid arthritis [42,43].

NF-κB p100
Th e NF-κB family comprises RelA (p65), RelB, c-Rel, NF-κB1 (p50 and its precursor p105), and NF-κB2 (p52 and its precursor p100). NF-κB activation is elicited by two major signaling pathways; the classical pathway mainly involves IκB kinase-β-induced IκBα degradation and subsequent RelA/p50 activation, and the alternative pathway involves NF-kappa-B-inducing kinase-induced p100 processing to p52 and RelB/p52 activation. Th ere is crosstalk between these two pathways, and NF-κB activation from these two pathways plays important positive roles in inducing osteoclastogenesis [2]. On the other hand, recent emerging evidence shows that NF-κB p100 functions as a negative regulator of osteo clastogenesis by binding to NF-κB complexes and preventing their nuclear translocation. Cytosolic accumulation of p100 impairs osteoclastogenesis, whereas p100 defi ciency leads to enhanced osteoclastogenesis that contributes to an osteopenic phenotype in vivo [44,45]. TNF-α, unlike RANKL, does not seem to activate the alternative NF-κB pathway effi ciently, as it induces an accumulation of p100 in osteoclast precursors via induction of TRAF3, thus limiting TNF-α-induced osteoclastogenesis [44]. TNF-Tg mice lacking NF-κB p100 exhibit more severe joint erosion than that of TNF-Tg littermates [44]. Although there is some controversy whether TNF-α positively regu lates osteoclastogenesis [44,46], these data suggest that blockade of NF-κB p100 processing might represent a novel therapeutic strategy for infl ammatory bone loss as occurs in RA.

Cytosolic phosphatase: SHIP1
SHIP1 (Src homology 2-containing inositol-5-phos phatase 1) is preferentially expressed in hematopoietic cells, including T and B lymphocytes, mast cells and macrophages. M-CSF induces tyrosine phosphorylation of SHIP1 and the association of SHIP1 with c-fms via the adaptor protein Shc, whereby SHIP1 specifi cally dephosphorylates phosphatidylinositol 3,4,5-triphosphate and thus inactivates phosphatidylinositide-3-kinase/Aktmediated signaling. Genetic evidence from SHIP1defi cient mice showed that SHIP1 negatively regulates osteoclast formation and function. Compared to wildtype mice, SHIP1-defi cient mice exhibit increased prolifera tion of osteoclast precursors with hypersensitivity to M-CSF and RANKL, and increased osteoclasts with prolonged survival and enhanced bone resorptive activity, thus leading to an osteoporotic phenotype [47]. SHIP1 suppresses osteoclastogenesis and bone erosions in K/BxN mouse serum-induced infl ammatory arthritis models [48]. Th e underlying mechanisms of the suppressive eff ect of SHIP1 on osteoclastogenesis involve negative regulation of M-CSF-dependent Akt activity and consequent negative regulation of D-type cyclins, upregulation of cyclin-dependent kinase inhibitor p27, and negative regulation of retinoblastoma and cell proliferation [48]. A recent study revealed a novel mechanism in which SHIP1 interacts with DAP12 (DNAX-activating protein of 12 kDa) via its SH2 domain, thereby directly blocking the binding and activation of phosphatidylinositide-3-kinase, and thus limiting TREM2-and DAP12-mediated co-stimulatory signaling for osteo clastogenesis [49]. It is also interesting to note the morphological and functional similarities between SHIP1 knockout osteoclasts and osteoclasts in patients with Paget's disease, and similar high IL-6 expression [47]. However, the possibility of SHIP1 involvement in Paget's disease requires genetic analysis and additional supporting evidence.

Notch signaling pathway
Th e Notch signaling pathway regulates cell proliferation, diff erentiation and survival. In mammalian cells, there are four Notch receptors (Notch 1 to 4) and fi ve notch ligands (Jagged1, Jagged2, Delta-like (DLL)1, DLL3, and DLL4). Ligation of Notch receptors by their ligands leads to proteolytic cleavage of Notch by ADAM family proteases that releases the extracellular domain followed by intramembranous cleavage by γ-secretase that releases the Notch intracellular domain. Th e Notch intracellular domain translocates to the nucleus, binds to the DNAbinding protein RBP-J (recombinant recognition sequence binding protein at the Jκ site; also named CSL or CBF1), and activates Notch target genes such as Hes and Hey. Induction of Notch ligand Jagged1 and expression of Notch receptors 1, 2, and 3 were observed during RANKL-induced osteoclastogenesis [50][51][52]. Some investi gators found that activation of the Notch signaling pathway inhibits RANKL-induced osteoclast diff erentiation [50,51], whereas others described the opposite [52]. Th e genetic evidence obtained by using bone marrow-derived macrophages from Notch 1/2/3 knockout mice or Notch 1 or Notch 3 knockout mice, however, solidify the fi nding that Notch negatively regulates osteoclastogenesis [51]. Th e osteoclast inhibi tory mechanisms include the suppression of osteoclast precursor proliferation by Notch, likely through inhibi tion of the expression of the M-CSF receptor c-Fms [51]. On the other hand, Notch also indirectly blunts osteoclastogenesis by aff ecting osteoblastic/stromal cells to decrease the OPG/RANKL ratio [51] or M-CSF gene expression [50]. However, it should be noted that the inhibitory eff ect of Notch on RANKL-induced osteoclastogenesis is modest since the mice with Notch 1/2/3-specifi c defi ciency in the osteoclast lineage do not exhibit signifi cant defects in physiological bone development [51]. In addition, Notch signaling plays an important role in proliferation, diff erentiation and expression of RANKL and OPG by osteoblast lineage cells [53][54][55], and thus indirectly regulates osteoclastogenesis in vivo. Th e role of the Notch pathway in infl ammatory bone resorption has not been investigated, and future studies in this area may reveal new opportunities for therapeutic intervention.

