A total of 45 pairs of BC and respective adjacent normal tissues were obtained from 45 female patients (mean age, 55.4 years; age range, 24-76 years) who underwent surgery at the Department of Breast Surgery, Xiangya Hospital, Central South University (Changsha, China) between January 2017 and February 2018. Adjacent tissue samples were obtained within 3 cm of the edge of the cancer tissue. The 45 paired BC samples were divided into TNBC (n=16) and non-TNBC (n=29) based on immunohistochemistry staining of estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor (HER-2), and according to the Chinese Society of Clinical Oncology Breast Cancer Diagnosis and Treatment Guidelines 2018 (31).

No preoperative radiotherapy, chemotherapy or other tumour treatments were performed prior to tumour resection. Informed consent was signed by all patients, and this research was approved by the Ethics Committee of Xiangya Hospital Affiliated to Central South University (approval no. 201803377). MCF10A, MCF7, T47D, MDA-MB-436, MDA-MB-231, MDA-MB-468, SK-BR-3 cell lines were purchased from American Type Culture Collection and identified by STR verification. The MDA‑MB‑436, MDA‑MB‑231 and MDA-MB-468 cells were TNBC cell lines. The cells were cultured in DMEM (Invitrogen; Thermo Fisher Scientific, Inc.) supplemented with 10% foetal bovine serum (FBS; Invitrogen; Thermo Fisher Scientific, Inc.), 1% streptomycin (Invitrogen; Thermo Fisher Scientific, Inc.) and 1% penicillin (Invitrogen; Thermo Fisher Scientific, Inc.) in a 5% CO₂ atmosphere at 37˚C.

Tumor tissue samples were fixed with 10% formalin for 24 h at 4˚C, embedded in paraffin and sectioned to a thickness on 2 µm. The sections were then depa-raffinized and blocked with 10% goat serum (Gibco; Thermo Fisher Scientific, Inc.) in PBS for 1 h at 4˚C. The slides were incubated with anti-ER (1:100; Abcam; cat. no. ab108398), anti-PR (1:100; Abcam; cat. no. ab32085) and anti-HER-2 (1:100; Abcam; cat. no. ab32085) for 12 h at 4˚C. Following washing with PBS, the sections were incubated with a goat anti-rabbit HRP-conjugated secondary antibody (1:10,000; cat. no. 4414; Cell Signalling Technology, Inc.) for an additional 30 min at 25˚C. Following staining with hematoxylin for 5 min at 25˚C, the slices were blocked with neutral gum at 25˚C for 24 h and images were obtained using a light microscope (magnification, ×100). Non‑specific primary antibody staining was assessed by substituting the primary antibody with PBS, and at least 10 fields were randomly selected from each section for evaluation. For scoring, ER negative and PR negative were defined as <1% of cells with positive ER and positive PR staining based on five independent fields. HER‑2^- or HER-2⁺ were considered to be negative for HER-2. HER-2^- was when the pathologist observed no staining or membrane staining in <10% of the tumour cells. HER-2⁺ was when faint or barely perceptible membrane staining was detected in >10% of tumour cells. A sample was considered HER-2⁺⁺ when the pathologist observed a weak-to-moderate complete membrane staining in >10% of the tumour cells. HER-2⁺⁺ samples were applied for fluorescence in situ hybridization, and the cases with amplification were defined as HER2-positive. The samples negative for all three markers were defined as TNBC, whereas all other cases were considered non-TNBC.

HER-2⁺⁺ samples were applied for fluorescence in situ hybridization. The ZytoLight^® SPEC ERBB2/CEN 17 Dual Color Probe (Zytomed, Systems GmbH), which employs two DNA probes, was used to perform the assay, according to the manufacturer's protocol. The following probes were used: HER‑2(ERBB2) gene specific probe, ZyGreen (excitation 503 nm/emission 528 nm)-labelled polynucleotides (~10.0 ng/µl), which target sequences mapping in 17q12-q21.1* (chr17:37,572,531-38,181,308) harbouring the ERBB2 gene region; and an α satellite probe targeting the centromere region of chromosome 17 (CEP17): ZyOrange (excitation 547 nm/emission 572 nm)-labelled polynucleotides (~1.5 ng/µl), which target sequences mapping in 17p11.1-q11.1 specific for the alpha satellite centromeric region D17Z1 of chromosome 17. In brief, specimens were incubated with pre‑treatment solution (99˚C for 20 min) and then digested with protease (37˚C for 10 min). The DNA probe was applied and hybridized to each section overnight at 37˚C. The slides were then washed, counterstained with DAPI, and observed by fluorescence microscopy. In each of the specimens of papillary carcinoma, the total numbers of HER2 and CEP17 signals were counted in 20 tumour cell nuclei, and the ratios of HER2 signals to CEP17 signals were calculated. Polysomy 17 was defined as the occurrence of a centromere copy number of three or more for chromosome 17 per cell.

