Human umbilical vein endothelial cells (HUVECs) were purchased from AllCells LLC (cat. no. H-001F-C) and cultured with HUVEC medium (cat. no. H-004; AllCells LLC) in a 0.25% gelatin-coated culture flask. HepG2, Bel7405 and MHCC-97H liver cancer cell lines were cultured in DMEM medium (cat. no. C11995; Gibco; Thermo Fisher Scientific, Inc.) and provided by Professor ZY Tang (Liver Cancer Institute, Fudan University, Shanghai, China), which were used in previous studies (22,26). These cell lines were characterized by DNA fingerprinting and isozyme detection. All cell lines used in the present study were regularly authenticated via morphologic observation and tested for the absence of mycoplasma contamination. Samples were last tested for mycoplasma in March 2017.

Cells were transfected with SVIL Stealth siRNA: E4 double stranded (ds)RNA, E5 dsRNA, E11 dsRNA and negative control dsRNA (all Invitrogen; Thermo Fisher Scientific, Inc.; all 40 nM) using Lipofectamine^® RNAi-MAX (Invitrogen; Thermo Fisher Scientific, Inc.). The dsRNA targeting sequences were as follows: E4 dsRNA, 5′-CUCACUUUGAAUGUAGAGAACCAUC-3′; E5 dsRNA, 5′-UUCUGCUGAAGUUAUAGGUUGGGUU-3′; E11 dsRNA, 5′-AGCAUAUUUAGAUUCCUUAUGGCUG-3′.

HepG2 cells were treated with or without 50 µg/l recombinant human VEGF (Novus Biologicals, LLC) at the indicated time-points.

The TCGA (https://tcga-data.nci.nih.gov/tcga/) cohort included normal tissues (n=50) and tumor tissues (n=347). The relationship between the expression of genes in liver cancer was analyzed using this database.

A liver cancer tissue microarray consisting of 173 pathological samples was purchased from US Biomax, Inc. Liver cancer tissue microarrays were immunohistochemically treated with antibodies against SVIL (dilution 1:1,000; cat. no. S8695; Sigma-Aldrich, Inc.) (24), cluster of differentiation (CD) 31 (dilution 1:1,500; product no. 3528; Cell Signaling Technology, Inc.), CD34 (dilution 1:50; product no. 3569; Cell Signaling Technology, Inc.), CD146 (dilution 1:500; cat. no. GTX60775; GeneTex, Inc.) and CD248 (dilution 1:500; cat. no. 564993; BD Biosciences). Tissue was sectioned and incubated in 65–75°C for 90 min. Samples were then placed in xylene in 25°C for 10 min and re-treated with xylene in 25°C for a further 10 min. The sections were sequentially placed in 100, 95 and 80% ethanol, after which purified water was used for 5 min. Samples were placed into a repair box with antigen repair solution (citrate buffer). A pressure cooker was heated under 1,600 W to automatically deflate for 2 min and samples were removed for 2 min for cooling. The antigen retrieval solution was discarded and sections were rinsed with PBS. The samples were permeabilized with 0.5% Triton X-100 (prepared in PBS) at room temperature for 20 min. Sections were transferred to a wet box and freshly prepared 3% hydrogen peroxide was added in 25°C for 10 min to remove endogenous peroxidase blocking solution. Samples were subsequently incubated for 10 min at room temperature and rinsed with PBS. During washing, slides were immersed in PBS 3 times for 3 min each and blotted dry. Normal goat serum (cat. no. AR0009; Boster Biological Technology co., Ltd) was added dropwise for blocking at room temperature for 30 min. Each slide was dropped with a sufficient quantity of diluted primary antibodies into a wet box and incubated overnight at 4°C. Rabbit/mouse secondary antibodies (MaxVision TM HRP-Polymer anti-Mouse IHC Kit; cat. no. 5001; MXB Biotechnologies) were subsequently added and incubated for 1 h at room temperature, after which slides were washed with PBS. DAB was developed for 3 min in 25°C and the degree of staining was assessed under a light microscope (magnification, ×100 and ×200). Samples were then rinsed again with PBS or tap water for 1 min. The samples were then counterstained with hematoxylin in 25°C for 3 min. Samples were subsequently rinsed with tap water for 1 min. Subsequently, dehydration, transparency, sealing, and microscopic examination were performed. The KF-PRO Digital Slide Scanning System (Kongfong Biotech International Co., Ltd.) was used to visualize the resulting signal.

