Bone remodeling and osteoporosis: RANKL-RANK-OPG signaling, bisphosphonates, and the Frost mechanostat

Intuition Beginner

Bone is not a stone scaffold inside the body — it is living tissue, and it rebuilds itself continuously. Old bone is taken down; new bone is laid down. The two processes run side by side from childhood through old age, and in a healthy adult they exactly balance: the amount of bone removed equals the amount of bone rebuilt, year after year.

Two specialist cell types do the work. Osteoclasts are large, multinucleated cells that dissolve old bone mineral and digest the protein scaffold beneath it — they are the demolition crew. Osteoblasts are smaller cells that lay down new protein matrix and seed it with calcium-phosphate mineral — they are the builders. Osteoclasts arrive first in any given patch of bone; osteoblasts follow and refill the cavity they left.

Osteoporosis is what happens when the balance tilts: removal outruns rebuilding, bone density falls, and the internal architecture thins until ordinary loads — a stumble, a cough, even a twist — break bones that should have held. About 200 million people worldwide have osteoporosis. One in three women and one in five men over fifty will suffer an osteoporotic fracture. The most dangerous is the hip fracture, which carries a one-year mortality of roughly 20 percent.

The molecular switch that decides how many osteoclasts the body makes is called the RANKL-RANK-OPG axis. Osteoblasts display a signal protein called RANKL on their surface. Osteoclast precursors carry its receptor, called RANK. When RANKL finds RANK, an osteoclast is born. To keep the system in check, osteoblasts also secrete a decoy molecule called OPG that mops up RANKL before it can signal. The balance between RANKL and OPG sets the demolition pace.

After menopause, the hormone estrogen falls. Estrogen normally suppresses RANKL and boosts OPG. Without it, RANKL rises, OPG falls, osteoclast numbers climb, and bone loss accelerates. The most-prescribed osteoporosis drugs — bisphosphonates (alendronate, zoledronate) and denosumab — work by either killing osteoclasts directly or blocking RANKL. The first large clinical trial proving that a drug prevents osteoporotic fractures, the FIT study reported in 1996, showed that alendronate cut hip fractures by 51 percent.

Bone also responds to the loads placed on it. Harold Frost, an orthopedic surgeon, proposed in 1964 that bone operates like a thermostat: it senses mechanical strain and adjusts its mass up or down to match — what he called the mechanostat. Astronauts in zero gravity lose roughly 1–2 percent of their bone mass per month; professional tennis players have arm bones 20–40 percent thicker on their racket side than on their other side. Bone listens to what you ask of it.

Visual Beginner

The defining picture has three cells in conversation across a patch of bone surface. On the left is the osteoclast — a large, irregular, multinucleated cell attached to bone like a suction cup, secreting acid and proteases into a sealed pocket beneath it, dissolving the mineral and digesting the matrix. On the right is the osteoblast — a smaller cuboidal cell depositing fresh protein (collagen) onto the repair surface, which then mineralises into new bone.

Buried inside the bone matrix above sit osteocytes — former osteoblasts entombed in small spaces (lacunae) connected by a network of tiny fluid-filled channels (canaliculi). These are the mechanosensors: when the bone bends under load, fluid flows through the canaliculi, the osteocyte's hairs bend, and a signal travels out to recruit osteoclasts or osteoblasts as needed.

The signaling arrows run as follows. The osteoblast-lineage cell displays RANKL on its surface (small triangles). The osteoclast precursor carries the matching receptor RANK (Y-shaped). Binding of RANKL to RANK triggers the precursor to fuse into a mature, multinucleated, bone-resorbing osteoclast. The osteoblast also releases OPG — free-floating decoy molecules that bind RANKL and prevent it from reaching RANK. The ratio of RANKL to OPG determines how many osteoclasts form. Estrogen acts on the osteoblast to suppress RANKL and increase OPG; menopausal estrogen loss removes both checks.

The picture completes itself with the remodeling cycle. A small swarm of osteoclasts digs a trench in old bone over about two weeks. They undergo apoptosis and depart. Osteoblasts arrive over the next four to six months and refill the trench with new bone. The whole unit — osteoclasts plus following osteoblasts — is called a basic multicellular unit (BMU). At any one time, roughly a million BMUs are working in the adult skeleton; each replaces a small parcel of bone every few years.

Worked example Beginner

A 68-year-old woman presents to her primary-care clinic after a fall from standing height onto her right hip. X-rays show a fracture of the right femoral neck (the neck of the thigh bone just below the hip joint) — a fragility fracture, because a fall from standing height should not break a healthy femur. She went through menopause at age 51 and has never taken hormone therapy or bone medication. The clinical picture is postmenopausal osteoporosis presenting with its most feared complication.

Step 1. Confirm low bone density with a DEXA scan. DEXA (dual-energy X-ray absorptiometry) measures bone mineral density at the spine and hip and reports it as a T-score: the number of standard deviations below the average bone density of a healthy 30-year-old of the same sex. Her T-score at the lumbar spine is $-3.1$ and at the femoral neck is $-2.8$. By the World Health Organization diagnostic criterion, T-score $-2.5$ or below is osteoporosis. Both sites are osteoporotic.

Step 2. Quantify fracture risk and start treatment. Her 10-year probability of a major osteoporotic fracture, calculated by the FRAX tool, is approximately 28 percent, and her hip-fracture probability is approximately 7 percent. Alendronate 70 mg once weekly is started. Alendronate is a nitrogen-containing bisphosphonate; taken up by osteoclasts during bone resorption, it inhibits an enzyme in the mevalonate pathway (farnesyl pyrophosphate synthase), depletes the cell of prenylated small GTPases, and triggers osteoclast apoptosis.

Step 3. Quantify the expected benefit using the FIT trial (Black et al., Lancet 1996). The FIT trial randomised 2027 postmenopausal women with low femoral-neck bone density and existing vertebral fractures to alendronate 5 mg/day (later increased to 10 mg/day) or placebo for up to three years. Results: hip fracture was reduced from 2.2 percent (placebo) to 1.1 percent (alendronate) — a 51 percent relative reduction. Clinical vertebral fractures were reduced by 47 percent; any clinical fracture by 28 percent. Translated to the patient in front of us, alendronate roughly halves her hip-fracture risk over the next three years.

What this tells us: a single drug targeting a single cell type (the osteoclast) reverses the demographic risk of osteoporotic fracture in the population that first defined the disease. The FIT trial was the proof, in 1996, that bone remodeling is pharmacologically tractable, and alendronate became the most-prescribed osteoporosis drug worldwide.

Check your understanding Beginner

Formal definition Intermediate+

Bone remodeling is the coupled, asynchronous replacement of old bone by new bone at discrete anatomic sites throughout the skeleton. The remodeling cycle runs in four phases ^{[Boron-Boulpaep 2017]}:

1. Activation. A mechanical or biochemical signal recruits a basic multicellular unit (BMU) to a site on the bone surface. The signal originates in osteocytes sensing microdamage or altered strain.
2. Resorption. Osteoclasts attach to the bone surface, seal a compartment beneath their ruffled border, acidify it (via the vacuolar H+-ATPase and chloride channel ClC-7), and dissolve hydroxyapatite mineral. Cathepsin K digests the demineralised collagen matrix. The resorption phase lasts approximately 2 weeks in cortical bone.
3. Reversal. Osteoclasts undergo apoptosis or depart. Mononuclear cells prepare the surface.
4. Formation. Osteoblasts are recruited (by coupling factors released from the resorbed matrix, including TGF-$\beta$, IGF-1, and sphingosine-1-phosphate), deposit unmineralised osteoid (collagen type I plus non-collagenous proteins), and seed mineralisation. The formation phase lasts approximately 4 to 6 months — roughly ten times longer than resorption.

