Zoledronic Acid Inhibits Both the Osteolytic and Osteoblastic Components of Osteosarcoma Lesions in a Mouse Model

Osteogenic sarcoma, commonly referred to as osteosarcoma (OS), is the most frequent primary malignancy of the skeleton, developing mainly before the age of 30 years (1, 2). Metastatic spread, preferentially to the lungs, is seen in 20% of presenting patients and is correlated with poor survival statistics (3–5). Bone lesions caused by OS are characterized based on their radiologic appearance and appear as either osteolytic, osteoblastic (osteosclerotic), or mixed (6). Osteolysis is a common manifestation associated with OS, even within predominantly osteoblastic lesions, and is mediated primarily by osteoclasts and their bone-resorbing activity (7, 8). Factors released from the bone are believed to stimulate tumor growth and tumor cells are in turn able to produce factors that stimulate osteoclast differentiation and activity. This results in the establishment of a host-tumor relationship, often termed “the vicious cycle” due to its progressively destructive nature (9). By mechanisms that are not well understood, tumor cells associated with osteoblastic lesions can stimulate osteogenesis (10, 11).

Bisphosphonates (BP) are commonly used for the prevention and treatment of various bone diseases associated with enhanced bone resorption, such as Paget's disease and osteoporosis (12, 13). In addition to their antiresorptive activity, there is growing evidence supporting the direct effects of BPs on cancer cells themselves, at least in vitro. In this respect, zoledronic acid (ZOL) exhibits the highest potency of its class. The antitumor activity of BPs has been shown in a wide variety of tumor cell types, including leukemia (14), breast cancer (15), prostate cancer (16), and OS (17, 18). These in vitro studies have shown that BPs can dose dependently inhibit proliferation and induce apoptosis in tumor cells. A reduction by BPs in tumor cell adhesion, invasion, and angiogenesis has also been reported, making BPs potentially attractive agents in the treatment of metastatic cancers (19).

Preclinical animal models of metastatic cancer have shown a reduction in tumor-induced osteolysis with ZOL treatment (20–22). Reports of animal models of prostate cancer have all shown reduced osteolysis with ZOL treatment, with conflicting results in osteoblastic lesions (23). Clinically, BP treatment is the current standard of care for palliative treatment of bone metastases (24). ZOL has proven to be effective in large nonrandomized clinical trials and is the first BP to show significant clinical benefit in patients with bone metastases from various primary tumors (19). Prolonged treatment with ZOL seems to be safe and well tolerated, and this combination of potency and safety makes it a useful adjuvant therapy in bone metastasis (25).

We report here that although ZOL treatment had a significant protective effect on OS-induced bone destruction, it did not inhibit primary tumor growth or reduce lung metastases and in some cases even promoted lung metastases in a xenograft mouse model that closely mimics the clinical outcome of patients with OS.

Cells and reagents. The human OS cell line K-HOS was obtained from the American Type Culture Collection. The KRIB (26) cell line was kindly provided by Dr. Berlin (Sahlgren University Hospital, Gothenburg, Sweden). Cells were cultured in DMEM supplemented with glutamine (2 mmol/L), penicillin (100 IU/mL), streptomycin (100 μg/mL), gentamicin (160 μg/mL), and 10% fetal bovine serum (Biosciences) in a humidified atmosphere containing 5% CO₂.

ZOL was generously provided by Novartis Pharma AG. The tetrapeptide caspase inhibitor zVAD-fmk was purchased from Calbiochem and geranylgeraniol (GGO) was purchased from MP Biomedicals.

Measurement of cell viability and apoptosis analysis. Cells (1 × 10⁴) were incubated for 72 h with increasing concentrations of ZOL (0.1-100 μmol/L). Cell viability was determined by staining with crystal violet and measuring absorbance at 570 nm wavelength. The experiments were done in quadruplicate and repeated at least thrice. Results of representative experiments are given as the mean ± SD. Apoptosis analysis, including 4′,6-diamidino-2-phenylindole (DAPI) staining, caspase activation, and caspase inhibition, was done as described previously (17).

