Breast cancer commonly causes osteolytic metastases in bone, a process that is dependent on tumor-stromal interaction. Proteases play an important role in modulating tumor-stromal interactions in a manner that favors tumor establishment and progression. Whereas several studies have examined the role of proteases in modulating the bone microenvironment, little is currently known about their role in tumor-bone interaction during osteolytic metastasis. In cancer-induced osteolytic lesions, cleavage of receptor activator of nuclear factor-κB ligand (RANKL) to a soluble version (sRANKL) is critical for widespread osteoclast activation. Using a mouse model that mimics osteolytic changes associated with breast cancer–induced bone metastases, we identified cathepsin G, cathepsin K, matrix metalloproteinase (MMP)-9, and MMP13 to be proteases that are up-regulated at the tumor-bone interface using comparative cDNA microarray analysis and quantitative reverse transcription-PCR. Moreover, we showed that cathepsin G is capable of shedding the extracellular domain of RANKL, generating active sRANKL that is capable of inducing differentiation and activation of osteoclast precursors. The major source of cathepsin G at the tumor-bone interface seems to be osteoclasts that up-regulate production of cathepsin G via interaction with tumor cells. Furthermore, we showed that in vitro osteoclastogenesis is reduced by inhibition of cathepsin G in a coculture model and that in vivo inhibition of cathepsin G reduces mammary tumor–induced osteolysis. Together, our data indicate that cathepsin G activity at the tumor-bone interface plays an important role in mammary tumor–induced osteolysis and suggest that cathepsin G is a potentially novel therapeutic target in the treatment of breast cancer bone metastasis. [Cancer Res 2008;68(14):5803–11]
As the most common type of cancer in women in the United States and the second leading cause of cancer-related death, breast cancer represents a serious health problem (1). Metastatic disease is responsible for most of breast cancer–related mortality, accounting for the sharp decline in 5-year survival observed as breast cancer progresses from regional to distant metastasis (1). In patients that ultimately succumb to breast cancer, nearly all are found to have bone metastases, showing the tropism of breast cancer cells to bone. These metastatic bone lesions, predominantly osteolytic in nature, carry severe consequences including hypercalcemia, pathologic fracture, and leukoerythroblastic anemia (2). As a result, bone metastasis dramatically increases the risk of mortality as well as significantly reduces the quality of life.
The mechanisms involved in breast cancer–induced osteolytic bone metastases have been extensively studied but remain unclear. As breast cancer cells establish secondary colonies in the bone microenvironment, a vicious cycle ensues (2). Tumor cells secrete soluble factors such as interleukin (IL)-1, IL-8, and parathyroid hormone–related peptide that activate osteoblasts and induce expression of receptor activator of nuclear factor-κB ligand (RANKL). RANKL expressed by osteoblasts can then interact with its receptor (RANK) on the cell surface of osteoclast precursors, leading to osteoclast differentiation and activation. Activated osteoclasts then mediate bone degradation through the release of proteases including matrix metalloproteinases (MMP) and cathepsins. As bone is degraded, sequestered growth factors are released that serve as growth and survival factors as well as chemoattractants for tumor cells, promoting the establishment of bone metastases (2). Thus, tumor-stromal interactions play a significant role in the establishment and regulation of osteolytic bone metastases.
Proteases play a major role in tumor-stromal interactions by modifying the bidirectional communication to favor tumor establishment and growth at both the primary tumor site and metastatic site (3–5). In a prostate cancer model, we have previously shown that MMP7 is capable of cleaving membrane-bound RANKL into a soluble version (sRANKL), which relieves the contact-dependent nature of osteoblast-osteoclast interaction, bypasses the rate-limiting step of the vicious cycle, and enhances osteoclast activation and subsequent osteolysis (6). To date, however, no such mechanism has been described in breast cancer.
In the present study, we examined the patterns of expression of proteases at the tumor-bone interface of mammary tumor–induced osteolytic lesions and investigated their functional significance in modulating RANKL-RANK signaling in osteoclast activation and osteolysis. We identified cathepsin G, cathepsin K, MMP9, and MMP13 to be proteases that are up-regulated at the tumor-bone interface. Moreover, we showed that cathepsin G is capable of shedding the extracellular domain of RANKL, generating active sRANKL that is capable of inducing differentiation and activation of osteoclast precursors. Furthermore, we showed that in vitro osteoclastogenesis is reduced by inhibition of cathepsin G in a coculture model and subsequently showed that in vivo inhibition of cathepsin G reduces mammary tumor–induced osteolysis. Thus, we have shown for the first time a role for cathepsin G at the tumor-bone interface of mammary tumor–induced osteolytic lesions and have uncovered a potential therapeutic target.
