TP53INP1 inhibits hypoxia‐induced vasculogenic mimicry formation via the ROS/snail signalling axis in breast cancer

Abstract Tumour protein p53‐inducible nuclear protein 1 (TP53INP1) is a tumour suppressor associated with malignant tumour metastasis. Vasculogenic mimicry (VM) is a new tumour vascular supply pattern that significantly influences tumour metastasis and contributes to a poor prognosis. However, the molecular mechanism of the relationship between TP53INP1 and breast cancer VM formation is unknown. Here, we explored the underlying mechanism by which TP53INP1 regulates VM formation in vitro and in vivo. High TP53INP1 expression was not only negatively correlated with a poor prognosis but also had a negative relationship with VE‐cadherin, HIF‐1α and Snail expression. TP53INP1 overexpression inhibited breast cancer invasion, migration, epithelial‐mesenchymal transition (EMT) and VM formation; conversely, TP53INP1 down‐regulation promoted these processes in vitro by functional experiments and Western blot analysis. We established a hypoxia model induced by CoCl2 and assessed the effects of TP53INP1 on hypoxia‐induced EMT and VM formation. In addition, we confirmed that a reactive oxygen species (ROS)‐mediated signalling pathway participated in TP53INP1‐mediated VM formation. Together, our results show that TP53INP1 inhibits hypoxia‐induced EMT and VM formation via the ROS/GSK‐3β/Snail pathway in breast cancer, which offers new insights into breast cancer clinical therapy.

nism of the relationship between TP53INP1 and breast cancer VM formation is unknown. Here, we explored the underlying mechanism by which TP53INP1 regulates VM formation in vitro and in vivo. High TP53INP1 expression was not only negatively correlated with a poor prognosis but also had a negative relationship with VE-cadherin, HIF-1a and Snail expression. TP53INP1 overexpression inhibited breast cancer invasion, migration, epithelial-mesenchymal transition (EMT) and VM formation; conversely, TP53INP1 down-regulation promoted these processes in vitro by functional experiments and Western blot analysis. We established a hypoxia model induced by CoCl 2 and assessed the effects of TP53INP1 on hypoxia-induced EMT and VM formation. In addition, we confirmed that a reactive oxygen species (ROS)-mediated signalling pathway participated in TP53INP1-mediated VM formation. Together, our results show that TP53INP1 inhibits hypoxia-induced EMT and VM formation via the ROS/GSK-3b/Snail pathway in breast cancer, which offers new insights into breast cancer clinical therapy. Vasculogenic mimicry (VM), a newly discovered blood supply model, is a vessel structure made of cancer cells and lacking endothelial cells. 4 A hypoxic microenvironment not only leads to VM formation but also accelerates tumour metastasis, increasing the risk of resistance to chemotherapy in breast cancer. 5,6 The expression of hypoxia-inducible factor-1a (HIF-1a) has been reported to be associated with VM in many cancers types, including liver cancer, 7 ovarian cancer, 8 colorectal cancer 9 and melanoma. 10 These observations suggest a new idea for inhibiting VM formation in hypoxia.
Epithelial-mesenchymal transition (EMT) has also been confirmed to be associated with breast cancer cell invasion and is a key step for VM formation. 11 Research has shown that HIF-1a can influence EMT-related gene expression, resulting in a cell phenotype change. 12 Furthermore, reactive oxygen species (ROS) accumulated in the mitochondria could activate the EMT transcription factor Snail to promote cancer progression and participate in the regulation of HIF-1a transcriptional activity. 13 We speculated that hypoxia-inducible EMT may be indispensable in VM formation in breast cancer.
Tumour protein 53-induced nuclear protein 1 (TP53INP1), also called stress-induced protein (SIP), is a tumour suppressor gene located on chromosome 8q22. 14,15 TP53INP1 is a major regulator of p53 in response to oxidative stress, including hypoxia. 16 TP53INP1deficient cells generate more intracellular ROS. 17 TP53INP1 participates in cancer progression in p53-independent and p53-dependent processes. 14,18 As a target gene of p53, TP53INP1 regulates p73 activity by binding to p53-responsive elements. [19][20][21] Moreover, it has been reported that TP53INP1 can modulate p21, 22 HIPK2, 23 SPARC 24 and TGF-b, 25 suggesting that TP53INP1 plays a critical role in cancer progression. TP53INP1 also negatively regulates the metastasis of malignant tumours including cancers of gliocytes, 26 breast, 27 the stomach, 28 liver 29 and pancreas. 30  2.2 | Immunohistochemistry and CD31/PAS double staining Immunohistochemistry was performed on paraffin-embedded sections of 100 breast cancer tissues. Tissues were dewaxed and rehydrated using graded concentrations of alcohol. Tissue sections were then pretreated by microwave retrieval, blocked with goat serum and incubated with primary and secondary antibodies. DAB staining was performed for appropriate durations, and all sections were counterstained with haematoxylin. After CD31 staining was performed, the tissue was incubated with 0.5% periodic acid for 15 minutes and Schiff's stain for 15 minutes for double staining. PBS was used in place of the primary antibodies for all negative controls. The staining intensity was scored based on four classes: 0 (negative), 1 (weak), 2 (medium) and 3 (high).

