Regulation of Skeletal Homeostasis

Abstract

Landmark advances in skeletal biology have arisen mainly from the identification of disease-causing mutations and the advent of rapid and selective gene-targeting technologies to phenocopy human disease in mice. Here, we discuss work on newly identified mechanisms controlling the remodeling of bone, communication of bone cells with cells of other lineages, and crosstalk between bone and vital organs as these relate to the therapeutic targeting of the skeleton.

Essential Points

• There is vibrant molecular crosstalk among osteoblasts, osteocytes, and osteoclasts during bone remodeling

• Three key pathways—Notch, bone morphogenetic protein (BMP), and Wingless-intregration-1 (WNT)—regulate osteoblastic bone formation, all of which are controlled epigenetically

• Neural surveillance of the skeleton comprises a neural arm driven mainly by sympathetic discharges and a neuroendocrine arm wherein pituitary hormones act directly to regulate bone remodeling

• Molecular interactions between bone and immune cells within bone marrow generate cytokines that influence bone remodeling, whereas osteoblasts modulate both erythropoiesis and blood vessel formation

• There has been new, valuable insight into interactions between bone and distantly located vital organs, including muscle, adipose tissue, pancreas, kidney, and immune organs

• Unraveling of the molecular circuitry through which bone communicates both locally and globally exposes new targets for the therapy of multiple diseases, including osteoporosis, obesity, and sarcopenia

Bone is a highly dynamic and purposefully organized composite consisting of a protein matrix, mainly comprising type 1 collagen, mineral, and cells of multiple lineages. In 1760, John Hunter, an orthopedic surgeon, proposed that the integrity of the vertebrate skeleton is maintained through the process of remodeling a precise spatiotemporal sequence in which packets of old bone are removed and are replaced with new bone (1). Kölliker (2) later described the process of bone resorption by multinucleated cells—osteoclasts. Following resorption, osteoblasts synthesize new bone and die by apoptosis, become bone lining cells, or get buried within the newly formed bone matrix and morph into osteocytes to provide mechanical cues for spatial synchronization (3–5). Dysregulated bone remodeling not only causes osteoporosis (6), a crippling public health hazard, but also other genetic and acquired skeletal disorders. Decade-long efforts in cataloguing gene mutations and the use of genetically modified mouse models to mimic human bone disease have led to a surge in our understanding of the genetic and epigenetic basis of the bone-remodeling process and provide new molecular targets potentially amenable to therapeutic targeting.

We have long known that multiple local cytokine circuits and systemic hormones, such as parathyroid hormone (PTH), regulate the delicate balance between bone resorption and bone formation. More recent studies have, however, positioned bone as a vital organ, with both paracrine and endocrine functions. Bone cells, namely osteoblasts, osteoclasts, and osteocytes, not only communicate among themselves but also talk to cells within the bone marrow, such as T and B cells, macrophages, adipocytes, and hematopoietic progenitors. Equally critical has been the discovery of a central surveillance exerted via the sympathetic nervous system (SNS), as well as by pituitary hormones (7). These studies have also highlighted certain endocrine functions of the skeleton in regulating muscle mass, adiposity, and global energy balance. The medical implications of these new discoveries are intriguing. With the consideration that osteoporosis can coexist with sarcopenia or postmenopausal obesity, the question arises as to whether it is possible in the future to treat two diseases with a single agent.

Cellular Crosstalk and Its Regulation During Bone Remodeling

Bone cell communication

New insights have been gleaned into mechanisms through which bone resorption and bone formation are coupled (Fig. 1) (8). TGF-β1 (TGFB1), released from bone matrix during resorption, is the primary inducer of the migration of bone marrow stromal cells to the site of bone resorption (9). The TGFB1 gradient, in fact, determines spatial localization of stromal cells, and disruption of this gradient in mice or in Camurati-Engelmann disease harboring an activating TGFB1 mutation results in poorly organized stromal cell recruitment, dysplastic bones, and fracture (9) (Table 1). Once stromal cell progenitors arrive into the resorption hemivacuole, insulin-like growth factor (IGF) 1, released from matrix degradation, initiates their differentiation into mature osteoblasts via the mechanistic target of rapamycin pathway (10) (Fig. 1). Mice with global or osteoblast-targeted genetic deletions of Igf1 receptor or Igf1 have an osteopenic skeleton, testifying to a fundamental role of the IGF regulatory system in skeletal remodeling (10). In addition, the extracellular matrix itself regulates the conversion of stromal cells to osteoblasts through multiple well-known interactions with type 1 collagen and importantly, noncollagenous proteins, including fibronectin, vitronectin, laminin, and the more recently described epidermal growth factor (EGF)–like repeats and discoidin domains 3 (11–14). Interestingly, the elasticity of the surrounding matrix can also direct stromal cell specification (15).

Figure 1.

Molecular coupling of bone resorption with bone formation. The resorption of bone matrix by proteolytic enzymes secreted by the osteoclast, such as cathepsin K (CTSK), liberates growth factors, notably TGF-β (Tgfb) and insulin-like growth factor 1 (Igf1), from bone. Hydroxyapatite dissolution by secreted acid releases Ca²⁺ locally, which triggers intracellular Ca²⁺ release to inhibit further enzyme secretion and bone resorption. Osteoclast precursors (or preosteoclasts), derived from hematopoietic stem cells (HSCs), become differentiated into mature resorbing cells through the actions of macrophage colony–stimulating factor (MCSF) and receptor activator for nuclear factor (NF)-κB ligand (Rankl). The latter acts on the stimulatory receptor Rank to trigger TNF receptor (TNFR)–associated factor 6 (Traf6)–mediated Nfkb and NF for activated T cells 2 (Nfat2) activation. Rankl also acts on the newly discovered inhibitory GPCR, leucine-rich repeat-containing GPCR 4 (Lgr4), and Nfat2 increases Lgr4 expression to ensure feedback control of Rankl action. In contrast, osteoblasts are derived from mesenchymal stem cells (MSCs) and are recruited to the site of bone resorption by released Tgfb. MSCs differentiate into mature, mineralizing osteoblasts under the influence of Igf1 and osteoblast-derived bone morphogenetic proteins (Bmp2/4), notch, and wingless-ints (Wnt). Osteoblasts also secrete semaphorin 3a (SEMA3A), which further stimulates osteoblast precursor proliferation and differentiation through Wnt but repels osteoclast precursors. Shown also are BMP, notch, and canonical Wnt signaling pathways in osteoblast-lineage cells. Also shown is epigenetic regulation by miRNAs, where known. ALK, activin receptor–like kinase; APC, adenomatous polyposis coli; ATF4, activating transcription factor 4; CBP/p300, CREB-binding protein/E1A binding protein p300; CK1, casein kinase 1; CN, calcineurin; Co-R, co-repressor; CSL, CBF1/Su(H)/Lag-1 transcription factor complex; DKK1, Dickkopf 1; DLL1-3, delta-like canonical Notch ligands 1-3; FZD, Frizzled; GSK3β, glycogen synthase kinase-3β; HDAC1, histone deacetylase 1; HES1-5, hairy and enhancer of split 1-5; LRP5/6, LDL receptor–related protein 5/6; MAML, Mastermind-like 1; mi, miRNA; NICD, NOTCH intracellular domain; NOG, noggin; OSX, osterix; PDGF-BB, platelet-derived growth factor BB; RUNX2; runt-related transcription factor 2; S1P, sphingosine-1-phosphate; SEMA3A, semaphorin 3a; SHN2, schnurri 2; SHN3, schnurri 3; SMAD, mothers against decapentaplegic homolog; SFRP2, secreted Frizzled-related protein 2; SOST, sclerostin; TCF/LEF1, transcription factor 7/lymphoid enhancer–binding factor 1; TNFα, tumor necrosis factor-α.

