Homotypic RANK signaling differentially regulates proliferation, motility and cell survival in osteosarcoma and mammary epithelial cells

The RANK [receptor activator of nuclear factor kappa B (NF-κB)] signaling pathway, which is activated in response to its ligand, RANKL, has been shown to regulate many cellular processes in a diverse set of tissues. Most notably, RANK signaling is essential for bone homeostasis because it controls the differentiation and activation of bone-resorbing osteoclasts (Bucay et al., 1998; Dougall et al., 1999; Kong et al., 1999). RANK also plays a crucial function in immunity through promotion of dendritic cell survival (Anderson et al., 1997), induction of thymic epithelial progenitor cell differentiation (Lee et al., 2008) and lymph node development (Dougall et al., 1999; Kong et al., 1999). In the mammary gland, RANK regulates progesterone-mediated responses during pregnancy and the estrous cycle (Fata et al., 2000; Joshi et al., 2010). In these normal biological systems, RANK activation is facilitated through paracrine and/or heterotypic RANK–RANKL interactions that generate regulatory feedback mechanisms that help maintain normal physiology. It is therefore not surprising that deregulation of RANK signaling underlies the progression of multiple human diseases (Dougall et al., 2007; Leibbrandt and Penninger, 2008). The development of molecular therapies aimed at inhibiting the RANK pathway has been significant in the past decade, and therefore a better understanding of this pathway and the molecular mechanisms directed by RANK is imperative for best utilization of these evolving drugs.

In cancer, the importance of RANKL has been highlighted in multiple reports that demonstrate its role in regulating tumor cell metastasis to bone (Morony et al., 2001; Feeley et al., 2006; Armstrong et al., 2008; Canon et al., 2008). However, recent studies have provided new insights into the involvement of RANK signaling in tumorigenesis beyond that of metastasis. It has been shown that RANK signaling in mammary epithelial cells dictates cell proliferation (Cao et al., 2001; Fernandez-Valdivia et al., 2009), as well as mammary stem cell expansion during the reproductive cycle (Joshi et al., 2010). Consistent with these findings, mouse models of progestin-driven mammary tumorigenesis demonstrate a crucial function of RANK in breast cancer (Schramek et al., 2010; Gonzalez-Suarez et al., 2010). A role for RANK signaling in primary bone cancer has also been acknowledged, where aberrant production of RANKL by tumor cells leads to unregulated osteoclast-mediated bone destruction and increased tumor burden (Lamoureux et al., 2007). Through the generation of a transgenic mouse model of osteosarcoma (Molyneux et al., 2010), we have obtained evidence for RANKL involvement in primary bone tumorigenesis. Specifically, we found that dysregulated PKA signaling due to the loss of the Prkar1a bone tumor suppressor gene resulted in copious amounts of RANKL and greatly accelerated tumorigenesis. Furthermore, aberrant RANK expression and signaling has been noted in human (Mori et al., 2007) and mouse (Wittrant et al., 2006) osteosarcoma cell lines. Together, these studies open the possibility that RANK, signaling through an autocrine or homotypic mechanism, regulates tumorigenic phenotypes in tumor cells of both epithelial and mesenchymal origin.

The RANK receptor is a type I transmembrane protein that exists as a homotrimer at the cell surface and binds to RANKL, which is a type II membrane protein sharing close homology to the tumor necrosis superfamily (TNFSF) members FasL, TRAIL and TNF (Anderson et al., 1997; Wong et al., 1997; Lacey et al., 1998). RANK signal transduction is mediated by the TNF receptor-associated factor (TRAF) family of adaptor proteins, with TRAF6 being the best described in regulating the downstream effectors NF-κB, Akt (PKB) and the mitogen-activated protein kinases (MAPKs) JNK1, p38 (MAPK14) and Erk1/2 (MAPK2) in osteoclasts and mammary epithelial cells (Wong et al., 1998; Darnay et al., 1999; Kobayashi et al., 2001; Jones et al., 2006; Lamothe et al., 2007). Of these effectors, NF-κB is activated upon RANK stimulation during bone remodeling and in other tissue systems, and has become a hallmark of the RANK-signaling pathway. However, other signaling pathways modulated by TRAF6 could serve as crucial effectors of RANK signaling during tumorigenesis. Identifying the role of specific pathways downstream of RANK activation in tumor cells will fill an important knowledge gap, and might provide opportunities to combine existing RANK inhibiting therapies with drugs targeting crucial effectors downstream of RANK.

In this study, we interrogate the importance of aberrant RANK expression in osteosarcoma cells and test the significance of homotypic RANKL–RANK activation in directing specific tumorigenic parameters using RANK-expressing osteosarcoma cells derived from primary mouse bone tumors, as well as systems engineered to ectopically express RANK or RANKL. We show in osteosarcoma cells that homotypic RANK activity promotes cell motility and anchorage-independent viability, but not proliferation. Overexpression of RANK in non-tumorigenic osteoblastic and mammary epithelial cells further demonstrates that RANK signaling differentially regulates motility, proliferation, anchorage-independent viability and growth in these two cell types. In-depth analyses of the major signaling pathways and small-molecule inhibitors provide evidence of the involvement of Erk1/2 and Akt in mediating specific tumorigenic properties downstream of RANK activation. Thus, cell-autonomous RANK signaling leads to distinct advantages for mammary epithelial and mesenchymal osteoblastic cells through activation of Erk1/2.