Materials and Methods
Animal model and tissue preparation. 4T1, Cl66, and Cl66M2 murine breast adenocarcinoma cell lines were used in this study. Tumor cells (1 × 105) mixed with growth factor–reduced Matrigel were implanted on the dorsal skin flap over the calvaria of female BALB/c mice. Tumor growth was monitored twice a week. Mice were sacrificed and necropsied for examination of osteolytic lesions at 4 wk postimplantation. At that time, the tumor and the underlying bone were divided into two pieces. One piece was used for separation of the tumor-bone interface from the tumor alone area for further analysis and the other piece was used for histology sections. All studies were done in accordance with the Institutional Animal Use and Care Committee of the University of Nebraska Medical Center. For histologic examination, tissues were fixed with periodate-lysine-paraformaldehyde at 4°C for 48 h. The tissues were then transferred into a decalcification solution (15% EDTA with glycerol, pH 7.4–7.5) for 4 wk. The tissue was then paraffin embedded and processed for further analysis.
Microarray analysis and real-time PCR. Calcified frozen sections were serially sectioned into 10-μm slices. Sections were then microdissected to separate the tumor-bone interface from the tumor alone area (Fig. 1A). Total RNA was extracted from both the tumor-bone interface and the tumor alone area and then amplified by using a probe amplification kit (Affymetrix). Genetic expression at the tumor-bone interface was compared with the tumor alone area using an Affymetrix Mouse Expression Array (430A). Analysis was done using Affymetrix Gene Chip Operating Software to generate raw expression data. A signal log algorithm was used to compare the quantity of a given transcript in two arrays: a baseline array, here the tumor alone area, and an experimental array, here the tumor-bone interface. The signal log ratio is calculated by comparing each probe pair on the experimental array to the corresponding probe pair on the baseline array and then considering the mean of the log ratios of probe pair intensities across the two arrays. The change in transcript level is then expressed as a log 2 ratio. Using this system, a signal log ratio of 1.0 indicates a 2-fold increase at the tumor-bone interface whereas a ratio of −1.0 indicates a 2-fold decrease. For each set of tissue from 4T1, Cl66, and Cl66M2, the fold change at the tumor-bone interface with respect to the tumor alone area was calculated and the genes were ranked from highest to lowest expression.
For real-time quantitative reverse transcription-PCR (RT-PCR) analysis, total RNA was isolated from tissue at the tumor-bone interface and tumor alone area using Trizol reagent (Invitrogen). Five micrograms of total RNA were used for reverse transcription. First-strand cDNA was generated using oligo(dT)18 (Fermentas) and Superscript II RT (Invitrogen). Two microliters of the resulting cDNA (1:10 dilution) were used in the real-time reactions with gene-specific primers. The following gene-specific primers were used: cathepsin G, 5′-GAGTCCAGAAGGGCTGAGTG-3′ and 5′-CCTTTCTCGCATTTGGATGT-3′; cathepsin K, 5′-CCAGTGGGAGCTATGGAAGA-3′ and 5′-CTCCAGGTTATGGGCAGAGA-3′; MMP9, 5′-CATTCGCGTGGATAAGGAGT-3′ and 5′-TCACACGCCAGAAGAATTTG-3′; MMP12, 5′-AATGGGCAACTGGACAACTC-3′ and 5′-ACCGCTTCATCCATCTTGAC-3′; MMP13, 5′-TCCCTGCCCCTTCCCTATGG-3′ and 5′-CTCGGAGCCTGTCAACTGTGG-3′; and glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5′-AGCCTCGTCCCGTAGACAAAA-3′ and 5′-GATGACAAGCTTCCCATCTCG-3′. Quantitative RT-PCR reactions were carried out with FastStart SYBR Green Master mix (Roche) and MyIQ iCycler (Bio-Rad). Fluorescence intensity was measured at the end of each elongation step as a means to evaluate the amount of formed PCR product. GAPDH was used as a reference to normalize the samples.