Retroviral infection of K-HOS cells with the triple reporter gene construct SFG-NES-TGL. Luciferase-expressing K-HOS cells were generated using the retroviral expression vector SFG-NES-TGL (27). Virus particle–containing supernatants were generated and filtered to remove any cellular debris and then used to infect K-HOS cells, as described previously (28). The retrovirally transduced K-HOS cells were grown as bulk cultures for 48 h and subsequently sorted for positive green fluorescent protein (GFP) expression using fluorescence-activated sorting (FACS; Aria, BD Biosciences). The cells were allowed to proliferate and the 10% of cells expressing GFP most strongly were obtained by FACS to generate the subline K-HOS-NES-TGL.

Mouse model of OS. Four-week-old female BALB/c nu/nu mice were housed under pathogen-free conditions in accordance with the guidelines approved by the Institute of Medical and Veterinary Science animal ethics research committee. The left tibia was wiped with 70% ethanol and a 27-gauge needle coupled to a Hamilton syringe was inserted through the tibial plateau with the knee flexed, and 1 × 10⁵ K-HOS-NES-TG cells, resuspended in 10 μL PBS, were injected into the marrow space of the proximal tibia. As a control, all animals were injected with PBS in the contralateral tibia. Mice were randomly assigned into three groups of 10 animals each, and starting 1 wk after cancer cell implantation, 100 μg/kg body weight ZOL was given as weekly, or as a single dose, equivalent to the 4 mg i.v. dose given to patients. Similarly, untagged parental KRIB cells were injected intratibially, and mice were randomly assigned into vehicle (n = 6) or the ZOL (once only; n = 6) treatment groups.

In vivo bioluminescence imaging. Noninvasive, whole-body imaging to monitor luciferase-expressing K-HOS-NES-SFG-TGL cells in mice was done weekly using the IVIS 100 Imaging system (Xenogen). Mice were injected i.p. with 100 μL of the d-luciferin solution at 250 mg/kg body weight (Xenogen) and then gas anaesthetized with isoflurane (Faulding Pharmaceuticals). Images were acquired for 0.5 to 10 s and the photon emission transmitted from mice was captured and quantitated in photons/second using Xenogen Living Image (Igor Pro version 2.5) software.

Micro–computed tomography analysis. Both the right and left tibiae of each animal were mounted in the CT specimen tube and placed securely into the SkyScan-1072 X-ray micro–computed tomography (μCT) scanner. Scanning was commenced with magnification set to give scan slices of 17.8 or 5.2 μm, respectively. Three-dimensional images were generated using Cone-Beam reconstruction and three-dimensional visualization (SkyScan). Using the two dimensional images obtained from the μCT scan, the growth plate was identified and 750 sections, starting from the growth plate/tibial interface and moving down the tibia, were selected. Bone volume (mm³) was generated and compared with the control tibia for each animal.

Histology. Tibiae were fixed in 10% buffered formalin followed by acid decalcification in 10% EDTA solution and 7% nitric acid at room temperature. Samples were paraffin embedded, sectioned longitudinally at 6 μm, and stained with H&E. Analysis was done on a Nikon Eclipse TE300 inverted microscope (Nikon Corp.). Tumor size, defined as the tumor area, was calculated from the section of the tibia that best represents the center of the tumor mass and expressed as an average tumor area per group in absolute units (mm²).

Statistical analysis. The continuous outcome bone volume was analyzed using mixed-model ANOVA to allow for clustering of mice (i.e., more than one observation per mouse). Post hoc pairwise comparisons were undertaken with no adjustment made for multiple comparisons.

