To investigate the expression of activin A in blood vessels and in tumor cells, immunohistochemical analyses with anti-activin A and anti-CD34 antibodies were performed with 95 cases of primary OSCC from The Oncology Center of Cascavel (Paraná, Brazil) and The UOPECCAN Cancer Hospital (Paraná, Brazil), which were previously described (19). The samples were obtained between December 2012 and October 2013. The age range of the patients was 31-84 years, with a median of 56 years. OSCC consecutive slices from paraffin-embedded samples were subjected to immunostaining for activin A and CD34. Briefly, after dewaxing in xylene for 30 min and hydration in graded alcohol solutions (100, 90, 80, 70, 60 and 50%; 2 min/wash), the 3-µm sections were treated with 3% hydrogen peroxide followed by antigen retrieval with 10 mM citrate buffer pH 6.0 in a pressure cooker for 15 min. Following washing with PBS, the sections were treated with 1% bovine serum albumin in PBS at room temperature for 1 h and then incubated overnight at 4°C with polyclonal rabbit activin A antibody (cat. no. HPA020031; 1:100; Sigma Aldrich-Aldrich) or monoclonal mouse CD34 antibody (cat. no. M7165; 1:400, clone QBEnd-10, Dako; Agilent Technologies, Inc.) at room temperature for 1 h followed by LSAB kit (LSAB+ system-HRP kit; Dako; Agilent Technologies, Inc.) or Advance HRP kit (Dako; Agilent Technologies, Inc.), respectively. Reactions were developed by incubating the sections with 0.6 mg/ml 3,3'-diami-nobenzidine tetrahydrochloride (Dako; Agilent Technologies, Inc.) and counterstained with hematoxylin at room temperature for 2 min. Control reactions were performed by omission of the primary antibodies.

Activin A immunoexpression was assessed in blood vessels and tumor cells by two independent pathologists, who were not aware of any clinical data. Activin A expression in blood vessels was determined by counting of activin A-positive and activin A-negative blood vessels in three magnifcation, x200 fields per sample with a light microscope, and samples were categorized into two groups based on median expression: Low (≤53% of activin A-positive blood vessels) and high (>53% of activin A-positive blood vessels). For tumor cells, the number of positive cells was graded according to the following: 1, 1-25% staining; 2, 26-50% staining; 3, 51-75% staining; and 4, 76-100% staining); and the intensity of staining was scored as follows: 0, negative; 1, weak staining; 2, moderate staining; and 3, strong staining. The two grades were added together, producing scores from 0 to 7 that were classified as low (scores 0-4) and high (scores 5-7) expression for comparative analysis. Written informed consent was obtained from each patient according to the declaration of Helsinki, and the study was approved by the Human Research Ethics Committee of the School of Dentistry, University of Campinas (approval no. 100/2012, October 10, 2012).

HUVECs were obtained from American Type Culture Collection and cultured as recommended in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F12 medium (DMEM/F12; Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% fetal bovine serum (FBS; Cultilab), 400 ng/ml hydrocortisone (Sigma-Aldrich; Merck KGaA) and antibiotic-antimycotic (penicillin, streptomycin and amphotericin B solution; Thermo Fisher Scientific, Inc.). The SCC-9 ZsGreen LN-1 cells, isolated from a meta-static cervical lymph node (20), were cultured in DMEM/F-12 (Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% FBS, 400 ng/ml hydrocortisone (Sigma-Aldrich; Merck KGaA) and antibiotic-antimycotic. shINHBA OSCC cells (SCC-9 ZsGreen LN-1 cells constitutively expressing a shRNA sequence against activin A) and the respective control (shControl OSCC cells; cells stably transfected with the same vector encoding a non-targeting scrambled shRNA sequence) were generated as previously described (11). All cells were cultured at 37°C in a humidifed atmosphere of 5% CO₂.