Table 1.

Mutations in Rare Genes With Large Effects on Bone Mass in the Bone Morphogenetic Protein, Notch, and Wnt Signaling Pathways Reveal New Targets for Bone Protection

	BMP	Notch	Wnt
Mutations	Fibrodysplasia ossificans progressiva	Hajdu-Cheney syndrome	Osteoporosis-pseudoglioma syndrome
	ACVR1R206H	NOTCH2 (14-point mutations in exon 34)	LRP5 (multiple point mutations)
	Osteogenesis imperfecta	Spondylocostal dysostosis	High bone mass
	BMP1F249L (protease domain)	DLL3 (14 mutations)	LRP5 (multiple point mutations)
	BMP1G12R (signal peptide)		Van Buchem disease: sclerosteosis
	Osteoarthritis		SOST (52 kb gene deletion)
	BMP5D6S1276
	BMP145′UTR
	Myhre syndrome
	SMAD4I500V
	SMAD4R496C
Drug targets	BMP2/4: fracture healing	Jagged 1: fracture healing	Anti-SOST antibody: osteoporosis
	BMP7: osteoarthritis		Anti-DKK1 antibody: multiple myeloma

Abbreviations: ACVR1, activin A receptor type 1; DLL3, Delta-like protein 3.

Once mature, osteoblasts not only lay down collagen and mineral but also secrete numerous molecules, one of which is semaphorin 3a (SEMA3A), a known axonal guidance molecule. SEMA3A has a dual function in bone remodeling: it augments osteoblastic activity through Wingless-intregration-1 (WNT) signals, but repels osteoclast precursors (16, 17). Recombinant SEMA3A thus uncouples remodeling in ovariectomized mice by stimulating new bone formation and inhibiting bone resorption with a net increase in bone mass (16). Another molecule, produced by osteoclasts, sphingosine-1-phosphate, also controls both bone formation and osteoclast precursor migration (18). Newly discovered “clastokines,” such as sphingosine-1-phosphate, provide evidence of bidirectional signaling between osteoclasts and osteoblasts.

Osteoblasts build new bone and regulate osteoclastogenesis. Seminal early in vitro studies provided proof that the effect of resorption stimulators, such as PTH or 1,25-dihydroxyvitamin D₃, was not direct. Instead, their actions required the osteoblast, which produced a soluble molecule that stimulated bone resorption (19). This molecule later turned out to be receptor activator for nuclear factor (NF)-κB ligand [RANKL; or TNF receptor (TNFR) superfamily 11 (TNFRSF11)], a TNF-α family member, which acts on an osteoclast-resident TNFRSF member, receptor activator for nuclear factor (NF)-κB [RANK (TNFRSF11A)], to promote the full differentiation of the osteoclast (20). Osteocytes embedded within bone matrix can also generate RANKL, which induces osteoclastic activity, as well as resorption in and around the osteocyte lacunae (21, 22) (Fig. 1). The deletion of RANKL selectively in osteocytes attenuates osteoclast numbers, which may explain how mechanical stimuli transmitted through osteocytes may direct site-specific osteoclastic resorption (21, 22).

The concept of osteoblast/osteocyte–osteoclast crosstalk led to the discovery of the anti-RANKL antibody, denosumab, currently in use for the treatment of osteoporosis and skeletal metastasis. It also led to studies on the mechanisms that would prevent excessive RANKL-mediated bone removal, for example, in people with activating TNFRSF11A mutations or after menopause when RANKL expression is high (23, 24). Almost every proresorptive hormone or cytokine stimulates RANKL expression, whereas RANKL inhibitors are far fewer. Osteoprotegerin (OPG; or TNFRSF11B), produced by osteoblasts and osteocytes, was first identified as a decoy molecule that would bind to and inhibit RANKL action (25). The WNT inhibitor secreted Frizzled-related protein 1 does the same, but the expression of both molecules is somewhat ubiquitous (26). Very recently, leucine-rich repeat-containing G protein–coupled receptor (GPCR) 4 (LGR4) has been identified as a partner for RANKL, which can downregulate osteoclast differentiation and bone resorption (27) (Fig. 1). The binding of RANKL to LGR4 triggers coupling with Gα_q that causes cytosolic Ca²⁺ release to inhibit NF for activated T cells (NFAT) 2 activation and osteoclastogenesis (27). NFAT2, together with activated NF-κB, in turn, induces LGR4 transcription (27) to create a negative-feedback loop to restrict RANKL action (28). There is therefore the potential for use of an extracellular domain of LGR4 to bind excessive RANKL as a means of preventing excessive bone resorption (27, 28).

Interacting genetic pathways regulating bone remodeling

Bone remodeling is regulated by several distinct but overlapping genetic programs, among which BMPs, NOTCH, and WNTs are associated with genetic bone diseases (Fig. 1). Several BMPs and the type 1 and 2 BMP receptors are expressed in osteoblasts. Noggin (NOG), a 90-kDa protein secreted by osteoblasts, binds to and competitively blocks the action of certain BMPs, namely BMP2 and BMP4. Mice, in which Nog is overexpressed selectively in osteoblasts, display accelerated age-related osteopenia as a result of reduced bone formation, implying that the BMP pathway is functional in adult bone formation (29). Osteoinduction by BMPs has long been used to accelerate fracture healing. However, this pathway cannot be exploited for osteoporosis treatment because of the potential for ectopic ossification, particularly as an activating mutation in one of the type 1 BMP receptors, activin A receptor type 1, causes a rare but crippling disease, fibrodysplasia ossificans progressiva (30) (Table 1).

NOTCH is another transmembrane protein from which its active intracellular domain can be cleaved by γ-secretases, such as presenilin 1/2 (Fig. 1). Early deletion of Notch1, using a paired related homeobox 1–Cre recombinase promoter, results in osteosclerotic bones in lethal embryos (31), whereas late deletion using a 2.3/3.6 kb fragment of the collagen, type 1, alpha 1 proximal promoter (Col2.3)–Cre recombinase line causes osteoporosis (32). In reverse, osteoporosis and osteosclerosis arise, respectively, when NOTCH intracellular domain overexpression is driven by an early (Col3.6) or late (Col2.3 or dentin matrix protein 1) promoter (33, 34). It is therefore not unexpected that human mutations that activate NOTCH2 during early embryogenesis are associated with Hajdu-Cheney syndrome, a devastating disease of bone loss and fractures (35, 36) (Table 1). In contrast, late activation of Notch signaling in adult mouse osteocytes triggers a profound, therapeutically relevant skeletal anabolic response that rescues both age- and ovariectomy-induced bone loss and promotes bone healing (37). With these new insights, the NOTCH pathway may become a future target, particularly as its ligand Jagged 1 stimulates bone formation (37).

