In human prostate to bone metastases and in a novel rodent model that recapitulates prostate tumor–induced osteolytic and osteogenic responses, we found that osteoclasts are a major source of the proteinase, matrix metalloproteinase (MMP)-9. Because MMPs are important mediators of tumor-host communication, we tested the effect of host-derived MMP-9 on prostate tumor progression in the bone. To this end, immunocompromised mice that were wild-type or null for MMP-9 received transplants of osteolytic/osteogenic-inducing prostate adenocarcinoma tumor tissue to the calvaria. Surprisingly, we found that that host MMP-9 significantly contributed to prostate tumor growth without affecting prostate tumor–induced osteolytic or osteogenic change as determined by microcomputed tomography, microsingle-photon emission computed tomography, and histomorphometry. Subsequent studies aimed at delineating the mechanism of MMP-9 action on tumor growth focused on angiogenesis because MMP-9 and osteoclasts have been implicated in this process. We observed (a) significantly fewer and smaller blood vessels in the MMP-9 null group by CD-31 immunohistochemistry; (b) MMP-9 null osteoclasts had significantly lower levels of bioavailable vascular endothelial growth factor-A164; and (c) using an aorta sprouting assay, conditioned media derived from wild-type osteoclasts was significantly more angiogenic than conditioned media derived from MMP-9 null osteoclasts. In conclusion, these studies show that osteoclast-derived MMP-9 affects prostate tumor growth in the bone microenvironment by contributing to angiogenesis without altering prostate tumor–induced osteolytic or osteogenic changes. Mol Cancer Res; 8(4); 459–70. ©2010 AACR.
Prostate to bone metastases induce mixed lesions containing areas of extensive bone destruction (osteolysis) and formation (osteogenesis) that are mitigated by the principal cells of the bone, osteoclasts and osteoblasts, respectively (1). The highjacking of the normal bone remodeling process by the metastatic prostate cancer cells results in an increase in growth factors and cytokines that subsequently can stimulate tumor growth, thus generating what has been called a “vicious cycle” (2). Therefore, understanding the molecular mechanisms that facilitate the interaction between the multiple cell types in the tumor-bone microenvironment can provide new targets for therapies that will be effective in controlling and/or curing prostate to bone metastases.
Matrix metalloproteinases (MMP) are a family of 23 enzymes that collectively are capable of processing extracellular matrix components including those that comprise the bone matrix (3). Given their role in bone matrix resorption, it is not surprising that osteoclasts express a large repertoire of MMPs. However, recent analyses have identified that osteoclast-derived MMPs can control cell behavior in the tumor microenvironment by regulating the bioavailability/bioactivity of nonmatrix-related molecules, for example, receptor activator of nuclear κB ligand and kit ligand (4, 5). Despite their role in bone matrix synthesis, osteoblasts also express several MMPs (6, 7). However, at this juncture, no studies about the effect of osteoblast-derived MMPs or MMPs derived from other cellular sources on osteoblast behavior in the pathologic context of bone metastasis have been documented.
In the tumor microenvironment, MMPs are often induced in the host compartment in response to the cancer cells (3, 8). In previous studies, roles for tumor derived MMPs in the prostate tumor–bone microenvironment have been described (9, 10). However, little is known about the contribution of host-derived MMPs to tumor growth or tumor-induced osteolytic/osteogenic changes. The analysis of MMP expression in human samples of prostate to bone metastasis and in an animal model that recapitulates the human clinical scenario of prostate tumor–induced osteolytic and osteogenic change (4) revealed that osteoclasts are a major source of MMP-9. In the current in vivo study, we addressed whether host-derived MMP-9 affects prostate tumor growth or prostate tumor–induced osteolytic or osteogenic changes. Although MMP-9 contributed to prostate tumor growth, we found that this observation was not due to an effect on tumor-induced osteolysis or osteogenic response. In pursuing the mechanism, we observed that osteoclast-derived MMP-9 could contribute to tumor growth by promoting angiogenesis in the tumor-bone microenvironment.
