RANKL-Independent Osteoclastogenesis: A Long-Standing Controversy

In this issue of JBMR, Tsukasaki and colleagues1 challenged the concept of RANKL-independent osteoclastogenesis, which was recently reported in Nature.2 This report will interest the readers of JBMR because of the long-standing debate regarding RANKL-independent osteoclastogenesis. Coordinated regulation of osteoclasts and osteoblasts is required for maintaining the skeletal homeostasis, and its imbalance is observed in various diseases such as rheumatoid arthritis (RA), postmenopausal osteoporosis, and tumor-induced bone diseases. RANKL (encoded by the Tnfsf11 gene), a member of the tumor necrosis factor (TNF) family, is an essential cytokine for osteoclast differentiation. RANKL binds to its receptor RANK (encode by the Tnfrsf11a gene) expressed by osteoclast precursor cells and activates various intracellular signaling cascades.3, 4 The indispensable role of RANKL/RANK pathways was evidenced by genetic findings that mice with targeted disruption of either the Tnfsf11 or Tnfrsf11a gene exhibit severe osteopetrosis due to the complete absence of osteoclasts, and mutations in the TNFSF11 gene cause autosomal recessive osteopetrosis in humans.5, 6 These findings clearly support the idea that the RANKL/RANK axis is indispensable for osteoclastogenesis in both physiological and pathological conditions. Nevertheless, the possibility of RANKL-independent osteoclastogenesis has fascinated researchers for long time, in particular under some disease conditions such as RA, and a number of studies supporting the presence of RANKL-independent osteoclastogenesis have been published: TNF-α,7-9 APRIL, BAFF, NGF, IGF I, IGF II,10 LIGHT,11 TGF-β,12 IL-6, IL-11,13 and SOFAT14 have been proposed as RANKL-independent osteoclastogenic factors. TGF-β is a cytokine with ubiquitous proliferation and differentiation activity in many types of cells, and is abundantly stored in bone.15 Although the role of TGF-β in bone milieu has been extensively studied, it is still controversial whether TGF-β promotes or inhibits osteoclastogenesis. It was reported that TGF-β directly acts on bone marrow macrophages and promotes osteoclastogenesis, whereas it inhibits osteoclastogenesis indirectly through the effect on bone marrow stromal cells or osteoblasts.16-19 We recently reported that although the presence of TGF-β is indispensable for RANKL-induced osteoclastogenesis, TGF-β alone did not induce osteoclastogenesis.20 In addition, Smad2/3 activated by TGF-β associates with c-Fos and induces their nuclear translocation.21 Therefore, TGF-β is an essential cooperative factor for osteoclastogenesis, but cannot substitute for RANKL. TNF-α, which shares many common downstream signaling molecules with RANKL, is the most well-studied cytokine as a possible substitute for RANKL. Because TNF-α plays an essential role in the pathogenesis of RA, a direct role of TNF-α in bone destruction in arthritic conditions has been proposed in many studies. Several studies have shown that TNF-α alone induces osteoclast differentiation from murine bone marrow cells or human peripheral blood-derived monocytes.7-9 On the other hand, other studies opposed the direct induction of osteoclastogenesis by TNF-α.22-24 Lam and colleagues23 reported that TNF-α stimulated osteoclastogenesis when whole bone marrow cells were cultured. On the other hand, TNF-α failed to generate osteoclasts from a pure population of myeloid cells (positive for c-Fms, CD11b, RANK, and F4/80, but not for VCAM-1, a marker for bone marrow stromal cells). In addition, when saturating levels of osteoprotegerin were added in whole bone marrow cultures, TNF-α–induced osteoclastogenesis was completely abrogated, indicating that the effect of TNF-α on osteoclastogenesis is dependent on the RANKL/RANK axis. Based on these results, they concluded that TNF-α promotes osteoclastogenesis only in the presence of “permissive levels” of RANKL. Therefore, the conflicting observations in previous studies might be attributed to the different levels of RANKL expressed in the stromal cells contaminated in various culture systems. In addition, Yao and colleagues25 and Zhao and colleagues26 reported that TNF-α induces strong inhibitors of osteoclastogenesis, which significantly limits its osteoclastogenic activity. It is important to point out that recent clinical findings that the progression of bone erosion is suppressed in RA patients treated with a fully human anti-RANKL antibody, denosumab, presented definite evidence that RANKL plays an essential role in bone erosion in RA.27 Recently, Cox and colleagues2 proposed that lysyl oxidase (LOX), the collagen cross-linking enzyme, promotes breast cancer bone metastasis through RANKL-independent osteoclastogenesis. They showed that long-term treatment with LOX