The identification of rare WNT gene variants with large effect size, notably low-density lipoprotein receptor–related protein 5 (LRP5) and sclerostin (SOST) mutations causing osteoporosis-pseudoglioma syndrome and sclerosteosis/Van Buchem disease, respectively, together with a more recent genome-wide association study (GWAS) and mouse genetic study, has established a key function for WNT signaling in skeletal homeostasis [e.g., Rivadeneira et al. (38)] (Table 1). The activation of Wnt and its downstream transcriptional regulator β-catenin in adult mice increases bone mass by enhancing stem cell renewal, preosteoblast proliferation and differentiation, and osteoblast and osteocyte survival. These properties, together with β-catenin–mediated osteoclast inhibition (39), underscore the clinical use of a new antibody, romosozumab, which blocks the osteocytic WNT inhibitor SOST. Romosozumab increases bone formation, inhibits bone resorption, and reduces vertebral fracture risk in women with postmenopausal osteoporosis (40). In contrast, the aberrant production of another WNT inhibitor Dickkopf 1 (DKK1) from myeloma cells underlies the reduced bone formation in this disease and is a target of a new monoclonal antibody (41). Both romosozumab and antibodies to DKK1 are potential therapeutic agents to be trialed in myeloma bone disease.

Epigenetic regulation by miRNAs

The BMP, NOTCH, and WNT pathways, as well as their target master transcription factors, runt-related transcription factor 2 (Runx2), osterix, activating transcription factor 4, and schnurri 2 (or HIV type I enhancer–binding protein 2), are also regulated epigenetically, mainly by noncoding microRNAs (miRNAs; Fig. 1). Conditional deletion of the miRNA processing endoribonuclease Dicer in osteoblast lineage cells shows that miRNAs are required for skeletogenesis, as well as postnatal bone growth, modeling, and remodeling (42). Specific functions have also been attributed to different miRNAs. For example, miR-2861 enhances Runx2 expression to stimulate osteoblastogenesis, whereas the miRNA cluster, comprising miR-23a, miR-30c, miR-34c, miR-133a, miR-135a, miR-205, and miR-217, does the opposite (43). Likewise, miR-29a attenuates the expression of the three WNT inhibitors—SOST, DKK1, and secreted Frizzled-related protein 2—and by stimulating bone formation, protects against glucocorticoid-induced bone loss (44, 45). Likewise, miR-874-3p, highly expressed during the phase of accelerated bone formation after weaning, induces osteoblast differentiation via histone deacetylase 1 by enhancing Runx2 transcription (46). Whereas there are fewer reports on the epigenetic control of osteoclasts (47), miR-16 and miR-378 inhibit osteoclastic bone destruction from tumor invasion.

Central Surveillance of Skeletal Homeostasis

Neural regulation

The discovery of the brain–bone connection has been validated rigorously by mouse genetic and human investigations (Fig. 2). Early studies showed that intracerebroventicular leptin reduced bone mass by increasing the release of noradrenaline to activate the osteoblast β-adrenergic receptor 2 (ADRB2) (48). A connection with bone resorption was established through findings that leptin deficiency, by reducing ADRB2 signaling, activating transcription factor 4, and cocaine and amphetamine-regulated transcript, can suppress Rankl-induced osteoclastogenesis (49). Mechanisms for central leptin effects on bone mass are distinct from those on appetite (50) (Fig. 2). Leptin acts on distinct sets of brainstem neurons through serotonin: on ventromedial hypothalamic 5-hydroxytryptamine receptor (HTR) 2c to affect bone mass and arcuate nuclear HTR1A and HTR2Bs to modulate appetite, respectively (51). At the level of sympathetic ganglia, enhanced sympathetic tone is initiated through the transcription factor Foxhead box protein O1 via its control of dopamine β-hydroxylase expression (52). Downstream of ADRB2, the effect of central leptin on osteoblast proliferation uses molecular clock genes, notably period and cryptochrome (53), similarly to the circadian rhythmicity noted with hematopoietic stem cell (HSC) egress (54).

Figure 2.

Neural and neuroendocrine control of bone remodeling. Bone formation is regulated by signals from the SNS, controlled centrally by leptinergic neurons via serotonin action on Htr2c in the ventromedial hypothalamus (VMH). Enhanced sympathetic tone is initiated through Foxhead box protein O1 (FoxO1) via its control of dopamine β-hydroxylase (Dbh) expression. Secretion of norepinephrine at the nerve terminals innervating osteoblasts is regulated by the endocannabinoid receptor, Cb1, and its ligand 2-arachidonoylglycerol (2AG). Norepinephrine activates the Adrb2 receptor to activate activating transcription factor 4 (ATF4) signaling, ultimately modulating expression of circadian clock genes, notably period (Per) and cryptochrome (Cry), to inhibit bone formation. Adrb2 also signals via the neuropeptide cocaine and amphetamin-regulated transcript (Cart) to modulate the Rankl section and, thus, osteoclast activation. Signals arising from IL-1 (Il1)-, cannabinoid receptor 2 (Cb2)–, and neuropeptide Y (Npy)–expressing neurons regulate bone remodeling, as do peripheral nicotinic acetylcholine (ACH) receptors (nAchr). Neurohypophyseal and pituitary hormones, namely growth hormone (Gh), adrenocorticotrophic hormone (Acth), follicle-stimulating hormone (Fsh), thyroid-stimulating hormone (Tsh), prolactin (Prl), oxytocin (Oxt), and arginine vasopressin (Avp), also regulate both osteoclasts and osteoblasts directly through GPCRs. Certain ligands—namely Oxt, TSHβ variant (TSHβv), and Acth—are also produced in bone marrow by macrophages and/or osteoblasts. ADRB2, β-adrenergic receptor 2; CART, cocaine and amphetamine–regulated transcript; HTR, 5-hydroxytryptamine receptor; Lep, leptin; MAPK, mitogen-activated protein kinase; Ser, serotonin; VEGF, vascular endothelial growth factor.

It is not surprising that with burgeoning new information, the neural surveillance of bone mass has turned out to be more complex than originally thought and includes other hypothalamic and peripheral relay networks, including the IL-1, melanocortin-4, neuropeptide Y, cannabinoid, and neuromedin U receptor systems (55–57). For example, cannabinoid receptor 2–deficient (Cb2^−/− or Cnr2^−/−) mice display age-related trabecular bone loss, phenocopying the osteoporosis associated with CNR2 gene polymorphisms [e.g., Ofek et al. (58)]. Consistent with this, a cannabinoid receptor (CB) 2 agonist displays anabolic activity in vivo (59). In contrast, prejunctional CB1 interacts with high concentrations of the endocannabinoid 2–arachidonoylglycerol, produced to regulate noradrenaline release in bone (60). Originating from the spinal cord, parasympathetic nerves release acetylcholine (ACH), which activates nicotinic but not muscarinic ACH receptors to inhibit bone resorption and increase bone mass (57). This latter pathway is believed to be regulated centrally by IL-1 (57).

