Osterix promotes the migration and angiogenesis of breast cancer by upregulation of S100A4 expression

Abstract As a key transcription factor required for bone formation, osterix (OSX) has been reported to be overexpressed in various cancers, however, its roles in breast cancer progression remain poorly understood. In this study, we demonstrated that OSX was highly expressed in metastatic breast cancer cells. Moreover, it could upregulate the expression of S100 calcium binding protein A4 (S100A4) and potentiate breast cancer cell migration and tumor angiogenesis in vitro and in vivo. Importantly, inhibition of S100A4 impaired OSX‐induced cell migration and capillary‐like tube formation. Restored S100A4 expression rescued OSX‐short hairpin RNA‐suppressed cell migration and capillary‐like tube formation. Moreover, the expression levels of OSX and S100A4 correlated significantly in human breast tumors. Our study suggested that OSX acts as an oncogenic driver in cell migration and tumor angiogenesis, and may serve as a potential therapeutic target for human breast cancer treatment.

differentiation and maturation. 4 The expression of OSX had been suggested to be limited to bone tissues; however, recent studies identified that OSX was overexpressed in various cancer tissues, such as osteosarcoma, prostate cancer and breast cancer. [5][6][7] Nevertheless, there have been few studies on the role of OSX in carcinogenesis. OSX expression in osteoblasts is regulated by various factors, including bone morphogenetic protein 2 (BMP2), Msh homeobox 2 (MSX2), myogenic differentiation (MYOD), and distal-less homeobox 5 (DLX5). [8][9][10] However, its downstream signaling remains largely elusive. Although a recent study indicated that OSX could alter the expression profile of several metastasis-associated genes, such as those encoding vascular endothelial growth factor (VEGF), matric metalloproteinase 9 (MMP-9), β-catenin, and E-cadherin in human breast cancer cells, 7 whether the expression of OSX is critical for cancer metastasis is unknown, and the underlying mechanism remains to be defined. S100 calcium binding protein A4 (S100A4) is involved in a variety of physiological functions, such as cell motility, adhesion, proliferation, and metastasis. [11][12][13][14][15] Mammary tumors from S100A4 transgenic mice displayed higher vessel density compared with nontransgenic animals. 16 Vessel density and S100A4 expression correlated positively in primary tumors from patients with breast cancer. 17 S100A4 is a well established marker and mediator of metastatic disease. 18,19 However, the upstream molecular signaling pathway involved in S100A4-mediated metastasis is less well defined.
In the present study, we investigated the role of OSX in the cell migration and tumor angiogenesis in breast cancer. We found OSX is highly expressed in metastatic breast cancer cells. Moreover, we identified that OSX could promote the breast cancer cell migration and tumor angiogenesis by increasing S100A4 expression, suggesting that OSX participates in breast cancer malignancy and may serve as a potential target for breast cancer therapy.  Cell culture   MCF 10A, MDA-MB-231, MCF7, T-47D, MDA-MB-468, HUVEC and   EA.hy926 cells were obtained from the American Type Cell Collection (Manassas, VA, USA). MDA-MB-231, T-47D, and EA.hy926 cells   were cultured in RPMI-1640 medium; and MCF7, MDA-MB-468 and HUVEC cells were grown in DMEM. All culture media were supplemented with 10% FBS and 1% penicillin/streptomycin. MCF 10A cells were grown in DMEM/F12 medium supplemented with 5% horse serum, 20 ng/mL of epidermal growth factor (EGF), 0.5 mg/mL of hydrocortisone, 100 ng/mL of cholera toxin, 10 μg/mL of insulin, and 1% penicillin/streptomycin. All cells were incubated in a humidified atmosphere with 5% CO 2 at 37°C.  Table S1. To knockdown OSX expression, several independent short hairpin RNAs (shRNAs) against the human OSX gene were ligated into vector pGV248-GFP (GeneChem), with a non-targeting control sequence (shNC) serving as the control. Sequences of the shRNAs targeting OSX are shown in Table S2. The constructed plasmids were transiently transfected into breast cancer cells. Quantitative real-time reverse transcription PCR (qRT-PCR) was used to detect OSX expression and to validate the transfection efficiencies.