To elucidate the role of homotypic and or autocrine RANK signaling in cancer, we took advantage of our recently reported transgenic mouse model of osteosarcoma designated as MOTO (Molyneux et al., 2010). Comparison of Rankl, Opg and Rank mRNA levels in normal bone, early osteosarcoma lesions and advanced osteosarcomas revealed elevated expression of Rankl and Rank at both early and late stages of bone tumor development (Fig. 1A). Specifically, Rank expression was significantly higher in 3 of 4 bones from 10-week-old MOTO mice; Rankl was elevated in all samples; whereas 3 of the 4 samples expressed significantly less Opg (Fig. 1A). In 28-week-old mice with advanced tumors, 8 or 12 of 13 tumors showed high Rankl and Rank levels, respectively (Fig. 1A), and 5 tumors within this cohort expressed Rank at ~50-fold higher levels than normal bone. Immunohistochemical analysis of RANK in MOTO tumors demonstrated specific expression of this protein in osteosarcoma cells, whereas RANK was absent in normal osteoblasts and osteocytes but was detectable in osteoclasts and macrophages residing in the bone marrow (Fig. 1A). Thus, MOTO osteosarcomas express high levels of Rank mRNA at initial stages of tumor development when no gross tumors are observed, as well as at more advanced stages of tumorigenesis. Furthermore, RANK protein is localized to osteosarcoma tumor cells and osteoclasts within the osteosarcoma tumors.

We have previously reported the expression status of Rankl and Opg in cell lines (designated Moto1.1, Moto1.2) derived from MOTO osteosarcomas (Molyneaux et al., 2010). In addition to these two cell lines, we analyzed Rank and Rankl expression in nine other MOTO-derived cell lines (designated Moto1.3–Moto1.10. and Moto2.1). We also included the immortalized but non-tumorigenic osteoblastic MC3T3-E1 and 7F2-OSB cells as controls in this analysis (Fig. 1B). RANKL was expressed at significantly higher levels relative to 7F2OSB osteoblasts in 4 of 11 MOTO cell lines (Moto1.2, Moto1.3, Moto1.4 and Moto2.1) whereas Rank mRNA was overexpressed in 5 of the 11 cell lines (Moto1.1, Moto1.2, Moto1.4, Moto1.5 and Moto2.1), indicating that the overexpression of these molecules is a common feature of osteosarcoma cells. MC3T3-E1 and 7F2-OSB cells express low amounts of Rankl and undetectable levels of Rank. RANK and RANKL protein expression was confirmed in Moto1.1 and Moto1.2 cell lines (Fig. 1B). Specifically, immunocytochemical analysis showed expression of RANK in both Moto1.1 and Moto1.2. In these cells, RANK protein displayed a diffuse cytoplasmic pattern that was more intense in Moto1.1 than in Moto1.2. MC3T3-E1 cells did not show any positive RANK signal. Further, immunoblot analysis of RANKL demonstrated elevated protein levels in Moto1.2 cells and low levels of RANKL in Moto1.1, 7F2OSB and MC3T3-E1 cells. These findings are consistent with the mRNA expression data of Rank and Rankl in these osteosarcoma cell lines.

We next established a system to study the effects of homotypic RANK signaling in normal osteoblasts by using gain- and loss-of-function strategies. We stably transduced the MC3T3-E1 and 7F2-OSB cell lines with a Rank overexpression vector generating MC3T3-pRANK and 7F2-pRANK cells. Overexpression of RANK in these cells was confirmed by immunoblotting using a RANK-specific antibody and by qRT-PCR analysis (Fig. 1C; supplementary material Fig. S1). To generate a loss-of-RANK scenario in MOTO osteosarcoma cells, Rank-specific shRNA knockdown expression constructs were stably transduced into Moto1.2 cells. The extent of knockdown was measured by qRT-PCR for four distinct Rank targeting constructs (pLKO.1-shRNA-2, -7, -8 and -10); pLKO.1-shRNA-7 was found to be the most effective and selected for loss-of-function experiments, whereas cells transduced with a non-specific sequence (pLKO.1) or GFP (pLKO.1-GFP) served as controls (Fig. 1D).

Because we have observed expression of both Rank and Rankl at early stages of tumor development in MOTO mice and in Moto osteosarcoma cells, we tested whether induction of RANK activity could regulate proliferation in Moto1.1 and Moto1.2. Increasing concentrations of recombinant mouse RANKL (RL) did not affect cell proliferation in either cell line assayed over 96 hours as shown in (supplementary material Fig. S1). Moto1.2 cells express endogenous RANKL; we therefore treated these cells with increasing concentrations of the RANKL inhibitor, OPG-Fc, which also failed to alter cell proliferation (supplementary material Fig. S1). Finally, stimulation of RANK-overexpressing MC3T3-E1 preosteoblasts and 7F2-OSB osteoblasts (MC3T3-pRANK and 7F2-pRANK) with RL did not affect cell proliferation (supplementary material Fig. S1). These studies show that RANK signaling does not regulate cell proliferation in either RANK-expressing osteosarcoma cells or in osteoblastic cells overexpressing RANK.

