Identification of Breast Cancer Cell Line-derived Paracrine Factors That Stimulate Osteoclast Activity¹

Most women with breast cancer receive treatment that seeks to irradicate cancer cells from the breast. Twenty-four percent of those patients with an apparent permanent elimination of cancer from the breast and a lack of evidence of skeletal metastases at the time of surgery will eventually develop signs or symptoms of breast cancer metastases involving the skeleton (1). This indicates that undetectable, microscopic bone metastases were present when the breast cancer was originally diagnosed, and it underscores the importance of understanding how these microscopic breast cancer deposits in bone develop into clinically relevant tumors. Three events must occur before women with breast cancer develop the signs or symptoms of skeletal metastases: (a) cancer cells must leave the breast, travel to bone, and occupy the osseous intramedullary compartment; (b) cancer cells within the osseous intramedullary compartment must induce bone destruction to provide space for tumor growth; and (c) clusters of cancer cells within the osseous intramedullary compartment must grow to form solid tumors. We are investigating the hypothesis that it is locally produced, tumor-derived paracrine factors that are driving the debilitating bone loss associated with metastatic cancer.

The prevalence of bone metastasis in breast cancer patients is highlighted by the fact that, at the time of autopsy, 70% of the women who die from breast cancer show metastases to bone (reviewed in Ref. 2). Tumor cells travel to other parts of the body by altering their phenotype to exploit the blood vasculature and lymph system for transport and deposit in other tissues. Once the tumor cells arrive in bone, they can begin to grow and actively alter their environment to maximize growth. As this growth proceeds, the tumors stimulate the destruction of large amounts of bone at the site of the tumor. This focal loss of bone weakens the skeletal structure and usually results in considerable pain, decreased mobility, hypercalcemia, and significant levels of skeletal fracture. Once tumor cells are deposited and begin to grow in the bone, curative therapy is problematical. For most of these patients, the goals of treatment aim to alleviate discomfort and prevent pathological fractures. Current treatments enable control of tumors in the breast, and patient deaths are more likely caused by metastatic cancer. Thus, therapies that limit tumor-driven bone destruction could greatly slow the progression of complications and suffering. Because a significant problem both in terms of patient suffering and in terms of promoting tumor progression is the result of tumor-driven osteolysis, tumor stimulation of osteoclastic bone resorption is an important target in studies seeking for ways to slow tumor progression.

Multiple growth factors and cytokines have been reported to influence osteoclast differentiation (reviewed in Refs. 3, 4). These include M-CSF 1,³ GM-CSF, IL-1 and -6, TGF-β, IGFs, TNF-α, and PTHrP. Much less is known concerning the influences of these factors on the resorption activity of mature osteoclasts. Osteoclasts have been reported to express receptors for M-CSF, TGF-β, IGFs, IL-1, IL-6, and PTH/PTHrP (5, 6, 7, 8, 9, 10). Thus, these factors could potentially impact the activity of the differentiated cells directly in addition to influencing differentiation. We have, therefore, sought to determine whether breast cancer cells stimulate osteoclast resorption directly and which growth factors secreted by these cells are candidate factors responsible for this stimulation.

MDA MD 231 cells were obtained from American Type Culture Collection (Rockland, MD) and subcultured in phenol red-free α Minimal Essential Medium (αMEM) obtained from Life Technologies, Inc., Gaithersburg, MD, supplemented with 10% fetal bovine serum at 37°C, 5% CO₂ until confluent. Cell monolayers were washed with sterile PBS (pH 7.4) and placed in αMEM supplemented with 0.25% (wt/v) BSA obtained from Sigma Chemical Co. (St. Louis, MO) for 3 days. At the time of collection, cellular debris was removed by centrifugation, and aliquots were frozen at −70°C until analyzed.

Osteoclasts were isolated from White Leghorn hatchlings that were maintained on a low-calcium diet for a period of 5 weeks (11). All of the animals were treated as humanely as possible and treatment followed the NIH and institutional guidelines for care and use of experimental animals. An osteoclast-directed monoclonal antibody, 121F (a gift from Dr. Philip Osdoby, Washington University, St. Louis, MO), coupled to immunomagnetic beads obtained from Dynal, Inc., was used to obtain cell populations that consisted of at least 90% pure osteoclasts and 10% or less unidentified mononuclear cells (12). The purified osteoclasts exhibited all of the phenotypic attributes of osteoclasts including multinucleation, attachment, and ruffled border formation when cultured with bone particles, and the ability to attach and form resorption pits when cultured on slices of cortical bone. Osteoclasts were cultured in phenol red-free αMEM supplemented with 0.25% (wt/v) BSA as described for individual experiments (see figure legends and below).