Transcriptional repressors: Ids, Eos, MafB, C/EBPβ, IRF-8, BcL6
Balanced osteoclast diff erentiation is precisely controlled and maintained by complex mechanisms at various levels. In the past two decades, extensive studies have focused on the activation of signaling cascades that lead to activation of transcription factors such as NF-κB, AP-1 and NFATc1 that promote osteoclast diff erentiation (Figure 2, right). More recently, accumulating evidence has revealed that transcriptional repressors expressed constitutively in osteoclast precursors function to oppose the action of RANK and to restrain osteoclastogenesis (Figure 2, left). Th us, in addition to activating positive signaling pathways, RANK needs to overcome the 'brakes' imposed on osteoclast diff erentiation by transcriptional repressors that include inhibitors of diff erentiation/DNA binding (Ids) [56,57], Eos [58], MafB (v-maf musculoaponeurotic fi brosarcoma oncogene family protein B) that is in turn induced by C/EBPβ (CCAATenhancer-binding protein β) [59], IRF-8 [38] and B cell lymphoma (Bcl)6 [60]. RANK signaling appears to overcome transcriptional repression of genes important for osteoclast diff erentiation and functions, at least in part, by downregulating expression of these transcriptional repressors. Th e need for removal of transcriptional repressors for osteoclast diff erentiation to occur highlights their critical roles in negative regulation of osteoclastogenesis.
Th e expression levels of the currently identifi ed negative transcription factors Id, Eos, MafB, IRF-8 and Bcl6 are downregulated by RANKL during osteoclastogenesis with diff erent kinetics. Ids, IRF-8 and MafB are decreased at the early stage of osteoclasogenesis, within 24 hours after RANKL stimulation, whereas Eos and Bcl6 expression appear to decrease at later time points. Forced expression of Id, MafB, IRF-8 or Bcl6 strongly inhibits RANKL-induced osteoclastogenesis in vitro. Eos targets Microphthalmia-associated transcription factor (MITF)/ PU.1 target genes for repression, whereas inhibition of NFATc1 induction by the other repressors represents a common mechanism of suppression of osteoclast diff erentiation. Id proteins associate directly with MITF to downregulate expression of osteoclast-associated recep tor (OSCAR) as well as NFATc1, without aff ecting the expression of TREM2, DAP12 or Fc receptor γ. MafB proteins interfere with the DNA-binding ability of c-Fos, MITF, and NFATc1, thereby inhibiting the transactivation of NFATc1 and OSCAR. IRF-8 binds to NFATc1 and suppresses its DNA binding ability and transcriptional activity, thereby inhibiting NFATc1 autoamplifi cation and expression of NFATc1 target osteoclast marker genes. Bcl6 directly binds to the promoters of NFATc1, dendritic cell-specifi c transmembrane protein (DC-STAMP) and cathepsin K, which are NFATc1 targets, to suppress osteoclastogenesis.
Defi ciency of IRF-8 [38], Id1 [57] or Bcl6 [60] in mice leads to enhanced osteoclast formation and diff erent extents of osteoporosis, indicating IRF-8, Id1 and Bcl6 play an inhibitory role in in vivo osteoclastogenesis and physiological bone metabolism. Th e role of MafB in physiological bone metabolism in vivo has not been reported. Expression of MafB, IRF-8 and Bcl6 is relatively selective for hematopoietic cells, whereas expression of Ids is observed in diverse cell types, including osteoblasts. Th us, the role of Ids seems to be more complex in vivo. Hypoxia-induced Id2 expression is found in rheumatoid arthritis synovial fi broblasts, and promotes synovial fi bro blast-dependent osteoclastogenesis [61]. Another study showed that overexpression of Id1 in prostate cancer cells has an important role in promoting prostate cancer-mediated osteoclast diff erentiation, probably via certain secreted factors [62]. Th erefore, the role of Id proteins during in vivo osteoclastogenesis in physiological and pathological conditions might be regulated by diff erent cells and dependent on a particular environment.
Th e role of IRF-8 in infl ammatory bone resorption was studied in vitro and in vivo [38]. Infl ammatory bone erosion stimulated by RANK signaling is enhanced by infl ammatory cytokines such as TNF-α that activate osteoclastogenesis directly or indirectly via activation of stromal cells and osteoblasts. IRF-8 defi ciency dramatically promotes TNF-α-induced osteoclastogenesis in vitro, and results in increased NFATc1 expression, indicat ing that IRF-8 has a suppressive role in TNF-αinduced osteoclastogenesis. IRF-8 defi ciency signifi cantly attenuates TLR-induced inhibition of osteoclastogenesis, suggesting IRF-8 plays an important part in the inhibitory mechanisms of TLRs. In a lipopolysaccharide-induced infl ammatory bone resorption model, IRF-8-defi cient mice exhibit enhanced osteoclast formation and more dramatic bone destruction than wild-type littermates. Th ese data indicate that this homeostatic role of IRF-8 may be important to limit bone resorption during acute infections and also in chronic infl ammatory conditions such as rheumatoid arthritis. IRF-8 expression is also downregulated during RANKL-induced human osteoclastogenesis and silencing of IRF8 mRNA in human osteoclast precursors with small interfering RNAs leads to enhanced osteoclast diff erentiation, indicating the function of IRF-8 in osteoclastogenesis is well conserved in humans and mice. Th e mechanisms by which the expression of these repressors is downregulated are largely unknown. Recently, the transcriptional repressor Blimp1 (B lymphocyteinduced maturation protein-1), which is induced by NFATc1 in response to RANKL stimulation, was shown to suppress the expression of IRF-8, MafB [63] and Bcl6 [60] (Figure 2). Blimp1 defi ciency attenuates downregulation of IRF-8, MafB and Bcl6 expression after RANKL stimulation, and thus Blimp1 promotes osteoclast diff eren tiation by suppressing expression of its repressors. Conversely, Bcl6 can regulate Blimp1 expression and IRF-8 can regulate Bcl6 expression. Th ese fi ndings suggest a complex network of transcriptional repressors that control osteoclast diff erentiation, and it will be important to identify RANKL-induced signaling pathways and upstream molecules that control this transcriptional network. It will be also interesting to clarify whether these transcriptional repressors mediate the eff ects of inhibitory cytokines and infl ammatory factors on osteoclasts. For example, factors that induce or maintain IRF-8 expression in the presence of RANKL would act to restrain osteoclast diff erentiation. IRF-8 expression is induced by IFN-γ, and augmented IRF-8 expression may contribute to the inhibitory eff ects of IFN-γ on osteoclastogenesis, and also to the well documented suppressive eff ects of TLRs on osteoclast precursor cells. Identifi cation of signaling pathways, additional factors, and mechanisms that regulate IRF-8 expression and function represents a promising approach to control infl ammatory bone loss.

