Osteoclasts are multinucleated myeloid lineage cells formed in response to macrophage colony-stimulating factor (M-CSF) and receptor activator of NF-κB ligand (RANKL) by fusion of bone marrow–derived precursors that circulate in the blood and are attracted to sites of bone resorption in response to factors, such as sphingosine-1 phosphate signaling. Major advances in understanding of the molecular mechanisms regulating osteoclast functions have been made in the past 20 years, mainly from mouse and human genetic studies. These have revealed that osteoclasts express and respond to proinflammatory and anti-inflammatory cytokines. Some of these cytokines activate NF-κB and nuclear factor of activated T cells, cytoplasmic 1 (NFATc1) signaling to induce osteoclast formation and activity and also regulate communication with neighboring cells through signaling proteins, including ephrins and semaphorins. Osteoclasts also positively and negatively regulate immune responses and osteoblastic bone formation. These advances have led to development of new inhibitors of bone resorption that are in clinical use or in clinical trials; and more should follow, based on these advances. This article reviews current understanding of how bone resorption is regulated both positively and negatively in normal and pathologic states. © 2013 American Society for Bone and Mineral Research.
Osteoclastic bone resorption coupled with bone formation helps to maintain skeletal integrity and mineral homeostasis, but it also is responsible for localized or generalized bone loss, which can weaken the skeleton and increase the risk of fracture.1, 2 Osteoclasts (OCs) are specialized myeloid lineage cells that function most effectively as multinucleated cells. Osteoclast precursors (OCPs) arise in the bone marrow and, like other leukocytes, circulate in the blood, from where they are attracted to bone remodeling units (BRUs) in response to chemokines, cytokines, and other factors elaborated at sites either destined for or already undergoing resorption.3, 4
Bone loss and fractures can be prevented by antiresorptive drugs.5 However, atypical femoral fractures have been linked to long-term treatment of osteoporosis with bisphosphonates, the most widely prescribed antiresorptive drugs,6 and osteonecrosis of jaw bones has been reported in some patients given intravenous bisphosphonates6 or denosumab, a human monoclonal antibody to receptor activator of NF-κB ligand (RANKL), the major osteoclastogenic cytokine.7 These adverse events have resulted in increasing reluctance of patients with osteoporosis to take bisphosphonates long-term and point to the need for new antiresorptive therapies designed to target osteoclasts.
Morphologic studies of human bone biopsy samples in the 1960s, 1970s, and 1980s revealed how bone remodeling maintains skeletal integrity and is disrupted in common diseases.2 Mouse and human genetic studies since the 1980s have identified numerous signaling pathways regulating OC formation and function1, 8 as well as some unanticipated functions of osteoclasts to regulate osteoblastic and other cells.9 These studies should lead to development of new drugs targeted specifically to these signaling pathways, which are reviewed in this article.
Regulation of Osteoclast Formation
Cytokines and transcription factors
OCs form by cytoplasmic, but not nuclear, fusion of precursors derived from myeloid progenitor cells that give rise also to macrophages and dendritic cells.9, 10 The progenitor cells differentiate into OCPs in response to macrophage colony-stimulating factor (M-CSF) and RANKL, but expression of several transcription factors, including PU.1, and a heterodimeric complex of microphthalmia-associated transcription factor (MITF) and transcription factor binding to IGHM enhancer 3 (Tfe3),11 is required earlier in myeloid progenitors to promote their differentiation.12 For example, PU.1 and MITF promote expression of c-fms (the M-CSF receptor),12 and mice deficient in these develop osteopetrosis,13 a condition characterized by radiographically dense long bones in which trabecular bone formed during endochondral ossification is not removed due to failure of OC formation or activity.
M-CSF is an essential osteoclastogenic cytokine expressed by osteoblast lineage cells. It promotes expression of RANK on OCP cell membranes, leaving the RANK+ cells primed to respond to RANKL.14 It mediates OCP proliferation, differentiation, and survival through extracellular signal-regulated kinase (ERK)/growth factor receptor bound protein 2 (Grb-2) and Akt/phosphoinositide 3-kinase (PI3K) signaling.14 M-CSF also signals through a complex comprised of phosphorylated DNAX-activating protein 12 (DAP12) and the nonreceptor tyrosine kinase, spleen tyrosine kinase (Syk),14 which is also activated by costimulatory signaling. Thus, M-CSF has important roles in all aspects of osteoclastic bone resorption.
RANKL/RANK and downstream signaling
RANKL is a member of the tumor necrosis factor (TNF) superfamily of proteins and is expressed by osteoblast lineage and other cell types, including T and B lymphocytes.15 In the absence of essential molecules that signal downstream of RANK, such as NF-κB and c-Fos, increased numbers of CD11b+ OCPs accumulate (as in RANK–/– mice16 and NF-κBp50/p52 double knockout [dKO] mice17) or are diverted down the macrophage lineage (as in c-Fos–/– mice).18 Thus, treatment of patients with anti-RANKL drugs could lead to accumulation of OCPs, which could differentiate into OCs after therapy is discontinued. Such a mechanism could perhaps account for the increase in serum bone resorption markers reported in some clinical trials following cessation of denosumab,19 but the precise mechanism remains to be determined.