Generation of RANKL protein. Full-length RANKL cDNA (NM_003701) with an NH2-terminal His-tag sequence cloned in pReceiver-B01 (GeneCopoeia) plasmid was used to generate RANKL protein using a TnT Quick Coupled Transcription/Translation system (Promega) per manufacturer's protocol using translation grade [35S]methionine (1,175 Ci/mmol at 10 mCi/mL). Two microliters of the reaction mixture were separated on a 12% SDS-polyacrylamide gel and subsequently visualized and photographed with a Typhoon 9410 Variable Mode Imager (GE Healthcare). The resulting translation product was purified with a His-Trap Crude 1-mL column (GE Healthcare) per manufacturer's protocol. The purified protein was then concentrated by using a Centriprep YM-10 spin column (Omicon) and quantified with the BCA Protein Assay kit per manufacturer's protocol (Pierce).
Protease function assays. To test the activity of each of the proteases, colorimetric substrates were used. One hundred nanograms of cathepsin G or cathepsin K were incubated with 500 μmol/L Suc-AAPF-pNA substrate (Biomol International) in enough buffer to make a 50-μL reaction. The 10× buffer contained 100 mmol/L CaCl2, 500 mmol/L HEPES, 10 mmol/L DTNB, and 0.5% Brij-35. Similar reactions were done for MMP9 and MMP13 with chromogenic MMP substrate (Biomol International). Reaction mixtures were incubated at 37°C for 60 min. Readings were taken on a BioTek ELx800 microplate reader every 15 min at 405 nm.
Protease-mediated cleavage of RANKL. Three micrograms of full-length RANKL were incubated overnight at 37°C alone or with 100 ng of cathepsin G, cathepsin K, MMP9, or MMP13 in enough buffer to make a 25-μL reaction. Reactions were terminated by adding EDTA to a final concentration of 10 mmol/L. The reaction mixture was then separated on a 12% SDS-polyacrylamide gel. The gel was then stained with 0.025% Coomassie brilliant blue (Bio-Rad) and photographed by using a MultiImage Light Cabinet (Alpha Innotech Corporation). Cathepsin G inhibition was tested by adding Na-Tosyl-Phe-chloromethylketone (TPCK; Sigma-Aldrich) at a final concentration of 1 mmol/L. TPCK was incubated with enzyme and buffer for 30 min at 37°C before adding RANKL. The reaction mixture was then incubated, separated, and analyzed.
Similar reactions were carried out and subsequently used for Western blot analysis. Reaction mixtures were separated on a 12% SDS-polyacrylamide gel and then transferred onto a polyvinylidene difluoride (PVDF) membrane (GE Healthcare) at 100 V for 1 h. The membrane was then washed with 0.1% Tween-TBS. The membrane was then blocked overnight in 5% bovine serum albumin in PBS. The next day, the membrane was incubated for 90 min with 1:500 anti-RANKL antibody that recognizes an epitope near the COOH terminus of RANKL (Santa Cruz Biotechnology) diluted in 0.1% Tween-TBS. After washing in 0.1% Tween-TBS, the membrane was incubated for 60 min with 1:1,500 antigoat-horseradish peroxidase diluted in 0.1% Tween-TBS. Finally, after washing with 0.1% Tween-TBS, the membrane was developed with an enhanced chemiluminescence (ECL) Plus Western Blotting Detection System (GE Healthcare) per manufacturer's protocol and imaged with a Typhoon 9410 Variable Mode Imager (GE Healthcare).
Osteoclast differentiation and activation assay. RAW 264.7 cells were seeded onto eight-well chambered slides at a density of 3,000 per chamber. Cells were grown in DMEM plus 10% fetal bovine serum (FBS), 1% vitamins, 1% l-glutamine, and 0.08% gentamicin (DMEM complete medium). The cells were incubated overnight at 37°C. The following day, the medium was changed and the cells were treated with medium alone, 50 ng/mL sRANKL, or 50 ng/mL cathepsin G–generated sRANKL. The cells were then allowed to incubate for 7 d with medium changes every other day. On the 7th day, cells were stained for tartrate-resistant acid phosphatase (TRAP) per manufacturer's protocol (Sigma-Aldrich). The number of osteoclasts per 250-μm field was counted for five fields per treatment per replicate. Fully mature, TRAP-positive multinucleated cells were counted as osteoclasts.