Possible miR-574-5p targets were predicted by bioinformatics analysis with TargetScan (version 7.2; http://www.targetscan.org/vert_72/).

Short hairpin RNA (shRNA)-negative control (NC), shRNA-BCL11A (shBCL11A), NC mimic and miR-574-5p mimic (for miR-574-5p overexpression) were purchased from Shanghai GenePharma Co., Ltd. The target sequences of the shRNA are presented in Table I. The miRNA mimics (2 µg, 100 nM) or shRNA (1 µg, 100 nM) were transfected into MDA-MB-468 or MDA-MB-231 cells using Lipofectamine^® 3000 (Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. For SOX2 and BCL11A overexpression, the pcDNA3.1 vector (Shanghai GenePharma Co., Ltd.) was used to construct the recombinant plasmids pcDNA3.1-SOX2 and pcDNA3.1-BCL11A, respectively. The recombinant vectors (1 µg, 100 nM) were transfected into MDA-MB-468 or MDA-MB-231 cells using Lipofectamine^® 3000, according to the manufacturer's protocol. A total of 48 h after transfection, cells were used for subsequent experiments.

The MTT experiment was applied to analyse cell viability. MDA-MB-468 and MDA-MB-231 cells were seeded in 96-well plates (5x10⁴ cells/µl). Subsequently, 0.5% MTT (Sigma-Aldrich; Merck KGaA) was added to each well and incubated for 4 h. Then, DMSO (150 µl) was added to each well and incubated for 10 min. The absorbance at 490 nm was detected using the Thermo Fisher Multiskan FC plate reader (Thermo Fischer Scientific, Inc.).

MDA-MB-468 and MDA-MB-231 cells in the logarithmic growth phase were serially diluted and then inoculated separately in culture dishes and cultured at 37˚C for 2 weeks. The cells were fixed with a mixture of acetic acid and formic acid (at a ratio of 1:3) for 15 min at 25˚C, stained with Giemsa solution for 15 min at 25˚C and then rinsed slowly with running water. Colonies were counted using a light microscope (magnification, x40). Five fields were randomly selected and counted for each well.

Cells were seeded into the upper chamber (1.0x10⁶ cells/chamber). RPMI-1640 medium (Gibco; Thermo Fisher Scientific, Inc.) with 10% FBS was added to the lower chamber. Cells were incubated in the chambers at 37˚C for 48 h. The lower chamber was stained with 0.1% crystal violet 30 min at 25˚C, and the cells were counted under a light microscope (magnification, x100). Five fields were randomly selected and counted for each well. For the cell migration assay, 24-well plates (8-µm pores; Corning, Inc.) without coating were used. For the invasion assay, the upper chambers were coated with Matrigel (1:8 dilution; BD Biosciences).

Total RNA was extracted using TRIzol (Thermo Fisher Scientific, Inc.). The PrimeScript RT Reagent kit with gDNA Eraser (Takara Bio, Inc.) and the SYBR Green qPCR Master mix kit (Takara Bio, Inc.) were used for RT (45 min at 50˚C) and qPCR, respectively. The thermocycling conditions were as follows: 96˚C for 6 min, followed by 40 cycles of 96˚C for 15 sec, 57˚C for 30 sec and 72˚C for 30 sec. GADPH was used as an internal control for SOX2 and BCL11A. U6 was used as an internal control for miRNA-574-5p. The 2^-ΔΔCq method (32) was used to calculate the relative expression level. The primer sequences are listed in Table II.