Lyophilized recombinant activin A (Sigma-Aldrich; Merck KGaA) and follistatin (R&D Systems, Inc.) were dissolved in culture medium, aliquoted and stored at ‑80°C. To assess the effect of activin A, cells were cultured in DMEM/F-12 containing 0 or 1 ng/ml of recombinant activin A for 24 h at 37°C. For follistatin treatment, cells were cultured at 100 ng/ml for 24 h at 37°C. Concentrations of activin A and follistatin were selected based on our previous study (11).

HUVECs grown in 12-well plates at a confiuence of 50% were incubated with control shRNA lentiviral particles (MISSION^® pLKO.1-puro Non-Mammalian shRNA Control; Sigma-Aldrich; Merck KGaA) or INHBA shRNA lentiviral particles (INHBA MISSION^® shRNA Lentiviral Transduction Particles, NM_002192; 5'A-GACCCATGTCCATGTTGT-3'; Sigma-Aldrich; Merck KGaA) at a multiplicity of infection of 1.5 in culture medium containing 8 µg/ml polybrene (Sigma-Aldrich; Merck KGaA) for 8 h. After washing with PBS, cells were cultured in fresh medium for an additional period of 48 h. Cells were then split in a 1:5 dilution, and cultured for 10 days in the presence of 1 µg/ml puromycin dihy-drochloride (Sigma-Aldrich; Merck KGaA) to select resistant cells. The effcacy of activin A‑knockdown was determined by reverse transcription-qPCR (RT-qPCR) and enzyme-linked immunosorbent assay (ELISA).

Total RNA was extracted using the RNeasy mini kit (Qiagen, Inc.), according to the manufacturer's protocol. Following DNase I treatment in order to eliminate genomic DNA contamination, 1 µg total RNA per sample was used to generate cDNA using Oligo-dT (Invitrogen; Thermo Fisher Scientifc, Inc.) and reverse transcriptase (Superscript II RT enzyme; Invitrogen; Thermo Fisher Scientifc, Inc.), according to the manufacturer's protocol. The resulting cDNAs were subjected to qPCR using specifc primers and SYBR Green PCR master mix (Applied Biosystems; Thermo Fisher Scientifc, Inc.) in a Real Time PCR thermal cycler (Applied Biosystems; Thermo Fisher Scientifc, Inc.). The thermocycling conditions were: 95°C for 1 min followed by 40 cycles 95°C for 15 sec and 60°C for 30 sec. Gene expression was determined using the 2^-ΔΔCq method (21) and PPIA was used as reference gene for data normalization. All reactions were performed in triplicate. Primer sequences are provided in Table SI.

HUVECs were plated at concentration of 25,000 cells per well in a 24-well culture plate in triplicate and cultured for 24 h. After washing, the cells were cultured in serum-free medium for an additional 24 h. The medium was then collected, centrifuged to remove foating cells, and used for analysis. The concentration of activin A was determined using a human activin A ELISA kit (cat. no. RAB0324; Sigma-Aldrich; Merck KGaA), according to the manufacturer's instructions.

HUVECs (40,000 cells per well) were seeded into wells of a 96-well plate coated with 50 µl Matrigel (BD Biosciences). After 12 h of incubation, tube formation was observed under an inverted microscope (magnifcation, x40; Nikon Corporation), photographed and analyzed using Motic Images Plus 2.0 (Nikon Corporation), as previously described (22).

HUVECs were plated in 96-well plates at a density of 10,000 cells per well in 100 µl medium containing 10% FBS. After 16 h, the cells were washed with PBS and cultured in serum-free medium for an additional 24 h to reach cell cycle synchronism. Following serum starvation, cells were treated for 24 h and proliferation rates were determined by measuring bromodeoxyuridine (BrdU) incorporation into DNA using a cell proliferation ELISA, BrdU (colorimetric) kit (cat. no. 11647229001; Roche Applied Science; Merck KGaA).