Isolated osteoclasts were plated on 1-mm² slices of bovine cortical bone. Bone slices were prepared as described previously (13). Samples were treated with vehicle or the indicated test substance as detailed in the figure legend. After 24 h of culture, the slices were placed in 1% (v/v) paraformaldehyde in PBS. The number of osteoclasts per mm² slice was determined for each slice as follows: the fixed slices were rinsed with water and stained for TRAP activity using a Sigma histochemical kit. Osteoclasts were identified as stained multinucleated cells. The number of pits per osteoclast was determined after removal of the cells. The pits, resulting from osteoclast activity, were stained with toluidine blue, counted by reflected light microscopy, and expressed as the number of pits per osteoclast as described previously (13, 14).

Osteoclasts were cultured on bone slices with either vehicle or the test substance as detailed in the figure legends. The conditioned media were harvested, and the amount of antigenic collagen fragments released was determined as described previously (15).

Cell pellet extracts and conditioned media were assayed. To standardize for relative cell number, the protein content of the solubilized cell pellet was determined using the Bio-Rad protein detection system. TRAP activity was measured using an assay based on the work of Hofstee (16). The initial rate of hydrolysis of o-carboxy phenyl phosphate was determined by following the increase in absorbency at 300 nm resulting from the liberation of salicylic acid. One unit is defined as the amount that hydrolyses 1 μmol of o-carboxy phenyl phosphate per min at 24°C (pH 5.0). The assay was performed in the presence of 1 mm tartrate. cathepsin B levels were measured by Na-CBZ-lysine p-nitrophenyl ester hydrolysis as measured by 520 nm absorbance as outlined by Barrett and Kirschke (17).

Recombinant human growth factors were purchased from R&D (Minneapolis, MN) and reconstituted in αMEM supplemented with 0.25% (wt/v) BSA at 1000-fold the concentration used in each experiment (see figure legends). Aliquots were stored at −70°C.

Conditioned media were collected as outlined above and 1 ml loaded onto a Superdex 75 molecular sieve column (Pharmacia, Piscataway, N.J.), which has a functional separation range of M_r 5,000–75,000 after pre-equilibration with αMEM using 3 bed volumes (150 ml) at 1 ml/min. Gel filtration separation of the sample is carried out with a flow rate of 0.5 ml/min with a back-pressure of 0.7 Mpa. Thirty 1-ml fractions were collected on ice and frozen immediately at −70°C until assayed for effects on osteoclast activity or growth factor quantitation.

IL-1, IL-6, M-CSF, GM-CSF, TGF-β1, TGF-β2, and TNF-α were quantitated using R&D Quantikine kits according to the instructions. IGF-II and PTHrP levels were analyzed by the method of de Leon and Asmerom (18).

Unless otherwise indicated in the figure legends, the results represent the mean ± SE of three separate experiments. The effect of treatment was compared with control values by one-way ANOVA; significant treatment effects were further evaluated by the Fisher’s least significant difference method of multiple comparisons in a one-way ANOVA. Tests were carried out using Apple software, obtained from Statview II (Abacus Concepts, Inc., Cupertino, CA).

Initially, we surveyed conditioned media from several well-characterized breast cancer cell lines for their effects on osteoclast resorption activity. As demonstrated in Fig. 1, conditioned media from each of these cell lines stimulated bone resorption, although the stimulatory level varied with the cell line. For subsequent studies, we have focused our studies on the cell line MDA MB 231 because this cell line has proven to cause osteolytic lesion in an in vivo animal model (19). To estimate the total volume of bone resorbed when osteoclasts are cultured in the presence of MDA MB 231 cell-conditioned media, we have used a newly developed assay that quantitates the amount of collagen peptide released. Osteoclasts were cultured on bone slices and treated with a series of dilutions of MDA MB 231 cell-conditioned media (Fig. 2,A). This analysis revealed that there was a dose-dependent effect of the conditioned media on osteoclast activity. Interestingly, the response was biphasic, with a maximal effect at a 0.1% dilution of the conditioned media and an inhibitor effect at the highest dose. Using the pit formation assay, we observed a biphasic effect of the conditioned media, but the peak stimulation in the number of pits per osteoclast was at a dilution of 1% conditioned media (Fig. 2,B). Again, the highest dilution was inhibitory. A similar pattern emerged when the number of osteoclasts per bone slice was assessed with the peak concentration at 0.1% conditioned media (Fig. 2,C). Having ascertained that MDA MB 231 cells produced a substance or substances that stimulated osteoclast activity, the conditioned media were assayed for the presence of a number of cytokines and growth factors (Table 1). Significant levels of IL-1, IL-6, M-CSF, GM-CSF, TGF-β1, TGF-β2, TNF-α, IGF-II, and PTHrP were measured in the conditioned media.

