No Cytokine on This Pathway Touches Bone

Not one immune cytokine on this pathway touches bone. TNF, IL-1β, IL-6 and IL-17A all act on osteoblasts, osteocytes and synovial fibroblasts, and it is those stromal cells that decide between RANKL and its soluble decoy osteoprotegerin. Everything below that decision — the precursor, the commitment, the sealing zone, the acid, the proteases — follows from a ratio set by cells that are not part of the immune system at all. Getting that architecture the right way round is the difference between an experiment that answers a question and one that measures a marker.

Osteoimmunology and the RANKL–OPG axis — 35 clickable nodes from the inflammatory and T-cell input through the stromal decision, osteoclast commitment, the resorption machinery and the formation arm. Open the interactive version to click any protein through to its ELISA kit, biosimilar or In Vivo antibody.

The immune system never talks to bone directly

IL-17A has no receptor on the osteoclast. What it does is induce RANKL on osteoblasts and synoviocytes, while separately recruiting neutrophils. IL-6 does the same, which is why tocilizumab slows radiographic progression rather than only relieving symptoms — an effect on structure that follows from an effect on a stromal cell. IL-23 sustains the Th17 population that supplies IL-17A, and it is the sharpest counterexample in the field: in the spondyloarthritides the IL-23 arm associates with new bone formation rather than erosion. Inflammation does not have a single direction of effect on bone.

TNF is the partial exception and the most therapeutically consequential one. It raises RANKL on stromal cells and it also acts on the osteoclast precursor directly, lowering the RANKL threshold at which that precursor commits. That direct contribution is the most plausible explanation for why anti-TNF treatment halts erosion faster than it resolves synovitis — two effects with different kinetics from one molecule. IL-1β acts later still, extending osteoclast survival after RANKL has already primed the cell rather than initiating the lineage, and pairing the IL-1β antibody with the IL-1R1 antibody separates ligand neutralisation from receptor blockade on the same axis.

Only the ratio means anything

Osteoprotegerin is a soluble decoy receptor for RANKL, and it is made by the same osteoblast lineage that makes RANKL itself. Neither concentration predicts turnover on its own. Reporting one without the other is the commonest error in this literature, and it is not a minor one — a rising RANKL with a rising OPG is a different biology from a rising RANKL with a flat one.

There is a second requirement that the field's shorthand quietly omits. RANKL cannot make an osteoclast unaided: M-CSF must first keep the precursor alive and induce RANK on it. Two signals, in a required order — which is why blocking CD115 and blocking RANKL are different experiments with different meanings. One removes the precursor pool before RANKL is ever relevant; the other removes the instruction. If the question is precursor supply, CSF1R blockade is the cleaner intervention.

RANKL itself comes from four sources with different regulation — osteoblasts, osteocytes, activated T cells and synovial fibroblasts — which is the structural reason blocking any single upstream cytokine only ever partly works. It is also why TGF-β, released from the bone matrix as it is resorbed, matters: the act of destroying bone liberates the factor that recruits osteoblasts to rebuild it. That is the coupling signal, and uncoupling it is what inflammation does.

The T cell is not uniformly pro-erosive

Th17 cells drive the axis, and RORγt is the transcription factor that makes a helper T cell osteoclastogenic — Th1 and Th2 are not. But IFN-γ from Th1 cells drives TRAF6 degradation and actively opposes osteoclastogenesis, which means depleting CD4 T cells wholesale can move bone loss in either direction depending on the model. That ambiguity is a reason to include IFN-γ neutralisation as a control in any T-cell-dependent erosion experiment rather than to avoid the depletion.

CTLA-4 is the most interesting node on the T-cell side because its effect is not a T-cell effect at all. CTLA-4-Ig binds CD80 and CD86 on the osteoclast precursor itself and suppresses differentiation directly, independently of any T cell — which is the mechanism behind abatacept's effect on bone. Three functional-grade clones are stocked, so the T-cell-mediated and precursor-direct contributions can be separated rather than conflated.

Commitment is a switch, not a dial

The osteoclast precursor is a myeloid cell marked by CD115 and CD11b, and until commitment it is a macrophage. c-Fos is the switch: c-Fos-deficient mice make macrophages where osteoclasts should be, from the same precursor under the same signals. That is about as clean a statement as developmental biology offers that this is a lineage decision rather than a graded activation state.