Transwell migration assays were performed in 6.5-mm inserts with 8-µm pore size (Corning, Inc.). Serum-starved cells (80,000 cells/well) were plated into the upper chamber in 200 µl serum-free DMEM/F12. As a chemoattractant in the lower chamber, depending on the assay, serum-free medium containing activin A or follistatin, and shINHBA OSCC- and shControl OSCC-conditioned medium were tested. Activin A or follistatin were also added in the upper chamber, in direct contact with the HUVEC cells, and in those situations 10% FBS was used in the lower chamber as chemoattractant. After 24 h, assessment of migration was performed by gently removing cells in the interior part of the insert with a cotton swab. Cells on the underside of the membrane were fxed in 10% formalin for 15 min and stained with 1% toluidine blue in 1% borax solution, with both procedures at room temperature. The excess dye was washed out and cells were then eluted in 1% SDS solution for 5 min. Absorbance was measured at 650 nm.

To analyze the expression of pro- and anti-angio-genic factors, RNA from HUVEC cells treated with 1 ng/ml of activin A for 24 h, alongside untreated controls, and from shControl-HUVECs and shINHBA-HUVECs were subjected to a human angiogenesis cDNA-based qPCR array (TaqMan^® Array Human Angiogenesis; Applied Biosystems; Thermo Fisher Scientific, Inc.), according to the manufacturer's instructions.

HUVECs were cultured in the presence of 1 ng/ml activin A for 24 h at 37°C and the expression of total VEGFA and of isoforms 121, 165 and 189 were evaluated by RT-qPCR. To assess the effects of activin A on the production and secretion of total VEGFA and VEGFA isoform 121, ELISA was used. The concentration of VEGFA was determined using the human VEGFA ELISA kit (cat. no. RAB0507; Sigma-Aldrich; Merck KGaA) and the isoform 121 was evaluated with the human VEGFA isoform 121 ELISA kit (cat. no. SEB851Hu; Cloud-Clone Corp.), according to the manufacturers' instructions.

Western blot analysis was used to determine the abundance of phosphorylated SMAD2/3. Cells were washed with cold PBS and lysed in cell lysis buffer (Cell Signaling Technology, Inc.) containing protease and phosphatase inhibitors. Following centrifugation, protein concentration was measured using a Bio-Rad Protein assay (Bio-Rad Laboratories, Inc.) according to the manufacturer's instructions. Total protein (30 µg per sample) was resolved by 10% SDS-PAGE under reducing conditions, and transferred to nitrocellulose membranes. The membranes were blocked with 10% non-fat dry milk in PBS containing 0.1% Tween-20 at room temperature for 2 h, rinsed in the same buffer, and incubated at room temperature for 2 h with the following primary antibodies: Polyclonal rabbit anti-phosphorylated SMAD2/3 (1:1,000; cat. no. 8828; Cell Signaling Technology, Inc.), polyclonal rabbit anti-SMAD2/3 (1:1,000; cat. no. 3102; Cell Signaling Technology, Inc.) and monoclonal rabbit anti-GAPDH (1:1,000; cat. no. 2118; Cell Signaling Technology, Inc.). Following washing, the membranes were incubated at room temperature for 1 h with polyclonal goat anti-rabbit IgG HRP-linked antibody (1:2,000; cat. no. 7074; Cell Signaling Technology, Inc.). Bands were detected using enhanced chemiluminescence western blotting system (GE Healthcare) and signals captured with an Alliance 9.7 instrument (UVITEC, Ltd.).

Associations between immunohistochemical expression of activin A in blood vessels and in tumor cells and clinicopathological parameters of the tumors were analyzed using a χ² test. Survival curves were constructed based on the Kaplan-Meier method and compared with the log-rank test. For multivariate survival analysis, the Cox proportional hazard model with a stepwise method including all parameters was employed. Disease-specific survival (DSS) time was the time from treatment initiation until death due to cancer or the last known date alive, and disease-free survival (DFS) was the time from treatment initiation until diagnosis of the first recurrence (local, regional or distant) or the date of last follow up information for those without recurrence. Spearman's correlation test was applied for determine the correlation between activin A expression in blood vessels and in tumor cells.

