Metastasis is the most relevant prognostic factor for survival in neuroblastoma patients, and bone is one of the target organs of metastasis in advanced neuroblastoma. Patients with bone metastasis are classified as stage IV, regardless of other clinical manifestations, since the presence of bone/bone marrow metastasis is responsible for poor prognosis despite high-dose chemotherapy and autologous hematopoietic stem cell transplantation.1 One possible reason for this ominous prognosis is the difficulty in reaching cancer cells that have colonized the bone/bone marrow microenvironment. This suggests the need for effective therapeutic modalities for patients with skeletal metastases based on a precise understanding of the pathophysiology of bone colonization in neuroblastoma. Two classical theories are generally advocated to explain the invasion of tumor cells in bone, i.e., the soil hypothesis2 and the circulation theory.3 The first suggests that proper factors favor the implantation of metastatic cells, including nutrients, oxygen tension, hormonal environment and hydrogen ion concentrations. The second assumes that development of cancer metastases is dependent on the blood volume inside a target organ. This hemodynamic theory alone cannot explain the high frequency of skeletal metastasis from some neoplasms because blood flow in bone is very poor. Thus, tumor cells may possess some peculiar properties that make them able to colonize and resorb bone.
Clinical and experimental evidence has shown that metastatic osteolysis depends on osteoclasts.4 Some tumors, including those of the breast,5 release parathyroid hormone–related peptide, which selectively favors tumor growth in bone. Moreover, cytokines and growth factors, matrix metalloprotease activity and neoangiogenesis are also involved in osteolysis.6, 7, 8, 9
A cytokine system belonging to the TNF family has been identified and extensively characterized. This system is capable of regulating proliferation, differentiation, fusion, activation and apoptosis of osteoclasts.10, 11, 12 This cytokine network includes RANKL, which exists in 2 forms: a cell membrane-bound variant and a secreted, soluble variant, which is generated by enzymatic cleavage of the membrane protein. RANKL binds 2 types of receptor: the first, known as “osteoclast differentiation and activation receptor” or RANK, is expressed in preosteoclasts and its binding triggers the cascade of intracellular events essential to complete osteoclastic differentiation and activation. The second is OPG, also reported to be an osteoclastogenesis inhibitory factor, which is a decoy receptor that limits the biologic activity of RANKL. OPG suppresses the differentiation of osteoclasts, inhibits their activation and induces apoptosis. Our current understanding of the pathogenesis of bone metastasis is that various osteotropic malignancies use RANKL-mediated mechanisms to trigger osteoclastic bone resorption.13, 14, 15 Michigami et al.16 found that cell interaction between neuroblastoma cells and bone marrow of immunocompromised mice induced RANKL expression in host cells, leading to osteoclast formation and bone metastasis. However, apart from the above-mentioned work, the molecular mechanisms responsible for metastatic osteolysis in neuroblastoma have been poorly investigated, and the exact nature of neuroblastoma–bone interactions remains largely unknown.
In the current study, we tested the hypothesis that neuroblastoma cells are able to induce osteoclastogenesis as they possess some of the mechanisms by which osteoblasts regulate osteoclastogenesis. We focused our attention on the capability of tumor cells to influence directly the RANKL-to-OPG ratio, which could be the peculiarity that enables neuroblastoma cells to colonize bone and induce osteolysis. Our aims were as follows: (i) to investigate if different neuroblastoma cell lines produce RANKL and OPG, (ii) to establish which form of RANKL is produced by neuroblastoma cell lines, (iii) to demonstrate that neuroblastoma cell lines are able to induce osteoclastic differentiation and (iv) to identify one or more molecules that revoke the osteoclastic phenotype induced by neuroblastoma cells.
