The osteosarcoma cell lines U-2 OS, SaOS-2, 143B, MG-63 and HOS, and the human hFOB1.19 osteoblast and mouse NIH3T3 fibroblast cell lines, were purchased from the American Type Culture Collection, and cultured in Dulbecco's modified Eagle's medium (DMEM, Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (HyClone; GE Healthcare Life Sciences) in a 37°C incubator containing 5% CO₂. MG-63 and U-2 OS cells were selected for subsequent experiments (transfection, migration and invasion assays, and luciferase assays) due to their increased GDF15 expression. For simulating extracellular secretion of GDF15, cells were treated with 50 ng/ml recombinant human GDF15 (rhGDF15; R&D Systems, Inc.) at 37°C for 24 h.

The present study was approved by the Ethics Committee of the Affiliated Foshan Chancheng District Center Hospital of Guangdong Medical University according to the 1975 Declaration of Helsinki (ethics approval no. IRB-ATT-001-24). All clinical samples were collected from the Affiliated Foshan Chancheng District Center Hospital of Guangdong Medical University. Patient written informed consent forms were obtained prior to the use of clinical specimens for research purposes. In the present study, 106 serum samples and 10 fresh tissues were collected from 106 patients (67 males and 39 females, aged 12–58 years), diagnosed pathologically with osteosarcoma, between January 2005 and December 2009. All patients received standard neoadjuvant chemotherapy, followed by resection of the tumor and postoperative chemotherapy. Clinical follow-up information, including Enneking staging (a system for staging bone cancer) (16), overall survival and pulmonary metastasis-free survival time, was available for all patients enrolled in the study. All serum samples (67 non-metastatic and 39 pulmonary metastatic osteosarcoma samples) were collected from the peripheral blood of patients with osteosarcoma prior to systemic treatment. The serum was obtained from blood samples by centrifugation at 1,000 × g for 10 min at 4°C, and then stored at −80°C and thawed prior to analyses. Fresh osteosarcoma samples for reverse transcription-quantitative PCR (RT-qPCR) and ELISA analyses, including 4 non-metastatic and 4 pulmonary metastatic osteosarcoma tissues, and 2 adjacent non-tumor soft tissues, were obtained from 10 of the 106 aforementioned patients with osteosarcoma at the time of surgical resection, and immediately frozen at −80°C until further use. In the present study, 106 serum samples and 10 fresh tissues were collected between January 2005 and December 2009. All patients received standard neoadjuvant chemotherapy, followed by resection of the tumor and postoperative chemotherapy.

Total RNA samples from osteosarcoma cell lines and fresh surgical osteosarcoma tissues were extracted using TRIzol^® reagent (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocols. The isolated RNA was pretreated with RNase-free DNase. Then, the mRNA levels of GDF15 were evaluated by RT-qPCR. Total RNA (2 µg) from samples was reverse transcribed to cDNA using M-MLV Reverse Transcriptase (Promega Corporation) according to the manufacturer's protocols. Briefly, total RNA (2 µg) was incubated with random primers, heated to 70°C for 5 min and then placed on ice. A mix containing 1X M-MLV reaction buffer, dNTPs (all 0.5 mM), Recombinant RNasin^® Ribonuclease Inhibitor (25 U) and M-MLV reverse transcriptase (200 U; Promega Corporation) was added to each sample at 37°C for 1 h. All samples, including cDNAs, gene-specific primers and SYBR-Green (Roche Applied Sciences) were heated to 95°C for 5 min and then amplified for 35 cycles consisting of 95°C for 30 sec, a 30 sec annealing step at a primer-specific annealing temperature, and 72°C for 45 sec. All reactions were then incubated at 72°C for 7 min and cooled to 4°C. qPCR analysis was performed on an ABI Prism 7500 Sequence Detection System (Applied Biosystems; Thermo Fisher Scientific, Inc.). The results were normalized to the expression of the housekeeping gene GAPDH. Relative expression levels were determined using the following formula: 2^{−[(Cq of gene)-(Cq of GAPDH)]}, where Cq represents the threshold cycle for each transcript (17). The oligonucleotide primers are listed in Table I.

The concentrations of GDF15 in serum and cell culture medium samples were determined using ELISA kits (1:250; cat. no. DGD150; R&D Systems, Inc.). ELISAs were conducted according to the manufacturer's instructions. Briefly, the supernatant was transferred to a well coated with GDF15 monoclonal antibody (cat. no. MAB957; R&D Systems, Inc.) and immunosorbed using biotinylated polyclonal anti-human GDF15 antibody (1:1,000; cat. no. BAF940, R&D Systems, Inc.) at room temperature for 1 h. The color development was catalyzed by horseradish peroxidase, and the absorption was detected at 450 nm. The protein concentration was determined by comparing the relative absorbance of each sample with the standards. Patients were assigned to high and low GDF15 expression groups based on the results of the ELISA. Patients with serum GDF15 concentrations <1 ng/ml were assigned to the low expression group; patients with serum GDF15 concentrations ≥1 ng/ml were assigned to the high expression group.

