Bone is a dynamic connective tissue that is continuously formed, resorbed, and reformed throughout life. Maintenance of healthy bone quality and quantity depends on the coordinated development and activity of bone-forming osteoblasts and bone-resorbing osteoclasts within local sites of bone remodeling called basic multicellular units (BMUs; Fig. 1). During skeletal remodeling, cells of the osteoblast lineage synthesize and secrete proteins that can initiate or control osteoclast differentiation . Similarly, cells of the osteoclast lineage secrete chemical signals that impact osteoblast differentiation . Intercellular communication between these two lineages therefore plays a key role in coordinating the development and function of key effectors and processes required for healthy bone remodeling. Disorders of bone loss such as osteoporosis are often associated with changes in these chemical signals and a consequent loss of the homeostatic control over balanced bone remodeling .
Within bone, osteoblasts are derived from mesenchymal stem cells (MSCs), while osteoclasts originate from hematopoietic stem cells (HSCs). Osteoblast differentiation proceeds through an initial organic phase where collagen matrix is produced, and a terminal inorganic phase where mineralization by crystals of hydroxyapatite occurs on the collagen scaffold . After completion of matrix secretion and mineralization, some osteoblasts become buried within the matrix and are termed osteocytes, while those that remain on the bone surface become flattened lining cells or undergo apoptosis. Importantly, MSCs are multipotent, and in addition to osteoblasts can give rise to several other distinct cell types including adipocytes, myocytes, chondrocytes, endothelial cells, and vascular smooth muscle cells (Fig. 2) . Among these cell lineages, adipogenic differentiation has particular relevance to bone homeostasis, as cells undergoing osteoblastogenesis can be diverted to become adipocytes, causing a loss of these cells from the osteoblast pool and diminishing overall bone formation potential . An imbalance in the regulation of osteoblast and adipocyte differentiation is commonly seen with conditions such as aging and diabetes mellitus, and can result in fatty bone marrow, impaired osteoblast renewal, and chronic bone loss [4-6]. In addition to the potentially deleterious reduction in osteoblast numbers, adipocytes secrete a spectrum of biologically active signaling molecules that can also impact bone homeostasis. Several of these molecules, including chemerin, resistin, visfatin, leptin, and adiponectin, have been reported to influence the development and function of osteoblasts and osteoclasts [7-12]. Here, we review the impact of bone marrow adipogenesis on the homeostatic coordination between bone formation and resorption, and highlight the potential of targeting bone marrow adipocytes and/or adipocyte-secreted signaling molecules in the therapeutic intervention of bone loss disorders.
Bone Remodeling Cycle
Bone remodeling can occur at random sites, as part of overall calcium and phosphate homeostasis, or at specific sites targeted for repair of structural damage. Regardless, bone remodeling generally follows tightly coupled sequential phases of activation, resorption, reversal, formation, and quiescence (resting), controlled by the spatial and temporal arrangement of cells within the BMU . Activation of bone remodeling begins with detachment of the bone surface lining cells, which, along with the associated capillary network, form a canopy at areas undergoing local remodeling (Fig. 1). This is followed by a resorption phase initiated by osteocyte- and osteoblast-secreted signaling proteins, including macrophage-colony stimulating factor (M-CSF) and receptor activator of nuclear factor kappa B ligand (RANKL). These signals stimulate the recruitment of HSCs, osteoclast differentiation, and bone resorption within the activated BMU . Although osteoblasts and their precursors are generally considered to be the major source of these two cytokines, targeted depletion of these cells in vivo did not affect bone resorption substantially, despite a complete ablation of osteoblasts and bone formation . In contrast, osteocyte-specific deletion of RANKL was found to increase bone mass by reducing osteoclast number and osteoclastic bone resorption, suggesting that osteocytes independently control the number of BMUs, and thereby the rate of bone remodeling, through production of this cytokine . Following RANKL-dependent HSC differentiation, osteoclasts polarize and become tightly bound to the old bone matrix through an interaction between integrin receptors present on the osteoclast membrane and an RGD (arginine, glycine, and asparagine) peptide motif of matrix proteins . This binding leads to the formation of a sealing zone beneath the osteoclasts in close proximity to the matrix, which facilitates proton transport into the microenvironment by a proton pump (H+-ATPase) located at the ruffled border of the osteoclasts. The resulting localized acidification (pH 4–5) dissolves the minerals in the matrix and exposes the organic component of the matrix for further degradation and resorption by osteoblast and osteoclast-secreted proteins. These include matrix metalloproteinases (MMPs), such as MMP-13 that cleave and degrade the unmineralized collagen and cysteine proteases such as cathepsin K which denature the helical structure of collagen contained within the sealing zone .