BD BioCoat Osteologic Coverslips (BD Biosciences) were placed in 24-well plates. RAW 264.7 cells were then seeded onto the plates at a density of 5,000/cm2. Cells were grown in DMEM complete medium. The cells were incubated overnight at 37°C. The following day, the medium was changed and the cells were treated with medium alone, 50 ng/mL sRANKL, or 50 ng/mL cathepsin G–generated sRANKL. The cells were then allowed to incubate for 7 d with medium changes every other day. On the 7th day, the medium was aspirated and 1 mL of bleach was added to each well. After 5 min, the bleach was aspirated and the coverslips were washed thrice with distilled water. The coverslips were then stained with hematoxylin for 2 min and then rinsed with running tap water for 10 min. Finally, the number and area of calcium phosphate matrix resorption pits (clear areas) on each 12-mm coverslip were measured.
Immunoblotting for cathepsin G. Fifty micrograms of protein from the tumor-bone interface and tumor alone area for each of the animals injected with Cl66 cells were separated on a 12% SDS-polyacrylamide gel and then was transferred onto a PVDF membrane (GE Healthcare). The membranes were immunoblotted with anti–cathepsin G antibody that recognizes an epitope near the NH2 terminus of cathepsin G (1:200; Santa Cruz Biotechnology) and anti–β-actin antibody (1:2,000; Santa Cruz Biotechnology) and developed with an ECL Plus Western Blotting Detection System (GE Healthcare) per manufacturer's protocol and imaged with a Typhoon 9410 Variable Mode Imager (GE Healthcare). The bands for both cathepsin G and β-actin were then quantified and compared using ImageQuant 5.1 (Molecular Dynamics).
Immunohistochemistry for cathepsin G. For the in vivo detection of cathepsin G, sections were deparaffinized and processed for antigen retrieval. Following nonspecific blocking, sections were then incubated with anti–cathepsin G antibody (Santa Cruz Biotechnology; 1:50 dilution) overnight at 4°C. The sections were then washed and subsequently incubated with biotinylated antigoat IgG antibody diluted to 1:500 for 1 h. Sections were then washed and incubated with avidin-biotin complex (Vector Laboratories) for 45 min. Next, the slides were developed with diaminobenzidine tetrahydrochloride substrate (Vector Laboratories) and counterstained with hematoxylin for 30 s. Antigoat IgG was added in lieu of primary antibody as control and no detectable staining was observed in these sections.
Cathepsin G inhibition coculture model. RAW 264.7 cells were seeded at a density of 3,000/cm2 on gelatin-coated six-well plates in DMEM plus 10% FBS, 1% vitamins, 1% l-glutamine, and 0.08% gentamicin (DMEM complete medium). Following overnight incubation, cells were incubated with DMEM complete medium alone, DMEM complete medium plus 25% (v/v) Cl66-conditioned medium with or without TPCK at a final concentration of 1 mmol/L, TPCK at a final concentration of 1 mmol/L plus commercially available sRANKL at 50 ng/mL, or DMEM complete medium containing Cl66 cells at a density of 3000/cm2 with or without TPCK. In the TPCK-treated groups, TPCK was added to a final concentration of 1 mmol/L every 24 h. After 4 d, the coverslips were stained for TRAP per manufacturer's instructions (Sigma-Aldrich) and multinucleated, TRAP-positive osteoclasts were counted.
In vivo inhibition of cathepsin G. Cl66 tumor cells (1 × 105) mixed with growth factor–reduced Matrigel were implanted on the dorsal skin flap over the calvaria of female BALB/c mice. Tumor growth was monitored twice a week. Beginning 7 d after tumor implantation, mice were injected s.c. with TPCK (Sigma-Aldrich) at 50 mg/kg/d (n = 6) or 50-μL DMSO as control (n = 6) for 21 d. Mice were sacrificed at day 31 postimplantation and necropsied for examination of osteolytic lesions. To calculate the bone destruction index, sections were stained with H&E. The bone destruction index was calculated by dividing the length of bone destruction by the length of the tumor-bone interface and multiplying by 100. All studies were done in accordance with the Institutional Animal Use and Care Committee of the University of Nebraska Medical Center.
Statistical analysis. For in vivo studies, the Wilcoxon signed rank test was used to analyze data. For in vitro studies, the Student t test was used. P < 0.05 was considered significant.
Protease gene expression at the tumor-bone interface. Mammary tumor cells with different metastatic potentials, 4T1 (high), Cl66 (moderate), and Cl66M2 (low), were transplanted into the calvaria of BALB/c mice. Histochemical analysis showed that all tumors exhibited tumor-induced osteolysis and osteoclast activation similar to that observed in breast cancer bone metastasis. Using cDNA microarray analysis, we examined the gene expression patterns at the tumor-bone interface compared with the tumor alone area (Fig. 1A). A whole-genome microarray was used and subsequently analyzed by sorting out proteases and protease inhibitors. The five most up-regulated protease genes were cathepsin G, cathepsin K, MMP9, MMP12, and MMP13 (Fig. 1B). The common up-regulation of these five genes in all three cell lines further suggested that they may play an important role at the tumor-bone interface.