At any one time, approximately $10^6$ BMUs are active in the adult human skeleton, each replacing a quantum of bone. The net effect on whole-body bone mass is determined by the per-BMU balance between resorbed and formed bone and by the BMU activation frequency.

Cell lineages. Two embryologically distinct lineages converge on bone:

• Osteoclasts are multinucleated giant cells derived from hematopoietic progenitors of the monocyte/macrophage lineage. Their differentiation requires two osteoblast-lineage-derived signals: M-CSF (macrophage colony-stimulating factor, acting through the c-Fms receptor) and RANKL (receptor activator of NF-$\kappa$B ligand, acting through the RANK receptor) ^{[Yasuda 1998 RANKL]}.
• Osteoblasts are derived from mesenchymal stem cells, also the precursors of chondrocytes, adipocytes, and myoblasts. Runx2 and Osterix are the master transcription factors driving osteoblast differentiation. Terminally differentiated osteoblasts may become osteocytes (the most numerous and longest-lived bone cell type; ~$10^{10}$ per adult skeleton), or lining cells on the bone surface, or undergo apoptosis.

The RANKL-RANK-OPG axis. This is the molecular switch controlling osteoclastogenesis. Osteoblast-lineage cells (osteoblasts, bone-lining cells, osteocytes) express RANKL on their surface as a transmembrane protein; a soluble form generated by metalloprotease cleavage also circulates. RANKL binds and activates RANK, a TNF-receptor-superfamily member on the surface of osteoclast precursors. RANK engagement recruits TRAF6, activates NF-$\kappa$B and the MAP kinases, and induces the master osteoclast transcription factor NFATc1 (NFATc1 auto-activates its own promoter, locking in the osteoclast fate) ^{[Lacey 1998 RANK]}. NFATc1 transcribes the osteoclast-specific machinery: tartrate-resistant acid phosphatase (TRAP), cathepsin K, the calcitonin receptor, and the ${\beta }_3$ integrin.

Osteoprotegerin (OPG) is a soluble decoy receptor — a TNF-receptor-superfamily member without a transmembrane domain — secreted by osteoblasts and many other cell types. OPG binds RANKL with higher affinity than RANK does and prevents the ligand-receptor engagement that drives osteoclastogenesis ^{[Simonet 1997 OPG]}. The ratio of RANKL to OPG, not the absolute concentration of either, sets the permissive state of osteoclast formation; the canonical example is the postmenopausal shift, in which estrogen loss raises RANKL and lowers OPG, doubling or tripling the ratio and producing sustained net bone loss.

Mechanosensing. Osteocytes are interconnected by gap junctions and by dendritic processes passing through the canaliculi. When bone is mechanically loaded, interstitial fluid flows through the lacuno-canalicular network, producing shear stress on osteocyte dendrites; this is transduced by integrin focal adhesions, primary cilia, and ion channels (including PIEZO1) into biochemical signals ^{[Frost 1964]}. The downstream effect is biphasic: high strain suppresses sclerostin (the SOST gene product, a Wnt antagonist), releasing Wnt signaling to drive osteoblast bone formation; low strain or disuse upregulates sclerostin and RANKL, tipping the balance toward resorption.

Bone-mineral density (BMD) and T-score. BMD is measured by DEXA in grams of hydroxyapatite per square centimeter of projected bone area. The T-score is the number of standard deviations above or below the young-adult mean BMD:

$$T=\frac{{\mathrm{BMD}}_{\text{patient}}-{\mathrm{BMD}}_{\text{young-adult mean}}}{{\sigma }_{\text{young-adult}}}.$$

The World Health Organization operational definitions are: $T\ge -1.0$ normal; $-2.5<T<-1.0$ osteopenia; $T\le -2.5$ osteoporosis ^{[Rosen 2019 ASBMR Primer]}.

Counterexamples to common slips

• Bone is an inert scaffold. False. Bone is metabolically active, vascularized, innervated, and continuously remodeled; the adult skeleton is replaced roughly every ten years. About 10 percent of the skeleton is actively being remodeled at any moment.

• Osteoclasts make their own decision to resorb. False. Osteoclast differentiation and activation require paracrine signals from osteoblast-lineage cells — the coupling hypothesis of Rodan and Martin (1981). Osteoclasts lack RANKL; they receive the signal from the mesenchymal compartment. This is why osteopetrosis (marble bone disease, characterised by absent osteoclast function) can be cured by bone marrow transplantation (replacing the hematopoietic lineage) but not by mesenchymal grafts.

• Bisphosphonates build new bone. False. Bisphosphonates inhibit osteoclasts and thereby slow bone loss. The bone-density gain seen in the first one to three years of therapy (typically 5 to 10 percent at the spine) reflects partial refilling of remodeling space — the BMUs in mid-cycle at the start of therapy complete their formation phase, while fewer new resorption cavities open — not true anabolic bone formation. The anabolic agents teriparatide (recombinant human PTH 1-34) and romosozumab (anti-sclerostin mAb) build new bone; bisphosphonates merely preserve what is there.

• Osteoporosis is a woman's disease. False as stated, although women predominate. One in three women over fifty will suffer an osteoporotic fracture, but so will one in five men. Men present on average ten years later than women (no equivalent of menopause), with secondary causes more common: hypogonadism, glucocorticoid therapy, alcohol excess, and androgen-deprivation therapy for prostate cancer are the leading contributors in men.

• Stopping denosumab is the same as stopping a bisphosphonate. False and dangerous. Bisphosphonates persist in bone for years (the half-life of alendronate in the skeleton is approximately 10 years), so their antiresorptive effect wanes slowly after discontinuation. Denosumab is a monoclonal antibody cleared from plasma in weeks; its discontinuation produces a rapid rebound of osteoclast activity (RANKL is no longer neutralised), with multiple vertebral fractures reported within 12 to 24 months of stopping. Denosumab discontinuation requires transition to a bisphosphonate to prevent the rebound.

• Alendronate is taken without specific instructions. False. Nitrogen-containing bisphosphonates have very poor oral bioavailability (approximately 0.7 percent under ideal conditions) and can cause severe esophagitis. They must be taken on an empty stomach, with a full glass of water, in the upright position, with nothing else by mouth for at least 30 minutes. Calcium, iron, coffee, and orange juice at the same time abolish absorption.

Key mechanism: RANKL-RANK-OPG axis and bone remodeling Intermediate+

Mechanism (the bone-balance sign theorem: the RANKL/OPG ratio sets the rate, the coupling efficiency sets the sign). At quasi-steady state, the rate of change of bone mass in any anatomic region is the difference between osteoblast-mediated formation and osteoclast-mediated resorption. The sign of this rate is determined by the per-BMU coupling efficiency (whether osteoblasts fully refill the cavity that osteoclasts dug), while the magnitude is set by the RANKL-RANK-OPG axis through the steady-state active-osteoclast number. Estrogen loss in menopause attacks both terms — raising RANKL, lowering OPG, and impairing coupling — which is why postmenopausal bone loss is rapid rather than the slow age-related attrition seen in estrogen-replete adults.