Clinically, compelling evidence shows that ADRB antagonists, such as propranolol, increase bone mass and reduce fracture risk in some people (61, 62), supporting the neural–bone connection. Recent studies show that sympathetic nerves that richly endow inner vestibular cells of the ear regulate bone remodeling peripherally and that bilateral vestibular lesions cause significant appendicular bone loss (63). This finding may suggest causality between vestibular dysfunction and severe osteoporosis, which often coexist in astronauts and in older individuals. Yet, another clinical insight stems from observations that reduced ADRB2 signaling can limit breast cancer metastasis to bone (64). In fact, there is growing evidence that autonomic nerves, both sympathetic and parasympathetic, can drive tumor growth and infiltration (65).

Regulation by pituitary hormones

Hormones from the pituitary gland have notable effects on the skeleton, challenging views on their solitary functions, such as master hormone secretion (Fig. 2). Both osteoclasts and osteoblasts possess abundant GPCRs for growth hormone (GH), thyroid-stimulating hormone (TSH), follicle-stimulating hormone (FSH), adrenocorticotropic hormone (ACTH), prolactin, oxytocin (OXT), and arginine vasopressin (AVP) (66–71). Certain pituitary hormones are themselves expressed by bone cells, underscoring paracrine regulation. ACTH is produced by macrophages (72), making it possible that its receptor, melanocortin 2, in bone is activated by local plus pituitary-derived ACTH. Likewise, macrophages synthesize a Tshβ splice variant (73, 74), whereas osteoblasts produce abundant OXT under estrogen control (75, 76). Indeed, GPCRs for pituitary hormones have primitive functions in coelenterates and bony fish (77). Their presence in tissues other than traditional endocrine targets and their coexistence with ligands thus come as no surprise. What does come as a surprise is that the skeleton is exquisitely sensitive to GPCR stimulation, arguably more than the primary target organs. That haploinsufficiency of TSH receptors in heterozygotic Tshr^+/− mice causes osteoporosis without affecting thyroid function (66) provides compelling evidence for such skeletal sensitivity. Likewise, in postmenopausal women, a single subcutaneous injection of low-dose recombinant TSH attenuates bone resorption within 2 days without increasing serum thyroid hormone levels (78).

GH plays a fundamental function in chondrocyte and bone cell function and modulates skeletal growth and modeling. Whereas a GH receptor is expressed in osteoblasts, the primary function of GH is exerted through its release of IGFs. IGF1, the predominant growth factor, is synthesized mainly in the liver, and 80% of the circulating form is bound to IGF-binding protein 3 and the acid labile subunit. Laron syndrome, which arises from loss-of-function mutations in the GHR gene, is characterized by dwarfism and osteoporosis (79). Likewise, Ghr^−/− mice display growth retardation and osteoporosis as phenotypic hallmarks (80). However, the Ghr^−/− phenotype is compensated by the overexpression of IGF1, and mice lacking both liver IGF1 and acid labile subunit, with depleted serum IGF1, show reduced bone growth and bone strength in the face of high circulating GH levels (81). Whereas these results suggest that the skeletal effects of GH require IGF1, there is equally compelling evidence that GH can act independently of IGF. In ovariectomized liver-deficient IGF1 mice, for example, GH reverses osteopenia (67). GH replacement also reverses increased adiposity in hypophysectomized rats, whereas IGF1 replacement does not (82). Together, the latter findings not only point to a direct action of GH on bone but also extend GH actions to the control of adiposity.

TSH displays both antiresorptive and anabolic actions. Antiresorptive actions of TSH are mediated via reductions in NF-κB and c-Jun N-terminal kinase signaling and inhibited TNF-α production (66, 83, 84). As proof, the latter is upregulated in osteoporotic Tshr^−/− mice (66), and importantly, the genetic deletion of Tnfα in Tshr^−/− mice reverses the osteoporosis (85). In contrast, the anabolic actions of TSH, noted in both rodents and people, are exerted through increased osteoblastogenesis mediated by WNT5A upregulation (84, 86–88). Of note are strong epidemiologic correlations among low serum TSH levels, increased bone resorption, low bone mineral density, and increased fracture risk in thyroid cancer and hypothyroid patients receiving TSH-suppressive therapy, as well as in normal women with osteoporosis who have low–normal euthyroid TSH levels [see review in Zaidi et al. (89)]. In contrast, patients with gain-of-function TSHR^D727E polymorphisms have high bone mass and reduced resorption [e.g., Albagha et al. (90)]. Definitive evidence for a contribution of low TSH signaling to thyrotoxic bone loss nonetheless comes from mouse studies: whereas wild-type hyperthyroid mice lose bone, the bone loss is greatly exaggerated in Tshr^−/− mice rendered hyperthyroid (91).

TSH action on the skeleton is opposed by FSH, which normally stimulates ovarian estrogen synthesis and secretion. FSH inhibits bone formation but accelerates bone resorption by stimulating osteoclasts via an FSH receptor (FSHR) isoform linked to guanine nucleotide-binding protein G_i subunit alpha 2 (68, 92). FSH-induced osteoclastogenesis is abolished in mice lacking immunoreceptor tyrosine–based activation motif (ITAM) adapter signaling molecules (93), suggesting an interaction between the FSHR and immune receptors. FSH also enhances RANK expression (94), and stimulates osteoclastogenesis indirectly through IL-1β, TNF-α, and IL-6 release (95, 96). When administered in vivo, it augments ovariectomy-induced bone loss, whereas an FSH antagonist attenuates bone loss postovariectomy or after FSH injection (97, 98).

As for a role for high circulating FSH levels in causing menopausal bone loss, the Study of Women’s Health Across the Nation documents strong estrogen-independent correlations between high FSH levels and hyper-resorption, most marked during the late perimenopause when the rate of bone loss is the steepest, but serum estrogen is normal (99, 100). Four Chinese cohorts, National Health and Nutrition Examination Survey III, and Bone Turnover Range of Normality have collectively shown strong correlations between high serum FSH levels and bone loss [see review in Zaidi et al. (89)]. Closest to causality, however, are observations that amenorrheic women with a high serum FSH level (∼35 IU/L) have greater bone loss than those with lower levels (∼8 IU/L) (101), that the effectiveness of estrogen therapy is related to the extent of FSH suppression (102), and that women harboring an activating FSHR rs6166 polymorphism have lower bone mass and high resorption markers (103).

With that said, it has often been difficult to tease out the action of FSH from that of estrogen on bone in vivo (104, 105). This is indeed expected, as FSH releases estrogen, and the actions of FSH and estrogen on bone are opposed but use nonoverlapping mechanisms. A recent study addressing the importance of high FSH in causing perimenopausal bone loss used 4-vinylcyclohexene diepoxide, an ovatoxin, to phenocopy the estrogen-replete, high-FSH, low-inhibin phase in rats. An impressive estrogen-independent decrement in bone density (5% to 13%) was noted during this phase (106). Likewise, an epitope-specific antibody against a peptide sequence of the receptor-binding domain of mouse FSHβ prevented bone loss after ovariectomy by inhibiting bone resorption and enabling new bone synthesis without affecting estrogen (107, 108).