We then used quantitative RT-PCR to confirm the up-regulation of these five genes at the tumor-bone interface. We confirmed the up-regulation of cathepsin G (Fig. 1C,-i), cathepsin K (Fig. 1C,-ii), MMP9 (Fig. 1C,-iii), and MMP13 (Fig. 1C,-v), but not MMP12 (Fig. 1C -iv), at the tumor-bone interface.
Cathepsin G cleaves RANKL to generate sRANKL. We previously reported that MMP7 plays a role in prostate tumor–induced osteolysis through the cleavage of cell-surface RANKL into a soluble form of RANKL that is capable of enhancing osteoclast differentiation and subsequent osteolysis (6). To determine if any of the proteases that we observed to be up-regulated at the tumor-bone interface play a similar role in mammary tumor–induced osteolysis, the activity of each protease against full-length RANKL was assessed.
Full-length RANKL protein was generated by using an in vitro transcription/translation kit. The product was visualized and the expected ∼34-kDa protein was seen (Fig. 2B). Although activity against the native substrate of each protease was observed (Fig. 2A,-i), only cathepsin G generated a soluble version of RANKL (∼24 kDa; Fig. 2C). To confirm cathepsin G–mediated cleavage of RANKL, we used TPCK, a cathepsin G inhibitor, before incubation with full-length RANKL protein. TPCK is an inhibitor of the chymotrypsin-like group of proteases and is a potent inhibitor of cathepsin G (7, 8). We first confirmed the ability of TPCK to inhibit cathepsin G activity against its native substrate, and TPCK effectively inhibited cathepsin G activity (Fig. 2A,-ii). Similarly, cathepsin G preincubated with TPCK did not cleave RANKL (Fig. 2D,-ii). Finally, Western blot analysis with an anti-RANKL antibody that recognizes an epitope near the COOH terminus confirmed that the cleaved protein was RANKL and that the soluble product generated was derived from full-length RANKL cleaved near the NH2 terminus within the extracellular domain (Fig. 2D -i).
Cathepsin G–generated sRANKL induces osteoclast differentiation and activation. The ability of cathepsin G–generated sRANKL to induce osteoclast differentiation and activation was assessed using the murine monocyte/macrophage cell line RAW 264.7. Fully mature, TRAP-positive multinucleated cells were counted as osteoclasts (Fig. 3B,-ii). Commercially available sRANKL (positive control) induced significantly higher osteoclast differentiation than medium alone (negative control; Fig. 3A). Numerous TRAP-positive cells were seen in the control sample but most were mononuclear (Fig. 3B,-i). Similar to the commercial sRANKL, cathepsin G–generated sRANKL induced significantly higher osteoclast formation than control (Fig. 3A). Although osteoclast differentiation was higher in the sRANKL-treated group, it did not differ statistically from the cathepsin G–generated sRANKL–treated group.
We then examined whether or not the osteoclasts produced by treatment with cathepsin G–generated sRANKL are active. Osteoclast activation was assessed by counting the number of calcium phosphate matrix resorption pits formed per BD BioCoat Osteologic Coverslip. Commercially available sRANKL (positive control) induced activation of osteoclasts with a significantly higher number of resorptive pits compared with medium alone (negative control; Fig. 3C and D). Similarly, cathepsin G–generated sRANKL yielded a significantly higher number of pits compared with control and was statistically indistinguishable from commercially available sRANKL (Fig. 3C and D). Together, these results indicate that cathepsin G–generated sRANKL is capable of inducing osteoclast differentiation and activation and suggest that increased cathepsin G activity at the tumor-bone interface can lead to increased osteoclast differentiation and subsequent osteolysis.
Up-regulation of cathepsin G at the tumor-bone interface. Recognizing that cathepsin G may make a significant contribution to the vicious cycle, we reexamined cathepsin G expression at the tumor-bone interface and further confirmed its up-regulation at the protein level by Western blot analysis comparing the tumor-bone interface to the tumor alone area. Cl66 tumors showed significant up-regulation of cathepsin G at the tumor-bone interface as compared with the tumor alone area (Fig. 4A and B). This protein level analysis confirmed the earlier cDNA microarray and quantitative RT-PCR analyses.