The intermediate or reversal phase is the least understood, but is generally regarded as a coupling stage between bone resorption and bone formation. During this phase, a poorly characterized population of mononucleated cells called reversal cells colonize the eroded bone surface created by the osteoclasts. Although the precise function of these cells is unknown, it is believed that they contribute to clearing the degraded matrix components from the resorption pits and preparing the eroded bone surface for new bone formation [2, 3]. Several cytokines and soluble factors are released from the degraded matrix during the resorption and reversal phases, including transforming growth factor beta (TGF-β), bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), platelet-derived growth factor, and insulin-like growth factors (IGFs). These paracrine signals drive the bone formation phase by promoting the recruitment of MSCs and preosteoblasts to the BMU, as well as stimulating cell proliferation, osteoblast differentiation, matrix deposition, and mineralization [4, 15, 16]. When bone formation is complete, the bone surface returns to a quiescent or resting state. Thus, bone remodeling is a tightly orchestrated process involving several cell types. Communication among these cells via paracrine signaling molecules is vital to the balanced and efficient function of the BMU.
Osteoblast/osteoclast Differentiation of Stem Cells and Their Developmental Cross-Talk
Bone MSCs are a multipotent and self-renewing cell type within the bone marrow that can give rise to various cell lineages, including osteoblasts and adipocytes. Several extracellular signaling molecules that are either secreted by osteoclasts or released from the matrix during bone resorption, including wingless-type MMTV integration site (Wnt) proteins, BMPs, IGFs, and FGFs, are known to promote osteoblastogenesis of MSCs by stimulating the expression and activity of key osteoblastogenic transcription factors in MSCs, including Dlx5, runx2, and osterix. These transcription factors are important regulators of genes critical for osteoblast differentiation and function including osteocalcin, type I collagen, alkaline phosphatase, osteopontin, and bone sialoprotein (BSP) [4, 17, 18].
Wnt signaling initiates a complex cascade of intracellular events that can be classified into canonical or noncanonical pathways, both of which generally promote osteoblastogenesis. For example, the canonical Wnts, such as Wnt3a, bind to frizzled receptors expressed by MSCs and activate co-receptors, low-density lipoprotein receptor-related proteins (LRP) 5, and LRP6, which inhibits the activation of Adenomatous polyposis coli/axin/glycogen synthase kinase 3 (GSK3) complexes. Inhibition of GSK3 results in stabilization of cytoplasmic β-catenin, which enters the nucleus and activates osteoblast gene transcription in association with other transcription factors such as lymphoid enhancer-binding factor/T-cell factor (reviewed in ref. 4). Noncanonical Wnts, including Wnt1, Wnt5a, and Wnt7b, signal through β-catenin-independent pathways involving calcium-dependent enzymes such as calcium-calmodulin-dependent protein kinase-II (CaMKII) or calcineurin. For example, Wnt5a promotes osteoblast differentiation over adipocyte differentiation by repressing peroxisome proliferator-activated receptor gamma (PPARγ) expression through a CaMKII-mediated signaling cascade . BMP signaling through type IB BMP receptor (BMPR-IB) plays a crucial role in mediating osteoblastogenic differentiation of MSCs by a Dlx5/Runx2-mediated pathway [4, 18]. FGFs and IGFs are potential inducers of both MSC adipogenic and osteoblastogenic differentiation, with the fate determination largely dependent on the presence of other extracellular factors that promote the adipocyte or osteoblast differentiation program (reviewed in ref. 4).