The above list contained many factors that could be working either alone or in combinations to stimulate osteoclast activity. To further define the list of candidate osteoclast stimulatory factors, MDA MB 231-conditioned media were passed over a molecular sieve column under nondenaturing conditions. The resultant fractions were sterile filtered and assessed for effects on osteoclast activity. The pattern of effects on the total amount of collagen peptide released into the media suggested that there were regions that stimulated resorption and, interestingly, regions that repressed osteoclastic activity (Fig. 3,A). When the number of pits per osteoclast was examined, there were several fractions that stimulated bone resorption but no regions that appeared to inhibit the number of pits formed per cell (Fig. 3,B). When the number of osteoclasts per slice was examined, fractions that had elevated collagen-releasing effects but no effect on the number of pits per slice contained more cells per slice (Fig. 3,C). Similar to the pattern observed when the amount of collagen peptide released was determined, there were fractions that had fewer osteoclasts per slide than control cultures. These assays have indicated that there were stimulatory and inhibitory conditioned media fractions. Both stimulatory and inhibitory fractions were assayed for the presence of the same growth factors that were identified in Table 1. As detailed in Table 2, TNF-α and IGF-II were present in fractions that stimulated resorption. GM-CSF, M-CSF, IL-6, TGF-β2, and PTHrP were present in the fractions that inhibited osteoclast activity. No IL-1 or TGF-β1 was measured in any of the active fractions. All of the active fractions were examined for cytokine and growth factor levels, and several of the active fractions contained no detectable levels of any of the factors examined.

We examined the effects of the above identified factors on osteoclast resorption activity (Fig. 4). IGF-II, PTHrP, and TNF-α each stimulated resorption activity by all of the parameters measured here. TGF-β2 stimulated the number of pits per osteoclast and the number of osteoclasts per slice, but the total volume of collagen peptide released was not significantly altered with treatment. In contrast, treatment with GM-CSF significantly inhibited resorption activity as measured by determining the number of pits per osteoclast, whereas there seemed to be no significant effects on the number of osteoclasts per slice or on the total amount of collagen released. IL-6 treatment increased the number of pits per slice; yet there was no significant effect on either the amount of collagen peptide released or on the activity per osteoclast. We were unable to detect any significant effect of M-CSF on resorption activity during the 24-h treatment period by any of the criteria measured here. Having measured effects of these growth factors on resorption activity, we next determined whether the factors likewise influenced lysosomal enzyme secretion. As demonstrated in Fig. 5, IGF-II, PTHrP, TNF-α, TGF-β2, and IL-6 stimulated cathepsin B and TRAP secretion when results were normalized for respective sample cell numbers. Treatment with GM-CSF or M-CSF had no effect on the secretion per osteoclast of cathepsin B or TRAP.

Having assessed the effects of treatment with individual growth factors on osteoclast resorption and lysosomal enzyme activity, we explored the influences of the combinations of growth factors present in the conditioned media fractions that had either stimulatory or inhibitory effects (Table 3). The combination of IGF-II, TGF-β2, PTHrP, and TNF-α approached the stimulatory level of 0.1% conditioned media with respect to effects on bone resorption and lysosomal enzyme secretion. Interestingly, the combination of GM-CSF, M-CSF, IL-6, PTHrP, and TGF-β2 inhibited these same parameters. Combining stimulatory IGF-II and TNF-α with these inhibitory factors resulted in stimulatory activity, but the level of stimulation did not approach that of the diluted conditioned media.