There has been increasing attention assigned to the role of RANKL in regulating cell motility (Jones et al., 2006; Mori et al., 2007; Armstrong et al., 2008; Chen et al., 2010; Hsu et al., 2010; Sabbota et al., 2010) and metastasis (Jones et al., 2006; Miller et al., 2008; Canon et al., 2008; Tan et al., 2011), with a significant emphasis on prostate and breast cancer cell metastasis to bone. Nonetheless, the role of RANK signaling through a homotypic scenario in the context of normal cell motility has yet to be investigated. Also, it is not clear whether co-expression of RANKL and RANK in cancer cells will promote their invasive capacity. We therefore measured the effect of RANK signaling on cell migration and invasion using Moto1.1 and Moto1.2 osteosarcoma cells, MC3T3-E1 preosteoblasts and 7F2-OSB osteoblasts. Comparison of baseline invasion of Moto1.1 and Moto1.2 cells showed that motility of both cell lines significantly differed (Fig. 2A). Specifically, Moto1.2 cells were more invasive than Moto1.1 cells. Upon treatment with RL, the invasive capacity of both tumor cell lines increased significantly; exogenous RL treatment in Moto1.1 cells lead to a ~sixfold increase in invasion, whereas this increase was twofold for Moto1.2 cells. This difference might arise in part from the presence of endogenous RANKL in Moto1.2 cells, resulting in basal RANK activity. Consistent with this, perturbation of endogenous RANKL by OPG-Fc in Moto1.2 cells significantly ablated cell migration and invasion as assessed by scratch-wound motility (data not shown) and Matrigel-coated Transwell invasion assays (Fig. 2B). Similarly, shRNA-mediated knockdown of endogenous RANK in Moto1.2 cells (pLKO.1-shRNA-7) inhibited cell invasion through both Matrigel and rat-tail collagen I (Fig. 2C). In non-tumorigenic MC3T3-E1 and 7F2-OSB cells overexpressing RANK (MC3T3-pRANK or 7F2-pRANK), treatment with RL promoted cell migration and invasion through collagen I matrix (Fig. 2D; supplementary material Fig. S2). Note that the baseline increase in migration and invasion in MC3T3-E1 cells was independent of RL treatment, suggesting that low levels of endogenous RANKL, leading to an autocrine or homotypic RANK-RANKL interaction, is sufficient to elicit a motile phenotype. These data clearly show that RANKL signaling induces motility in osteoblastic cells.

The role for RANK signaling in promoting dendritic cell and osteoclast survival has been well established (Dougall et al., 1999; Lum et al., 1999); however, its effects on mediating tumor cell survival or resistance to apoptosis is less clear. To address whether a RANKL stimulus could provide a survival mechanism in osteosarcoma cells, we performed anchorage-independent cell viability assays on Moto1.1, Moto1.2 and MC3T3-pRANK cells. Addition of RL to Moto1.1 and Moto1.2 osteosarcoma cells led to an increase in cell viability when cultured on ultra-low-attachment microplates over a 72 hour time course (Fig. 3A). Although the kinetics of cell viability varied among these two cell lines, RANK signaling provided a significant survival advantage to both Moto1.1 and Moto1.2 cells, whereas other genetic inherent differences between these cancer cell lines might account for these differences. Furthermore, treatment of MC3T3-pRANK cells with RL resulted in increased cell viability over the entire 72 hour time course; the histogram representing the 48 hour time-point highlights this observation. Even in RL-unstimulated MC3T3-pRANK cells, a baseline increase in anchorage-independent cell survival was evident, again suggesting the significance of endogenous RANKL in activating RANK through an homotypic mechanism.

One means by which cells facilitate anchorage-independent viability is through resistance to anoikis (Westhoff and Fulda, 2009). To investigate whether RANK signaling could provide a resistance mechanism to anchorage-independent cell death, we analyzed levels of active caspase-3 by immunoblot analysis of protein lysates obtained from Moto1.1 and Moto1.2 cells cultured for 48 hours on ultra-low-attachment plates. Moto1.1 and Moto1.2 cells showed lower cleaved caspase-3 levels following RL treatment, and this observation was more pronounced in Moto1.1 than in Moto1.2 cells (Fig. 3B). Intriguingly, MC3T3-pRANK cells cultured on ultra-low-attachment plates displayed a remarkable reduction in cleaved caspase-3 levels in the presence or absence of RL (Fig. 3B). These data strongly implicate RANK in providing a mechanism for resistance to anoikis in osteosarcoma and RANK-expressing osteoblastic cells.

Next, we performed a standard transformation assay by subjecting Moto1.2 pLKO.1-shRNA-7 and MC3T3-pRANK cells to soft-agar colony formation. Osteosarcoma Moto1.2 cells transduced with control vector (pLKO.1) grew numerous large colonies over 14 days and knockdown of endogenous RANK in pLKO.1-shRNA-7 cells significantly compromised their colony-forming ability (Fig. 3C). MC3T3-pRANK cells overexpressing RANK did not form colonies upon RL treatment, as assessed over 21 days in culture (data not shown). Together, these results show that constitutive RANK signaling in osteosarcoma cells provides a transformation advantage. However, the establishment of a RANK-signaling system in non-transformed osteoblastic cells is not sufficient to promote anchorage-independent growth and/or cell transformation, but does provide a survival advantage.

The above series of experiments identified that RANK regulates properties of cell transformation, i.e. cell motility and anchorage-independent viability of osteosarcoma and non-tumorigenic osteoblastic cells. These findings in mesenchymal cells prompted us to ask whether a RANKL–RANK homotypic mechanism would similarly impact epithelial cells. We selected non-transformed and non-tumorigenic mammary epithelial NMuMG cells, as counterparts to MC3T3-E1 cells, because the importance of RANK in mammary epithelial cell growth, differentiation and tumorigenesis is now evident. The widely used NMuMG cells have been derived from normal murine mammary epithelial tissue and retain characteristics of primary mammary epithelial cells (David et al., 1981; Soriano et al., 1995). Our measurements of the relevant molecules by qRT-PCR showed that NMuMG cells express Rank, the RANKL decoy inhibitor Opg, but not Rankl mRNA, relative to adult murine mammary gland (Fig. 4A). We then established a system to study homotypic RANK signaling. The mammalian expression vector (pcDNA3.1) containing a full-length mouse Rankl cDNA was stably transfected into NMuMG cells to overexpress RANKL (NMuMG-pRL). We verified ectopic RANKL expression by qRT-PCR using Rankl-specific Taqman primers and by immunoblot analysis using an anti-RANKL antibody (Fig. 4A).