CHP-212, SJ-N-KP and NB-100 cell lines, derived from primary neuroblastomas,17 and SH-SY5Y, derived from bone marrow metastasis of neuroblastoma, were used. Cells were maintained in MEM supplemented with 10% FCS (Life Technologies, Grand Island, NY), antibiotics (Life Technologies) and L-glutamine (2 mM). The human preosteoclastic FLG29.1 cell line was used in coculture experiments because its reactivity to RANKL, as well as to RANKL plus OPG, has been well recognized.18 FLG29.1 cells were cultured in RPMI-1640 (GIBCO BRL, Gaithersburg, MD) supplemented with 10% FCS (GIBCO BRL) and 100 μg/ml of gentamicin (Sigma, St. Louis, MO). Cells were maintained at 37°C in a 5% CO2 humidified atmosphere.
Osteoclast-like cell formation assay
FLG29.1 cells were seeded in 12-well culture plates or in chamber slides at a 0.5 × 106/ml density onto a semiconfluent layer of neuroblastoma cells (SH-SY5Y and SJ-N-KP). As a negative control, cells were resuspended in fresh culture medium and placed at equal density. As a positive control, FLG29.1 cells were induced to differentiate by addition of PMA (Sigma) at a final concentration of 0.1 μM. After 48 hr, the suspension-growing FLG29.1 and neuroblastoma cells were collected separately. To examine if the culture medium of neuroblastoma cell lines favored osteoclastogenesis, osteoclast precursors were obtained from PBMCs of healthy donors following the methods described by Taranta et al.19 Briefly, PBMCs were plated onto chamber slides (3 × 106/cm2) and incubated at 37°C. After 1 hr, nonadherent cells were removed and fresh medium containing 25% of supernatant of confluent SH-SY5Y was added. Cultures were maintained for 7 days. As a negative control, cells were cultured with fresh culture medium only; as a positive control, 25 ng/ml of recombinant macrophage colony-stimulating factor (Sigma), 30 ng/ml of human recombinant RANKL (Preprotech, London, UK) and 100 nmol/l of parathyroid hormone (Sigma) were added to the culture medium. At the end points, FLG29.1 cells and PBMCs were analyzed for TRAP activity using a commercial kit (Sigma): TRAP-positive FLG29.1 cells and PBMCs containing 2 or more nuclei were considered to be osteoclast-like cells.
Treatment of SH-SY5Y neuroblastoma cells with biologic modifiers of RANKL activity
Two ODNs, phosphorothioate in the first and the last base, were synthesized to complement specific sequences encoded by RANKL mRNA (TIB MolBiol, Genoa, Italy). The 22 mer ODN1 (5′-CTCTGCTGGCGCGGCGCATGGC-3′) was designed to block the transmembrane region of RANKL variant 1 mRNA (residues 154–175, Genbank NM003701.2).
The 26 mer ODN3 (5′-GCAGTGAGTGCCATCTTCTGATATTC-3′) was designed to complement the RANKL ectodomain (residues 389–414 of RANKL variant 1 and 108–133 of RANKL variant 2 mRNA, Genbank NM 033012.2) in the “stalk” region. An 18 mer fully degenerated ODN (dODN, 5′-NNNNNNNNNNNNNNNNNN-3′, where N was G, C, A or T) was used as the most suitable control for non-antisense effects.20 An artificial cationic lipid, DOTAP, was used to enhance both uptake and stability of ODNs in the cells. Before use, ODNs were preincubated at 37°C for 15 min with 10 μM DOTAP (Liposomal Transfection Reagent; Roche, Mannheim, Germany).21
Preliminary experiments were conducted to investigate the effect on neuroblastoma cells, and 10 μM concentration was chosen because it was the highest concentration with no toxic effect. SH-SY5Y neuroblastoma cells (0.2 × 105/ml) were seeded in Petri dishes and ODNs added after 24 hr of culture. After 3 days, half a dose of the antisense was added to compensate for the decrease in activity. After the second administration, 0.25 × 106/ml FLG29.1 cells were placed onto a semiconfluent layer of SH-SY5Y neuroblastoma cells. After 48 hr, FLG29.1 and adherent neuroblastoma cells were collected separately, counted in Bürker's chambers and used for extraction of total RNA and proteins. Cytospins were prepared to evaluate TRAP activity and supernatants collected and stored at –20°C until use.