During endochondral ossification, growth plate chondrocytes express RANKL, RANK, and osteoprotegerin (OPG).20 1,25,(OH)2D3, bone morphogenetic protein 2 (BMP2), and Wnt/β-catenin signaling20–22 regulate RANKL expression by these cells to attract OCPs to growth plates and facilitate rapid removal of newly formed bone, thus preventing osteopetrosis.13 Hypertrophic chondrocytes are the major source of RANKL during endochondral ossification, not osteoblastic cells, as had been thought previously, and osteocytes in bone are the major source of RANKL during bone remodeling and in response to mechanical stress.23–25
Unlike c-fms, RANK lacks intrinsic kinase activity to phosphorylate and activate downstream signaling molecules. Rank recruits TNF receptor-associated factors (TRAFs), particularly TRAFs 1, 2, 3, 5, and 6, which function as adapter proteins to recruit protein kinases.26, 27 Of these, only TRAF6 appears to have essential functions in osteoclastic cells.26, 27 RANK/TRAF6 signaling activates four main pathways to induce OC formation: NF-κB; c-Jun N-terminal kinase (JNK)/activator protein-1 (AP-1); c-myc; and calcineurin/nuclear factor of activated T cells, cytoplasmic 1 (NFATc1); and two others to mediate osteoclast activation (sarcoma [Src]) and mitogen activated protein kinase kinase 6 [MKK6]/p38/MITF) and survival (Src and ERK),26–28 which are discussed in the “Regulation of osteoclast activation” section. TRAF2 positively and TRAF3 negatively regulates OC formation (see below).
NF-κB is a family of transcription factors that includes the signaling proteins, reticuloendotheliosis viral oncogene homolog-A (RelA), p50, Rel B, p52, and c-Rel, which induce expression of genes involved in normal and aberrant immune responses, cell division, differentiation, and movement, and carcinogenesis through canonical and noncanonical pathways.29, 30 Requirement of NF-κB in osteoclastogenesis was discovered unexpectedly when NF-κB p50/p52 dKO mice failed to thrive at weaning as a result of the absence of tooth eruption associated with osteopetrosis because the mice did not form OCs.31, 32 The defect in OC formation in NF-κB dKO OCPs is rescued by either c-Fos or NFATc1 retroviral constructs,33 indicating that they act downstream of NF-κB.
NF-κB appears to cooperate with NFATc2 (which is not required for OC formation) to induce expression of NFATc1, with NF-κB p50 and p65 being recruited to the NFATc1 promoter within 1 hour of treatment of OCPs with RANKL, resulting in transient autoamplification of NFATc1 expression.34 It remains to be determined what the precise role of this transient autoamplification of NFATc1 is and which genes are activated in response to it, but it may be downregulation of constitutively active repressors of RANK signaling35 (see “Constitutively-expressed transcriptional repressors of RANK signaling”).
In the canonical pathway, RANKL binding to RANK leads quickly to formation of a complex on the intracellular cytoplasmic portion of RANK that contains a number of proteins, including TRAF6 and transforming growth factor β (TGFβ)-activated kinase-1 (TAK1), which induce activation of inhibitor of nuclear factor kappa-B kinase subunit gamma (IKKγ; also called NF-κB essential modulator [NEMO]). This leads to phosphorylation and subsequent activation of IKKβ, which phosphorylates IκB, an inhibitory NF-κB family protein that holds p65 and p50 heterodimers in an inactive state in the cytoplasm. IκB consequently undergoes rapid degradation by the 26S proteasome, resulting in release of p65 and p50 and their translocation to nuclei where they prevent apoptosis of OCPs, thus allowing them to continue differentiating.36, 37 Mice with deletion of IKKβ in OC lineage cells have impaired OC formation and osteopetrosis.36 Interestingly, a constitutively active IKKβ (IKKβ-SS/EE) expressed in OCPs induces their differentiation into OCs in the absence of RANK or RANKL treatment,38 further emphasizing the importance of NF-κB signaling in OC formation.
Activation of the noncanonical pathway occurs more slowly, typically within 3 to 4 hours of RANKL treatment through the activity of NF-κB–inducing kinase (NIK). This leads to processing of the precursor molecule, p100, to p52, which typically signals in association with RelB. In unstimulated cells, newly synthesized NIK gets bound by TRAF3, leading to NIK proteasomal degradation.39 NF-κB activation by RANKL recruits a TRAF/cellular inhibitor of apoptosis 1 (cIAP) E3 ligase complex to RANK leading to cIAP1/2 activation by TRAF2. This targets TRAF3 for ubiquitination and degradation allowing NIK to accumulate and activate IKKα, which phosphorylates p100 and leads to its proteasomal processing to p52 and subsequent nuclear translocation of RelB/p52 heterodimers.29, 36
NIK, IKKα, and p100 do not have a required function for basal OC formation.29, 36 However, they do appear to have regulatory roles in RANKL-enhanced or TNF-enhanced OC formation in pathologic states. For example, intratibial injection of murine melanoma cells caused localized osteolysis in wild-type (WT), but not in NIK–/– mice.36 In contrast, TNF-transgenic (TNF-Tg) mice crossed with p100–/– mice developed earlier and more severe joint inflammation and bone erosion than TNF-Tg mice, indicating that p100 limits TNF-induced OC formation and inflammation.40 These studies suggest that strategies to inhibit NIK or increase p100 could reduce bone loss in inflammatory and metastatic bone disease. Preclinical studies with a peptide that inhibits NF-κB signaling by binding to NEMO reduced osteoclastogenesis and bone erosion in inflammatory arthritis.41 However, to date there have been no clinical studies reported with this agent.