Osteoclasts/osteoclast precursors are the major source of cathepsin G. To determine the source of cathepsin G at the tumor-bone interface, immunohistochemistry was used to assess the source of cathepsin G in vivo. Osteoclasts and osteoclast precursors stained strongly positive for cathepsin G (Fig. 4C,-i), whereas tumor cells at the tumor-bone interface stained moderately positive (Fig. 4C,-i). Tumor cells away from the tumor-bone interface in the tumor alone area stained weakly positive (Fig. 4C,-ii). Normal bone did not show any immunoreactivity to cathepsin G (Fig. 4C -iii). These data agree with our in vitro data (Supplementary data) that tumor cells have minimal baseline expression of cathepsin G that is up-regulated as they interact with osteoclast precursors, but osteoclasts are the major source of cathepsin G at the tumor-bone interface.
Inhibition of cathepsin G blocks osteoclastogenesis in vitro. RAW 264.7 cells treated with 25% Cl66-conditioned medium showed significantly higher osteoclast differentiation compared with control (Fig. 5A), suggesting that a soluble factor produced by Cl66 cells is capable of inducing osteoclast differentiation. To determine whether cathepsin G is responsible for the generation of this soluble factor, Cl66 cells were treated with TPCK while collecting conditioned medium. Cl66-conditioned medium from cells treated with TPCK induced significantly lower osteoclast differentiation compared with Cl66-conditioned medium from untreated cells (Fig. 5A). Importantly, RAW 264.7 cells treated with both TPCK and sRANKL showed significant osteoclast formation, showing that reduction in osteoclast formation in the TPCK-treated samples is not merely due to death of osteoclasts secondary to TPCK toxicity (Fig. 5A). Similarly, RAW 264.7 cells cocultured with Cl66 cells showed significantly higher osteoclast differentiation than RAW 264.7 cells alone (Fig. 5B). TPCK also reduced osteoclastogenesis in the coculture but did not eliminate it (Fig. 5B). Taken together, these results indicate that cathepsin G is responsible for the generation of a soluble factor capable of inducing osteoclast differentiation and that inhibition of cathepsin G reduces osteoclast differentiation.
Inhibition of cathepsin G reduces osteolysis in vivo. In vitro inhibition of cathepsin G reduced osteoclastogenesis, and thus we sought to determine whether inhibition of cathepsin G in vivo would reduce mammary tumor–induced osteolysis. No difference was observed in tumor growth or growth kinetics between TPCK-treated mice and negative control mice (data not shown). However, TPCK treatment significantly reduced tumor-induced osteolysis from an average bone destruction index of 35.3% in the DMSO-treated (negative control) mice to 13.2% in the TPCK-treated mice (Fig. 6A). Figure 6B shows the marked reduction in osteolysis at the tumor-bone interface of TPCK-treated mice compared with the control mice.
Proteases play an important role in modulating tumor-stromal interactions in a manner that favors tumor establishment and progression. Although several studies have examined the role of proteases in modulating the bone microenvironment (9–12), little is currently known about their role in tumor-bone interaction during osteolytic metastasis. In this study, we have identified potentially important proteases in mammary tumor–induced osteolytic lesions, including cathepsin G, cathepsin K, MMP9, and MMP13.
Because RANKL is one of the key molecular players in a variety of osteolytic lesions that has previously been shown to be modified by proteases to favor tumor progression (6, 13, 14), we sought to determine if any of the up-regulated proteases were capable of modulating RANKL signaling. The importance of RANKL is in its ability to signal through RANK on preosteoclasts to induce differentiation and activation leading to bone resorption. The rate-limiting step to the vicious cycle is the RANKL-RANK signaling pathway. Because both RANKL and RANK are membrane-bound proteins, this interaction requires physical cell-to-cell contact between osteoblasts and osteoclast precursors in order to occur. Because RANKL, a member of the tumor necrosis factor family of cytokines, stimulates RANK via cell-to-cell contact between osteoblasts and osteoclast precursors, osteoclast differentiation and activation occurs with induction of TRAP activity, calcitonin binding, and actin-ring formation characteristic of osteoclasts (15, 16). The centrality of RANKL to the vicious cycle makes it a molecule that could potentially be up-regulated to establish osteolytic lesions, and in fact, we have found that RANKL is up-regulated at the tumor-bone interface of mammary tumor–induced osteolytic lesions.3
K.C. Nannuru, M. Futakuchi, A. Sadanandam, T.J. Wilson, M.L. Varney, L.J. Myers, X. Li, E.G. Marcusson, and R.K. Singh. Unpublished data.