A chimeric form of rhOPG/Fc was employed to neutralize in vitro the activity of RANKL (Recombinant Human OPG/TNFRSF11B/Fc Chimera 805-OS-100; R&D Systems, Minneapolis, MN). Preliminary experiments were conducted to investigate the effect of OPG/Fc on neuroblastoma and FLG29.1 cells. Serial dilutions between 100 and 0.1 ng/ml neither stimulated cell growth nor showed toxic effects. According to the previous experience of Simonet et al.,22 a 100 ng/ml concentration of OPG/Fc protein was used in coculture experiments as this inhibited completely osteoclastogenesis in vitro. SH-SY5Y (0.5 × 105/ml) was seeded in a culture flask: after 24 hr, culture medium was changed and 0.25 × 106 FLG29.1 cells plus 100 ng/ml of rhOPG/Fc were added. Cocultures were maintained for 48 hr and treated as described above. rhOPG/Fc (100 ng/ml) was added to the culture medium of adherent PBMCs challenged with 25% of SH-SY5Y supernatant, and the effects on osteoclast formation were seen after 7 days.
A polyclonal antibody antihuman sRANKL was employed to neutralize in vitro the activity of RANKL (goat antihuman sRANKL, Preprotech). A concentration range of 0.4–0.9 μg/ml neutralizes the biologic activity of sRANKL. As preliminary experiments showed that serial dilutions between 0.5 and 1 μg/ml neither stimulated cell growth nor showed toxic effects on neuroblastoma and FLG29.1 cells, 1 μg/ml was chosen to perform the cocultures (see above).
Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA integrity was assessed by running aliquots on agarose, and quality was checked by evaluating an A260/A280 ratio higher than 1.7. Reagents for reverse transcription were contained in the Advantage for RT-PCR Kit (Clontech, Palo Alto, CA). All cDNAs were stored at –70°C until use. Diluted cDNA (5 μl) was amplified to detect expression of OPG and RANKL (Table I). Amplification was performed in a DNA thermal cycler (GeneAmp 9600; Perkin-Elmer, Norwalk, CT) under the following conditions: denaturation at 94°C for 5 min, followed by 35 cycles at 94°C for 30 sec, 55°C for 30 sec, 72°C for 60 sec and final extension at 72°C for 10 min.
Table I. List of primers
Primers (source or NCBI sequence viewer)
Product size (bp)
β-Actin amplimer set (Clontech)
RANKL variant 1 (NM_0030701)
RANKL variant 2 (NM_033012.2)
RANKL variant 2 (NM_033012.2)
Cathepsin K (NM_000396)
Real-time PCR analysis
Table I lists the primers designed for the real-time PCR. A conventional PCR was performed to obtain a product of amplification suitable for the construction of standard curves with the real-time PCR procedures. Amplification products were purified with the QIAEX II Gel Extraction kit (Qiagen). The amount of eluted product was assessed spectrophotometrically. Real-time PCR was performed using the Light Cycler instrument (Roche Applied Science, Monza, Italy). Standard curves were constructed by applying a model of linear regression, where the independent variable was the known concentration of template (range of log10 dilution) and the dependent variable was the corresponding value of threshold cycle. cDNA (1.5 μg) was amplified in a final reaction volume of 20 μl containing 2 μl of Master Mix (FastStart DNA Master SYBR Green I; Roche Applied Science). The presence of more peaks in the melting curve indicated that the template was below the detection limit: in this case, the value of the cycle threshold could not be considered reliable, and the concentration was considered arbitrarily log−1 of the lower concentration in the standard curve. Detection limits for 1.5 μg of cDNA were as follows: β-actin, 5.5e−14 μg; RANKL variant 1, 1.15e−8 μg; RANK, 1.93e−9 μg; c-src, 2.34e−8 μg; c-fos, 9.95e−8 μg; 3.71e−10 μg. Lack of the cycle threshold, as in the blank sample (water), indicated absence of template: in this case, the concentration was considered equal to 0. The β-actin housekeeping gene was used as reference for the relative quantification of the gene of interest, which was expressed as the ratio of target to reference concentration.