NFATc1 and costimulatory signaling
NFAT transcription factors regulate immune responses as well as cardiovascular, muscle, and neuronal and other cell functions.42 NFATc1 is activated in OCPs by being dephosphorylated by calcineurin, a phosphatase, which is activated by calcium-calmodulin signaling34, 43 mediated by phospholipase Cγ (PLCγ), which plays a key role by releasing calcium from stores within the cytoplasm.34, 44 NFATc1 is also activated through PLCγ by costimulatory signaling, which is initiated by ligand binding to immunoglobulin (Ig)-like receptors, such as triggering receptor expressed in myeloid cells-2 (TREM-2) and osteoclast-associated receptor (OSCAR).34 These receptors are expressed on OCPs and they recruit adaptor molecules, such as Fc receptor common γ subunit (FcRγ) and DAP12, leading to phosphorylation of immunoreceptor tyrosine-based activation motifs (ITAMs) within these adaptor proteins and activation of downstream signaling. NFATc1 is involved in all aspects of osteoclast formation and activation and seems like a prime target for anti-osteoclast therapy. Indeed, the immunosuppressive agents and calcineurin/NFATc1 inhibitors, tacrolimus (FK506) and cyclosporine-A, prevent bone loss in inflammatory arthritis because they reduce the inflammation and associated bone resorption.45 However, NFATc1 also positively regulates expression of osterix, a key osteoblastogenic transcription factor, and the net effect of these inhibitors in normal mice is bone loss.45
The ligands for most costimulatory receptors remain unidentified, but OSCAR is activated in OCPs by portions of exposed collagen fibers in resorption lacunae.46 Activation of NFATc1 through RANK and OSCAR in turn induces increased OSCAR expression on OCPs in a positive feedback loop.47 Expression of OSCAR and RANKL is increased in the synovium of joints of patients with rheumatoid arthritis (RA).48 Thus, costimulatory signaling likely enhances OC formation and bone resorption mediated by RANKL through this and other mechanisms in RA.
T and B lymphocytes and osteoimmunology
The recognition that RANKL is expressed not only by osteoblastic cells, but also by T and B cells and synoviocytes in inflammatory bone diseases and that RANK signaling is involved in immune responses, lymph node formation, and B cell maturation27, 44 spawned the new field of osteoimmunology.49 However, the contributions of T and B cells to the increased osteoclastogenesis in inflammatory bone disease are complex. For example, although T helper (Th) cells express RANKL, T regulatory cells (Tregs) inhibit OC formation through cytotoxic T lymphocyte antigen 450, 51 and production of interleukin (IL)-4 and IL-1035 and Th1 cells express interferon-γ (IFN-γ), which inhibits OC formation. Both T cell types are present in inflamed joints of RA patients and the effects of T cells overall appear to be inhibitory.52 Th17 cells are the major subset of RANKL-expressing Th cells in inflamed synovium of RA patients.53 In inflamed synovium, they54 and mast cells55 express IL-17, which induces OC formation mainly by increasing RANKL expression by synovial fibroblasts,52 although one study has reported direct induction of OC formation by IL-17 from human monocytes.56 Th17 cells are also inflammatory and cause increased expression of TNF, IL-1, and IL-6 by synovial fibroblasts, which in turn increase their expression of RANKL.52 Interestingly, adoptive cotransfer of a subset of CD11b–/loLy6Chi (lymphocyte antigen 6 complex) OCPs with CD4+ T cells from arthritic mice markedly decreased the severity of arthritis in recombination activating gene 2 (Rag2–/–) recipient mice, suggesting that these subpopulations of OCPs and T cells can be anti-inflammatory.57 There are also conflicting data about B cell expression of RANKL.43, 51, 58 Further study will be required to explain how these complex positive and negative functions of immune cells lead overall to increased bone resorption in inflammatory bone disease, but they point to additional mechanisms to limit bone resorption.
T cells have also been implicated in ovariectomy (Ovx)-induced bone loss in mice, but these findings also are somewhat controversial.15 For example, T cell–deficient nude mice appear to be protected from bone loss after Ovx in some, but not all studies.15 Estrogen inhibits differentiation of Th17 cells, but the role of IL-17 in Ovx-induced bone loss is unclear because there are conflicting findings of the effects of Ovx on bone loss in IL-17 receptor–deficient mice.15 Estrogen also increases Treg numbers; but it also regulates T cell production of TNF by inhibiting expression of IL-7, which promotes OC formation. In contrast, estrogen deficiency expands the pool of TNF-producing T cells, whereas transgenic mice overexpressing Tregs are protected against Ovx-induced bone loss.15, 59 Some of the discrepancies among these studies may result from differences in the strains of mice used, in study design, or to the positive effects of one set of T cells being negated by those of another set, as appears to occur in RA.
