Malignant B-Lymphoid Cells with Bone Lesions Express Receptor Activator of Nuclear Factor-κB Ligand and Vascular Endothelial Growth Factor to Enhance Osteoclastogenesis

Certain types of B-lymphoid neoplasms, such as lymphoma of bone origin and multiple myeloma, exclusively develop and expand in the skeleton. These tumors cause lytic bone lesions, which lead to the debilitating clinical symptoms including intractable bone pain and disabling fractures. In those destructive bone lesions, osteoclasts seem to surround tumor cells and actively resorb bone. Several bone-resorbing factors, including parathyroid hormone–related protein (1), interleukin (IL)-1β (2–4), and IL-6 (3–5), have been implicated as causative factors for bone destruction in multiple myeloma, and recent reports including those from ourselves indicate that a C-C chemokine, macrophage inflammatory protein-1α and macrophage inflammatory protein-1β, plays an important role in the development of lytic bone lesions by multiple myeloma (6–10). However, the pathogenesis as well as the mechanism of bone destruction by B-cell lymphoma of bone origin is largely unknown.

Binding of receptor activator of nuclear factor-κB ligand (RANKL; refs. 11–13) to its receptor (RANK; ref. 14) is essential for the enhancement of osteoclast differentiation, activation, and survival, whereas its decoy receptor, osteoprotegerin (15), inhibits RANKL-RANK signaling. Along with RANKL, macrophage colony-stimulating factor (M-CSF) is required to enhance osteoclastogenesis (16). Recent reports showed that vascular endothelial cell growth factor (VEGF) can substitute for M-CSF and that VEGF and RANKL in combination potently stimulate osteoclastogenesis (17). Growing evidence indicates that tumor cells stimulate osteoclastogenesis by enhancing RANKL expression and suppressing osteoprotegerin in the surrounding cells or by themselves to form destructive bone lesions in multiple myeloma and metastatic bone diseases (18–23). Expression of RANKL by tumor cells has been reported in adult T-cell leukemia (24), pre–B-cell leukemia (25), multiple myeloma (26–28), and prostate cancer (29, 30). Furthermore, the expression levels of RANKL on tumor cells have been found to correlate with the severity of lytic bone lesions in multiple myeloma (27, 28) and hypercalcemia in adult T-cell leukemia (24), suggesting a causative role for RANKL on tumor cells in the development of lytic bone lesions. Although RANKL is shown to be expressed on CD20-positive B cells in the bone marrow (31) and induced by activated B cells (32), RANKL expression on B-lymphoid neoplasms other than pre–B-cell leukemia and multiple myeloma has been poorly examined. Here, we show that a portion of B-lymphoid tumor cell lines as well as primary B-lymphoid tumor cells of bone origin aberrantly express RANKL and VEGF, and directly induce osteoclastogenesis from osteoclast progenitors via cell-to-cell contact. The present results also show that a close interaction between B-lymphoid tumor cells and osteoclasts augments mutual growth and activity.

Reagents. The following reagents were purchased from the indicated manufacturers: recombinant human (rh) soluble RANKL (sRANKL) from Pepro Tech (London, United Kingdom); rhM-CSF, rhVEGF, mouse monoclonal antibodies against human RANKL (TRANCE), IL-1β, IL-6, and CD68 from R&D Systems (Minneapolis, MN); anti-human RANKL, RANK, osteoprotegerin, VEGF, parathyroid hormone–related protein monoclonal antibody from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); antihuman CD20 monoclonal antibody from Becton Dickinson (Mountain View, CA); antitartrate-resistant acid phosphatase (TRAP) antibody from Novocastra (Newcastle-upon-tyne, United Kingdom); antihuman macrophage inflammatory protein-1α monoclonal antibody from Chemicon International, Inc. (Temecula, CA). Recombinant human osteoprotegerin was a kind gift from Dr Tadaaki Higashio (Saitama Medical University, Saitama, Japan).

