Bone is a common site of cancer metastasis. Breast, prostate, and lung cancers show a predilection to metastasize to bone. Recently, we reported that the chemokine interleukin 8 (IL-8) stimulates both human osteoclast formation and bone resorption. IL-8 mRNA expression was surveyed in a panel of human breast cancer lines MDA-MET, MDA-MB-231, MDA-MB-435, MCF-7, T47D, and ZR-75, and the human lung adenocarcinoma cell line A549. IL-8 mRNA expression was higher in cell lines with higher osteolytic potential in vivo. Human osteoclast formation was increased by MDA-MET or A549 cell-conditioned medium, but not by MDA-MB-231. Pharmacologic doses of receptor activator of nuclear factor-κB (RANK)-Fc or osteoprotogerin had no effect on the pro-osteoclastogenic activity of the conditioned medium; however, osteoclast formation stimulated by conditioned medium was inhibited 60% by an IL-8-specific neutralizing antibody. The data support a model in which tumor cells cause osteolytic bone destruction independently of the RANK ligand (RANKL) pathway. Tumor-produced IL-8 is a major contributor to this process. The role of secreted IL-8 isoforms was examined by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry, which detected distinct IL-8 isoforms secreted by MDA-MET and MDA-231 cells, suggesting different pro-osteoclastogenic activities of the two IL-8-derived peptides. These data indicate that (a) osteoclast formation induced by MDA-MET breast cancer cells and A549 adenocarcinoma cells is primarily mediated by IL-8, (b) cell-specific isoforms of IL-8 with distinct osteoclastogenic activities are produced by tumor cells, and (c) tumor cells that support osteoclast formation independent of RANKL secrete other pro-osteoclastogenic factors in addition to IL-8.
Bone is a common site of cancer metastasis. Several tumors show a particular predilection for metastasis to bone, including breast, prostate, lung, thyroid, and renal cancers (1). Of the 4 million people who die in the United States each year, approximately one quarter die from cancer and 70% of these have either breast, lung, or prostate cancer (1). Consequently, there are >350,000 people in the United States who die each year with bone metastases (1).
Tumors cause two distinct but overlapping types of skeletal lesions when they spread to bone, either lytic or blastic, with many tumors demonstrating the pathologic features of both (1). After tumor cells find their way to the bone marrow, there are numerous cell types present in the microenvironment that are involved in the maintenance of immune and inflammatory responses and whose activity may be modulated by the cytokines and/or growth factors secreted by tumor cells (1). The progression of osteolytic bone metastases requires the establishment of functional interactions between metastatic cancer cells and bone cells (2), which are mediated by soluble regulators of osteoclast formation, activity, and survival (1).
The number of osteoclasts is increased at metastatic sites in breast cancer patients (3). Tumor cells can interact with osteoblasts, which, in turn, stimulate osteoclasts to differentiate from hematopoietic precursors in the bone marrow, resulting in increased osteoclastic bone resorption (4). Cancer cells are known to produce a variety of stimulators of bone resorption, such as parathyroid hormone-related protein (PTHrP), interleukin 1 (IL-1), IL-6, IL-8, IL-11, colony-stimulating factor-1 (CSF-1), granulocyte macrophage-CSF, transforming growth factor (TGF)-β1, TGF-β2, tumor necrosis factor-α, insulin-like growth factor II, and hepatocyte growth factor (5–14). The secretion of some but not all of these factors by cancer cells regulates the expression of receptor activator of nuclear factor-κB ligand (RANKL) on the surface of stromal osteoblasts, thereby increasing osteoclast-mediated bone resorption (4, 5). Osteoclasts are also stimulated by tumor products (15, 16), but not usually by direct tumor cell secretion of RANKL (4). The bone marrow microenvironment is further enriched by growth factors released during osteoclastic bone resorption, which may support proliferation and growth of tumor cells or alter their phenotype in bone (17). In addition, tumor-derived products, such as dkk-1, can also inhibit osteoblast differentiation, thus also contributing to bone loss (18).
