Receptor Activator of Nuclear Factor-κB Ligand Expression by Human Myeloma Cells Mediates Osteoclast Formation in Vitro and Correlates with Bone Destruction in Vivo¹

MM⁴ is characterized by the presence of a monoclonal population of terminally differentiated MPC that home to the BM (1, 2). A predominant clinical feature of MM patients is bone destruction, which accounts for much of the morbidity and mortality associated with MM. Although some patients exhibit diffuse osteopenia, most present with multiple discrete lytic lesions adjacent to nests of MPC (3, 4, 5). The precise molecular mechanisms responsible for osteolysis are unknown. However, it is clear from histomorphometric studies that osteolysis is caused by an uncoupling of the normal process of bone remodeling, leading to an increase in OC activity and bone resorption (6, 7). Several known OAFs are produced locally by MPC or by stromal cells in response to MPC (5, 7, 8). These include lymphotoxin (9, 10), IL-1 (11, 12, 13), TNF-α (13), IL-6 (14, 15), parathyroid hormone-releasing protein (16, 17), and MIP-1α (18). Although previous studies have shown that MPCs produce all of these OAFs, their contribution to the osteolytic disease is unclear.

The TNF-ligand family member RANKL has been shown to play a critical role in OC development and activation (19, 20). A secreted form of RANKL has also been identified (21). The activity of RANKL is antagonized by its decoy receptor, OPG (22), a soluble TNF-receptor family member that acts by binding to RANKL and preventing its ligation with RANK (22). OPG-knockout mice develop extensive osteoporosis because of an increased number of mature OCs, indicating that OPG plays a significant role in the regulation of osteoclastogenesis (23). Thus, the RANKL:OPG ratio is a critical determinant of OC formation (24), and a number of reports now demonstrate that this ratio is significantly altered in malignancy-associated osteolysis by factors produced by tumor cells (24, 25).

The role of RANKL in MM has been controversial. Clearly, OAFs produced by MPC are capable of up-regulating RANKL expression by stromal cells (26, 27, 28). However, a recent study has demonstrated that human MM cell lines or primary MPC obtained from MM patients do not express RANKL on their cell surface (28). Moreover, Roux et al. (29) demonstrated RANKL expression by immunohistochemistry in MM patient biopsy material in BM stromal cells but not in MPC. Contrary to these findings, Sezer et al. (30, 31) recently showed RANKL expression immunohistochemically in MPC in patient biopsy material.

In the present study, we demonstrate by RT-PCR and flow cytometry that CD38⁺⁺⁺ human MPCs express RANKL but lack detectable expression of OPG at both the transcriptional and protein level. Furthermore, when cultured in vitro with peripheral blood-derived OC precursors, these cells were capable of supporting OC formation and subsequent bone resorption in a RANKL-dependent manner, in the complete absence of a stromal osteoblast layer. Our studies also demonstrate that high expression of membrane-associated RANKL by MPC was directly correlated with the presence of osteolytic bone lesions in patients with myeloma. These studies therefore suggest that the accumulation of OCs and formation of osteolytic lesions adjacent to nests of MPC is attributable, at least in part, to MPC-derived RANKL.

Human PBMCs were obtained from normal volunteers and isolated on separation gradients (Lymphoprep, Nycomed Pharma, Oslo, Norway). After centrifugation at 400 × g for 30 min at 4°C, the buffy layer was removed with a transfer pipette and washed three times in HBSS (Life Technologies, Inc., Gaithersburg, MD), containing 5% FCS (CSL Limited, Victoria, Australia). Isolated PBMCs were resuspended at a concentration of 1 × 10⁷ cells/ml in α-MEM medium containing 10% (α-MEM-10) and used for OC coculture experiments. Cryopreserved BMMC samples from MM patients were collected by the Transfusion Laboratory, Institute of Medical and Veterinary Science, South Australia. The use of all normal donor and patient-derived material was used after informed donor consent and approved by the Ethics committees of the Institute of Medical and Veterinary Science and the Royal Adelaide Hospital.