The data presented here demonstrate that breast cancer cells secrete multiple growth factors that have the ability to stimulate osteoclast-mediated bone loss. We have shown that all of the cell lines examined secrete factors that stimulate osteoclast resorption activity. For the remaining studies, we elected to examine the MDA MB 231 cell-conditioned media because these have proven to be highly metastatic to bone using an animal model system pioneered by Nakai et al. (19). We have examined three different resorption parameters for these studies. Quantitation of the total amount of bone removed was achieved by determining the amount of collagen peptide released into the media during the resorption process. This assay detects the total amount released whether it is due to increased numbers of osteoclasts, increased number of pits generated by each osteoclast, or increased pit size. Analysis of the number of pits per osteoclast indicated the activity per cell, and the calculation of the number of osteoclasts per bone slice indicated the number of cells that were present. Changes in this last parameter could be due to a number of different effects including decreased apoptosis or increased binding to bone. Resolution of the mechanisms by which the number of osteoclasts present were altered is not revealed by these studies and remains to be resolved with further experimentation. The conditioned media effects were seen at surprisingly low concentrations and were biphasic. Our studies revealed that all of the measures of osteoclast activity, including the total amount of bone removed, the activity of each osteoclast, and the number of osteoclasts bound to the bone, exhibited this biphasic response. There are several possible reasons for this; among the possibilities: depletion of important factors by using spent media or a toxic metabolic waste build up. Depletion of important factors is unlikely given the identification of inhibitory fractions after chromatography in fresh media. Any accumulated toxic metabolic products would elute as small molecules in late fractions. The inhibitory fractions were in midrange (estimated sizing between M_r 40,000 and 100,000), indicating that this is not a likely explanation either.

Not surprisingly, there were many growth factors and cytokines present in the conditioned media. To better define the growth factors in the conditioned media that were active in stimulating osteoclast activity, the media was fractionated using an approach that was not disruptive to native protein conformations and interactions. Analysis of these fractions with the three different resorption activity measurements has revealed an interesting pattern of responses that varies according to which resorption criteria was examined. As outlined in Table 4, fractions 3, 4, and 5 increased the total amount of bone removed and the activity of each osteoclast. In these fractions, the number of osteoclasts per slice was decreased; thus, the stimulation in resorption activity is likely to be due to the elevation in activity of the individual osteoclast. In contrast, fractions 6 and 7 significantly stimulated the number of cells per slice, whereas there was no significant effect on the activity per osteoclast. Thus, the elevation in the amount of total bone excavated is likely to be, at least in part, the result of an increase in the number of cells on the bone slices, counterbalancing the lack of any effect on the activity per osteoclast. Because fraction 6 contained TNF-α, we examined the effects of TNF-α on osteoclast numbers. The data presented here supports the theory that TNF-α elevates the number of osteoclasts found on bone after short-term treatment, and we are presently pursuing the mechanisms of this effect. Fractions 9 and 10 caused an increase in the activity of each osteoclast, whereas the number of osteoclasts per slice was decreased. These combined to slightly stimulate bone resorption. Fraction 10 contains IGF-II, and our data demonstrate that the major influence of IGF-II is on the activity of individual osteoclasts, supporting the theory that IGF-II may be important in tumor-driven stimulation of osteoclast activity. When the amount of collagen peptide released was examined, fractions 23 and 24 stimulated bone resorption. In these fractions, there was no effect observed on the activity level of the osteoclasts and very small stimulation in the number of osteoclasts present. It may be that the elevation in collagen peptide released in these samples was due to each osteoclast generating a larger resorption pit. Interestingly, fractions 11 through 20 inhibited the amount of bone removed. In these fractions, the activity of each osteoclast was stimulated, whereas the number of osteoclasts was depressed. There are many growth factors present in these fractions, and our data demonstrate that the interactions of these growth factors results in repressed bone resorption and lysosomal enzyme secretion. The observed decrease in total bone loss may be the result of a decrease in osteoclast binding, an elevation in apoptosis, or may also be due to shallower pits being generated. The precise nature of this observation and the interactions of these growth factors will also require further study. This effect of repressing resorption by these fractions may at least in part explain the biphasic nature of the dilution curve observed above. If substances are present in the media that repress resorption, the higher concentrations of them could repress activity despite the presence of the stimulatory agents.

