The human ovarian cancer cell line C13K was a gift from the Department of Obstetrics and Gynecology, Ottawa University, Department of Cell and Molecular Research Center. SKOV3 and A2780 and breast cancer line MDA-MB-231 were purchased from ATCC and cultured according to their guidelines. All the above cell lines used in this study were cultured in RPMI-1640 medium supplemented with 5% fetal bovine serum. The cells were used for the experiments within 20 passages. Human umbilical vein endothelial cells (HUVECs) were purchased from Collection of Biological Center, Wuhan University and cultured in endothelial cell medium (ECM; ScienCell) with 5% FBS and endothelial growth medium supplements.

The Smad4-siRNA sequences used were as follows: GUACUUCAUACCAUGCCGATT and UCGGCAUGGUAUGAAGUACTT. Alternatively, Lipofectamine (Invitrogen)-mediated Smad4-siRNA Oligo transfection was used to knock down Smad4 in HUVECs. Cells were harvested 2 days later for expression analysis or in vitro tube formation.

Female athymic nude (nu/nu) mice (4-week-old) were purchased from SLAC Laboratory Animal Co. Ltd. (Shanghai, China) for studies approved by the Committee on the Ethics of Animal Experiments of Tongji Medical College. The mice were maintained in the accredited animal facility of Tongji Medical College. C13K tumor cells (3×10⁶) and 1×10⁶ HUVEC were washed, suspended in 50 µl of PBS, and co-injected subcutaneously into nude mice. Tumor volumes were measured using a slide caliper every 3 days according to the formula: volume = (larger diameter) × (smaller diameter)² / 2 (10).

Specimens from normal ovary (14 cases), ovarian carcinoma (19 cases) and normal endometrium (7 cases) were acquired by surgeries as approved by the Ethics Committee of the Medical Faculty of Tongji Medical College (Wuhan, China). The tumor specimens were acquired from patients with cancer who had not undergone preoperative radiotherapy or chemotherapy. Tissue sections were subjected to immunohistochemical (IHC) analysis using the avidin-biotin complex (ABC) Vectastain kit (Zsgb-Bio, Beijing, China) according to the manufacturer's protocol. Anti-human CD34 (Abcam, ab81289), and anti-human SMAD4 (R&D Systems, AF2097) antibodies were used as primary antibodies. Briefly, slides were scanned at low power and the areas with the highest density of CD34-positive vessels were identified. The pathological analyses were done double-blinded.

Cells were pre-treated with siRNA as needed. Total proteins were harvested with RIPA buffer. Immunoblotting was performed according to manufacturer's instructions. The relative expression level was quantified using Image-Pro Plus.

HUVECs and primary pericytes were mixed and replated to 48-well plates precoated with a thin layer of Matrigel (BD Biosciences) in culture medium containing 5% FCS, and allowed to form tube-like structures for 4–6 h. Measurement was performed as described (11).

In vitro permeability assay were performed as described (12). The concentration of FITC-conjugated dextran (MW40,000, Sigma) was determined with an EnVision fluorescence multiwell plate reader (BD) using a fluorescein filter pair [Ex (l) 480 nm; Em (l)535 nm]. The percentages of control were quantified.

RNA was extracted from HUVECs by TRIzol reagent (Invitrogen) and reverse transcribed by using an mRNA selective PCR kit (Takara). Real-time PCR was performed with Roche LightCycler 2.0 system using a SYBR Green assay. Primers used were as follows: smad4 sense, 5′-CAGGATCAGTAGGTGGAAT AGC-3′; antisense, 5′-TCTTTGATGCTCTGTCTTGGG-3′. INSR sense, 5′-GGAAGTTACGTCTGATTCGAGG-3′; antisense, 5′-TGAGTGATGGTGAGGTTGTG-3′. FYN sense, 5′-ACTATGAAGCACGGACAGAAG-3′; antisense, 5′-TGCTGGGAATGTAACCTGTC-3′. PTPRM sense, 5′-CGATGAGGTGAAGGTGTTAGG-3′; antisense, 5′-ACTGGAAGGT AGCAAACTGG-3′. PTPRJ sense, 5′-CTAGTCCAATTCCTGACCCTTC-3′; antisense, 5′-AGCTTTCACCATCCTCACTG-3′. VCAN sense, 5′-CACTCTAATCCCTGTCGTAATGG-3′; antisense, 5′-ATGTCTCGGTATCTTGCTCAC-3′. C CND2 sense, 5′-TGAGGAACAGAAGTGCGAAG-3′; antisense, 5′-TGGTCTCTTTGAGTTTGGAGG-3′. Control-siRNA and Smad4-siRNA HUVECs were subjected to microarray analysis, which was performed using human Gene 1.0 ST array (Affymetrix).

Data were evaluated using a Student's t-test, two-tailed. p<0.05 and p<0.01 was considered statistically significant. The error bars on graphs represent the mean ± standard deviation (SD).