Derivation. Let $B$ denote the bone mass per unit anatomic region (in grams of hydroxyapatite, or equivalently the DEXA-derived BMD in $g/c{\mathrm{m}}^2$). Let $N_{oc}$ and $N_{ob}$ denote the numbers of active osteoclasts and osteoblasts per unit bone-surface area. Let $k_r$ denote the per-osteoclast bone-resorption rate and $k_f$ the per-osteoblast bone-formation rate (in $g/cell/day$). Then

$$\frac{dB}{dt}=k_f{N}_{ob}-k_r{N}_{oc}.$$

The active osteoclast population obeys a quasi-steady balance between RANKL-driven differentiation and apoptosis:

$$\frac{d{N}_{oc}}{dt}=\alpha \cdot S_{\mathrm{RANK}}-{\delta }_{oc}{N}_{oc}=0⟹N_{oc}^{\ast }=\frac{\alpha }{{\delta }_{oc}}{S}_{\mathrm{RANK}},$$

where $\alpha$ is the pre-osteoclast differentiation constant and ${\delta }_{oc}$ the mature-osteoclast apoptosis rate. The RANKL signaling strength $S_{\mathrm{RANK}}\in [0,1]$ follows competitive-inhibition kinetics: OPG is a soluble decoy receptor that competes with RANK for RANKL binding, so

$$S_{\mathrm{RANK}}=\frac{R}{R+K_R(1+O/K_O)},$$

where $R$ is the surface concentration of RANKL on osteoblast-lineage cells, $O$ is the local soluble OPG concentration, $K_R$ is the RANK-RANKL dissociation constant, and $K_O$ is the OPG-RANKL dissociation constant ^{[Yasuda 1998 RANKL; Simonet 1997 OPG]}.

Coupling (Rodan-Martin hypothesis). Osteoblast recruitment to a remodeling site is driven by coupling factors released from resorbed matrix and secreted by active osteoclasts (TGF-$\beta$ mobilised from the demineralised matrix, IGF-1, sphingosine-1-phosphate, cardiotrophin-1, and several others). In the simplest linear approximation, the steady-state osteoblast number is proportional to the osteoclast number:

$$N_{ob}^{\ast }=\beta N_{oc}^{\ast },$$

where $\beta$ is the per-osteoclast coupling efficiency — a dimensionless number equal to the ratio of recruited osteoblasts to active osteoclasts in a BMU. In a healthy adult, $\beta \cdot k_f/k_r\approx 1$ (formation refills the resorption cavity), and the net bone balance is zero.

Substituting $N_{oc}^{\ast }$ and $N_{ob}^{\ast }$ into the bone-mass balance,

$$\frac{dB}{dt}=(k_f\beta -k_r)\cdot \frac{\alpha }{{\delta }_{oc}}\cdot \frac{R}{R+K_R(1+O/K_O)}.$$

This is the bone-balance sign theorem. The sign of $dB/dt$ is set by the coupling-efficiency term $(k_f\beta -k_r)$ alone, while the magnitude is modulated by the RANKL-RANK-OPG signaling strength $S_{\mathrm{RANK}}$. In the healthy estrogen-replete adult, $k_f\beta \approx k_r$ and $dB/dt\approx 0$. In the postmenopausal state, three independent insults compound:

1. Estrogen directly suppresses RANKL transcription and promotes OPG transcription in osteoblast-lineage cells. Estrogen loss raises $R$ (often doubling or tripling surface RANKL) and lowers $O$ (typically halving OPG), so the ratio $R/O$ rises sharply. $S_{\mathrm{RANK}}$ increases; $N_{oc}^{\ast }$ increases; the remodeling rate rises.
2. Estrogen promotes osteoblast survival and coupling-factor responsiveness. Without it, $\beta$ drops because osteoblast recruitment to the resorption cavity is less efficient — the cavity is incompletely refilled.
3. Estrogen promotes osteoclast apoptosis. Without it, ${\delta }_{oc}$ drops, extending osteoclast lifespan and increasing the depth of each resorption event.

All three terms tilt $dB/dt$ negative. The compounded effect is the rapid postmenopausal bone-loss rate of approximately 1 to 2 percent of BMD per year in the first five years after menopause, compared with approximately 0.3 to 0.5 percent per year in the estrogen-replete adult.

Pharmacological corollary. Every approved osteoporosis therapy targets one of the three control terms in the bone-balance equation:

• Bisphosphonates (alendronate 1995, risedronate, zoledronate 2001) bind hydroxyapatite on the bone surface with very high affinity and are ingested by osteoclasts during resorption. Nitrogen-containing bisphosphonates inhibit farnesyl pyrophosphate synthase in the mevalonate pathway, depleting prenylated small GTPases (Ras, Rho, Rac) required for osteoclast cytoskeletal function and survival, and inducing apoptosis ^{[Russell 2008 bisphosphonates]}. In the bone-balance equation, this raises ${\delta }_{oc}$ and lowers the effective $k_r$ — bone loss slows or stops, BMD rises by 5 to 10 percent as remodeling space closes.
• Denosumab (anti-RANKL monoclonal antibody, 2010) acts as an exogenous OPG mimic, raising the effective $O$ and suppressing $S_{\mathrm{RANK}}$ — the steady-state osteoclast number drops sharply, resorption falls, BMD rises by 6 to 9 percent at the spine over three years.
• Teriparatide (recombinant human PTH 1-34, 2002) and abaloparatide (PTHrP analog, 2017) act on the formation side directly, raising $k_f$ when given intermittently — the only approved anabolic (bone-building) class. Romosozumab (anti-sclerostin mAb, 2019) raises $k_f$ by releasing Wnt-driven bone formation, also anabolic.

Bridge. The bone-balance sign theorem builds toward the pharmacology catalog in 35.07.04 — every osteoporosis drug enters this single equation at a specific term, and the chapter-closing synthesis of chronic-disease epidemiology appears in 35.03.01 where osteoporosis aggregates into a global burden of 8.9 million fractures per year. The foundational reason bone remodeling is dynamically analogous to the HPT axis in 18.07.04 is exactly that both are adaptive-tissue feedback loops with a single dominant controllable gain — TSH-receptor activation there, RANKL-RANK signaling here. Putting these together identifies bone mass with the time-integral of a sign-flipping balance equation, and the central insight is that menopausal estrogen loss tilts the sign by attacking three independent terms simultaneously (raising RANKL, lowering OPG, impairing coupling). The bridge is the RANKL/OPG ratio itself — a single dimensionless number summarising the state of osteoclastogenesis in the same way the TSH set-point summarises the state of thyroid feedback — and the pattern generalises to all coupled-resorption-formation systems in mineralised tissue (dental remodeling, vascular calcification, heterotopic ossification), each governed by a local variant of the same equation.