The functional importance of RANK signaling in NMuMG cells was next tested using methods identical to those applied to the osteosarcoma cells. RANKL treatment has been reported to promote proliferation of mammary epithelial cells cultured ex vivo (Kim et al., 2006). Beyond mammary epithelium, recent evidence shows proliferative effects directed by RANKL in other tissues, including the hair follicle pilosebaceous gland (Duheron et al., 2011) and thymus (Lee et al., 2008). Addition of RL stimulated proliferation of NMuMG cells in a dose-dependent manner over 72 hours of culture (Fig. 4B). Furthermore, NMuMG-pRL cells overexpressing RANKL displayed a significant increase in proliferation at 48 and 72 hours compared with control NMuMG cells expressing β-galactosidase (NMuMG-pLacZ) (Fig. 4B). Treatment with OPG-Fc reversed the proliferative effect of RL in parental cells, and of transfected RANKL in NMuMG-pRL cells, thus demonstrating specificity of RANKL-induced mammary epithelial cell proliferation. This result was in contrast to that seen in osteosarcoma and RANK-overexpressing osteoblastic cells.

We next determined whether stimulation of RANK in mammary epithelial cells altered cell motility and invasion. Treatment of NMuMG cells with increasing concentrations of RL produced a negligible effect on cell migration or invasion (supplementary material Fig. S3). Consistent with this finding, the invasion of NMuMG-pRL cells through Matrigel did not differ from control cells (supplementary material Fig. S3). We conclude that homotypic activation of RANK in mammary epithelial cells does not alter migration or invasion.

The role of RANK signaling was finally tested in anchorage-independent viability and growth. Recombinant mouse RANKL treatment led to a significant increase in cell viability of NMuMG cells cultured on ultra-low-attachment plates over a 72 hour time course, whereas concomitant treatment with the inhibitor OPG-Fc reversed this effect (Fig. 4C). Intriguingly, RL treatment did not result in a reduced level of cleaved caspase-3 in NMuMG cells (Fig. 4D), which we observed in both Moto1.1 and Moto1.2 osteosarcoma cells and RANK-overexpressing MC3T3 preosteoblasts (Fig. 3B). However, subjecting NMuMG-pRL cells to soft-agar colony assays clearly demonstrated that overexpression of RANKL generated multiple colonies over 14 days (Fig. 4E). This growth was suppressed by treatment with OPG-Fc, suggesting that activation of RANK is sufficient to elicit cell transformation in mammary epithelial cells.

A number of signaling pathways have been implicated downstream of RANK activity. To identify the signal transduction pathways responsive to RANK stimulation in MOTO osteosarcoma, MC3T3-E1 and NMuMG cell lines, we performed immunoblot analysis for NF-κB p65, Akt, and the MAPKs Erk1/2, JNK and p38. Specifically, cells were serum starved and then stimulated with RL for 0, 5, 10, 15, 30 and 60, or up to 120 minutes. First, a time-course analysis following RL treatment of Moto1.2 cells showed a sharp peak of induction in phosphorylated Erk1/2, phosphorylated p38 and phosphorylated JNK within 5 minutes, whereas the phosphorylated p65 signal also increased but showed a gradual induction over 60 minutes. Treatment with RL had no effect on phosphorylated Akt. NMuMG cells displayed a few distinctions in the kinetics of RANKL response when compared to Moto1.2. Specifically, Erk1/2, JNK and p38 and Akt were all induced within 5 minutes following RL treatment. Interestingly, NF-κB, which is the most studied pathway downstream of RANK, showed a negligible induction as measured by phosphorylated p65 levels (Fig. 5A).

We next determined the consequence of de novo RANK activation in preosteoblasts and mammary epithelial cells. Comparison of MC3T3-pRANK with control MC3T3-pLVX cells showed that RL stimulation essentially captured the same repertoire of signal transduction responses described in above section, with the main difference being a gradual induction of phosphorylated Akt over 60 minutes post treatment (Fig. 5B). With respect to epithelial cells, NMuMG-pRL cells endogenously expressing RANK coupled with ectopic RANKL established an autocrine RANK signaling system. When cultured in low serum (1% FBS), we observed an elevated baseline of phosphorylated Akt, whereas levels of phosphorylated Erk1/2 and phosphorylated p65 remained unchanged, although phosphorylated Erk1/2 levels increased in response to OPG-Fc in NMuMG-pLacZ control cells (Fig. 5C). The specificity of RL in inducing these signaling proteins was determined through inhibition of RANK signaling by OPG-Fc. Addition of OPG-Fc dampened Akt phosphorylation in NMuMG cells whereas incorporating OPG-Fc treatment to MC3T3-pRANK cells negated activation of phosphorylated Erk1/2 (Fig. 5C; supplementary material Fig. S4). The signaling pathways found to be activated by RANKL are summarized in Fig. 5D.