A further twist to the role of T cells in Ovx-induced bone loss is that OCs can function as antigen-presenting cells and thus can behave as immune cells to activate T cells.10 For example, they express Fc receptor common γ subunit (FcRγ), major histocompatibility complex (MHC) molecules, CD40, and CD80 (60), just like dendritic cells60 and express a wide range of cytokines. Therefore, OCs could participate in Ovx-induced T cell proliferation and activation along with or in place of dendritic cells. This positive role could be negated, however, because OCs can also inhibit T cell proliferation and suppress T cell production of TNF and IFNγ.61 These positive and negative effects of immune cells, cytokines, estrogen, and estrogen deficiency emphasize the fact that even in pathologic conditions there are mechanisms to limit excessive tissue destruction.
RANKL/RANK mutations cause osteopetrosis in humans
Kindreds with RANKL or RANK deletion mutations have marked osteopetrosis and appear to lack palpable lymph nodes.62, 63 However, obvious immunodeficiencies have not been reported in any of them, suggesting that they may have compensatory mechanisms that maintain normal immune responses. Interestingly, mice deficient in RANK specifically in B cells have normal B cell development.64 This may be important for patients being treated with anti-RANKL drugs, such as denosumab, because these findings in mice suggest that they might not interfere with B cell maturation. To date, no significant increase in infections or other signs of impaired immune responses have been reported in patients in clinical trials of denosumab.19
Activating mutations in the RANK gene are responsible for a number of rare bone diseases, including familial expansile osteolysis, and expansile skeletal hyperphosphatasia,65 in which there is increased localized, rather than generalized OC formation and bone resorption. This focal involvement has similarities to adult Paget's disease, many cases of which have mutations in genes encoding molecules that signal downstream of RANK,66 but it is not known why skeletal involvement is not diffuse in these diseases.
Recruitment of Osteoclast Precursors to Remodeling Sites
OCPs circulate in the blood from where they are attracted to BRUs.4 Morphologic studies have identified a thin canopy of connective tissue covering resorption sites, and small vessels pass through it into BRUs, bringing cells and nutrients.67 Resorbing OCs pass the products of resorption through their cytoplasm within lysosomes into the extracellular space within BRUs and from there these products enter efferent vessels and the circulation.68
Chemokine attraction of OCPs
Stroma-derived factor-1 (SDF-1) is a chemokine that mediates leukocyte migration, and its local concentration can determine the location of cells. For example, in TNF-mediated inflammatory arthritis, TNF inhibits SDF-1 production by marrow cells, leading to mobilization of OCPs from the marrow69 and to increased numbers of them in the bloodstream70 from where they can be attracted to inflamed joints by high SDF-1 concentrations.
OCPs are attracted to the bloodstream by sphingosine-1 phosphate (S1P), a bioactive sphingolipid with numerous functions, including regulation of cell motility, proliferation and survival.71 S1P is secreted from red blood cells and platelets, resulting in higher concentrations in serum than in the marrow. OCPs express S1P receptors (S1PRs) 1 and 2. Signaling through these receptors can have different effects. For example, S1PR1 signaling chemo-attracts OCPs from the marrow to the blood, whereas S1PR2 signaling appears to chemo-repel them back to the marrow.71 FTY720, an S1PR1 agonist, but not an S1PR2 agonist, prevented Ovx-induced bone resorption in mice, whereas more OCPs were attached to bone surfaces in S1PR1–/– mice, associated with increased OC formation and bone resorption.71
These findings raise the possibility that low serum S1P levels or mutations in S1PRs could be associated with increased bone resorption and osteoporosis in some patients. S1P levels are increased in the synovial fluid of patients with RA, and FTY720 significantly reduced joint destruction and inflammation in mice with inflammatory arthritis.71 Collectively, these findings suggest that drugs that can promote OCP migration to the bloodstream or prevent them from leaving it could reduce bone loss in common bone diseases. However, S1P has complex roles in inflammation and cytokine expression, with positive and negative regulatory functions, which may make use of agonists challenging.71
Regulation of Osteoclast Activation
Inactive, quiescent bone surfaces are covered by bone lining cells, the cytoplasm of which must be retracted, exposing matrix proteins such as vitronectin, before bone resorption can begin.72 Lining cells communicate with osteocytes within the bone and with osteoblastic cells in the marrow through their extensive dendritic processes, which could mediate signaling that leads to the cytoplasmic retraction,73 but the mechanisms involved remain unknown. OCs form a tight sealing zone with the exposed bone surface using actin filament-rich podosomes that are surrounded by adhesion, signaling, and adaptor molecules, protein tyrosine kinases, and actin-associated molecules, such as vinculin, talin and paxillin, which are involved in multiple facets of cell movement in normal and pathologic conditions.74
Integrin-mediated and cytokine-mediated OC attachment to matrix and activation
OCs attach to the bone surface mainly through the vitronectin receptor, αVβ3 integrin,75 in association with kindlin-3,76 a member of a family of proteins that are recruited to integrin adhesion sites in platelets and leucocytes. Kindlin-3 activates β3 integrin during the early events in OC activation. Humans with kindlin-3 gene mutations and kindlin-3–/– mice are osteopetrotic because their OCs do not form podosomes.77
Integrin binding to bone matrix proteins, such as osteopontin and bone sialoprotein, activates αVβ3 and recruits Src tyrosine kinase by standard outside-in signaling.78 Src phosphorylates Syk, which recruits the costimulatory ITAM protein, Dap12, and SH2 domain-containing leukocyte protein of 76 kDa (Slp76), and these function as an adaptor protein complex for Vav3, a guanine nucleotide exchange protein that activates the small guanosine-5′-triphosphatase (GTPase) Rho family members, Cdc42 and Ras-related C3 botulinum toxin substrate (Rac).78 αVβ3 and c-fms interact physically, and by inside-out signaling through αVβ3 cause a structural change in αVβ3, which is required for its activation.78 Interestingly, β3 integrin–/– mice have only mild osteopetrosis, perhaps because other integrins can substitute for it in OCs.78 RANK is also physically linked to αVβ3 by Src, forming a complex that activates Syk, Slp76, Vav3, and Rac, and in this respect is similar to αVβ3/Src interaction.78 Because the OC is the only cell that forms a ruffled membrane to resorb bone, it is possible that components of this activation step could be targets of future antiresorptive drugs.
