The materials and reagents used in this study were purchased from the following providers. The Periodic Acid-Schiff (PAS) kit was purchased from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany); Matrigel was obtained from BD Biosciences (Franklin Lakes, NJ, USA). Primary anti-VEGFA (sc-152), anti-VEGFR1 (sc-316), anti-VEGFR2 (sc-505) and anti-β-actin (sc-8432) polyclonal antibodies; VEGFA small interfering RNA (siRNA); VEGFR1 siRNA; and VEGFR2 siRNA were purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA). Anti-LMP1 antibody was purchased from Dako/Agilent (M0897; Agilent Technologies, Inc., Santa Clara, CA, USA); anti-CD34 antibody was purchased from Abcam Corp. (cat. no. ab81289; Cambridge, UK); and the HistoMouse-SP Broad Spectrum DAB kit was purchased from Invitrogen-Zymed (Thermo Fisher Scientific, Inc., Waltham, MA, USA).

HK1 is an LMP1-negative NPC cell line, and HK1-LMP1 is a stable LMP1-integrated cell line (20). Cells were grown in Gibco™ RPMI-1640 medium (Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS). Cells at 60–70% confluence were transfected with siRNA using Invitrogen™ Lipofectamine™ 2000 (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions.

A Matrigel tube formation assay was developed for testing tubular structure formation. Briefly, we transfected HK1-LMP1 cells with targeted siRNA (LMP1, VEGFA, VEGFR1, or VEGFR2) and control siRNA (CON). Cells were trypsinized after 24 h and centrifuged at 600 × g for 5 min. Approximately 1×10⁵ cells were seeded in each well of a 24-well plate with 200 µl of embedded Matrigel. Next, the cells were incubated for 8 h, and the extent of tubular structure formation was examined using an inverted microscope (CKX41SF; Olympus Corp., Tokyo, Japan).

Cells were harvested and lysed at 4°C for 15 min in lysis buffer, and the protein concentration was determined by the Bradford protein assay (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer's instructions. Proteins were then separated by SDS-PAGE (4–20% Mini-PROTEAN^® TGX™ Precast Protein Gels (Bio-Rad Laboratories) and transferred to a nitrocellulose membrane (GE Healthcare, Piscataway, NJ, USA). Membranes were blocked with TBS-T (20 mM Tris-HCl, pH 7.4, 137 mM NaCl and 0.1% Tween-20) containing 5% non-fat milk for 1 h at room temperature (RT). Then, the membranes were incubated with primary antibodies (LMP1, dilution 1:250; VEGFA, dilution 1:250; VEGFR1, dilution 1:500; VEGFR2, dilution 1:500), followed by incubation with horseradish peroxidase-conjugated secondary antibody (cat. nos. sc-2005 or sc-2004; Santa Cruz Biotechnology) for 1 h at RT and visualization with an enhanced chemiluminescence detection kit (Pierce ECL; Thermo Fisher Scientific, Inc., Pittsburgh, PA, USA). The relative protein levels were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA).

The cells were fixed and permeabilized with cold methyl alcohol (−20°C) for 10 min, and then blocked in 5% donkey serum in PBS for 1 h and incubated with the primary antibody in PBS containing 1% BSA at 4°C overnight. The cells were washed 3 times with PBS, and incubated for 30 min with fluorochrome-conjugated secondary antibody (cat. nos. A-11001 or A-21207; Thermo Fisher Scientific, Inc., Waltham, MA, USA) for 30 min. For fluorescence analysis, cell samples were visualized on a laser scanning confocal microscopy with appropriate emission filters (LSM 510 META; Carl Zeiss, Oberkochen, Germany).

The NPC tissue array was purchased from Pantomics (Richmond, CA, USA), and the NPC paraffin-embedded tumor tissue samples, clinical details, and follow-up data were obtained from the Pathology Department of Xiangya Hospital from 2009 to 2015. The institutional review board of the Xiangya Hospital Ethics Committee approved the use of human samples in this study. IHC staining was performed using a HistoMouse-SP Broad Spectrum DAB kit (Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to standard protocols. The signal was detected using a diaminobenzidine solution. A semi-quantitative evaluation of the positivity of each protein by IHC was performed using a previously defined method (21). The percentage of positive cells was divided into five grades (percentage scores): 0, ≤10%; 1, 11–25%; 2, 26–50%; 3, 51–75%; and 4, >75%. The intensity of staining was divided into four grades (intensity scores): 0, no staining; 1, light brown; 2, brown; and 3, dark brown. Staining positivity was determined by the following formula: Overall score = percentage score × intensity score. The stained sections were independently examined by two of the authors (Z.Z. and B.L.). Scores of 3 to 12 were considered positive for LMP1 and VEGFR1 expression. For CD34/PAS double staining, after IHC staining for CD34 as described above, the sections were washed with running water for 6 min, incubated with PAS for 15 min, and counterstained with hematoxylin. To quantify the differences in the density of VM, we calculated the total number of VMs in five fields for each tumor dot or section.

Statistical analyses were performed using the Student's t-test. The Kaplan-Meier method was used to estimate progression-free survival, and the log-rank test was used to evaluate differences between survival curves. A P-value of <0.05 was considered statistically significant.

Highly aggressive tumor cells form patterned networks of matrix-rich tubular structures when cultured on Matrigel (4). In this experiment, we used an in vitro tube formation assay to evaluate the tubular structure formation ability of different NPC tumor cells. The results showed that it was almost impossible for the LMP1-negative NPC cells, HK1, to form tubular structures in Matrigel, whereas the stable LMP1-integrated cells, HK1-LMP1, were able to form tubular structures. The tubular structure forming ability of the HK1-LMP1 cells decreased by approximately one-half after treatment with LMP1 siRNA compared with the control siRNA (CON) group (P<0.05) (Fig. 1A). Thus, the data indicated that LMP1 might contribute to tumor VM formation in NPC.

