Correlation of SOX4, SOX17, VE-cadherin and vasculogenic mimicry with prognosis in esophageal squamous cell carcinoma

Background: This study to explore the relationships between SOX4, SOX17, vascular endothelial (VE)-cadherin and VM with the occurrence and development of ESCC. Methods: SOX4, SOX17, and VE-cadherin expression as well as VM in tissues were determined by immunohistochemistry (IHC). The cell invasion, migration, and proliferation were determined after silencing of VE-cadherin. SOX4, SOX17, and VE-cadherin protein and mRNA expression were quantied by Western blotting and qRT-PCR analyses, respectively. Results: The expression of SOX17, SOX4, and VE-cadherin were signicantly correlated with clinical characteristics of ESCC. After VE-cadherin silencing, the cell invasion, migration, and proliferation were decreased, while SOX17 levels were increased and SOX4 levels were decreased. Conclusions: SOX17, SOX4, and VE-cadherin are involved in the development of ESCC. SOX4 survival of ESCC to determine exact pathophysiological mechanism linking SOX4, SOX17, VE-cadherin, and VM.

In SOX17-and SOX4-positive cells, aky or granular brownish yellow staining was seen in the nucleus or cytoplasm, while VE-cadherin-positive cells showed aky or granular brownish yellow staining of the cell membrane and cytoplasm. The staining results included the proportion of positive cells and staining intensity [25]. The proportion of positive cells refers to the percentage of positive cells among the total observed cells of the same species: 0 (≤10%); 1 point (11~25%); 2 points (26%~50%); 3 points (51%~75%); and 4 points (> 75%). Staining intensity was graded as 0, 1, 2 and 3 points for no staining, light yellow, brownish yellow and tan yellow staining, respectively. The points for percentage of positive cells and staining intensity were multiplied, and the mean value was calculated to decide the staining results as follows: 0~3 was considered negative, and 4~12 was taken as positive.
For all CD34-stained immunohistochemical sections, DAB color development was performed, and the color development reaction was stopped by washing with owing water for 1 min. The cells were rinsed with water for 2 min and then stained with periodic acid-Schiff (PAS) for 15~30 min. The cells were again rinsed with distilled water thrice for 1 min each. VM was detected by the presence of tumor cells around the PAS-positive and CD34-negative tubes with few necrotic tumor cells and in ammatory cells in ltrating the surrounding tissues, and absence of red blood cells in the lumen of the tubes. Endothelium-lined normal vessels were identi ed by the presence of CD34-positive endothelial cells in their wall.

Cell lines and cell culture
The human ESCC cell lines EC9706 and Eca109 were grown in Dulbecco's Modi ed Eagle's Medium (DMEM, Hyclone, USA), supplemented with 10% fetal bovine serum (FBS, Gibco, USA) in 5% CO 2 at 37℃. The cell lines were divided into three groups: control group, control siRNA group, and VE-cadherin siRNA group. In the control group, EC9706 and Eca109 cells were not treated. In the control siRNA group, EC9706 and Eca109 cells were infected with the empty plasmid. In the VE-cadherin siRNA group, EC9706 and Eca109 cells were infected with a lentivirus encoding precursor VE-cadherin or vector and treated with puromycin for 2 weeks to obtain stably transfected cells.
siRNA transfection EC9706 and Eca109 cells were seeded into six-well plates and transfected with VE-cadherin siRNA and a negative control (NC, GenePharma, China) using Lipofectamine 2000 (Invitrogen, USA) according to the manufacturer′s instructions. The siRNA sequence was: CCAUUGUGCAAGUCCACACAUTT (forward, 5′-3′) and AUGUGGACUUGCACAAUGGTT (reverse, 5′-3′). The cells were subjected to analysis as described in the Results section.

Western blotting
An appropriate amount (250~500 mg) of fresh tissue or properly preserved tissue was immersed in 1 ml of strong radio-immunoprecipitation assay (RIPA) buffer containing phenylmethylsulfonyl uoride (PMFS) was added. An electric homogenizer was used to produce the homogenate. The samples were collected after addition of lysis buffer and placed on ice for cracking for 20~30 min. Then centrifugation was performed at 12000 rpm for 10 min. ESCC cells were lysed in RIPA buffer with proteinase inhibitors. Protein concentrations were quanti ed using a BCA assay kit (Beyotime Biotechnology, China).
Wound healing assays Two cell lines were seeded overnight in six-well plates followed by transfection with VE-cadherin siRNA or NC siRNA. When cells reached greater than 90% con uency, the tip of a pipette was used to make a wound, and the detached cells were rinsed away with phosphate-buffered saline (PBS). Images of the wounds were taken at 0 h and 24 h.
MTT assay ESCC cell lines were seeded overnight into 96-well plates at 5×10 3 cells per well. Subsequently, the cells were transfected with VE-cadherin siRNA for 72 h. Cell viability was measured by MTT assay as described previously [25].