Materials and Methods
All experiments involving animals were conducted after review and approval by the office of animal welfare at Vanderbilt University. Double null immunocompromised recombinase activating gene-2 and MMP-9 mice (c57BL/6 background) were generated as previously described (11, 12). Rat prostate adenocarcinoma tissue was provided by Dr. Mitsuru Futakuchi, Nagoya Medical School, Nagoya, Japan (13). Human samples of prostate to bone metastasis were provided by Dr. Robert L. Vessella, University of Washington, Seattle, WA. All reagents were obtained from Sigma-Aldrich except where specified.
Surgical Procedure and Measurement of Tumor Growth
Six-week-old immunocompromised wild-type (WT; n = 9) and MMP-9 null (n = 9) mice were anesthetized, and using a pair of scissors, a small incision between the ears was made in the dermis of the scalp. The subcutaneous tissue was separated from the underlying calvaria using a blunt-ended scissors to form a pocket. To promote tumor-bone interaction, the periosteum was removed from the calvaria to expose the calvarial bone. Equal volumes of rat prostate adenocarcinoma tissue (0.1 mm3) was inserted into the pocket, which was closed with wound clips. Tumor measurements with calipers (Fine Science Tools) were made on a weekly basis. The length, width, and height of the tumor were used to calculate the tumor volume. All animal studies were repeated on three independent occasions with similar sized groups.
In vivo Imaging of Tumor-Induced Osteolysis and Osteogenesis
To assess extent of tumor-induced osteolysis in three-dimensions, microcomputed tomography scanning (Siemens Preclinical) was used. Mice were imaged using X-ray tube settings of 80 kVp and 0.5 mA for 300 ms per projection, 360 projections at weeks 1, 2, and 3 posttumor transplantation. Data were reconstructed into three-dimensional images with voxel sizes of 0.1 × 0.1 × 0.1 mm.
With a different set of mice, microsingle-photon emission computed tomography (Micro-SPECT) imaging was also done to measure changes in bone formation at weeks 1, 2, and 3 posttumor transplantation. Each animal received 1 mCi of 99mTechnetium-Methylene DiPhosphonate through tail vein injection and 1.5 h later were subjected to micro-SPECT imaging on the NanoSPECT/CT system (Bioscan, Inc.). Subsequently, 24 projection views were acquired in a helical scan mode using a nine-pinhole (1.4 mm diameter) collimator on each of the four camera heads—a total of six scanner positions with 60-s acquisition per position. Tomographic images were reconstructed from the projection data using an ordered subsets expectation maximization algorithm provided with the scanner at an isotropic voxel size of 0.3 mm. To aid in interpretation of the micro-SPECT images, low-dose microcomputed tomography images were acquired along with each micro-SPECT image on the same scanner without moving the subject. X-ray tube settings were 45 kVp and 0.17 mA, and the data were reconstructed into images with 0.2 mm isotropic voxel size that were inherently registered with the micro-SPECT images based on the geometric calibration of the system.
For the quantitative analysis of bone formation and bone destruction in individual tumor-bearing WT and MMP-9 null animals over time, we took the following approach. The serial scans of individual animals at different time points were coregistered with each other, a process that was achievable due to the rigid structures of the skull such as the eye orbits and upper teeth staying constant over time. Using the Amira software (Visage Imaging), the images taken with micro-SPECT and microcomputed tomography were superimposed using an isosurface thresholding method, allowing for the visualization of the changes in the calvarial bone at each time point. Using a smaller region of interest in the registered images, bone volumes were determined by calculating the number of image voxels exceeding a chosen threshold. This threshold was kept constant from mouse to mouse allowing for comparative analysis. Individual micro-SPECT images were converted to percentage injected dose per gram body weight by dividing voxel values by the injected dose for that study, thus facilitating comparisons between animals. These radiotracer uptake ratios were used to normalize the data pertaining to bone formation between time points in individual animals and the animals in each group being studied.