To identify the molecular differences between tumor-associated BECs and their normal BEC counterparts, blood vessels were isolated using in situ laser capture microdissection and verified by the detection of the mRNA of specific markers. Then, the gene expression profiles of tumor BECs and normal BECs were analyzed using a cDNA microarray as described (13). Since we focused on smad in this study, SMAD4, among the top 10 genes, was chosen for further investigation.

Immunohistochemical analysis of serial sections showed that SMAD4 colocalized with the blood vessel marker CD34. (Fig. 1A). Compared with blood vessels in normal tissues (normal ovary and normal endometrium), the expression of SMAD4 protein in tumor-associated blood vessels was significantly decreased (Fig. 1A). In addition, we showed SMAD4 expression intensity in Fig. 1B). These results suggest that tumor blood vessels might functionally differ from normal blood vessels due to loss of SMAD4 expression.

Research on SMAD4 in ovarian cancer is scarce. Western blotting and immunofluorescence assays were used to detected SMAD4 expression. As shown in Fig. 2A, SMAD4 is expressed universally in 3 ovarian tumor cell lines (SKOV3, A2780 and C13K), human umbilical vein endothelial cells (HUVECs) and one breast cancer cell line (MDA-MB-231) (Fig. 2A). Immunofluorescence analysis of SMAD4 protein are distributed in the nuclear and cytoplasm in HUVEC. In order to simulate the low expression of SMAD4, siRNA was used to interfere the SMAD4 expression in HUVEC, at 48 h, compared with control, the SMAD4 expression was decreased 90% (Fig. 2B).

We next investigated the role of SMAD4 in regulation of HUVEC tube formation and migration. TGF-β (5 ng/ml or 400 pg/ml) treated HUVECs with siRNA targeting SMAD4 exhibit a decrease in tube formation and migration (Fig. 3A and B). This result raised the possibility that SMAD4 plays a very important role in angiogenesis.

As SMAD4 deletion decrease the angiogenesis in vitro, we next determined whether the effect also occur in vivo. We transduced HUVECs with SMAD4-siRNA or control-siRNA and co-inject with C13K subcutaneous in nude mice. Tumor volume was periodically measured, and tumor weight was determined upon dissection at the end of the experiment. Tumor xenografts with SMAD4-deficient HUVECs did not display a significant reduction in volume and tumor burden (Fig. 4A), but they were unexpected associated with a remarkable increase of spontaneous metastatic dissemination to the liver (Fig. 4B). The metastatic potential was quantified by scoring micrometastasis in the liver. Moreover, we found that blood vessels in the tumor with SMAD4-deficient HUVECs were obviously reduced (Fig. 4C). We further investigated whether there was any defect in mural cell coverage. As shown in Fig. 4D, the control tumor vasculature was completely enveloped by mural cells, which were identified by NG2 immunostaining. In comparison, a local smooth muscle cell-coating deficiency was observed in the vessel of Smad4-deficient tumors. These results clearly showed that loss of Smad4 in tumor BECs resulted in defective EC-mural cell contact, which in turn might contribute to decreased mechanical stability, then allow the tumor cells to cross the blood vessel barrier easier and consequently tumor metastasis.

To explore EC-pericyte interaction in vitro, we used an in vitro BBB model (13) by coculturing HUVECs with the primary pericytes. When monolayer HUVECs were seeded on the Transwell membrane in tracer permeability assays, there was no apparent difference between siSMAD4 and siNC HUVECs (Fig. 5A and B). Coculture with pericytes reduced the permeability in the control HUVECs to a greater degree than in Smad4-deficient HUVECs, thus, the Smad4-deficient EC barrier supported by pericytes was notably weaker than that of the control.

To further validate the defect in EC-pericyte interaction, we performed an in vitro three-dimensional coculture system, in which both ECs and pericytes are morphologically stretched and coordinately formed capillary-like structures (14). With the same primary pericytes in the coculture, Smad4-siRNA HUVECs demonstrated significantly impaired tube-forming capacity compared to the siNC HUVECs. Morphologically, Smad4-siRNA HUVECs showed inefficient elongation and connection with pericytes (Fig. 5C). These results strengthen the opinion that loss of Smad4 in ECs impairs EC-pericyte interaction.

To identify potential Smad4 target genes that regulate blood vascular integrity in ovarian tumor ECs, a microarray assay was performed to compare the gene expression profiles of HUVECs cells transfected with Smad4-siRNA and control-siRNA. Using GO function enrichment and pathway enrichment analysis, we further verified some expression changes of the genes involving PTCH2, ATL3, PTPR and FYN (Fig. 6A). Of particular interest was the remarkable increase in FYN in Smad4-siRNA HUVECs. We further examined the FYN expression by real-time PCR. As shown in Fig. 6B, the decrease of SMAD4 in the HUVECs induced a remarkably increased FYN expression. These results demonstrated that impaired EC-pericyte interaction in Smad4-deficient ECs might be largely due to the increased FYN expression.