The specialized bone marrow niches containing MSCs also contain multipotent HSCs , which are characterized by the property of self-renewal and the ability to differentiate into all the mature blood cell types (Fig. 2). Prior to undergoing terminal differentiation, HSCs first undergo a process of lineage restriction that gives rise to either a lymphoid or myeloid progenitor that loses the capability to proceed down the alternate lineage . Lineages derived from lymphoid progenitors include T-lymphocytes, B-lymphocytes, and natural killer (NK) cells, while myeloid progenitors give rise to erythrocyte, megakaryocyte, granulocyte, and monocyte–macrophage lineages . Under appropriate conditions, cells of the monocyte–macrophage lineage can further differentiate to bone-resorbing osteoclasts . Importantly, bone HSCs are commonly found in close vicinity to sites of active bone remodeling. This is achieved through expression of calcium-sensing receptors in HSCs that enables their engraftment to microenvironmental niches closest to the bone lining surfaces where a very high calcium concentration prevails due to the activity of osteoclasts and osteoblasts in an active BMU .
Within the microenvironment niche, MSCs and osteoblasts synthesize and secrete paracrine signaling molecules, including growth factors, cytokines, and chemokines that are critical for HSC self-renewal and differentiation into osteoclasts [1, 14, 23]. Foremost among the secreted paracrine signaling molecules is RANKL, a type II homotrimeric transmembrane protein of the tumor necrosis factor (TNF)-family of cytokines that is expressed by osteoblasts and osteocytes as either a soluble or membrane bound protein [14, 24]. RANKL is a pro-osteoclastogenic factor that serves as a ligand activator of RANK, a type I homotrimeric transmembrane protein expressed by osteoclast precursors . RANK stimulation results in the recruitment of TNF-receptor-associated factor 6 (TRAF6), an adaptor protein that triggers a complex signaling cascade involving nuclear factor kappa B (NF-κB), mitogen-activated protein kinases (MAPKs), and c-fos, which ultimately stimulates nuclear factor of activated T cells, cytoplasmic 1 (NFATc1 or NFAT2). Convergent activation of NFAT2 is also mediated through calcium–calmodulin-dependent activation of calcineurin, a phosphatase that dephosphorylates and promotes nuclear translocation of NFAT2. NFAT2 is generally regarded as the master regulator of osteoclast differentiation, as it promotes the expression of several genes required for osteoclast differentiation and function, including tartrate resistant acid phosphatase (TRAP), cathepsin K, osteoclast-associated receptor (OSCAR), calcitonin receptor, and integrin b3 (itgb3) .
The rate of osteoclast differentiation can be altered by osteoblast secretion of osteoprotegerin (OPG), a soluble decoy receptor that binds to RANKL and blocks its interaction with RANK. As such, the ratio of RANKL to OPG within the bone serves as an indicator of the net stimulus for osteoclastogenesis . RANKL and OPG are differentially expressed in osteoblasts depending on the stage of maturation (Fig. 3). RANKL expression is detectable in MSCs, but increases progressively with the early onset of differentiation, declines with maturation, and returns to baseline in mineralizing osteoblasts [26, 27]. In contrast, OPG expression increases progressively throughout osteoblast maturation and mineralization. Extracellular signals that promote MSC osteoblast differentiation, such as Wnts and BMPs, increase OPG expression and decrease RANKL production by osteoblast precursors . Conversely, stimuli that suppress osteoblastogenesis, such as the endogenous antagonist of Wnt signaling, including sclerostin down-regulate OPG and up-regulate RANKL expression . These studies suggest that the switch to osteoclast differentiation during the bone resorption phase is determined by the proportion of immature versus mature osteoblasts within the microenvironmental niche of the BMU. Consistent with this, the ratio of RANKL/OPG expression in mouse calvarial cells is altered depending on the stage of osteoblast differentiation, such that the immature osteoblasts stimulated osteoclast differentiation and bone resorption, while mature osteoblasts prevented osteoclastogenesis and resorption . In addition, several factors that influence bone remodeling are known to mediate their action by targeting the OPG/RANK/RANKL system rather than the direct stimulation or suppression of bone cell differentiation. For example, parathyroid hormone, IL-6, prostaglandin E2, and glucocorticoids act by affecting the production of OPG and RANKL by osteoblasts, and thereby indirectly influence osteoclastogenesis through a paracrine mode of action [30, 31].
In addition to osteoblast-derived factors, a number of signaling molecules produced by osteoclasts and osteoclast precursors are known to participate in the coordination of bone formation and resorption. RANKL stimulation of osteoclasts induces the activity of sphingosine kinase 1 (SPHK1), an enzyme that catalyzes the phosphorylation of sphingosine to sphingosine 1-phosphate (S1P) . S1P is a lipid signaling molecule that acts in a paracrine fashion on preosteoblasts to stimulate osteoblastogenesis, and in an autocrine/paracrine fashion on preosteoclasts to inhibit osteoclastogenesis [32, 33]. Similarly, many other osteoclast-derived molecules, including Wnt10B and BMP6, have also been identified to exhibit pro-osteoblastogenic effects .