Primary lymphoma samples from patients. Lymphoma samples in bone were obtained from two patients with primary bone lymphoma manifesting bone destruction (cases 1 and 2) and from eight patients with B-cell lymphoma (four follicular lymphoma, two splenic marginal zone lymphoma, and two mantle cell lymphoma) secondarily infiltrating to the bone marrow without apparent bone destruction. Lymph node tissues were also obtained from 19 patients with nodal lymphoma without bone involvement (10 diffuse large, 5 follicular, and 4 marginal cell lymphoma), and skin specimens from two patients with diffuse large lymphoma. Case 1 showed intraspinal mass in the 11th thoracic vertebra (T11) and T12 with a compression fracture of T11. The spinal lesions were resected because of a rapid progression of spinal cord compression. Pathologic studies on the resected specimen revealed diffuse infiltration of large B-lymphoid cells. A diagnosis of diffuse large B-cell lymphoma was made. Case 2 showed lytic bone lesions at multiple sites in the limbs and the right clavicle. Open biopsy of the osteolytic lesion of the left radius yielded the histologic diagnosis of diffuse large B-cell lymphoma.

Informed consent was obtained from all the patients, and all procedures involving human specimens were done according to the protocol approved by the Institutional Review Board for human protection.

Cells and cultures. Human B-lymphoid cell lines, ARH77, IM-9, Raji, Daudi and Ramos, human myeloma cell lines, U266 and RPMI8226, a human osteoblast-like cell line, MG-63, and a murine preosteoclastic cell line, RAW264.7 were obtained from the American Type Culture Collection (Rockville, MD). Human myeloma cell lines TSPC-1 and OPC were established in our laboratory as previously described (6). Mononuclear cells were isolated by Ficoll-Hypaque density gradient centrifugation from heparinized blood drawn from healthy volunteers and resected tumors in the bone and lymph nodes. Primary B-lymphoid tumor cells were further purified by immunomagnetic cell sorting using an antihuman CD19 antibody conjugated to magnetic beads (MACS system, Miltenyi Biotec, Bergisch, Gladbach, Germany).

Immunohistochemistry. Tissue sections were prepared from 10% formalin-fixed, paraffin-embedded resected tumors and bone marrow aspirates. They were stained with H&E (Sigma, St. Louis, MO). The tissue sections were deparaffinized, rehydrated, and treated with 3% H₂O₂ for 30 minutes to inactivate endogenous peroxidase activity. After blocking with DAKO protein block (DAKO, Carpinteria, CA), sections were incubated with anti-CD20, IL-1β, IL-6, CD68, RANKL, VEGF, parathyroid hormone–related protein and TRAP antibodies, followed by the detection with the standard detection system (LSAB kit; DAKO) containing peroxidase-conjugated secondary antibodies and diaminobenzidine as a chromogen. The cells were visualized with a microscope (BX50; Olympus, Tokyo, Japan) using UPlanFl ×40 objective lens. Images were recorded with a Olympus CCD camera (SC35, Olympus) and Viewfinder Light software (Pixera Corporation, Los Gatos, CA), and digitally processed using Photoshop software (Adobe, San Jose, CA).

Flow cytometry. Cells were incubated with saturating concentrations of FITC-conjugated antibodies on ice for 1 hour. For indirect fluorescence staining, cells were incubated first with primary antibodies on ice for 1 hour, washed, and then incubated with FITC-conjugated secondary antibodies on ice for 30 minutes. Samples were analyzed by flow cytometry using EPICS-Profile (Coulter Electronics, Hialeah, FL).

Cocultures of primary bone lymphoma cells with rabbit bone cells. Rabbit bone marrow cells were prepared from unfractionated bone cells according to previously described procedures with slight modification (33). In brief, minced long bones of 5-day-old male white rabbits were agitated by vortexing and bone particles were removed by sedimentation for 30 seconds in Eagle's α-MEM modification (Invitrogen, Carlsbad, CA). After centrifugation at 300 × g for 3 minutes, two thirds of the supernatant from the top were removed to enrich osteoclast precursors. Thus, obtained fractions of rabbit bone cells were seeded in 96-well plates containing a bovine dentine at 5 × 10³ cells/well in α-MEM containing 10% fetal bovine serum (Whittaker Bioproducts, Inc., Walkersville, MA). Isolated lymphoma cells at 2 × 10³ cells/well were cocultured. At day 4, the number of TRAP-positive multinucleated cells was counted. Subsequently, cells were blushed off and bone slices were stained with hematoxylin for 5 minutes to visualize resorption pits. For quantification, the number of mesh squares covering the pits was counted using a microscope with a mesh glass installed in the ocular lens.