Total RNA was extracted from patient-derived BMMC using TRIzol reagent (Life Technologies, Inc., Gaithersburg, MD), as recommended by the manufacturer. Patient BMMCs were sorted on the basis of their expression of CD38, and 2.5–5 × 10⁵ cells were collected for RNA extraction.

First-strand cDNA was synthesized from 1–2 μg of total RNA using a reverse transcription kit, Superscript II (Invitrogen Life Technologies, Inc., Carlsbad, CA). cDNA was then used as a template for amplification by semiquantitative PCR using AmpliTaq Gold DNA polymerase (Perkin-Elmer, Norwalk, CT), as described previously (25). DNA was amplified under the following cycling conditions: denaturation 94°C/1 min, annealing at 58–62°C/1 min, and extension at 72°C/1 min for 22 cycles for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and 30–40 cycles for other primer pairs, such that all products could be assayed in the log-linear phase. This was followed by a final extension step at 72°C for 10 min. Most of the primer sequences used and their expected product sizes are as described previously (25). Primer sequences used to amplify and detect sRANKL and TM-RANKL isoforms are shown in Table 1. PCR amplification products were resolved by electrophoresis on a 2% w/v agarose gel and visualized by SYBR Gold (Molecular Probes, Eugene, OR) staining at 570 nm. The relative amounts of PCR products were determined by quantitating the intensity of the bands using a Fluorimager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA), as described previously (25). Negative controls, which had no cDNA added to the reaction mixture, were performed with all PCR reactions.

Expression of RANKL was determined by flow cytometry. BMMCs were blocked in blocking buffer comprised of HBSS supplemented with 20 mm HEPES (pH 7.35) and 5% (v/v) FCS, 2% (v/v) normal human serum, and 0.4% (v/v) BSA and stained with a mouse anti-RANKL antibody (TRANCE; MAB626, R&D Systems, Inc., Minneapolis, MN) for 1 h on ice. After washing twice in HBSS supplemented with 5% (v/v) FCS (HHF), cells were incubated with a 1:50 dilution of biotinylated goat antimouse IgG antibody (Southern Biotechnology Associates, Inc., Birmingham, AL) for 45 min on ice. Cells were washed again and incubated on ice for 30 min with a 1:100 dilution of streptavidin antibody conjugated to FITC (Caltag Laboratories, Burlingame, CA). Purified mouse IgG2_b (1A6.11; a kind gift from Prof. L. K. Ashman, University of Newcastle, Australia) was used as an isotype-matched negative control for the TRANCE antibody. In two-color stains, anti-CD38 directly conjugated to phycoerythrin (BD PharMingen, San Diego, CA) was added in the final incubation step with streptavidin-FITC. Purified mouse IgG1 (BD PharMingen) was used as the isotype-matched control for CD38. All samples were fixed in “FACS Fix” (1% v/v formaldehyde, 2% w/v D-glucose, and 0.02% w/v sodium azide in PBS) and analyzed on an Epics-XL-MCL flow cytometer (Beckman Coulter, Hialeah, FL), on the basis of their forward and side light scatter properties and their FITC and phycoerythrin fluorescence.

Cells were blocked in blocking buffer, washed twice in PBS, and fixed for 20 min in 1% paraformaldehyde at room temperature. Cells were then washed twice in HHF and 0.1% saponin (Sigma Diagnostics, Inc., St. Louis, MO) to permeabilize the membrane. Cells were immunostained as described above, with all washes and antibody dilutions in HHF and 0.1% saponin. Intracellular RANKL expression was detected with a rabbit anti-RANKL polyclonal antibody (Sc-6073; Santa Cruz Biotechnology, Santa Cruz, CA). The isotype-matched negative control was purified rabbit immunoglobulin with no human reactivity. The rabbit anti-RANKL antibody was used to examine the expression of cytoplasmic RANKL protein, because the epitope identified by the monoclonal anti-TRANCE antibody was paraformaldehyde-fixation sensitive.