Osteoclastogenic cells also express membrane-bound Ephrin ligands, which interact with transmembrane receptor tyrosine kinase Eph receptors on osteoblast progenitor cells through direct osteoclast-osteoblast contact (Fig. 1) . This cell–cell interaction is essential for activating a bidirectional signaling cascade that involves receptor-mediated forward signaling in osteoblasts, and ligand-mediated reverse signaling in osteoclasts . The forward signal in osteoblasts stimulates bone formation while the reverse signal in osteoclasts mediates the negative feedback inhibition of osteoclastogenesis that terminates osteoclast differentiation and bone resorption at the end of the resorptive phase .
In addition to these stimuli for bone formation produced by osteoclasts, osteoclasts also produce sclerostin and Semaphorin 4D (Sema4D), which inhibit osteoblast differentiation (Fig. 3) [35, 36]. Sclerostin is a protein encoded by the SOST gene, and is known to suppress osteoblast differentiation by inhibiting the pro-osteoblastogenic actions of BMP6 and BMP7 on MSCs and pre-osteoblasts . Semaphorin 4D (Sema4D), also known as cluster of differentiation 100, is an osteoclast protein that exists in both a soluble and transmembrane form. The anti-osteoblastogenic effects of this protein are mediated through interaction with Plexin-B1, a receptor expressed by osteoblast precursors. The Sema4D/Plexin-B1 interaction has been shown to block bone formation through activation of Ras homolog family member A (RhoA)-mediated suppression of IGF-1 signaling .
Factors released from the bone matrix as a consequence of osteoclast activity during the resorptive phase also play a critical role in signal coupling during the reversal phase . For example, matrix-derived TGF-β can act on osteoblasts to suppress RANKL secretion and thereby reduce osteoclastogenesis in an indirect fashion . TGF-β also acts directly on osteoclasts to promote the secretion of cytokines, such as leukemia inhibitory factor (LIF), and the chemokine CXCL16, both of which contribute to the recruitment of MSCs and preosteoblasts to resorption pits . In addition, TGF-β also promotes osteoblastogenic differentiation of MSCs by stimulating the secretion of Wnt10b from osteoclasts .
The Impact of Adipocyte Differentiation on Osteoblast and Osteoclast Differentiation
Among the differentiation fates of bone marrow MSCs (Fig. 2), the adipogenic and osteoblastogenic programs are the most closely related, and play an important role in regulating bone mass homeostasis (reviewed in ref. 4). Several extracellular signaling proteins have overlapping roles in MSC adipogenesis and osteoblastogenesis by modulating the expression and/or activity of adipocyte-specific (e.g., PPARγ) or osteoblast-specific (e.g., Runx2 and osterix) transcription factors. Some of these factors play opposing roles with respect to lineage determination, while others function in complementary fashion. For example, both canonical and noncanonical Wnt signaling have been shown to promote osteoblastogenesis and at the same time actively repress adipogenesis. Conversely, several BMPs including BMP-2, BMP-4, BMP-6, BMP-7, and BMP-9 can stimulate both the adipogenic and osteoblastogenic pathways depending on the receptor activated. For example, activation of the type IA BMP receptor (BMPR-IA) on MSCs induces PPARγ expression and promotes adipocyte differentiation, while activation of the type IB BMP receptor (BMPR-IB) induces Runx2 expression and favors osteoblastogenesis (reviewed in ref. 4). Insulin and IGF1, which bind with the insulin receptor or IGF receptor, respectively, activate a common cascade of signaling events involving insulin receptor substrates 1 and 2, and are essential for both adipogenic and osteoblastogenic differentiation of MSCs. Likewise, the FGFs are also well established to promote both adipogenic and osteoblastogenic differentiation of MSCs .