In vitro osteoclastogenesis. RAW264.7 cells were seeded in 96-well plates at 5 × 10⁴ cells/well and cultured for 96 hours in α-MEM containing 10% fetal bovine serum in the absence or presence of 50 ng/mL sRANKL. For coculture experiments, B-lymphoid cells (2 × 10³ cells) were added to the wells precultured with RAW264.7 cells. After 4 days, the cultured cells were stained for TRAP using Leukocyte Acid Phosphatase kit (Sigma) and the number of TRAP-positive multinucleated cells was counted.

In vitro osteoclast generation from peripheral blood mononuclear cell (PBMC) was done as previously described (34). Briefly, an adherent cell population was collected and cultured at 1 × 10⁵ cells/mL in α-MEM containing 10% fetal bovine serum supplemented with 50 ng/mL sRANKL and 500 units/mL M-CSF. Media were replenished twice a week. After 2 weeks, most of adherent cells were still mononuclear and became TRAP positive, and the adherent cells were harvested using 0.05% trypsin/0.53 mmol/L EDTA (Invitrogen). The cells were replated onto 24-well plates at 5 × 10⁴ cells/well, and cocultured with Daudi cells (1 × 10⁴ cells/well). A membrane filter (Intercell TP, Kurabo, Osaka, Japan) was used to prevent cell-to-cell contact. After 5 days, the number of TRAP-positive multinucleated cells was counted.

Collection of conditioned media and cytokine measurement. The B-lymphoid cell lines were cultured in α-MEM with 1% fetal bovine serum at 1 × 10⁶ cells/mL. Conditioned media were harvested at day 2. Human sRANKL and osteoprotegerin were measured by sRANKL and osteoprotegerin kit (Biomedica GmbH, Wien, Austria), respectively; VEGF and M-CSF levels were determined by hVEGF and hM-CSF kit (Quantikine, R&D), respectively.

Migration assays. Cell migration assays were done using membrane filters with 8 μm pore–sized transmigration chambers (Chemotaxicell, Kurabo, Japan). Peripheral blood–derived TRAP-positive mononuclear cells were placed on upper chambers in 400 μL of α-MEM containing 1% fetal bovine serum. Lower chambers were filled with 800 μL of the same medium. Conditioned media from Daudi was added at 10% to either upper or lower chambers or both. Anti-VEGF antibody was added at 20 μg/mL. As a positive control, recombinant VEGF was added to lower chambers at 10 ng/mL. The plates were incubated at 37°C for 4 hours in a 5% CO₂ incubator. After 4-hour incubation, nonmigrated cells on the upper side on membrane filters were removed with a cotton swab and the cells migrated to the bottom of filters were fixed with formalin and stained with Giemsa solution (Merck, Tokyo, Japan). The filters were dried, cut out, and mounted with permount solution (Thomas Scientific, Swedesboro, NJ) on a glass slide. Cells were viewed under a microscope and cell numbers were counted.

Analyses of messenger RNA expression. Total RNA was extracted from cells using TRIzol reagent (Invitrogen). For reverse transcription-PCR, 2 μg total RNA were reverse-transcribed with Superscript II (Invitrogen) in a 20 μL reaction solution. Two microliters of the 20 μL reaction solution were used for the subsequent PCR analyses with cycles of 95°C for 30 seconds, 56°C for 30 seconds, and 72°C for 30 seconds. Primers used are as follows: 5′-GGATCACAGCACATCAGAGCAGAG-3′ (nucleotides 590-613) and 5′-GTAAGGAGGGGTTGGAGACCTCG-3′ (nucleotides 1,079-1,057) for human RANKL and 5′-TGTCTTCACCACCATGGAGAAGG-3′ (nucleotides 340-362) and 5′-GTGGATGCAGGGATGATGTTCTG-3′ (nucleotides 762-750) for human glyceraldehyde-3-phosphate dehydrogenase. Amplified products were dissolved in a 2% agarose gel and visualized with ethidium bromide staining.

