SRGN-TGFβ2 regulatory loop confers invasion and metastasis in triple-negative breast cancer

Patients with triple-negative breast cancers (TNBC) are at a high risk for a recurrent or metastatic disease, and the molecular mechanisms associated with this risk are unclear. Proteoglycan serglycin (SRGN) proteins are involved in tumor metastasis, but their role in TNBC has not yet been elucidated. This study investigates the SRGN gene expression and how it regulates TGFβ2 and the downstream signaling of TGFβ2 in TNBC cells and tissues. Our results show that SRGN mRNA and protein expression levels were significantly higher in TNBC cell lines and tumor tissues than that in non-TNBC cells and tissues. We inhibited SRGN expression and protein secretion using shRNA and we observed this inhibited the invasive motility of TNBC cancer cells in vitro and metastasis of TNBC cancer cells in vivo. SRGN protein treatment increased the expression and secretion of transforming growth factor-β2 (TGFβ2) by activating CD44/CREB1 signaling and promoted epithelial-to-mesenchymal transition in TNBC cells. Moreover, TGFβ2 treatment increased the mRNA and protein expression of the SRGN gene by activating Smad3 to target the SRGN relative promoter domain in TNBC cells. Our findings demonstrate that SRGN interacts with TGFβ2 which regulates TNBC metastasis via the autocrine and paracrine routes. SRGN could serve as a potential target for development of agents or therapeutics for the TNBC.


INTRODUCTION
Breast cancer is one of the most common malignancies for women. Although the treatment of breast cancer has greatly improved over the past few decades, there are still around 50 million people worldwide that die from breast cancer every year. The persistence and prevalence of breast cancer may be due to its high genetic heterogeneity, since patients who exhibit the same clinical and histological tumor morphology may have divergent molecular genetic characteristics. These divergences can affect the treatment and prognosis of breast cancer, 1-3 however, the molecular events involved in the treatment resistance have not yet been fully elucidated.
Breast cancer is commonly divided into four types: Luminal A (ER + , PR + , HER2 − and low Ki67 expression), Luminal B (ER + , PR + , HER2 − /+ , high Ki67 or any level expression), HER2 overexpression (ER − , PR − , HER2 + ), or basal-like. Approximately 80% of basal-like breast cancer are ER − , PR − and HER2 − , which are called TNBC − (triple-negative breast cancer). TNBC is a very heterogeneous group of cancers that is aggressive, has a high risk of relapse, has poor prognosis, and often occurs in women under 50 years of age. Due to lack of endocrine therapy and HER2 targeted therapy, the main clinical treatments of TNBC rely on chemotherapy which causes therapeutic resistance. 4 Therefore, it is important to determine the molecular mechanism underpinning metastasis and recurrence of TNBC.
Serglycin (SRGN) is a low molecular weight glycoprotein involving in breast cancer metastasis. SRGN can be secreted from cells and integrated into the extracellular matrix. SRGN consists of a core protein with 158 amino acids that is attached with various mucopolysaccharides (GAGs). The core protein forms three function regions: signal peptide (amino acid residues 1-27), N-terminal (amino acid residues 28-76), and C-terminal (amino acids 77-158). The C-terminal of SRGN contains multiple serine and glycine repeat regions which bind to the GAGs. 5,6 Studies have shown that SRGN is mainly expressed in blood cells, endothelial cells, tumor cells, and embryonic stem cells. SRGN has an important role in the storage and secretion of a variety of proteases, chemokines and cytokine. 7,8 The effect of SRGN in cancer was first found as a marker to distinguish myeloid leukemia from lymphoid leukemia. 9 In multiple myeloma, high expression of SRGN can inhibit the complement activity, which can help tumor cells to escape from immune surveillance. 10 Elevated SRGN expression can be used as a prognostic indicator of liver cancer. 11 Recent studies have shown that SRGN can induce an epithelial-mesenchymal transition (EMT) in nasopharyngeal carcinoma cells and aggressive breast cancer cells, promoting tumor cell invasion and metastasis. 12,13 Lung metastasis of breast cancer is significantly inhibited in SRGN-deficient mice. 14 It is suggested that SRGN may be highly associated with breast cancer metastasis, possibly through regulating EMT activation. In this study, we investigated the expression and biological functions of SRGN gene in TNBC breast cancer cells and observed which signal pathways were affected.