Pituitary hormones also appear to oversee the intergenerational transfer of calcium from the mother to enable fetal skeletogenesis during pregnancy and skeletal mineralization during postnatal life (109). Both phases of procreation, namely pregnancy and lactation, are characterized by excessive maternal bone resorption and bone loss, which is reversed promptly upon weaning (109). PRL, a primary regulator of lactation, increases resorption by reducing OPG to enhance calcium bioavailability (110). Unlike PRL, whereas mediating milk ejection in nursing mammals, the major action of OXT is on bone formation. It acts on osteoblastic OXT receptors, which translocate to the nucleus to enhance osteoblast differentiation and mineralization (69, 111). Oxt^−/− and Oxt receptor–deficient mice thus display severe osteoporosis as a result of a bone-forming defect, and OXT injections increase bone mass (69). However, that Oxt^−/− pups have hypomineralized skeletons and that pregnant Oxt^−/− moms show reduced bone formation establish a permissive role for OXT in fetal skeletal mineralization (112). OXT also interacts with osteoblastic receptors for AVP (113), another posterior pituitary hormone with opposing skeletal actions (70). The independent antianabolic and proresorptive actions of AVP assume importance in conditions of chronic hyponatremia, such as syndrome of inappropriate antidiuretic hormone secretion, known to be associated with profound osteoporosis. Elevated AVP levels may increase bone resorption to mobilize sodium from its extensive skeletal reservoir and in doing so, cause bone loss.

Crosstalk Within the Local Bone Marrow Niché

Bone cell–immune cell communication

Cortical and trabecular bone are both shrouded with bone marrow that contains immune cells, erythrocytes, megakaryocytes, endothelial cells, and adipocytes (Fig. 3). Whereas work in the early 1970s pointed to a relationship between bone and immune molecules, including TNF-α, IL-1, IL-6, and interferon (IFN)-γ (114), studies showing that osteoclasts themselves bear immune receptors, notably osteoclast associated, Ig-like receptor, triggering receptor expressed on myeloid cells, signal regulatory protein beta 1, and paired Ig-like receptor A, laid the framework for an osteo-immune interface [see review in Takayanagi (115)]. Two immune receptor adapter proteins, namely ITAM in DNAX-activating protein of 12 kD and Fc receptor γ subunit, are activated upon RANKL–RANK interaction, which in turn leads to the recruitment of spleen tyrosine kinase, Ca²⁺ release, calcineurin activation, and in cooperation with activator protein 1, the critical step of NFAT2 autoamplification (115). That Dap12^−/− mice fail to lose bone upon ovariectomy testifies to the fundamental importance of this immune pathway (93). Likewise, there is evidence downstream for reduced osteoclastogenesis and bone loss in Cnaα^−/− mice (116, 117). The latter evidence also supports the premise that calcineurin inhibitors, such as tacrolimus, when used after organ transplant, directly cause acute, rapid, and severe bone loss (118). Additionally, as noted previously, T cells produce proresorptive and antiresorptive cytokines that, in concert, optimize overall bone remodeling (Fig. 3). For example, IFN-γ inhibits, whereas TNF-α stimulates osteoclastogenesis. Low IFN-γ production from T helper 17 cells is thus permissive to the bone loss in autoimmune arthritis and hyperparathyroidism (119, 120). Finally, recent work from the McCauley laboratory (121) highlights the importance of bone marrow macrophages in regulating osteoblast apoptosis, thereby providing a further link between immune surveillance and bone remodeling.

Figure 3.

Crosstalk of bone cells within bone marrow regulates bone remodeling, adipogenesis, hematopoiesis, and angiogenesis. Bone cells interact with multiple bone marrow cells, including HSCs, MSCs, adipocytes, T cells, erythroid precursors, and CD31^hiendomucin (Emcn)^hi endothelial cells of H-type vessels. Osteocytes and osteoblasts communicate with osteoclasts by secreting Rankl, which acts on the Rank receptor that interacts with two ITAM-containing immune receptors, DNAX-activating protein of 12 kD (Dap12) and Fc receptor γ subunit (Fcrγ), as well as with Lgr4, a negative regulator. Osteoclasts are also controlled both directly and via the osteoblast by multiple pro-osteoclastic and antiosteoclastic cytokines, prominently Tnfα, IL-1, IL-4, IL-6, IL-17, and Ifnγ. White adipocytes, prominently seen in hypogonadal states, are MSC derived and secrete adipokines, notably adiponectin, which interacts with osteoblastic receptors. Adipocytes also secrete leptin (Lep), which not only interacts with adipocyte and osteoblast leptin receptors (Leprs) but also crosses the blood-brain barrier to regulate the hypothalamic relay of sympathetic signaling to bone. In addition to osteoblastic suppression exerted through Adrb2, sympathetic nerves innervating MSCs regulate the HSC niché and hematopoiesis via C-X-C motif chemokine ligand 12 (Cxcl2) release that primarily causes HSC egress. Through the activation of a hypoxia-inducible factor 1a (Hif1a) pathway in response to hypoxia, MSCs also produce erythropoietin (Epo) to cause erythrocytosis but also, through notch and NOG, to regulate angiogenesis. OPG, osteoprotegerin; Th17, T helper 17 cell.

Bone regulates blood and blood vessel formation

Bone formation and blood formation are critically intertwined, with the osteoblast having a primary role in controlling hematopoiesis (Fig. 3). Hematopoietic precursors reside in close proximity to stromal osteoprogenitor cells within a niché (122). Osteoprogenitor cells regulate this niché by producing molecules, such as Wnt and Jagged 1, that determine HSC renewal, maturation, and survival (123, 124). These signals are compounded by the circadian rhythmicity of HSC egress occurring in antiphase with expression of the C-X-C motif chemokine ligand 12 (125). The latter is dependent on the circadian secretion of noradrenaline by the sympathetic nerves, which in turn, acts on osteoprogenitor cell ADRBs (54). Disruption of sympathetic signaling, for example, in acute myeloblastic leukemia dysregulates quiescent niché cells, causing stromal cell expansion and commitment to the osteoblast lineage at the expense of periarteriolar cells that normally maintain the HSC lineage (126).

Despite being richly endowed with blood vessels, there is a notable oxygen gradient in bone and bone marrow. In response to hypoxia, stromal osteoprogenitor cells within the niché also promote erythropoiesis by producing erythropoietin (127). Expansion of the erythroid lineage, in turn, protects mice from stress-induced anemia. Hypoxia-inducible factor activation also initiates both angiogenesis and new bone formation, which are tightly coupled through angiocrine, Notch, and NOG signals (128). The inducible disruption of Notch signaling selectively in endothelial cells thus impairs both angiogenesis and osteogenesis (128). Crosstalk is facilitated by a subtype of the cluster of differentiation (CD)31^hiendomucin^hi vessel that generates a distinct TGFB3-rich microenvironment to maintain perivascular osteoprogenitors (129). Importantly, both angiogenesis and osteogenesis decline during aging, resulting in profoundly reduced blood vessel and bone formation (129).