Exercises Intermediate+

Advanced results Master

Theorem 1 (Frost mechanostat — Frost 1964). Harold Frost, an orthopedic surgeon at Henry Ford Hospital, proposed that bone remodeling is governed by a negative-feedback control system in which the controlled variable is bone mass and the sensed variable is mechanical strain ^{[Frost 1964]}. Frost's mechanostat partitions the strain range into four windows: disuse (below approximately 50 to 200 microstrain, net bone loss), adapted (approximately 200 to 1500 microstrain, steady balance), mild overload (approximately 1500 to 3000 microstrain, net bone gain via modeling-based formation), and pathologic overload (above approximately 3000 to 4000 microstrain, microdamage accumulation). The molecular identity of the mechanosensor was solved only in the 2000s with the identification of osteocyte primary cilia, integrin focal adhesions, ion channels (PIEZO1/2, TRPV4, connexin-43 hemichannels), and most consequentially the load-induced suppression of sclerostin (the SOST gene product) releasing canonical Wnt signaling to drive osteoblast bone formation. Frost's mechanostat generalises Wolff's 1892 "law of bone transformation" (Wolff, Das Gesetz der Transformation der Knochen) from a qualitative observation to a quantitative feedback control system with a defined set-point and gain.

Theorem 2 (Rodan-Martin coupling hypothesis — Rodan & Martin 1981). Greg Rodan and T. John Martin proposed that osteoclast-mediated resorption is controlled not by osteoclasts themselves but by cells of the osteoblast lineage, which produce the permissive signals (then unknown, later shown to include M-CSF and RANKL) for osteoclast differentiation ^{[Rodan-Martin 1981]}. The hypothesis was based on two empirical observations: (a) osteoblasts and osteoclasts are never found in direct physical contact (an osteoid seam always separates them), suggesting an indirect paracrine mechanism; and (b) every known agent that stimulates resorption (PTH, IL-1, TNF-$\alpha$, $1,25(\mathrm{OH}{)}_2{\mathrm{D}}_3$, prostaglandin E2) acts on osteoblast-lineage cells, not on mature osteoclasts. The Rodan-Martin hypothesis predicted that the osteoclast-differentiation factor would prove to be a membrane-bound or soluble protein of osteoblast origin — a prediction vindicated 17 years later by Yasuda's cloning of RANKL (Theorem 4). The coupling hypothesis also implied that bone remodeling is coupled: formation follows resorption in time, with osteoblast recruitment driven by coupling factors released from the resorbed matrix (TGF-$\beta$, IGF-1, sphingosine-1-phosphate) and secreted by osteoclasts themselves (cardiotrophin-1, SEMA4D). The pathological uncoupling of resorption and formation is the proximate mechanism of every form of osteoporosis.

Theorem 3 (osteoprotegerin — Simonet 1997). William Simonet and colleagues at Amgen, cloning novel TNF-receptor-superfamily members, identified a secreted protein whose transgenic overexpression in mice produced a profound increase in bone density (osteopetrosis), and whose deletion produced severe osteoporosis; they named it osteoprotegerin (OPG, "bone-protecting protein") ^{[Simonet 1997 OPG]}. OPG is a soluble decoy receptor — a TNF-receptor-superfamily member lacking a transmembrane domain and a death domain — that binds an unknown ligand (later identified as RANKL; Theorem 4) and prevents it from activating its cognate receptor on osteoclast precursors. The Simonet paper established that the osteoclast-differentiation pathway was a tractable ligand-receptor-target axis and triggered the search for the ligand; it also identified OPG as the first known endogenous antagonist of osteoclastogenesis. OPG-Fc (a fusion protein of OPG with the Fc domain of IgG1, extending plasma half-life) was the conceptual ancestor of denosumab (Theorem 7).

Theorem 4 (RANKL cloning — Yasuda 1998; Lacey 1998 RANK). The osteoclast-differentiation factor long postulated by the Rodan-Martin hypothesis was cloned in 1997-1998 by two groups. Yasuda, Suda, and colleagues at Asahi Chemical Industry and Osaka University isolated the factor from a stromal-cell expression library, showing that it was identical to three previously-cloned immune-system proteins (TRANCE/RANKL/ODF), and named it osteoclast differentiation factor (ODF), later unified as RANKL ^{[Yasuda 1998 RANKL]}. Lacey, Boyle, and colleagues at Amgen, working from the OPG-decoy-receptor side, identified RANKL's receptor on osteoclast precursors as RANK (Receptor Activator of NF-$\kappa$B), a TNF-receptor-superfamily member, and showed that RANKL binding to RANK was necessary and sufficient for osteoclastogenesis in vitro and in vivo ^{[Lacey 1998 RANK]}. The RANKL-RANK-OPG axis was thus identified as a single cytokine-receptor-decoy system, ancestral to the immune system (RANKL is expressed in lymph nodes and mediates dendritic-cell survival) and repurposed in bone for osteoclast regulation. The RANKL and OPG knockouts produced the predicted phenotypes: RANKL-knockout mice and OPG-transgenic mice are osteopetrotic (no osteoclasts); OPG-knockout mice are severely osteoporotic (unrestrained osteoclasts).

Theorem 5 (FIT trial — Black et al. 1996, alendronate). The Fracture Intervention Trial, run by Dennis Black, Steven Cummings, and the FIT Research Group, was the first large randomised placebo-controlled trial to demonstrate that a bisphosphonate prevents osteoporotic fractures ^{[Black 1996 FIT]}. The trial enrolled 2027 postmenopausal women with low femoral-neck BMD and existing vertebral fractures, randomised to alendronate 5 mg/day (later increased to 10 mg/day) or placebo for up to three years. Results: clinical vertebral fractures reduced 47 percent (relative risk 0.53, 95 percent CI 0.32 to 0.88); clinical hip fractures reduced 51 percent (RR 0.49, 95 percent CI 0.23 to 0.99); any clinical fracture reduced 28 percent. The FIT trial established alendronate as first-line osteoporosis therapy and was the basis of the 1995 FDA approval (alendronate was the first nitrogen-containing bisphosphonate approved for osteoporosis; etidronate, a non-nitrogen bisphosphonate, had been approved earlier but had weak antiresorptive activity and significant mineralisation defects at therapeutic doses). A second FIT arm (Black et al. JAMA 1998) extended the finding to women without baseline vertebral fractures but with low BMD, showing significant hip-fracture reduction in the lowest-BMD subgroup ($T<-2.5$). The FIT trial's design — fracture outcomes rather than BMD surrogate endpoints — became the regulatory template for every subsequent osteoporosis drug approval.

Theorem 6 (HORIZON — Black et al. 2007, zoledronate). Once-yearly intravenous zoledronic acid 5 mg was shown in the HORIZON-Pivotal Fracture Trial (Black, Delmas, Eastell, Reid et al. N. Engl. J. Med. 357:1799, 2007) to reduce clinical vertebral fracture by 77 percent, hip fracture by 41 percent, and nonvertebral fracture by 25 percent over three years in postmenopausal women with osteoporosis. Zoledronate (approved by the FDA in 2001 for hypercalcaemia of malignancy and in 2007 for osteoporosis) is the most potent nitrogen-containing bisphosphonate in clinical use (FPPS inhibition constant in the subnanomolar range). The once-yearly intravenous route solved the compliance problem of daily oral alendronate (which requires fasting, upright posture, and 30 minutes of nothing-by-mouth) and is the standard therapy for patients who cannot tolerate oral bisphosphonates or in whom adherence is poor. Zoledronate is also the only osteoporosis drug with proven mortality benefit in hip-fracture patients (Lyles et al. N. Engl. J. Med. 357:1793, hip-fracture-only cohort: 28 percent reduction in all-cause mortality over two years).

