Hypoxia promotes vasculogenic mimicry formation by vascular endothelial growth factor A mediating epithelial‐mesenchymal transition in salivary adenoid cystic carcinoma

Abstract Objectives To investigate the role of hypoxia in vasculogenic mimicry (VM) of salivary adenoid cystic carcinoma (SACC) and the underlying mechanism involved. Materials and methods Firstly, wound healing, transwell invasion, immunofluorescence and tube formation assays were performed to measure the effect of hypoxia on migration, invasion, EMT and VM of SACC cells, respectively. Then, immunofluorescence and RT‐PCR were used to detect the effect of hypoxia on VE‐cadherin and VEGFA expression. And pro‐vasculogenic mimicry effect of VEGFA was investigated by confocal laser scanning microscopy and Western blot. Moreover, the levels of E‐cadherin, N‐cadherin, Vimentin, CD44 and ALDH1 were determined by Western blot and immunofluorescence in SACC cells treated by exogenous VEGFA or bevacizumab. Finally, CD31/ PAS staining was performed to observe VM and immunohistochemistry was used to determine the levels of VEGFA and HIF‐1α in 95 SACC patients. The relationships between VM and clinicopathological variables, VEGFA or HIF‐1α level were analysed. Results Hypoxia promoted cell migration, invasion, EMT and VM formation, and enhanced VE‐cadherin and VEGFA expression in SACC cells. Further, exogenous VEGFA markedly increased the levels of N‐cadherin, Vimentin, CD44 and ALDH1, and inhibited the expression of E‐cadherin, while the VEGFA inhibitor reversed these changes. In addition, VM channels existed in 25 of 95 SACC samples, and there was a strong positive correlation between VM and clinic stage, distant metastases, VEGFA and HIF‐1α expression. Conclusions VEGFA played an important role in hypoxia‐induced VM through regulating EMT and stemness, which may eventually fuel the migration and invasion of SACC.


| INTRODUC TI ON
Salivary adenoid cystic carcinoma (SACC) is one of the most common malignant cancers of the salivary glands accounting for approximately 30% of all salivary malignant tumours. [1][2][3] As an epithelial malignant tumour, SACC is characterized by early haematogenous dissemination, high incidence of lung metastasis and perineural spread. 4,5 The growth of SACC is quite slow and the 5-year survival rates are 70%-90%. However, the 5-year survival rates of patients with distant metastasis are only 20%. 6 Hence, it is necessary to address the underlying molecular mechanisms regulating metastatic dissemination of SACC.
Vasculogenic mimicry (VM), proposed by Maniotis in 1999, was defined as a type of vessel-like structures lined with tumour cells without endothelial cells. 7 VM had been found to evaluate in various aggressive cancers, including colorectal cancer, 8 breast cancer, 9 melanoma, 10 and head and neck squamous cell carcinoma, 11 suggesting that it was a novel hallmark of cancer. As newly defined mechanism to supply oxygen and nutrition to tumour cells, the high expression of VM was also viewed as a risk factor for poor prognosis, low survival, and invasion and metastasis in cancer patients. 12 Recently, increasing evidence has showed that hypoxic microenvironment not only accelerated tumour invasion and metastasis, but also led to VM formation 13,14 and the expression of hypoxia-inducible factor-1α (HIF-1α) was associated with VM in many cancers types, including breast cancer, 15 ovarian cancer 16 and colorectal cancer. 17 Moreover, Ahluwalia et al reported that HIF-1 by hypoxia condition adjusted vascular endothelial growth factor A (VEGFA) expression at the transcriptional level. VEGFA was a downstream target of HIF-1, and angiogenesis produced by hypoxia was usually VEGFA-dependent. 18 And hypoxia induced EMT by regulating VEGFA. However, it is still not clear whether VEGFA is involved in the hypoxia microenvironment mediating VM formation. 19,20 In the previous study, our group has observed that CD133 + stem-like SACC cells contributed to the migration and invasion of SACC through inducing VM formation. 21 Here, we used a three-di-

| Ethics statement
All studies carried out on human specimens were approved by the

| Cell culture
The high metastatic potential cell line, SACC-LM, and poor metastatic potential cell line (SACC-83) were obtained from the State Key Laboratory of Stomatology, Sichuan University. All cell lines were maintained in high glucose DMEM (HyClone, USA) (10% foetal calf serum, HyClone), 100 U/mL penicillin (Beyotime Biotechnology, China) and 0.1 mg/mL streptomycin (Beyotime Biotechnology) and incubated at 37°C at 5% CO 2 . Change the frequency of culture solution according to the growth rate and medium colour. Cells were passaged when cells covered 80%-90% of the bottom.
inhibited the expression of E-cadherin, while the VEGFA inhibitor reversed these changes. In addition, VM channels existed in 25 of 95 SACC samples, and there was a strong positive correlation between VM and clinic stage, distant metastases, VEGFA and HIF-1α expression.
Conclusions: VEGFA played an important role in hypoxia-induced VM through regulating EMT and stemness, which may eventually fuel the migration and invasion of SACC.