After incubation for 30 min at 37°C, 2×10⁴ cells were added to a 96-well plate coated with 35 µl Matrigel at a concentration of 8.8 mg/ml. Following incubation for 6 h (liver cancer cells) or at the indicated time-points (0, 2, 4, 6 and 8 h) (HUVEC cells) at 37°C, three non-overlapping light microscopic images were randomly obtained at low-power magnifications (magnification, ×100). Total tube length and the number of branching points formed by endothelial or liver cancer cells per field were measured using Angio Tool 64 0.6a software (National Cancer Institute).

Cells were lysed in RIPA lysis buffer (cat. no. P0013K; Beyotime Institute of Biotechnology) to obtain a protein lysate. Then the protein quantification was determined by BCA. Each 8% SDS gel was infused with 40 µg of protein product. Total protein was separated by SDS-PAGE and transferred to nitrocellulose membranes. Samples were then blocked with 4% skim milk powder in Tris-buffered saline with Tween-20 (TBST) for 1 h at room temperature. Primary antibodies were diluted in 4% skim milk powder with TBST and incubated overnight at 4°C. Membranes were probed with targeted primary antibodies: SVIL (dilution 1:1,000; cat. no. S8695; Sigma-Aldrich, Inc.), β-tubulin (dilution 1:5,000; cat. no. EM0103; HuaBI), JNK (dilution 1:1,000; product no. 9252; Cell Signaling Technology, Inc.), p-JNK (dilution 1:1,000; product no. 9255; Cell Signaling Technology, Inc.), p38 (dilution 1:1,000; product no. 9212; Cell Signaling Technology, Inc.), p-p38 (dilution 1:1,000; product no. 9211; Cell Signaling Technology, Inc.), VE-cadherin (dilution 1:1,000; product no. 2500; Cell Signaling Technology, Inc.). After washing with TBST three times, the membranes were incubated with horseradish peroxidase-conjugated anti-rabbit IgG antibodies (cat. no. 7074s; Cell Signaling Technology, Inc.) for 1 h at 25°C. Proteins were visualized using an ECL luminescent solution (cat. no. 180-5001; Tanon Science and Technology Co., Ltd.).

For cell viability, cells were seeded in 96-well plates and treated with MTT (5 mg/ml) for 4 h at 37°C. DMSO (150 µl) was subsequently added and plates were measured at 450 nm (CMax Plus; Molecular Devices).

For cell proliferation, cells were cultured in a 12-well plate at 37°C. Subsequently, 50 µM EdU (cat. no. C10310-3; Guangzhou RiboBio Co., Ltd.) was added for 2 h, after which plates were analyzed according to the manufacturer's protocol. For the BrdU assay, 10–20,000 cells were added per well to a 12-well plate. After cells reached a density of 30–40%, transient transfection was performed and the culture medium was changed after 4–6 h. Following 48 h of culture, 10 µm BrdU was added to each well plate and incubated for 4 h. Subsequently, the samples were fixed in 4% paraformaldehyde precooled on ice for 20 min, after which slides were washed three times with PBS. Samples were then further incubated with PBS containing 1.5 mol HCl for 10 min at room temperature and 0.2% Triton X-100 for 10 min. PBS was used to wash samples in triplicate. BrdU antibodies were diluted to 1:1,000 and 300 µl was added to each well. After incubation at 4°C overnight, samples were washed three times with PBS and cell nuclei were stained with 0.5 µg/ml DAPI at 25°C for 2 min. Distilled water was used to wash off the PBS, glycerin was used to seal the samples and nail polish was sealed around the film.

For cell migration, 2×10⁵/100 µl cells were placed in a Transwell chamber and cultured for 16 h at 37°C. Samples were fixed with 4% paraformaldehyde at 4°C for 15 min, stained with 0.1% crystal violet at 25°C for 10 min and counted using a light microscope (magnification, ×100).

To determine the degree of cell spreading, samples were seeded in 12-well plates coated with 10 µg/ml fibronectin and observed at the indicated time-points. Briefly, 800 µl of diluted Fibronectin (10 µg/ml) was added to each well of the 12-well plate, and coated at room temperature for 4 h. The transfected cells were digested and resuspended, the cell density was adjusted to 4–5×10⁵ cells after counting the cells, and adding them to the coated 12-well plate. After cells adhered to the wall, cells were observed and images were captured every 30 min.

Secreted VEGF was quantified using a VEGF ELISA kit (cat. no. DVE00; R&D Systems, Inc.). The absorbance of each well was measured at a range of wavelengths based on the manufacturer's protocol. A Microplate ELISA Analyzer (CMax Plus; Molecular Devices, LLC) was used to obtain absorbance data.