|
The expression level of OSX was lowest in #1shRNA group. Therefore, #1shRNA was selected as the optimum shRNA for lentivirus packaging.

| Stable transfections
High-titer lentivirus was packaged in HEK 293T cells. The viral particles were collected by centrifugation at 48 hours post-transfection, and applied to MDA-MB-231 cells in the presence of 5 μg/mL polybrene for 48 hours. Cells were selected using puromycin (3 μg/mL) for 2 weeks. Single colonies were screened by limiting dilution. Gene knockdown and overexpression were confirmed by Western blotting.
| 1117 labeling were the same as those previously described. 20 The labeled peptides were applied to an LTQ-Orbitrap instrument (Thermo was then harvested and centrifuged at 500 g for 10 minutes. Cell-free conditioned medium (CM) was used for subsequent experiments.

| Tubule formation assay
The microtubule formation assay was performed in 96-well plates coated with 50 μL of matrigel (BD Biosciences, Bedford, MA, USA). HUVEC and EA.hy926 cells were seeded at 1 × 10 5 cells per well and incubated with CM at 37°C for 24 hours. Five random selected fields of view were captured using a microscope.
Tube lengths were assessed by drawing lines along the tube-like structure and measuring the lengths of the lines in pixels using Image J software.

| Nude mouse xenograft model
Cell aliquots (100 μL) were mixed with matrigel, and the mixture was immediately engrafted into the fourth inguinal mammary fatpad of 6-week old female BALB/c nu/nu mice (n = 8 each group).
Tumor sizes were measured every 3 days from the sixth day post-

| Statistical analysis
Statistical analysis was performed using SPSS 19.0. Results were expressed as the mean ± the standard deviation. Comparisons between two groups were analyzed using a two-sided Student's t test. Correlations between OSX expression and clinicopathological characteristics of the patients with breast cancer were examined using the chi-squared test. The relationship between OSX and S100A4 expression levels was assessed using Spearman correlation analysis. P < 0.05 was considered statistically significant.

| Data availability
All data supporting the findings of this study are available from the corresponding author upon reasonable request.

| OSX promotes breast cancer cell migration and vascular tube formation
Although OSX has been shown to be involved in the regulation of some metastasis-associated genes, its expression in breast cancer cell lines with different metastatic potencies has not been well surveyed.
We initially investigated the expression levels of OSX in a set of breast cancer cell lines with different metastatic features. As shown in Figure Figure 1E and Figure S1C).

| OSX promotes breast cancer angiogenesis in vivo
To confirm the in vitro findings, we evaluated the effect of OSX on angiogenesis in the CAM model. As shown in Figure 2A,  (Table 1 and Table S5). Among them, nine proteins, including ANXA4, 25 EPS8L2, 26 HCCR1, 27 HLA-DPB1, 28 HLA-DRA1, 29 HPRT1, 30 LMP2, 31 S100A4 32 and TOX, 33 have been reported to be involved in cancer progression as shown in Figure 3A. Growing evidence indicates that elevated S100A4 protein levels are associated with the progression and angiogenesis of several malignant tumors, including breast cancer, 17 non-small cell lung cancer, 18 prostate cancer, 19 and colon cancer. 34 Therefore, we hypothesized that OSX might exert its effects on cell migration and angiogenesis by regulating S100A4 expression in breast cancer. To test this hypothesis, the changes in S100A4 expression were further assessed by qRT-PCR and western blotting. OSX KD significantly reduced both the mRNA and protein levels of S100A4, whereas OSX overexpression had the opposite effect ( Figure 3B and Figure S2).
We then evaluated the involvement of S100A4 in OSX-induced cell migration and angiogenesis of breast cancer. Three siRNA oligonucleotides targeting different sites in the mRNA of S100A4 were tested and S100A4-siRNA-1 was selected for further experimental because of its highly efficient knockdown of S100A4 expression ( Figure S3). S100A4 expression was restored by transfection of a construct expressing S100A4 in shOsx cells, while it was depleted using S100A4-siRNA-1 in OE-Osx cells. There was no change in the expression of endogenous OSX ( Figure 3C). Restored expression of S100A4 significantly rescued OSX-shRNA-suppressed cell migration and capillary-like tube formation. In contrast, S100A4 KD impaired OSX-induced cell migration and capillary-like tube formation, as determined by transwell migration and tube formation assays ( Figure 3D,E). These data strongly suggested that OSX-induced cell migration and capillary-like tube formation was partially mediated by S100A4.
Interestingly, the decrease in CD44 and VEGF expression was also abolished after restoring the expression of S100A4 in shOsx cells, and increased CD44 and VEGF levels were downregulated in S100A4 depleted OE-Osx cells ( Figure 3F and Figure S4). There was no change in the expression of β-catenin after restoring S100A4 in shOsx cells or in S100A4-depleted OE-Osx cells (data not shown).
These data suggested that S100A4 induces cell migration and angiogenesis partially via modulation of CD44 and VEGF in breast cancer cells.