Effect of ZOL on cell proliferation and apoptosis of K-HOS cells. The effect of ZOL on cell proliferation and apoptosis was tested in the K-HOS human OS cell line. Treatment with ZOL for 72 hours resulted in a dose-related inhibition of cellular proliferation of K-HOS cells in monolayer cultures (Fig. 1A). Exposure to concentrations of ZOL >10 μmol/L resulted in marked cytotoxicity, with cells detaching from the substratum and cell death occurring within 48 to 72 hours. Morphologic evidence of apoptosis, including chromatin condensation, nuclear fragmentation, and the formation of dense rounded apoptotic bodies, was seen at doses greater than the half-maximal effective dose of 15 μmol/L (data not shown). ZOL-induced apoptotic cell death was concomitant with a dose-related increase in caspase-3–like activity (Fig. 1B). To investigate the role of caspase activation in the ZOL-mediated apoptosis of K-HOS cells, the effect of caspase inhibition on the induction of apoptosis was examined. Cells were incubated for 72 hours with 15 μmol/L ZOL either alone or in the presence of the broad specificity caspase inhibitor zVAD-fmk. Consistent with our previously published data in other OS cell lines (17), the inhibition of caspase activity failed to protect K-HOS cells from ZOL-induced cell death, suggesting that ZOL-induced apoptosis of K-HOS cells is independent of caspase activation (Fig. 1C). To examine the involvement of the mevalonate pathway on the ZOL-mediated apoptosis of K-HOS cells, we assessed the ability of GGO, an intermediate of the mevalonate pathway, to protect cells from ZOL-induced apoptosis. The addition of 300 μmol/L GGO significantly protected K-HOS cells from ZOL-induced apoptosis, whereas GGO alone had no effect on cell number (Fig. 1C). Lysates from control untreated cells, and cells treated with 30 μmol/L ZOL, were collected after 48 hours and analyzed by Western blotting for the presence of prenylated and the nonprenylated forms of the small GTPase, Rap1A. Untreated cells expressed only the prenylated form of Rap1A, whereas in the ZOL-treated cells both prenylated and unprenylated Rap1A were present (Fig. 1D). These results are consistent with the importance of the mevalonate pathway as an intracellular target for the effects of ZOL.

Effect of ZOL on the development, progression, and metastatic spread of human OS in a mouse model. A limitation in measuring tumor burden in bone is that it is not possible to accurately assess the progression of tumor growth by palpation, as in soft tissue tumors, before they break through the cortical bone. To overcome this limitation, we used noninvasive bioluminescence imaging (BLI), which enabled sensitive real-time in vivo tracking of OS growth in bone and its metastatic spread. The parental K-HOS human OS cells (29) were retrovirally infected with a triple-fusion protein reporter construct encoding herpes simplex virus thymidine kinase, GFP, and firefly luciferase. After infection, K-HOS cells were enriched for high-level expression of GFP by two rounds of fluorescence-activated cell sorting, thus generating the subline K-HOS-NES-TGL (Fig. 2A). The selected cells exhibited a 1,000-fold induction of luciferase activity when analyzed in vitro. When compared with the parental noninfected cells, K-HOS-TGL cells were equally responsive to ZOL treatment on both cell survival and kinetics of caspase activation (data not shown).

When injected into bone, these tumor cells produce mixed osteolytic/osteoblastic lesions directly in the area of injection. As in the human disease, K-HOS cells from the intraosseous tumors reproducibly form pulmonary metastases that are easily quantifiable with BLI, and by histologic inspection of the lungs, 5 to 6 weeks after cancer cell inoculation. K-HOS-NES-TGL cells were injected intratibially in 4- to 6-week-old BALB/c nu/nu female mice. Thirty animals were inoculated and animals were then assigned randomly into three groups: (a) the control untreated group, (b) the ZOL weekly treated group, and (c) the ZOL single dose. To allow establishment of tumor cells within the bone environment, ZOL treatment was commenced 7 days after cancer cell transplantation and ZOL was then administered s.c. at a dose of 100 μg/kg either once weekly until sacrifice or as a single dose. Results showed an exponential increase of photon emission associated with an increase in tumor burden, which was clearly evident from day 7 onwards in both the weekly and single dosing regimens (Fig. 2B). This trend was not significantly different from that of the vehicle-treated animals, indicating that ZOL treatment has no measurable effect on tumor burden at the primary site (Fig. 2C).