OC ruffled border formation and bone resorption
Inside sealing zones formed by podosomes, the OC surface area facing the bone is increased significantly by the ruffled border membrane, which is formed by accumulation of cytoplasmic lysosomal secretory vesicle fusion with the cytoplasmic membrane and requires expression of Src.79 Vesicle fusion is promoted by the small GTPase Ras-associated protein RAB7 (Rab7) and synaptotagmin VII, a calcium-sensing molecule, and by proteins involved in autophagy and extracellular protein secretion, including Autophagy 5 (Atg5), Atg7, Atg4B, and light chain 3 (LC3).80, 81 Accordingly, synaptotagmin VII−/− osteoclasts have severely defective ruffled border formation.82 H+ and Cl− ions pass through the ruffled border and form HCl to dissolve the mineral component of bone, and proteolytic enzymes, particularly cathepsin K, are secreted to degrade the matrix.75 H+ ions are secreted through the V-type H+ ATP6i proton pump complex, whereas Cl− ions pass through a chloride channel encoded by chloride channel 7 (ClCN7).
Src phosphorylates proteins involved in OC activation, including Syk, proline-rich tyrosine kinase 2 (Pyk2), cortactin, and Casitas B-lineage lymphoma (c-Cbl), which has ubiquitin ligase activity.83 It also mediates RANKL-induced survival signaling in vitro,84 but src−/− OCs have normal survival in vivo,79, 83 perhaps because other Src family members substitute for it. Src is overexpressed in many cancers, in which it plays positive roles in proliferation, invasion, and metastasis, and is thus a therapeutic target in both OCs and tumor cells in metastatic bone disease.83 Small molecular inhibitors of Src have been developed, and of these saracatamib is currently being investigated in metastatic prostate cancer, with some promising results as an adjuvant to standard chemotherapy.83 To date no Src inhibitors have been studied in osteoporosis clinical trials.
High OC nuclear numbers correlate with more aggressive resorption, as is seen in Paget's disease and giant cell tumor of bone. OCP fusion is regulated by dendritic cell–specific transmembrane protein (DC-STAMP), Atp6v0d2, OC-STAMP, and CD9.85 Atp6v0d2 is a subunit of V-ATPase, a component of the V-type H+ ATP6i proton pump complex, which also is involved in OCP-mediated inhibition of osteoblast precursor formation,86 one of a number of unanticipated roles for OCPs and OCs in the regulation of bone formation.9
NFATc1 and c-Fos play major roles in OCP fusion and activation and in conjunction with MITF and PU.1 variously regulate expression of a number of genes, including, DC-STAMP, OC-STAMP, OSCAR, tartrate-resistant acid phosphatase, cathepsin K, V-ATPase-d2, and the calcitonin receptor.12, 87, 88 Vitamin E (α-tocopherol) also regulates OCP fusion by inducing DC-STAMP expression through activation of mitogen-activated protein kinase 14 and MITF.89 Importantly, administration of α-tocopherol to rats at doses taken by some humans as dietary supplements increased OC numbers in the animals and reduced bone mass, suggesting that excessive Vitamin E consumption could adversely affect bone health.89
OC-rich osteopetrosis in humans due to defects in genes regulating OC functions
Most cases of osteopetrosis in humans result from mutations in genes involved in matrix demineralization and dissolution. These include: T-cell immune regulator 1 (TCIRG1), which encodes the α3 subunit of the H+ ATPase involved in proton generation; carbonic anhydrase II, which catalyzes hydration of CO2 to H2CO3 to provide a source of H+; ClCN7, through which chloride ions pass; Pleckstrin homology domain-containing family M member 1, which encodes a vesicle-associated protein linked to small GTPase signaling; and cathepsin K.13, 90, 91 Humans with cathepsin K mutations develop an osteochondrodysplasia, called pycnodysostosis, the features of which include osteopetrosis, dwarfism, and defects of the craniofacial bones. In contrast, cathepsin K−/− mice have osteopetrosis, but no other bone defects,5 suggesting a more complex role for the gene in humans and raising the possibility that cathepsin K inhibitors could have adverse effects on nonstable fracture healing, in which some features of endochondral ossification are recapitulated.