Although a number of factors have overlapping roles in promoting osteoblastogenic/adipogenic differentiation of MSCs, in general these lineages are mutually exclusive and entail active suppression of intracellular pathways conducive to the alternate lineage. As such, a shift in MSC differentiation towards the adipogenic program is believed to be a major contributor to the decline of osteoblastogenesis and bone formation potential that is characteristic of bone-loss disorders such as osteoporosis (reviewed in ref. 4). A similar shift of MSC differentiation to the adipogenic lineage has also been linked to the reduced bone formation associated with diabetes mellitus and chronic therapeutic use of steroids and thiazolidinediones, both of which are known to promote MSC adipogenesis [4, 5]. Consistently, when the endogenous expression of the key driver of adipocyte differentiation, PPARγ was reduced or deleted in MSCs, it resulted in abrogation of adipogenesis along with a reciprocal increase in osteoblastogenesis .
In addition to changes in the levels of endogenous signaling molecules and exposure to therapeutic drugs, alterations in the collagen composition of the bone matrix are also believed to have an influence on MSC differentiation [41, 42]. For example, it has been observed that culturing MSCs on denatured collagen is not permissive for osteoblastogenesis, but promotes adipocyte differentiation . Thus, a shift in MSC differentiation to favor adipogenesis may be linked to deficiencies or changes in the structural conformation of collagen that are known to occur with disorders such as osteoporosis [41, 42].
In addition to driving MSC adipogenesis, PPARγ has been reported to bind cognate response elements within the c-fos promoter in HSCs, and thereby promote osteoclastogenesis through the induction of c-fos . Consistent with these findings, conditional selective deletion of osteoclast PPARγ increased bone volume coincident with reduced bone resorption in mice . This type of osteoblast-independent induction of osteoclastogenesis may lead to abnormal and excessive resorption at bone remodeling sites, because the stimulus for osteoclastogenesis is derived from a cell type that is not associated with the formative phase, but is rather associated with suppressed bone formation. In this fashion, the normal signaling and cross-talk between bone-forming osteoblasts and bone-resorbing osteoclasts are decoupled by pro-adipogenic factors, leading to a disruption of normal bone homeostasis. The tight regulation between osteoclastogenesis and osteoblastogenesis may therefore be lost due to factors that drive MSC adipogenesis and suppress osteoblastogenesis as well as direct activation of HSC osteoclastogenesis in a PPARγ-dependent fashion (Fig. 4).
Role of Adipocyte-Derived Soluble Factors in the Cross-Talk Between Osteoblasts and Osteoclasts
In addition to the impact of adipocyte differentiation, the release of proinflammatory cytokines and other signaling proteins by mature adipocytes can also affect bone homeostasis. For example, experimental evidence indicates that leptin acts directly on precursor cells to promote osteoblast differentiation and proliferation, inhibit adipogenesis, and suppress osteoclastogenesis [9, 44] (Fig. 4). Moreover, changes in leptin secretion associated with increased numbers of adipocytes in the bone marrow of osteoporotic subjects have been shown to disrupt the normal paracrine coordination between osteoblasts and osteoclasts [44, 45]. However, the overall impact of leptin on bone remodeling remains somewhat unclear, due to conflicting experimental results, and the observation that white adipose tissue-derived leptin in the CNS has been linked to both increased and reduced bone mass . Another adipocyte-derived signaling molecule, adiponectin, has been reported to have a direct negative effect on MSC osteoblastogenesis . Consistent with this finding, culture of osteoblasts with adipocyte-conditioned media was reported to decrease expression of the osteoblastogenic transcription factor runx2, an effect that was abrogated by knockdown of adiponectin receptor 1 . Although these studies provide evidence that adiponectin has a negative impact on osteoblastogenesis, it is important to note that others have reported contradictory results suggesting that adiponectin can promote osteoblastogenesis . Conflicting results have also been reported with respect to HSC osteoclastogenesis. For example, while some studies observed that adiponectin can directly act on HSCs and suppress their differentiation into mature osteoclasts , adiponectin was also reported to indirectly promote osteoclastogenesis by inducing RANKL expression and repressing OPG expression in osteoblasts . Similar to leptin, the situation is further complicated by conflicting reports regarding the action of white adipose tissue-derived adiponectin. For example, some studies have reported an inverse relationship between serum adiponectin levels and bone mineral density (BMD) in aging and postmenopausal women . However, serum adiponectin levels were also found to be positively correlated with BMD in patients with type 2 diabetes mellitus, suggesting a protective role of adiponectin against bone loss in these patients .