Conclusion
Osteoclastogenesis in vivo is mediated by various factors, including cytokines, signaling molecules and transcription factors that directly aff ect osteoclast precursors and/ or indirectly mediate osteoclastogenesis by targeting other cell populations, such as osteoblastic/stromal cells, synovial cells and T cells. In the latter case, the balance of RANKL versus OPG is often regulated to modulate osteo clastogenesis. Both direct and indirect eff ects need to be studied to fully understand the regulation of osteoclastogenesis. In addition, many infl ammatory factors also infl uence osteoblast diff erentiation/function and osteoblastic bone formation, for example, the induction of Wnt pathway inhibitors Dickkopf (DKK) proteins and Frizzled-related proteins in infl ammatory arthritis [64,65]. Regulation of osteoblast diff erentiation will impact on RANKL/OPG expression [66,67] and anabolic function and thus play an important part in physiological and pathological bone turnover in vivo; discussion of osteoblast diff erentiation is beyond the scope of this review.
It is interesting that the eff ects of most direct inhibitors are highly dependent on the timing of exposure and inhibit most strongly when present prior to or shortly after RANKL administration ( Figure 1). Strikingly, exposure of pre-osteoclasts to TLR ligands and GM-CSF several days after the RANK-mediated osteoclast diff eren tiation program has been initiated actually results in increased osteoclastogenesis and bone resorption, possibly by mechanisms related to increased cell survival. Another attractive explanation for this timing pheno menon could be related to the downregulation of transcriptional repressors such as IRF-8 at the early stage of osteoclastogensis, thereby diminishing the suppressive function of infl ammatory factors that utilize these repressors to suppress osteoclastogenesis.
One key principle that we have tried to develop is that the extent of infl ammatory bone resorption is often deter mined by the balance between opposing factors. Th is includes not only the balance between positive osteoclastogenic factors and negative regulators, but also opposing eff ects of individual factors on diff erent cell types. A striking example of opposing eff ects is off ered by TLR ligands that promote osteoclastogenesis by activating RANKL expression on stromal cells, yet at the same time restrain the amount of bone resorption by directly inhibiting early osteoclast precursors. In acute infection or chronic infl ammatory diseases such as rheumatoid arthritis, osteoclastogenic factors, including RANKL, TNF-α and IL-1, are often predominant and/or osteoclast precursors in the infl ammatory microenvironment are refractory to inhibitors of osteoclastogenesis, such as IL-27, leading to excessive and pathologic bone resorption. Th us, identifi cation of additional mechanisms and factors that increase the potency of repressors or restore cellular responses to suppressive factors may represent eff ective therapies for bone loss.