All in vitro assays were performed at least three times in triplicate, and the results are presented as the mean ± standard deviation. Differences were compared using Mann-Whitney's U test or one-way analysis of variance with post-hoc comparisons based on Tukey's multiple comparison test. All statistical analyses were performed using GraphPad Prism v6.01 (GraphPad Software, Inc.). P<0.05 was considered to indicate a statistically significant difference.

To investigate whether activin A expression is associated with clinicopathological features of patients with OSCC, immunohistochemistry was performed in 95 cases of human OSCC. To quantify activin A abundance in blood vessels, immunohistochemistry for anti-CD34, a classical blood vessel marker, was performed concurrent with anti-activin A in serial sections (Fig. 1). Activin A was observed as a cytoplasmic stain with variable distribution and intensity in the endothelial cells of the blood vessels and in the tumor cells (Fig. 1). Expression was also observed in some inflammatory cells and cancer-associated fibroblasts. A significant correlation between activin A expression in endothelial cells and in tumor cells was detected (r=0.6; P<0.0001; data not shown). In order to categorize samples regarding age and expression level, the median values were applied to separate samples into two groups. As presented in Table I, no significant associations were observed between activin A expression in blood vessels and the clinicopathological features of patients with OSCC. However, a significant association was identified between activin A expression in tumor cells and margin status (P=0.02; Table I).

Univariate survival analysis based on log-rank test revealed a significant association of DSS with clinical stage of the tumor (P=0.001), histological grade (P=0.003), activin A expression in blood vessels (P=0.0002) and activin A expression in tumor cells (P=0.01; Table II). High expression of activin A in tumor cells was a significant marker of reduced DSS, with a 5-year survival of 30% for the patients with high activin A expression compared with 56.3% for those with low activin A expression (P=0.01; Fig. 2C). For activin A expression in the blood vessels, the reduction in DSS was even more significant, with a 5-year survival of 24.8% for patients with high activin A expression compared with 63.8% for those with low activin A expression (Fig. 2A). DFS revealed a significant association with activin A expression in blood vessels (P=0.03; Table II), but not in tumor cells (Fig. 2D; Table II). The expression of activin A in the blood vessels was associated with a 5-year DFS of 43.1% for patients with high expression compared with 65.7% for patients with low expression (Fig. 2B). Multivariate Cox regression analysis confirmed that clinical stage [hazard ratio (HR), 2.02; 95% confidence interval (CI), 1.09-3.73; P=0.03), activin A expression in blood vessels (HR, 2.47; 95% CI, 1.30-4.71; P=0.006) and activin A expression in tumor cells (HR, 1.63; 95% CI, 1.08-2.45; P=0.02) are independent risk factors for DSS of patients with OSCC (Table III). Furthermore, high expression of activin A in the blood vessels was determined to be an independent indicator of DFS (HR, 2.09; 95% CI, 1.07-4.08; P=0.03; Table III).

In order to strengthen the prognostic information provided by these independent factors, activin A expression levels in blood vessels and in tumor cells were combined and analyzed for both univariate and multivariate survival. Combination of clinical stage with activin A expression in blood vessels and with activin A expression in the tumor cells were also evaluated. In all combinations, the discriminatory ability to predict survival of patients with OSCC was largely improved (Table IV). Activin A expression in tumor cells was not individually associated with DFS in either univariate or multi-variate analysis, but when combined with activin A expression in the blood vessels and with clinical stage, a significant association was observed, revealing a prognostic discrimination. Collectively, these findings suggest that the expression levels of activin A in both blood vessels and tumor cells could be used as risk factor to predict poor prognosis of OSCC.