It has also become clear that endothelial cells can directly convert to bone-forming cells. They acquire a stemlike phenotype upon the overexpression of constitutively active activin receptor-like kinase 2 or treatment with TGFB2 and BMP4 (130). These latter cells are potentially differentiable to osteoblasts or adipocytes (130). There is also evidence that ectopic bone in the early phase of traumatic heterotopic ossification arises from endothelial cells through inhibition of endothelial-to-mesenchymal transition by miR-630 that targets snail family transcriptional repressor 2 (131). Finally, it has been speculated that osteoblastic metastases from prostate cancer may have an endothelial cell origin. Thus, BMP4 overexpression in nonosteogenic C4-2b prostate cancer caused ectopic bone formation (132).

Bone marrow adipocytes

Advancing age in both sexes and menopause populate bone marrow abundantly with adipocytes, the focus of recent attention (133) (Fig. 3). Whereas the origin of these cells is not well established, lineage tracing supports a mesenchymal origin with a signature that includes osteogenic marker genes, including osterix, Runx2, and Lepr (134). Osteogenesis and adipogenesis are often reciprocally related. Deletion of the leptin receptor (LEPR) from stromal cells using a paired related homeobox 1 promoter enhances osteogenesis and inhibits adipogenesis (135). The reverse is true with leptin that activates Janus kinase 2/signal transducer and activator of transcription 3 signaling in stromal cells. Notwithstanding their unclear origin, bone marrow adipocytes, which increase dramatically in hypogonadal states, are not inert; they secrete adipokines, importantly adiponectin, and fatty acids that can act on osteoblasts in the niché (136). More recently, it has been demonstrated that marrow adipocytes differ from peripheral fat cells in their capacity to generate RANKL. Osteocytes also influence marrow adipogenesis by inhibiting WNT/LRP5 signaling, such as SOST (137).

Signals That Integrate Bone With Vital Organs

Initial evidence for crosstalk between bone and any vital organ, notably the kidney, came from the finding that excess bone-derived fibroblast growth factor (FGF) 23 acted on the renal FGF receptor 1–Klotho complex to increase phosphate excretion, causing a mineralization defect in tumor-induced osteomalacia (138). Advances in integrative physiology have since unleashed unique, hitherto unrecognized, connections between bone and other vital organs with important therapeutic potential (Fig. 4). The idea that bone itself is an endocrine organ is rapidly undergoing validation.

Figure 4.

Distant interactions between bone and vital organs. In addition to the central nervous system and pituitary, a number of vital organs have connections with bone cells, mainly osteoblasts. Uncarboxylated osteocalcin is thought to interact via a G protein–coupled receptor, family C, group 6, member A (GPCR6A) to enhance insulin secretion from pancreatic β-cells, improve peripheral glucose tolerance by acing on white adipocytes, increase muscle mass, elevate testosterone secretion, stimulate brain development, and improve cognition. White adipocytes produce adiponectin and leptin (Lep); the latter not only acts on osteoblast LEPRs but also crosses the blood-brain barrier to regulate the SNS, which in turn, causes lipolysis through Adrb1/2/3 activation. Osteocytes produce fibroblast growth factor (Fgf)23, a hormone that interacts with the Fgf receptor (Fgfr) and Klotho in the renal tubule to increase phosphate secretion; the kidney, in turn, generates active 1,25-dihydroxyvitamin D₃ (1,25-D₃)to stimulate bone resorption and increase Ca and P absorption from the gut. Less is known about muscle–bone connectivity, except that irisin, produced during exercise, stimulates cortical bone mass, and osteoblastic osteocalcin is thought to maintain muscle mass. Adrb, β-adrenergic receptor; Dlk1, delta-like noncanonical Notch ligand 1; Insr, insulin receptor; VMH, ventromedial hypothalamus.

Bone–muscle connection

Bone and muscle work as a single functional unit, with osteoporosis occurring in tandem with sarcopenia. Yet, mechanisms through which muscle and bone connect are just beginning to be clarified using both cell-specific gene deletion and surgical denervation strategies. Prior studies focused on deletion of pleotropic genes, such as the chaperone methylation gene methyltransferase 21c, originally identified through GWAS on patients with inclusion body myositis and Paget’s disease of bone (139). However, more recently, the deletion of membrane-bound transcription factor peptidase, site 1 selectively in osteocytes was shown to trigger a muscle regeneration gene program (paired box 7, myogenin, myogenic differentiation 1; Notch, and myosin heavy chain 3) and to increase the mass and contractility of slow-twitch soleus muscle (140). In contrast, skeletal muscle-specific deletion of the brain and muscle Arnt-like protein-1 (aryl hydrocarbon receptor nuclear translocator-like) gene, which encodes a molecular clock transcription factor, was found to impair muscle function and to result in bone and cartilage defects (141).

The functional interdependence of muscle and bone has been further established by experimentally transecting the spinal cord. Spinal cord injury is a major cause of severe osteoporosis and progressive muscle loss, and the two together lead to a precipitous increase in fracture risk. Interestingly, electrical stimulation of denervated muscle in rats reverses the increases in bone resorption and elevations in antianabolic Wnt pathway genes, including Dkk1, Sfrp2, and Sost (142). Whereas this may translate into possible anabolic effects of electrical stimulation (or exercise) in increasing bone mass, a phenomenon not readily observed in people, the establishment of true, non-neural connectivity between bone and skeletal muscle has propelled a search for coupling molecules. Irisin, a myokine released upon muscle use, stimulates osteoblast differentiation and increases cortical bone mass and strength at low doses (143). In contrast, osteocalcin, a bone-specific peptide, increases muscle mass, as well as improves glucose sensitivity (144).

Bone, adipose tissue, and global energy metabolism

Bone remodeling and adipose tissue remodeling are coupled functionally through a complex neuroendocrine circuit that involves the brain, pituitary gland, adipose depots, and the skeleton (Fig. 4). Sympathetic signals not only decrease bone formation but also drive lipolysis via three adipocytic ADRB subtypes. Adipocytes, in turn, produce leptin, which as a sensor of peripheral fuel status, crosses the blood-brain barrier to act on ventromedial hypothalamic nuclei receptors to trigger sympathetic relay (Fig. 2). Adipocytes also produce adiponectin in response to peroxisome proliferator–activated receptor γ activation, which in part, enhances sympathetic tone to cause bone loss (145). The latter, to an extent, accounts for the fractures associated with the treatment of individuals with a selective peroxisome proliferator–activated receptor γ agonist, such as rosiglitazone. Type 2 diabetes, treated with glitazones, often shows enhanced bone resorption and suppressed bone formation, as marrow adipogenesis is enhanced. In contrast, several appetite-regulating hormones, including peptide Y, ghrelin, and adiponectin, link bone loss with weight loss during caloric restriction in conditions, such as anorexia nervosa (136).

We have shown recently that the inhibition of FSH signaling in mice by a blocking antibody to FSHβ not only improves bone mass but also dramatically lowers body fat in all compartments, including bone marrow, causes beiging of white adipose tissue, and induces a thermogenic response (146). These effects were noted on mice pair fed on a high-fat diet following ovariectomy and when mice were allowed to consume chow ad libitum (146). These antiobesity actions were phenocopied in male Fshr haploinsufficient mice in which the antibody failed to reduce body fat, confirming its in vivo action via FSH inhibition. We have also documented that fat tissue has FSHRs, the stimulation of which initiates reduced cAMP activation and uncoupled protein 1 expression (146, 147). These effects of FSH inhibition on the skeleton and fat are critical as high circulating FSH levels track with rapid bone loss and the onset of weight gain during the perimenopausal transition (99, 100).