HepG2 cells were transiently transfected with siRNA, after which total RNA was extracted using TRIzol^® (cat. no. 15596026; Invitrogen; Thermo Fisher Scientific, Inc.). cDNA was subsequently obtained using an RT kit (cat. no. AH311-02; Beijing Transgen Biotech Co., Ltd.) and mRNA levels of associated factors were determined via RT-PCR with cDNA as a template. The reaction was performed in 2X EasyTaq^® PCR SuperMix (cat. no. AS111-11; TransGen Biotech) and the thermocycling conditions were followed according to the instructions. The primer sequences were as follows: 5′ to 3′: Cyclooxygenase 2 (COX2) forward, ATGATTGCCCGACTCCCTTG and reverse, CCCCACAGCAAACCGTAGAT; EphA2 forward, AGCATCAACCAGACAGAGCC and reverse, AGCATGCCCTTGTACACCTC; N-cadherin forward, CCTGGATCGCGAGCAGATAG and reverse, CCGTGGCTGTGTTTGAAAGG; VE-cadherin forward TCAAGCCCATGAAGCCTCTG and reverse, CCGGTCAAACTGCCCATACT; MMP-2 forward, AACACCTTCTATGGCTGCCC and reverse, GCCGTACTTGCCATCCTTCT; MMP-9 forward, TCCTTATCGCCGACAAGTGG and reverse, AGCGGTCCTGGCAGAAATAG; MMP-12 forward, AACCAACGCTTGCCAAATCC and reverse, GGCCCGATTCCTTGGAAGTT; MMP-14 forward, ACATCTTCCTGGTGGCTGTG and reverse, GTACTCGCTATCCACTGCCC; MMP-25 forward, AAGCGAACCCTGACATGGAG and reverse, CGCCTTCCCATAGAGTTGCT; Mig-7 forward, AGAGGAAAACTGAGGCTGCC and reverse CCGAGTGACAATCTGGGCTT.

Healthy nude mice (20 male; age, 3–4 weeks old; weight, 12–14 g) were purchased at the Institute of Model Animal Research of Nanjing University and cultured in an SPF environment. Cells (1×10⁷/200 µl) were injected into the right side of the back of nude mice. After one week, tumor size was measured (~10 mm³), and in vivo SVIL siRNA (50 µl; 10 nmol) was injected directly into tumor tissue. Injections were administered three times in the first week and then twice a week, for 4 consecutive weeks. The weight of mice and tumor growth was subsequently measured. RNA interference (RNAi) modified with 2′-OMe (Guangzhou RiboBio Co., Ltd.) with the following targeting sequences were utilized: Negative control, 5′-AAUUCUCCGAACGUGUCACGU-3′; E4 RNAi, 5′-CUCACUUUGAAUGUAGAGAACCAUC-3′; E5 RNAi, 5′-UUCUGCUGAAGUUAUAGGUUGGGUU-3′; E11 RNAi, 5′-AGCAUAUUUAGAUUCCUUAUGGCUG-3′. Animal experiments were performed according to the guidelines of the Animal Use and Care Committees at Hefei Institutes of Physical Science, CAS.

Data are presented as the mean ± standard deviation. Continuous variables were analyzed using an unpaired Student's t-test for comparisons between two groups. P<0.05 was considered to indicate a statistically significant difference.

To determine the expression of SVIL in liver cancer tissue, SVIL levels were analyzed in 50 normal liver tissues and 347 liver cancer tissues obtained from The Cancer Genome Atlas (TCGA) database. The results revealed that SVIL was significantly increased in liver cancer compared with normal liver tissue (Fig. 1A). Immunohistochemical staining of liver cancer further indicated that SVIL levels were significantly increased in liver cancer, and positively associated with liver cancer stage (Fig. 1B).

In addition to hepatoma cells, SVIL was expressed in tumor vessels. Tumor vessels of liver cancer primarily include EDV formed via endothelial cells and VM formed via tumor cells (25,27). The expression of SVIL, CD31, CD34, CD146 and CD248 were assessed in serial liver cancer tissue sections via immunohistochemical staining. The results revealed that in the liver cancer samples, certain SVIL-labeled cells were co-located with CD31 and CD34 endothelial cells, exhibiting close proximity to CD146- or CD248-positive pericytes (smooth muscle cells for microvessels; Fig. 1C). Additionally, the results demonstrated that certain SVIL-labeled cells were present on neovascular-like structures formed by non-endothelial cells, presenting as CD34⁻/PAS⁺ and therefore indicating that SVIL may be expressed in vascular mimetic structures (Fig. 1D). The VM structure accounted for ~40% after determining the number of EDVs labeled with CD34 and the number of tumor blood vessels marked via SVIL (Fig. 1D).