We examined all of these fractions to determine the presence of IL-1, IL-6, M-CSF, GM-CSF, TGF-β1, TGF-β2, TNF-α, IGF-II, and PTHrP. In the stimulatory fractions, we detected TNF-α in fraction 6 and IGF-II in fraction 10. None of the other fractions, whether stimulatory or inhibitory contained these factors. As indicated in Table 2, some of the inhibitory fractions contained GM-CSF, M-CSF, IL-6, TGF-β2, and PTHrP. Other inhibitory fractions contained no detectable levels of the factors examined. This is intriguing, and we are presently pursuing the content of these fractions with more extensive studies. None of the fractions contained detectable levels of TGF-β1 or IL-1, which suggests the possibility that these factors were diluted by the fractionation sufficiently to be below the detection limits of the assays (TGF-β1: <7 pg/ml; IL-1: 0.5 pg/ml) or they were present in untested, therefore inactive, fractions.

Because several of the factors present in the inhibitory fractions have been shown to stimulate osteoclastic resorption, we examined whether the factors identified in the stimulatory fractions were capable of stimulating osteoclasts in our system. Our studies revealed that human IGF-II and TNF-α both stimulated the activity of the avian osteoclasts. Because these were identified in two of the stimulatory fractions, it seems likely that these factors are breast cancer-derived factors that are involved in stimulating osteoclast-mediated bone resorption. Hou et al. (7) have demonstrated that purified rabbit osteoclasts have IGF-I receptors, bind IGF-I with high affinity, and respond to IGF-I treatment with decreased apoptosis. In contrast with these finding, others have shown that IGFs either have no effect or stimulate only if osteoblasts are present (20, 21). TNF-α stimulates osteoclast differentiation, but the effects on mature cells have not been extensively studied (3, 4). TNF-α receptor 1 knockouts seem to have normal bone, which suggests that TNF-α has little role in normal bone development but does not preclude a role in pathological bone loss (22). IGF-II, TGF-β, PTHrP, and TNF-α each individually stimulated osteoclast activity. The addition of these factors together (in concentrations similar to that found in diluted 231-conditioned media) stimulated osteoclast activity to a level approaching that of the diluted conditioned media.

Interestingly, the effects of the factors identified in the inhibitory fractions were more complex. Individually, several of the factors stimulated the activity of the isolated osteoclasts, whereas neither GM-CSF nor M-CSF stimulated activity. Indeed, GM-CSF decreased the activity of individual osteoclasts. There is little data available on GM-CSF effects on mature osteoclast activity, but M-CSF has been implicated in suppressing apoptosis (23). IL-6 significantly increased the number of osteoclasts per slice, whereas it had no effect on the activity per cell or the amount of collagen peptide released. Because the major effects of IL-6 seem to be on osteoclast differentiation and we are examining highly purified mature osteoclasts, a lack of effect of IL-6 on resorption activity is not surprising (24). In seeming contradiction to these results, it has been shown that mature osteoclasts have IL-6 receptors and that IL-6 reverses calcium-induced decreases in bone resorption (25). Direct effects of PTHrP on osteoclast activity have not been reported, but PTH, which uses the same receptor, seems to directly alter F actin distribution and cytosolic pH (8, 26). TGF-β influences on osteoclast activity are somewhat mixed, with demonstrations of stimulation and inhibition of resorption and also a stimulation of apoptosis (27, 28). Thus, studies of direct growth factor effects on resorption activity are in their infancy, inasmuch as highly purified authentic cells are not routinely used for these studies, and receptor identification studies are just now being undertaken. In our studies, the combination of the factors found in the fractions that decreased bone resorption (GM-CSF, M-CSF, IL-6, PTHrP, and TGF-β) were inhibitory, supporting the idea that the inhibitory factors present were able to overcome the stimulatory effects of some of the components of the mixture. When IGF-II and TNF-α were added to this inhibitory mixture, there was stimulation of resorption activity, but the level did not approach the stimulatory effects of the conditioned media. This leads us to conjecture that there are other stimulatory factors being secreted by the tumor cells that we have not identified. This possibility is strengthened by the presence of stimulatory fractions in which we were unable to detect the factors we are studying.

Taken together, the data presented here demonstrate that metastatic breast cancer tumors are likely to produce multiple factors that have diverse effects on osteoclast bone resorption activity. The effect of some of the secreted factors in suppressing bone resorption was unexpected, but our data clearly show that the overall effect of the combination of inhibitory and stimulatory factors was stimulatory. This is based on both the conditioned media studies and the effects of combined stimulatory and inhibitory purified growth factor studies. These data support the possibility that IGF-II and TNF-α are likely to be key factors secreted by metastatic breast cancer tumors responsible for stimulating bone resorption activity.

The tireless work of Ryan Sportel in the resorption and lysosomal enzyme assays is gratefully acknowledged.