Introduction
Osteosarcoma (OS) is the most common malignant bone tumor in adolescents and childhood. Despite combination treatment with wide resection and chemotherapy, lung metastases occur in ~40–50 % of OS patients [
1,
2], resulting in poor prognosis [
2]. Elucidation of the underlying mechanisms and new targets for the treatment of lung metastasis are strongly needed to improve prognoses for OS patients. The metastatic cascade in malignant tumor involves a multi-step progression: detachment and local invasion at the primary site, entry into the circulation (intravasation), survival in the bloodstream, migration though the endothelium (extravasation), and colonization of target organs [
3‐
7]. The details of the metastatic process remain mysterious due to difficulties in studying cell behavior at sufficiently high spatial and temporal resolution in vivo [
5]. We have performed detailed dynamic analysis of each step of the metastatic cascade using the LM8 mouse spontaneous highly metastatic OS cell line [
8] and the parental Dunn line [
9]. This syngeneic metastatic model offers the benefit of allowing us to investigate the biological features of metastasis through comparison of results for LM8 and Dunn cells against a consistent genetic background. We have already reported that LM8 cells show high secretion vascular endothelial growth factor (VEGF) [
8], a fibroblastic morphology with abundant filopodia and invasive properties [
8,
10] and migration ability for approximately two times in migration assays using a Boyden chamber [
10,
11], which showed activated Cdc42 and autophosphorylation of focal adhesion kinase compared to Dunn cells [
10].
The presence of circulating tumor cells (CTCs) in the bloodstream fits very well with this cascade, and is involved in intravasation, survival in circulation and extravasation steps, consequently detection of CTCs has long been considered as a possible tool for assessing the aggressiveness of tumors and subsequent development in distant organs [
7,
12,
13]. VEGF as a key protein involved in the angiogenic switch in tumors [
14], is also known as a vascular permeability factor and has been reported to decrease biological barrier function, promoting vascular permeability and extravasation via VEGF–VEGFR interactions [
15,
16]. Kaya et al. [
17,
18] reported serum VEGF levels as a predictor of lung metastasis and poor prognosis in patients with OS. This study aimed to clarify critical steps toward pulmonary metastasis using LM8 and parental Dunn, in order to seek new candidate molecules for suppressing the development of lung metastasis.
Materials and methods
Reagents
Pazopanib was kindly gifted from GlaxoSmithKline (London, UK). Recombinant murine VEGF165 was purchased from PeproTech (Rocky Hill, NJ). FITC-dextran was purchased from TdB Consultancy AB (Uppsala, Sweden). The collagen gel culture kit was purchased from Nitta Gelatin (Osaka, Japan). The Mouse VEGF Quantikine enzyme-linked immunosorbent assay kit was purchased from R&D Systems (Minneapolis, MN). CellTrackerTM Red CMTPX was purchased from Invitrogen (Carlsbad, CA).
Cells
The LM8 highly metastatic OS cell line was derived from the Dunn OS cell line through eight repeated cycles of the procedure described by Poste and Fidler [
19]. Mouse aortic endothelial cells (mAEC) were purchased from Angio-Proteomie (Boston, MA). LM8 and Dunn cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen, Carlsbad, CA) supplemented with 10 % fetal bovine serum (FBS), and mAEC were maintained in DMEM supplemented with 20 % FBS. Cells were cultured at 37 °C in a fully humidified incubator under 5 % CO
2.
CTCs culture
Forty-microliter peripheral blood samples were collected from the tail vein and 1 μl of EDTA (0.5 μmol) added. These samples were maintained in DMEM supplemented with 10 % FBS and penicillin (100 U/ml)-streptomycin (100 μg/ml). All cells were cultured at 37 °C in a fully humidified incubator under 5 % CO2.