SRGN mRNA and protein expression increases in breast cancer cells and tissues
SRGN has been demonstrated to induce proliferation, migration and invasion in breast cancer cells, 13 but its expression in human breast cancer tissues and differential expression in different cell lines has not been reported. We therefore monitored SRGN mRNA expression using real-time PCR and measured protein expression by western blot in six breast cancer (BC) cell lines. SRGN mRNA ( Figure 1a) and protein (Figure 1b) expression levels were significantly higher in the TNBC cells (MDA-MB-231 and BT549 cell), than other subtypes of BC cells (MCF7, T47D, BT474 and SKBR3 cells) (P o 0.05, P o0.01 vs MCF cells). We also measured the SRGN protein level in the supernatant of cultured cells by ELISA and found that the supernatant SRGN level in TNBC was significantly higher than other types of BC cells (P o 0.05, P o 0.01 vs MCF cells, Figure 1c). We further measured mRNA expression of the SRGN gene using RT-PCR in tissues from 106 cases of TNBC and 320 cases of other BC types (non-TNBC) and we found that TNBC tumors contained significantly higher SRGN mRNA levels compared with other BC types (P o0.001, Figure 1d). We searched our observed mRNA expression pattern of the human SRGN gene against the Breast Cancer Gene-Expression Miner v4.0 database (http://bcgenex.centregauducheau.fr) and again found that mRNA expression of SRGN gene was significantly higher in basal-like and TNBC tumors than that in non basal-like and non-TNBC tumors (P o 0.001, Figure 1e).
We wanted to understand the association between SRGN gene expression and breast cancer progression and metastasis, and collected a panel of breast cancer tissue samples and divided them into a positive lymph node metastases group (Node positive) (n = 135) and a negative lymph node metastases group (Node negative) (n = 142). The SRGN mRNA levels were significantly higher in the 'Nodal positive' group than that in the 'Nodal negative' group (P o0.001, Figure 1f). These results suggest that SRGN gene is highly expressed TNBC cells and tissues compared with other BC types and associated with the metastatic phenotype of BC.

SRGN promotes TNBC cells migration, invasion and metastasis
Although one study investigated the biological functions of SRGN in breast cancer cells by overexpression of serglycin, 13 Figure 2D). The Transwell invasion assay showed that significantly more MCF7-SRGN stable cells (MCF7-SRGN) invaded to the lower chamber compared with the MCF7-NC stable cells (P o 0.01, Supplementary Figure 2E,F). These results suggest that SRGN overexpression promotes breast cancer cells invasion and metastasis.