How therefore does bone signal back to fat and consequently, to energy homeostasis? The bone matrix protein, osteocalcin, the production of which from osteoblasts is regulated by insulin signaling in its uncarboxylated form, is thought to act upon GPCR family C, group 6, member A receptors to stimulate β-cell proliferation and insulin secretion (148, 149). Thus, osteocalcin (bone γ-carboxyglutamate protein)-deficient and GPCR family C, group 6, member A–deficient mice display decreased β-cell replication, glucose intolerance, and insulin resistance (148–150). The osteocalcin effect on β-cells is, in turn, regulated centrally by leptin-induced sympathetic activity and locally via delta-like noncanonical Notch ligand 1, an osteocalcin-regulated β-cell gene (151, 152).

Therapeutically relevant, however, are the improvements, albeit limited, in glucose tolerance and insulin resistance, noted with intermittent osteocalcin given to mice on a high-fat diet (153). In contrast, muscle insulin resistance worsens when insulin receptor deletion in osteoblasts reduces plasma uncarboxylated osteocalcin (as a result of impaired pH-dependent decarboxylation) (154, 155). Nonetheless, such worsening of insulin resistance has not yet been reported in patients with type 2 diabetes, who are also often on bisphosphonates, drugs known to reduce resorption and thus, lower serum uncarboxylated osteocalcin. However, if validated in people, the osteocalcin–metabolism connection should provide a vital link in the integrative circuit that regulates bone mass, body composition, and energy use, although in this case, the master regulator is the skeleton.

Bone and distant immunity

Evidence has been mounting for crosstalk between bone and distant immune organs. For example, widespread immunologic dysfunction—particularly, defective thymic T cell egress and systemic cytokine dysregulation—has been implicated in the profound bone loss associated with certain monogenic disorders, such as Gaucher disease (156, 157). However, more intriguing are recent studies that implicate the gut microbiome in postmenopausal osteoporosis. Estrogen deficiency in mice causes increased gut permeability; immune cell activation; and RANKL, TNF, and IL-17 production in the intestine (and bone marrow), resulting in bone loss (158). This chain of events does not occur upon ovariectomy of germ-free mice, and probiotics promptly reverse these effects (158, 159). A question therefore arises of whether simple, over-the-counter medicines, such as probiotics, can in the future be tested for the prevention or treatment of postmenopausal bone loss.

Therapeutic Targeting of Bone

Current therapies for osteoporosis

Despite the fact that over 100 million individuals across the globe have osteoporosis, with over 2 million fractures annually, the armamentarium of osteoporosis therapies pales in comparison with that for other public health hazards of a similar or lesser magnitude, including cardiovascular disease, diabetes, and cancer. Osteoporosis-related fractures have also rocketed over the past decade as a result of an overall increase in lifespan. Furthermore, it has been gleaned that increases in bone resorption begin as early as 2 years before menopause (100). There is thus a need for new therapies to reduce not only the risk of fracture in patients that have already lost bone but also to prevent the active loss of bone early in menopause.

Current therapies for treating and/or preventing osteoporosis include antiresorptive agents, namely estrogen, raloxifene, bisphosphonates, and denosumab, and the skeletal anabolic agents, teriparatide and abaloparatide, all of which reduce the risk of fracture [see review in Pazianas et al. (160); (Fig. 5)]. Estrogens and raloxifene, a selective estrogen receptor modulator (SERM), are generally used early in menopause, as they prevent hyper-resorption of bone by inhibiting the increased osteoclastogenesis. Unlike estrogenic molecules, bisphosphonates are imbibed into osteoclasts, where they inhibit the enzyme farnesyl pyrophosphate synthase to reduce resorption. As a result of their persistence in bone, bisphosphonates have been implicated in rare complications, including osteonecrosis of the jaw and atypical femoral fractures, although there is no evidence for a causal association (161). In contrast, denosumab, an injectable monoclonal antibody, does not reside in bone and by blocking the action of RANKL on the RANK receptor, prevents osteoclast formation, function, and survival.

Figure 5.

Current and near-future therapies to prevent and treat bone loss in osteoporosis. Therapies to prevent and/or treat osteoporosis either act to suppress bone resorption (termed antiresorptive or anticatabolic agents) or stimulate new bone formation (anabolic agents). Either action allows bone formation to exceed bone resorption either in absolute or relative terms and therefore, has a net positive effect on bone mass. Estrogen (E₂) and the SERM raloxifene (and other similar drugs in the pipeline) act on the estrogen receptor α (ERα) in a tissue-specific manner. Thus, raloxifene is proestrogenic in reducing osteoclastogenesis and hence, reduces the risk of vertebral fractures, as noted in the Multiple Outcomes of Raloxifene Evaluation Trial. In contrast, it is antiestrogenic in reducing breast epithelial cell proliferation with potent effects in reducing ER⁺ breast cancers in the Study of Tamoxifen and Raloxifene and Continuing Outcomes Relevant to Evista (raloxifene) Trials. It is therefore used for osteoprotection during early menopause when osteoclastogenesis is high, particularly in patients with a high risk of breast cancer. The currently approved bisphosphonates (alendronate, risedronate, ibandronate, and zoledronic acid) display a high avidity to bone hydroxyapatite with which their N-atoms form H-bonds. When an osteoclast resorbs bone, the drug is released, enters the osteoclast by pinocytosis, and inhibits the enzyme farnesyl pyrophosphate synthase (FPPS). The blocking of farnesylation of small GTP-binding proteins is the basis of their potent antiresorptive actions in osteoporosis and skeletal metastasis. Besides inhibiting FPPS, bisphosphonates also interact with other targets, such as EGFRs, and may therefore exert anticancer actions directly. They also have known antiangiogenic actions, likely arising from a weak action on vascular EGFRs. Denosumab is a human monoclonal antibody that binds to and prevents the interaction of osteoblast- and osteocyte-derived RANKL to the osteoclast receptor RANK. In doing so, denosumab reduces osteoclastogenesis, inhibits bone resorption by mature cells, and induces osteoclast apoptosis. It reduces the risk of fracture at all sites (Fracture Reduction Evaluation of Denosumab in Osteoporosis Every 6 Months Trial) and is used for high-risk osteoporosis. The CTSK antagonist, odanacatib, was recently withdrawn from active development as a result of potential off-target effects; nonetheless, the phase III clinical studies do prove that CTSK is a valuable therapeutic target for osteoporosis. The only approved anabolic agent teriparatide or recombinant N-terminal fragment 1-34 of PTH [PTH(1-34)] acts directly on the PTH/PTH-related protein (PTHrP) receptor on the osteoblast to stimulate its bone-forming action with a substantial reduction of fracture risk when given intermittently. The new, likely-to-be-approved anabolic PTHrP(1-34) (abaloparatide) mimics the action of PTH mechanistically but displays a more rapid effect in reducing vertebral fracture risk in clinical trials that compare either abaloparatide or teriparatide against placebo (Abaloparatide Comparator Trial In Vertebral Endpoints Trial). Romosozumab, currently in clinical development, is a humanized monoclonal antibody to the WNT inhibitor SOST. By removing SOST from the LRP5/6 receptor, romosozumab triggers WNT signaling to a therapeutic advantage in osteoporosis.