Because primary MPCs do not have the capacity to divide or survive for extended periods in vitro, and the temporal requirements for the exposure of developing OC precursors to RANKL (32), we derived a method for generating OC in coculture with MPC. First, normal PBMCs were plated onto slices of sperm whale dentine (1 × 10⁶ cells/slice) in α-MEM-10 medium in 96-well plates. Cells were allowed to adhere for 1 h at 37°C, and nonadherent cells were removed by washing three times in HHF. Adherent PBMCs were cultured on the dentine slices in 0.2 ml of α-MEM-10 supplemented with 25 ng/ml recombinant human M-CSF (Genetics Institute, Cambridge, MA) for 1 week, and the media were replenished every 3 days. On day 7 of the culture, MM patient BMMCs were stained for CD38 and CD45 antigens and sorted on the basis of surface expression. Each of the sorted populations (CD38⁻, CD38⁺, CD38⁺⁺⁺45⁺, and CD38⁺⁺⁺45⁻) were added to wells at a concentration of 2 × 10⁴ cells/well. For each population, there were four conditions (each condition was performed in triplicate): (a) no exogenous treatment; (b) soluble, recombinant human RANKL/IL-1β (each 2 ng/ml; R&D Systems, Inc.) added on day 16 of culture to activate and prolong the survival of any OC formed, because in our experience, the MPCs in these populations would be poorly viable at this time and therefore be suboptimal to “support” OC development; (c) rhOPG (50 ng/ml) added on day 7; and (d) rhOPG (50 ng/ml) and RANKL/IL-1β (each 2 ng/ml) added on days 7 and 16, respectively. As a negative control, PBMCs were cultured in the absence of sorted cells and in conditions (a) to (d). This was to provide a baseline level of spontaneous OC formation attributed to PBMCs alone. As a positive control, PBMCs were treated with 50 ng/ml soluble RANKL in medium containing 25 ng/ml recombinant human M-CSF. For each condition, the medium was changed every 3 days for the duration of 21 days, and the dentine slices and wells were stained for the presence of TRAP-positive cells, and the dentine slices were examined for the presence of resorption lacunae. The numbers of TRAP⁺ cells were scored by light microscopic analysis, and TRAP⁺ cells with three or more nuclei were scored as positive. Results were enumerated in triplicate wells ± SE, and the significant differences between treatments were determined using Student’s t tests (two tailed, paired).

The appearance of large, multinucleated cells staining positive for TRAP was indicative of the formation of OCs in coculture. Cells were fixed in 4% (w/v) gluteraldehyde in PBS and stained using a commercially available Leukocyte Acid Phosphatase kit (Sigma Chemical Co., St. Louis, MO).

To assess bone resorption by OCs after coculture, dentine slices were treated with 6 m ammonium hydroxide for 2 h and ultrasonicated for 30 min to remove cell debris, washed in 70% ethanol, and dried overnight. The dentine slices were then mounted on stubs, carbon coated, and visualized on a Philips XL-20 scanning electron microscope, as described previously (33). The number of identifiable contiguous resorption pits (resorption sites) were determined for each dentine slice. The area of each resorption site was quantitated from SE images, using QUANTIMET image analysis software (Leica, Cambridge, United Kingdom) as described previously (34).

Total serum calcium measurements were performed as part of routine pathological assessment of MM patients, using the arsenazo III dye method (35), and analyzed on an Olympus 5400 general chemistry analyzer.

Femora, tibiae, skull, lumbar vertebrae, and pelvis were radiographed using a Faxitron X-ray system (Hewlett Packard, McMinnville, OR), as part of routine diagnostic skeletal survey. Patients were scored as either “+” or “−” depending on the presence or absence of osteolytic lesions, respectively. In addition, using a modification of the Durie and Salmon method for skeletal involvement (36), patients were scored according to the following system: 0 = no osteolytic lesions; 1 = one or more osteolytic lesion in one site; or 2 = multiple lesions in multiple sites. Significant differences between groups were determined using the Mann-Whitney U test.

For semiquantitative RT-PCR analysis, experiments were performed in triplicate, and the data are represented as mean ± SE. For comparisons between molecular and radiographic parameters, the Mann-Whitney U test and Student t test were used to determine the statistical difference between two nonparametric and parametric populations, respectively.