The presence of the RANKL protein on the SH-SY5Y cell membrane was detected by indirect immunofluorescence. Briefly, a single-cell suspension (1 × 106 cells/ml) was incubated with 0.2 μg/ml of MAb specific for the extracellular domain of RANKL protein (Alexis, Lausanne, Switzerland), washed and stained with 10 μl of FITC-conjugated goat antimouse immunoglobulins (Dako, Glostrup, Denmark). A flow cytometer equipped with an argon laser (EPICS XL-MCL; Beckman Coulter, Hialeah, FL) was used. The percentage of fluorescent cells was calculated out of 10,000 events.
Immunoenzymatic assay of secreted proteins.
A multiple sandwich enzyme immunoassay was performed for the quantitative determination of OPG and sRANKL released from neuroblastoma cells. For OPG detection, the following antibodies were used (R&D Systems): mouse antihuman OPG (2 μg/ml) and goat biotinylated antihuman OPG (100 ng/ml). The standard reference curve was set using rhOPG/Fc chimera with a range of 31.2–1,000 pg/ml. For RANKL detection, the following antibodies were used (Preprotech): goat antihuman RANKL (2 μg/ml) and biotinylated goat antihuman RANKL (0.2 μg/ml). The standard reference curve was set using recombinant human sRANKL (Preprotech) with a range of 0.01–50 ng/ml.
Binding between proteins and respective antibodies was detected by the following sequence of reactions: anticytokine biotinylated antibody, streptavidin–peroxidase conjugate and the substrate reacting with peroxidase, namely, tetramethylbenzidine. The reaction was stopped with 0.18 M H2SO4 and the optical density read by a spectrophotometer with a filter of 450 nm wavelength; dedicated software extrapolated the cytokine concentration. Detection limits of the immunoassay were 1 pg/ml for OPG and 0.05 ng/ml for sRANKL. Values below the threshold of test sensitivity were considered arbitrarily as log−1 of the detection limit.
Immunoenzymatic assay of cellular proteins.
For protein extraction, cells were washed twice with PBS and lysed in buffer containing 25 g/100 ml SDS, 25 g/100 ml TRIS HCl 0.5 M (pH 6.8). Cells were scraped off the flask and the lysates applied in a sterilized tube, boiled for 3 min and then sonicated to have a liquid solution, which was centrifuged at 20,000g for 20 min. The protein concentration of each sample was measured using a commercial kit based on the bicinchoninic acid method, following the manufacturer's instructions (BCA Protein Assay Reagent; Pierce, Rockford, IL).
A multiple sandwich enzyme immunoassay was performed to quantify the concentration of RANKL protein in neuroblastoma cells and the amount of RANK in FLG29.1 cells. Protein lysate (5 μl) was diluted in PBS containing 1% BSA. For RANKL detection, procedures and reagents were the same as for the immunoenzymatic assay of soluble proteins but the standard reference curve was set by diluting recombinant proteins in PBS containing 1% BSA and 5% SDS, to evaluate the interference of SDS. For RANK detection, the following antibodies were used (R&D Systems): mouse antihuman RANK (1 μg/ml) and goat biotinylated antihuman RANK (0.1 μg/ml). The standard reference curve was set using recombinant RANK/Fc chimera (R&D Systems) with a dilution range of 15.6–1,000 pg/ml. Protein lysates were diluted 10-fold in TRIS-buffered saline containing 0.1% BSA and 0.05% Tween-20. The presence of protein was detected as described in the previous paragraph. The detection limit was 0.1 pg/ml. The results of all assays were expressed as (target protein/total protein/number of cells)/ml.
Calculations and statistical analysis
All calculations and statistical analyses were performed using StatView 5.01 for Windows (SAS Institute, Cary, NC). Results were expressed as arithmetic means ± SE. ANOVA and the Bonferroni-Dunn multiple-comparison test were applied to detect the effects of different anti-RANKL treatments. p ≤ 0.05 was considered statistically significant.