SRGN increases TGFβ2 levels and activated EMT
A previous study demonstrated that SRGN can induce an epithelial-mesenchymal transition (EMT) in nasopharyngeal carcinoma cells 12 and the cytokine transforming growth factor β (TGFβ) is known to trigger the EMT process. 15 In addition, the E-cadherin to N-cadherin switch is a biomarker for EMT. The increased vimentin expression is associated with a migratory phenotype, and up-regulated fibronectin expression is observed in EMT. 16 We wanted to know whether SRGN activates EMT in BC cells through regulating TGFβ2 gene expression. Western blots showed that E-cadherin was significantly increased in MDA-MB-231-shSRGN2# cells stably expressing shRNA2# of SRGN gene (shSRGN2#), but was significantly decreased in MCF7 cells overexpressing SRGN gene compared with their control groups, respectively ( Figure 3a). In contrast, the protein expression of vimentin, N-cadherin, and fibronectin was significantly decreased in MDA-MB-231-shSRGN2# stable cells (shSRGN2#), but was significantly increased in MCF7-SRGN stable cells overexpressing SRGN gene (SRGN) compared with their own control groups (Figure 3a). These findings suggest that SRGN activates EMT.
TGFβ2 can promote the expression of CREB1 in tumor cells 23 and CD44 is a classic marker of breast cancer stem cells. 17 We wanted to know whether SRGN can regulate CD44 TGFβ2 protein expression and CREB1 phosphorylation. We therefore tested CERB1 protein expression and phosphorylation in MDA-MB-231 cells to observe the effects of exogenous SRGN on CD44 and TGFβ2 protein expression. The exogenouse SRGN was the supernatant collected from the cultured MCF7-SRGN stable cells. The CD44 gene expression was silenced by a siRNA of the CD44 gene (siRCD44). MDA-MB-231 cells were treated with the supernatant (sup) collected from the cultured MCF7-SRGN stable cells, with or without 50 and 100 nM siRCD44 for 48 h. Western blots showed that the supernatant from cultured MCF7-SRGN stable cells significantly increased CD44 and TGFβ2 protein SRGN  We want to know how the TGFbeta regulates SRGN gene expression in TNBC cells and performed a bioinformatic assay, which predicted that SMAD3 may have five binding sites in the SRGN promoter. We therefore tested which binding site has a key role in regulating SRGN gene expression. Western blots showed that TGFβ2-stimulated SRGN protein expression correlated with the increases in Smad3 protein phosphorylation in MDA-MB-231 cells (Figure 4e). The bioinformatic analysis predicted five potential SMAD2/3 binding sites (GTCT/AGAC) in the SRGN promoter region ( −2000 to +200). To validate it, five different 5'deletion sequences containing different Smad3 binding sites were cloned into the promoter-luciferase reporter vector pGL4. The constructs were transfected into 293 T cells. Results from this deletion analysis showed that the Smad3 binding sites 1, 4, 5 were essential for the promoter activity upon treatment with 10 ng/ml TGFβ2 (Figure 4f). The pLuc vector containing the SRGN promoter with mutated 1, 4 or 5-binding site of Smad3 (Samd3-1-mut, Samd3-4-mut, Samd3-5-mut) or wild-type Samd3 binding site (Samd3-1-wt, Samd3-4-wt, Samd3-5-wt) was co-transfected with the pRL-TK Renilla luciferase vector with or without the pCMV-SMAD3 vector (pCMV-Samd3) or control vector (pCMV-control) into 293 T cells. The luciferase activity assay showed that pCMV-SMAD3 vector transfection significantly increased luciferase activity in cells transfected with pLuc vector containing wild-type Samd3 binding site compared with transfection with pCMV-control vector (P o0.05) (Figure 4g). Also, in cells transfected with pCMV-Smad3 vector, the luciferase activity was significantly decreased when co-transfected with pLuc vector containing wild-type Samd3 binding site compared with co-transfected with pLuc vector containing mutated Samd3 binding site (Figure 4g). To further verify the binding sites of SMAD2/3 in SRGN promoter region, we performed CHIP-PCR, which also confirmed that SMAD2/3 could bind to site 1,4,5 in SRGN promoter region (Figure 4h (Figure 5i), in BC patients without lymph node metastasis (Nodal Free) (P o0.01) (Figure 5j), and TNBC patients without lymph node metastasis (Nodal Free) (P o 0.05) (Figure 5k), respectively. These data suggest that SRGN up-regulation correlates with TGFβ2 in triple-negative breast cancers which may lead to eventually metastasis.