Recently, we reported that the α chemokine IL-8 is a potent and direct activator of osteoclastic differentiation and bone resorption (2). The mechanism of action of this chemokine does not require activation of the RANKL pathway, but involves the expression and activation of the specific IL-8 receptor (CXCR1) on the surface of osteoclasts and their precursors (2).
The aggressive human breast cancer line MDA-MET secretes high levels of IL-8, forms large bone tumors in vivo, and induces extensive osteolysis following intracardiac injection of cells into nude mice, whereas MDA-MB-231 cells, obtained from the American Type Culture Collection (ATCC, Manassas, VA), secrete little or no IL-8, and fail to colonize the skeleton (5). A549 lung adenocarcinoma cells also form lytic tumors following intracardiac injection. The ability of MDA-MET and A549 cells to cause bone metastases in vivo and to support osteoclast development in vitro was examined, with the goal of defining the mechanism of tumor-stimulated osteolysis. Our data show that tumor growth mediated by osteolysis can be substantially driven by IL-8, as well as other non-RANKL tumor-secreted factors.
Materials and Methods
Reagents. Weak cation exchange protein chips and other surface-enhanced laser desorption/ionization (SELDI)–related materials were obtained from Ciphergen Biosystems (Fremont, CA). Tissue culture plastics were supplied by Falcon (Lincoln Park, NJ). All other analytic grade reagents were purchased from Sigma (St. Louis, MO) or Fisher (Springfield, NJ). All tissue culture media and reagents were supplied by Life Technologies, Inc. (Grand Island, NY). Recombinant human RANKL, recombinant human RANK-Fc chimera, recombinant murine macrophage CSF-1, anti-mouse RANKL antibody, and recombinant human IL-8 were purchased from R&D Systems (Minneapolis, MN). Accu-Prep used for isolation of human mononuclear cells was purchased from Accurate Chemical and Scientific Corporation (Westbury, NY).
A mouse monoclonal IL-8 antibody used in in vitro osteoclast cultures was kindly provided by Dr. Rakesh Singh (University of Nebraska Medical Center, Omaha, NE). Additional anti IL-8 antibodies used in the in vitro osteoclast cultures were as follows: rabbit polyclonal IL-8 antibody (Abcam, Cambridge, United Kingdom), mouse monoclonal IL-8 antibody, and goat polyclonal IL-8 antibody (R&D Systems). Goat anti-mouse secondary antibodies were obtained from Pierce (Rockford, IL).
Cell lines and culture conditions. The MDA-MB-231 cells and MDA-MET cell lines were maintained in DMEM and supplemented with 10% fetal bovine serum (FBS) at 37°C in sterile culture dishes (as described; ref. 5). MDA-MET cells were derived from a weakly osteolytic MDA-MB-231 variant by in vivo selection (5). MDA-MET cells produce osteolytic lesions (100%) within 4 weeks of inoculation in the circulation or tibia of athymic nude mice (5) and grow effectively in the mammary fat pad (19). Breast cancer cell lines (MCF-7, T47D, BT549, parental MDA-MB-231, MDA-MB-361, MDA-MB-435, MDA-MB-436, and ZR75-1) were obtained from ATCC and maintained as described above. Two variants of MDA-MB-231 cells, indicated Sa and Ky, were obtained from Dr. T. Yoneda (University of Texas Health Science Center, San Antonio, TX) and Dr. V. Vetvicka (University of Louisville, Louisville, KY), respectively. MDA-MB-231 Sa cells are those used by Guise (6) to show the role of PTHrP via neutralizing monoclonal antibody. MDA-MB-231 Ky cells have been tested in vivo and produce much less aggressive osteolytic lesions and behave like conventional ATCC cells. The other cell lines are those reported by Yin et al. (20) and have also been tested for individual bone metastases frequencies via the intracardiac model.
A549 is a human lung adenocarcinoma cell line with type II pneumocyte characteristics (21). The cells were obtained from the ATCC and transfected with pcDNA3 empty vector. Single cell clones resistant to G-418 antibiotic were isolated by limiting dilution. Cell lines were maintained in Ham's F12K plus 2% glutamine, supplemented with 10% FBS at 37°C in sterile culture dishes.
