FOXM 1 induces Vasculogenic mimicry in esophageal cancer through β-catenin /Tcf4 signaling

Objective To investigate the role of FOXM1, β-catenin and TCF4 in esophageal cancer (EC) and their relationship to VM (Vasculogenic Mimicry). Methods CCK-8 were performed to examine EC cell proliferation in FOXM1 silenced cells. EC cell migration and invasion were investigated through wound healing and Transwell assays, respectively. The formation of pipe like structures were assessed in 3D cultures. The expression of Foxm1, β-catenin, Tcf4 and E-cadherin were investigated through western blot, RT-qPCR and immunohistochemistry (IHC) staining. The relationship between FOXM1 expression, clinic-pathological features, and overall survival (OS) were further analyzed. Results A loss of FOXM1 expression correlated with the OS of ESCC patients. FOXM1 silencing led to a loss of cell growth and suppressed cell migration and invasion in ESCC cells. VM structures were identified in ESCC tissues and human EC cell lines. Mechanistically, FOXM1 was found to promote tumorigenesis through the regulation of β-catenin, Tcf4, and E-cadherin in EC cells, leading to the formation of VM structures. Conclusions These findings highlight FoxM1 as a novel therapeutic target in ESCC.


Background
Esophageal Carcinoma (EC) is an invasive malignancy of the digestive system and the 6th leading cause of tumorrelated mortality. EC originates from the mucosa or glands of the esophagus [1,2]. It is estimated that 17,290 patients in the United States (US) alone were diagnosed with esophageal cancer in 2018, from which 15,850 individuals died [3]. Two major histopathological subtypes of EC exist, including esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma. ESCC is the predominant histological subtype in China, accounting for~90% of cases.
The mechanisms governing EC development remain poorly defined. ESCC is characterized by a high degree of malignancy, metastasis and recurrence, and poor prognosis [4]. Due to the low efficiency of current diagnostics, most ESCC patients present lymph node or distant metastasis at advanced stages. Although the therapeutic effects of surgery combined with radiotherapy and chemotherapy have improved the outcome of ESCC patients, the 5-year OS rates are ≤25% [5,6]. A further understanding of the molecular mechanisms of ESCC development are required for new diagnostic markers and therapeutic strategies.
Wnt signaling is key to cancer development, particularly ESCC. Approximately 90% of ESCC cases show dysregulated Wnt signaling, leading to the nuclear translocation and aggregation of β-catenin through inactivating APC mutations or activating β-catenin mutations, followed by the combination of T-cell factor (TCF) and lymphoid enhancing factor transcription factors (LEF) to activate the transcription of downstream genes, including c-Myc and cyclin D1 [17]. Studies has revealed that FOXM1 regulates is a novel target of Wnt signaling and indispensable for β-catenin/TCF4 transactivation. In glioma, FOXM1 interacts with β-catenin to enhance its nuclear translocation and ability to promote self-renewal and tumorigenesis. In osteosarcoma, FOXM1 and βcatenin directly interact to promote TCF4 binding to the Wnt gene promoter, enhancing Wnt activity [23]. Yoshida et al. found that the deletion of FOXM1 in mice prevented colorectal tumorigenesis through the regulation of β-catenin/TCF4/E-cad signaling and the suppression of EMT [24].
Vasculogenic mimicry (VM) is a novel blood-supply system first proposed in invasive human uveal melanoma and metastatic cutaneous melanoma by Maniotis et al. in 1999 classified VM into two sub-types: tubular and pattern matrix. Tumors modify their blood supply through the ECM to form a vessel like pipeline, representing a tumor microcirculation mode that is independent of vascular endothelial cells [26]. Through VM, tumor tissues can contact host blood vessels for oxygen and nutrition, increasing tumor metastasis and invasion [26][27][28]. Red blood cells (RBCs) are visible in the VM channel in which no necrotic hemorrhage is evident. Subsequently, fluorescence dyes and activated carbon particles injected into host blood vessels can be detected in the VM channel through laser scanning confocal microscopy [26][27][28]. Periodic acidschiff (PAS) positive ECM and tumor cells are arranged in the VM channel with endothelial cells and markers that cannot be detected using light microscopy, transmission electron microscopy or immunohistochemistry [29]. Increasing evidence suggest that cancer stem cells (CSCs) and EMT play a vital role in the tumor vasculature [30][31][32]. EMT is a temporary and reversible transition from epithelial cells to mesenchymal cells (such as myofibroblasts) that refers to the removal of the differentiated epithelial cells, a loss of polarity, reduced contact with surrounding cells and the extracellular matrix, and the enhanced mobility and migration of epithelial cells. Epithelial marker such as E-cadherin are gradually lost during EMT, whilst mesenchymal phenotypes such as vimentin, N-cadherin, alpha-SMA, Snail and slug increase [33]. In this study, we investigated the roles of FOXM1, β-catenin, TCF4, and E-cadherin in EC to verify the role of FOXM1 in the formation of VM structures. We herein describe a novel anti-angiogenic strategy to target EC tumor vascularization.