Drug candidates have been developed to inhibit the proton pump and cathepsin K in OCs, but most of these have failed because of adverse effects in clinical trials. However, odanacatib, a cathepsin K inhibitor, showed promise in phase 2 clinical trials for the treatment of postmenopausal osteoporosis.92, 93 It inhibits bone resorption, but does not appear to inhibit bone formation to the same extent as other antiresorptive drugs, by mechanisms that remain to be explained, but which could involve OC-stimulated bone formation.94
Negative Regulation of OC Formation
Osteoprotegerin (OPG) is the major negative regulator of bone resorption.26, 27 OPG is secreted by osteoblastic and other cell types and binds to RANKL as a decoy receptor, thus preventing its interaction with RANK. Most of the factors, including growth factors and cytokines, that upregulate expression of RANKL also increase OPG expression, typically to a lesser extent, and thus the net effect is increased bone resorption.95 Other cytokines, such as, IL-4 and IL-13, which are produced by Th2 lymphocytes, enhance OPG and inhibit RANKL expression in osteoblastic cells and therefore suppress osteoclastogenesis.35 As will be seen later (in “Negative regulation by proinflammatory cytokines”), there are multiple additional mechanisms whereby cytokine signaling can limit OC formation and activation.
Interestingly, some of the signaling that regulates osteoblast formation also regulates OPG expression. For example, Wnt/β-catenin canonical signaling, which is required for osteoblast formation, positively regulates OPG expression in osteoblasts,96 whereas Wnt 5a-induced noncanonical signaling in osteoblasts positively regulates OC formation through receptor tyrosine kinase-like orphan receptor (Ror) proteins expressed in OCPs.97 This is a potentially important target for therapeutic intervention because a soluble form of Ror2 acted as a decoy receptor of Wnt5a and prevented bone destruction in mouse models of arthritis.97 In addition, Jagged1/Notch1 signaling, which regulates the number of mesenchymal stem cells (MSCs) and osteoblast differentiation, alters the OPG/RANKL expression ratio in stromal cells to inhibit OC formation.98
These findings are consistent with a model in which immature osteoblastic cells interact with osteoclastic cells through Wnt 5a near the cutting edges of BRUs to promote OC formation, whereas more mature osteoblastic cells express OPG through Wnt canonical and Notch signaling to inhibit osteoclastogenesis and promote OC apoptosis through OPG near the reversal site in BRUs, where osteoblastic cells can differentiate into matrix-forming osteoblasts.
Loss-of-function mutations of TNFRSF11B, the gene encoding OPG, occur in humans and account for most cases of juvenile Paget's disease.99 The mutation results in OPG deficiency and unopposed RANKL-induced bone resorption with osteoporosis and long-bone and vertebral deformities during childhood; the phenotype is similar to that seen in OPG−/− mice.100
Some inhibitory mechanisms are actually mediated by RANKL signaling itself within OCPs. For example, activation of c-Fos also induces expression by OCPs of IFN-β, which binds to the IFN-α/β receptor on OCPs, leading to inhibition of OCP differentiation by posttranscriptional reduction in c-Fos protein.101 Mice deficient in IFN-β or in a component of the IFN-α/β receptor have severe osteopenia due to increased OC formation and activity, emphasizing how important this mechanism is.101 IFN-β is used to prevent disease flares in multiple sclerosis with significant efficacy. Although some studies have reported beneficial effects on bone mineral density, patient numbers have been low, and this warrants further study.
Ephrin/Eph and semaphorin/neuropilin/plexin signaling (and osteoclast regulation of osteoblasts)
Ephrins and semaphorins are widely expressed molecules that control communication between cells, including neurons and axons during nervous system development, and endothelial cells and lymphocytes during immune responses and angiogenesis.102–104 These molecules are also expressed in bone and regulate interactions between and functions of osteoclastic and osteoblastic cells.105–107 For example, RANKL-induced c-Fos/NFATc1 signaling increases expression of the ligand, ephrinB2, on the surface of OCPs. Reverse signaling through this ligand when it binds directly to the Eph4 receptor on osteoblastic cells downregulates c-Fos and NFATc1 expression to limit OC formation; forward signaling through Eph4 stimulates osteoblast precursor differentiation by inhibiting the small GTPase, RhoA.105 Decreased ephrinA1 and EphA1 expression was identified in bones of patients with metastatic prostate cancer108 and giant cell tumor of bone109 by mRNA microarray analysis, implicating reduced Ephrin-Eph signaling in osteolytic bone disease.
Semaphorins (Semas) are expressed widely as secreted and membrane-associated proteins; the latter signal through plexins and the former through neuropilins (Nrps). Sema3A is secreted by osteoblasts and OCs, and its binding to Nrp1 on OCPs inhibits RANKL-induced OC formation by inhibiting ITAM and RhoA signaling.110 It also binds to Nrp1 on mesenchymal precursors to stimulate osteoblast and inhibit adipocyte differentiation through canonical Wnt/β-catenin signaling. Accordingly, Sema3A and Nrp1−/− mice have osteoporosis with reduced bone formation. Importantly, treatment of mice with Sema3A inhibited bone resorption and increased bone formation in normal mice and enhanced bone regeneration in a mouse cortical bone defect model.110 Sema4D is membrane-bound and binds to plexin1 on target cells. It is expressed by OCs and inhibits osteoblast differentiation and function by activating RhoA-ROCK, which inhibits insulin-like growth factor-1 signaling.111 Consistent with these findings, sema4D−/− and plexin1−/− mice have high bone mass due to increased bone formation.111 Sema6d is membrane-bound, and by binding to plexin-A1 on OCPs induces OC formation through Trem-2/DAP12/PLCγ-induced NFAT activation as well as podosome formation through Rac-GTP generation in OCs. Accordingly, plexin1−/− mice have marked osteopetrosis, but normal osteoblast function.107 Sema7A is expressed by osteoclasts and osteoblasts and induces monocyte production of free radicals, IL-6, and TNF, suggesting that it may play a role in inflammatory bone disease.107 It promotes OC formation and OCP fusion as well as osteoblast migration in vitro, but full understanding of its role in bone awaits generation of Sem7a−/− mice.