X-ray images of tumor-bearing tibiae taken just before sacrifice confirmed the presence of osteolysis in all animals within the control vehicle group. A characteristic radiodense “sunburst” configuration, consistent with spicular new bone formation, was also evident, extending perpendicularly to the periosteum and into the surrounding soft tissues (Fig. 2D). In contrast to radiography, which only provides two-dimensional images, μCT offers the advantage of generating three-dimensional images that can be quantified. Reconstructed three-dimensional images of μCT scans showed extensive bone remodeling, characterized by large osteolytic and osteoblastic activity in the tumor-bearing tibiae (Fig. 2D). Cross-sections of the μCT images of the normal non–tumor-bearing tibiae revealed a smooth surface of the cortical bone with an intact endocortical trabecular network. In contrast, significant cortical bone destruction and loss of trabecular architecture were apparent in all tumor-bearing tibiae, representing the osteolytic component of OS lesions. In addition, marked spicular new bone formation, extending from the periosteum, was clearly evident in all tumor-bearing tibiae, confirming the radiographic data (Fig. 2D). Histologic analysis showed that the spicular new bone was mineralized woven bone and was not in the tumor mass but was an extension of, and in contact with, the periosteum and cortex (Fig. 2D).

In animals treated with either the single or weekly dosing regimen of ZOL, qualitative assessment of bone architecture showed remarkable conservation of the tibiae, with no radiographic or μCT evidence of cortical or trabecular bone destruction (Fig. 3A). In addition, the periosteal new bone formation was significantly reduced by the single dose of ZOL and was almost absent in animals receiving the weekly dose regimen (Fig. 3A). Closer examination of the radiographic images showed that in all of the ZOL-treated animals, from both treatment regimens, but not in the vehicle-treated animals, the bones were highly radiodense, reflecting increases in mineral content, likely due, in these young growing animals, to ongoing bone formation and inhibition of bone resorption. This effect of ZOL and other BPs was striking in the distal femurs and proximal tibiae and was not restricted to the tumor site, being also observed in the contralateral non–tumor-bearing tibiae, as reported in several other studies (Fig. 3A and B; refs. 30–35).

In addition to the qualitative assessment of bone architecture, morphometric parameters were quantified from the three-dimensional μCT data generated after termination of the K-HOS experiments using the high-resolution desktop μCT. We compared the left tumor-bearing tibiae with the contralateral right non–tumor-bearing tibiae at a selected region, beginning at the growth plate and extending downwards 750 × 5 μm slices. This region encompassed all of the OS lesions. In the untreated animals, there was a mean net increase of 36% in the total bone volume in the left tumor-bearing tibiae in comparison with the contralateral right non–tumor-bearing tibiae of the same animals (L = 3.5 ± 0.4 mm³ versus R = 2.6 ± 0.48 mm³), which was a composite of osteolytic and osteoblastic activity by the tumors (Fig. 3C). With ZOL treatment, the total bone volume increased significantly in both the tumor-bearing and non–tumor-bearing tibiae when compared with vehicle (Fig. 3C). However, the percentage increase in bone volume induced by the tumors was significantly reduced in both the single and weekly ZOL dosing regimen, compared with the untreated control, suggesting that inhibition of osteoclastic activity by ZOL leads also to inhibition of tumor-induced osteoblastic activity (Fig. 3D). The effect of ZOL on trabecular bone parameters was also quantified from the μCT data. Table 1 shows that the derived histomorphometric parameters were consistent with the increased bone seen with ZOL, with increased trabecular number and bone surface and decreased trabecular space. Trabecular thickness was not affected by ZOL treatment. The effects of ZOL on trabecular structure were found in both the tumor-bearing tibiae and the contralateral non–tumor-bearing tibiae (Table 1).

At the end of the experiment, the lungs from each group of animals were excised and tumor burden, as a function of photon counts per second, was quantified ex vivo using BLI (Fig. 4A). The mean luciferase activity of lungs from mice treated with both the weekly and single dosing regimen of ZOL was significantly higher than that of the vehicle-treated animals (Fig. 4B). These results indicate that, in addition to ZOL having no effect on OS burden within bone, it actually increased the propensity of K-HOS cells to metastasize to lungs.