Cell invasion was performed by BD BioCoat™ Matrigel™ Invasion
Chamber (BD Biosciences, USA) according to the manufacturer's protocol. Briefly, cells (5 × 10 4 cells) were plated on the top chamber in serum-free DMEM and 10% FBS in DMEM in the bottom chamber.
Cells were counted at a 100× magnification in standard microscopy 24 hours later.

| Immunofluorescence
Cells were washed with 1× PBS for 3 minutes, fixed with 4% paraformaldehyde for 20 minutes and washed for three times with 1× PBS for 3 minutes. Then, cells were treated with 0.5% Triton X-100 for 20 minutes at room temperature (this step was omitted when the antigen is expressed on the cell membrane) and washed with 1× PBS.
Blocking was performed by normal goat serum for 30 minutes, fol-

| Vasculogenic mimicry assays
Wells of 24-well plate were coated with Matrigel basement membrane matrix (BD). It was allowed to polymerase at room temperature for 1 hour and 37°C for 30 minutes. Cells were resuspended and seeded into a well at a density of 1 × 10 5 /mL, and then, the experimental group was incubated in a hypoxic incubator at 37°C, 94% N 2 , 5% CO 2 and 1% O 2 , while the control group was cultured in a normoxic incubator at 37°C, 5% CO 2 and 21% O 2 . Cell morphology and formation of cord-like structures were observed using fluorescence inverted microscope (Olympus, Japan). Forty-eight hours later, the length of tubes per field was quantified by counting in five randomly chosen 100× scopes. After 48 hours, the expression of VE-cadherin and VEGFA was observed by immunofluorescence microscopy.

| Scanning electron microscopy (SEM)
Cells cultured on a coverslip were rinsed with ice-cold 1× PBS (pH 7.2) and fixed with 2% glutaraldehyde for 3 hours at 4°C. After washing with 1× PBS, the cells were fixed with 1% OsO4 for 2 hours and dehydrated in ethanol. The samples were then critical-point dried and sputter-coated with gold. Samples were examined, and images were acquired using a HITACHI S-520 scanning electron microscope.

| Confocal laser scanning microscopy (CLSM)
The cells were fixed in 4% paraformaldehyde (PFA) in 1× PBS at room temperature for 20 minutes. The cells were first pre-treated and then blocked in 1× PBS with 0.5% Triton X-100 and 5% BSA for The images were collected with the Volivity software containing Acquisition, Quantitation and Visualization modules.

| Flow cytometry (FCM)
The pre-treated cells were harvested and washed twice with FCM buffer (PBS with 5% FBS and 0.1% NaN3). Cells were resuspended in PBS and incubated with PE-anti-human CD144 (BD Biosciences) for 30 minutes at 4°C.

| Tumoursphere formation assay
For the tumoursphere formation assay, SACC-LM cells were seeded at a density of 800 cells/ mL of tumoursphere formation medium in ultra-low attachment 12-well plates for 12 days. The tumourspheres formed (spherical, non-adherent cell-masses >90 μm in diameter) were photographed and counted under inverted phase contrast microscope.

| Western-blot
The isolated cells were lysed in lysis buffer. The protein concentration was then determined. Equal amounts of protein were analysed by SDS-PAGE, followed by electrophoretic transfer to PVDF membranes (Millipore Corp., Billerica, MA, USA). The membrane was blocked for 1 hour with 5 mL blocking buffer (5% skim milk in TBST) and incubated overnight at 4°C with 5% skim milk diluted primary antibody: Rabbit anti-Human VE-cadherin antibody (1:500 dilutions, ProteinTech); Rabbit anti-Human VEGFA antibody (1:500 dilutions, ProteinTech); and Rabbit anti-Human GAPDH antibody (1:500 dilutions, ProteinTech). The next day, the membranes were washed three times with TBST and incubated with 2% skim milk diluted secondary antibody for 2 hour. After washing, the immunoreactive protein bands were visualized using the BIO-RAD gel imaging system. GAPDH was used as an internal control.

| Real-Time PCR (RT-PCR)
Total RNA was extracted from cells using Trizol (Invitrogen, USA). Then, we synthesized cDNA using a reverse tran-
The tissue sections were then deparaffinized and dehydrated followed by incubation in 3% hydrogen peroxide for 10 minutes. Slides were stained with primary antibodies at 4°C overnight after blocking with appropriate serum. Corresponding secondary antibodies were used for 1h at room temperature. Targeted molecules were detected following DAB staining for immunohistochemistry. Tissues were sectioned for CD31 immunohistochemical staining, followed by PAS staining. Slides were finally counterstained with haematoxylin.
PBS was used as the primary antibody for the negative controls.