Receptor activator of nuclear factor-κB ligand expression and osteoclastogenesis by primary bone B-cell lymphoma cells. To clarify whether lymphoma cells aberrantly express RANKL to cause bone destruction, we first immunohistochemically examined RANKL expression in bone tissue samples obtained from patients with primary bone lymphoma and nodal lymphoma infiltrating to the bone marrow without bone destruction. RANKL was found to be expressed by B-lymphoid tumor cells only from patients with primary bone lymphoma (Fig. 1A). We further investigated RANKL expression by lymphoma cells in B-cell non–Hodgkin's lymphoma without bone destruction. We found RANKL expression by lymphoma cells in only 2 of 19 lymph node samples from patients with systemic nodal lymphoma. We also examined eight bone marrow specimens secondarily infiltrated by lymphoma cells, but there was no detectable RANKL expression by lymphoma cells (data not shown). No RANKL expression was detectable in two cutaneous lymphoma specimens (data not shown). We also did immunohistochemical staining of RANK and osteoprotegerin in the same B-cell lymphoma sections in addition to RANKL. However, we were able to detect neither RANK nor osteoprotegerin immunoreactivity in lymphoma cells in any sections studied (data not shown), suggesting very little or no expression of RANK and osteoprotegerin at protein levels in B-cell non–Hodgkin's lymphoma. In primary bone lymphoma, tumor cells showed positive staining for VEGF along with RANKL (Fig. 1A). No immunoreactivity was detected for other known osteoclastogenic factors, including IL-1β, IL-6, and parathyroid hormone–related protein (data not shown). TRAP- and CD68-positive osteoclasts were observed adjacent to eroded bone surfaces, indicating active bone resorption in these sites (Fig. 1B). Interestingly, lymphoma cells were surrounded by a number of TRAP- and CD68-positive mononuclear cells, namely preosteoclasts or immature cells of osteoclastic lineage (Fig. 1B). These observations suggest a close cell-to-cell interaction between lymphoma cells and such cells of osteoclastic lineage. In flow cytometric analyses, RANKL was expressed on the surface of lymphoma cells from a patient with primary bone lymphoma, whereas tumor cells from a patient with nodal B-cell lymphoma did not express RANKL on their surface (Fig. 1C). Functional activity of RANKL expressed on the tumor cells was corroborated by coculturing the osseous lymphoma cells with an osteoclast-enriched fraction of rabbit bone cells on dentine slices. The RANKL-expressing tumor cells potently enhanced TRAP-positive multinucleated cell and pit formation, which were almost totally abrogated by addition of excess osteoprotegerin (Fig. 1D).

Expression of receptor activator of nuclear factor-κB ligand by B-lymphoid cell lines. To further clarify whether B-lymphoid tumor cells aberrantly express RANKL, we next examined RANKL expression in B-lymphoid tumor cell lines exhibiting mature B-cell or plasma cell phenotypes. Among the cell lines tested, we found constitutive expression of RANKL mRNA in ARH77, IM-9, and Daudi B-cell lines and faintly in TSPC-1 myeloma cell line (Fig. 2A). Cell surface expression of RANKL protein was further confirmed by flow cytometry. RANKL protein was strongly expressed on IM-9 and Daudi cells, which paralleled with the mRNA levels (Fig. 2B). We did not detect sRANKL immunoreactivity in culture supernatants of these cell lines, suggesting that cell-to-cell contact is required for RANKL binding to its receptor, RANK. Although osteoprotegerin, a soluble inhibitor of RANKL, is also known to be produced by various types of tumor cells (35, 36), none of the B-lymphoid cell lines secreted detectable levels of osteoprotegerin in culture supernatants. Thus, RANKL expressed on the surface of the tumor cells acts via cell-to-cell contact without inhibition by osteoprotegerin.