These studies and other studies9 have revealed complex interactions between osteoclastic and osteoblast cells and that OCs and OCPs have important positive and negative regulatory roles in normal and pathologic bone remodeling that had not been anticipated a decade ago. Further studies will be required to determine exactly where and when these interactions take place in remodeling units to initiate and subsequently stop both resorption and formation, but they suggest that drugs could be developed to enhance or restrict some of these interactions to increase bone mass.
Constitutively-expressed transcriptional repressors of RANK signaling
There are also constitutive mechanisms to inhibit basal OC formation. For example, in the absence of RANKL stimulation, B-cell leukemia/lymphoma 6 (Bcl6) is recruited to the NFATc1, cathepsin K, and DC-STAMP promoters to inhibit osteoclastogenesis. In contrast, RANKL induces removal of Bcl6 from these promoters and its replacement by NFATc1, to mediate osteoclastogenesis.112 Bcl6 is one of a group of constitutively-expressed transcriptional repressors in OCPs, including interferon regulatory factor-8 (IRF-8), Eosinophilia (Eos), and v-maf musculoaponeurotic fibrosarcoma oncogene family protein B (MafB).35 Their expression is downregulated by RANK signaling. They are also direct targets of B lymphocyte–induced maturation protein 1 (Blimp-1), deletion of which in OCs results in osteopetrosis due to upregulation of Bcl6 and impaired osteoclastogenesis.112 In contrast, Bcl6−/− mice have increased OC formation and severe osteoporosis. Thus, RANKL/RANK activation of NFATc1 in OCPs not only promotes osteoclastogenesis directly, but it also facilitates it indirectly by repressing expression of negative regulators.
Negative regulation by proinflammatory cytokines
TNF-induced and RANKL-induced translocation of NF-κB RelA/p50 complexes to nuclei induces increased expression of p100, which is efficiently processed to p52 by NIK in response to RANKL, but not to TNF.40 Thus, in conditions such as RA, in which TNF expression is increased, excess p100 protein becomes available to bind to RelB and/or RelA and in this way p100 can inhibit NF-κB signaling and limit OC formation.40 This accumulation of p100 is accompanied by an increase in TRAF3 in the cells and this in part explains why TNF only modestly increases OC formation in WT mice and does not induce OC formation when given to either RANKL−/− or RANK−/− mice.40 However, when these RANKL−/− or RANK−/− mice are also deficient in p100, TNF induces OC formation in vivo.40 TNF also limits OC formation in RA by inducing expression of IRF-8 and the Notch-induced DNA binding molecule, recombinant recognition sequence binding protein at the Jκ site35 in OCPs and by promoting secretion by OCPs and OCs of TNF-stimulated gene 6 (TSG-6), which acts synergistically with OPG to inhibit OC activity by an autocrine mechanism.113 Although T cells express RANKL in rheumatoid joints to potentially increase bone resorption, they also secrete IFN-γ, which degrades TRAF6 in OCPs, and in this way T cells can also limit OC formation.114
During immune responses, anti-inflammatory cytokines, including IL-10, are elaborated to help resolve inflammation, and some of these can inhibit OC formation. For example, IL-10 inhibits expression of c-Fos, c-Jun, TREM-2, and NFATc1 in OCPs.35 IL-4 inhibits bone resorption by several mechanisms in addition to its effects on OPG expression. For example, it suppresses expression of RANK, NFATc1, and c-Fos, as well as NF-κB, mitogen-activated protein kinase (MAPK), and calcium signaling during osteoclastogenesis. In combination with granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-4 cleaves c-fms from the surface of OCPs to suppress osteoclastogenesis.35
During costimulatory signaling, ITAM-bearing proteins typically partner with proteins containing an immunoreceptor tyrosine-based inhibitory motif (ITIM),35 some of which promote and others inhibit immune responses and osteoclastogenesis. For example, Ly49Q promotes osteoclastogenesis by competing with the murine paired Ig-like receptor, subfamily B (PIR-B), for association with the SH2 domain-containing tyrosine phosphatase 1 (SHP-1).115 In contrast, the human inhibitory immunoglobulin-like receptor, leukocyte immunoglobulin-like receptor, subfamily B (LILRB), and paired immunoglobulin-like receptor B (PIR-B) recruit SHP-1 to negatively regulate osteoclastogenesis.115 An ITIM is present in the cytoplasmic tail of DC-STAMP, whose surface expression declines during osteoclastogenesis. DC-STAMP is phosphorylated on its tyrosine residues and physically interacts with SHP-1 and CD16, which is an ITAM-associated protein,116 and by binding to this complex DC-STAMP might actually limit OC formation.116 Finally, Toll-like receptor signaling is activated in monocytes by microbial products at sites of inflammation leading to enhanced immune responses, but in OCPs it leads to inhibition of their differentiation by causing the cells to shed the extracellular domain of c-fms.35 These advances in understanding of how bone and immune cells interact could lead to development of new therapies to inhibit both inflammation and osteoclastogenesis, but given the complexity involved, this is likely to be challenging.