RANKL expression in normal and myeloma BM was determined by dual color immunofluorescence staining, using CD38 as a marker of MPC. CD38 staining of myeloma BM revealed three characteristic subpopulations: (a) mature myeloma (and normal) plasma cells express high levels of CD38 and are designated here as CD38⁺⁺⁺; (b) monocytoid cells express intermediate levels of CD38 and are designated here as CD38⁺; and (c) CD38⁻ cells include a variety of other hematopoietic lineages, including non-B lymphoid cells present in the BM. As seen in Fig. 1, membrane-bound RANKL was highly expressed by CD38⁺⁺⁺ MPCs in the myeloma BM. The majority of the MPCs were found to express RANKL, although the staining was heterogeneous. Normal BM, on the other hand, contains few plasma cells (<2%), as indicated in Fig. 1, where there is a corresponding lack of the CD38⁺⁺⁺ population in the normal BM. However, in normal marrow, there was a notable shift in both the CD38⁺ and CD38⁻ populations when intracellular RANKL expression was examined in cells isolated from five normal donors (data not shown). These staining patterns were confirmed by substituting the CD38 antibody with an antibody directed against CD138 (syndecan-1), an additional, independent marker of MPC (data not shown). Fig. 2 shows MPC cells from 15 patients, each with varying degrees of RANKL expression, determined by gating on CD38⁺⁺⁺ cells. In 9 of 15 patients tested (UPN 1, 3, 4, 11, 12, 13, 14, 16, and 17), RANKL was strongly expressed on the surface of MPC. In the remaining 6 patients, membrane RANKL was weakly expressed. Similar results were observed when intracellular RANKL expression was examined (data not shown).

To determine the gene expression profile of various cell populations in MM patients, RT-PCR was performed on patient-derived FACS-sorted cells. Cells were sorted on the basis of their CD38 expression giving rise to lymphoid (CD38⁻), monocytoid (CD38⁺), and MPC (CD38⁺⁺⁺) populations (Fig. 3,A). RANKL mRNA expression was consistently detected in the CD38⁺ and CD38⁺⁺⁺ populations (Fig. 3,B). When primers were used to specifically amplify either the sRANKL or TM-RANKL isoforms, the CD38⁺⁺⁺ MPCs were found to express a message for both isoforms. Conversely, CD38⁺ cells were found to express predominantly the TM-RANKL isoform. OPG mRNA expression was either very weak or not found in CD38⁺⁺⁺ MPC (Fig. 3,B). Consistent with data for normal B-lineage cells (37), RANK mRNA was also expressed in CD38⁺⁺⁺ MPC (Fig. 3 B). As expected, RANK mRNA was also found in monocytoid cells (CD38⁺ CD14⁺), which are likely to include OC precursors in these BM aspirates (data not shown).

To test whether the RANKL expressed by MPC was functional and of likely in vivo significance, OC-forming cocultures were performed. A limitation of this type of study is the poor survival and viability of primary MPC when these are cultured in vitro. To address this problem, MM patient-derived BMMCs were stained for both CD38, a marker for MPC, and CD45, a marker for immature MPC (Fig. 4,A). CD38⁺⁺⁺45⁺ MPCs would be expected to have increased survival capability and thus would be most likely to support OC formation in an in vitro setting. Furthermore, the requirements of OC precursors for exposure to RANKL have been demonstrated to be temporal, such that exposure between days 7 and 14 of a 21-day culture period has been shown to be necessary and adequate for human OC formation from PBMC precursors grown in the presence of M-CSF (32). Therefore, various subpopulations of BMMCs were sorted by FACS and cocultured with PBMCs pretreated with M-CSF for 7 days, as described in “Materials and Methods.” To demonstrate the formation of morphologically identifiable OCs, OCs were defined as TRAP⁺ cells with at least three nuclei. PBMCs cultured in M-CSF in the absence of MM cell subpopulations gave rise to few, if any, morphologically identifiable OCs (Fig. 4,B). However, CD38⁺⁺⁺45⁺ MPCs were observed to support the formation of a significant number of TRAP⁺ OCs with three or more nuclei (P < 0.03; Fig. 4,B). We reasoned that the coculture conditions were still suboptimal because of the relatively poor survival of even the CD38⁺⁺⁺45⁺ MPCs. Therefore, to prolong the survival and activity of any OCs formed in this assay system, recombinant RANKL and IL-1β were added to replicate wells at day 16 of the coculture at concentrations which promote the survival and the activity of existing OCs but not de novo OC formation. Indeed, the addition of this cytokine cocktail to PBMCs cultured alone did not result in the formation of any additional OCs (Fig. 4,B). However, on the addition of RANKL and IL-1β to the coculture with CD38⁺⁺⁺45⁺ MPCs, significantly more large, multinucleated (5–30 nuclei) TRAP⁺ OCs were formed (P < 0.03; Fig. 4,B). To demonstrate that OC formation in this system occurred through the RANKL/RANK signaling pathway, recombinant OPG was added to cultures on day 7, at the time of MPC addition. In all instances, OPG abolished the formation of identifiable OCs (P < 0.005; Fig. 4 B), suggesting that the process by which MPCs support osteoclastogenesis is mediated through RANKL.