All cell lines used were tested to be Mycoplasma-free and subcultured by trypsinization in 0.5% trypsin (Sigma) and 0.5 mmol/L EDTA in HBSS without calcium or magnesium in a laminar flow hood during their logarithmic phase of growth.
Forty-eight-hour conditioned medium (containing serum) from MDA-MB-231, MDA-MET, and A549 cells were collected, diluted 50% in α-MEM (MDA-MET and MDA-231) or Ham's F12K (A549), and added to cultures of human peripheral blood mononuclear cells (PBMC; as described below).
In vitro proliferation of MDA-MET and MDA-MB-231 cells was determined using the 2,3-bis[2-methoxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide inner salt (XTT) assay (Sigma), which measures the mitochondrial activity of living cells as a readout for cell growth. Both cell lines were plated in six-well multiwell dishes (50,000 cells per well), and proliferation was measured daily using the XTT assay according to the instructions of the manufacturer.
Direct intratibial injection of tumor cells. MDA-MB-231 and MDA-MET cells were grown to subconfluence. Fresh medium was added 24 hours before injection. On the day of the injections, cells were harvested with 0.2% EDTA and 0.02% trypsin, washed twice in PBS, and resuspended in PBS at a concentration of 106/mL. Four-week-old athymic female nude mice were purchased from Taconic Farms (Taconic, NY) and housed in an approved animal facility with protocols approved by the University of Arkansas for Medical Sciences Institutional Animal Care and Use Committee.
Before injection, the animals were deeply anesthetized with a 1:1:4.6 solution of xylazine/ketamine/PBS (administered i.p. at 0.033 mL mixture/10 g body weight). The mice were placed in a prone position. After gently palpating the bones in the legs, the upper end of the tibia was identified as the site for injection. Ten microliters (10,000 cells) were injected into the right tibia of nude mice using a 50 μL syringe and 28 gauge needle. PBS only was injected in the left leg as control. After injection, the mice were placed on a heating pad to recover from anesthesia and then returned to their cages.
The mice were followed for a period of 4 weeks, sacrificed, and X-rayed (Kodak X-Omat, Kodak, Rochester, NY) as described below. Dissected legs were processed for histopathology.
Tumor growth in bone was monitored by weekly X-ray of deeply anesthetized nude mice (4 weeks old at time of injection) using an AXR minishot 110 X-ray cabinet (Associated X-ray Corporation, East Haven, CT) at 3 mA, 33 kV for 20 seconds using Kodak X-Omat TL film, and processed on a Kodak X-Omat RP automated film processor.
Radiologically affected and unaffected long bones were excised, fixed in 10% neutral-buffered formalin for 2 days, and decalcified in 5% formic acid with agitation until deemed clear by the ammonium oxalate end point test (5). The decalcified specimens were then dehydrated through graded ethanol, cleared in methyl salicylate, embedded in paraffin, sectioned (5 μm), and stained with H&E or for tartrate-resistant acid phosphatase (TRAP) as described previously (5).
RNA isolation and analysis. Total RNA was isolated from the tumor cell lines in vitro using the Qiagen RNeasy Midi kit (Qiagen, Inc., Valencia, CA) according to the instructions of the manufacturer. Extracted RNA was quantitated by spectrophotometry and tested for integrity by agarose gel electrophoresis as described previously (5). RNA from all cell lines was analyzed for PTHrP and IL-8 (5) expression by reverse transcription-PCR (RT-PCR) with human-specific primers designed from Genbank files BC013615 (IL-8) and J03580 (PTHrP). Human PTHrP: 364 bp amplicon, 55°C annealing temperature, sense primer: 5′-CTCAGGAAACTAACAAGGTGG; antisense: 5′-GCAGTTTCATAGAGCAATGG. Human IL-8: 384 bp amplicon, 60°C annealing temperature, sense primer: 5′-CTCTCTTGGCAGCCTTCCTG; antisense: 5′-CAACCCTACAACAGACCCAC as described previously (5).