By identifying proteases differentially expressed as a result of tumor-bone interactions, we generated a list of proteases potentially involved in the modulation of RANKL-RANK signaling through the generation of sRANKL. We selected the top five proteases for further evaluation, which included cathepsin G, cathepsin K, MMP9, MMP12, and MMP13. Using quantitative RT-PCR to confirm up-regulation at the tumor-bone interface, we were able to eliminate MMP12 from the list, leaving four up-regulated proteases for further evaluation. Interestingly, RANKL has been shown to induce the expression of both cathepsin K and MMP9 in osteoclast precursor cells (17), and our data agree with this report (data not shown), suggesting that up-regulation of these two proteases may be an effect of enhanced RANKL signaling rather than a cause. Nonetheless, all four proteases were tested for their ability to cleave RANKL. Of these, only cathepsin G was capable of generating sRANKL. Western blot analysis suggests that the cleavage site is near the NH2 terminus of the RANKL protein because the longer ∼24-kDa fragment was detected with an anti-RANKL antibody that recognizes an epitope near the COOH terminus. The cleavage site and molecular weight of this larger fragment are similar to those observed in MMP7 cleavage of RANKL (6). Furthermore, the sRANKL generated by cathepsin G is an active product capable of inducing osteoclastogenesis. Osteoclasts are derived from the hematopoietic monocyte/macrophage lineage of cells. For this reason, we used the murine monocyte/macrophage cell line RAW 264.7 to test the activity of cathepsin G–generated sRANKL. Treatment of these cells with cathepsin G–generated sRANKL increased the number of differentiated and activated osteoclasts as shown by an increased number of multinucleated, TRAP-positive cells and the production of resorptive pits when cultured on artificial bone matrix.
The production of sRANKL by cathepsin G circumvents cell-cell contact–dependent signaling between membrane-bound RANKL and RANK, which allows increased osteoclastogenesis and subsequent bone resorption. Thus, cathepsin G can be added to the list of proteases known to be capable of cleaving RANKL, which includes MMP3, MMP7, a disintegrin and metalloproteinase (ADAM)-17, and ADAM19 (6, 18, 19). The generation of sRANKL is one way that cathepsin G may make a significant contribution to the vicious cycle. However, outside of this report, little is currently known about cathepsin G in the pathologic microenvironment of bone metastases.
Although the functional role of cathepsin G is well characterized in a variety of pathologic conditions including rheumatoid arthritis (20), coronary artery disease (21), periodontitis (22), and ischemic reperfusion injury (23), this is the first report of its involvement in bone metastasis. Interestingly, however, the source of cathepsin G in rheumatoid arthritis is macrophages (20). Because osteoclasts are derived from the monocyte/macrophage lineage of cells, this represents a potential source for cathepsin G in the pathologic bone microenvironment. Currently, the other known source of cathepsin G is neutrophils. Cathepsin G is expressed at high levels on the plasma membrane of neutrophils bound to heparin and chondroitin sulfate after neutrophils degranulate (24). Whether or not neutrophils are an important source of cathepsin G at the tumor-bone interface is not currently known.
To evaluate the source of cathepsin G in the bone microenvironment, immunohistochemistry was done. These studies supported the in vitro conclusion (Supplementary data) that osteoclasts likely make a significant contribution to the observed up-regulation of cathepsin G at the tumor-bone interface. Osteoclasts stained strongly positive for cathepsin G, whereas tumor cells at the tumor-bone interface showed moderate positivity. Osteoclast precursors that interspersed within the tumor cells also stained strongly positive. Further away from the tumor-bone interface in the tumor alone area, positivity for cathepsin G among the tumor cells became weaker and more sparse. Thus, this supports that tumor cells have low, baseline expression of cathepsin G that is up-regulated through interactions with osteoclast precursors but that osteoclasts are the major source of cathepsin G at the tumor-bone interface.
