The integrity of the adult skeleton must be maintained by tightly coupled bone-remodelling processes: old bone is resorbed by cells called osteoclasts and is replaced by new bone, which is synthesized by cells called osteoblasts1. Osteoblasts become buried within the newly formed bone and morph into a third cell type, osteocytes. Osteoblasts and osteocytes secrete the signalling protein RANKL, which activates the RANK receptor on haematopoietic stem cells, triggering their differentiation into osteoclasts and so promoting osteoclast function2,3. But it is less clear how osteoclasts signal to osteoblasts to modify bone formation. In a paper in Nature, Ikebuchi et al.4 provide compelling evidence that osteoclasts regulate osteoblasts using the same RANKL–RANK system acting in reverse — in this setting, it is the RANKL ligand that modulates intracellular signalling, rather than the RANK receptor.
The authors first confirmed previous data5 indicating that osteoclasts release small extracellular vesicles that harbour RANK on their surface (Fig. 1a). They showed that overlaying a layer of isolated RANK-bearing vesicles on a layer of mouse osteoblasts in culture activated the expression of the differentiation-promoting genes Col1a1, Runx2 and Osx in the cells. The presence of RANK-containing vesicles also triggered mineral deposition by osteoblasts, both in vitro and in mouse skulls in vivo. Mineral deposition, indicative of bone formation, was promoted by the Runx2 protein.
Runx2 is known6 to be activated by a signalling pathway involving the proteins PI3K, Akt and mTOR. Indeed, Ikebuchi et al. found that Runx2 activation in mouse osteoblasts was blocked by the mTOR-inhibitor molecule rapamycin. The authors then showed that RANKL activates the PI3K–Akt–mTOR pathway through a region in its cytoplasmic tail that is rich in the amino acid proline; this proline-rich domain interacts with kinase enzymes that activate PI3K. Mutations in crucial proline residues within this domain abrogated the effects of RANK–RANKL reverse signalling.
Finally, Ikebuchi and colleagues established that RANKL monomers are not activated by RANK. Instead, crosslinking of individual RANKL proteins to produce multimers is a prerequisite for downstream signalling. The authors showed that genetically modified anti-RANKL antibodies containing structures called leucine zippers, which are known to induce trimer formation, could activate RANKL reverse signalling in osteoblasts (Fig. 1b).
These intriguing data reveal a plausible physiological explanation for why RANK would be released from osteoclasts as cargo in extracellular vesicles5; they also establish that RANKL can act as a signal-transducing molecule at the cell surface, rather than merely as a secreted signalling protein. More broadly, this is the first demonstration of signal transduction by vesicular cell-surface receptors interacting with membrane-bound ligands. Finally, and perhaps most importantly, the study provides a mechanism by which osteoclasts in the process of resorbing bone communicate with nearby osteoblasts to regulate the extent to which new bone is formed. It remains unclear whether RANK-containing vesicles can activate cell-surface RANKL on osteocytes, which might not be in close proximity to osteoclasts.
On the therapeutic front, the anti-RANKL antibody denosumab is widely used to prevent and treat both osteoporosis and skeletal problems caused by the spread of cancers to bone. By preventing RANKL–RANK forward signalling, denosumab inhibits bone resorption by osteoclasts. But, in doing so, it transiently lowers bone formation, because of the tight coupling between osteoclasts and osteoblasts7. Ikebuchi and colleagues’ newly engineered anti-RANKL antibody could potentially uncouple resorption and formation, and so be a more-effective alternative to denosumab. Notably, the authors found that, in mice whose ovaries had been removed (a model for post-menopausal osteoporosis), the antibody reduced bone resorption, but did not suppress bone formation.
It is difficult to speculate on whether Ikebuchi and co-workers’ anti-RANKL antibody would have untoward side effects. This is certainly possible, because RANK and RANKL have many roles elsewhere in the body. For instance, immune cells called T cells express RANKL, and RANK is found on dendritic cells, with which T cells interact to trigger immune responses; in this setting, the interaction between RANK and RANKL enhances T-cell immunity8. The proteins have also been identified in the brain, notably in some neurons and neuron-supporting cells, in brain-specific immune cells called microglia, and in brain tissue deprived of oxygen9. They have also been implicated as mediators of fever-related responses to infections10.
That said, a decade’s worth of clinical studies using denosumab has revealed no discernible effects on the immune system or the brain. There is also no evidence yet for reverse RANK signalling in immune cells or neurons. Nonetheless, given that nearly all nucleus-bearing cells release extracellular vesicles, further studies of RANK–RANKL signalling in non-skeletal tissues are imperative. Given nature’s propensity for reusing principles of intercellular signalling in many ways, the possibility of reverse signalling by other receptors must also be explored.
Nature 561, 180-181 (2018)