Collectively, the results indicated that SVIL served a role in liver cancer angiogenesis, particularly in EDV and VM development.

The expression of SVIL was upregulated during angiogenesis in HUVEC cells (Fig. 2A). To determine the potential role of SVIL in HUVEC angiogenesis, Stealth RNAi™ dsRNAs were used to target sequences within the SVIL coding exon 4 (E4 dsRNA), coding exon 5 (E5 dsRNA), and coding exon 11 (E11 dsRNA). As described previously, each Stealth siRNA that targeted the different axons of SVIL in HUVEC cells reduced the level of each isoform by ≥75%. Furthermore, transfection of SVIL-specific RNAi resulted in a 25% (E4 dsRNA), 40% (E5 dsRNA) and 55% (E11 dsRNA) reduction in tube formation during EDV angiogenesis (Figs. 2B and S1A).

The roles of SVIL in HUVEC migration, spread and proliferation were assessed in the present study as these biological functions are important to HUVEC angiogenesis (28). The results revealed that SVIL knockdown inhibited HUVEC migration (Figs. 2C and S1C), spread (Figs. 2D and S1B), viability (Fig. 2E) and proliferation (Fig. 2F), to different degrees. Thus, the results indicated that SVIL served an important role in HUVEC angiogenesis, particularly regarding cell migration, spread and proliferation.

VM formation may be a primary factor for the failure of traditional anti-vascular treatment (6). The present study therefore analyzed the role of SVIL in the progression of VM development. The results revealed that SVIL expression levels were positively associated with VM formation in liver cancer. This may have been due to HepG2 cells expressing greater quantities of SVIL, gaining a stronger ability to develop VM when compared with other liver cancer cells (Fig. 3A). SVIL knockdown with specific dsRNA resulted in a reduction in VM formation, to varying degrees (Figs. 3B and S2A). Similar to the results of HUVEC angiogenesis, SVIL knockdown in HepG2 cells demonstrated decreased cell migration (Figs. 3C and S2B), spread (Figs. 3D and S2C), viability (Fig. 3E), and proliferation (Figs. 3F and S2D). The results indicated that SVIL served an important role in VM formation, potentially by regulating the survival, migration and proliferation of hepatoma cells.

Increasing evidence has indicated that VEGF is crucial for the development of VM and can be secreted from tumor cells (17,29). In the present study, it was demonstrated that SVIL knockdown significantly suppressed VEGF secretion in the culture media of HepG2 cells, indicating that there was an interaction between VM and tumor cells which may lead to tumor progression (Fig. 4A).

The MAPK signaling pathway has been demonstrated to mediate cell survival and migration, and serve an important role in VM formation (11,30,31). It was demonstrated in a previous study that SVIL was associated with MAPK activation (22,32,33). As hypothesized, SVIL knockdown increased p38 activation and the phosphorylation level of JNK compared with the control group (Fig. 4B). Furthermore, cellular apoptosis was increased (Figs. 4C and S3A) and cell population was decreased (Fig. S3B). The results indicated that SVIL may regulate tumor cell survival and p38 activation to ensure VM development.

As VEGF secretion and p38 signaling are indispensable for VM formation (6,11,34), the present study investigated potential crosstalk between the VEGF and p38 signaling pathways. The results revealed that VEGF abolished the SVIL knockdown-induced reduction of tube formation in VM (Figs. 4D and S4) and downregulated p38 phosphorylation levels. However, the phosphorylation level of JNK did not change, indicating that p38 phosphorylation may be downstream of SVIL-mediated VEGF secretion (Fig. 4B).

As previously reported, VE-cadherin, N-cadherin, COX-2, EphA2, MMPs and Mig-7 are involved in VM formation (6,35,36). In the present study, SVIL downregulation induced alterations of the VM formation transcriptional network, particularly via VE-cadherin, MMP9, MMP12 and Mig-7. With the addition of VEGF, the expression levels of these transcriptional factors were rescued by SVIL knockdown (Fig. 4E and F). Collectively, the results suggested that SVIL promoted VM development by activating the VEGF-p38 axis and inducing VM-associated transcriptional factors.

As tumor angiogenesis promotes the proliferation, growth, invasion and metastasis of hepatoma cells (1,2,37), the present study investigated whether SVIL knockdown inhibited tumor growth in vivo. Hepatoma cells were injected into the right side of the back of each nude mouse. After a week, in vivo SVIL siRNA was injected directly into the tumor tissue. The results revealed that injection of in vivo SVIL siRNA significantly inhibited tumor growth, particularly E5 RNAi and E11 RNAi (Fig. 5A and B). However, mouse weight did not differ compared with the negative control RNAi-treated mice (Fig. 5C).