alone induced human osteoclasts with greater efficiency than RANKL. The effect of LOX was abrogated by the treatment with catalase, a reactive oxygen species (ROS)-neutralizing enzyme, suggesting that LOX-induced osteoclastogenesis is mediated by ROS production. The effect of LOX was considered RANKL-independent because soluble RANKL was not detected in the culture supernatant by ELISA assay. The concept that LOX directly stimulates osteoclast formation is very attractive because global quantitative analysis of the hypoxic secretome identified LOX as significantly associated with bone tropism and relapse. However, the concept was challenged by the study of Tsukasaki and colleagues,1 who tried to reproduce the capacity of LOX to induce osteoclastogenesis with the same experimental settings used in the paper by Cox and colleagues.2 Unexpectedly, LOX failed to induce osteoclast differentiation in the absence of RANKL although LOX promoted induction of endogenous RANKL expression in bone marrow stromal cells. To exclude the involvement of endogenous RANKL in LOX-induce osteoclastogenesis, they analyzed the effect of LOX using RANKL-deficient or RANK-deficient mice, and again, LOX did not induce osteoclastogenesis. Long-term treatment with LOX synergistically enhanced RANKL-induced osteoclastogenesis in cells derived from wild-type mice, but not in RANKL-deficient cells. LOX upregulates RANKL expression on bone marrow stromal cells via the production of ROS. Finally, in vivo effects of LOX were examined using RANKL-deficient mice, and LOX injection did not rescue the osteopetrotic phenotype of RANKL-deficient mice. Based on these findings, they concluded that LOX stimulates osteoclastogenesis only in the presence of RANKL. Similar to the controversy in TNF-α–induced osteoclastogenesis, RANKL expression in contaminated stromal cells might explain the discrepancy between the two studies. Cox and colleagues2 proposed a model in which LOX directly induces osteoclastogenesis in a RANKL-independent manner, whereas Tsukasaki and colleagues1 showed that LOX indirectly promotes osteoclastogenesis via upregulation of RANKL expression on stromal cells (Fig. 1). However, it remains unclear why endogenous RANKL induced by LOX in stromal cells could not support osteoclastogenesis in the culture system used by Tsukasaki and colleagues.1 The molecules proposed as RANKL-independent osteoclastogenic factors (TNF-α, LOX, etc.) are considered to play important roles in bone destruction in cooperation with RANKL under pathological conditions such as RA and cancer metastasis, and thus can be therapeutic targets. To elucidate the precise mechanism underlying the osteoclastogenesis in these disease conditions, it is important to clarify whether the molecules exert the osteoclastogenic activity independently of RANKL or just synergize with RANKL. Because the involvement of stromal cells is not negligible, as Tsukasaki and colleagues1 and previous studies reported,23, 28, 29 it is difficult to propose RANKL-independent osteoclastogenesis without the validation in RANKL-deficient or RANK-deficient conditions. However, it should be noted that Li and colleagues30 reported that injection of TNF-α in the calvaria induced a small number of osteoclasts in RANK knockout mice, indicating the possibility that high local concentrations of proresorptive factors such as TNF-α can induce osteoclastogenesis even in the absence of RANK signaling. However, they also showed that TNF-α injection failed to induce hypercalcemia or radiographic signs of bone resorption as observed in wild-type mice,30 and Zhao and colleagues26 reported that TNF-α induced osteoclastogenesis in RANK-deficient mice in the absence of recombinant recognition sequence binding protein at the Jκ site (RBP-J). These results suggest that TNF-α can induce osteoclastogenesis only under limited circumstances. Most recently, Reynaud and colleagues31 reported that LOX promotes colorectal cancer bone metastasis by accelerating RANKL-induced osteoclastogenesis. Consistent with the observations by Tsukasaki and colleagues,1 they also showed that LOX alone failed to induce osteoclastogenesis, but enhanced it in combination with RANKL. These results further supported that LOX is not a substitute for RANKL but a cooperative factor of RANKL-dependent osteoclastogenesis. Needless to say, these studies never deny the importance of LOX in bone destruction associated with cancer metastasis. It would be interesting to explore whether blocking both LOX and RANKL more efficiently inhibits bone metastasis than LOX alone. The controversy reminds us how difficult it is to prove RANKL-independent osteoclastogenesis in a strict way and so far there seems to be no evidence that proves a molecule replacing RANKL. The author states that he has no conflicts of interest.