| Statistical analysis
The data were recorded as mean ± SD (standard deviation) and evaluated by SPSS 21.0. Differences between groups were analysed by Student's t test or chi-squared test. P < 0.05 was considered to indicate a statistically significant result.

| Hypoxia contributed to migration and invasion, and EMT of SACC cells
To investigate the role of hypoxia in SACC, we applied wound healing and transwell invasion assays to observe migration and invasive abilities under hypoxia. As shown in Figure 1A

| VEGFA was crucial for hypoxia-induced VM of SACC cells in vitro
As we known, VEGFA plays a key role in tumour-associated angiogenesis, but its role in vasculogenic mimicry remains unclear. In

| VEGFA resulted in hypoxia-induced VM formation by inducing EMT in SACC
For epithelial tumour cells, the capability of mimicking endothelial functions played an important role in VM formation. 22 Thus, we  Figure 4A,B). It was documented that EMT endowed epithelial cancer cells with the self-renewal capacity. 23 And our group previously

F I G U R E 3 VEGFA was involved in VM formation of SACC-LM and SACC-83 cells in vitro. A, Immunofluorescence staining and RT-PCR
assessed the effect of hypoxia on VEGFA protein and mRNA expression in SACC cell lines, respectively. The data showed that hypoxia enhanced the level of VEGFA of SACC cells in 2D culture in both protein and mRNA levels (magnification, ×100; bars, 100 μm, *P < 0.05). B, VEGFA expression in SACC-LM and SACC-83 cells was assessed by immunofluorescence staining and RT-PCR. The data showed that VEGFA was overexpression accompanied by high expression of VE-cadherin under hypoxia (magnification, ×100; *P < 0.05). C, CLSM (magnification, ×200) was performed to assess the role of VEGFA on VE-cadherin of SACC-LM cells. The data showed that VEGFA enhanced the level of VE-cadherin while bevacizumab suppressed its expression. D, Western blot was performed to assess the role of VEGFA on VE-cadherin of SACC-LM cells. The data showed that VEGFA enhanced the level of VE-cadherin while bevacizumab suppressed its expression. E, RT-PCR (*P < 0.05) was performed to assess the role of VEGFA on VE-cadherin of SACC-LM cells. The data showed that VEGFA enhanced the level of VE-cadherin while bevacizumab suppressed its expression proved that CD133 + , the stemness marker, was positively associated with VM formation in SACC specimens. 21 Therefore, we inferred that VEGFA also contributed to the acquisition of stem cell phenotype in SACC cells. Immunofluorescence staining revealed that compared to the control group, SACC-LM treated with exogenous VEGFA showed the increase of stemness marker-CD44 and ALDH1 expression in VM-forming cells. On the contrary, bevacizumab led to a significant decrease in CD44 and ALDH1 expression of tube-forming cells ( Figure 4C). However, the flow cytometry showed that the expression of CD144, another stemness marker, had no difference among the three groups ( Figure 4C). We assessed the sphere formation of SACC-LM under the stimulation of exogenous VEGFA or bevacizumab. And we observed that the exogenous VEGFA heightened the sphere-forming capacity while bevacizumab inhibited the tumoursphere formation ( Figure S1C). Collectively, these experiments suggested that VEGFA promoted hypoxia-induced VM formation, which may be mediated by the EMT process and cancer stem cells (CSCs) in SACC cells in vitro.

| Presence of VM in SACC tissues correlated with high expressions of VEGFA and HIF-1α
The presence of VM was authenticated by the presence of PASpositive loops and/or contained red blood cells that were negative for the endothelial cell marker CD31, while the endothelial vascular channels were positive for CD31 staining ( Figure 5A). As shown in Table 1

| D ISCUSS I ON
Hypoxia, a common characteristic of solid tumours, has been reported to activate migration, invasion and VM formation in several tumour models including melanoma, 23 oral squamous cell carcinoma 24 and glioma. 25 Here, we found that hypoxia accelerated VM EMT, the reversible dedifferentiation process of polarized epithelial cells, has been shown to play an important role in acquiring endothelial-like properties to form vessel-like structures in epithelial cancers. 35 Meng et al 36  Taken together, in present study, we confirmed that hypoxia accelerated the VM of SACC cells, and VEGFA was a critical downstream molecule that mediated the hypoxia-controlled EMT inducing VM in SACC. In human SACC tissues, the presence of VM was positive correlation with the advanced clinic stage and distance metastasis, as well as the levels of VEGFA and HIF-1α in SACC tissue. These indicated that hypoxia may serve as an inducer of VM in SACC by targeting VEGFA-mediated EMT, which provided a new sight in the search the therapeutic target of SACC.

ACK N OWLED G M ENTS
The present study was supported by National Natural Science

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.