Below RANK the adaptor is TRAF6, and it is an unusually clean single point of failure in a pathway otherwise full of redundancy — TRAF6-deficient mice are osteopetrotic and no other TRAF substitutes. NF-κB follows, and it is necessary and completely non-specific: it is shared with every inflammatory receptor on the same cell, so an NF-κB read-out cannot tell you that a precursor is becoming an osteoclast rather than an activated macrophage.

NFATc1 is the master regulator and it autoamplifies — it induces its own promoter, so past a threshold the programme completes whether or not RANKL is still present. This is the mechanistic reason anti-resorptive treatment prevents new osteoclasts rather than switching off existing ones, and why the clinical effect lags the pharmacology by weeks. The calcium flux NFATc1 requires comes from OSCAR, a co-stimulatory receptor signalling through ITAM adaptors — co-stimulation, on a cell with no antigen receptor, borrowed wholesale from immune signalling. That borrowing is what makes this field osteoimmunology rather than bone biology.

Acid first, protease second

The machinery below commitment is mechanical, and the order matters. Integrin αvβ3 forms the sealing zone that isolates a patch of bone surface from the extracellular space; without it the acid and the proteases disperse instead of concentrating, and resorption fails despite a fully differentiated osteoclast. Osteopontin in the sealing zone is both the integrin ligand and, separately, a macrophage-recruiting cytokine — a structural protein and an immune signal in one molecule, which is this pathway in miniature.

Carbonic anhydrase II then supplies the protons that dissolve the mineral, and only after that does cathepsin K reach the collagen. The sequence is not a detail: carbonic anhydrase II deficiency causes osteopetrosis just as cathepsin K deficiency does, which locates the demineralisation step precisely. MMP-9 works alongside cathepsin K and lets the cell migrate through the matrix it has degraded — though in an inflamed joint it is also made by every infiltrating neutrophil, making it the least osteoclast-specific marker available.

Erosion is uncoupling, not just resorption

This is the point most worth carrying away. TNF raises RANKL and induces DKK1 at the same time, and DKK1 suppresses RUNX2, the osteoblast master regulator. So the Wnt-dependent repair arm is switched off in the same stroke that accelerates the resorptive one. That uncoupling, rather than the resorption itself, is what makes inflammatory bone loss permanent — and it is why an anti-resorptive alone halts erosion without restoring what was lost.

Sclerostin is the other brake, made by osteocytes to restrain Wnt signalling. Mechanical loading suppresses it, which is how the osteocyte translates strain into bone mass, and romosozumab works by blocking it. Between them, sclerostin and DKK1 are the reason the formation arm deserves its own band on this map rather than a footnote.

Which reagents this map is made of

Eleven of the thirty-five nodes carry a functional-grade antibody, and on this pathway they sit exactly where the decisions are made rather than scattered across it. RANKL is blockable in both species: a functional-grade anti-mouse RANKL antibody for models, and the VivoGenie anti-human RANKL antibody — the denosumab sequence in functional grade — for human systems. That makes this one of the few maps in the library where the intervention and the mechanism are the same molecule.

Around it: CD115 and CD11b on the precursor; TNF, stocked ultra-low endotoxin, which here is a requirement rather than a refinement because LPS drives osteoclastogenesis through TLR4 on its own and would be scored as the cytokine's effect; IL-1β and IL-1R1 as a ligand-versus-receptor pair; and CD4, IFN-γ, CTLA-4 and pan-TGF-β on the T-cell and coupling side.

Said plainly, what is not blockable in this range: IL-6, IL-17A, IL-23, OPG, M-CSF, RANK itself, sclerostin and DKK1. What compensates is that the measurement layer is unusually complete — cathepsin K, TRAP, MMP-9, CAII, osteopontin, CTX-I, RUNX2, bone ALP and osteocalcin. The design this range supports is therefore to intervene above the membrane and read out below it, which on a pathway where the decision is made by a stromal cell and executed by a myeloid one is exactly the right division of labour.

Two related maps in this library are worth reading alongside this one and are not redrawn here: the IL-6 axis, and the TNF receptor superfamily — RANK is a TNFRSF member and behaves like one, including in its dependence on receptor clustering. NFATc1 and the commitment machinery below it appear here as the bone-specific consequence rather than as a second treatment of NF-κB signalling.

For research use only; not for use in diagnostic or therapeutic procedures. Explore the full library of interactive pathway diagrams.