Since tube morphogenesis is a critical step in the formation of blood vessels, the present study first assessed the importance of activin A for the tubulogenesis of HUVECs. Cells were treated with recombinant activin A, the activin A-antagonist follistatin or were transduced with lentivirus carrying shINHBA. shINHBA-transfected HUVECs demonstrated a significant reduction in both activin A mRNA and protein levels in comparison with cells transfected with shControl (Fig. S1). A signifcant increase in the HUVEC tubulogenesis was observed in cells treated with 1 ng/ml activin A (P<0.01; Fig. 3A). Conversely, follistatin significantly reduced the tubulogenic activity of HUVECs (P<0.01; Fig. 3A). In addition, shINHBA-transfected HUVECs generated a significantly lower number of tubes compared with shControl-transfected HUVECs (P<0.05; Fig. 3B). Similar inhibition was observed when HUVECs were cultured in the presence of conditioned medium harvested from shINHBA OSCC cells compared with shControl OSCC cells (P<0.01: Fig. 3C).

To improve understanding of the role of activin A in the events that control tubulogenesis, the effects of activin A on the proliferation and migration of HUVECs were evaluated. Compared with untreated cells, activin A significantly increased the proliferation of HUVECs (P<0.05; Fig. 4A). Conversely, the proliferation of HUVECs was decreased, but not at significant level, after 24 h treatment with 100 ng/ml follistatin (Fig. 4A), and by gene silencing of activin A (P<0.01; Fig. 4B). Treatment of HUVECs with conditioned medium collected from shINHBA OSCC cells also significantly decreased HUVEC proliferation in comparison with conditioned medium from shControl OSCC cells (P<0.05; Fig. 4C).

The migration of HUVECs was not affected by activin A or follistatin directly in the upper chamber or when used as chemoattractant in the lower chamber of the Transwell system (Fig. 5A and B). However, activin A-knockdown significantly increased the migration of HUVEC cells (P<0.01; Fig. 5C). When conditioned medium collected from shINHBA OSCC cells was used as a chemoattractant, a significant increase in migration of HUVECs was observed (P<0.01; Fig. 5D).

To identify putative activin A-target genes, the expression of angiogenesis-related genes in HUVEC cells treated with activin A and in shINHBA-transfected HUVECs was profiled. Up- and downregulated genes were defined as those with an expression level >2.0- or <2.0-fold different in an average of 3 independent experiments. Table SII presents the up- and downregulated genes in HUVECs treated with activin A compared with untreated HUVEC cells and in shIN-HBA-transfected cells compared with shControl-transfected cells. VEGFA was markedly upregulated by activin A treatment (29.6-fold), whereas it was downregulated in HUVEC shINHBA-transfected cells compared with shControl-transfected HUVECs (-3.0-fold).

Considering that VEGFA was markedly regulated by activin A and given that VEGFA is the main regulator of angiogenesis (13,14), RT-qPCR was performed to determine the effects of activin A on the expression of total VEGFA and its major isoforms. The results demonstrated that compared with the untreated control, activin A significantly increased the expression of total VEGFA (P<0.01) and the pro-angiogenic isoform VEGFA121 (P<0.01; Fig. 6A). A significant decrease of VEGFA165 was detected in cells treated with activin A (P<0.001; Fig. 6A). Accordingly, significantly higher levels of total VEGFA (P<0.01) and VEGFA121 (P<0.05) were secreted by HUVECs following treatment with activin A in comparison with the control (Fig. 6B and C).

Activin A signaling is mediated by the SMAD pathway and VEGFA is positively regulated by SMAD2/3 (4,13,14). The present study therefore examined whether activin A stimulates SMAD2/3 phosphorylation in HUVECs. After 30 min of stimulation, increased phosphorylated SMAD2/3 levels were observed in HUVECs treated with 1 ng/ml activin A (Fig. 6D).