ZEB1‐mediated vasculogenic mimicry formation associates with epithelial–mesenchymal transition and cancer stem cell phenotypes in prostate cancer

Abstract The zinc finger E‐box‐binding homeobox 1 (ZEB1) induced the epithelial–mesenchymal transition (EMT) and altered ZEB1 expression could lead to aggressive and cancer stem cell (CSC) phenotypes in various cancers. Tissue specimens from 96 prostate cancer patients were collected for immunohistochemistry and CD34/periodic acid–Schiff double staining. Prostate cancer cells were subjected to ZEB1 knockdown or overexpression and assessment of the effects on vasculogenic mimicry formation in vitro and in vivo. The underlying molecular events of ZEB1‐induced vasculogenic mimicry formation in prostate cancer were then explored. The data showed that the presence of VM and high ZEB1 expression was associated with higher Gleason score, TNM stage, and lymph node and distant metastases as well as with the expression of vimentin and CD133 in prostate cancer tissues. Furthermore, ZEB1 was required for VM formation and altered expression of EMT‐related and CSC‐associated proteins in prostate cancer cells in vitro and in vivo. ZEB1 also facilitated tumour cell migration, invasion and clonogenicity. In addition, the effects of ZEB1 in prostate cancer cells were mediated by Src signalling; that is PP2, a specific inhibitor of the Src signalling, dose dependently reduced the p‐Src527 level but not p‐Src416 level, while ZEB1 knockdown also down‐regulated the level of p‐Src527 in PC3 and DU‐145 cells. PP2 treatment also significantly reduced the expression of VE‐cadherin, vimentin and CD133 in these prostate cancer cells. Src signalling mediated the effects of ZEB1 on VM formation and gene expression.

usually diagnosed at advanced stages of disease, which are more likely to metastasize to other organs and have high mortality rates. 2 This is because treatment options for advanced PCa are limited; chemoradiation therapy and hormone therapy have limited effectiveness. 3 To date, the exact aetiology of PCa remains to be defined, and primary risk factors include obesity, old age, race and family history. [4][5][6][7] The development of PCa is usually related to prostatic intraepithelial neoplasia 8 due to the silencing of tumour suppressor genes and/or activation of oncogenes. [9][10][11] However, further research on PCa pathogenesis and molecular mechanisms could help identify biomarkers for early cancer detection and prediction of treatment responses and prognosis, as well as novel treatment strategies for the future control of PCa.
Zinc finger E-box binding homeobox 1 (ZEB1) is the critical epithelial-mesenchymal transition (EMT) activator and up-regulates tumour cell plasticity and the EMT to acquire cancer stem cell properties. 12 A recent study showed that ZEB1 repression by radiation enhanced lung adenocarcinoma cell migration and invasion capacity, as well as the EMT, 13 while other previous studies showed that ZEB1 played an important role in the formation of vasculogenic mimicry (VM) in colorectal and breast cancers. 14,15 In PCa, ZEB1 was demonstrated to closely associate with EMT-induced metastasis and stemness maintenance. 16,17 Moreover, the oncogene Src signalling was involved in the EMT and acquisition of cancer stem cell (CSC) phenotypes. [18][19][20] For example, ZEB1 expression promoted lung cancer cell EMT through the activation of Fak/Src signalling, 21 whereas the Src inhibition reduced the mammosphere formation and tumorigenesis potential of breast cancer stem cells. 18 In addition, blockage of Src signalling compromised VM formation in malignant glioma cells. 22 Accumulating evidence indicates that tumour EMT is relevant for the acquisition and maintenance of CSC characteristics and that it contributes to VM formation. 23,24 VM is a newly defined mechanism to supply nutrition to tumour cells and was described as the fluid-conducting channel formed by highly aggressive tumour cells without endothelial cells. 25 Tumour cells capable of VM formation have the commonality of a stem celllike, transendothelial phenotype after tumour tissues undergo hypoxia. 26 VM formation has been linked to an unfavourable outcome of various human cancers. 27,28 In PCa, high levels of tumour tissue VM are associated with higher tumour Gleason score, TNM and metastasis. 29 Thus, this study investigated the role and the potential contribution of ZEB1 in VM formation, EMT and CSC phenotypes in PCa tissues and the underlying molecular events in vitro. We hope to provide information regarding the role of ZEB1 in PCa development and progression.