Isolation and culture of peripheral blood mononuclear cells and differentiation to osteoclasts. Peripheral blood was collected from healthy donors approved by the University of Arkansas for Medical Sciences Institutional Review Board with heparin anticoagulant, in the presence of 200 ng/mL RANK-Fc, to minimize any priming of osteoclast progenitors by endogenous RANKL as described (2). Blood was diluted in sterile PBS (1:1) in a sterile hood. The blood-PBS solution was slowly layered over Accu-Prep solution (Accurate Chemical and Scientific) and then centrifuged at 400 × g in swing buckets for 30 minutes at 21°C. The PBMC layer was collected and washed in five to six volumes of PBS, isolated by centrifugation at 140 × g, and resuspended in α-MEM containing 10% FBS. Cells were counted with a hemocytometer and plated in 48-well tissue culture plates at a concentration of 0.5 million cells in 0.5 mL volume per well. Macrophage CSF-1 (25 ng/mL) was added to all groups. RANKL (25 ng/mL) was used as a positive control. Cultures were maintained at 37°C and half-feeds were done thrice per week and terminated on the day 10. Medium was aspirated and the cells were fixed with 10% formalin. TRAP staining was done (Sigma) for quantitation of TRAP-positive multinucleated cells. TRAP-positive cells having more than three nuclei were counted as osteoclasts in the entire well with four wells per treatment. Cell counts were averaged and the results were expressed as the number of TRAP-positive multinucleated cells per well per treatment group (n = 4 per treatment; ref. 2).
Surface-enhanced laser desorption/ionization time of flight mass spectrometry analysis. IL-8 expression in serum-free conditioned medium was evaluated by SELDI time-of-flight mass spectrometry (SELDI-TOF-MS). Approximately 50 μL of 24-hour conditioned medium were mixed with an equal volume of binding buffer [100 mmol/L sodium acetate (pH 4.5); 0.2% Triton X-100 in PBS] and applied to a weak cation exchange chip (Ciphergen Biosystems). The chips were incubated overnight at 4°C with constant horizontal shaking to maximize interaction of the sample with the chip surface. Unbound protein/peptides were removed by washing thrice with buffer (5 minutes) under identical conditions. The chips were then rinsed, air-dried, and a saturated solution of energy-absorbing matrix (3,5-dimethoxy-4-hydroxycinnamic acid in 50% acetonitrile/5% trifluoroacetic acid) was added to each spot on the chip. All solution additions and washes were done using a Biomek 2 robotic liquid handling system (Beckman Coulter, Fullerton, CA). Each chip was analyzed in a Ciphergen SELDI PBSII reader as described (22).
In SELDI-TOF-MS analysis, a nitrogen laser (337 nm) desorbs the protein/3,5-dimethoxy-4-hydroxycinnamic acid mixture from the array surface, enabling the detection of the specific proteins/peptides captured by the array. The mass spectra of proteins/peptides in conditioned medium were generated using an average of 100 laser shots at a laser intensity of 220 to 240 arbitrary units. The mass-to-charge ratio of each of the proteins/peptides captured on the array surface was determined according to externally calibrated standards: human angiotensin 1 (1.29 kDa), [Glu1]fibrinopeptide B (1.57 kDa), porcine dynorphin A (209-225; 2.15 kDa), adrenocorticorticotropic hormone (1-24; 2.94 kDa), human β-endorphin (61-91; 3.47 kDa), bovine insulin (5.73 kDa), and bovine ubiquitin (8.56 kDa).
Statistical analysis. Data were analyzed by ANOVA with Bonferroni's post-test and by Student's t test. P < 0.05 between groups was considered significant.
MDA-MET cells form tumors following direct intratibial injection. MDA-MET cells produce tumors in bone following intracardiac injection into nude mice, whereas the clone of MDA-MB-231 cells we tested lacks this capacity (5). MDA-MET cells form large tumor foci with extensive bone destruction when injected directly into the tibia of nude mice (Fig. 1A). MDA-MB-231 cells, on the other hand, do not form tumors in vivo, lack the ability to proliferate in the tibia, and do not induce bone destruction (Fig. 1A). The development of the large MDA-MET tumor in this model is accompanied by increased osteoclastic bone resorption and increased recruitment of osteoclasts to the tumor-bone interface (Fig. 1B).