Immunohistochemistry, Cytochemistry, and Histomorphometry
After sacrifice, rodent tissues were fixed overnight in 10% buffered formalin and decalcified for 3 wk in 14% EDTA at pH 7.4 at 4°C with changes every 48 to 72 h. Tissues were embedded in paraffin and 5-μm-thick sections were cut. Human prostate to bone metastasis samples (n = 10) were provided by Dr. Vessella. For MMP-9 and tartrate resistant acid phosphatase (TRAcP) localization, the following technique was used. Sections were rehydrated through a series of ethanols and then rinsed in TBST (10 mmol/L Tris at pH 7.4, 150 mmol/L NaCl) with 0.05% Tween 20. For antigen retrieval, slides were immersed in a 20-μg/mL solution of proteinase K (Sigma-Aldrich) according to the manufacturer's instructions for 10 min at room temperature. Following washing in TBS, tissue sections were blocked using standard blocking criteria for 1 h at room temperature. MMP-9 (Oncogene) antibodies at a dilution of 1:100 were added in blocking solution overnight at 4°C. Slides were washed extensively in TBST (TBS with 0.05% Tween-20) before the addition of a species-specific fluorescently labeled secondary antibody (Alexafluor 568 nm, Invitrogen) diluted 1:1,000 in blocking solution for 1 h at room temperature. Slides were washed in TBS and then equilibrated in an acetate buffer as described (14). The ELF97 TRAcP stain (Invitrogen) was diluted 1:1,000 in acetate buffer and slides were incubated for 15 min at room temperature. Following washing, slides were aqueously mounted in media (Biomeda Corp) containing 2 μmol/L 4′,6-diamidino-2-phenylindole for nuclear localization. Angiogenesis in the tumor microenvironment was assessed by CD-31 (BD Pharmingen) immunohistochemistry using a standard immunohistochemistry protocol as previously described (4).
For histomorphometry, at least three nonserial sections from multiple animals in each group were stained with H&E using standard protocols or TRAcP staining as described to assess osteoclast/osteoblast number per ×20 field at the tumor-bone interface using Metamorph. The extent of osteolysis was calculated in multiple sections from each animal using a “bone destruction index,” which refers to the length of osteolysis per length of cranial bone beneath the transplanted tumors, and was determined using Metamorph. The chaotic and woven nature of pathologic bone was easily distinguished from the remaining laminar calvarial bone by H&E. The area of pathologic bone at the tumor-bone interface was calculated using Metamorph.
Osteoclastogenesis and Aortic Ring Assays
CD11b-positive myeloid precursors were isolated from the bone marrow of 6-wk-old immunocompromised mice that were WT or null for MMP-9 using the MACS separation system as per manufacturer's instructions (Miltyni Biotec). After isolation, 1 × 106 myeloid cells/osteoclast precursors were resuspended in 1 mL of αMEM medium containing 10% FCS seeded into each well of a 48-well plate. The following day, WT and MMP-9 null groups were treated with osteoclast differentiation factors, 75 ng/mL recombinant receptor activator of nuclear κB ligand (R&D systems), and 25 ng/mL of macrophage colony–stimulating factor (R&D Systems). A separate group of WT and MMP-9 null cultures (n = 6 per group) were treated with 100 ng/mL recombinant active MMP-9 (Calbiochem). Osteoclast differentiation medium with or without MMP-9 was changed every 48 h. After 10 d of culture, cells were fixed with ice-cold methanol for 5 min, rinsed in 1xPBS, and then subjected to colorimetric staining for the osteoclast marker, TRAcP, as per manufacturer's instructions (Sigma-Aldrich). For the collection of conditioned media, the same procedure was used with the exception that on day 10, groups subjected to media alone or osteoclast differentiation media were carefully rinsed with 1xPBS. Serum-free αMEM (200 μL) was added to each well and the medium was allowed to condition for 24 h. Levels of vascular endothelial growth factor (VEGF)-A were quantitated by ELISA as per manufacturer's instructions (R&D Systems).