Given the importance of Erk1/2 and Akt in regulating cell proliferation (Meloche and Pouysségur, 2007; Cheng et al., 2005) and our findings that these signaling pathways are activated by RANK in mammary epithelial cells, we examined their significance as downstream effectors of RANK-induced proliferation. The PI3K inhibitor wortmannin and MEK1 inhibitor U0126 were selected to block activation of Akt and Erk1/2 in NMuMG cells. The effectiveness of inhibition was verified by western blotting following RL treatment (supplementary material Fig. S5). We observed that RL stimulated NMuMG proliferation as expected, and the co-treatment of RANKL with either wortmannin or U0126 abrogated this induction, demonstrating that both pathways underlie RANK-mediated cell proliferation (Fig. 6A).

Next, we tested the importance of both Akt and Erk1/2 signaling pathways in RANK-regulated invasion using Moto1.2 cells. Matrigel-coated Transwell chamber assays were performed following RL treatment in combination with wortmannin, U0126 and OPG-Fc (Fig. 6B). The PI3K inhibitor wortmannin showed no effect on RANK-induced invasion through Matrigel, whereas the Erk1/2 inhibitor U0126 clearly reduced invasion; these two inhibitors exerted little effect in the absence of RL. Strikingly, OPG-Fc completely prevented invasion. Representative images of invaded cells at the 24 hour time point are depicted (Fig. 6B). These data show that Erk1/2, but not Akt, facilitates RANK-dependent cancer cell invasion, whereas the powerful inhibition of invasion by OPG-Fc indicates that additional mechanisms are recruited by RANK during the process of cell invasion.

Lastly, we sought to determine whether inhibition of Akt or Erk1/2 in NMuMG and Moto1.1 cells affected the pro-survival mechanism elicited by RANK. In both cell types, U0126 treatment in combination with RL significantly inhibited cell viability over the entire time course of anchorage-independent culture (Fig. 6C). Wortmannin, however, transiently inhibited RANKL-induced NMuMG survival and had no effect on Moto1.1 cells. Taken together, we conclude that the Erk1/2 MAPK pathway is the predominant signaling effector underlying RANKL-induced anchorage-independent cell survival.

In this work we have investigated the functional importance of RANK signaling in osteosarcoma cells, and also in non-tumorigenic mesenchymal and epithelial cell lines engineered to ectopically express RANK or RANKL. Our approaches were generally designed to study homotypic RANK signaling through ligand and receptor production by the same cell type, with some experiments extending to autocrine RANK signaling with cells expressing both ligand and receptor. This is in contrast to the well-described heterotypic interaction among RANKL and RANK expressed by distinct cell types. The gain- and loss-of-function studies demonstrate that RANK activation differentially impacts individual tumor cell characteristics of epithelial and osteoblastic origin, and identify the involvement of Erk1/2 as a major downstream effector of RANK. We also uncover the ability of RANK overexpression to impact cell transformation and anchorage-independent viability of MC3T3 preosteoblasts and NMuMG epithelial cells. This indicates that de novo RANK expression and activation might be sufficient for cell transformation. The implications of RANK signaling on tumorigenic characteristics through homotypic, autocrine and heterotypic (paracrine) mechanisms in osteosarcoma and mammary epithelial cells are modeled in Fig. 7.

The TNFR superfamily member RANK is widely expressed in a number of tissues, whereas its ligand has a more restricted expression pattern. RANKL is abundantly expressed in normal thymus, lymph node and lung, and at lower levels in spleen, bone marrow and mammary gland (Boyce and Xing, 2007). In autoimmune disorders such as osteoarthritis, osteoporosis and bone loss induced by cancer metastasis, RANKL is aberrantly produced by synovial cells (Nakano et al., 2004), activated T cells (Kotake et al., 2001) and bone stromal cells (Kitazawa and Kitazawa, 2002). In these degenerative diseases, paracrine RANKL–RANK signaling is recognized to have a pivotal role in osteoclast-mediated bone destruction. Consistent with this concept, we observed paracrine RANK-mediated effects when the Moto1.1 cell line was treated with exogenous RANKL, which provides a strong stimulus for cell motility and survival. Beyond this, we demonstrated homotypic RANK function. First, we found that primary osteosarcoma bone lesions expressed significant levels of both RANKL and RANK at early stages of tumorigenesis, and that this expression pattern becomes exaggerated in late-stage tumors. Second, we show that RANK localizes to osteoblastic tumor cells residing within osteosarcomas. Last, through the use of Moto1.2 cells that co-express RANK and RANKL and through strategies that inhibit endogenous RANKL (OPG-Fc and RANK shRNA), we confirm the importance of homotypic and autocrine RANK function. Similar effects of RANK activation on SaOS-2 human osteosarcoma cell line have been reported (Akiyama, et al., 2010). We propose that autocrine or homotypic RANK signaling is crucial for early stages of osteosarcoma development, and that this in conjunction with heterotypic signaling between osteoclasts, osteoblasts and tumor cells, operates during tumor progression. These complex interactions create a vicious cycle to accentuate bone tumor burden.