Theorem 7 (FREEDOM trial — Cummings et al. 2009, denosumab). The FREEDOM trial (Cummings, Martin, McClung, Siris, Eastell, Reid, Delmas et al.) randomised 7868 postmenopausal women with osteoporosis to denosumab 60 mg subcutaneously every six months or placebo for three years ^{[Cummings 2009 FREEDOM]}. Results: new radiographic vertebral fracture reduced 68 percent (from 7.2 percent placebo to 2.3 percent denosumab; RR 0.32); hip fracture reduced 40 percent (RR 0.60); nonvertebral fracture reduced 20 percent. Denosumab, a fully human monoclonal antibody against RANKL developed by Amgen, was the first osteoporosis drug whose mechanism targets the bone-balance equation at the signaling-strength term ($S_{\mathrm{RANK}}$) rather than at the osteoclast-apoptosis or osteoblast-formation term. Denosumab was FDA-approved in 2010 under the trade name Prolia; it is the only osteoporosis drug whose molecular target is a single named cytokine-receptor interaction. The FREEDOM Extension study (Papapoulos et al. Lancet Diabetes Endocrinol. 2015) demonstrated sustained BMD gains over 8 to 10 years of continuous therapy — an unprecedented duration of antiresorptive effect — but also documented the rebound bone loss on discontinuation that has become the principal safety concern of the drug.

Theorem 8 (sclerostin and the molecular mechanostat — van Bezooijen 2004; Lewiecki 2018 romosozumab). The molecular link between osteocyte mechanosensing and osteoblast activation is sclerostin, the product of the SOST gene expressed almost exclusively in osteocytes ^{[van Bezooijen 2004 sclerostin]}. Sclerostin is a secreted Wnt antagonist that binds the Wnt co-receptors LRP5/6 and prevents canonical Wnt signaling, thereby suppressing osteoblast bone formation. Mechanical load suppresses SOST expression in osteocytes (Robling et al. 2008 J. Biol. Chem.), releasing Wnt signaling and driving formation — the molecular implementation of the Frost mechanostat's set-point. Loss-of-function mutations in SOST cause sclerosteosis and van Buchem disease (massive overgrowth of bone skeleton-wide), confirming sclerostin as the dominant tonic inhibitor of bone formation. Romosozumab, a humanised monoclonal antibody against sclerostin developed by Amgen and UCB, was the first anabolic agent whose mechanism is the dual action of increasing bone formation (Wnt release) while transiently decreasing resorption; in the ARCH trial (Saag et al. N. Engl. J. Med. 377:1417, 2017) romosozumab reduced clinical vertebral fracture by 48 percent versus teriparatide over a median 33 months. Romosozumab was approved by the FDA in 2019 with a black-box warning for cardiovascular events, the principal remaining controversy in clinical osteoporosis.

Synthesis. The eight theorems trace the canonical path of a tissue-remodeling system solved across forty years: a feedback-control hypothesis (Frost 1964), a coupling architecture (Rodan-Martin 1981), a soluble decoy receptor (Simonet 1997), the matching ligand and receptor identified as a TNF-superfamily cytokine (Yasuda 1998 RANKL; Lacey 1998 RANK), the regulatory-approved first-in-class drugs targeting the osteoclast (Black 1996 alendronate; Black 2007 zoledronate; Cummings 2009 denosumab), and the molecular identity of the mechanostat (van Bezooijen 2004 sclerostin; Lewiecki 2018 romosozumab). The foundational reason bone remodeling is solved at this depth is exactly that the RANKL-RANK-OPG axis is the simplest cytokine-receptor-decoy architecture that nonetheless generates the full disease taxonomy — postmenopausal, glucocorticoid-induced, male, disuse, malignancy-associated — and admits every major pharmacological class. The central insight is that the RANKL/OPG ratio is the single dimensionless number summarising the state of osteoclastogenesis, and putting these together identifies the osteoporotic phenotype with a sign-flip of the bone-balance equation driven by estrogen loss.

The same pattern recurs in every adaptive-tissue feedback loop studied at comparable depth: the HPT axis (cortisol, 18.07.04), the calcium-PTH-vitamin D axis (parathyroid chief-cell calcium sensing), and the renin-angiotensin-aldosterone axis. The bridge is the RANKL-RANK-OPG architecture itself, a TNF-superfamily cytokine repurposed from the immune system to drive osteoclastogenesis — and the pattern generalises to the broader TNF-superfamily signaling roles in immune regulation, lymph-node organogenesis, and mammary-gland development. Bone remodeling is the cleanest example of how a single cytokine-receptor-decoy module can implement a controllable gain in a tissue-maintenance feedback loop, and the same architecture appears wherever a tissue must continuously replace itself in response to changing load.

Full proof set Master

Proposition 1 (bone-balance sign theorem). Let $B$ denote the bone mass per unit anatomic region; let $N_{oc}$ and $N_{ob}$ denote the numbers of active osteoclasts and osteoblasts per unit bone-surface area; let $k_r$ and $k_f$ denote the per-cell resorption and formation rates; let $\alpha$ and ${\delta }_{oc}$ denote the pre-osteoclast differentiation rate and the mature-osteoclast apoptosis rate; let $R$ and $O$ denote the surface RANKL concentration and soluble OPG concentration; let $K_R$ and $K_O$ denote the RANK-RANKL and OPG-RANKL dissociation constants; and let $\beta$ denote the osteoclast-osteoblast coupling efficiency. Then

$$\frac{dB}{dt}=(k_f\beta -k_r)\cdot \frac{\alpha }{{\delta }_{oc}}\cdot \frac{R}{R+K_R(1+O/K_O)},$$

and in any region with active remodeling (positive RANKL signaling), $\mathrm{sign}(dB/dt)=\mathrm{sign}(k_f\beta -k_r)$.

Proof. Bone mass changes by formation minus resorption. At any instant,

$$\frac{dB}{dt}=k_f{N}_{ob}-k_r{N}_{oc}.$$

The osteoclast population at quasi-steady state satisfies $d{N}_{oc}/dt=\alpha S_{\mathrm{RANK}}-{\delta }_{oc}{N}_{oc}=0$, where $S_{\mathrm{RANK}}$ is the RANKL signaling strength given by competitive-inhibition kinetics (OPG as a soluble decoy competing with RANK for RANKL):

$$S_{\mathrm{RANK}}=\frac{R}{R+K_R(1+O/K_O)}\in (0,1).$$

So $N_{oc}^{\ast }=(\alpha /{\delta }_{oc})S_{\mathrm{RANK}}$. By the Rodan-Martin coupling hypothesis, osteoblast recruitment is proportional to active osteoclasts: $N_{ob}^{\ast }=\beta N_{oc}^{\ast }$, with $\beta$ the dimensionless coupling efficiency. Substituting,

$$\frac{dB}{dt}=k_f\beta N_{oc}^{\ast }-k_r{N}_{oc}^{\ast }=(k_f\beta -k_r)\cdot \frac{\alpha }{{\delta }_{oc}}\cdot \frac{R}{R+K_R(1+O/K_O)}.$$

The second and third factors are strictly positive in any region with active remodeling ($R>0$). Therefore $\mathrm{sign}(dB/dt)=\mathrm{sign}(k_f\beta -k_r)$, independent of the RANKL signaling strength. This is the bone-balance sign theorem: the sign of the bone-mass trajectory is set by the coupling-efficiency term alone, while the magnitude is modulated by RANKL-RANK-OPG signaling. $\square$

Corollary (menopausal sign-flip). In a healthy estrogen-replete adult, $k_f\beta \approx k_r$ and $dB/dt\approx 0$. Estrogen loss produces three independent tilts — $R\uparrow$ (RANKL transcription derepressed), $O\downarrow$ (OPG transcription unstimulated), and $\beta \downarrow$ (impaired osteoblast recruitment) — all in the direction of $k_f\beta -k_r<0$. The compounded effect is the rapid postmenopausal bone-loss rate of approximately 1 to 2 percent BMD per year, versus 0.3 to 0.5 percent per year in the estrogen-replete adult.