In addition to osteoblasts, all three cell lines (4T1, Cl66, and Cl66M2) used in this study express RANKL as well as cathepsin G (data not shown). The constitutive, low-level expression of cathepsin G by the tumor cells may be a mechanism by which they constitutively produce low levels of sRANKL. Thus, as tumor cells enter the bone microenvironment, they produce low levels of sRANKL, which triggers osteoclast differentiation and increased cathepsin G production. As the production of cathepsin G is then amplified, sRANKL production is increased, osteoclast differentiation is subsequently increased, and sequestered growth factors are released from the bone matrix that favor tumor progression. Thus, we are left with the potential for a new model of the osteolytic vicious cycle in which cathepsin G plays a central role.
Cathepsin G is potentially a central player in the vicious cycle, and thus, it represents an exciting potential therapeutic target in the treatment of mammary tumor–induced osteolysis. To test the effects of cathepsin G inhibition in vitro, we used a coculture model. Cl66-conditioned medium induced osteoclastogenesis in RAW 264.7 cells, suggesting that Cl66 cells produce a soluble factor capable of inducing osteoclastogenesis. Given that Cl66 cells express both cathepsin G and RANKL, this soluble factor is likely sRANKL. Thus, we would expect TPCK treatment to reduce the production sRANKL, which would reduce the concentration of sRANKL in the Cl66-conditioned medium and reduce osteoclastogenesis. As expected, TPCK abrogated the production of this soluble factor and reduced osteoclastogenesis, suggesting that cathepsin G is responsible for the generation of this soluble factor. Furthermore, Cl66 cells cocultured with RAW 264.7 cells also induced osteoclastogenesis that is reduced with TPCK treatment. Osteoclastogenesis is reduced but not eliminated because signaling via full-length, membrane-bound RANKL still occurs even with TPCK treatment. Thus, this is proof-in-principle that cathepsin G inhibition has the potential to reduce osteoclastogenesis and subsequent osteolysis in vivo. Based on these data, we subsequently sought to determine if inhibition of cathepsin G in vivo would reduce mammary tumor–induced osteolysis. TPCK treatment of mice significantly reduced osteolysis at the tumor-bone interface. In fact, TPCK-treated mice showed a 63% reduction in the bone destruction index. We did not observe a difference between TPCK-treated mice and control mice in terms of tumor growth or growth kinetics, suggesting that TPCK is not toxic to tumor cells and is not inducing apoptosis. In addition, our in vitro data suggest that TPCK is not toxic to osteoclasts. Thus, the mechanism by which TPCK inhibits osteolysis is likely via interruption of signaling between tumor cells and the bone microenvironment, and our data suggest that the pathway that is interrupted is the generation of sRANKL.
This reduction in osteolysis shows the central role that cathepsin G plays in mammary tumor–induced osteolysis in vivo. TPCK is a potent inhibitor of cathepsin G but is not specific to only cathepsin G (8). TPCK is also capable of inhibiting chymotrypsin, papain, bromelain, and ficin (7). Although TPCK is not specific for cathepsin G, given its ability to potently inhibit cathepsin G, taken with the lack of toxicity to tumor cells shown by no observable difference in tumor growth or growth kinetics, the lack of toxicity to osteoclasts as shown in vitro by treatment of RAW 264.7 cells with both TPCK and sRANKL, and the proposed mechanism by which inhibition of cathepsin G would reduce sRANKL and specifically decrease osteolysis, we believe that inhibition with TPCK adequately shows the role of cathepsin G in mammary tumor–induced osteolysis.
In conclusion, this study shows that cathepsin G plays an important role in mammary tumor–induced osteolytic lesions by contributing to the vicious cycle. It is significantly up-regulated at the tumor-bone interface and is capable of generating sRANKL, which potentially enhances osteoclast activation and osteolysis. We have also shown that inhibition of cathepsin G in vitro and in vivo reduces osteoclastogenesis and subsequent osteolysis. Thus, we have shown a central role for cathepsin G in the establishment of mammary tumor–induced osteolytic lesions. Further studies are needed to reveal the potential subsequent roles of cathepsin G, in addition to generation of soluble RANKL, in the bone tumor microenvironment. This study reveals cathepsin G as an appealing therapeutic target in mammary tumor–induced osteolysis.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
T.J. Wilson is a Howard Hughes Medical Institute Research Training Fellow.
Grant support: Grant CA72781 and Cancer Center Support Grant P30CA036727 from the NIH (R.K. Singh), Nebraska Department of Health and Human Services, Nebraska Research initiative (R.K. Singh), and the Howard Hughes Medical Institute Research Training Fellowship (T.J. Wilson).
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 Michelle Varney for careful review and critique of the manuscript.