| OSX induces migration and angiogenesisrelated genes expression in vivo
To further validate the effects of S100A4 on OSX-induced cell migration and angiogenesis in vivo, OSX KD or overexpressing cells were injected into the fourth mammary fat-pad of BALB/c-nu/nu mice. Tumor volumes were measured every 3 days when they were palpable. Tumors from the shOsx group were significantly smaller than those from control group. However, there was no significant difference in tumor volume between the OE-Osx and control group (data not shown). As expected, the expression of S100A4 was markedly reduced in tumors from the shOsx group, and significantly increased in tumors from the OE-Osx group ( Figure 4A,B). In addition, immunohistochemical staining and western blotting analysis showed that the expression levels of migration-related proteins CD44, β-catenin, the endothelial markers CD31 and CD34, and the vascular/lymphatic marker VEGF were all decreased in tumors from the shOsx group and increased in tumors from the OE-Osx group ( Figure 4C,D). Collectively, these data showed that knockdown of OSX inhibited migration and angiogenesis by downregulating S100A4, CD44, β-catenin, CD31, CD34, and VEGF levels in vivo, whereas overexpression of OSX had the opposite effect.

| OSX expression is positively correlated with S100A4 levels in breast cancer tissues
Using immunohistochemical staining, we examined whether OSX expression is correlated with S100A4 expression in breast cancer samples. As shown in Figure 5A, high expression of OSX was detected in 86 cases, among which 71 cases exhibited high expression of S100A4 (82.6%). Meanwhile, there were 26 cases with low expression of OSX, among which 23 cases showed low expression of S100A4 (88.5%). By contrast, there was no statistically significant correlation between OSX expression and the patients' age, tumor size, or histological sub-type (ER, PR, or HER2 status) (Table S6).
These results strongly indicated that the expression levels of OSX are significantly and positively correlated with those of S100A4 in breast cancer tissues. Nevertheless, in a study based on a mouse osteosarcoma model, Osx was down-regulated and its expression was negatively associated with metastatic potency. 35 This discrepancy may be caused by the context dependence of the specific cell lines used in each study.