To substantiate these observations, we did a second independent experiment using parental untagged K-HOS cells. Twenty animals were inoculated and animals were then assigned randomly into two groups: (a) the control untreated group and (b) the ZOL weekly treated group. The rate at which the parental K-HOS cells established tumors in the tibia was 85% of injected animals (8 in the untreated control group and 9 in the ZOL-treated group). In the absence of BLI, successful implantation was manifested by visible tumors in all animals by 3 weeks after cancer cell transplantation, as tumors penetrated the cortex and extended into the surrounding soft tissue. Tumor burden at the primary site and metastatic spread to the lungs were assessed histologically. Tumor area was calculated from histologic sections of the tibia that best represented the center of the tumor mass and was expressed as an average tumor area per group in absolute units (mm²). Tumor volume was also measured from μCT analysis software, as described in Materials and Methods. As with the first experiment, ZOL treatment had no effect on tumor burden at the primary site (data not shown). However, multiple pulmonary macrometastases were detected in both groups. The frequency of animals with lung metastases was 50% (4 of 8) for the control vehicle group and 100% (9 of 9) for the ZOL weekly treated group (Fig. 4C). In addition, the size of lung metastases was significantly greater in the ZOL-treated group, with tumor area occupying 7% of the total lung area compared with only 2% in the control vehicle group (Fig. 4D).

To determine whether the observed increase in pulmonary metastases was a general phenomenon, or peculiar to the K-HOS cell line, we extended our studies using a second OS cell line (KRIB), described previously by Berlin et al. (26) as a highly metastatic line when transplanted intratibially. As observed with K-HOS cells, treatment with ZOL in vitro decreased KRIB cell proliferation in a dose-dependent manner, which was concomitant with a dose-dependent increase in caspase-3 activity (Fig. 5A). DAPI staining of nuclei confirmed induction of apoptosis with ZOL treatment (Fig. 5B). ZOL administration s.c. to KRIB tumor-bearing mice, once only at the clinically relevant dose of 100 μg/kg, inhibited both the osteolytic and osteoblastic components of KRIB lesions, as assessed qualitatively by three-dimensional μCT (Fig. 5C). However, as we observed with K-HOS cells, ZOL was without effect on tumor burden at the primary site of the tibia (Fig. 5D). Unlike K-HOS cells, treatment neither reduced nor increased significantly pulmonary metastasis formation in this model (Fig. 5D), suggesting that different OSs may respond differently to ZOL treatment in vivo.

In seeking to understand the molecular mechanisms by which ZOL is cytotoxic in OS cells, we have shown that the ZOL-mediated induction of OS cell apoptosis was due to the inhibition of protein prenylation and this was concomitant with an increase in caspase-3 activity. However, although caspase inhibition blocked the ZOL-induced activation of caspases, it did not prevent ZOL-induced apoptosis. These results are consistent with our previously published data on the in vitro cytotoxic activity of ZOL in a panel of human OS cell lines, in which we provided evidence that the induction of cell death of human OS cells by ZOL resembled “anoikis,” a special mode of apoptosis that occurs when adherent cells detach or lose particular attachment contacts with the extracellular matrix that confer survival signals to the cells (17).