| Transient and stable transfections
The small interfering RNA kit was purchased from RiboBio (Guangzhou, China) and contained two efficient siRNA sequences targeting ZEB1 (siZEB1#1, 5 0 -GGCAAGTGTTGGAGAATAA-3 0 and siZEB1#2,  (Guangzhou, China). The shRNA sequences were the same as the siRNA sequences. The assay was performed as previously described. 32

| RNA isolation and qRT-PCR
Total RNA was isolated from cells using an E.Z.N.Aâ HP total RNA kit (OMEGA, Norcross, GA, USA) and reversely transcribed into cDNA using a RevertAid first strand cDNA synthesis Kit  Table SI. Relative mRNA levels of each gene were analysed in each sample using the 2 ÀDDCt method against GAPDH mRNA.

| Western blot
Protein extraction and Western blot were performed according to a previous study 29

| Tumour cell three-dimensional culture
This assay was performed to assess the capacity of tumour cells to form VM as described previously. 29 Briefly, we first coated 96-well plates with growth factor-reduced Matrigel (BD Biosciences, Bedford, MA) at 50 lL/well. We then seeded tumour cells at a density of 4 9 10 4 cells per well and incubated them at 37°C for 4 hours.
After that, we counted the number of tube-like structures in three randomly selected microscopic fields. The data were expressed as the mean AE SD for data analysis.

| Wound-healing assay
Cells were seeded in a six-well plate and transfected with ZEB1 siRNA or plasmid for 48 hours. When the cells reached approximately 95% confluence, scratch wounds were made across the monolayer cells using a 200 lL pipette tip as described previously. 33 After washed with PBS, the cells were further cultured in a complete growth medium for up to 48 hours, and the wound healing was photographed at various time-points under an inverted microscope (Olympus, Tokyo, Japan) for three randomly selected sites per well.

| Tumour cell invasion assay
Tumour cell invasion capacity was assessed using Transwell cell culture inserts with 8-lm membrane pores that were pre-coated with Matrigel (BD Biosciences, Bedford, MA, USA) and performed as described previously. 14 The experiment was performed in triplicate and repeated at least once.

| Colony formation assay
Tumour cell clonogenic ability was assessed using a colony formation assay as described previously with minor revisions. 34 In brief, PCa cells were transiently transfected with ZEB1 siRNA or plasmid and then seeded in six-well plates at a density of 500 cells per well and cultured for 15 days. Colonies were then fixed in 70% ethanol and stained with 0.5% crystal violet. Colonies with 50 cells or more were counted under an inverted microscope, and the data were expressed as the mean AE SD of three independent experiments.

| In vivo tumour xenograft assay
This study was approved by the Institutional Animal Care and Use Committee (IACUC) of The First Affiliated Hospital, Sun Yat-sen University (Guangzhou, China). Specifically, 12 male 6-week-old BALB/c nude mice were purchased from Nanjing Biomedical Research institute of Nanjing University (Nanjing, China) and maintained in a specific pathogen-free (SPF) "barrier" facility and housed under controlled temperature and humidity and alternating 12-hour light and dark cycles. The mice will receive SPF mouse chow and be allowed to drink sterile water ad libitum. For the assay, we firstly generated a stable ZEB1-silenced PC3 cell subline; the mice were then randomly divided into two groups, that is an shControl group and shZEB1 group and subcutaneously injected with 5 9 10 6 cells in 100 lL volume into the right armpit.
Tumour growth was monitored and recorded every 7 days for 28 days with calliper. The tumour volume was calculated using the following formula: volume = (length [mm] 9 width 2 [mm])/2. Four weeks later, mice were killed and tumour cell xenograft samples were resected and fixed in 10% buffered formalin for further experiments.

| Statistical analysis
All statistical analyses were performed using SPSS 17.0 software (SPSS, Chicago, IL, USA). Depending on data sets, Student's t test, the chi-square test, Fisher's exact test and Spearman correlation analysis were applied to evaluate the significant associations among categorical variables. A P < .05 was considered statistically significant.