Short-hairpin RNA (shRNA) against GDF15 (sense sequence: 5′-CTATGATGACTTGTTAGCCAA-3′) in a Plko.1-puro vector was commercially purchased (Sigma-Aldrich; Merck KGaA). Retroviruses containing pLKO.1-puro-GDF15-Ri or pLKO.1-puro-vector were generated by transfecting 293FT cells (Invitrogen; Thermo Fisher Scientific, Inc.). Virus particles were then examined by spectrophotometry at 260 nm and the infectivity of adenovirus stocks was evaluated using an Adeno-X Rapid Titre kit (Clontech Laboratories, Inc.), which determines infectious replication in 293FT cells. The virus particle/infectious focus-forming units (IFU) ratio was then calculated. The virus stocks used in the paper were GDF15-Ri (IFU ratio 43). MG-63 or U-2 OS cells (2×10⁵) were seeded and infected with 500 µl/day retroviral particles for 3 days. The stable cell lines were identified following treatment with 0.5 µg/ml puromycin for 10 days. The (CAGAC) 12/pGL3 TGF-β/SMAD-responsive luciferase reporter plasmid and control plasmids (Clontech Laboratories, Inc.) were used to quantitatively evaluate the transcriptional activity of TGF-β signaling components. All transfection experiments were performed using the Lipofectamine^® 3000 reagent (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocols, and cells were collected 24 h later for subsequent experiments.

Nuclear fractions were prepared using the Nuclear Extraction kit (Active Motif, Inc.) according to the manufacturer's protocols. Western blotting was conducted according to a standard method, as previously described (18). Briefly, cells were lysed, and protein sample (20 µg/lane) was separated via 10% SDS-PAGE and transferred to PVDF membranes. Following blocking with 5% (w/v) skim milk for 1 h at room temperature, membranes were incubated overnight at 4°C with primary antibodies. Antibodies against phosphorylated (p)-mothers against decapentaplegic homolog (SMAD)-2 (1:1,000; cat. no. 3108), p-SMAD3 (1:1,000; cat. no. 9520), SMAD2 (1:1,000; cat. no. 5339), and SMAD3 (1:1,000; cat. no. 9513) were purchased from Cell Signaling Technology, Inc. The GDF15 monoclonal antibody (1:1,000; cat. no. MAB957) was purchased from R&D Systems, Inc. Anti-α-tubulin (1:2,000; cat. no. T6199, Sigma-Aldrich; Merck KGaA) and anti-P84 (1:1,000; cat. no. ab102684; Abcam) antibodies were used as loading controls. Then the membrane was incubated with a secondary antibody for 1 h at room temperature. The secondary antibodies, goat anti-rabbit immunoglobulin G (1:2,000; cat. no. 31460, Pierce; Thermo Fisher Scientific, Inc.), and goat anti-mouse immunoglobulin G (1:2,000; cat. no. 31430, Pierce; Thermo Fisher Scientific, Inc.), were used in this study. Protein bands were visualized using an enhanced chemiluminescence kit (EMD Millipore).

Wound-healing and cell invasion assays. Wound-healing and cell invasion assays were conducted as described previously (19) to determine the migratory and invasive abilities of osteosarcoma cells. For the wound healing assay, in brief, 2×10⁵ cells were seeded in 6-well plates, and the cell layer at ~90% confluence was wounded using a sterile tip. The extent of wound closure was evaluated using a light microscope (magnification, ×100) following incubation at 37°C for 24 h. For the invasion assay, in brief, 2×10³ cells suspended in serum-free DMEM were seeded into the upper Transwell chambers coated with Matrigel, whereas DMEM with 20% FBS was seeded into the lower chambers. Then, cells were incubated at 37°C for 24 h. The cells were fixed at 4% paraformaldehyde for 20 min and stained with 1% crystal violet for 1 min (both at room temperature), and the number of cells that reached the lower membrane was counted using a light microscope (magnification, ×200). All experiments were performed in triplicate.

Luciferase assays were conducted as described previously (20). Briefly, cells (1×10⁴) were seeded in triplicate in 48-well plates, and cultured at 37°C for 24 h. Then, 100 ng of (CAGAC) 12/pGL3 reporter luciferase plasmid or control luciferase plasmid, and 3 ng pRL-TK Renilla plasmid (Promega Corporation) were transfected into shRNA- or rhGDP15-treated cells using Lipofectamine^® 3000 (Thermo Fisher Scientific, Inc.) according to the manufacturer's protocols. The luciferase and Renilla signals were measured 24 h following transfection using a Dual Luciferase Reporter Assay kit (Promega Corporation) according to the manufacturer's protocols; luciferase activity was normalized to Renilla activity.