To test the function of the newly formed OCs, cocultures were also performed where the cells were cultured on dentine slices, and the level of resorption was assessed by scanning electron microscopy after 3 weeks. Consistent with the patterns of TRAP staining, cultures of PBMC alone, in the presence or absence of recombinant RANKL and IL-1β added at day 16 of the culture, gave rise to few recognizable resorption lacunae (Fig. 5,B). In contrast, coculture of PBMCs with CD38⁺⁺⁺45⁺ MPCs resulted in the formation of TRAP+ OCs[Fig. 5,A(a) ], which gave rise to small but clearly defined resorption lacunae [Fig. 5,A (b) and Fig. 5,B]. However, with the addition of RANKL and IL-1β on day 16 of culture, much more extensive resorption was observed, with pits of ∼10–15 μm in diameter, indicating that the OCs formed were activated or survived longer and were better able to resorb bone [Fig. 5,A(d) ]. The increased size and degree of multinucleation of the formed OC in the presence of added RANKL and IL-1β [Fig. 5,A (c)] are consistent with reports that these cytokines are also involved in the late stages of OC fusion and subsequent function (38). The addition of recombinant OPG at day 7 (at the time of MPC addition) inhibited the formation of TRAP+ cells [Fig. 5,A (e)] and the resorption of dentine by these OCs [Fig. 5,A (f)]. Cocultures with CD38⁺⁺⁺45⁻ mature MPCs, which have a very poor survival potential in vitro, gave rise to few resorption lacunae, even on the addition of RANKL and IL-1β on day 16 (Fig. 5,B). This strongly implies that the critical stages of OC formation in this culture system are supported by viable (CD38⁺⁺⁺45⁺) MPC and not simply the exogenous recombinant RANKL and IL-1β. Consistent with our previous findings (34), only 15–20% of the TRAP+ cells were capable of creating a resorption site on the dentine slice. However, the number of resorption sites corresponded closely with the relative number of TRAP+ cells in each condition examined (compare Fig. 4,B with Fig. 5,B). As a positive control in this system, a high concentration of recombinant RANKL (50 ng/ml) was added from day 7 of the culture of PBMC alone. This gave rise to OC phenotypically identical to those derived from the coculture of PBMCs with CD38⁺⁺⁺45⁺ MPCs when 25-fold less RANKL was added at day 16 of the culture [Fig. 5,A (g)]. The resorption pits in the positive control were generally larger (638.2 +/− 74.48 μm²; mean +/− SE, n = 76), when compared with those generated by coculture with CD38⁺⁺⁺45⁺ MPCs [193.87 +/− 16.13 μm², mean +/− SE, n = 78; compare Fig. 5,A(d) with Fig. 5 A (h)]. This difference in pit characteristics between the positive control and pits formed in coculture likely reflects the poor viability of primary MPC in vitro, even under the pro-survival and activation conditions used here, but could also be caused by suboptimal relative cell densities or other aspects of the assay. Nevertheless, this is the first demonstration that purified MPCs have been shown to participate in de novo OC formation, in the absence of any other stromal cell layer.