Aortic ring assays were done as described (15). After isolation and embedding of WT aortic rings in-type I collagen, the explants were treated with 2.5% mouse serum/αMEM including 20% conditioned media derived from WT or MMP-9 null osteoclast cultures or 10 ng/mL of VEGF-A164 (R&D Systems) as a positive control. The medium was changed daily for 9 d. Photomicrographs were also recorded on a daily basis and the distance of sprout outgrowth was determined by Metamorph.
For in vivo data, statistical analysis was done using ANOVA and Bonferroni multiple comparison tests. A value of P < 0.05 was considered significant. Data are presented as mean ± SD.
MMP-9 Is Primarily Localized to Osteoclasts in the Prostate Tumor–Bone Microenvironment
Previously, we identified that several MMPs including MMP-9 were highly expressed in the prostate tumor–bone microenvironment (4).5 Because changes in MMP expression at the level of gene transcription often are not reflected at the level of the translated protein product, we addressed whether MMP-9 protein was detectable in the prostate tumor–bone microenvironment and what the cellular source of MMP-9 was. Our results indicate that MMP-9 was localized to the stromal compartment of the human (n = 10) and rodent (n = 25) prostate tumor–bone microenvironment, whereas the prostate cancer cells were largely negative (Supplementary Figs. S1 and S2; Fig. 1A-D). Analysis of ×40 photomicrographs (n = 10) of the rodent tumor–bone microenvironment revealed that 7.3% of the cell total (as assessed by counting 4′,6-diamidino-2-phenylindole–stained nuclei) was positive for MMP-9 by immunofluorescence. Using multinuclearity and TRAcP as markers for mature osteoclasts, we determined that within the population of MMP-9–positive cells, 58.7% (4.26% ± 1.99% SD) were osteoclasts. These results show that osteoclasts are a major source of MMP-9 in the tumor-bone microenvironment and are in keeping with our studies examining the localization of MMP-9 in human and murine breast/mammary osteolytic tumor–bone microenvironments (16).
Host MMP-9 Promotes Tumor Growth in the Bone Microenvironment
Because osteoclasts are primarily responsible for bone resorption, we next determined whether the ablation of host MMP-9 would affect prostate tumor progression in a bone microenvironment. To this end, equal volumes of rat prostate adenocarcinoma tissue were transplanted to the calvaria of immunocompromised 6-week-old mice that were either WT (n = 9 per time point) or null for MMP-9 (n = 9 per time point). Tumor volumes were measured on a weekly basis for 3 weeks using calipers. Our results show that the tumor volume in the WT animals was significantly higher compared with the MMP-9 null animals at the week 3 time point [1,746 ± 250.2 mm3 (WT) versus 1,262 ± 207.6 (MMP-9−/−) mm3; P < 0.05; Fig. 2], suggesting that host-derived MMP-9 contributes to prostate tumor growth in the bone microenvironment. This effect was consistently observed in three independently repeated experiments.
Host MMP-9 Does Not Affect Prostate Tumor–Induced Osteolysis
Next, we determined the effect of host MMP-9 on prostate tumor–induced osteolysis because we observed that osteoclasts are a major source of MMP-9 in the prostate tumor–bone microenvironment; the concept of the vicious cycle dictates that osteoclast-mediated bone resorption is critical for tumor growth (2); and MMP-9 null mice have been shown to have a delay in osteoclast recruitment to centers of ossification during bone development (17, 18). Using in vivo microcomputer tomography imaging, prostate tumor–induced osteolysis was imaged over a 3-week time period and segmentation analysis using the Amira software allowed for the quantitation of osteolysis. The results from three independent experiments with small numbers of animals in each group (n = 3) revealed the presence of more bone in the MMP-9 null animals at the week 3 time point, i.e., less osteolysis, but this difference was not statistically significant [5,617 ± 208.2 mm3 (WT) versus 7,057 ± 1,443 mm3 (MMP-9−/−); P > 0.05; Fig. 3A].