Suspension culture
Poly-hydroxyethyl methacrylate (poly-HEMA) (Sigma, St. Louis, MO) was solubilized in 95 % methanol (30 mg/ml). To prepare poly-HEMA-coated dishes, 25-μl aliquots of poly-HEMA solution were placed onto 96-well dishes and dried in a tissue culture hood. Five thousand trypsinized LM8 or Dunn cells per 100 μl of medium were plated onto 96-well poly-HEMA-coated dishes and cultured for 24–48 h.
Cell proliferation assay
Cell proliferation was measured using CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI) according to the protocols provided by the manufacturer.
Measurement of caspase-3/7 activity
Caspase-3/7 activation was measured using the Caspase-Glo 3/7 Luminescence Assay (Promega) according to the protocol from the manufacturer.
Three dimensional culture of LM8 and Dunn
To assess cell morphology and proliferation in 3D collagen matrix, LM8 and Dunn cells were cultured using a 3D collagen cell culture system (Nitta Gelatin, Osaka, Japan). Briefly, collagen gel solution was prepared according to the instructions from the manufacturer. Collagen gel concentration at the bottom layer was 2.4 mg/ml, as while 1.2 or 2.4 mg/ml for the upper layer. After polymerization of collagen gel in the bottom layer, cells were suspended in collagen gel solution. Cell suspensions were added to the dish, then immediately transferred to a 37 °C incubator for 60 min to initiate polymerization of collagen. After formation, the collagen gel was covered with culture media.
Transendothelial migration assay
Transendothelial migration assay was performed using 24-well HTS FluoroBlok™ 8.0-μm colored PET membrane inserts (BD Biosciences). Forty-five thousand mAEC were applied to the upper chamber after coating the upper surface of the membrane with 30 μg/ml of fibronectin. After 24 h culture of mAEC, LM8 or Dunn cells were applied to the upper chamber after LM8 or Dunn cells were stained using CellTracker™ Red CMTPX according to the manufacturer’s protocol. After 12 h after application, LM8 or Dunn cells had migrated through the endothelial layer and PET membrane pores were detected using fluorescent light (excitation, 577 nm; emission, 602 nm) from the bottom side of the membrane. Attachment of LM8 or Dunn cells to mAEC and mitosis of mAEC were monitored using time-lapse photography with a Cool Snap Cf camera (Roper Scientific, Ottobrunn, Germany) and an IX70 inverted microscope (Olympus, Tokyo, Japan).
RNA isolation, reverse transcription, and polymerase chain reaction (PCR)
Total RNA was purified using the TRIzol reagent (Invitrogen). Total RNA (1 μg) was used as a template for reverse transcription using a High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA) according to the instructions from the manufacturer. PCR was performed with Taq DNA Polymerase (Promega, Madison, WI) using the indicated primer sets (forward and reverse respectively) for the following genes: Vegf-a, 5′-CATGCGGATCAAACCTCAC-3′ and 5′-TTCTGGCTTTGTTCTGTCTTTC-3′; Vegfr1, 5′-CTCGGGTGTCTGCTTCTCA-3′ and 5′-CTCAGCCTTTTGTCCTCCTG-3′; Vegfr2, 5′-TTTGGCAAATACAACCCTTCAGA-3′ and 5′-GCAGAAGATACTGTCACCACC-3′ and Gapdh, 5′-AGGTCGGTGTGAACGGATTTG-3′ and 5′-GCAGAAGATACTGTCACCACC-3′.
Permeability assay
The mAEC (4.5 × 104) were plated in 24-well Bio-Coat cell migration chambers (diameter, 6.4 mm; pore size, 8.0 μm, BD Biosciences, Bedford, MA), grown for 24 h, and serum starved for 4 h. DMSO (2 μl), pazopanib (1 μM), VEGF (100 ng/ml) and a combination of pazopanib (1 μM) and VEGF (100 ng/ml) were added to the upper chamber (total 500 μl) containing 100 μg/ml of FITC-labeled dextran (2,000 kDa). At 0, 5, 15, 30, and 60 min, 100 μl of media was removed from the lower chamber and the amount of FITC-labeled dextran present was determined using a fluorescence spectrophotometer (1420 Multilabel Counter; PerkinElmer, Norwalk, CT). Values represent the mean of results from triplicate trials for all experiments.