To further examine the likely in vivo relevance of RANKL expression by MPCs, we conducted a blind study to correlate expression with parameters of bone loss in MM patients. Radiographs of multiple bones (femora, tibiae, skull, lumbar vertebrae, and pelvis) were assessed for the presence of radiologically detectable bone lesions as part of routine diagnostic skeletal survey. In initial studies, patients were scored as either positive or negative with regard to the presence or absence of radiologically detectable lytic lesions, respectively. When these data were plotted with reference to the expression of TM-RANKL (Fig. 6,A), it was evident that the expression of high levels of TM-RANKL by CD38⁺⁺⁺ MPCs was associated with the presence of osteolytic bone lesions (P = 0.01). Using a modification of the Durie and Salmon system (36), patients were further subdivided and scored according to those who possessed no radiographic lesions (score = 0), those who possessed one or more lesions in one site (score = 1), and those who possessed more than one lesion in multiple sites (score = 2). Correlations were then sought between osteolytic score and expression of RANKL by MPCs derived from the same patients as described above (Fig. 6,B). High levels of TM-RANKL expression by CD38⁺⁺⁺ MPCs were associated with the presence of multiple osteolytic bone lesions (P = 0.01; Fig. 6 B). Together with the above in vitro data, this strongly implies that the expression of RANKL by MPCs has a functional role in the osteolysis observed in MM. No correlation was observed when the level of TM-RANKL expression was compared with the level of total serum calcium on the day of BM acquisition (data not shown). Furthermore, there was no difference between the levels of serum calcium in patients with osteolytic lesions and those without (data not shown), consistent with the findings of Durie et al. (39).

MM is a B-cell malignancy that has profound effects on the skeleton. Bone destruction is mediated by OAFs, which are produced locally by MPCs, or by osteoblasts/BM stromal cells in response to MPCs. These factors stimulate OC formation and survival and ultimately increase overall OC activity. A number of MPC-derived OAFs, including IL-6, IL-1β, TNFα, MIP-1α, and parathyroid hormone-releasing protein, have now been shown to act by up-regulating the expression of RANKL and/or down-regulating the expression of OPG in osteoblastic stromal cells (27, 28). Although RANKL has been implicated in the pathogenesis of MM in an indirect fashion (27, 28), the expression of RANKL by human MPC themselves has been the subject of considerable controversy (29, 30, 40). Our study demonstrates unequivocally the direct expression of RANKL by MPCs.

Using more sensitive techniques than have been used previously (28, 29), we found that RANKL was expressed at both the mRNA and protein level in patient-derived MPCs. We have shown that RANKL was detected by flow cytometry at varying levels in all patient-derived MPCs. In some patients, RANKL was highly expressed on the MPC cell surface, whereas others had very little membrane-associated RANKL but had higher levels of cytoplasmic RANKL. This difference in expression is potentially caused by the cleavage of membrane-bound RANKL by TACE, a recently identified metalloproteinase, which is known to cleave RANKL and other TNF family members from the cell surface (41). By RT-PCR, we found that TACE is expressed by MPC,⁵ raising the possibility that in MM, TACE may cleave membrane-bound RANKL from the surface of MPC. TACE-like activity may account in part for the lack of immunostaining for RANKL observed in MPC observed by others (40). We have also shown using isoform-specific RT-PCR primers that MPCs express the secreted isoform of RANKL, as well as the transmembrane isoform. Functionally, there is no specific difference between the various RANKL isoforms (41, 42); however, it may be speculated that differences might be observed at the microenvironmental level, with respect to the focal or disseminated nature of the resulting lytic lesions. Consistent with our study, Sezer et al. (30) recently reported the expression of RANKL in cells morphologically identified as MPCs in patient biopsy material using immunohistochemical methods. Our findings using two independent anti-RANKL antibodies and RT-PCR analysis of highly purified FACS-sorted subsets confirm and extend these earlier studies.