OSs have a variable bone-forming ability but are destructive by virtue of their ability to expand in bone by inducing osteoclast-mediated bone resorption. The effect of BPs on OS cells has not been extensively studied, although studies with alendronate and pamidronate reportedly showed proapoptotic effects on canine (18) and human (36, 37) OS cells in vitro. Two recent studies have reported on the effect of ZOL in animal models of mouse and rat OS. For example, Ory et al. (38) used mouse OS cells administered i.v. to show that ZOL treatment suppressed lung metastases and prolonged the overall survival of OS-bearing mice. Heymann and coworkers (30) also showed that ZOL completely prevented lung metastases and caused reduction in osteolytic lesions in a rat model, in which intact rat OS tissue was implanted s.c. and in contact with the tibial bone surface. OS originates in the bone so that the clinical relevance of animal models of OS with ectopic tumor implantation, such as i.v. injections or s.c. implantation, is not clear. Unlike these studies, our experimental model, in which OS cells are transplanted directly into the bone marrow cavity, more closely simulates the normal progression of OS and allows for the direct evaluation of ZOL treatment of OS lesions normally seen in cancer patients. Here, we show that ZOL is able to inhibit the development and progression of not only osteolytic but also the osteoblastic component of primary OS lesions. New bone formation, the so-called “sunburst spiculation,” was seen in mice with xenografts after the tumor had breached the cortical bone. This is a distinctive feature of OS and is observed radiographically in ∼40% of OS (39). The type of reactive new bone as shown in this study did not take place within the soft tissue of the tumor mass but was an extension of, and in contact with, the periosteum and bone cortex. Although breaching bone integrity (e.g., by drilling) can result in ectopic bone formation, we have not seen such periosteal reaction when highly osteolytic breast cancer (MDA-MB-231; ref. 40) or multiple myeloma (RPMI-8226; ref. 41) tumors breach the cortex after intratibial injection. This indicates that this effect is not the result of periosteal contact by the tumors per se but is likely to be caused by factors produced specifically by OS cells acting on osteoblast precursors within the periosteum, leading to osteogenesis.

Despite the protective effects of ZOL on OS-induced bone destruction, ZOL treatment had no effect on tumor burden at the primary site, with either the weekly ZOL dosing regimen or the clinically compatible single dose. More importantly, ZOL failed to reduce lung metastases of KRIB cells and actually promoted lung metastases of K-HOS cells. In contrast to our observations, Dass and Choong (42) have recently shown, in a similar mouse model implanted with SaOS-2 cells, that ZOL administration twice weekly at 120 μg/kg resulted in primary tumor growth inhibition, reduction in lung metastases, and dramatic decrease in osteolysis. The reasons for these different observations are not clear but may relate to cell type and, in particular, the dose of ZOL administered, which in this case was ∼8-fold higher than the clinically compatible dose used in our study.

It remains unclear as to how ZOL might promote pulmonary metastases of K-HOS cells in this model. The tumor-bone microenvironment is a complex system influenced by many factors and cell types. We have shown that ZOL treatment results in increased amounts of bone at the tumor site and this is likely to alter the spatial relationship and interaction of tumor cells with the bone stroma. One could speculate that the ZOL-mediated increase in trabecular density and spatial constraints could “force” cancer cells to escape the marrow cavity, thus increasing metastatic spread. In an earlier study, we showed that ZOL treatment increased cell detachment from the substratum and provided evidence that the induction of cell death of human OS cells by ZOL resembled anoikis, a special mode of apoptosis that occurs when adherent cells detach or lose particular attachment contacts with the extracellular matrix (17). It is possible that the same may occur in vivo and that cancer cells exposed to ZOL may gain metastatic potential as they lose attachment. Many studies (14–16, 18, 43–46), including our own (17), have provided in vitro data to suggest that ZOL has a direct effect on tumor cells, as it inhibited tumor cell proliferation and induced apoptosis. However, whether cancer cells are exposed to similar levels of ZOL in vivo is a contentious issue. In contrast to some previous reports, the present study showed that ZOL did not decrease tumor burden at the primary site, suggesting that the protective effects of ZOL on bone were not mediated by a direct effect on tumor cells themselves but were the result of inhibition of bone resorption.

In conclusion, our data show that clinically relevant doses of ZOL, while protecting the bone from OS-induced bone destruction, do not inhibit primary tumor growth or reduce lung metastases but in some cases can promote lung metastases. Caution is therefore required if and when the clinical application of the BP class of antiresorptives is considered in OS.

No potential conflicts of interest were disclosed.

We thank Dr. Jonathan Green (Novartis, Basel, Switzerland) for helpful discussions, Dr. Peter Self (Adelaide Microscopy Centre) for his technical assistance in μCT analysis, and Prof. Peter Choong for his assistance with the intratibial injections.