Moreover, the western blotting results showed that VEGFA was highly expressed in the HK1-LMP1 cells compared to the level in the HK1 cells (Fig. 1B). Furthermore, when HK1-LMP1 cells were treated with LMP1 siRNA, LMP1 expression was reduced compared to that noted in the CON group, and this reduction was accompanied by decreased VEGFA expression. Meanwhile, the data of the immunofluorescence assay showed that, there was a higher VEGFA protein expression in HK1-LMP1 cells compared to that in the HK1 cells, and the co-localization of LMP1 and VEGFA was in the cytoplasm of NPC cells. Following inhibition of LMP1 expression by siRNA, both LMP1 and VEGFA expression was reduced compared to the CON group, and the co-localization of LMP1 and VEGF become very weak (Fig. 1C). These results indicated that EBV-LMP1 is involved in the tubular structure formation in NPC cells and is related to VEGFA expression.

Furthermore, we investigated the contribution of VEGFA to tubular structure formation in LMP1-positive NPC cells. The data showed that VEGFA siRNA also strongly inhibited the formation of tubular structures compared with the CON group in the in vitro tube formation assays (P<0.05) (Fig. 2A). Meanwhile, VEGFA was highly expressed in HK1-LMP1 cells compared to that noted in the HK1 cells, and VEGFA siRNA significantly reduced VEGFA expression compared with that noted in the CON group (Fig. 2B). These findings indicated that VEGFA is involved in VM formation in LMP1-positive NPC cells.

VEGFA performs its biological function mainly through binding and activating its receptor VEGFRs, VEGFR1 and VEGFR2, which have been shown to be involved in VM formation in different tumor cells (6,12). Thus, we determined which VEGFR is required for the tubular structure formation in human LMP1-positive NPC cells. The data showed that VEGFR1 siRNA also strongly inhibited the formation of tubular structures compared with the CON group in in vitro tube formation assays (P<0.05) (Fig. 3A). Furthermore, VEGFR1 was highly expressed in HK1-LMP1 cells compared to that observed in the HK1 cells. Moreover, VEGFR1 siRNA significantly reduced VEGFR1 expression compared with the CON group (Fig. 3B). These findings indicated that VEGRA/VEGFR1 signaling was involved in VM formation in LMP1-positive NPC cells. In contrast, HK1-LMP1 cells were transfected with VEGFR2 siRNA, which decreased the expression of VEGFR2. However, we did not observe any changes in the main geometrical features of the tubular structures that formed compared with the CON group (Fig. 4). These findings indicated that VEGFA/VEGFR2 signaling was not involved in the tubular structure formation in LMP1-positive NPC cells. According to the above western blotting results, VEGFA and VEGFR1 were significantly increased in LMP1-positive cells compared with LMP1-negative cells, whereas VEGFR2 did not change substantially. Therefore, we hypothesized that LMP1-VEGFA functions in the formation of VM through VEGFR1 rather than VEGFR2, which may be related to the proteins that are regulated by LMP1.

The VM channels consist of a basement membrane with a lining of tumor cells on the external wall and do not contain endothelial cells (ECs) on the inner wall. Thus, PAS-positive and CD34-negative mature tumor vessels form a patterned network in the VM morphology (2). In further support of the preferential association of EBV-LMP1 and its downregulation of VEGFA/VEGFR1 signaling with VM in NPC, we examined the expression levels of LMP1 and VEGFR1 in a commercial NPC tissue array, and VM was detected in the tumor tissue array using CD34/PAS double staining. As shown in Fig. 5A, the VM channels were composed of NPC cells and were PAS-positive, CD34-negative (indicated by black arrows), while the endothelial-dependent vessels were positive for both CD34 and PAS (indicated by red arrows). These results showed that VM occurred in EBV-LMP1-positive NPCs and that hardly any VM occurred in LMP1-negative NPCs. There was a positive correlation between LMP1 expression and VM formation according to the Pearson correlation coefficient results (r=0.397, P=0.002). Meanwhile, the data showed a significant correlation between LMP1 and VEGFR1 expression (correlation coefficient=0.481, P=0.000) (Fig. 5B). These results indicated that LMP1 expression was associated with VEGFR1 and VM formation in NPCs.

The presence of VM in malignant tumors is associated with increased patient mortality (3). Thus, we determined whether the LMP1-mediated upregulation of VM is associated with the prognosis of NPC patients. We examined the expression levels of LMP1 and VM formation in tumor tissue samples from 40 NPC patients. These patients were successfully followed up (a median follow-up period of 5.61 years) (Table I). The positive staining rate of LMP1 in NPC tissues was 70% (28/40). The positive staining rate of VM in NPC tissues samples was 57.5% (23/40). There was a significant correlation between LMP1 expression and VM formation according to the Pearson correlation coefficient (r=0.372, P=0.018). These results indicated that LMP1 expression is associated with VM formation in NPC tumor tissues, consistent with the results from the commercial NPC tissue array (Fig. 6A and B).

Using the clinical follow-up data, we retrospectively analyzed the prognostic significance of VM formation on the progression-free survival time (PFS) of 40 NPC patients, who received radiation therapy or concomitant chemoradiotherapy. As shown in Fig. 6C, the results of the Kaplan-Meier method analysis with log-rank test revealed a statistically significant difference in PFS between the 23 patients in the VM-positive group (78.3% with expression of LMP1, median survival time of 37.12 months) and the 17 patients in the VM-negative group (58.9% with expression of LMP1, median survival time of 54.81 months). These results not only indicated that VM was upregulated along with LMP1, but also suggested that VM formation was associated with a worse clinical outcome following therapy.