Because the MDA-MET tumor cells secrete high levels of IL-8, which is known to stimulate cell growth (5, 23), the effect of a neutralizing IL-8 antibody on MDA-MET and MDA-MB-231 cell growth in vitro was examined (Fig. 2). The effects of IL-8-neutralizing antibody on MDA-MET and MDA-MB-231 cell proliferation were evaluated by XTT assay as described previously (5). Cell proliferation (Fig. 2A and B) was unaffected by the antibody. Neither MDA-MET nor MDA-MB-231 cells express detectable amounts of the IL-8-specific receptor CXCR-1 by fluorescent-activated cell sorting (data not shown), suggesting that changes in tumor growth in vivo were unrelated to direct IL-8 effects on tumor proliferation (2) and associated with the effects of IL-8 on the bone microenvironment.
Interleukin-8 is expressed by multiple tumor cell lines. Because a general role for IL-8 in cancer bone metastases has not been established and to exclude the possibility that IL-8 expression was unique to MDA-MET cells, RNA from cancer cell lines was tested for IL-8 mRNA expression by RT-PCR. Extending our previous observations (5), Fig. 3 shows that breast cancer (lanes B-L) and a lung cancer line (lane M) commonly express IL-8 mRNA and/or PTHrP mRNA. MDA-MB-231Sa cells (lane D; ref. 6) express low levels of both IL-8 and PTHrP, and both mRNAs are increased by TGF-β treatment (lane E). The expression of IL-8 and PTHrP mRNA by MDA-MET cells (5) is shown for comparison (lane N). The osteolytic phenotypes in vivo and other characteristics of the breast cancer lines tested have been reported previously (20). MDA-MB-231Sa, MDA-MB-231Ky, BT549, and MDA-MD-435Sa cell lines cause osteolytic lesions in vivo. Normal breast tissue (lane A) expresses PTHrP as expected and low levels of IL-8. Prostate cancer cell lines PC3 and LnCaP also express IL-8 (data not shown).
Because lung cancers commonly cause bone metastases (1), the A549 lung adenocarcinoma cell line was examined further. A549 cells expressed PTHrP and IL-8 mRNAs in amounts similar to those made by MDA-MB-231 cells (Fig. 3, lane M). By PCR, these cells also expressed mRNAs for adrenomedullin, IL-11, and platelet-derived growth factor B chain (data not shown). Together, these data show that IL-8 expression is common in tumor cell lines and may correlate with their osteolytic phenotype in vivo (5, 20).
When A549 cells were inoculated into the left cardiac ventricle, osteolytic bone lesions were observed in 10 of 10 mice by Faxitron X-ray at 9 weeks (data not shown). Parallel results were obtained with mice receiving another independent A549 cell clone (data not shown). Histologic examination of the same bone lesions (data not shown) revealed abundant bone-resorbing multinucleated osteoclasts at the tumor/bone interface, consistent with a mechanism in which the tumor cells secrete locally high concentrations of osteoclastogenic factors, such as IL-8.
Based on these data and our previous studies (2, 5), we postulated that the difference in the in vivo colonization and growth in bone of both the A549 and MDA-MET tumor cells was due to differences in their ability to stimulate osteoclast formation via IL-8, prompting us to study the effect of A549, MDA-MB-231, and MDA-MET conditioned medium on the stimulation of osteoclast formation in vitro.
Conditioned medium from A549 and MDA-MET but not MDA-MB-231 cells supports osteoclast differentiation. Forty-eight-hour conditioned medium (containing serum) from MDA-MB-231, MDA-MET, or A549 cells diluted 50% in α-MEM (MDA-MET and MDA-MB-231) or Ham's F12K (A549) were added to cultures of human PBMCs. MDA-MET and A549 conditioned medium stimulated osteoclast formation, whereas MDA-231 conditioned medium did not (Fig. 4). These data suggest that MDA-MET and A549 cells secrete factor(s) that stimulate osteoclast formation, whereas MDA-231 cells make little or none of these factors or secrete factors that are inhibitory to osteoclast formation.