Importantly, we have shown that primary MPCs, in addition to expressing RANKL, are capable of directly supporting osteoclastogenesis in vitro. Immature MPCs (CD38⁺⁺⁺45⁺), when cocultured with PBMCs, were able to support TRAP⁺, multinucleated OC formation. The addition of recombinant OPG to cocultures simultaneously with the addition of MPCs completely blocked the formation and activity of OC in these cocultures, showing that OC formation was RANKL mediated. A problem with using patient-derived cells of this type is their limited survival and proliferative capacity in vitro. This was overcome by adding the MPC to the PBMC on day 7 of culture. By day 16, the majority of MPC had undergone apoptosis, because the culture conditions described here were only able to promote their survival for 48–96 h (data not shown). However, this brief window of exposure of OC precursors to MPC-expressed RANKL and the other factors important in the early stages of OC differentiation proved to be adequate for triggering OC maturation, consistent with findings using recombinant RANKL to generate OC from PBMCs (32). To complete the process of differentiation and further activate developed OC (38), low concentrations of recombinant RANKL and IL-1β were added to some cocultures on day 16, concentrations which were inadequate to support OC differentiation in the absence of stromal cells. This resulted in a significant increase in the number of fully functional multinucleated TRAP⁺ OC, which were able to effectively resorb bone. This profound increase in the number of TRAP⁺ cells was unique to the CD38⁺⁺⁺45⁺ MPCs cocultured with PBMCs. The CD38⁺⁺⁺45⁻ MPCs have a more mature phenotype and less proliferative potential (43), undergo rapid apoptosis when cultured ex vivo, and may, in part, explain why these cells are unable to support osteoclastogenesis. Therefore, in addition to merely supporting the survival of the osteoclast precursors, our studies indicate that MPCs are also able to promote the entire process of OC formation in a RANKL-dependent manner.

OPG is a key regulator of OC formation by functioning as a decoy receptor for RANKL (22). We have shown by RT-PCR and ELISA that OPG is not expressed by patient-derived MPCs. This is consistent with studies by Shipman and Croucher (44), who also demonstrate that patient-derived cells lack OPG expression and, moreover, cause a down-regulation of stromal cell-derived OPG. This is a significant finding, because in vivo, the relative ratio of RANKL to OPG has been shown to be an important parameter of both physiological bone remodeling (45) and an important determinant for the progression of bone disease in MM (27, 28). Therefore, the localized expression of membrane-bound and soluble RANKL by MPC, without the compensatory expression of OPG, will contribute to an overall shift in the ratio of RANKL to OPG, in favor of OC formation.

The expression of RANKL by MPC is likely to be a highly significant finding from a clinical perspective. In the cohort of 22 MM patients in this study, the level of expression of RANKL by MPC was significantly higher in cells from patients with multiple identifiable osteolytic lesions than those from patients without osteolytic lesions. Serum calcium levels on the other hand were not informative with respect to RANKL expression, consistent with the findings of Durie et al. (39), who reasoned that calcaemia in MM may fluctuate for a number of reasons distinct from pathological bone loss, including calcium absorption and its excretion rate (39). Together, our data strongly suggest that in addition to other known OAFs, the expression of RANKL by MPC may play a significant role in the focal formation of OCs. This is supported by the in vitro OC formation data presented here and points to a possible direct pathophysiological role for MPC-derived RANKL in MM. Our data are consistent with the recent findings of others (30, 31, 46), who also show that RANKL expression correlates closely with osteolytic bone disease in patients with MM.

Taken together with existing data, our results imply that MM is a disease in which the tumor cell, the MPC, in addition to producing autocrine growth factors, deregulates the normal process of bone remodeling, in favor of OC formation, by a number of means: (a) by producing and secreting RANKL themselves; (b) by inducing the expression of RANKL and down-regulation of OPG expression by osteoblasts and stromal cells via the production of other OAFs; and (c) by replacing the marrow space normally occupied by OPG-expressing lineages of cells with MPCs that do not express OPG. Consistent with a number of recent studies (47, 48, 49, 50), our study suggests the possibility of blocking RANKL-mediated osteoclastogenesis by using recombinant OPG as a treatment for MM-induced osteolysis, which may significantly reduce the morbidity associated with this disease.

We thank the Genetics Institute for their provision of recombinant human M-CSF for this study. We also thank the staff of the Detmold Imaging Facility, Andrew (Sandy) Macintyre, and Alan Bishop for their assistance with cell sorting. Finally, we thank Dr. Stan Gronthos for his critical and insightful comments.