| Western blotting analysis
Protein was extracted using SDS lysis buffer and transferred to PVDF membranes. After the membranes were blocked with 5% skim milk for 1 hour, they were incubated with primary antibodies overnight at 4°C, which was followed by incubation with secondary antibodies for 2 hours. Bands were visualized using a C-Digit Blot Scanner (Gene Company) and analysed with ImageJ software. b-actin (sc1616-R, 1:1000; Santa Cruz) was used as a protein loading control. The antibodies are listed in Table S1.

| Immunofluorescence staining
Cells in the different groups were seeded onto glass slides. When the cells were grown to 50%-60% confluency, they were washed with PBS and fixed with cold methanol at À20°C. The cells were permeabilized with 0.1% Triton X-100 in PBS for 20 minutes and blocked with 5% FBS in PBS at room temperature for 30 minutes.
The cells were incubated with primary antibodies for 1 hour at 37°C and overnight at 4°C. The cells were then incubated with secondary antibodies for 1 hour and washed with PBS. Nuclear staining with DAPI (Sigma) was then performed. Slides were viewed under a fluorescence microscope (Nikon, Japan).

| Wound-healing assay
Cells were seeded in 6-well plates. When cells reached confluency, a wound was created using a 100-lL sterile pipette tip and photographed (0 hour). The rate of gap closure was measured at different time-points. Each experiment was performed three times.

| Cell invasion assay and cell migration assays
Migration assays were performed with breast cancer cells (1 9 10 5 ) that were added to the upper chamber with serum-free medium, and DMEM with 10% FBS was added to the bottom chamber in 24-well plates. After the cells were incubated for 24 hours, they were fixed with methanol and stained with crystal violet for 20 minutes. Invasion assays were performed as with the migration assays except that the transwells chambers were coated with Matrigel before the cells were seeded in the upper chamber. These cells were counted using an inverted light microscope (Nikon). Each experiment was performed three times.

| Three-dimensional (3-D) cultures
For 3D culture, 96-well plates were coated with Matrigel (BD, USA) on ice, and the gel was incubated for 1 hour at 37°C. A breast cancer cell suspension (10 5 cells/well) with or without CoCl 2 (150 lmol/L) was added into the plates when the gel solidified, and the plates were placed into a 37°C 5% CO 2 incubator for 24 hours.
Channel-like structure was selected in random fields and filmed under phase contrast microscopy (1009); each group experiment was performed at least three times.

| ROS measurement
Breast cancer cells were seeded at a density of 2 9 10 5 cells/well in a 6-well plate and were treated with drugs for 48 hours at 37°C in 5% CO 2 . DCFH-DA (20 lmol/L; Sigma D6883) was diluted in culture medium and incubated for 30 minutes in a humidified atmosphere.
Cells were washed three times with PBS, suspended in flow tubes, and then analysed with a FACS Accuri C6 (BD Biosciences America).
ROS generation was measured relative to the control, and 1 9 10 5 cell events were analysed for each group.