Osteoclastic differentiation from RAW264.7 cells in the presence of receptor activator of nuclear factor-κB ligand–expressing B-lymphoid cell lines. To clarify whether RANKL aberrantly expressed on B-lymphoid tumor cells plays any functional role, we cocultured RAW264.7 preosteoclastic cells with IM-9 and Daudi cells. As shown in Fig. 3A, RAW264.7 cells differentiated into TRAP-positive multinucleated cells upon RANKL stimulation alone even in the absence of stromal cells/osteoblasts. The cocultures with IM-9 or Daudi cells potently enhanced TRAP-positive multinucleated cell formation from RAW264.7 cells. The enhancement of osteoclastogenesis by IM-9 and Daudi cells was mostly abrogated by addition of osteoprotegerin, indicating a critical role for RANKL expressed on these cells in the enhancement of osteoclastogenesis.

Osteoclast formation from human peripheral blood mononuclear cell–derived preosteoclasts through cell-to-cell contact with Daudi cells. The osteoclastogenic activity of RANKL expressed on B-lymphoid tumor cells was further confirmed by coculturing with human preosteoclasts derived from PBMC. Addition of sRANKL alone substantially enhanced formation of multinucleated cells from PBMC-derived preosteoclasts (Fig. 3B). These multinucleated cells possessed the hallmark of osteoclasts, including expression of calcitonin receptors and ability to form pits on dentine slices as well as positive staining for TRAP (data not shown). Similarly, when Daudi cells were cocultured with PBMC-derived preosteoclasts, osteoclast formation was even more enhanced (Fig. 3B). Again, exogenous osteoprotegerin mostly abrogated the osteoclast formation enhanced by Daudi cells as well as sRANKL. Furthermore, inhibition of cellular contact by membrane filters strongly suppressed the osteoclast formation. These observations show that cell-to-cell contact is required for the osteoprotegerin-inhibitable enhancement of osteoclast formation by Daudi cells.

Vascular endothelial growth factor secretion by B-lymphoid cell lines. Although the presence of both RANKL and M-CSF is required for osteoclastogenesis, B-lymphoid tumor cells do not costitutively produce M-CSF. Because VEGF can substitute for M-CSF in osteoclastogenesis (17), we examined the secretion of VEGF by B-lymphoid tumor cells. As shown in Table 1, all B-lymphoid cell lines, including IM-9 and Daudi, constitutively secreted VEGF into culture supernatants. Thus, by expressing both RANKL and VEGF, B-lymphoid tumor cells can enhance osteoclastogenesis by themselves even in the absence of stromal cells/osteoblasts.

Transmigration of peripheral blood mononuclear cell–derived preosteoclasts by Daudi cells. We next examined how RANKL-expressing tumor cells can effectively interact with osteoclast precursors. For this purpose, we conducted transmigration assays to clarify whether tumor cells can attract preosteoclasts toward tumor cells. When Daudi cell–conditioned media was added to chambers beneath transmigration pore membranes, a substantial number of PBMC-derived preosteoclasts migrated to the lower side of the membranes (Fig. 4). When Daudi cell–conditioned media was added to both upper and lower chambers or to only upper chambers, downward transmigration was not enhanced. These results are consistent with the assumption that Daudi cell–derived soluble factors induce chemotaxis of preosteoclasts toward Daudi cells. VEGF is known to exert chemotactic effects on various types of cells, including cells of osteoclastic lineage (37, 38). Because VEGF is secreted by B-lymphoid tumor cells (Table 1), we next examined chemotactic effects of tumor-derived VEGF on preosteoclasts. As shown in Fig. 4, rhVEGF potently enhanced transmigration of preosteoclasts, and a neutralizing antibody against VEGF significantly suppressed transmigration induced by Daudi cell–conditioned media. These results show that tumor cell–derived VEGF, at least in part, contributes to chemoattraction of osteoclast precursor cells.