Proposition 2 (the mechanostat as a stable integral controller). Let $B$ denote the bone mass in a fixed anatomic region, $F$ the daily mechanical load imposed on that region, $E$ the elastic modulus, $A=cB$ the bone cross-sectional area (with $c$ a geometric constant), $\epsilon =F/(EA)$ the resulting strain, and ${\epsilon }_{\text{set}}$ the mechanostat strain set-point. If the bone-mass dynamics are $dB/dt=\gamma (\epsilon -{\epsilon }_{\text{set}})$ with $\gamma >0$ a remodeling-rate constant, then the steady-state bone mass is $B^{\ast }=F/(cE{\epsilon }_{\text{set}})$, the equilibrium is exponentially stable, and the time constant is $\tau =B^{\ast }/(\gamma {\epsilon }_{\text{set}})$.

Proof. Substituting $\epsilon =F/(EcB)$ into the mechanostat ODE,

$$\frac{dB}{dt}=\gamma \left(\frac{F}{EcB}-{\epsilon }_{\text{set}}\right).$$

At equilibrium $dB/dt=0$, so $F/(Ec{B}^{\ast })={\epsilon }_{\text{set}}$, i.e., $B^{\ast }=F/(cE{\epsilon }_{\text{set}})$. Linearise around $B^{\ast }$: write $B=B^{\ast }+\delta B$ with $\delta B$ small. Expand

$$\frac{F}{EcB}=\frac{F}{Ec{B}^{\ast }}\cdot \frac{1}{1+\delta B/B^{\ast }}\approx {\epsilon }_{\text{set}}(1-\delta B/B^{\ast }),$$

to first order in $\delta B/B^{\ast }$. Then

$$\frac{d(\delta B)}{dt}=\gamma {\epsilon }_{\text{set}}(-\delta B/B^{\ast })=-\frac{\gamma {\epsilon }_{\text{set}}}{B^{\ast }}\delta B,$$

with solution $\delta B(t)=\delta B(0)\exp (-t/\tau )$ and $\tau =B^{\ast }/(\gamma {\epsilon }_{\text{set}})$. The mechanostat is a first-order integral controller, stable for any $\gamma >0$. $\square$

The mechanostat therefore frames bone mass as the controlled variable of a feedback loop whose set-point is mechanical strain. Microgravity, bed rest, and spinal-cord injury lower $F$ and so lower $B^{\ast }$ (disuse osteoporosis). Resistance exercise, weight-bearing activity, and racket-arm sport raise $F$ and so raise $B^{\ast }$ (mechanical hypertrophy of bone). The set-point ${\epsilon }_{\text{set}}$ is approximately 1000 microstrain in healthy cortical bone, set by the osteocyte sclerostin-Wnt signaling that implements the mechanosensor (Theorem 8). Drugs targeting the mechanostat directly (anti-sclerostin mAbs) lower the effective ${\epsilon }_{\text{set}}$ by removing the Wnt brake, raising $B^{\ast }$ at constant load — the first bone-anabolic mechanism that does not depend on osteoclast suppression.

Connections Master

• Skeletal muscle physiology — excitation-contraction coupling 18.04.01. The chapter-opening unit on skeletal-muscle physiology supplies the load-bearing framework for the mechanostat: the daily mechanical strain on bone is set overwhelmingly by muscle force, not by gravity. The skeletal-muscle unit develops peak forces of roughly $30N/c{\mathrm{m}}^2$ of physiologic cross-section, and these forces, transmitted across the tendon-bone insertion, dominate the strain signal sensed by osteocytes. Sarcopenia (age-related muscle loss) produces disuse-pattern bone loss through the mechanostat exactly as bed rest does; conversely, resistance training builds bone mass in the loaded regions in proportion to the load increase. This unit supplies the cellular machinery (osteoclast-osteoblast-osteocyte) that converts the muscle-imposed strain into a remodeling response.

• Thyroid hormones and metabolic regulation — HPT axis 18.07.04. Bone remodeling and the HPT axis are the two cleanest instances of adaptive-tissue feedback control in vertebrate physiology. The HPT axis is the textbook endocrine loop: three tiers, one hormone pair, one receptor, logarithmic-gain feedback. Bone remodeling is the textbook mechanical-feedback loop: a cytokine-receptor-decoy module (RANKL-RANK-OPG) implementing the controlled gain, with osteocyte mechanosensing setting the set-point. The two systems are control-theoretic duals of one another: endocrine axis versus mechanical axis, hormonal gain versus RANKL/OPG-ratio gain. Thyroid hormone itself also acts directly on bone (T3 upregulates osteoclast and osteoblast activity; hyperthyroidism produces high-turnover osteoporosis and hypothyroidism produces low-turnover impaired growth), so the HPT unit is also a direct metabolic regulator of the bone-remodeling cells considered here.

• Chronic disease survey — cardiovascular, diabetes, cancer 35.03.01. Osteoporosis aggregates into the global chronic-disease burden alongside cardiovascular disease, type 2 diabetes, and cancer. Worldwide, approximately 200 million people have osteoporosis; one in three women and one in five men over fifty will sustain an osteoporotic fracture; roughly 8.9 million osteoporotic fractures occur annually. Hip fracture carries approximately 20 percent one-year mortality — comparable to many cancers — and the global cost of osteoporotic fracture care exceeds $\$40$ billion per year. The chronic-disease survey provides the population-level framing into which the cellular-molecular mechanism of this unit feeds.

• Cytochrome P450 pharmacogenomics and precision dosing 35.07.04. The CYP450 pharmacology framework is the canonical reference for hepatic metabolism and drug-drug-interaction prediction, and bisphosphonates are the instructive negative case that proves the rule. Nitrogen-containing bisphosphonates are not metabolised at all — they are not substrates for any CYP enzyme — and are excreted unchanged by the kidney via glomerular filtration. The bisphosphonate class thereby escapes both hepatic first-pass pharmacokinetics and CYP-mediated drug-drug interactions entirely. Denosumab, as a monoclonal antibody, is similarly outside CYP pharmacology (cleared by reticuloendothelial recycling via FcRn). The CYP-pharmacology unit provides the framework against which the bisphosphonate exception is defined; the contrast highlights how the route of administration and clearance mechanism determine a drug's interaction profile and dosing constraints.