| Differential VM formulation and ZEB1 expression in PCa tissues
In this study, we first assessed VM levels in PCa tissue specimens following our previous study. 29 Tissue specimens underwent immunohistochemical CD34/periodic acid-Schiff double staining and haematoxylin and eosin staining to identify VM structures, which are defined as (1). the PAS-positive or PAS-negative loops with red blood cells, (2). they are negative for the endothelial cell marker CD34 immunostaining, and (3). they are surrounded by tumour cells.
However, the blood vessels are formed by the endothelia that are F I G U R E 1 Immunohistochemistry, haematoxylin and eosin staining and immunohistochemical CD34/periodic acid-Schiff double staining. Paraffin-embedded prostate cancer tissue specimens were double stained with periodic acid-Schiff stain, haematoxylin and eosin staining or immunostained. A, Identification of VM with immunohistochemical CD34 and periodic acid-Schiff double staining. Please refer to the method section for the criteria to identify VM; that is, a channel lined by tumour cells without CD34 staining is considered VM. This structure is indicated by a red arrow on the left panel (magnified in inset). The endothelium-dependent vessel (indicated by a black arrow) that indicates positive CD34 staining is presented on the right panel. Note that red blood cells can be observed in the lumens of both VM and endotheliumdependent vessels (magnified in inset). B, Haematoxylin and eosin staining. VM channel was surrounded by tumour cells. C, Immunohistochemistry. ZEB1 is strongly expressed in the nuclei or cytoplasm, and the expression of ZEB1 was higher in the VM-positive sample (left) than the VMnegative samples (right). D, The expression of EMT-related proteins (E-cadherin, vimentin) and the CSC-associated protein CD133 in prostate cancer tissue specimens. E-cadherin is strongly expressed on the cell membrane, and vimentin is primarily located in the extracellular matrix while CD133 is primarily cytoplasmic staining (Original magnification, 9200; scale bar, 20 lm, for insets, 10 lm) positive for CD34 immunostaining ( Figure 1A and B). As summarized in Table 1, VM was detected in 20 (20.8%) of 96 PCa specimens, and the presence of VM was significantly associated with higher Gleason score, TNM stage, and lymph node and distant metastases, but not with the patient's age (Table 1). For example, VM was significantly higher in PCa with a Gleason score of ≥8 (43.3%, 13/30) compared with a Gleason score of ≤7 (10.6%, 7/66). VM was more prevalent in PCa with ≥T3 stage (41.1%, 14/34) than that with ≤T2 stage (9.6%, 6/62).
We then immunostained ZEB1 protein in the duplicated PCa tissue sections and scored high vs low expression of ZEB1 protein in these samples. To compare VM positivity, we divided these tissue samples into two groups. As shown in Figure 1C Table 2). The expression of ZEB1 protein was also associated with higher Gleason score, TNM stage, and lymph node and distant metastases (Table 1). These data indicated that ZEB1 could regulate tumour VM formation, invasion and metastasis.

CSC-related proteins in PCa tissues
We then associated VM formation with tumour cell EMT and cancer stem cell phenotypes in PCa consecutive tissue sections. In Figure 1D, we found that VM-positive specimens were more likely to express a high level of vimentin and CD133 protein but lacked E-cadherin expression. Notably, there was a significant association between the expression of EMT markers (E-cadherin vs vimentin) and the presence of VM (P = .045 and P = .036, respectively; Table 3) in PCa tissues.
Similarly, the presence of VM was also associated with the expression of a CSC marker, CD133 (P = .003).

| ZEB1 association with the expression of EMT/CSC-related proteins in PCa tissues
In Figure 1D, we also found that ZEB1-expressed PCa cells had down-regulated E-cadherin expression, which was an inverse association (r = À0.375; Table 3). However, ZEB1 expression was demonstrated to associate with vimentin and CD133 expression (r = 0.367 and r = 0.482, respectively; Table 3). Taken together, there could be interplay between ZEB1, EMT/CSC and VM formation.

| ZEB1 regulated VM formation and expression of EMT-related and CSC-associated proteins
To assess and confirm the role of ZEB1 in VM formation, we first measured ZEB1 expression in PCa cell lines (PC3, DU-145 and LNCaP). Similar to the result of our previous study, 29 we found that androgen-independent PC3 and DU-145 cells could form typical vessel-like tubes in the three-dimensional culture but that androgendependent LNCaP cells could not (Figure 2A). We also found that LNCaP expressed the lowest level of ZEB1 protein as that by PC3 and DU-145 (Figure 2A). We then knocked down ZEB1 expression in PC3 and DU-145 cells using siRNA, whereas overexpressed ZEB1 expression in LNCaP cells was performed using plasmid. As shown in Figure 2B and C, ZEB1 expression was significantly reduced after ZEB1 siRNA transfection, and the number of tubular structures was also remarkably decreased. However, LNCaP cells unexpectedly failed to form tubular structures after ZEB1 overexpression (

| Reduction of PCa cell migration, invasion and clonogenicity after ZEB1 knockdown
We then evaluated the effect of ZEB1 knockdown on the regulation of tumour cell migration and invasion capacity. The wound-healing assay showed that ZEB1 knockdown significantly reduced PC3 and DU-145 cells migration ( Figure 3A), while the Transwell invasion assay showed that the down-regulation of ZEB1 expression significantly down-regulated tumour cell invasion capacity ( Figure 3B). Similarly, ZEB1 knockdown significantly suppressed the colony formation of PC3 cells by 55.0% and DU-145 cells by 73.3% ( Figure 3C). wound-healing assay demonstrated that ZEB1 overexpression significantly induced LNCaP cell migration ( Figure 4A), while the Transwell invasion assay showed that the up-regulated ZEB1 expression significantly enhanced tumour cell invasion capacity ( Figure 4B). Furthermore, the overexpression of ZEB1 remarkably up-regulated the clonogenic potential of LNCaP cells ( Figure 4C).