| DISCUSSION
A mass spectrometry-based analysis identified 19 differentially expressed proteins in shOsx and OE-Osx cells, compared with their respective controls. We decided to focus on S100A4 because it is a mediator of tumor cell migration and angiogenesis. [12][13][14] The results of in vivo xenograft mouse experiments also supported our in vitro observation that OSX regulates S100A4. Importantly, restored expression of S100A4 significantly rescued OSX-shRNA-suppressed cell migration and capillary-like tube formation, while S100A4 KD inhibited OSX-induced cell migration and capillary-like tube formation. These data strongly suggested that OSX-induced cell migration and capillary-like tube formation were partially mediated by S100A4.
Moreover, OSX expression was significantly positively correlated with the level of S100A4 in breast cancer tissues. An increase in S100A4 protein expression has been correlated with a worse prognosis for patients with different types of cancer including breast, colon, gastric, lung, hepatocellular, and pancreatic cancer, 34,[36][37][38][39] which suggested that OSX could be used as a target gene to improve cancer prognosis.
As shown above, S100A4 was differentially expressed at the mRNA and protein levels between OSX KD and OSX overexpressing cells, compared with their respective control cells. The differential expression of S100A4 might occur by three possible mechanisms: differential mRNA transcription, mRNA stability, or protein stability of S100A4. OSX belongs to the specificity protein (SP) family that is presumed to function by binding directly to DNA promoter elements via an SP1-like DNA-binding domain. 4 The human S100A4 gene contains four exons, two of which are non-coding at the 5′ UTR position 14 ; the transcription of the S100A4 gene is controlled by both positive and negative regulatory elements located within the first intron, which is bound by several transcript factors. 40,41 Bioinformatic analyses indicated that the core region of the S100A4 promoter contains five potential OSX-binding sites (data not shown).
We speculated that OSX probably regulates the transcription of S100A4 by binding to its promoter. Nevertheless, further investigation is required to uncover the detailed mechanisms by which OSX regulates S100A4 expression.
A previous study revealed that OSX was associated with the expression of a number of metastasis-associated genes such as VEGF, MMP-9, β-catenin, and E-cadherin. 7,42 VEGF is one of the most potent endothelial cell mitogens and plays a crucial role in tumor growth, angiogenesis and metastasis. [43][44][45] It has been reported that VEGF is the direct target gene of OSX in osteoblasts. 42 In our study, we found that OSX positively regulated VEGF expression in breast cancer cells. More interestingly, the decrease in VEGF expression  Data represent the means ± SD of three independent experiments. CM, conditioned medium. **P < 0.01. ****P < 0.0001 was abolished after restoring the expression of S100A4 in shOsx cells, and increased VEGF was downregulated in S100A4 depleted OE-Osx cells. It has been reported that S100A4 alters the neovascularization ability in tumors by regulating VEGF. 46 A significant relationship between S100A4 and VEGF expression was also demonstrated in clear renal cell carcinoma, gastric carcinoma and pancreatic cancer. [47][48][49] Therefore, OSX induces angiogenesis at least in part through the S100A4-VEGF pathway ( Figure 5B). In our study, we also found OSX positively regulated CD44 expression, and the decrease in CD44 expression was abolished after restoring the expression of S100A4 in shOsx cells while increased CD44 expression was downregulated in S100A4 depleted OE-Osx cells. CD44 participates in many cellular processes, including the regulation of cell survival, migration, and adhesion through the binding of its major ligand, hyaluronic acid. 50 Aberrant overexpression of CD44 correlates with the metastatic potential of several malignant tumors, such as prostate cancer, breast tumors and chondrosarcoma .22,51,52 S100A4 induced the re-distribution of CD44 and enhanced the cell surface expression of CD44, thereby inhibiting cell-cell and cellmatrix adhesion in B16 murine melanoma cells. 53 In addition, transfection with S100A4 siRNA significantly reduced the expression of CD44 in osteosarcoma cells. 54 These data suggested that OSX might induce migration partly through the S100A4-CD44 pathway (Figure 5B).

F I G U R E 4 Effects of osterix (OSX) on the expression of migration and angiogenesis-related genes in vivo. (A)
Cells were mixed with matrigel and engrafted into the fourth inguinal mammary fat-pads of nude mice. Six weeks later, the tumor samples were harvested. Immunohistochemical analysis was used to detect the expression levels of OSX and S100A4 in nude mice tumors. (B) Relative mRNA and proteins expression levels of OSX and S100A4 were determined by qRT-PCR and western blotting analysis, respectively, in nude mice tumors. (C) Immunohistochemical staining analysis was used to detect the expression levels of CD44, β-catenin, CD31, CD34 and vascular endothelial growth factor (VEGF) in nude mice tumors. (D) The protein expression levels of CD44, βcatenin, CD31, CD34 and VEGF were analyzed by western blotting analysis in protein samples from nude mice tumors. *P < 0.05. **P < 0.01. ****P < 0.0001 F I G U R E 5 Correlations between osterix (OSX) and S100A4 expression levels in breast cancer tissues. (A) Representative immunohistochemical peroxidase staining for OSX and S100A4 in breast cancer tissues (upper) and statistic data of OSX and S100A4 expression levels (lower). (B) Schematic diagram representing the role of OSX in cell migration and tumor angiogenesis in breast cancer In conclusion, this study revealed that OSX could potentiate breast cancer cell migration and tumor angiogenesis by up-regulating S100A4 expression in vitro and in vivo. Augmented CD44 and VEGF in breast cancer cells are associated with OSX-mediated cell migration and angiogenesis. Overall, our study suggested that OSX participates in breast cancer malignancy and might serve as a potential target for breast cancer therapy.