Histomorphometry analysis of the bone destruction index at the week 3 time point also revealed no difference in the extent of osteolysis between the WT (n = 9) and MMP-9 null (n = 9) groups [0.4398 ± 0.1331 (WT) versus 0.34833 ± 0.0752 (MMP-9−/−) bone destruction index; P > 0.05; Fig. 3B]. Furthermore, no difference in the number of multinucleated TRAcP-positive osteoclasts was observed between the WT and MMP-9 null groups [5.67 ± 2.236 (WT) versus, 8.02 ± 2.693 osteoclasts per ×20 field (MMP-9−/−); P > 0.05; Fig. 3C]. Taken together, these in vivo and histologic analyses show that host MMP-9 does not contribute to prostate tumor–induced osteolysis.
Host MMP-9 Does Not Affect Prostate Tumor–Induced Osteogenic Response
Our data show that osteoblasts are not a major source of MMP-9 in the in vivo tumor-bone microenvironment (Fig. 1). However, given the interdependency between osteoblasts and osteoclasts with respect to the bone remodeling process, we next determined whether the ablation of host MMP-9 could affect prostate tumor–induced osteogenic changes
Before and at weekly intervals after transplantation of the prostate tumor tissue, WT and MMP-9 null mice were injected with 99mTC-MDP, which selectively concentrates in areas of active bone remodeling (19). Using in vivo micro-SPECT imaging, the osteogenic responses were measured in individual animals in each group over time and whereas lower values were obtained in the MMP-9 null group compared with WT control, these differences did not reach statistical significance [2.9 × 10−5 ± 1.5 × 10−5 (WT) versus 1.9 × 10−5 ± 0.94 × 10−5 (MMP-9−/−) micro-SPECT activity ratio; P > 0.05; Fig. 4A]. Similarly, histomorphometry examining the extent of pathologic bone formation, which was discerned by H&E staining (Fig. 4B) of multiple sections from multiple animals (n = 9 per group), revealed no differences in prostate tumor–induced osteogenesis [5.4 × 104 ± 2.3 × 104 μm2 (WT) versus 5.4 × 104 ± 2.4 × 104 μm2 (MMP-9−/−); P > 0.05; Fig. 4C]. Furthermore, the number of osteoblasts rimming the pathologic bone also showed no difference between the WT and MMP-9 null mice [83.9 ± 28.9 (WT) versus 126.3 ± 21.23 (MMP-9−/−) osteoblasts per ×20 field; P > 0.05; Fig. 4D]. Collectively, these data show that host MMP-9 does not affect prostate tumor–induced osteogenic change.
Host MMP-9 Contributes to Angiogenesis in the Prostate Tumor–Bone Microenvironment
Our studies indicate that host MMP-9 contributed to tumor growth. But it seemed that this observation was independent of prostate tumor–induced osteolytic and osteogenic change because neither were significantly attenuated in the MMP-9 null mice compared with the WT controls. Several studies have shown that osteoclasts are important mediators of angiogenesis (17, 20), whereas MMP-9 has been identified as playing a key role in regulating the bioavailability of the angiogenic factor VEGF-A164 in the developing bone and in various tumor microenvironments (18, 21). Therefore, we next examined whether differences in angiogenesis existed within the WT and MMP-9 null prostate tumor–bone microenvironments.