We have previously reported that chemerin, a novel adipokine [53, 54], was expressed and secreted at the high levels by MSC-derived adipocytes, with other bone cell types, including osteoblasts, MSCs, and HSCs secreting less, but functionally meaningful amounts . Functionally, the interaction of chemerin with its receptor, chemokine-like receptor 1 (CMKLR1), was found to promote MSC adipogenesis [12, 55]. Moreover, abrogation or knockdown of CMKLR1 expression increased the expression of osteoblast markers, including osterix, alkaline phosphatase, and type I collagen, even in the context of persistent adipogenic stimuli . A direct interaction between PPARγ and chemerin was found to play a central role in the cross-talk between adipogenic and osteoblastogenic differentiation of MSCs [12, 55]. Among the bone cell types examined, CMKLR1 expression was highest in HSCs, and neutralizing extracellular chemerin in HSC culture media using a blocking antibody inhibited osteoclastogenic differentiation . Taken together, this experimental evidence suggests that a blockade of chemerin/CMKLR1 signaling can improve osteoblastogenesis by suppressing the adipogenic pathway in MSCs, while simultaneously reducing osteoclast differentiation (Fig. 4). However, similar to other adipokines, the overall relevance of chemerin to bone homeostasis and pathophysiology is presently unclear. For example, MSCs derived from CMKLR1 knockout mice do not exhibit the dramatic loss of adipogenic potential that is observed in tissue culture with acute shRNA-induced knockdown (Muruganandan, Sinal et al. unpublished data). Further studies using inducible CMKLR1 knockout mouse models will be required to investigate the apparent discordance between the effects of acute versus chronic (i.e., conventional knockout mouse) loss of chemerin/CMKLR1 signaling on bone cell differentiation.
Other adipokines including omentin-1, resistin, and visfatin have also been shown to modulate osteoblast and/or osteoclast differentiation [8, 56, 57]. Omentin-1 promoted the proliferation of osteoblasts  and indirectly suppressed osteoclast differentiation by inducing OPG production, and suppressing RANKL production, by osteoblasts in a co-culture system . Consistent with these in vitro effects, adenoviral gene delivery of omentin-1 in ovariectomized mice ameliorated the bone loss caused by estrogen deficiency . Resistin has been reported to promote osteoblast proliferation . However, the overall impact on bone remodeling is unclear, as resistin has also been reported to promote osteoclastogenesis via activation of NF-κB signaling . Visfatin was observed to promote osteoblast proliferation and collagen secretion in an insulin receptor-mediated fashion , as well as block osteoclast differentiation by suppressing RANK, NFATc1, and cathepsin K expression . However, in patients with inflammatory bowel disease, serum visfatin levels were found to be negatively correlated with BMD, suggesting the need for additional studies to understand the in vivo role of visfatin on bone metabolism .
In addition to these adipokines, adipocytes also express RANKL and OPG, and when co-cultured with HSCs, support osteoclast differentiation. Moreover, adipocyte RANKL expression is induced markedly by dexamethasone, suggesting a potential mechanistic link between the chronic use of synthetic steroids and bone loss [59, 60].
Therapeutic Targeting of Osteoblast–Osteoclast Communication
Highly orchestrated chemical communication between osteoblasts and osteoclasts is critical for maintaining the balance between bone formation and bone resorption. Disruption of this balance due to aging, or pathological conditions such as diabetes mellitus, postmenopausal estrogen deficiency, or long-term steroidal therapies, can result in increased bone resorption and decreased bone formation. The current first line of treatment for this type of bone loss is the bisphosphonate class of drugs (e.g., zoledronate), which stabilize or improve bone mass by impairing osteoclast activity, and inducing osteoclast apoptosis. However, direct targeting of osteoclasts disrupts the normal paracrine communication that occurs between bone cell types, due to a loss of osteoclast-derived signaling molecules that are necessary for osteoblastogenesis. As such, the imbalance in cell–cell signaling caused by antiresorptive drugs can ultimately suppress the formation of both osteoblasts and osteoclasts, resulting in a suppressed state of bone remodeling and a poor quality bone [37, 61]. In the presence of antiresorptive drugs, osteoblast-mediated bone formation cannot be fully recovered even by co-treatment with efficacious anabolics, such as parathyroid hormone . Thus, treatments that restore the normal paracrine signaling between bone cell types, rather than targeting the development and/or action of a single-cell type, are desirable.