Mouse model
Specific pathogen-free 5- to 6-week-old C3H/He mice were used in this study (SLC, Shizuoka, Japan).Pazopanib solution, prepared as described previously [
20], was orally administered at 100 mg/kg/day for 21 days. In order to generate a primary tumor resection model of OS, LM8 cells (1 × 10
7 cells/200 μl of PBS) were injected into the subcutaneous tissue of the backs of syngeneic C3H mice. The primary tumor was resected under anesthesia at 11 days after injection. C3H mice were estimated as likely to die from pulmonary metastasis by 5–6 weeks after primary tumor resection. Thus, for ethical reasons, histological evaluations, count of pulmonary metastatic foci and CTCs culture were performed at 3 weeks after primary tumor resections. Serum VEGF concentration was measured using a Mouse VEGF Quantikine ELISA kit (R&D Systems), according to the instructions from the manufacturer. Serum pazopanib concentrations were measured at the institute of GlaxoSmithKline. All animal experiments were approved by the institutional animal experiments review committee, and all animals were euthanized with diethyl ether at the end of experiments.
Phase contrast microscopy
Fluorescence and phase-contrast images were obtained with an IX70 microscope (Olympus). Images were analyzed and processed for presentation using brightness and contrast adjustment with ImageJ software (NIH Software, Bethesda, MD).
Statistical analysis
The significance of differences was evaluated using a two-sided Student’s t test for biological assays, and two-sided Mann–Whitney’s U test for animal experiments. Values of P < 0.05 were considered statistically significant. All analyses were performed using Excel software (Microsoft, Redmond, WA) and Statcel3 software (OMS Publishing, Saitama, Japan).
Discussion
In this study, we reported that LM8 displayed several superior biological characteristics in all metastatic steps rather than just a single step, by assessing the details of the metastatic cascade in a step-by-step manner using experimental systems to mimic in vivo conditions.
Previous studies have reported that isolation or detection of living CTCs was too difficult, because phenotype would be changed between primary site and CTCs in their quite different microenvironments, so the characteristics of these cells have remained unclear [
7,
12,
13]. In this study, we were able to culture living CTCs chronologically not only from mice with LM8 but also with Dunn. These results corresponded to the clinical reports that showed identification of CTCs in the peripheral blood of patients with early-stage localized prostate and breast cancer, as well as metastatic cancers using CellSearch systems (Veridex, North Raritan, NJ), as correlating with response to chemotherapy in patients with metastatic cancer [
23‐
29]. These methods measured only CTCs number but not analyzed its characterization. By contrast, our method could detect not only living CTCs but also analyze biological feature—higher proliferation ability of CTCs from LM8 compared to cells from primary site in suspension culture, while CTCs from Dunn did not show such a character. So far, there is a fear that phenotype of CTCs may be changed from original status while culturing in 2D condition. To clarify the precise mechanism, further improvement and combination with single cell analysis method would be required.
During the colonization step, LM8 also showed higher proliferation ability compared to Dunn in conditions that mimicked in lung environment (−150 Pa), but not in stiffer condition (800–1,200 Pa) [
21,
22]. In a mouse fibrosarcoma model, the proliferation rate of highly metastatic potential cell lines was higher than that of non-metastatic or intermediate potential cell lines in stiffer condition [
30]. Thus, our result opposed the previous report. The precise mechanism which LM8 preferred lung-mimicking condition than Dunn was unknown, and further examination should be required to clarify the mechanism.