DISCUSSION
TNBC is difficult to treat due to its high metastasis rate, its susceptibility to relapse, and the lack of information regarding its specific targets. This study found that glycoprotein SRGN is highly expressed in the TNBC tumor cells. SRGN enhances TGFβ2 expression and matured TGFβ2 extracellular secretion and subsequently activates EMT through binding CD44 and SRGN interacts with TGFβ2 in TNBC Z Zhang et al phosphorylation of CREB1. At the same time, TGFβ2 is highly expressed in TNBC and regulates SRGN transcriptional expression by binding to TGFβ receptor, which subsequently activates the downstream SMAD2/3 and positively regulates SRGN expression ( Figure 6). Therefore, activation of the SRGN-TGFβ2 loop in TNBC is essential for maintaining the high metastatic potential of TNBC.
SRGN is a small molecule glycoprotein mainly distributing in the cell and cell membrane, but it can also be secreted and integrated into the extracellular matrix. Studies have shown that SRGN promotes cell migration by binding to its receptor in CD44 in blood cells. 18 CD44 is thought to be a classic marker of breast cancer stem cells and has an important role in tumor stem cell adhesion, invasion, metastasis, apoptosis resistance, chemoresistance, and EMT formation. 17 For example, the binding of CD44 to its classical ligand, hyaluronic acid (HA), activates protein kinase C, which in turn activates transcription factor SMAD2/3, and subsequently has an important role in EMT formation. 19,20 Here, we observe that SRGN can promote the expression of TGFβ2 in TNBC cells through binding to its CD44 ligand, which subsequently activates CREB1. TGFβ2 is known activator of EMT which can explain the effect of SRGN in the formation of EMT. In addition, SRGN can also mediate the maturation and secretion of MMP9 precursors. 21 This suggests that SRGN may also promote metastasis and EMT though regulating MMP9 and other metal matrix protease molecules. Moreover, TGFβ family members have multiple disulfide bonds, and one of them is mainly involved in the intermolecular linkage. 22 This explains our finding that immunoprecipitation results in SRGN directly bound to TGFβ2 through the disulfide bond, which can promote the partial secretion and maturation of TGFβ2.
A previous study showed that the 5′-UTR of SRGN contains multiple (−80) ETS and (−70) CRE sites which mainly regulate the expression of SRGN, though different stimulus conditions can regulate different sites. 23 In the present study, we found that TGFβ2 regulates the expression of SRGN mainly through the (−745) SMAD3 site. Some studies have shown that TGFβ2 can promote the expression of CREB1 in tumor cells through activating the TGFβ receptor and form a self-activating loop. 24,25 In addition to our findings on the SRGN-TGFβ2 loop, we find that TGFβ2, but not TGFβ1, may have a crucial role in TNBC. Therefore, we postulate that TGFβ2 may be an ideal target for future targeting therapy. In addition, our study demonstrates that SRGN does not affect cell growth in vitro, but decreases the ability of tumor formation in vivo (Supplementary Figure 1), suggesting that tumor microenvironment has a great impact on SRGN tumorigenesis.
Previous studies demonstrated that SRGN is mainly involved in immune response, such as promoting the TNFα secretion from macrophages, and inhibiting classical and non-classical complement activation pathways. 8 This implies that SRGN may have an important role in tumor immune escape, which may diminish the effect of antibody-dependent complement killing of the current tumor targeted therapy. In addition, TGFβ2 also functions as an important immunosuppressive factor in the formation of tumor immunosuppressive microenvironment. 26 In this study, the SRGN-TGFβ2 loop may be an important mechanism for TNBC tumor cells to escape immune surveillance. This is worthy of further study.
In summary, our finding of a SRGN-TGFβ2-positive feedback loop highlights a new target for the therapy of refractory and high recurrence of TNBC breast cancer.

MATERIALS AND METHODS Reagents
Antibodies for E-cadherin, N-cadherin, vimentin, fibronectin, CREB1, pCREB1, pSMAD2/3, and CD44 were purchased from Cell Signaling Technology (Beverly, MA, USA). Antibodies for TGFβ1, TGFβ2, and β-actin were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). The antibody for SRGN and anti-Rabbit FITC second antibody were purchased from Abcam (Cambridge, MA, USA). The recombinant human TGFβ1 and TGFβ2 protein was obtained from R&D Systems (Minneapolis, MN, USA). The TGFβ receptor I kinase inhibitor SD208 was from Selleckchem