| Reverse transcriptase polymerase chain reaction(RT-PCR)
Total RNA from the adhesive breast cancer cells in a 6-well plate was extracted with TRIzol Reagent according to manufacturer's protocol (TIANGEN, China) and was reverse transcribed into cDNA using the RT-PCR Kit (TIANGEN, China). RT-PCR was performed with a gradient thermal cycler (Gene Company Limited, HK). The PCR conditions were 95°C for 30 seconds followed by 40 cycles of 95°C for 30 seconds, 60°C for 1 minute and 72°C for 30 seconds following. PCR products were electrophoretically on a Sepharose gel, visualized with a gel document system (Syngene, UK) and analysed by ImageJ software. The sequences of primer are listed in Table S2.

| Statistical analysis
All experiments were repeated independently at least three times.
Values are presented as the means AE SD. Analyses were performed with SPSS 22.0 (SPSS Inc, Chicago, IL, USA) statistical analysis software. Student's t test was used to determine differences between two groups. Measurement data among three or more groups were compared by ANOVA. The survival of each group was evaluated by  Figure 1A and S1). We found that 73 of 100 breast cancer samples had accompanying pericarcinous tissues; of these, the high expression of TP53INP1 was 49.3% (36/73) in breast cancer tissues and 65.8% (48/73) in matched adjacent tissues. Thus, TP53INP1 expression was lower in breast cancer tissues than in pericarcinous tissues (P = .045; Table 2). Studying further the correlation between TP53INP1 expression and clinical prognosis, we found that TNM stage exhibited a significant difference between the high TP53INP1 expression group and the low expression group; low TP53INP1 expression was found in 13 of 21 TNM stage III/IV samples (P = .005). Similarly, low TP53INP1 expression was positively correlated with lymphatic metastasis (P = .015) and triple-negative status (P = .012; Table 1). Meanwhile, we observed that TP53INP1negative tumour cells had relationship with VM formation. Only 10 samples with CD31-negative staining and PAS-positive staining showed VM tubes, but 54 samples in the TP53INP1 overexpression group showed no VM tubes (P = .009; Table 1), indicating that the expression of TP53INP1 was negatively correlated with VM. In addition, Kaplan-Meier survival analysis showed that low TP53INP1 expression correlated with poor overall survival (P = .044), and high VM expression was correlated with poor overall survival (P = .013; Figure 1C). The group with a combination of TP53INP1 low expression and VM positivity had a worse prognosis than the other groups (P = .001; Figure S2A). These results suggest that TP53INP1 expression may be associated with VM in breast cancer.  Figure S2C). In addition, we also observed a correlation of TP53INP1 expression with the expression of HIF1-a in breast cancer by Pearson analysis (r = À.0293, P = .003; Figure S2D). These data showed that TP53INP1 was negatively correlated with VE-cadherin, Snail and HIF-1a expression (Table 3). Together, these results indicated that TP53INP1 may be correlated with hypoxia-induced VM.  showed high expression of VE-cadherin, which is consistent with the above results ( Figure 2E). These data demonstrated that TP53INP1

| TP53INP1 inhibits breast cancer cell VM formation in vitro
inhibited VM formation in breast cancer cells.

| TP53INP1 attenuates the migration and invasion capacity, and TP53INP1 down-regulation is associated with EMT
We sought to determine whether TP53INP1 overexpression would be sufficient to inhibit EMT in breast cancer cells. We first observed that silencing TP53INP1 in MCF-7 cells resulted in a spindle-like, fibroblastic morphology, whereas there was no obvious morphology change in TP53INP1-overexpressing MDA-MB-231 cells ( Figure 3A). Western blot results showed that the expression of E-cadherin and GSK-3b was up-regulated, whereas vimentin, Snail and phospho-GSK-3b levels were decreased by the overexpression of TP53INP1 ( Figure 3B). The immunofluorescence results were consistent with the Western blotting results (Figure S4). Next, we have been suggested that TP53INP1 could inhibit BCC migration in vitro. The results showed that compared with the control, overexpressing TP53INP1 resulted in fewer migrated cells. However, TP53INP1-silenced MCF-7 cells migrated faster than controls ( Figure 3C). Wound-healing assays also confirmed these results; TP53INP1 decreased the healing rate of the wound ( Figure 3D). In general, these results indicated that TP53INP1 not only attenuates breast cancer cell migration and invasion but also inhibits EMT, which suggests that TP53INP1 is related to VM formation.