Historical & philosophical context Master

The modern understanding of bone as a dynamically remodeling tissue controlled by a signaling axis was the work of a forty-year arc that began with a control-engineering hypothesis, ran through a cytokine-receptor-decoy discovery, and culminated in regulatory-approved targeted drugs. Harold Frost, an orthopedic surgeon at Henry Ford Hospital in Detroit, proposed in 1964 that bone remodeling obeys a negative-feedback control law in which the controlled variable is bone mass and the sensed variable is mechanical strain — what he called the mechanostat ^{[Frost 1964]}. Frost's hypothesis, developed in a series of papers and the 1964 monograph The Laws of Bone Structure, was a quantitative refinement of Julius Wolff's 1892 "law of bone transformation" (Das Gesetz der Transformation der Knochen), which had observed qualitatively that bone architecture adapts to mechanical load. Frost provided the set-point, the gain, and the four strain windows (disuse, adapted, mild overload, pathologic overload) that made the law operationally testable. The molecular identity of the mechanosensor remained open for forty years.

The cellular architecture of remodeling was clarified in 1981 by Greg Rodan and T. John Martin in a hypothesis paper in Endocrine Reviews ^{[Rodan-Martin 1981]}. They proposed that osteoclasts do not regulate their own differentiation but are controlled by paracrine signals from cells of the osteoblast lineage. The Rodan-Martin coupling hypothesis rested on the observation that every known stimulator of bone resorption (PTH, IL-1, TNF-$\alpha$, $1,25(\mathrm{OH}{)}_2{\mathrm{D}}_3$) acts on osteoblasts, not on mature osteoclasts, and that osteoblasts and osteoclasts are never found in direct physical contact. The hypothesis predicted that the osteoclast-differentiation factor would prove to be a protein of osteoblast origin — a prediction vindicated seventeen years later.

The molecular-clinical synthesis began in 1997 with the discovery of osteoprotegerin by William Simonet and colleagues at Amgen, cloning a novel secreted TNF-receptor-superfamily member whose transgenic overexpression produced osteopetrosis in mice and whose deletion produced osteoporosis ^{[Simonet 1997 OPG]}. OPG was identified as a soluble decoy receptor for an unknown ligand. The ligand was cloned in 1998 by two independent groups: Yasuda, Suda, and colleagues at Asahi Chemical Industry and Osaka University identified it from a stromal-cell library, showing identity with three previously-cloned immune-system proteins (TRANCE/RANKL/ODF), and named it osteoclast differentiation factor ^{[Yasuda 1998 RANKL]}; Lacey, Boyle, and colleagues at Amgen, working from the OPG-decoy side, identified its receptor on osteoclast precursors as RANK, a TNF-receptor-superfamily member, and demonstrated that RANKL activation of RANK was necessary and sufficient for osteoclastogenesis ^{[Lacey 1998 RANK]}. The RANKL-RANK-OPG axis was thus identified as a single cytokine-receptor-decoy system ancestral to the immune system and repurposed for bone. The 1997-1998 discovery was the foundation for the targeted-drug era.

The clinical-trial lineage that established osteoporosis as a pharmacologically tractable disease is anchored by three trials. The Fracture Intervention Trial (Black, Cummings, and the FIT Research Group, Lancet 1996) demonstrated that alendronate reduced hip fracture by 51 percent and clinical vertebral fracture by 47 percent in postmenopausal women with existing vertebral fractures ^{[Black 1996 FIT]}; it was the basis of the 1995 FDA approval of alendronate (Merck, Fosamax), the first nitrogen-containing bisphosphonate for osteoporosis and the drug that became the most-prescribed osteoporosis therapy worldwide. The HORIZON trial (Black, Delmas, Eastell, Reid et al., N. Engl. J. Med. 357:1799, 2007) extended the bisphosphonate class to once-yearly intravenous zoledronic acid (approved 2001 for hypercalcaemia of malignancy, 2007 for osteoporosis) with a 41 percent hip-fracture reduction and a 28 percent all-cause-mortality reduction in the hip-fracture cohort. The FREEDOM trial (Cummings, Martin, McClung, Siris et al., N. Engl. J. Med. 361:756, 2009) demonstrated that denosumab, a fully human monoclonal antibody against RANKL developed by Amgen, reduced vertebral fracture by 68 percent and hip fracture by 40 percent over three years ^{[Cummings 2009 FREEDOM]}; denosumab was FDA-approved in 2010 as the first osteoporosis drug whose mechanism targets a single named cytokine-receptor interaction.

The molecular identity of Frost's mechanostat was solved in the 2000s with the demonstration that osteocyte sclerostin, the product of the SOST gene (van Bezooijen, Papapoulos et al. J. Biomed. Mater. Res. 68A:324, 2004), is a secreted Wnt antagonist whose expression is suppressed by mechanical load, releasing canonical Wnt signaling to drive osteoblast bone formation ^{[van Bezooijen 2004 sclerostin]}. Loss-of-function mutations in SOST cause sclerosteosis (massive skeletal overgrowth), confirming sclerostin as the dominant tonic inhibitor of bone formation. Romosozumab, a monoclonal antibody against sclerostin developed by Amgen and UCB, became the first dual-action anabolic antiresorptive agent, FDA-approved in 2019 with a black-box warning for cardiovascular events. The same molecular framework — RANKL-RANK-OPG for the antiresorptive class, anti-sclerostin for the anabolic class — continues to anchor osteoporosis drug development.

Bibliography Master

Primary literature.

@article{Frost1964,
  author = {Frost, H. M.},
  title = {The Laws of Bone Structure},
  journal = {Charles C. Thomas, Springfield, IL},
  year = {1964},
}

@article{Frost1964Mechanostat,
  author = {Frost, H. M.},
  title = {Mathematical elements of lamellar bone remodeling as a self-controlled process},
  journal = {Henry Ford Hosp. Med. Bull.},
  volume = {12},
  year = {1964},
  pages = {499--511},
}

@article{RodanMartin1981,
  author = {Rodan, G. A. and Martin, T. J.},
  title = {Role of osteoblasts in hormonal control of bone resorption: a hypothesis},
  journal = {Endocr. Rev.},
  volume = {2},
  year = {1981},
  pages = {209--213},
}

@article{Simonet1997,
  author = {Simonet, W. S. and Lacey, D. L. and Dunstan, C. R. and Kelley, M. and Chang, M.-S. and Luthy, R. and Nguyen, H. Q. and Wooden, S. and Bennett, L. and Boone, T. and Shimamoto, G. and DeRose, M. and Elliott, R. and Colombero, A. and Tan, H.-L. and Trail, G. and Sullivan, J. and Davy, E. and Bucay, N. and Renshaw-Gegg, L. and Hughes, T. M. and Hill, D. and Pattison, W. and Campbell, P. and Sander, S. and Van, G. and Tarpley, J. and Derby, P. and Lee, R. and Amgen EST Team and Boyle, W. J.},
  title = {Osteoprotegerin: a novel secreted protein involved in the regulation of bone density},
  journal = {Cell},
  volume = {89},
  year = {1997},
  pages = {309--319},
}

@article{Yasuda1998,
  author = {Yasuda, H. and Shima, N. and Nakagawa, N. and Yamaguchi, K. and Kinosaki, M. and Mochizuki, S. and Tomoyasu, A. and Yano, K. and Goto, M. and Murakami, A. and Tsuda, E. and Morinaga, T. and Higashio, K. and Udagawa, N. and Takahashi, N. and Suda, T.},
  title = {Osteoclast differentiation factor is a ligand for osteoprotegerin/osteoclastogenesis-inhibitory factor and is identical to {TRANCE/RANKL}},
  journal = {Proc. Natl. Acad. Sci. USA},
  volume = {95},
  year = {1998},
  pages = {3597--3602},
}