We next investigated whether the effect on osteoclasts of MDA-MET and A549 conditioned medium was due to IL-8 production by the tumor cells using commercially available IL-8-neutralizing antibodies at 200 μg/mL. The antibodies significantly reduced the number of osteoclasts induced by conditioned medium from MDA-MET (Fig. 5A) or A549 cells (Fig. 5B). The TRAP-positive multinucleated cells expressed αvβ3 integrin and resorbed bone (data not shown), confirming them as osteoclasts as we have shown previously (2).
However, IL-8 antibody treatment did not fully suppress osteoclast formation to control levels (Fig. 5), suggesting that MDA-MET and A549 cells support osteoclast formation by IL-8-dependent and IL-8-independent mechanisms. The IL-8-independent component may be related to the heparan sulfate fragments secreted by the tumor cells (19).
We next examined the IL-8-independent component of the pro-osteoclastogenic activity secreted by MDA-MET and A549 cells. RANKL, a central regulator of osteoclast formation and activity (24), was an obvious candidate factor for the IL-8-independent component. RANK-Fc (200 ng/mL), a soluble antagonist of RANKL (or osteoprotogerin; data not shown), was added to MDA-MET conditioned medium but did not block MDA-MET conditioned medium–induced osteoclast formation (Fig. 5C). However, the same dose of RANK-Fc efficiently blocked osteoclast formation stimulated by 50 ng/mL soluble recombinant RANKL (Fig. 5C). Similarly, the addition of RANK-Fc did not block A549 conditioned medium–induced osteoclast formation (Fig. 5D), whereas it did effectively block RANKL-stimulated osteoclast formation (Fig. 5D). These data suggest that secreted factor(s), other than RANKL, are responsible for the IL-8-independent effects on osteoclast formation.
Based on our previous microarray analysis (5), we tested MDA-MET cells for the secretion of known stimulators of osteoclast formation. The expression of TGF-β, IL-6, IL-1, and CSF-1 were examined by Western blot (data not shown) and did not differ significantly from their expression by MDA-MB-231 cells. The expression of PTHrP mRNA and protein by MDA-MET and MDA-MB-231 has been reported previously (5, 6, 25). The protein expression data agree completely with our published gene expression profiles of the MDA-MET and MDA-MB-231 breast cancer cells (5).
The results suggest that these factors are not responsible for the dramatically increased osteoclast formation caused by MDA-MET compared with MDA-MB-231 conditioned medium. The addition of specific antibodies to TGF-β, IL-6, or IL-1 to the conditioned medium from MDA-MET cells did not suppress osteoclast formation compared with the significant suppression by IL-8-neutralizing antibody (Fig. 5E). Collectively, these data suggest that the pro-osteoclastogenic activity of MDA-MET conditioned medium is associated with IL-8 and other, as yet unidentified, secreted molecules. Parallel lack of inhibition by antibodies other than against IL-8 was seen with A549 conditioned medium (data not shown).
Distinct IL-8 isoforms secreted by MDA-MET and MDA-MB-231 may explain the differences in osteoclastogenesis stimulated by these two closely related cell lines. Different amino terminal variants of IL-8 have been previously described (26, 27). Having shown the involvement of IL-8 in tumor osteolysis (2) and that the IL-8 secreted by MDA-MET cells (and A549 cells) was pro-osteoclastogenic (Fig. 5), we next sought to determine the identity of the IL-8 isoform secreted by MDA-MET and MDA-MB-231 cells. By using SELDI-TOF-MS, we examined serum-free conditioned medium protein expression. SELDI is based on the integration of chemically modified array surfaces with TOF-MS detection using the Ciphergen protein chip reader (model PBSII; ref. 22, 28, 29). This technology was selected because of its power of resolution, high reproducibility, ease of use, and femtomole detection limits.