| Inhibiting TP53INP1 promotes hypoxiainduced EMT and VM formation in breast cancer cells, and its expression decreases hypoxia-induced migration and invasion
To further evaluate the influence of TP53INP1 on hypoxia-induced VM formation in breast cancer, we again assessed migration and invasion. TP53INP1-overexpressing cells showed decreased migration and invasion, whereas shTP53INP1 MCF-7 cells migrated more rapidly than control cells in a hypoxic model ( Figure 4A). Next, we found that TP53INP1-overexpressing breast cancer cells formed very few tube-like structures in the hypoxic environment, whereas the control group formed more VM channels. However, TP53INP1 knock-down facilitated the formation of VM channels compared with the control with CoCl 2 treatment ( Figure 4B). These results were consistent with those in normoxia. We also revalidated that breast cancer cells exposed to hypoxia had increased ability to form channels and increased invasive capacity. As shown by the results of Western blotting assays, we found the up-regulation of E-cadherin and GSK-3b, and the down-regulation of phospho-GSK-3b, vimentin and Snail in CoCl 2 -treated TP53INP1-overexpressing cells compared to control cells. The expression of VE-cadherin and HIF-1a was also reduced. However, MCF7-shTP53INP1 cells had opposite results, which promoted EMT-related protein expression. Meanwhile, we also verified that EMT and VM-related protein levels in breast cancer cells increased in a hypoxic environment compared with those in a normoxic environment ( Figure 4C). To further verify that HIF-1a inhibits TP53INP1 expression, we performed a rescue experiment in which HIF-1a was down-regulated in TP53INP1-overexpressing cells and TP53INP1-silenced cells. Cell with shHIF-1a treatment showed increased TP53INP1 expression compared to the control group. The protein levels of VE-cadherin and MMP2 decreased in TP53INP1overexpressing cells independent of shHIF-1a treatment. Cells with TP53INP1 knock-down that were treated with shHIF-1a or CoCl 2 exhibited increased VE-cadherin and MMP2 protein levels ( Figure S5).
These results suggested that TP53INP1 inhibits hypoxia-induced downstream effectors that promote EMT and VM formation in a hypoxic microenvironment, and TP53INP1 may be mediated by HIF-1a in breast cancer cells.

| A ROS-mediated signalling pathway participates in the inhibition by TP53INP1 of hypoxia-induced VM formation and EMT in vitro
To confirm that TP53INP1 inhibits EMT and VM formation by mediating the activation of ROS signalling in response to hypoxia, we first and CoCl 2, and NAC was confirmed to inhibit the increase in the CoCl 2 -induced EMT and VM-related protein levels ( Figure S6). These data indicated that the role of ROS could be affected by TP53INP1, which promotes EMT and VM formation in breast cancer cells.

| TP53INP1 inhibits breast cancer growth, invasion and VM formation in the TA2 mouse model
To further clarify the relationship between TP53INP1 and VM formation in breast cancer, we explored the effect of TP53INP1 plasmid infection in vivo. First, we selected two of five spontaneous TA2 breast cancers; the TP53INP1 expression in these lines showed obvious differences by Western blotting. The group that was consistent with the MDA-MB-231 cell line was considered as the high-metastasis group and was infected with the TP53INP1-overexpressing plasmid; conversely, the other group was considered the low-metastasis group and was infected with the shTP53INP1 plasmid ( Figure 6A). Ten days after injecting plasmid, the tumour sizes were obviously smaller in the TP53INP1-overexpressing group than in the control group. In contrast, after injection with the shTP53INP1 plasmid, the tumour growth in TA2 mice was greater than that in the control mice (P < .05, Figure 6B). In addition, based on the results of endomucin/PAS double-staining, we confirmed that VM channel formation decreased in the TP53INP1-overexpressing mice. However, more VM channel formation was observed in the TP53INP1-silenced group than in the control group ( Figure 6C,D).