@article{Yasuda1998Endo,
  author = {Yasuda, H. and Shima, N. and Nakagawa, N. and Mochizuki, S. and Yano, K. and Fujise, N. and Sato, Y. and Goto, M. and Yamaguchi, K. and Kuriyama, M. and Kanno, T. and Murakami, A. and Tsuda, E. and Morinaga, T. and Higashio, K.},
  title = {Identity of osteoclastogenesis inhibitory factor ({OCIF}) and osteoprotegerin ({OPG}): a mechanism by which {OPG/OCIF} inhibits osteoclastogenesis in vitro},
  journal = {Endocrinology},
  volume = {139},
  year = {1998},
  pages = {1329--1337},
}

@article{Lacey1998,
  author = {Lacey, D. L. and Timms, E. and Tan, H.-L. and Kelley, M. J. and Dunstan, C. R. and Burgess, T. and Elliott, R. and Colombero, A. and Elliott, G. and Scully, S. and Hsu, H. and Sullivan, J. and Hawkins, N. and Davy, E. and Capparelli, C. and Eli, A. and Qian, Y.-X. and Kaufman, S. and Sarosi, I. and Shalhoub, V. and Senaldi, G. and Guo, J. and Delaney, J. and Boyle, W. J.},
  title = {Osteoprotegerin ligand is a cytokine that regulates osteoclast differentiation and activation},
  journal = {Cell},
  volume = {93},
  year = {1998},
  pages = {165--176},
}

@article{Black1996FIT,
  author = {Black, D. M. and Cummings, S. R. and Karpf, D. B. and Cauley, J. A. and Thompson, D. E. and Nevitt, M. C. and Bauer, D. C. and Genant, H. K. and Haskell, W. L. and Marcus, R. and Ott, S. M. and Torner, J. C. and Quandt, S. A. and Reiss, T. F. and Ensrud, K. E.},
  title = {Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures},
  journal = {Lancet},
  volume = {348},
  year = {1996},
  pages = {1535--1541},
}

@article{Black2007HORIZON,
  author = {Black, D. M. and Delmas, P. D. and Eastell, R. and Reid, I. R. and Boonen, S. and Cauley, J. A. and Cosman, F. and Lakatos, P. and Leung, P. C. and Man, Z. and Mautalen, C. and Mesenbrink, P. and Hu, H. and Caminis, J. and Tong, K. and Rosario-Jansen, T. and Krasnow, J. and Hue, T. F. and Sellmeyer, D. and Eriksen, E. F. and Cummings, S. R.},
  title = {Once-yearly zoledronic acid for treatment of postmenopausal osteoporosis},
  journal = {N. Engl. J. Med.},
  volume = {356},
  year = {2007},
  pages = {1809--1822},
}

@article{Cummings2009FREEDOM,
  author = {Cummings, S. R. and Martin, J. S. and McClung, M. R. and Siris, E. S. and Eastell, R. and Reid, I. R. and Delmas, P. and Zoog, H. B. and Austin, M. and Wang, A. and Kutilek, S. and Adami, S. and Zanchetta, J. and Libanati, C. and Siddhanti, S. and Christiansen, C.},
  title = {Denosumab for prevention of fractures in postmenopausal women with osteoporosis},
  journal = {N. Engl. J. Med.},
  volume = {361},
  year = {2009},
  pages = {756--765},
}

@article{Russell2008,
  author = {Russell, R. G. G. and Watts, N. B. and Ebetino, F. H. and Rogers, M. J.},
  title = {Mechanisms of action of bisphosphonates: similarities and differences and their potential influence on clinical efficacy},
  journal = {Osteoporos. Int.},
  volume = {19},
  year = {2008},
  pages = {733--759},
}

@article{vanBezooijen2004,
  author = {van Bezooijen, R. L. and Roelen, B. A. and Visser, A. and van der Wee-Pals, L. and de Wilt, E. and Karperien, M. and Hamersma, H. and Papapoulos, S. E. and ten Dijke, P. and Lowik, C. W.},
  title = {Sclerostin is an osteocyte-expressed negative regulator of bone mass, but not a classical {BMP} antagonist},
  journal = {J. Exp. Med.},
  volume = {199},
  year = {2004},
  pages = {805--814},
}

@article{Robling2008,
  author = {Robling, A. G. and Niziolek, P. J. and Baldridge, L. A. and Condon, K. W. and Allen, M. R. and Alam, I. and Mantila, S. M. and Gluhak-Heinrich, J. and Bellido, T. and Harris, S. E. and Turner, C. H.},
  title = {Mechanical stimulation of bone in vivo reduces osteocyte expression of {Sost/sclerostin}},
  journal = {J. Biol. Chem.},
  volume = {283},
  year = {2008},
  pages = {5866--5875},
}

@article{Saag2017ARCH,
  author = {Saag, K. G. and Petersen, J. and Brandi, M. L. and Karaplis, A. C. and Lorentzon, M. and Thomas, T. and Maddox, J. S. and Fan, M. and Meisner, P. D. and Grauer, A.},
  title = {Romosozumab or alendronate for fracture prevention in women with osteoporosis},
  journal = {N. Engl. J. Med.},
  volume = {377},
  year = {2017},
  pages = {1417--1427},
}

@article{Lyles2007,
  author = {Lyles, K. W. and Col{\'o}n-Emeric, C. S. and Magaziner, J. S. and Adachi, J. D. and Pieper, C. F. and Mautalen, C. and Hyldstrup, L. and Recknor, C. and Nordsletten, L. and Moore, K. A. and Lavecchia, C. and Zhang, J. and Mesenbrink, P. and Hodgson, P. K. and Abrams, K. and Orloff, J. J. and Horowitz, Z. and Eriksen, E. F. and Boonen, S.},
  title = {Zoledronic acid and clinical fractures and mortality after hip fracture},
  journal = {N. Engl. J. Med.},
  volume = {357},
  year = {2007},
  pages = {1799--1809},
}

Textbook and monograph.

@book{BoronBoulpaep2017,
  author = {Boron, W. F. and Boulpaep, E. L.},
  title = {Medical Physiology},
  edition = {3rd},
  publisher = {Elsevier},
  year = {2017},
}

@book{RosenASBMR2019,
  author = {Rosen, C. J. and Bouxsein, M. L. and Compston, J. E. and Rosen, V.},
  title = {Primer on the Metabolic Bone Diseases and Disorders of Mineral Metabolism},
  edition = {9th},
  publisher = {Wiley-ASBMR},
  year = {2019},
}

@book{MarcusOsteoporosis2021,
  author = {Marcus, R. and Feldman, D. and Dempster, D. W. and Luckey, M. and Cauley, J. A. and Madore, M. A.},
  title = {Osteoporosis},
  edition = {5th},
  publisher = {Elsevier},
  year = {2021},
}

@book{GuytonHall2021,
  author = {Hall, J. E. and Hall, M. E.},
  title = {Guyton and Hall Textbook of Medical Physiology},
  edition = {14th},
  publisher = {Elsevier},
  year = {2021},
}