Human ESCC samples
From June 2012 to June 2015, 216 cases of paraffin embedded ESCC and normal adjacent esophageal mucosal tissues were collected during surgery in the first affiliated hospital to Bengbu Medical College (Anhui, China). The clinical data of ESCC patients are shown in Table 1. None of the patients accepted radiotherapy or chemotherapy preoperatively. Informed consent was obtained from each patient. The study was approved by the Bengbu Medical College ethics committee.

Reagents
Rabbit monoclonal antibodies against human Foxm1, βcatenin, and Tcf4 were purchased from Abcam, USA. Anti-CD34 antibodies were purchased from Lab Vision Company (USA). Elivison™ plus immunohistochemical assays and diaminobenzedine (DAB) chromogenic reagent kits were purchased from Maixin Biology Company (Fuzhou, China). Periodic Acid Schiff (PAS) reagents were produced in the Department of Clinical Pathology, First Affiliated Hospital of Bengbu Medical College.

Immunostaining and CD34 + PAS double staining
All paraffin embedded specimens were immobilized in 10% formalin. All sectioned specimens (4 μm thick) were waxed, dehydrated with a gradient alcohol series, and washed in PBS (pH 7.2) for 10 min. Sections were incubated with 3% H 2 O 2 methanol at room temperature for 10 min to inhibit endogenous peroxidase activity. Samples were boiled in 1.0% citrate (pH 6.0) for 2 min at high pressure for antigen retrieval. Slides were washed in PBS and blocked in goat serum at room temperature for    Abcam, USA) primary antibodies overnight at 4°C, followed by the appropriate secondary antibodies for 30 min. All CD34 labeled specimens were stained with PAS to characterize the glycolic basement membrane of the vascular endothelial cells, in addition to vascular-like (VM) channels [24,25]. After rinsing with PBS, freshly prepared 3, 3 '-diaminobenzidine (DAB) solution was added to the sections for 5 min, and hematoxylin was back-dyed, dehydrated, air-dried and loaded.

Cell culture
Human Eca109 and Ec9706 cells were obtained from the Chinese Academy of Sciences (Shanghai, China). Cells were cultured in the basic medium supplemented with 10% fetal bovine serum (Gibco; Thermo Fisher Scientific, Inc., Waltham, MA, USA) at 37°C. Cells were passaged using 0.05% trypsin (Invitrogen, Carlsbad) and seeded into 6-or 96-well plates.

Real-time quantitative PCR
Trizol (Invitrogen; Thermo Fisher Scientific, Inc) was used to isolate total RNA from transfected cells and cDNA was reverse-transcribed using MMLV kits (Promega Corporation). SYBR green Premix Ex Taq II (Qiagen GmbH) and One Step plus RT-PCR systems (Applied Biosystems; Thermo Fisher Scientific, Inc.) were used for qPCR. The ABI Prism 7500 sequence detection system (Applied Biosystems; Thermo Fisher Scientific, Inc.) was used for all PCRs. Reaction conditions were as follows: total volume 25 μl: 10 μl SYBR Premix Ex Taq (2×), 1 μl forward Primer; 1 ul reverse primer; 0.5 μl ROX Reference Dye II (50×)* 3, 2 μl cDNA and 8 μl ddH 2 0. PCR parameters included: initial denaturation at 95°C for 10 min, followed by 40 cycles of 95°C for 5 s; 63°C for 30 s; and 72°C for 30 s; final extension: 72°C for 5 min. Relative gene expression was detected using the 2 ΔΔCq method (ΔΔCq = ΔCq ESCC ΔCq corresponding normal tissues) [34]. Primers sequences are shown in (Table 2). Gene expression was normalized to GAPDH as an internal standard. All PCRs were performed in triplicate.

Western blotting
Cells were lysed in RIPA buffer (Solarbio Technology Co, Ltd., Beijing) and protein concentrations were determined via BCA assay. Proteins (20 μg

Wound healing assays
Wound healing assays were used to evaluate the ability of Eca109 cells to migrate. Scratches were introduced into 6well plates with a 10 μl pipette tip (Eppendorf, Hamburg, Germany) and cells were grown in DMEM plus 1% fetal bovine serum at 5% CO 2 at 37°C. Cells in and around the scratch area were imaged at 0 h, 20 h and 30 h. Wound healing was monitored on a light microscope.