We further show that establishment of a RANK overexpression system in normal osteoblasts is able to enhance tumorigenic properties including migration and anchorage-independent survival but not soft-agar colony growth. These data suggest that RANK activation provides substantial advantages to cancer cells, but might not be sufficient for overt transformation of normal osteoblastic cells. However, clinical correlations have been made showing a positive relationship between RANK and RANKL expression, and outcome of human osteosarcoma patients, such that RANK-positive tumors exhibit a poor chemo-response and increased tumor burden (Mori et al., 2007; Lee et al., 2011), supporting a role for RANK signaling in primary bone tumorigenesis. Of notable interest is our observation that activation of RANK led to a greater cell survival advantage in NMuMG cells, as demonstrated by its dual ability to enhance anchorage-independent viability and soft-agar growth. This apparent distinction between the ability of RANK activity to promote a transformative phenotype in normal NMuMG, but not in MC3T3-E1 cells, was unexpected, but highly relevant. Gonzalez-Suarez and colleagues (Gonzalez-Suarez et al., 2007) showed that overexpression of RANK in the epithelial cell compartment of the mammary gland led to profound hyperplasia and impaired alveolar differentiation, although overt tumor formation did not occur. Recently, Santini and co-workers (Santini et al., 2011) provided a correlation between breast cancers expressing high levels of RANK and poor patient outcome. In this study, RANK was shown to be preferentially expressed in basal type breast cancers with a predisposition to metastasis. As mentioned previously, RANK has also been assigned a crucial role in promoting progestin-mediated breast cancer (Schramek et al., 2010; Gonzalez-Suarez et al., 2010). Taken together, the RANK signaling pathway is emerging as important in primary mammary tumorigenesis, although whether RANK or RANKL can be considered as proto-oncogenes is still premature. Further studies are required to elucidate whether aberrant expression of RANKL in primary, non-immortalized mammary epithelial cells is sufficient to initiate tumorigenesis.

At least seven signaling pathways are activated by RANK in osteoclasts; four directly mediate osteoclastogensis (IKK–NF-κB, JNK1, Myc, calcineurin–NFAT), two regulate osteoclast activity (Src–Akt, p38) and two promote osteoclast survival (Src–Akt, Erk1/2) (Boyce and Xing, 2007; Walsh and Choi, 2003). How these signaling molecules relate to osteoclasts is now well understood. The function of RANK in conditions other than bone homeostasis and immunity is exemplified in the mammary gland, where IKKa–NF-κB signaling has been shown to be a crucial downstream pathway in regulating epithelial cell proliferation (Cao et al., 2001; Schramek et al., 2010; Tan et al., 2011). Our thorough examination of the signaling pathways activated through RANK showed specific kinetics of phosphorylation for Erk1/2, Akt, JNK, p38 and p65–NF-κB. Because Akt and Erk1/2 family members have been widely implicated in tumorigenic parameters such as cell migration (Kim et al., 2001; Schäfer et al., 2004) and survival (Bao and Strömblad, 2004), even extending to some osteosarcoma cell lines (Wittrant et al., 2006; Mori et al., 2007; Akiyama et al., 2010), their significance was further probed with specific biochemical inhibitors. RANK activation in NMuMG cells resulted in increased cell proliferation, and inhibition of either Akt or Erk1/2 countered this effect. Furthermore, we found that the effects of RANK activation on osteosarcoma cell migration and survival were mediated through Erk1/2. Our data provide new evidence for the functional dependence of RANK on Erk1/2 signaling for regulating several tumor cell phenotypes.

Src is known to activate both Erk1/2 and Akt (Scheid and Woodgett, 2003; Walsh and Choi, 2003) and RANK stimulation leading to Src-mediated PI3K–Akt–NF-κB or Erk1/2 activation is facilitated by the adaptor protein TRAF6 (Wong et al., 1999; Xing et al., 2001). Activation of MAPKs and NF-κB through TRAF6 requires distinct protein-interacting regions (Kobayashi et al., 2001). For example, site-directed mutation of the first zinc-finger domain of TRAF6 abrogates NF-κB activation whereas transduction of MAPK remains unaffected. Thus, it seems that although RANK can stimulate both Erk1/2 and NF-κB, the exact mechanism diverges at the level of TRAF6, as modeled in Fig. 5D. Hence, RANK-directed effects might be channeled through distinct downstream signaling effectors depending upon the cell type.

This study highlights the advantages conferred by homotypic and autocrine RANK signaling on individual tumor cell types and delineates the downstream effectors of RANK signaling. RANK has been the focus of several recent studies, which define its fundamental importance in mammary tissue biology and mammary tumors. Moreover, RANKL has been shown to be regulated by the loss of a tumor suppressor in a genetic bone tumorigenesis model, beyond its deregulation in several human diseases. Accordingly, the broad potential of RANK therapy has garnered considerable pharmaceutical interest with the subsequent development of RANK inhibitors. These agents will achieve their optimal effects if we better understand cell-autonomous RANK signaling in addition to heterotypic interactions mediating its paracrine effects. This study provides new cellular and molecular insights into the causal effects of RANK on tumorigenic properties and identifies the MAPK pathways that can be targeted to curb pathologies linked to RANK signaling.