Growth enhancement of B-lymphoid cell lines by osteoclasts. In addition to osteoclastogenesis induced by a close interaction between B-lymphoid tumor cells and cells of osteoclastic lineage, we asked whether such a cellular interplay also enhances tumor growth and survival. To answer this question, we cocultured Daudi cells with PBMC-derived osteoclasts. Proliferation of Daudi cells was found to be enhanced in the presence of osteoclasts (Fig. 5). The growth-promoting effect of osteoclasts was also observed in other B-lymphoid tumor cells, including Ramos, OPC, TSPC-1, U266, and RPMI8226 (data not shown), suggesting that osteoclasts have growth-promoting effects on a majority of B-lymphoid tumor cells.

The present study shows that RANKL is expressed in some B-lymphoid cell lines as well as in B-lymphoid tumor cells from patients with bone destructive lesions, but not in those with bone marrow infiltration without destructive bone lesions. Such RANKL-expressing B-lymphoid tumor cells also produced VEGF, a substitute for M-CSF, thereby efficaciously inducing osteoclastogenesis even in the absence of stromal cells. For the development of bone lesions by prostate cancer (29) and humoral hypercalcemia by squamous cell carcinoma (39), secretion of soluble form of RANKL has been reported. However, immunoreactivity for RANKL was not detected in conditioned media of any of B-lymphoid cell lines. In addition, the enhancement of osteoclastogenesis by these tumor cells was abrogated by an addition of osteoprotegerin or an inhibition of cell-to-cell contact by membrane filters. These results support the notion that RANKL expressed on the surface of B-lymphoid tumor cells enhances osteoclastogenesis upon its cell-to-cell contact-mediated binding to RANK on preosteoclastic cells.

Although tumor cells express RANKL on their surface, osteoclast formation may not be enhanced unless enough number of osteoclast precursors get close contact with tumor cells. The present results show that conditioned media of B-lymphoid tumor cell cultures enhanced chemoattraction of preosteoclasts and that an anti-VEGF antibody mostly abrogated the enhancement. Furthermore, immunohistochemical examination of bone specimens from patients with primary bone lymphoma with bone destructive lesions revealed that RANKL- and VEGF-positive tumor cells were surrounded by CD68- and TRAP-positive cells of osteoclastic lineage. Thus, it is suggested that B-lymphoid tumor cells secrete VEGF to enhance osteoclastic bone resorption by not only enhancing osteoclastogenesis in cooperation with RANKL but also recruiting osteoclast precursors via its chemoattractant activity.

We observed RANKL expression in two lymph node samples of 19 patients with non–Hodgkin's lymphoma without bone involvement, suggesting that B-cell lymphoma cells rarely express RANKL when it develops in lymph nodes. We did not detect RANKL immunoreactivity in lymphoma cells infiltrating to the bone marrow in eight patients with systemic nodal lymphoma showing no apparent bone lesions. Interestingly, however, a case with follicular lymphoma infiltrating to the bone marrow was reported to show RANKL expression by lymphoma cells and enhancement of osteoclastic bone destruction (40). Therefore, RANKL is aberrantly expressed only in a limited number of B-cell non–Hodgkin's lymphoma clones, but once expressed it may promote osteoclastic bone destruction particularly when such lymphoma cells arise in or infiltrate to the bone marrow. However, the number of the analyzed specimens is small and more extensive study is required to establish clinical relevance of RANKL expression to bone destructive lesions in B-cell non–Hodgkin's lymphoma.

The present study also showed that cell-to-cell contact between B-lymphoid tumor cells and cells of osteoclastic lineage enhances not only osteoclastogenesis but also the growth of lymphoma cells. Thus, whereas B-lymphoid tumor cells stimulate bone destruction by enhancing osteoclastogenesis, the growth of B-lymphoid tumor cells harboring in bone is enhanced by cell-to-cell interaction with osteoclasts. Taken together, the present observations suggest that B-lymphoid tumors of bone origin expand by a cellular interplay between lymphoma cells and cells of osteoclastic lineage and that there is a vicious cycle between bone destruction and tumor expansion in bone.

We thank Momoko Nitta, Hiroe Amou, and Asuka Oda for their expert technical assistance.