Transwell chamber assays
Transfected EC cells were seeded into Transwell chambers (Costar; Corning Inc., Corning, NY, USA). For invasion assays, chambers were coated with Matrigel (BD, USA) and cells were inoculated in serum free medium (5 × 10 4 cells). The chamber below included DMEM containing 10% fetal bovine serum. After cell infiltration for 20 h, cells were washed, fixed, and stained with Calcein-AM or crystal violet for 20 min. Invading cells were counted and imaged (Hanrong Company, Shanghai).

3D cultures
Preserved matrix glue was dissolved into a liquid state at 4°C overnight. Micropipette tips and 24-well cell culture plates were precooled on ice, and matrix colloidal solution was added to the cells on ice. Cells were trypsinized into single-cell suspensions at a density of 5 × 10 5 /ml after. Twelve hour. Tubular structure arrangements and cell integrity then imaged across five visual fields (top, bottom, left, right and center) using an inverted microscope (× 200). The number of tubular structures were counted, and the mean value of each visual field recorded. Experiments were repeated on a minimum of 3 occasions.

Statistical analysis
All data are expressed as the mean ± stand deviation (SD). A one-way ANOVA followed by Tukey's post-hoc test was used for inter-group comparisons. A student's t-test was used to evaluate differences. GraphPad version 5 for Windows (GraphPad Software, USA) was all statistical analysis. Counting data were analyzed using a Chi-square test for group comparisons. Multiple logistic regression analysis was used to determine the factors influencing metastasis. Kaplan-Meier assays with log-rank tests were used for univariate OS analysis. Cox regression models were used for multivariate OS analysis. P-values < 0.05 were deemed statistically significant.

Results
Association between VM, FOXM1, β-catenin, TCF4 and Ecadherin expression and the clinicopathological features of EC To investigate the role of VM, FOXM1, β-catenin, TCF4 and E-cadherin during ESCC progression, immunohistochemical analysis of both ESCC and corresponding adjacent normal tissues were performed and compared to the clinicopathological features of the patients. VM+ was determined as small vessels with channels in ESCC that were PAS+ but CD34−. The VM structures included circular tubular, straight lines, and networks in the ESCC samples (46.76%; 101/216) that were higher than corresponding adjacent normal tissue (0%; 0/216; P < 0.001; Fig. 1a, b, c). The VM in ESCC was related to LNM, serosal invasion, and PTNM stage. No association with patient gender, age, anatomical location, gross morphology, degree of differentiation or tumor diameter were observed ( Table 1).
As is shown in Fig. 2c, β-catenin localized to the nucleus and cytoplasm and its positive expression significantly increased in ESCC tissue (66.20%, 143/216) compared to corresponding adjacent normal tissue (18.06%, 39/ 216; P < 0.001; Fig. 2d). The rate of βcatenin + expression was related to lymph node metastasis, serosal infiltration, and PTNM stage. No relationship was observed between β-catenin expression and patient gender, age, anatomical location, gross morphology, degree of differentiation and tumor diameter ( Table 1).
As is shown in Fig. 2e, TCF4 + was mainly observed in the nucleus and correlated with the high expression of ESCC tissue (61.11%, 132/216) compared to corresponding adjacent normal tissue (16.67%, 36/ 216; P < 0.001; Fig. 2 F). Immunostaining indicated that TCF4 + expression was related to LNM, serosal invasion, and PTNM stage, but not to patient gender, age, anatomical location, gross morphology, the degree of differentiation or tumor diameter ( Table 1).
As is shown in Fig. 2g, e-cadherin expression was mainly confined to the membrane and cytoplasm, and positive expression was lower in ESCC tissues (38.43%, 83/ 216) compared to corresponding adjacent normal tissue (82.41%, 178/216; P < 0.001; Fig. 2h). Immunostaining demonstrated that E-cad + expression was inversely related to LNM, serosal invasion, and PTNM stage, but not with patient gender, age, anatomical location, gross morphology, the degree of differentiation or tumor diameter (Table 1).

Survival prognosis
The OS was identified as the period from the initial time from surgery to death or at last follow-up. KM survival analysis showed that the median OS (27.526) and OS rates [17.82% (18/101)] in the VM+ group were      (Fig. 3). Multivariate analysis indicated that VM, Foxm1, βcatenin, Tcf4, E-cadherin, LNM, serosal infiltration, and PTNM stage were independent prognostic factors for ESCC (Table 4).