Establishment of stable cell lines
The plenshSRGN-GFP lentiviral vector overexpressing shRNA1 (5′-gaactacttccaggtgaatcc-3′) and shRNA2 (5′-ggaacaggattaccaactagt-3′) to silence  (e) Histological scores of SRGN staining in breast cancer patients with and without lymph node metastasis. **P o0.01 between two groups. (f) Histological score of TGFβ2 staining in breast cancer patients with and without lymph node metastasis. **P o0.01 between two groups. (g) Correlation analysis of the expression of SRGN and TGFβ2 protein in TNBC tissues. (h) Histological scores of SRGN staining in TNBC patients with and without lymph node metastasis. **Po 0.01 between two groups. (i) ELISA assay of serum SRGN concentration in patients with different types of breast cancer. **P o0.01, ***P o0.001 between two groups. (j) ELISA assay of serum SRGN concentration in breast cancer patients with and without lymph node metastasis. **P o0.01 between two groups. (k) ELISA assay of serum SRGN concentration in TNBC patients with and without lymph node metastasis. *Po 0.05 between two groups. Scale bar, 50 μM.
SRGN gene expression, pEZ-SRGN lentiviral vector overexpressing SRGN gene, and control lentiviral vectors (NC) were purchased from Genecopoeia (Guangzhou, China). MDA-MB-231 cells at 70% confluence were transfected with plenshSRGN-GFP lentiviral vector and control vector using BLOCK-iT Lentiviral Pol II miR RNAi system (Invitrogen) by following the user manual. Twenty-four hours after transfection, fresh medium containing 2 μg/ml of puromycin (Invitrogen) was added to cells and refreshed every three days for three weeks. Single colonies were then selected, identified, and continuously cultured. This procedure produced two SRGN shRNA stable expression cell lines and a control stable cell line, which we call MDA-MB-231/shSRGN1#, MDA-MB-231/shSRGN2#, and MDA-MB-231/shNC, respectively. MCF7 cells were transfected with pEZ-SRGN lentiviral vector to overexpress SRGN gene and transfected with pEZ-NC lentiviral to express the blank vector as control as described above. Stable cells were selected using 1.5 μg/ml of puromycin. The produced MCF7 cells stably expressing SRGN were named MCF-SRGN. The control MCF7 stable cell line was called MCF7-NC.

Cell lysis and western blot
Cultured cells were washed twice before lysis with ice-cold 1 × PBS (phosphate-buffered saline). Cells were then lysed using RIPA (radioimmunoprecipitation assay) buffer supplemented with proteases and phosphatases inhibitors (Sigma-Aldrich, Sigma, St Louis, MO, USA). Protein concentration in lysates was determined by a Pierce BCA protein assay kit (Thermo Scientific, Waltham, MA, USA). Equal amount of protein between samples was separated by electrophorese on SDS-PAGE gel and transferred to polyvinylidene difluoride membranes (Merck Millipore, USA). Membranes were incubated with 5% non-fat milk in TBS (Tris-buffered saline) buffer for an hour at room temperature and then incubated with primary antibodies overnight at 4°C. The membranes were washed three times with 1 × TBS buffer before incubation with LICOR 680 nm or 800 nm fluorescent secondary antibodies for one hour. After washing with 1 × TBS buffer, the membranes were scanned on a LICOR Odyssey system. The acquired images were analyzed with Image Studio Version 4.0 software according to manufacturer's instructions.

ELISA (enzyme-linked immunosorbent assay)
The SRGN concentration in the serum of breast cancer patients and the supernatant of serum-free cultured cells was measured for 48 h using SRGN ELISA Kit (CUSABIO, China). The concentrations of TGFβ1 and TGFβ2 in the supernatants of cell culture were measured using Human TGF-β1 Quantikine ELISA Kit (PDB110B) and Human TGF-β2 Quantikine ELISA Kit (PDB250) (R&D Systems), respectively according to the manufacturer's instructions.