Immunohistochemical staining for CD-31, a widely used marker for tumor vasculature (Fig. 5A), revealed a significantly lower number of blood vessels in the prostate tumor–bone microenvironment of the MMP-9 null mice (n = 17) compared with the WT (n = 25) controls [4.5 ± 0.91 (WT) versus 2.94 ± 1.14 (MMP-9−/−) CD-31–positive blood vessels per ×20 field; P < 0.05; Fig. 5B]. Furthermore, we observed that the diameter of the blood vessels within each tissue section was also significantly smaller in the MMP-9 null group [124 ± 72.22 μm (WT) versus 89.33 ± 58.33 μm (MMP-9−/−); P < 0.05; Fig. 5C]. Because MMP-9 can potentially be derived from other cellular sources in the prostate tumor–bone microenvironment, we tested whether MMP-9 affected the ability of osteoclasts to directly influence angiogenesis. Initially, we observed that MMP-9 ablation (n = 7) did not affect the ability of osteoclast precursors to undergo osteoclastogenesis compared with WT (n = 6) controls [153.8 ± 19.45 (WT) versus 173.9 ± 24.2 (MMP-9−/−) TRAcP-positive multinucleated osteoclasts per 48-well chamber; P > 0.05, 6A and B]. However, analysis of the conditioned media by ELISA revealed that MMP-9 was critical for mediating VEGF-A164 bioavailability because MMP-9 null osteoclast conditioned media (n = 7) had significantly lower levels of VEGF-A164 compared with WT (n = 6) controls [185 ± 17.5 pg/mL VEGF-A164 (WT OCL) versus 123.5 ± 17.56 pg/mL VEGF-A164 (MMP-9−/− OCL); P < 0.05; Fig. 6C]. Furthermore, the addition of recombinant MMP-9 to the MMP-9 null osteoclast cultures significantly enhanced the amount of bioavailable VEGF-A164 [123.5 ± 17.6 pg/mL VEGF-A164 (MMP-9−/− OCL) versus 219 ± 15.6 pg/mL VEGF-A164 (MMP-9−/− OCL plus rMMP-9); P < 0.05, Fig. 6C]. Reverse transcription-PCR analysis of VEGF-A164 isoform expression revealed no significant difference between WT and MMP-9 osteoclasts (data not shown), thus indicating that osteoclast-derived MMP-9 can regulate the bioavailability of VEGF-A164. This conclusion was further supported by the observation that conditioned media derived from the MMP-9 null osteoclast cultures was not as efficient as conditioned media derived from WT osteoclasts in promoting angiogenic sprouting using an aortic ring assay [1,654 ± 81.75 μm (WT) versus 1,087 ± 175.9 μm (MMP-9−/−) distance of sprout invasion at day 7; P < 0.05; Fig. 6D]. These observations agree with published roles for MMP-9 controlling angiogenesis and we suggest that a defect in osteoclast-mediated vascularization of the prostate tumor–bone microenvironment is the mechanism underlying the observed decrease in tumor volume in the MMP-9 null animals at week 3.
In human and rodent samples of the prostate tumor–bone microenvironment, we identified that osteoclasts are a rich source of the proteinase, MMP-9. Given that MMPs are involved in matrix remodeling and skeletal development, the observation that MMPs are expressed by osteoclasts in the prostate tumor–bone microenvironment is perhaps not surprising. However, the major novelty and conclusions of the current study are coherent with a theme that has been emerging from the MMP field of research over the past decade, i.e., that MMPs can have a profound effect on multiple aspects of tumor-host communication by regulating the bioactivity and bioavailability of growth factors and cytokines, in this case, VEGF-A164 and angiogenesis.