Transwell migration and invasion assay
Cell migration and invasion were evaluated by Transwell assay as previously described [26]. Brie y, transfected ESCC cells were seeded in 24-well plates with 8-µm-pore-size chamber inserts (Corning, USA). The upper chambers were coated with Matrigel (BD Biosciences, USA) before cell seeding. After incubation for 48 h, invading and migrating cells on the bottom surface of each chamber were stained with Giemsa solution and photographed. The migrating cells were then counted in ve random elds for quanti cation.

Statistical analyses
The continuous and categorical data are presented as mean ± standard deviation (SD) and frequency (percentage), respectively. Comparisons of quantitative data between two and multiple groups were conducted by Student′s t test and one-way analysis of variance (ANOVA), respectively, using GraphPad Prism 8.0.
Kaplan-Meier curves with log-rank tests were used for univariate overall survival (OS) analysis. Cox regression models were used for multivariate OS analysis. Differences were considered to be statistically signi cant if P<0.05.

SOX17, SOX4, and VE-cadherin expression and VM are associated with clinical characteristics of ESCC
The positivity rates for SOX17 expression in the normal esophageal mucosa and ESCC samples were 83.3% and 41.4%, respectively, and the difference was statistically signi cant (P<0.05). There was no signi cant correlation between SOX17 protein expression and clinicopathological characteristics such as gender, age or tumor location (P>0.05). SOX17 protein expression showed inverse correlations with tumor size, grade of differentiation, depth of invasion, and pathological tumor node metastasis (pTNM) stage. The rate of SOX17 expression was lower among cases with lymph node metastasis (LNM) than among those without LNM (P<0.05). The rate of SOX4 protein expression in the normal esophageal mucosa was signi cantly higher than in ESCC (30% vs. 60%, P<0.05). The SOX4 expression rate showed positive correlations with the grade of differentiation, depth of in ltration, LNM, and pTNM stage of ESCC (P<0.05). The rate of VE-cadherin protein expression in ESCC was signi cantly higher than in the normal esophageal mucosa (51.9% vs. 3.3%, P<0.05). Moreover, VE-cadherin protein expression showed no correlation with patient age and gender. VE-cadherin protein expression was positively correlated with tumor size, grade of differentiation, LNM, pTNM stage, and depth of in ltration of ESCC (P<0.05). The positive rate of VM in ESCC was 50% (105/210), while no VM was found in the test group ( Figure 1). VM positivity was not correlated with gender, age, tumor location, or histological grade (P<0.05), but did correlate with tumor size, gross type, in ltration depth, LNM, and pTNM stage (P<0.05). The above results are summarized in Table 1 and Fig. 1A-H.

Survival analysis
The 5-year OS rate in the ESCC group was 37.1% (78/210). The OS of patients with SOX17 expression was signi cantly better than that of patients without SOX17 expression (P<0.001; Table 3, Figure 2A). On the other hand, the OS of patients with SOX4 expression, VE-cadherin expression, and VM was signi cantly lower than that of patients negative for these factors (P<0.001; Table 3, Figure 2B-D). On Cox regression model analysis, various factors such as gender, age, tumor type, tumor location, tumor diameter, histological grade, LNM, depth of invasion, pTNM stage, VM, SOX17 expression, SOX4 expression, and VE-cadherin expression were identi ed as prognostic factors. It was found that the expression of SOX17, SOX4, and VE-cadherin as well as VM were independent risk factors affecting the long-term prognosis of ESCC patients ( Table 4) Additionally, the wound healing speed of cells in the VE-cadherin siRNA group was signi cantly slower than those in the corresponding control groups ( Figure   4E-G).

Silencing VE-cadherin reduced EC cell proliferation
The proliferative abilities of EC9706 and Eca109 cells were signi cantly weakened after silencing of VE-cadherin (P<0.05, Fig. 4H). No signi cant difference was observed between the control group and the control siRNA group (Fig. 4H).

Silencing VE-cadherin increased SOX17 expression and decreased SOX4 expression
After transfection, the VE-cadherin protein expression was signi cantly lower in the VE-cadherin siRNA group than in the control groups (P<0.05). Moreover, SOX17 protein expression was signi cantly upregulated and SOX4 protein expression was downregulated in the VE-cadherin siRNA group (Fig. 5).