In vitro migration and invasion assays
An in vitro scratch assay and Transwell assays were used to evaluate the migration and invasion abilities of the tumor cells, respectively. Briefly, cells were grown at monolayer in 12-well plates (2 × 10 5 ) for overnight. An artificial scratch was created in cells, and cell debris was removed by washing with 1 × PBS. Cell migration was photographed and the width of the wound was measured. Cell invasion was evaluated by a Transwell system. The polycarbonate filters (8-μm pore size) were pre-coated with Matrigel Matrix (BD Biosciences, Franklin Lakes, NJ, USA), and reconstituted at 37°C for 30 min. Cells (1 × 10 5 ) were suspended in 150 μl of serum-free RPMI 1640 medium and added into the upper chamber, while 600 μl of complete medium was added to the lower chamber. The cells that migrated through the matrigel and adhered onto the lower chamber after 24 h of incubation were fixed with 4% paraformaldehyde for 20 min, and then stained with Mayer's hematoxylin (Sigma-Aldrich, USA), and counted under microscope (five fields per chamber).
Immunoprecipitation assays MDA-MB-231 cells were washed twice with ice-cold 1 × PBS, and then lysed with RIPA buffer supplemented with protease inhibitors. Protein concentrations were determined as described above. One milligram of lysates was incubated with protein G-agarose and anti-SRGN or TGFβ2 antibody for 4 h at 4°C. Beads were washed 3 times with immunoprecipitation (IP) buffer (150 mM NaCl, 25 mM Tris-HCl pH 7.5, 0.1% Nonidet P-40), and proteins were eluted with 1.5 × SDS-PAGE sample buffer. Samples were analyzed by Western blot with anti-SRGN or anti-TGFβ2 antibody.

Flow cytometry
Cultured MDA-MB-231 cells were treated with or without different concentrations of SD208 inhibitor for 48 h. The suspended cells (1 × 10 6 ) were stained with SRGN (1:50) antibody. After 30 min incubation on ice, the cells were washed with staining buffer (BD Biosciences) and then incubated with FITC-labeled second antibody for 30 min on ice in the dark. After washing, the cells were resuspended in staining buffer. Data were acquired using a BD CantonII Flow Cytometer (BD Bioscience) and analyzed using FlowJo software (Ashland, OR, USA).
with the pLuc vector and the pRL-TK Renilla luciferase vector with or without the pCMV-SMAD3 vector using Lipofectamine 2000 (Invitrogen). After 48 h, luciferase activity was determined using a Dual-Luciferase Reporter Assay System (Promega). The firefly luciferase activity was calculated as the mean ± s.d. after normalization with the Renilla luciferase activity.

CHIP-PCR
The ChIP assay was performed using the EZ-CHIP chromatin immunoprecipitation kit (Merck Millipore, Germany). Briefly, 1% formaldehyde was added to the cultured cells to cross-link the chromatin proteins to DNA. After incubation for 10 min at room temperature, the cells were washed and then scraped off with ice-cold PBS containing protease inhibitors. Cells were pelleted, resuspended, and subjected to sonication to get about 200-1000 base pairs of DNA. After removing cell debris, the samples were diluted 10-fold in ChIP dilution buffer containing protease inhibitors. Five-μg of anti-H3 antibody (positive control was provided with the kit), or anti-pSMAD2/3 antibody (Cell Signaling Technology) were added to the chromatin solution and incubated overnight at 4°C with rotation. Protein G-agarose was then added and incubated at 4°C for two hours. The protein/DNA complexes were eluted with ChIP elution buffer. DNA was released form protein/DNA complexes by incubation with 5M NaCl at 65°C for 4 h. The DNA was purified and 50 μl of DNA was obtained for each treatment. 0.2 μl of DNA from each group was used as a template for PCR. Primers for the SRGN promoter containing putative SMAD3 binding sites included forward primer: 5′-AAACTCCTCCCTCCTATCAA-3′, reverse primer: 5′-AGCCTATCATACATCCTTGC-3′ for site 1; forward primer: 5′-TGTATTTATTGTATAACTTT-3′ and reverse: 5′-TGCCTGTAGTCCCAGCTA-CT-3′ for site 2; forward primer: 5′-CTCGCCACCA CGCCCGGCTA-3′ and reverse primer: 5′-GGCTCCATATTAAAGTTATA-3′ for site 3; forward primer: 5′-GTTTGCTGGGCACGGTGGCT-3′ and reverse primer: 5′-CCTGAGTAGCT-GGGATTACA-3′ for site 4; forward primer: 5′-AAGAAGTTGGCGTGCAGCTG-3′ and reverse primer: 5′-ATGGACCACAGGGCTTACAG-3′ for site 5. The primers amplify human GAPDH gene used forward primer: 5′-TACTAGCG-GTTTTACGGGCG-3′ and reverse prime: 5′-TCGAACAGGAGGAGCAGA-GAGCGA-3′. The PCR conditions were as follows: one cycle at 95°C, 5 min; 32 cycles at 95°C, 20 s, 60°C, 30 s, and 72°C, 30 s; and then one cycle at 72°C for 7 min. PCR samples were electrophoresed on 2% agarose gels and stained with ethidium bromide. In addition, real-time PCR was carried out according to standard protocols using an ABI7500 with SYBR Green detection (Applied Biosystems).