Expression of RANKL and OPG in neuroblastoma cell lines
Preliminary experiments showed that in all of the neuroblastoma cell lines large amounts of both RANKL and OPG genes were present (Fig. 1a,b). Only RANKL variant 1 was detected, whereas no amplification of RANKL variant 2 could be obtained (data not shown). Immunoenzymatic determination of proteins corresponding to transcripts was performed in the supernatant of the culture of the aforementioned cell lines (Table II). The highest release of RANKL was detected in SH-SY5Y cells, even if all cell lines produced it. Further analysis confirmed that both isoforms of RANKL protein were detectable on SH-SY5Y. The cell membrane form was detected by flow cytometry, and the highest proportion of positive cells was found after 24 hr of culture (Fig. 1c), whereas the amount of soluble protein increased over time, showing maximal release after 144 hr (Fig. 1d).
Table II. Immunoenzymatic Determination of RANKL and OPG in the culture supernatant
10.4 ± 9
10.1 ± 6
3.6 ± 3
28.8 ± 23
4.7 ± 3
13.6 ± 10
33.9 ± 13
32.5 ± 3
1.1 ± 0.4
0.9 ± 0.6
1,676 ± 275
0.0004 ± 0.0002
Despite the presence of large amounts of transcripts, OPG, which is the natural antagonist of RANKL, inducer of osteolysis, was present in tiny amounts in the culture supernatant (Table II). In particular, high levels of sRANKL were found in SH-SY5Y, whereas OPG release was often below the detection limit. The SH-SY5Y and SJ-N-KP cell lines were selected with the aim of assaying their ability to induce differentiation in coculture experiments with FLG29.1 preosteoclasts (Fig. 2a–c). SJ-N-KP was chosen because it released both RANKL and OPG, whereas SH-SY5Y released only RANKL. The mean percentage ± SE of TRAP-positive cells in FLG29.1 control cultures was 3.1 ± 1%. Coculture with SH-SY5Y induced a strong TRAP-positive reaction in FLG29.1 cells, which was similar to that induced by PMA, while no significant differences were observed between control cultures and cocultures of FLG29.1 and SJ-N-KP. Additional experiments were carried out by cultivating PBMCs in medium containing 25% of the supernatant of confluent SH-SY5Y cells in the absence of osteoclastogenic factors. In this case, qualitative analysis with low-magnification imaging (×10 objective) was more informative than quantitative results, which were poorly comparable because of the biologic diversity among the PBMC donors (number, size and intensity of the TRAP reaction of osteoclasts). After 7 days, some giant cells and many multinucleated cells, strongly positive for TRAP, were observed (Fig. 2d,e). This result shows that SH-SY5Y cells are able to induce an osteoclast-like phenotype by paracrine activity.
Another step in the assessment of differentiation status was to analyze the expression of RANK, the receptor of RANKL in cocultures of SH-SY5Y and FLG29.1. In control culture, FLG29.1 showed a transient presence of RANK transcripts at 12 hr, while addition of PMA induced strong expression of RANK, which persisted up to 48 hr. SH-SY5Y induced expression of RANK at 48 hr at levels that were comparable to those of PMA, whereas SJ-N-KP proved less effective (Fig. 2f).
Biologic modifiers of RANKL activity interfere with the capability of neuroblastoma cells to induce preosteoclastic differentiation
Different molecules, including OPG, anti-RANKL antibody and ODNs, were used to inhibit the capability of SH-SY5Y cells to favor osteoclastic differentiation. Preliminary tests excluded that these molecules had a significant effect on tumor cell proliferation (data not shown). Quantitative variation of RANKL expression was investigated in neuroblastoma cells, whereas the effect of RANKL blockade was studied on FLG29.1 cells by analyzing expression of RANK, c-src, c-fos and cathepsin K. In all experiments dODN, used as a control for non-antisense effects, did not induce significant changes (data not shown). In other hands, the lack of nonspecific effects due to OPG/Fc was proven,22 and we previously demonstrated the specific effect of anti-RANKL antibody: in an experimental model, where osteoblast–osteoclast cooperation was tested, the neutralizing antibody inhibited osteoclastogenesis induced by either a cocktail of differentiating agents, including RANKL, or the conditioned medium of human osteoblasts, while whole goat immunoglobulins did not affect it (data not shown).