Osteoblast-derived RANKL and osteoclast-derived Sema4D are key factors that act in a highly coordinated fashion to control the balance between bone formation and bone resorption [1, 36]. In addition to stimulating osteoclast differentiation, RANKL also promotes osteoclast secretion of Sema4D, which functions in a negative feedback loop to suppress osteoblastogenesis. Consistent with this, stimulation of osteoblast bone formation in ovariectomized mice by an anti-Sema4D antibody was shown to increase bone formation without affecting osteoclast bone resorption , suggesting that osteoblast–osteoclast communication is maintained by the anti-Sema4D antibody. Additionally, coupling factors that are released as a result of osteoclast activity during matrix resorption including TGF-β stimulate osteoblastogenesis and bone formation [15, 16]. In the context of cellular communications mediated by osteoblasts, the primary stimulus for osteoclastogenesis, RANKL, and the blocker of this process, OPG, are both derived from these cells depending on the stage of differentiation [26, 27]. Experimental and clinical studies have reported that estrogen-deficient females or androgen-deficient males have increased RANKL in both peripheral circulation and bone marrow plasma, resulting in increased bone resorption, bone loss, and osteoporosis [62, 63]. Under these conditions, RANKL neutralization has been found to increase bone mass, and an anti-human RANKL neutralizing antibody (denosumab) is currently being used clinically for the treatment of osteoporosis . In contrast to antiresorptives such as alendronate , RANKL neutralization preserves the bone anabolic actions of PTH , suggesting that the neutralization antibody maintains osteoblast–osteoclast communication.
Induction of bone marrow adipogenesis and a marked accumulation of bone marrow adipocytes have long been known to occur with senile osteoporosis  and diabetic osteopenia . In agreement with this, regulators of MSC differentiation within the bone marrow have been found to be shifted towards an induction of adipogenesis and a suppression of osteoblastogenesis in patients with osteoporosis . MSCs from osteoporotic patients produce lower TGF-β and type I collagen, which favors adipogenic differentiation, and lower levels of pro-osteoblastogenic signals, including osterix and LRP6, compared to healthy individuals . Harnessing the ability of adipogenic cells to trans-differentiate to the osteoblastogenic lineage  may provide another novel dimension to existing bone loss treatments. In addition to inducing bone formation through increased osteoblastogenesis, this approach could also improve the osteoblast-derived RANKL/OPG cytokine production that can further activate normal bone remodeling. Therefore, the conversion of adipocytes into osteoblasts may enhance the cross-talk between osteoblasts and osteoclasts besides alleviating the disruption induced by the effects of adipocyte-derived soluble factors. However, to fully understand the role of adipogenesis on osteoblast–osteoclast communication, further studies are required to investigate the influence of bone marrow adipocytes and/or adipogenic signals on the secretion of signaling molecules from pre- and mature osteoclasts.
In this review, we emphasize bone marrow adipocyte as a potential regulatory system that influences the interactions between osteoblasts and osteoclasts during bone remodeling. The major impact of bone marrow adipogenesis on bone remodeling arises from the fact that both osteoblasts and adipocytes derive from MSCs, which exhibit mutually exclusive differentiation into one lineage or the other. Beginning from this reciprocal association to bone formation, the increase in the accumulation of fat cells and/or their secreted molecules in bone marrow has been shown to affect several aspects of bone remodeling, including the development of osteoclasts, synthesis of the catabolic factor, RANKL and its antagonist, OPG, and the tight coupling of cellular signaling between osteoblasts and osteoclasts. The remarkable activation of osteoclastogenesis by the adipogenic transcription factor PPARγ in HSCs is a good example of the functional relevance of adipogenic targets even in non-adipogenic bone cell types. Therefore, the abrogation of signaling through this adipogenic target may induce a beneficial shift in bone remodeling by both stimulating osteoblastogenesis rather than adipogenesis, and suppressing bone resorption through a reduction in PPARγ-directed osteoclastogenesis. However, there is a lack of knowledge about the impact of adipogenic factors and/or adipocyte secreted molecules on the release of osteoclast-derived signaling molecules that deserves further studies to establish the therapeutic value of bone marrow adipocytes as targets for improving bone remodeling.