Mouse osteosarcoma tumor samples from 10- and 28-week-old MOTO (murine osteocalcin promoter-driven T-antigen osteosarcoma) transgenic mice were harvested and snap-frozen in liquid nitrogen before RNA or protein extraction. The MOTO transgenic mouse model of osteosarcoma has been described previously (Molyneux et al., 2010). Moto osteosarcoma cell lines Moto1.1, Moto1.2 and Moto2.1, propagated from MOTO tumors, have been described in detail elsewhere (Molyneux et al., 2010). The MOTO cell lines Moto1.3, Moto1.4, Moto1.5, Moto1.6, Moto1.7, Moto1.8, Moto1.9 and Moto1.10 were propagated from MOTO tumors as described (Molyneux et al., 2010). Mouse osteoblastic cells (7F2OSB, MC3T3-E1) and the mammary epithelial cell line (NMuMG) were obtained from ATCC (CRL-12557, CRL-2593, CRL-1636). Ongoing cultures were maintained in DMEM (Moto1.1, Moto1.2, NMuMG) or αMEM (7F2OSB, MC3T3-E1) containing 25 mM glucose, L-glutamine, antibiotics (100 U/ml penicillin, 100 μg/ml streptomycin) and supplemented with 10% FBS. Recombinant mouse RANKL (RL) and OPG-Fc were purchased from R&D Systems Proteins (Minneapolis, MN). The MEK1/2 inhibitor U0126 and PI3K inhibitor wortmannin were generously provided by Vuk Stambolic, Ontario Cancer Institute, Toronto, Canada. For cell-signaling experiments, cells were washed in PBS and starved in DMEM medium for 24 hours; cells were treated with 0, 10, 30 or 100 ng/ml recombinant mouse RANKL or 200 ng/ml OPG-Fc for the indicated times. In experiments using the pharmacological inhibitors U0126 (10 μM) or wortmannin (0.2 μM), cells were pre-treated in medium containing these inhibitors for 1 hour before RANKL stimulation; control cells were pre-treated in medium containing 1% DMSO.

Total RNA was prepared from MOTO bones (flushed to remove bone marrow, flash frozen and pulverized) or cell lines using TRIzol (Invitrogen, Carlsbad, CA). RNA purity was confirmed using a NanoDrop Spectrophotometer (Thermo Scientific, West Palm Beach, FL) and by running RNA on formaldehyde agarose gels to observe the integrity of 18S and 28S ribosomal RNA species. One microgram of RNA was reverse-transcribed using a first-strand cDNA synthesis kit (GE Healthcare, Pittsburgh, PA) and subjected to qPCR (ΔΔC_T) analysis, using an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems, Carlsbad, CA) and TaqMan gene expression assay mix containing unlabeled PCR primers and FAM-labeled TaqMan MGB probes (RANKL Mm01313944_g1, OPG Mm01205928_m1, RANK Mm01286484_m1, β-actin 4352933E). All raw data were analyzed using Sequence Detection System software version 2.1 (Applied Biosystems). The threshold cycle (C_T) values were used to calculate relative RNA expression levels. Values were normalized to endogenous β-actin transcripts.

Bone and osteosarcoma tissues were fixed in 10% formalin, decalcified with Immunocal (Decal Chemical, Tallman, NY), and sectioned at 5 μm. For cell immunocytochemical analysis, cells cultured on eight-well chamber slides were fixed in 4% paraformaldehyde. RANK immunostaining was performed following HRP-AEC System protocol (R&D Systems) with a few modifications. Specifically, antigen retrieval in tissue sections was achieved by Proteinase K (Dako, Carpinteria, CA) digestion for 10 minutes at 37°C followed by 7 minutes incubation at room temperature in a humidified chamber. Anti-RANK antibody (R&D Systems, clone AF692) diluted as 1:100 in 5% horse serum was incubated overnight at 4°C. Biotinylated horse anti-goat secondary antibody (BA9500 Vector Laboratories, Southfield, MI) diluted 1:1000 in PBS containing 2% mouse serum was incubated on sections at room temperature for 60 minutes.

Cells were washed in PBS and incubated in RIPA cell extraction buffer (20 mM Tris-HCl, pH 7.6, 1% Triton X-100, 0.1% SDS, 1% NP-40, 1% sodium deoxycholate, 5 mM EDTA, 50 mM NaCl) supplemented with 200 μM Na₃VO₄, 2 mM PMSF, and an appropriate dilution of Complete Mini, EDTA-free protease inhibition cocktail tablets (Roche, Indianapolis, IN), for 30 minutes. Protein concentrations were determined using a BCA kit (Pierce Chemicals, Rockford, IL). For immunoblotting, 30 μg of cell protein lysate was resolved by SDS-PAGE and transferred to nitrocellulose membranes. The membranes were probed using the following mouse antibodies against: FLAG M2 (Sigma), RANK (AF692; R&D Systems), RANKL (L300), phosphorylated JNK (p-JNK, Thr183/Tyr185), JNK, phosphorylated p44/42 MAPK (p-Erk1/2, Thr202/Tyr204), p44/42 MAPK (Erk1/2), phosphorylated p38 (p-p38, Thr180/Tyr182), p38 MAPK, phosphorylated Akt/PKB (p-Akt, Ser473), Akt, phosphorylated p65-NF-κB (Ser536), p65-NF-κB and cleaved caspase-3 (all from Cell Signaling Technology, Boston, MA). The blots were stripped and reprobed with an HRP-conjugated monoclonal antibody directed against mouse β-actin (Santa Cruz Biotechnology, Santa Cruz, CA).

A full-length mouse RANKL cDNA (GenBank Accession No. NM011613) was purchased from Origene (Rockville, MD) and cloned into the mammalian expression vector pcDNA3.1 at the NotI and BamHI restriction sites (Invitrogen, Carlsbad, CA). A clone (NMuMG-pRL) containing the RANKL cDNA in the forward orientation was subsequently identified by DNA sequencing. A pcDNA3.1 expression vector containing the β-galactosidase gene (pcDNA3-LacZ; Invitrogen) was used to determine transfection efficiency and served as a control for these studies. Stable transfections were performed to establish multiclonal NMuMG cells constitutively expressing RANKL. One μg/ml of plasmid DNA was transfected using Lipofectamine 2000 transfection reagent (Invitrogen) according to the manufacturer's protocol. Selection began 48 hours post transfection using 400 μg/ml G418 sulphate in DMEM; after selection, cells were maintained in medium containing 100 μg/ml G418 sulphate.