FoxM1 analysis
The expression of FoxM1 in Ec9706 cells was~46% lower than that of Eca109 cells (P < 0.05, Fig. 4a). Eca109 cells were therefore selected for subsequent analysis. Cells were transfected with FoxM1-siRNA and silencing was confirmed at both the mRNA and protein levels (P < 0.05, Fig. 4b).
FoxM1 silencing inhibits the proliferation of EC cells CCK-8 assays were used to investigate the effects of FoxM1 silencing in Eca109 cells. The results showed that the OD absorbance values in the FoxM1-siRNA group increased compared to NC groups (P < 0.05, Fig. 5) suggesting that FoxM1 silencing inhibited cell proliferation. No differences were observed between NC and blank control groups (p > 0.05).

FoxM1 silencing inhibits the invasion and migration of EC cells
Transwell migration assays showed that the number of cells penetrating the cell membrane (98.00 ± 10.97) in the FOXM1-silenced groups were lower than those of the NC (296.00 ± 6.66) or blank control groups (283.70 ± 12.41). Cell migration also significantly declined (p < 0.05, Fig. 6). No significant differences were observed between NC and blank control groups (P > 0.01). The levels of cell membrane penetration (76.33 ± 2.40) were lower in the FoxM1-silenced group compared to NC (270.67 ± 1.45) and blank control groups (260.33 ± 3.18). Cell invasion ability also significantly declined (P < 0.05, Fig. 6). No significant differences were observed between NC and blank groups (P > 0.01). Wound healing assays showed that FoxM1 silencing led to slower wound closure rates compared to the control group (P < 0.05, Fig. 6).

Effects of Foxm1 silencing on β-catenin, Tcf4, and Ecadherin expression
The expression of Wnt related proteins in each group were investigated by western blot and qRT-PCR. The results showed that the expression FoxM1 and Tcf4 in FoxM1 silenced groups were significantly lower than blank-and negative control groups (P < 0.05, Fig. 4d-e). The expression of E-cadherin in the FoxM1-siRNA group was higher than that of blank-or negative control groups (P < 0.05). No significant differences in total β-catenin expression were observed (P > 0.05). The nuclear localization of β-catenin decreased in the FoxM1-siRNA group (P < 0.01, Fig. 4c). These findings confirmed that FoxM1 silencing prevent the EMT of tumor cells.

Assessment of 3D tubular structures
We assessed 3D tubular structures to investigate cell pipeline formation. Blank-and NC groups could simulate the characteristics of vascular endothelial cells and could connect on the matrix gel. After 12 h, a typical vascular network structure was formed, with single multi-ring connections. However, in the FoxM1-siRNA group, tube was significantly inhibited, and annular structures were broken, most of which were incompletely closed linear structures. The quantity of tubular structures in the FOXM1-SiRNA group were also significantly lower than the control group (14.0 ± 1.2 vs. 30.0 ± 1.2, 27.7 ± 1.5, P < 0.05, Fig. 7).