Effects of biologic modifiers of RANKL activity on RANKL expression.
As already mentioned, in basal conditions, the quantity of variant 1 RANKL transcripts in SH-SY5Y cells was high. On the contrary, RANKL was not expressed in FLG29.1 cells: real-time PCR showed a lack of transcripts, and consequently the protein was undetectable in both cell lysate and culture medium. FLG29.1 promoted further the transcription of variant 1 RANKL in neuroblastoma cells. Treatment with ODNs significantly reduced RANKL transcript levels. Treatment with OPG did not induce marked changes in the amount of transcripts, while treatment with anti-RANKL antibody lowered gene expression by 3 orders of magnitude, which was an unexpected result since the antibody could hamper protein activity but not its gene expression (Fig. 3a). The presence of FLG29.1 cells induced a slight increase in the amount of synthesized protein (Fig. 3b). According to the gene expression results, ODNs were effective at reducing the amount of protein. Treatment with either OPG or anti-RANKL antibody did not hamper the cell protein, and that was not surprising since both treatments were antagonistics toward the soluble protein. Figure 3c shows how the biologic modifiers of RANKL activity influenced sRANKL. Treatment with ODNs did not have a marked effect on secreted protein. Treatment with OPG did not affect the release of sRANKL, even if the protein could be recognized by the antibody used in the immunoenzymatic method; but it was not bioavailable because it was bound to OPG. The significant reduction in sRANKL after treatment with anti-RANKL antibody represents a positive control: the protein was not detected because of the competition between the neutralizing antibody and the antibody used for the immunoenzymatic assay, which were identical.
Effects of biologic modifiers of RANKL activity on expression of genes involved in osteoclastic differentiation.
In SH-SY5Y cells, expression of osteoclastic differentiation–related genes, (RANK, c-fos, c-src and cathepsin K) was below the detection limit or null, but the ratio with β-actin was higher than 1. The reason for this discrepancy is that the sensitivity of the methods to detect the genes of interest was much lower than that for β-actin analysis; indeed, their detection limit was higher than that of the housekeeping gene. When the amplified product was below the detection limit, it was considered arbitrarily as log−1 of the lower concentration of the standard curve: consequently, the ratio vs. the housekeeping gene was always higher than 1. High levels of RANK transcript and protein were detectable in FLG29.1 cell cultures, in basal conditions and in coculture with SH-SY5Y cells (Fig. 4a,b). PMA increased RANK expression significantly but did not produce significant changes in the amount of protein, which occurred later, as previously described.17 Introduction of anti-RANKL antibody produced a significant reduction in both mRNA and protein. OPG and ODN1 showed a similar effect: they decreased the amount of RANK transcript, even if not significantly, while protein synthesis was not affected.
Neuroblastoma cells increased slightly the expression of osteoclastic differentiation–related genes, such as c-fos and cathepsin K, while the amount of c-src transcript was one logarithmic decade higher than that induced by PMA (Fig. 5a–c). Anti-RANKL antibody induced a significant reduction in the expression of all examined genes. The action of OPG was similar but less effective, especially on c-src expression. Pretreatment of SH-SY5Y cells with ODNs did not change the ability of tumor cells to induce, in osteoclast precursors, the expression of genes related to differentiation.
The effects of RANKL inhibitors on osteoclastic differentiation are shown in Figure 6. A strong TRAP-positive reaction was observed when FLG29.1 was cocultured with SH-SY5Y. Following addition of OPG, and even more with anti-RANKL antibody, in FLG29.1, the TRAP activity disappeared almost completely. The paracrine activity of the SH-SY5Y supernatant on the osteoclastic differentiation of adherent PBMCs was not affected by OPG, while the anti-RANKL antibody completely prevented formation of TRAP-positive cells. Pretreatment of SH-SY5Y cells with ODNs did not prevent differentiation of FLG29.1 or PBMCs.