For the generation of a RANK overexpression system, a full-length mouse RANK cDNA construct (No. NM009399) was purchased from Origene and cloned into the lentiviral expression vector pLVX-puro (Clontech, Mountain View, CA). Generation of virus particles was achieved using the Lenti-X HT packaging system (Clontech) following the manufacturer's protocol. MC3T3 cells were infected with virus for 24 hours, after which cells were provided fresh medium and cultured for an additional 48 hours. At this point, transduced cells were selected in complete αMEM containing 2 μg/ml puromycin for 72 hours. After selection, cells were maintained in αMEM containing 10% FBS.

A set of four lentiviral pLKO.1 short hairpin (sh)-RNA expression vectors expressing shRNA sequences directed against mouse RANK under the control of the U6-promoter were generously provided by Jason Moffat, RNAi Consortium, The University of Toronto, Canada. These constructs were designated pLKO.1-shRNA-2, -7, -8 and -10. Vectors containing a scrambled shRNA (pLKO.1) or GFP (pLKO.1-GFP) insert served as controls. Lentiviral particles were generated using standard molecular biology procedures. Briefly, 293T packaging cells were seeded at 1.3×10⁵ cells/ml in 6 cm tissue culture plates containing DMEM supplemented with 10% FBS and antibiotics. After 24 hours, cells were transfected with the packaging plasmid pCMV-dR8.91 (Addgene; Cambridge, MA, USA), the envelope plasmid VSV-G/pMD2.G (Addgene) and one of the four RANK-directed pLKO.1-shRNA vectors. After 18 hours, cell medium was removed and replaced with fresh DMEM medium containing 10% FBS and cells were cultured for a further 24 hours, after which medium containing lentivirus was collected and used to infect target cells. Selection of transduced cells was performed as described above using Puromycin.

Cells were seeded at 1×10³ cells/well into opaque 96-well microplates containing 100 μl of DMEM with 5% FBS. Cells were cultured for 0, 24, 48, 72 or 96 hours in the presence of recombinant mouse RANKL, OPG-Fc, U0126 or wortmannin, after which cell proliferation was measured using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Sunnyvale, CA) following the manufacturer's instructions on a FLUOstar Optima plate reader (BMG; Offenburg, Germany). Soft-agar colony formation assays were performed by seeding cells in six-well culture plates within 0.35% agar containing DMEM medium supplemented with 10% FBS over a base layer of 0.5% agar dissolved in DMEM. Complete DMEM medium (2 ml) containing 10% FBS and supplemented with or without recombinant mouse RANKL or OPG-Fc was placed on top of agar cultures and replaced every 3 days. On days 14 or 21 after seeding, cells were fixed in 100% methanol and stained with 0.05% crystal violet. Colony formations were imaged using a digital camera mounted to a light microscope (25× objective).

Cell motility assays were performed using uncoated or coated (growth factor reduced Matrigel, rat-tail collagen I; BD Biosciences, San Diego, CA) Transwells fitted with Millipore membranes (BD Biosciences; 6.5 mm filters, 8 μm pore size) as described (Beristain et al., 2011). Briefly, 2×10⁴ cells/200 μl of DMEM supplemented with 1% BSA were plated in the upper chambers and cultured for the indicated times. Lower chambers contained 500 μl of DMEM supplemented with 10% FBS and both upper and lower chamber medium contained the indicated concentrations of recombinant RANKL or inhibitors. Cells from the upper surface of the Millipore membrane were removed by gentle swabbing, whereas transmigrated cells attached to the membrane were fixed in 4% paraformaldehyde and stained with eosin. The filters were rinsed with water, excised from the Transwells, and mounted upside down onto glass slides. Cell invasion was determined by counting the number of stained cells in 10 randomly selected, non-overlapping fields at 100× magnification using a light microscope. Cell invasion was tested in triplicate wells, on three independent occasions. For scratch-wound migration assays, Moto1.2 cells were cultured in 24-well plates and allowed to grow to confluency, when a wound was made across the well with a yellow pipette tip. The medium was then changed to DMEM containing 10% FBS with or without 200 ng/ml OPG-Fc. The wound was photographed immediately and again 20 hours after scraping to document cellular migration across the wound.

Anchorage-independent growth was achieved by culturing 5×10³ cells/well in ultra-low-adherence 96-well microplates (Costar, New York, NY) containing DMEM supplemented with 10% FBS for 0, 2, 8, 24, 48 or 72 hours. Cells were treated in combination with recombinant mouse RANKL, OPG-Fc or the inhibitors U0126 and wortmannin. Cell viability was measured using the CellTiter-Glo Luminescent Cell Viability Assay (Promega) according to the manufacturer's instructions. For analysis of cleaved caspase-3, cells were cultured for 48 hours on 10 cm ultra-low-attachment culture plates (Costar) in DMEM containing 10% FBS supplemented with or without recombinant mouse RANKL (30 ng/ml). After which, suspended cells were pelleted by centrifugation, washed in ice-cold PBS and subjected to RIPA buffer protein extraction.

Data are reported as mean ± s.d. or s.e.m. All calculations were carried out using GraphPad Prism software (San Diego, CA). Comparisons were made using two-tailed Student's t-test and ANOVA. Cellular invasion indices were analyzed by one-way ANOVA followed by the Tukey multiple comparison test. The differences were accepted as significant at P<0.05.

The authors thank Jason Moffat, for providing reagents and advice. The authors also thank Paul Waterhouse and Sam Molyneux for critical reading of the manuscript.