Teriparatide or recombinant human N-terminal fragment 1-34 of PTH [PTH(1-34)], when injected intermittently, acts on the osteoblast PTH receptor 1 (PTH1R) to stimulate new bone synthesis and hence, reduce fracture risk in severely osteoporotic patients. This anabolic action of intermittent PTH contrasts with the profound bone loss seen in patients with hyperparathyroidism and in rats infused with PTH (162, 163). The latter response to continuous PTH has been explained by prolonged activation of the PTH1R (164), but more recent evidence provides a firm molecular basis via the selective activation of protein kinase Cδ (165). A newly approved anabolic therapy, abaloparatide, is a human recombinant PTH-related protein (PTHrP) analog that has been shown to act more rapidly than teriparatide and to reduce spine and hip fracture risk (166). One explanation for this difference could arise from the distinct modes of binding of PTH vs PTHrP to PTH1R. Studies using peptide analogs have shown that whereas PTH binds to the PTH1R–G protein–coupled state, PTHrP prefers the G protein–uncoupled state and can hence produce cumulatively greater signaling from G protein–uncoupled → PTH1R–G protein–coupled state isomerization (167). Even this elegant explanation may be simplistic, however, considering the complexity of dynamic atomistic-level interactions among PTH, PTHrP, and the PTH1R for which long-range molecular dynamics will be necessary.

Identifying new druggable targets

The discovery of genetic diseases arising from rare variants with large bone mass effects (Table 1) and their subsequent validation in mice continues to yield potentially valuable targets for therapeutic interrogation. Indeed, the WNT–bone connection initially emerged from the discovery of rare disease-causing mutations within the SOST and LRP5 genes that caused clinically severe high and low bone mass syndromes (Table 1). Likewise, the development of odanacatib, a cathepsin K (CTSK) inhibitor, arose from the discovery of rare loss-of-function mutations of the CTSK gene that caused dense bones in pycnodysostosis. With that said, the genetic architecture of human osteoporosis largely remains unresolved. The ∼30 GWAS studies, to date, have yielded sparse, new information on bone genes, in essence confirming loci associated with the RANKL, OPG, and WNT, among other known pathways (168). Nonetheless, new genes, such as matrix extracellular phosphoglycoprotein, SRY-box 4, spectrin beta, and nonerythrocytic 1, are beginning to be discovered by combining GWAS with gene expression and proteomic profiling (169). More impressive is the prowess of exome sequencing to identify common disease-causing variants, including epigenetic regulators (170, 171). Finally, targeting bone genes with miRNAs toward an anabolic advantage has also begun, particularly as certain miRNA signatures correlate with bone density and fracture risk in postmenopausal women (172).

Future challenges and opportunities

The translating of gene discoveries into druggable targets for osteoporosis has been equally challenging. It has required careful pharmacological testing, followed by relatively lengthy efficacy and safety trials with fracture reduction end points. Promising drugs, such as odanacatib and romosozumab, have fallen through postphase III, as a result of unexpected off-target signals. In addition to off-target effects, the focusing on any newly discovered gene that displays interactions with other genes in bone mass regulation, such as CD40 and CD40 ligand (173), is likely to have unpredictable outcomes. Equally challenging is when rare and common variants interact epistatically, such in the skull disorder craniosynostosis, where rare de novo SMAD6 mutations interact with common BMP1 variants (174). Even more difficult to predict are gene-to-environment interactions. Interestingly, alcohol consumption interacts with the rs2077647 SNP of the estrogen receptor 1 gene, which might explain variations in the effectiveness of estrogens and SERMs in osteoporosis patients.

These challenges have shifted the focus from de novo drug development to the repurposing of existing US Food and Drug Administration–approved drugs for new uses. Historically, repurposing began with discovery of aspirin as an antiplatelet agent and has now moved to a relentless search of databases, such as the connectivity map, and chemical libraries. For example, meclizine, an ERK antagonist, is being tested to enhance growth rates in children with achondroplasia (175). In contrast, bone-active bisphosphonates have been shown to bind directly to EGF receptors (EGFRs) to inhibit tumor growth, findings that can potentially redirect their use for the primary treatment of lung, breast, gastrointestinal, head and neck, and other cancers (176).

In conclusion, there is burgeoning interest in the skeleton, not only as an endocrine target but also as a true endocrine organ with both intercellular and interorgan crosstalk. The use of genetically modified mice has allowed precise insights into hormonal circuitry and to recapitulate various forms of human osteoporosis in mice. New therapeutic targets have thus unmasked and our prediction that therapeutic development is likely to shift in focus from using a single agent for a single disease to the concomitant therapy of coexisting disorders. For example, osteoporosis and the accompanying visceral obesity in postmenopausal women could potentially be treatable by a drug that builds new bone and reduces body fat. Toward this, our anti-FSH antibody shows promise (146, 147).

Glossary

Abbreviations

^−/−

deficient

ACH

acetylcholine

ACTH

adrenocorticotropic hormone

ADRB2

β-adrenergic receptor 2

AVP

arginine vasopressin

BMP

bone morphogenetic protein

CB (CNR)

cannabinoid receptor 2

CD

cluster of differentiation

CTSK

cathepsin K

DAP12

DNAX-activating protein of 12 kD

DKK1

Dickkopf 1

EGFR

epidermal growth factor receptor

FGF

fibroblast growth factor

FSH

follicle-stimulating hormone

FSHR

follicle-stimulating hormone receptor

GH

growth hormone

GPCR

G protein–coupled receptor

GWAS

genome-wide association study

HSC

hematopoietic stem cell

HTR

5-hydroxytryptamine receptor

IFN

interferon

IGF

insulin-like growth factor

ITAM

immunoreceptor tyrosine–based activation motif

LEPR

leptin receptor

LGR4

leucine-rich repeat-containing G protein–coupled receptor 4

LRP5

lipoprotein receptor–related protein 5

miRNA

microRNA

MSC

mesenchymal stem cell

NF-κB

nuclear factor κB

NFAT2

nuclear factor for activated T cells 2

NOG

noggin

OPG

osteoprotegerin

OXT

oxytocin

PTH

parathyroid hormone

PTH(1-34)

N-terminal fragment 1-34 of parathyroid hormone

PTH1R

parathyroid hormone 1 receptor

PTHrP

parathyroid hormone–related protein

RANK

receptor activator for nuclear factor κB

RANKL

receptor activator for nuclear factor κB ligand

Runx2

runt-related transcription factor 2

SEMA3A

semaphorin 3a

SERM

selective estrogen receptor modulator

SNS

sympathetic nervous system

SOST

sclerostin

TNFR

TNF receptor

TNFRSF11

TNF receptor superfamily 11

TSH

thyroid-stimulating hormone

WNT

Wingless-intregration-1