Promotion of osteoclast apoptosis
One of the earliest reports of OC apoptosis was in response to estrogen and was mediated by increased expression of TGFβ by bone marrow cells.117 Estrogen also induces OC apoptosis by increasing Fas-ligand expression in OCs,118 and estrogen receptor alpha signaling also inhibits expression of genes regulating OC activity, without affecting OCP proliferation or fusion.119 Thus, estrogen can inhibit bone resorption by limiting OC lifespan and activity. Interestingly, TGFβ appears also to support OC survival directly through TAK1/MEK/AKT-mediated activation of NF-κB in OCs.120 In addition to its expression by marrow stromal cells, TGFβ is also released from bone matrix during resorption and is activated by the acid environment under the ruffled border. Thus, these opposite effects of TGFβ on OC survival may reflect the site and source of its production.
Bisphosphonates also induce OC apoptosis in vitro and in vivo,121 in part by inhibiting the activity of enzymes in the mevalonate pathway122 and promoting caspase cleavage of mammalian sterile 20-like (Mst) kinase 1, which is a proapoptotic substrate for the apoptosis effector enzyme, caspase 3.123 However, some amino-bisphosphonates appear to inhibit bone resorption without inducing OC apoptosis,124 consistent with their differing affinities for the binding sites of target enzymes.125 Despite proven efficacy in many pathologic settings, bisphosphonates have only modest antiresorptive activity in patients with RA,126, 127 which may be related to high TNF levels in their joints and blood and to glucocorticosteroid therapy. For example, TNF attenuates bisphosphonate-induced apoptosis by upregulating B-cell lymphoma-extra large (Bcl-XL) expression in OCPs and OCs,128 and glucocorticosteroids can inhibit OC apoptosis,129 although the mechanism remains to be determined. Denosumab130 and raloxifene131 induce OC apoptosis, but other antiresorptive drugs, including calcitonin68 and the cathepsin K inhibitors odanacatib and ONO-5334, which are in phase 3 clinical trials, do not.94
OCPs are recruited continuously to the cutting edges of resorption lacunae to maintain a population of relatively young resorbing OCs at this site, whereas older OCs undergo apoptosis predominantly in reversal sites of resorption lacunae,132 where high extracellular calcium concentrations resulting from bone resorption133 and OPG released by osteoblastic cells can induce OC apoptosis. However, OPG can also bind to and inhibit TNF-related apoptosis-induced ligand (TRAIL), which induces OC apoptosis. OPG appears to reduce apoptosis of human OCs in vitro by inhibiting this mechanism,134 but further studies are required to determine if this mechanism has a functional role in vivo.
Prevention of osteoclast apoptosis
An early effect of RANKL signaling in OCP differentiation is upregulation of c-Jun N-terminal kinase (JNK) signaling, which rather surprisingly was found to induce apoptosis of NF-κB p65-deficient OCPs by activating BH3 interacting domain (Bid) and caspase 3.37 These findings indicate that p65 plays an important role to prevent OCP apoptosis. They also show that p65 is not required for expression of genes that regulate osteoclastogenesis. Enhanced OC survival is an important component of bone resorption and is increased by cytokines, including M-CSF, RANKL, TNF, IL-1, and vascular endothelial growth factor A (VEGF-A), which prevent OC apoptosis by upregulating Rho family small G-protein Ras/Rac1/Erk and phosphatidylinositol 3-kinase (PI3K)/mammalian target of rapamycin (mTOR)/ribosomal protein S6 kinase beta-1 (RPS6KB1) signaling.135 Withdrawal of these cytokines rapidly induces OC apoptosis due to reduced expression of the antiapoptotic protein, Bcl-2.136 M-CSF prevents OC apoptosis by a number of mechanisms, including: activating MITF, which increases Bcl-2 expression135–137; increasing the proteasomal degradation of Bcl-2-like protein 11 (Bim) by c-Cbl, an ubiquitin ligase; upregulating expression of Bcl-XL,138 which inhibits cleavage of procaspase-9; and inhibiting the activity of caspases 3 and 9, which initiate apoptosis. Deletion of Bcl-XL in OCs resulted in increased OC apoptosis, but surprisingly the mice had increased, rather than decreased, bone resorption. This was associated with increased c-Src activity and expression of vitronectin and fibronectin by OCs, resulting in enhanced integrin-mediated activation of the cells139 and suggesting that Bcl-XL also inhibits OC resorptive activity. Bim is a proapoptotic Bcl-2 family member whose expression is downregulated by IL-3 signaling through the Raf/Erk and/or PI3K/mTOR pathways. Bim−/− mice have decreased OC activity, despite increased OC survival.136 Thus, although enhanced OC survival is in general associated with increased bone resorption and vice versa, these two activities can be uncoupled.