Discussion
The formation of VM can provide blood supply for the rapid proliferation of tumors, relieve the ischemic and hypoxic micro-environment around the tumors, and further accelerate the invasion and metastasis of tumors, which in uences the clinical stage and long-term prognosis of cancer patients [5,27]. In this study, we con rmed the existence of VM in ESCC by IHC analysis of tumor tissues. At the same time, we found the VM was closely associated with the depth of invasion, pTNM stage, and LNM. The above ndings are consistent with those of previous studies [28]. On survival analysis, the presence of VM was an independent poor prognostic factor in ESCC patients. VE-cadherin, as an adhesion protein, can mediate the adhesion of cells to each other and maintain the further formation of tumor blood vessels. In the present study, VE-cadherin was found to be highly expressed in ESCC, and its positive expression was directly related to the depth of invasion, occurrence of LNM, and pTNM stage in ESCC. VE-cadherin expression was found to be an independent poor prognostic factor for ESCC patients. In vitro experiments suggested that high expression of VE-cadherin can accelerate ESCC invasion and metastasis, similar to the ndings of previous reports [28]. Notably, after siRNA-mediated interference of bcl-2 expression in EC9706 cells under hypoxic conditions, the expression of VM-related molecules, such as VE-cadherin and matrix metalloproteinase 2 (MMP2), was signi cantly inhibited and VM generation was signi cantly reduced [29]. Moreover, VE-cadherin downregulation in melanoma is associated with the loss of VM formation [30]. Heinolainen et al. [31] and Han et al. [32] speculated that VE-cadherin might be an important determinant of VM in EC. Based on the ndings of the present study, we rmly believe that VE-cadherin promotes the formation of VM in ESCC.
The SOX17 transcription factor has been known to have tumor suppressive function in ESCC [35][36][37]. SOX17 overexpression suppresses colony formation and cell migration/invasion in ESCC cell lines. In addition, SOX17 overexpression was found to inhibit tumor growth and metastasis in an ESCC xenograft model [33][34][35]. In the present study, we found signi cantly higher SOX17 expression in ESCC compared to the normal esophageal epithelium, con rming the tumor suppressive function of SOX17. A previous study demonstrated that hypermethylation of the promoter of the SOX17 gene leads to silencing of SOX17 protein expression in ~50% ESCC patients [33]. In the present study, the low expression of SOX17 was signi cantly correlated with tumor differentiation, depth of invasion, LNM, and pTNM stage, suggesting that SOX17 acts as a tumor suppressor gene in ESCC. Moreover, SOX17 expression was an independent predictor of prognosis in this study.
The present study found that SOX4 protein and mRNA expression levels were signi cantly increased in ESCC. Moreover, the high expression of SOX4 promoted the invasion of ESCC. Schilham et al. also found that SOX4 overexpression is closely associated with ESCC metastasis [36]. SOX4 has been found to be involved in tumorigenesis and tumor invasion in different types of cancers through the activated Wnt, transforming growth factor (TGF)-β, Hedgehog and Notch pathways or the regulation of microRNA expression [30]. Zilong et al. speculated that miR-212-3p targets SOX4 to inhibit the Wnt/β-catenin signaling pathway and promote apoptosis in ESCC [37].
VM is closely related to tumor growth, invasion, metastasis, and long-term prognosis of cancer patients [38][39][40]. In the present study, we con rmed that VM formation in ESCC was positively correlated with high SOX4 expression and low SOX17 expression. These ndings indicate that VE-cadherin may promote the formation of VM in ESCC by affecting the expression levels of SOX4 and SOX17. In a previous study, we con rmed that SOX4 may promote the formation of VM by promoting EMT in ESCC [23]. Studies have shown that in a hypoxic environment, hypoxia-inducible factor (HIF)-2α, a VM-initiating factor, is activated, which increases VE-cadherin transcription. VE-cadherin, in turn, induces repositioning of EphA2 to the cell membrane. Furthermore, PI3K is activated by VEcadherin and EphA2 simultaneously. The activated PI3K regulates the activation of the pre-gene of membrane type 1 matrix metalloproteinase (MT1-MMP). The combination of MT1-MMP and MMP2 promotes the fragmentation of laminin 5γ25γ2 (LN-5γ2) chains into fragments (5γ2 and 5γ2χ), and increased levels of these two fragments in the extracellular micro-environment eventually leads to the formation of a VM net-like structure [41].
In conclusion, the present study found low SOX17 expression, high SOX4 expression, and high VE-cadherin expression in ESCC. Moreover, the expression of these proteins was closely associated with VM in ESCC. We believe that development of targeted therapies to suppress SOX4 expression or enhance SOX17 expression may impair the formation of VM, thereby prolonging the survival of ESCC patients. Future studies are required to determine the exact pathophysiological mechanism linking SOX4, SOX17, VE-cadherin, and VM.      Comparison of in vitro proliferation of EC9706 and Eca109 cells (H). *P<0.05 vs control or control siRNA.