Our data show that MMP-9 is primarily localized to osteoclasts in a model of prostate tumor–induced osteolytic and osteogenic response (Fig. 1). We observed that the ablation of host MMP-9 significantly reduced tumor growth. Bone is a rich source of growth factors such as transforming growth factor-β and insulin-like growth factors, and the concept of the vicious cycle of tumor-bone interaction dictates that the resorption of the mineralized bone matrix by osteoclasts is critical for the release of these growth factors, thereby resulting in the stimulation of tumor growth (2). However, analysis of the amount of bone destruction between the WT and MMP-9 null groups revealed no difference, nor in the number of osteoclasts present at the tumor-bone interface. These results are consistent with previous reports from our group and others showing that host/osteoclast-derived MMP-9 does not seem to affect the extent of tumor-induced osteolysis (16, 22). Importantly, these studies do not rule out roles for MMP-9 in other types of cancer-induced bone disease. For example, in a recent study examining multiple myeloma progression, we observed that host MMP-9 significantly contributed to myeloma-induced bone destruction (23). These findings underscore the rationale for examining the contribution of individual MMPs in specific disease contexts, a conclusion that is consistent with reports examining the roles for MMPs in other diseases. For example, epithelial expression of MMP-3 is capable of initiating mammary gland tumorigenesis but in the context of skin cancer progression, leukocyte-derived MMP-3 has a protective effect (24, 25).
Osteoblasts are essential mediators of the vicious cycle and human to prostate to bone metastases are hallmarked by extensive areas of osteogenesis. Despite osteoblasts expressing a wide variety of proteinases including MMP-2 and MMP-14 that are important for normal osteoblast function (26, 27), it seemed that in our model system and in human samples of prostate to bone metastasis, that osteoblasts do not express MMP-9 in vivo. However, given the close relationship between osteoblasts and osteoclasts, we assessed if osteoclast-derived MMP-9 could affect the prostate tumor–induced osteogenic response. This is the first report to test the contribution of a host-derived MMP to tumor-induced osteogenesis. Although micro-SPECT and histomorphometry did not reveal a significant difference in prostate tumor–induced osteogenesis, our study shows the feasibility of examining the role of tumor or host-derived MMPs in osteogenic bone remodeling using micro-SPECT.
We acknowledge that the model used in the current study has limitations in that it is not reflective of the process of metastasis, true intraosseous growth, or an anatomic site that metastatic human prostate cancer cells typically metastasize to. However, the model does have several advantages including (a) the generation of a mixed osteogenic/osteolytic lesion that is more reflective of human prostate to bone metastases compared with the solely lytic type lesions generated by human prostate cancer cell lines such as PC-3; (b) its use of a straight forward surgical technique, and (c) it allows for the rapid interrogation of hypotheses in vivo (∼3 weeks). Furthermore, our studies with this model are complemented by Cher and colleagues (22) examining the contribution of host MMP-9 to prostate tumor–induced tibial osteolysis with the human PC-3 prostate cancer cell line. Given the caveats of the model used in the current study, it is possible that MMP-9 can contribute to other steps of metastasis that are not taken into account in the current study such as extravasation and survival/establishment, the latter of which is an important role for host-derived MMP-9 in early lung metastasis (12). Therefore, roles for host MMP-9 in other aspects of the metastatic cascade cannot be ruled out.
Although MMP-9 did not affect prostate tumor–induced osteolysis or osteogenesis, we did observe that osteoclast-derived MMP-9 significantly contributed to prostate tumor growth in the bone microenvironment. Although studies in the breast tumor–bone microenvironment have shown that MMP-9 may facilitate this process by transforming growth factor-β (28), our data identify that osteoclast-derived MMP-9 can affect angiogenesis in the prostate tumor–bone microenvironment. Recently, reports have indicated proangiogenic roles for osteoclasts based on findings that (a) osteoclasts express several proangiogenic factors such as VEGF-A (29); (b) the close proximity between osteoclasts and endothelial cells in areas of bone remodeling (30); (c) evidence that osteoclasts can directly stimulate angiogenesis in vitro (31); and (d) clinical data showing that bisphosphonates, potent antiosteoclast therapies, also significantly halt angiogenesis in pathologic bone diseases such as bone metastasis (32). MMP-9 has also been shown to be important in mediating the “angiogenic switch” by controlling the bioavailability of VEGF-A164 (21). Analysis of the developing bone in MMP-9 null mice showed a delay in osteoclast invasion and angiogenesis in primary ossification centers (18). Our data show that MMP-9 null osteoclast cultures generate significantly less bioavailable VEGF-A164 compared with WT controls. Analysis of VEGF-A164 expression at a mRNA level revealed that MMP-9 null osteoclast cultures expressed similar levels of VEGF-A164 (data not shown). Therefore, it seems that MMP-9 is important in regulating the bioavailability of VEGF-A164 that is generated by the osteoclast cultures and we suggest that the decreased angiogenesis observed in the MMP-9 null tumor–bone microenvironment is, in part, due to an inefficiency of MMP-9 null osteoclasts to generate bioavailable VEGF-A164.