Representative mass spectra for MDA-MET and MDA-MB-231 cell serum-free conditioned medium revealed significant differences in IL-8 isoforms (Fig. 6A). MDA-MB-231 cells secrete small amounts of predominantly an 8.4 kDa isoform (Fig. 6A) with minimal amounts of a higher molecular weight isoform (8.9 kDa) detectable. In contrast, the osteolytic MDA-MET cells secrete large amounts of an 8.9 kDa isoform (Fig. 6A) and little 8.4 kDa isoform (Fig. 6A). The addition of recombinant human IL-8 (R&D Systems; 200 ng/mL) to the conditioned medium from MDA-MET cells increased the intensity of both peaks (8.4 and 8.9 kDa; Fig. 6A), demonstrating that both peaks are IL-8 related and that commercially available rhIL-8 from R&D Systems also contains the two isoforms. Pretreatment of MDA-MET conditioned medium with an anti-IL-8-specific antibody eliminated both IL-8-related peaks (Fig. 6B). The SELDI-TOF-MS profiles of both MDA-MET and MDA-MB-231 conditioned medium were unaffected by the addition of protease inhibitors, thereby excluding the possibility that the truncated form of IL-8 secreted by MDA-231 cells is due to an over abundance (or activity) of a particular degradative enzyme (data not shown).
IL-8 is expressed by a number of cancer cell lines in vitro. A correlation is observed between tumor cell expression of IL-8 and metastatic potential (5, 23, 30–34). In addition, serum levels of IL-8 are increased in patients with breast cancer and seem to be an independent prognostic indicator for postrelapse survival (35, 36). Similar results have been reported for prostate cancer (37) and squamous cell carcinoma patients (38). Another study has shown that the combined administration of a neutralizing IL-8 antibody with an epidermal growth factor receptor antibody decreased metastases by human breast carcinoma xenografts in severe combined immunodeficient mice (33). In addition, breast cancer cells can induce the release of lysophosphatidic acid from activated platelets, which, in turn, promotes tumor cell proliferation and the lysophosphatidic acid–dependent secretion of IL-6 and IL-8, enhancing tumor growth and osteolysis (39). Collectively, these data and our published results (5, 40) suggest that IL-8 expression may be an important mediator of osteoclastogenesis and bone destruction once breast cancer cells are seeded in the bone microenvironment. The clinical correlation of IL-8 with both tumor aggressiveness and patient survival (35–38) supports an important contribution of IL-8 to osteolytic metastases (5, 40).
Further supporting the hypothesis that IL-8 is an important regulator of tumor growth in bone, three neutralizing anti-IL-8 antibodies were tested on osteoclast formation. All three anti-IL-8 antibodies significantly reduced osteoclast formation by either MDA-MET or A549 conditioned medium, suggesting that IL-8 is the major osteoclastogenic factor made by both MDA-MET and A549 cells. These tumor cells also make other RANKL-independent stimulators of osteoclast formation in addition to IL-8.
The major physiologic regulator of osteoclast formation is membrane-bound RANKL (24), which can be released in soluble form from tumor cells (24, 41) and can directly stimulate osteoclast formation in the absence of supporting stromal cells. However, the addition of the inhibitor RANK-Fc to medium conditioned by MDA-MET or A549 cells did not suppress osteoclast formation, demonstrating that soluble RANKL is not produced by MDA-MET or A59 cells and is thereby not responsible for their non-IL-8 osteolytic activity. These data also suggest that tumors that induce osteolysis, such as breast and lung, may not require RANKL for this activity. Data from a recent phase 1 study using a potent RANKL inhibitor, the decoy receptor osteoprotogerin (42), in cancer patients has shown that the potency for suppression of bone resorption is lower than anticipated. This may be associated with the elevated RANKL present in these patients, or based on the observations described here, reflect the secretion by human tumors of inducers of osteoclast formation, such as IL-8, that are independent of the RANKL pathway.