The physiological role of VEGF has been reported that neovascularization during embryonic development, skeletal growth, and reproduction. In addition, overexpression of VEGF in tumor could stimulate vasculogenic and angiogenic switches [
14]. We have previously shown that VEGF level secreted from LM8 was higher than that from Dunn [
8]. Here, we speculated that VEGF from tumor cells increased number of endothelial cell mitosis, resulting in opening cell–cell junction among endothelial cells, following to enhanced transendothelial migration. In fact, transendothelial migration was reduced by blocking the VEGF–VEGFR signal with pazopanib (Votrient; GlaxoSmithKline), multiple TKI targeting VEGFRs [
20]. This compound has been approved by the FDA for the treatment of advanced renal cell carcinoma (2009) and soft tissue sarcoma (2011). We first checked the effect of pazopanib in vivo tumor bearing mouse model (Supplementary Fig. 7). Administration of pazopanib significantly reduced only metastatic foci in lung without affecting primary site growth. To elucidate the effect on lung metastasis in clinical status more clearly, we next prepared a primary tumor resection model (Fig.
5a). We finally confirmed that pazopanib administration significantly reduced lung metastatic foci and the parallel correlation among CTCs colony number, metastatic foci and serum VEGF concentration. In this experiment, mean serum pazopanib concentrations were 106.1 ± 37.4 μM at 8–9 h after final oral administration. In terms of human pharmacokinetics, daily dosing at 800 mg results in a geometric mean C
max of 132 μM. Thus, our experimental dose of pazopanib was comparable to the dose used clinically in humans.
Lastly, we focused on parallel correlations among lung metastases, the number of CTCs and serum VEGF concentration. The source of VEGF was considered based on the following three points in this experiment. First, at the time of primary tumor resection, median colony number 11 per 40 μl blood sample (Fig.
1b, bottom), thus the number of CTCs in whole blood (~1 ml) was estimated ~275. It is unlikely that VEGF secreted from only 275 cells influence total concentration of VEGF in blood. Second, because pazopanib did not show anti-tumor effects under suspension conditions (Supplementary Fig. 8), CTCs might be died in circulation before CTCs overcome the extravasation and colonization steps. In fact, Pantel and Brakenhoff [
31] have reported that CTCs survived for only a short period of time. Third, the existence of tumor burden is important for CTCs as tumor cell resource. Several reports have been shown that the bone marrow as an important reservoir of tumor cells for many types of malignancy [
31,
32]. However, we did not detect tumor cells in bone marrow in mice with LM8 even in 5 weeks after tumor transplantation (Table
1). Thus, metastatic sites would be considered as a tumor cell resource, and might supply new CTCs and VEGF. Based on these inferences, (1) CTCs could not pass through endothelium (extravasation) by blocking VEGF–VEGFR interaction using pazopanib, (2) reducing metastatic foci in lung, (3) decreasing amount of VEGF and number of CTCs from metastatic site, (4) blocking extravasation step as vicious cycle.
So far, VEGF–VEGFR blockage therapies using monoclonal antibodies or small molecules approved by the FDA for the treatment of various cancers have proven effective for extending progression-free survival [
33‐
35]. On the other hand, increasing micro-metastasis in certain cancers has been arisen as an issue of great concern [
36,
37], because previous preclinical drug development did not consider the impact on metastasis [
38]. In this study, we propose that anti-VEGF therapy using small molecules could offer a potentially useful strategy for inhibiting transendothelial migration in vitro and lung metastasis in vivo from OS.
Acknowledgments
Pazopanib was a generous gift from GlaxoSmithKline Co. We thank GlaxoSmithKline Co for measuring serum Pazopanib concentrations. We also thank Dr. Ueda (Department of Orthopaedic surgery, Osaka National Hospital, Japan) for helpful discussion on experimental data, Drs. Takenaka, Outani, Yasui, Imura (Department of Orthopaedic surgery, Osaka University Graduate School of Medicine, Japan) and Dr. Sasagawa (Department of Biology, Osaka Medical Center of Cancer and Cardiovascular Diseases, Japan) for technical support. This study was supported in part by the Program of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (23659734, to KI), the Grant-in-Aid for Scientific Research (B) and Exploratory Research (23390372, to KI), the Grant-in-Aid for Scientific Research (C) (23592202, to KY), and the Grant-in-Aid for Scientific Research (C) (24592233, to NN).