Although osteoclasts are a major source of MMP-9 in the prostate tumor–bone microenvironment as observed by immunofluorescent localization, other cells throughout the stromal compartment also stained positively for MMP-9. Although we did not investigate the identity of the cellular sources, we posit based on the literature that these cell types are most likely composed of macrophages and neutrophils that are potent mediators of the angiogenic process (12, 33-36). Therefore, although we suspect that osteoclast-derived MMP-9 is important in mediating angiogenesis, roles for other MMP-9–positive host cells cannot be excluded. Surprisingly, it appeared that in our model and in human samples of prostate to bone metastasis (n = 10) that tumor cells were largely negative for MMP-9, but given the small sample size, roles for tumor-derived MMP-9 cannot be excluded at this juncture. Collectively however, our data point toward MMP-9 as being a clinically relevant target for the inhibition of angiogenesis in the tumor-bone microenvironment.
The rationale for the use of selective MMP inhibitors for the treatment of lytic breast and prostate to bone metastases is supported by several independently performed preclinical studies obtained with animal models of the disease. The treatment of mice bearing lytic breast or prostate bone metastases with the broad spectrum MMP inhibitors BB-94 or GM6001 could prevent tumor growth and tumor-induced osteolysis (37-39). Thus far, no studies have examined the affect of MMP inhibitors on osteogenic bone metastases. To translate MMP inhibitors for human clinical use, selective inhibition of metalloproteinases is a prerequisite to avoid the described drawbacks of the original MMP inhibitors (40). To this end, SB-3CT, an MMP inhibitor with heightened selectivity for MMP-2 and MMP-9 is effective in preventing PC-3 prostate tumor progression with decreased tumor growth, vascularization, and osteolysis being reported (41). Based on our findings, these data suggest that SB-3CT–mediated MMP-9 inhibition prevents tumor and host-mediated angiogenesis and points toward a potentially important role for MMP-2 in controlling osteolysis, which has thus far not been investigated. Therefore, fine-tuning the specificity of MMP inhibitors may be a relevant approach for the development of therapies for the treatment of bone metastases.
In conclusion, we have identified that osteoclasts are a major source of MMP-9 in the human and rodent prostate tumor–bone microenvironment. However, we have shown that host MMP-9 does not affect tumor-induced osteolytic or osteogenic responses in our animal model but does contribute to tumor growth in the prostate tumor–bone microenvironment by mediating angiogenesis through regulating the bioavailability of matrix-sequestered VEGF-A164. These observations are in keeping with the growing studies that show differential roles for host MMPs in mediating tumor-host interaction.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Views and opinions of, and endorsements by the author(s), do not reflect those of the U.S. Army or the Department of Defense. We thank Kathy J. Carter for assistance with the aortic ring assay.
Grant Support: Department of Defense under award number W81XWH-07-1-0208 (C.C. Lynch). Partial support was also provided by the NIH under grants R25-CA136440 (L.C. Johnson), S10-RR23784, P30-CA068485, and U24-CA126588 (T.E. Peterson). Human samples of prostate to bone metastases were collected under award number PO1-CA85859 and P50-CA097186 (R.L. Vessella).
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.
- ©2010 American Association for Cancer Research.