Factors such as IL-6, IL-1, PTHrP, and TGF-β have been reported to stimulate osteoclast formation directly from precursors (10, 40, 43–45). Western blots of MDA-MET and MDA-MB-231 conditioned medium were probed with antibodies against IL-1, IL-6, CSF-1, and TGF-β1. The factors were expressed equally by both cell lines, suggesting that the concentration of these factors in MDA-MB-231 conditioned medium was insufficient to stimulate osteoclast formation. Similarly, treatment of A549 cell conditioned medium with pharmacologic doses of RANK-Fc had no effect on osteoclast formation. These data clearly show that the tumor cell lines (MDA-MET and A549) secrete potent osteoclastogenic factors that do not include RANKL.
IL-8 is an autocrine stimulator of tumor cell proliferation and invasion (46, 47). However, we found that MDA-MB-231 and MDA-MET cell proliferation was not affected by IL-8 because the cells do not express the IL-8 receptors CXCR1 or CXCR2 (5). Collectively, these data suggest that IL-8 is not an autocrine growth factor for MDA-MET cells and that IL-8 acts in vivo on neighboring cells in the bone marrow microenvironment.
This intriguing idea is supported by the dramatic difference in MDA-MET and MDA-MB-231 growth following intratibial implantation. A major difference between the two related breast cancer cell lines is the enhanced ability of MDA-MET over parental MDA-MB-231 cells to stimulate osteolysis and grow in the skeleton although we cannot exclude an advantage in homing to bone for MDA-MET cells because unpublished observations6
T. Kieber-Emmons and L.J. Suva, unpublished observations.
MDA-MET and A549 cells have a growth advantage in bone, which is not due to a simple proliferative advantage but may be related to the ability of MDA-MET to stimulate osteoclastic bone resorption (2) via the secretion of IL-8 and other RANKL-independent factors. Osteolytic MDA-MET cells secrete full-length IL-8, whereas nonosteolytic MDA-MB-231 cells secrete a truncated (6-77) form of the protein. These observations imply that the secreted isoforms of IL-8 have distinct osteoclastogenic activities.
Different amino terminal variants of natural IL-8 have been previously described (26). Full-length IL-8 (1-77) and truncated IL-8 (6-77) have been identified as the major forms derived from endothelial cells or fibroblasts and leukocytes, although tumor-specific isoforms have not yet been identified. Additional isoforms that have been identified are 2 to 77, 7 to 77, 8 to 77, and 9 to 77 (27, 48). These specific isoforms are produced via enzymatic cleavage, with neutrophil gelatinase B (matrix metalloproteinase-9) being the only reported mediator (27). In ligand binding assays, the shorter forms of IL-8 have been shown to have slightly higher affinity than the full-length form (27), provided the Glu-Leu-Arg motif (found at position 9-11) remains intact (27). The relative activity of IL-8 isoforms on osteoclast formation and bone resorption has not been determined. The IL-8 isoforms secreted by metastatic versus primary human tumor cells remain unknown and are currently under investigation.
Tumor growth in bone depends on osteoclast activity to increase bone resorption and release growth factors sequestered in the bone matrix. Osteolytic tumor cells accomplish this by secreting factors that increase the number of mature osteoclasts and the activity of individual osteoclasts. An understanding at the molecular level of the mechanisms supporting and stimulating tumor osteolysis may provide important insight into more effective therapies for this devastating complication of cancer, including ones that target specific IL-8 isoforms.
Note: A.G. Margulies is a Virginia Clinton Kelley/Fashion Footwear Association of New York Research Fellow in Diseases of the Breast and B. Walser was the recipient of a National Cancer Institute Partners in Research studentship at University of Arkansas for Medical Sciences.
Grant support: NIH grant RO1 DK54044 (D. Gaddy); Arkansas Breast Cancer Research grant (L.J. Suva and D. Gaddy); for work at the University of Virginia: Gerald D. Aurbach endowment, Mellon Institute of the University of Virginia Cancer Center, and NIH grants RO1 CA69158/RO1 DK065837 and U.S. Army grants DAMD17-99-1-9401 (T.A. Guise) and DAMD17-98-1-8245/DAMD17-02-1-0586 (J.M. Chirgwin); and Spanish Ministry of Science and Education, BFU2004-02838/BFI (A. Martínez).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. David Findlay, the 2004 Webber Orthopaedic Scholar, for insightful discussions and critical review of the manuscript.