Discussion
Esophageal cancer (EC) is a prevailing tumor of the digestive system and a primary cause of cancerrelated mortality. The high heterogeneity of EC has led to a paucity of available biomarkers. To develop novel approaches to diagnosis and treatment, an improved understanding of the mechanism(s) of ESCC are required. Accumulating evidence demonstrates that FOXM1 is involved in the occurrence and development of the VM structures observed in EC, but the molecular mechanisms underlying these effects have not been elucidated. In this study, the outcome of candidate biomarkers were comprehensively evaluated to ensure their effectiveness. Highly invasive tumors have pipe-like structures, which are independent of vascular endotheliocytes and are directly formed by tumor cells, permitting contact with host blood vessels through their own deformation and ECM reconstruction to gain a blood source [26,35]. In this study, VM' structures were observed in 101 of 216 patients with ESCC, but not in corresponding adjoining normal tissue, confirming the presence of VM in ESCC. Upon PAS and CD34 double staining, VM channels form in straight lines, curves or branches, and only one layer of PASpositive substances can be observed between tumor cells and the blood flow, with no surrounding endotheliocytes. This leads to tumors more prone to recurrence and migration causing poor outcomes and high mortality rates [36][37][38][39]. This also demonstrates that VM is closely related to LNM, serosal invasion and the PTNM stage of ESCC. Survival analysis further indicated that the negative expression of VM in ESCC patients correlates with longer survival times. We conclude that tumors with VM' structures show poor differentiation, low clinical staging, and are more prone to lymph node metastasis [40][41][42][43]. FOXM1+ expression was higher in ESCC tissues than control tissues. The rates of FOXM1+ expression in ESCC were related to LNM, serosal invasion, and PTNM stage. Survival analysis further demonstrated that FOXM1 negative groups of ESCC patients had longer survival times than positive expression groups. Βcatenin + expression was also significantly higher in ESCC tissues than in control tissue. The rates of βcatenin + expression positively associated with LNM, serosal invasion, and PTNM stage. Survival analysis also indicated that ESCC patients in the β-catenin− group had longer survival times than the positive expression group. No relationship between β-catenin + expression and patient gender, age, anatomical location, gross morphology, the degree of differentiation and tumor diameter were observed. TCF4 + expression was higher in ESCC vs. control specimens. TCF4 + expression was also related to LNM, serosal invasion, and PTNM stage. Survival analysis indicated that the TCF4 negative expression group of ESCC patients had longer survival times than the positive expression group. E-cadherin + expression was significantly lower in ESCC tissue compared to control tissue and inversely related to LNM, PTNM stage and serosal infiltration. Survival analysis also showed that ESCC patients in the E-cadherin + group had longer survival times than the negative expression group. Multivariate analysis identified that LNM, serosa invasion, PTNM stages, VM, FoxM1, β-catenin, Tcf4 and E-cadherin were all independent predictors of ESCC. In clinical practice, LNM, PTNM stages and serosal invasion are considered to predict patient outcomes. In the same way, we identified a positive correlation between VM, FoxM1, β-catenin and Tcf4 and a negative relation with E-cadherin. These results indicate that the upregulation of VM, FoxM1, βcatenin, Tcf4 and the inhibition of E-cadherin plays a key role in the malignant transformation of EC cells in the esophagus, and in the progression, metastasis and invasion of ESCC. The results were in agreement with previous studies [44][45][46].
Increasing evidence has demonstrated that FoxM1 accelerates tumor invasion and migration by regulating eukaryotic extension factor 2 kinase eEF2 in triple negative breast cancer [47]. MicroRNA 630 (mi-RNA-630) inhibits the invasion and migration of gastric cancer cells by targeting FoxM1 [48]. FoxM1 increases the invasion and migration of colorectal cancer by binding to the HSPA5 promoter. Here, we found that the FoxM1 silencing inhibited cell growth and invasion, indicating that FoxM1 functions as a tumor promoter in ESCC. In addition, the network tube structure of FoxM1-siRNA groups was significantly lower than negative control or blank control groups. This suggests a positive correlation between FoxM1 and VM in EC.
The question remains as to how FOXM1 regulates tumor cell growth, invasion and the formation of VM structures in ESCC? The dysregulation of Wnt signaling is closely related to the biological behavior of various tumors. Aberrant activation of Wnt inhibits FoxM1 phosphorylation through the GSK3-Axin complex and interacts with the deubiquitinating enzyme USP5, leading to the deubiquitination of FoxM1 and the maintenance of its stability. FoxM1 accumulates in the nucleus and promotes the recruitment of β-catenin to target gene promoters of the Wnt pathway, leaving the β-catenin/TCF4 complex non-competitively inhibited by ICAT, thereby activating Wnt signaling [49]. Studies have shown that in osteosarcomas, FoxM1 and β-catenin directly bind to the cytoplasm and nucleus, enhancing the nuclear transcription of β-catenin, promoting its binding with TCF4 on the Wnt target gene promoter, enhancing the activity of the Wnt pathway [23]. Jiang et al. [50] also reported that the downregulation of β-catenin inhibits the proliferation of human nasopharyngeal carcinoma cells and reduces their invasion ability. Blocking the translocation of β-catenin into the nucleus using CGP57380 to inhibit the MNK-EIF4e axis inhibits the proliferation and progression of nasopharyngeal carcinoma cells,   and reduces their ability to migrate, invade and metastasize [51]. We further showed that when FoxM1 was silenced, the expression of Tcf4 decreased, whilst the expression of E-cad increased. Despite the comparable levels of β-catenin, its nuclear translocation markedly decreased in the FoxM1-siRNA group. This further demonstrated that FoxM1 activates β-catenin by activating Wnt signaling, enhancing the activation of the TCF/LEF promoter, leading to a loss of E-cadherin expression, intercellular adhesion ability, and the formation of VM structures.

Conclusions
This study highlights the influence of FOXM1 on the occurrence of ESCC that results in the formation of VM' structures through the β-catenin/TCF4 axis in ESCC.