The microarray gene expression profiles in GSE9508 (21) were downloaded from the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/). The datasets were analyzed for the expression of GDF15. Information on the clinical characteristics, including metastasis status, were obtained from the respective clinical information files.

Gene expression was graphed into heatmaps using MeV version 4.9 software (http://mev.tm4.org). The pseudocolours represent the intensity scale of fold change of GDF15-Ri vs. vector, or GDF15-Ri + rhGDF15 vs. GDF15-Ri in MG-63 and U-2 OS, generated by log2 transformation.

All statistical analyses were conducted using SPSS version 17.0 software (SPSS, Inc.). Data are presented as the mean ± standard deviation of three independent experiments. The associations between GDF15 protein expression and the clinicopathological characteristics of osteosarcoma patients were analyzed using χ² tests. Survival curves were generated using the Kaplan-Meier method and survival was analyzed using the log-rank test. The significance of various variables for survival was analyzed to predict prognostic value using the Cox proportional hazards regression model in univariate and multivariate analyses. For the comparison of two groups, P-values were calculated with a Student's t-test. One-way analysis of variance followed by a Newman-Keuls multiple comparison test was used to perform comparisons between multiple groups. P<0.05 was considered to indicate a statistically significant difference.

GDF15 expression was first investigated in human osteosarcoma tissues by analyzing the GEO dataset GSE9508, containing gene expression profiles for metastatic and non-metastatic osteosarcoma biopsies. Notably, the mRNA levels of GDF15 were significantly increased in metastatic osteosarcoma tissues compared with in non-metastatic osteosarcoma tissues (Fig. 1A). Similar results were obtained when analyzing the mRNA and protein expression of GDF15 in 4 non-metastatic osteosarcoma tissues, 4 pulmonary metastatic osteosarcoma tissues and 2 adjacent non-tumor soft tissues (Fig. 1B). RT-qPCR, western blot and ELISA analyses were conducted to determine GDF15 mRNA and protein expression in five osteosarcoma cell lines (MG-63, HOS, U-2 OS, Saos-2 and 143B), human hFOB1.19 osteoblasts and mouse NIH3T3 fibroblasts. As presented in Fig. 1C and D, GDF15 was expressed at high levels in osteosarcoma cell lines compared with FOB and NIH3T3 cell lines.

To further investigate the clinical relevance of serum GDF15 levels in patients with osteosarcoma, a total of 106 serum samples were examined for GDF15 expression by ELISA (Table II). χ² tests revealed that high serum levels of GDF15 were significantly associated with vital status (P=0.010) and pulmonary metastasis (P<0.001) in patients with osteosarcoma (Table III); however, there were no significant associations between GDF15 protein expression and other clinicopathologic characteristics.

Kaplan-Meier survival analysis and log-rank tests indicated that patients with osteosarcoma possessing high serum levels of GDF15 exhibited significantly poorer OS and reduced PMFS compared with those with low serum GDF15 levels (Fig. 2A and B). In addition, multivariate survival analyses revealed that serum GDF15 expression (P=0.004) was an independent prognostic factor for poor OS (Table IV), and that serum GDF15 expression (P<0.001) and Enneking stage (P=0.009) were independent prognostic factors for short PMFS (Table V). Thus, these results suggested that serum GDF15 may be a valuable prognostic marker in osteosarcoma.

To explore the potential oncogenic functions of GDF15 in osteosarcoma cells, migration and invasion assays were performed. MG-63 and U-2 OS cells (selected for their high GDF15 expression) were transduced with lentiviruses carrying GDF15-shRNA or vector control, with downregulated expression demonstrated via RT-qPCR, western blot and ELISA analyses (Fig. 3A and B). As presented in Fig. 3C and D, GDF15-shRNA significantly suppressed the migration and invasion of MG-63 and U-2 OS osteosarcoma cells, whereas treatment with rhGDF15 rescued the migratory and invasive abilities of GDF15-silenced cells. Collectively, these results suggested that GDF15 knockdown inhibited the metastatic ability of osteosarcoma cells.

GDF15 is a member of the TGF-β superfamily (22), and TGF-β signaling is an important pathway for maintaining metastatic phenotypes in osteosarcoma (23). Therefore, whether GDF15 promotes the migratory and invasive abilities of osteosarcoma cells via the activation of the TGF-β signaling pathway was subsequently investigated. As presented in Fig. 4A-C, silencing GDF15 notably inhibited p-SMAD2/3 nuclear translocation, TGF-β signaling activity and downstream target gene expression, whereas rhGDF15 reversed these effects in GDF15-silenced cells. These findings indicated that GDF15 knockdown suppressed the TGF-β signaling pathway, and that the TGF-β signaling may be involved in the GDF15-induced metastasis of osteosarcoma cells (Fig. 4D).