Histological analysis
Tissues were fixed in 10% neutral buffered formalin for 48 h and then transferred to 70% ethanol before embedding. Tissues were sectioned at 4-μm, mounted on DakoFlex slides (Dako, Denmark), and stained with haematoxylin & eosin.
All reagents used for immunohistochemistry were obtained from Beyotime Institute of Biotechnology (Beijing, China). Sections (4-μm) were deparaffinized in xylene. Endogenous peroxidase was blocked with 3% hydrogen peroxide in deionized water for 20 min. Antigen was retrieved in citrate buffer (10 mM, pH 6.0) at 95°C for 30 min. Sections were stained with SRGN (1:100) and TGFβ2 (1:100) antibody for 1 h at 37°C, followed by biotinylated secondary antibody for 30 min, reaction with horseradish peroxidase for 30 min, and visualization with hydrogen peroxide-activated diamino benzidine. After washing with 1 × PBS, sections were counterstained with hematoxylin, dehydrated using ethanol, cleared with xylene, and mounted in mounting medium. Sections treated without primary antibodies were used as negative controls. The investigators who performed histological analysis were blinded to the group allocation during the experiment and when assessing the outcome.

Animal experiments
The immunocompetent female BALB/c mice (10-11 weeks old) were purchased from Guangdong Animal Center (China) and randomly grouped. The mice were first injected with 1 × 10 5 of MDA-MB-231/ shSRGN or MDA-MB-231/shNC stable cells via tail vein (N = 8). After 5 weeks, the rats were killed to observe pulmonary metastases. In addition, 1 × 10 6 of MDA-MB-231/shSRGN or MDA-MB-231/shNC stable cells were injected subcutaneously into the right posterior of the mice to observe tumor growth. Tumor volume was calculated using V = a 2 × b/2 (a = width, b = length). Tumor growth and metastasis were observed by two investigators blinding to the experimental design through recording GFP signals using a small animal live-body imager (Bruker in vivo Fx pro, USA). Briefly, mice were anesthetized (inhalation) and placed on the imaging platform. The fluorescent signal was detected with a filter of 520 nm and an excitation of 480 nm. Tumor size and location was observed. All animal experiments were approved by the Experimental Animal Ethics Committee of Guangzhou Medical University.

Statistics
Statistical analyses were performed using a GraphPad Prism version 5.0 software (GraphPad Software Inc., La Jolla, CA, USA). A two-tailed Fisher's exact test was used to determine if the frequency distribution were statistically significant. Comparison of treatments was performed using one-way analysis of variance with Newman-Keuls post-test or a paired two-way Student's t-test. Differences were considered statistically significant at values of P o0.05.