| Src signalling mediation of ZEB1-induced VM
formation and gene expression ZEB1 and Src kinase were shown to modulate PCa cell metastatic phenotypes. 35 Hence, we investigated whether Src contributed to ZEB1-dependent VM formation and found that ZEB1 knockdown down-regulated the level of p-Src 527 in both PC3 and DU-145 cell lines but dramatically enhanced the level of p-Src 416 in PC3 and decreased level of p-Src 416 in DU-145 ( Figure 5A). We thus assessed the role of Src signalling using PP2, a specific inhibitor of Src signalling.
As shown in Figure 5B, we observed that PP2 dose dependently reduced the p-Src 527 level but not the p-Src 416 level in PC3 and DU-145 cells. In parallel, the tubular structures gradually disappeared following PP2 treatment in a dose-dependent manner ( Figure 5C). In addition, the expression of vimentin and CD133 was also partially reduced, whereas E-cadherin expression was increased after treatment of these PCa cell lines with 10 lmol/L PP2 ( Figure 5D). The data indicated that phosphorylation at Tyr-527 of Src signalling was required for VM formation and gene expression. Hence, we next transiently transfected the Src plasmid into stable ZEB1 knockdown cells and observed that the overexpression of Src restored VM formation ( Figure 6A and B). As presented in Figure 6C, similarly, a reoccurrence of VM behaviour was accompanied with the up-regulation of vimentin and CD133 and down-regulation of E-cadherin. Taken together, these data indicated that Src signalling mediated ZEB1-induced VM formation and gene expression and may act by activating Tyr-527.

| Depletion of ZEB1 restrained tumour growth and VM formation in vivo
To further confirmed the effect of ZEB1 on PCa VM, shControl or shZEB1 PC3 cells were subcutaneously injecting into nude mice. We observed that average tumour volume was significantly decreased in shZEB1 group in comparison with the shControl (P < .05, Figure 7A and B). Moreover, tumour xenografts of the shZEB1 group were growing more slowly than those of the shControl group (P < .05, Figure 7C). Immunohistochemical analysis of tumour cell xenografts with the CD34/PAS double staining showed that VM was more prevalent in the shControl (4/6) than that in the shZEB1 group (0/6, P < .05, Figure 7D). Compared with the corresponding control, expression of vimentin and CD133 was also decreased, whereas E-cadherin expression was increased in the shZEB1 group ( Figure 7E). Notably, VM is a known novel vascular network pattern that is formed by tumour cells but not endothelial cells in various cancers, including breast cancer, glioma and PCa. 25,29,36,37 The VM structure was firstly reported by Maniotis et al, 25,29,36,37 40 In the current study, we further confirmed our previous data showing that VM was significantly associated with higher Gleason score, advanced TNM stages and tumour metastasis. 29 We then explored the molecular mechanism underlying VM formation by investigating the association between ZEB1 expression and VM. As a transcription factor, ZEB1 is a well-known EMT inducer that plays a vital role in tumour initiation, tumour cell plasticity and the acquisition of stemness. 12 44 The tumour cell EMT refers to cancer cells losing epithelial features but acquiring a mesenchymal phenotype, which is also known as a phenomenon of cell plasticity. 45 50,51 Thus, further study is needed to address this discrepancy.
These data may indirectly indicate that the phosphorylation of the p-Src 527 site was able to functionally activate Src. Furthermore, Src was not only required to maintain cancer stem cell properties but also participated in the pathway that controls the EMT. 52,53 Unsurprisingly, the expression of vimentin and CD133 was decreased while the expression of E-cadherin was increased after the inhibition of Src. Taken together, our data identify the role of Src signalling in ZEB1-dependent EMT and CSC properties as well as their role in VM formation.

| CONCLUSIONS
Our current study was the first to reveal that ZEB1 played an important role in PCa VM formation in vivo and in vitro. Mechanistically, Src signalling mediated the effects of ZEB1 in PCa cells. Thus, this study provided a novel insight into the molecular mechanism of VM formation and may be used as novel therapeutic targets in controlling VM-positive PCa.

ACKNOWLEDG EMENTS
This work was supported in part by grants from the National Natural

CONFLI CT OF INTERESTS
All authors declared there was no conflict of interests involved in this study.

S U P P O R T I N G I N F O R M A T I O N
Additional Supporting Information may be found online in the supporting information tab for this article.