Our study has already been approved by the Ethical Committee and Institutional Review Board of Fudan University Shanghai Cancer Centre (FDUSCC). The methods were performed in accordance with the approved guidelines. All the participants signed written informed consent forms. The tumour samples were collected between May 2014 and May 2015. The breast cancer samples and matched para-carcinoma samples were collected from 39 patients pathologically diagnosed with breast cancer who undergone surgery. The samples were frozen immediately in liquid nitrogen after surgery.

Total RNA was extracted using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) and reverse transcribed with PrimeScript RT Reagent kit (Takara Biotechnology Co., Ltd., Dalian, China). Subsequently, RT-qPCR was performed with SYBR Premix Ex Taq (Takara Biotechnology Co., Ltd.) using ABI Prism 7900 instrument (Applied Biosystems; Thermo Fisher Scientific, Inc.) according to the manufacture's protocol. The thermocycling condition was pre-denaturation at 95°C for 30 sec, denaturation at 95°C for 5 sec, annealing at 60°C for 30 sec, with 40 cycles. The relative gene expression was calculated using the 2^−ΔΔCq method (12). The primer sequences used in this study are as follows: FOXP2: forward 5′-AACAACAGCAGGCTCTCCAG-3′, and reverse 5′-GGCACCTGCAGTGGTCTC-3′; GAPDH: forward 5′-GGTGGTCTCCTCTGACTTCAACA-3′, and reverse 5′-GTTGCTGTAGCCAAATTCGTTGT-3′.

The cells in this study were obtained from the Shanghai Cell Bank, Type Culture Collection Committee of Chinese Academy of Science (Shanghai, China). The cells for experiments were passaged for less than 6 months. MDA-MB-231, MDA-MB-231BO, MDA-MB-231HM and MDA-MB-468 (all triple negative breast cancer) cells were grown in Leibovitz L-15 medium (BasalMedia, Shanghai, China). 293T, MCF7 (luminal positive breast cancer) and BT549 (triple-negative breast cancer) cells were cultured in high-glucose Dulbecco's modified Eagle's medium (DMEM; HyClone; GE Healthcare Life Sciences, Logan, UT, USA). SKBR3 (HER2 positive breast cancer), T47D, ZR7530 and ZR75-1 (all luminal positive breast cancer) were maintained in RMPI-1640 medium (HyClone; GE Healthcare Life Sciences). All the medium was supplemented with 10% FBS, 100 IU/ml penicillin, and 100 mg/ml streptomycin. All the cells were maintained with 5% CO₂ at 37°C. The MDA-MB-231HM was developed from the parental MDA-MB-231 cell line via the tail vein in mice for four cycles. We have patent application for the cell line (patent no. 200910174455.4) which exhibited increased lung metastasis compared to parental cells. The MDA-MB-231BO cells, which have highly metastatic potential to bone, was obtained from Dr Toshiyuki Yoneda (University of Texas, Houston, TX, USA). We have done several studies with these cell lines (4,13).

Human FOXP2 cDNA was purchased from fulenGen and then subcloned into the pCDH-CMV-MCS-EF1-Puro lentiviral vector. The cloned primer sequence is as follows: FOXP2: forward 5′-CCGGAATTCGCCACCATGATGCAGGAATCTGCGAC-3′, and reverse 5′-CGCGGATCCTCATTCCAGATCTTCAGATA-3′.

FOXP2 shRNAs and the negative control were purchased from GeneChem and expressed in the GV248 backbone. The target sequences are as follows: shRNA NC 5′-TTCTCCGAACGTGTCACGT-3′; shRNA1: 5′-TTAACAATGAACACGCATT-3′; shRNA2: 5′-AGCAAACAAGTGGATTGAA-3′.

293T cells were cotransfected with lentiviral vectors and the packaging vectors pCDH (GV248), psPAX2 and pMD2G. Virus-containing medium was collected 48 h after the transfection and added to the cancer cells.

MDA-MB-231 and MDA-MB-231BO cells (3.6×10⁴) were plated on 96-well plates (Essen ImageLock; Essen Biosciences, Ann Arbor, MI, USA), and a wound was scratched with wound scratcher (Essen Instruments). Wound confluence was monitored with Live-Cell Imaging System and software (Essen Biosciences). The wound closure was observed after 28 h by comparing the mean relative wound density of at least three biological replicates in each experiment.

Cell migration and invasion was examined with Transwell chambers (BD Biosciences, Franklin Lakes, NJ, USA). The cells were seeded in the upper chamber in serum-free medium. The cell density was 5×10⁵ for migration assay and 1×10⁶ for Matrigel-coated invasion assay. Medium supplemented with 20% serum was added in the lower chamber. The cells were incubated for 15 to 20 h. After that, the cells remained on the upper chamber were removed with cotton swabs. The cells on the lower surface of the membrane were stained with 0.4% methanol and 0.25% crystal violet.