In the current study, we investigated the peculiarities that enable neuroblastoma cells to colonize bone and induce osteolysis. In particular, we tested the hypothesis that tumor cells might induce osteoclastogenesis by simulating the mechanisms by which osteoblasts regulate osteoclast development. We focused our attention on the ability of neuroblastoma to produce RANKL and OPG. The role that these molecules have in the control of bone remodeling is well known:23 the binding of RANKL to RANK is essential for differentiation of osteoclast precursors and their fusion into multinucleated cells, as well as for activation and survival of mature osteoclasts. OPG plays a crucial role in the homeostasis of the entire system by blocking the effects of RANKL, so the balanced RANKL-to-OPG ratio in the bone microenvironment represents a fundamental requirement for physiologic bone remodeling.
To investigate the hypothesis that the RANKL-to-OPG ratio might be directly influenced by neuroblastoma, we analyzed in different cell lines if and to what extent both proteins were expressed. Neuroblastoma cell lines were derived from primary tumors (CHP212, SJ-N-KP, NB100) or bone marrow metastatic sites (SH-SY5Y).17 All cell lines showed a large amount of both OPG and RANKL mRNA. sRANKL protein was detectable in all cell lines, but despite the large amount of transcripts, poor release of OPG protein was observed. This RNA–protein discrepancy was an unexpected result since the same amount of OPG transcript was found in primary human osteoblasts and that was matched with high levels of released protein, which exceeded 1,000 pg/ml.24 Two cells lines were selected to perform the in vitro osteoclastic differentiation assay, to prove that neuroblastoma cells could directly influence osteoclastogenesis because of unbalanced production of RANKL and OPG. SH-SY5Y was chosen because it showed the highest levels of sRANKL as well as the lowest amount of OPG, while SJ-N-KP released both proteins: the former enabled osteoclastic differentiation, while the latter did not show the same capability.
Our current results partially agree with other authors' findings on the involvement of RANKL in the pathogenesis of bone metastasis from neuroblastoma. In a murine model, Michigami et al.16 found that immunocompromised mice developed bone metastasis after s.c. injection of neuroblastoma cells. Contrary to our study, the cell lines they used did not show RANKL and OPG transcripts so that a conditioned medium containing the OPG protein blocked effectively osteoclast formation in coculture of neuroblastoma and bone marrow cells. Our results show that some neuroblastoma cells possess the genetic machinery for expression of both the above-mentioned proteins but some posttranscriptional events hamper the synthesis of OPG. Our first step confirmed that neuroblastoma cells may induce osteoclastic differentiation per se and suggested that an unbalanced RANKL-to-OPG ratio could play a crucial role in determining metastatic osteolysis. Pro-osteoclastic activity in SH-SY5Y cells derived from bone marrow metastasis of neuroblastoma, i.e., a site in direct contact with the bone microenvironment, may not be a chance finding.
A further aim was to establish which form of RANKL was produced by the SH-SY5Y cell line. RANKL exists in 2 forms: a cell membrane-bound variant that is produced by the majority of tissues and a secreted form. The latter may be generated from cleavage of a cellular form by TACE25 or may be a primary secreted form that has been described in activated T lymphocytes as well as in a squamous carcinoma cell line.26 Our findings demonstrated that in SH-SY5Y both RANKL forms were detectable and active: indeed, osteoclast formation was induced not only by the juxtacrine cell–cell interaction but also by the paracrine activity of the culture medium. Analysis of the transcript from sRANKL did not reveal amplification. It is therefore likely that the secreted form of RANKL found in the supernatant of SH-SY5Y comes from protein cleavage of variant 1. TACE, the enzyme responsible for RANKL release, has been identified in tumor and is able to induce differentiation of monocytes to osteoclasts.27 In our hands, TACE was detectable in >20% of SH-SY5Y cells, while no positive cells were found in the SJ-N-KP line (data not shown). Moreover, chromosomal mapping places TACE on human chromosome 2p25,28 near the MYCN gene (2p24); the latter is one of the most important prognostic factors in neuroblastoma, and its amplification is strongly associated with advanced stages of disease and poor prognosis. Given the large size of the MYCN amplicon, coamplification of additional genes is possible, expression of which may contribute to the aggressive phenotype and help to explain why not all patients with MYCN amplification have poor prognosis.