| DISCUSSION
TP53INP1 is a p53-driven oxidative stress response protein that is induced in many different stress situations, such as oxidative stress, radiation and inflammation. 31,32 TP53INP1 was shown to be associated with homeodomain-interacting protein kinase-2 (HIPK2) to mediate the antioxidant activity of p53. 16 Recent studies have shown that EMT, in which the cell-cell junction is lost and cells lose the epithelial phenotype and acquire mesenchymal features, is a critical step in the process of cancer metastasis. 42 Additionally, EMT plays a crucial role in VM formation. 11 In our study, we tested whether EMT participates in the inhibition of VM formation by TP53INP1. We found that EMT protein marker levels changed in these TP53INP1-transfected cell lines HIF-1a is activated by the chemical inducer CoCl 2 , which forces cancer cells to adapt to a hypoxic environment, speeding up VM formation and EMT. 10 In our CoCl 2 -induced hypoxia cell model, we confirmed that overexpressing TP53INP1 hindered breast cancer invasion and the formation of VM vessels; hypoxia-induced breast cancer cells overexpressing TP53INP1 had decreased expression of VM and EMT-related proteins. We also found that the suppression of HIF-1a causes a reduction in VE-cadherin and MMP2 in hypoxiainduced overexpressing-TP53INP1 cells. Cells cotransfected with shTP53INP1 and shHIF-1a also exhibited decreased levels of VEcadherin and MMP2 in hypoxic conditions, suggesting that TP53INP1 is needed for hypoxia-induced VM formation. We speculated that HIF-1a acts on one gene locus of the TP53INP1 promoter.
However, the relationship between TP53INP1 and HIF-1a must be studied further.
Accordingly, we needed to determine which pathway was mediated by TP53INP1 to inhibit hypoxia-induced EMT and VM formation. ROS can promote cancer progression and metastasis by regulating signalling pathways and intracellular metabolism in response to hypoxia. 43,44 EMT is induced through the activation of many transcription factors, such as Snail, MAPK and PI3K. 43,45,46 We found high ROS levels in TP53INP1-deficient cell lines. Moreover, the inhibition of ROS generation by NAC inhibited VM channel formation and VM-related protein expression. Glycogen synthase kinase 3b (GSK3b), a Ser/Thr protein kinase, is an important node; restricting its activation in epithelial cells can limit cancer progression and metastasis. 47 Research has suggested that oxidative stress activates GSK3b, which is then involved in inducing the activation of Snail in a hypoxic environment, and the phosphorylation of GSK-3b promotes the stabilization of Snail. 48 The levels of GSK-3b and phosphorylated GSK-3b were measured to evaluate whether GSK-3b is involved in VM formation and EMT via the ROS-mediated pathway.
We concluded that GSK-3b likely participates in the process. Therefore, TP53INP1 can be reasonably assumed to suppress hypoxiainduced breast cancer EMT and VM formation via the GSK-3b/Snail pathway. However, we did not rule out other pathways; additional detailed studies must be performed.
In conclusion, we identified a novel function of TP53INP1 that is associated with EMT and VM formation in breast cancer. TP53INP1 inhibits hypoxia-induced EMT and VM formation via the ROS/GSK-3b/Snail pathway in breast cancer, especially in triple-negative breast cancer. Therefore, targeting TP53INP1 may be a promising antiangiogenesis strategy for treating this disease.

CONFLI CT OF INTEREST
The authors declare no conflict of interest.

AUTHOR CONTRI BUTION
XLZ conceived the study. YW and DFZ designed the experiments.
YW, HZS and DF carried out the experiments and data collection.