Whole cell extracts were isolated with Pierce T-PER (Tissue Protein Extraction Reagent; Thermo Fisher Scientific, Inc.) containing protease inhibitor cocktail tablets (Roche Diagnostics, Indianapolis, IN, USA) and phosphatase inhibitors (Roche Diagnostics). Proteins (30 µg) were resolved by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred onto a polyvinylidene fluoride membrane (Pall Corporation, Port Washington, NY, USA). The membranes were blocked in 5% BSA and incubated in various primary antibodies followed by the corresponding horseradish peroxidase-conjugated secondary antibodies (Proteintech Group, Inc., Chicago, IL, USA). The primary antibody against FOXP2 (21608) and vimentin (40477) was obtained from Santa Cruz Biotechnology, Inc. (Dallas, TX, USA), fibronectin (66042-1-Ig) and GAPDH (60004-1-Ig) from Proteintech Group, Inc., N-cadherin (610920) from BD Biosciences, p-SMAD3 (ab52903) from Abcam (Cambridge, UK), SMAD3 (1735-s) from Epitomics (Burlingame, CA, USA), SMAD4 (9515p) from Cell Signaling Technology, Inc. (Danvers, MA, USA) and TGFβR1 (sc-398) from Santa Cruz Biotechnology, Inc. Antibody binding was detected using chemiluminescence (Amersham Imager 600; GE Healthcare Life Sciences), according to the manufacturer's instructions.

SPSS 22.0 software (IBM Corp., Armonk, NY, USA) was used to conduct the statistical analysis. Data were expressed as means ± standard deviation. The student's t-test was used to evaluate the differences between two groups and one-way analysis of variance (ANOVA) was used among multiple groups. LSD test was used after ANOVA to compare the differences between two groups. P<0.05 was considered to indicate a statistically significant difference. Graphs were created with GraphPad Prism 5 (GraphPad Software, Inc., La Jolla, CA, USA).

The information of FOXP2 expression and patients' survival was acquired from an online Kaplan-Meier plotter (http://kmplot.com/analysis/). This online tool was based on data of 1,809 patients downloaded from GEO (Affymetrix HGU133A and HGU133+2 microarrays) (14). The low and high expression was divided by median expression. In Kaplan Meier analysis (Fig. 1), the expression level of FOXP2 was significantly correlated to the patients' relapse-free survival (RFS, P=0.0047, HR=0.79; Fig. 1B), but not to overall survival (OS, P=0.18, Fig. 1A). Among patients with different molecular subtypes, low FOXP2 expression only predicted higher risk of relapse among luminal A (P=0.0037) and basal-like breast cancer patients (P=0.014). The results suggested that low expression of FOXP2 may be associated with recurrence and metastasis in some breast cancer patients.

According to the data from oncomine (www.oncomine.org), the expression of FOXP2 was significantly lower in breast cancer than normal breast tissue (P<0.001; Fig. 2A). The results remained same both in invasive ductal carcinoma and invasive lobular carcinoma (Fig. 2B and C). However, there was no significant difference among patients with and without metastasis (Fig. 2D). We also evaluate the expression level of FOXP2 in different breast cancer cell lines (Fig. 2E and F) as well as 39 patients' samples (Fig. 2G). Interestingly, FOXP2 was significantly less expressed in breast cancer tissue compared to adjacent control tissue (P=0.0005). The results were consistent with the data from the public database. Table I summarized the clinicopathological characteristics of the patients. In this cohort, the FOXP2 expression was not significantly associated with these clinical features except age.

In addition, we also investigated the frequency of FOXP2 alterations, including amplification and mutations, in breast cancer (Fig. 2H) from cbioprotal (www.cbioportal.org) (15). We found that among 4,786 invasive breast cancer cases from 9 studies, 144 patients had amplification (3%) and 17 (0.4%) had mutation of this gene.

To validate the function of FOXP2 in breast cancer metastasis and proliferation, we knocked down FOXP2 in MDA-MB-231 and overexpressed it in MDA-MB-231BO and MDA-MB-231HM breast cancer cells via lentivirus infection. The efficiency of infection was confirmed by western blot analysis (Fig. 3A). We then found that the depletion of FOXP2 did not have significant effects on cell proliferation (Fig. 3B). But the wound-healing assay and Transwell assay showed that the knockdown of FOXP2 could promote cell migration and invasion in vitro (P=0.001; Fig. 3C and D). Meanwhile, the cell migration and invasion was inhibited after the overexpression of FOXP2 (P<0.001). Taken together, the results suggested that FOXP2 may play a certain role in breast cancer as a tumor suppressor gene.

Epithelial-mesenchymal transition (EMT) was known as one of crucial mechanisms in cancer progression and metastasis (16). In order to further investigate the mechanisms of FOXP2 inhibiting breast cancer metastasis, we detected the expression level of several EMT-associated biomarkers in MDA-MB-231 and BT549 (Fig. 4A). The results revealed that suppression of FOXP2 promoted the expression of vimentin and fibronectin in MDA-MB-231. In BT549 cell line, fibronectin and N-cadherin was inhibited after overexpression of FOXP2. The results suggested that FOXP2 could inhibit EMT in breast cancer in vitro.

The activation of TGFβ/SMAD signaling was known as one of the critical pathways in EMT process and cancer metastasis (17,18). Therefore, we also detected the expression of some important proteins in TGFβ signaling (Fig. 4B). We found that upregulation of FOXP2 in BT549 could markedly decrease the phosphorylation of SMAD3, enhanced the expression of SMAD4, TGFβR1, ZEB1 and Snail. Meanwhile, suppression of the gene in MDA-MB-231 could cause the reverse effects except ZEB1. The expression of Twist and p-SMAD2 did not seem to show significant differences in the cell lines in this research. The phenomenon indicated that inhibition of FOXP2 induce EMT process in breast cancer probably via TGFβ/SMAD signaling pathway.