The final goal was to identify one or more molecules that interfere with RANKL/RANK binding and to revoke the osteoclastogenesis induced by neuroblastoma cells. The use of molecules with an inhibitory effect, namely OPG, RANKL-neutralizing antibody and ODNs, allowed the functional role of RANKL in inducing the osteoclastic phenotype to be increased. Anti-RANKL inhibited the differentiation of osteoclast precursors induced by SH-SY5Y cells: RANK, c-src, c-fos and cathepsin K transcripts and the RANK protein decreased, and the generation of TRAP-positive mononuclear giant cells was inhibited. Besides inhibiting the differentiation of preosteoclasts, anti-RANKL also produced a reduction in RANKL transcripts. Because the task of the antibody is to neutralize sRANKL, effects on gene expression should not be found. A possible explanation is that the antibody binds not only the soluble protein but also the cell membrane-bound form: the cell may be induced to synthesize more RANKL so that when the analysis was performed a lower relative amount of transcript was observed.
The key role of RANKL in the pathogenesis of neuroblastoma skeletal metastases was further confirmed by the functional study, but it was not clear why OPG was able to inhibit the juxtacrine activity of RANKL but ineffective at blocking the paracrine signaling.
The incomplete efficacy of OPG as an inhibitor of osteoclastic differentiation could depend on the insufficient concentration of this antagonistic molecule in the experimental system. One hypothesis is that SH-SY5Y cells could have the capability of neutralizing OPG, as already described for other tumors.29 Some authors have demonstrated that myeloma cells adopt mechanisms to inhibit the production of OPG or its availability in the bone microenvironment. Syndecan-1, a proteoglycan heparan sulfate expressed on the membrane of myeloma cells, binds inactive OPG by the same binding domain as heparin; then, OPG is internalized and degraded by lysosomes.30 Syndecans can be cleaved from the cell surface and released into the extracellular microenvironment by the action of secretase: cells expressing soluble syndecan-1 metastasize more extensively than cells bearing surface syndecan-1.31 Since various forms of proteoglycan heparan sulfate have been identified in neuroblastoma cells,32 it is reasonable to hypothesize that the neuroblastoma also produces and releases molecules capable of binding and neutralizing OPG. This might explain the scarce amount of OPG protein in the culture medium of the various neuroblastoma cell lines, on the one hand, and the poor efficacy of OPG at blocking paracrine RANKL activity, on the other. Furthermore, similar to what has been said for TACE, syndecan-1 shares the same MYCN locus on human chromosome 2.33 Treatment with ODNs was very effective at producing target RNA hydrolysis by RNAse-H: reduced expression of RANKL transcript and cell lysate protein was observed, whereas there were no changes with regard to soluble protein and the capability of SH-SY5Y cells to induce osteoclastic differentiation. The ineffectiveness at blocking fully the expression of the target could be due to insufficient time from ODN administration. In our experimental model, RANKL should be derived from the shedding of the membrane protein, which might be synthesized before treatment with ODNs so that the decrease of RANKL transcript did not influence the amount of the released protein.
Taken together, our findings confirm that neuroblastoma cells are able to induce osteoclastogenesis via RANKL and suggest that the RANKL expression associated with lack of the decoy receptor OPG could be a peculiarity of some tumors that makes them able to colonize bone and induce metastatic osteolysis. Moreover, our results suggest that RANKL could be a relevant target in the adjuvant therapy of bone metastatic neuroblastoma as proper neutralization revokes completely osteoclastic differentiation.