Total protein of MDA-MB-468 or MDA-MB-231 cells was extracted using RIPA buffer (Invitrogen; Thermo Fisher Scientific, Inc.). The protein concentration was measured using a BCA Protein assay kit (Takara Bio, Inc.). Protein (30 µg) was separated by 10% SDS-PAGE at 40 V for 4-5 h and then transferred to a PVDF membrane (Sigma-Aldrich; Merck KGaA). The PVDF membrane was stained with 1x Ponceau S solution (Beijing Solarbio Science & Technology, Co., Ltd.) for 5 min. The membranes were then blocked with 5% non-fat milk at room temperature for 1 h and incubated with the following primary antibodies: Anti-SOX2 (1:1,000; cat. no. 3579), anti-BCL11A (1:1,000; cat. no. 75432), anti-Slug (1:1,000; cat. no. 9585), anti-N-cadherin (1:1,000; cat. no. 13116), anti-vimentin (1:1,000; cat. no. 5741), anti-E-cadherin (1:1,000; cat. no. 14472), anti-SKIL (1:1,000; cat. no. 4973), anti-TAZ (1:1,000; cat. no. 83669), anti-CTGF (1:1,000; cat. no. 86641) and anti‑GAPDH (1:10,000; cat. no. 5174) at 4˚C overnight. All primary antibodies were from Cell Signaling Technology, Inc. Then, the membranes were washed with TBST buffer and incubated with goat anti-rabbit IgG antibody tagged with horseradish peroxidase (1:10,000; cat. no. 4414; Cell Signaling Technology, Inc.) at room temperature for 2 h. The membranes were scanned using an Odyssey Imaging system and analysed with the Odyssey v2.0 software (LI-COR Biosciences).

The binding elements of miRNA-574-5p in BCL11A and SOX2 were predicted by Starbase (http://starbase.sysu.edu.cn). The wild-type (WT) BCL11A, mutant (Mut) BCL11A, WT SOX2 and Mut SOX2 were synthesized and inserted into the pGL3 vector (YouBio) to construct pGL3-BCL11A-WT, pGL3-BCL11A-Mut, pGL3-SOX2-WT and pGL3-SOX2-Mut, respectively, which were obtained from Shanghai GenePharma Co., Ltd. Then, the recombinant plasmids were transfected into cells with NC mimic or miR-574-5p mimics using Lipofectamine 3000 (Invitrogen; Thermo Fisher Scientific, Inc.). A Dual-Luciferase™ Reporter (DLR™) assay system (Promega Corporation) was applied for luciferase activity analysis. The activity of normalized reporter was expressed as the firefly luciferase value divided by the renilla luciferase value, and normalized to the control vector activity.

This research was approved by the Ethics Committee of Xiangya Hospital affiliated to Central South University (approval no. 201803245). In total, 20 male BALB/c nude mice (4-6 weeks old) were obtained from SLAC Laboratory Animal Co. Ltd, (Hunan, China). Mice were raised in pathogen‑free conditions at 20‑25˚C, 40‑70% humidity, 15 times/h ventilation and a 12 h light/12 h dark cycle, with access to food and water ad libitum. For the xenograft assay, the mice were transplanted subcutaneously with MDA-MB-231 cells (transfected with NC mimic or miR-574-5p mimics) into the right flank and then housed until the tumour volume reached ~200 mm³. Tumour size was measured every 5 days using the following formula: Length x width² x 0.5. On the 30th day, the mice were sacrificed and the tumour tissue was photographed. For the in vivo metastasis assay, 5x10⁵ MDA-MB-231 cells (transfected with NC mimic or miR-574-5p mimics) were injected into the tail veins of nine mice. On the 30th day, the mice were sacrificed and the lung tissues were photographed. The number of pulmonary nodules was counted.

Haematoxylin-eosin staining was used for pulmonary migration analysis. Tissues were fixed by 4% formalin at 25˚C for 24 h, embedded in paraffin and cut into 5‑µm slices. The slices were stained with hematoxylin (Beyotime Institute of Biotechnology) for 5 min at 25˚C and eosin (Beyotime Institute of Biotechnology) for 3 min at 25˚C. The slices were then blocked with neutral gum at 25˚C for 24 h and imaged using a light microscope (magnification, x100).

Statistical analysis was performed using GraphPad Prism 6.0 software (GraphPad Software, Inc.). Three independent experiments were performed, and data are expressed as the mean ± standard deviation. Student's t-test was performed for comparisons between two groups, while one-way ANOVA followed by Tukey's post-hoc test was performed for multiple groups. P<0.05 was considered to indicate a statistically significant difference.

The RT-qPCR results revealed that the expression level of miR-574-5p was significantly lower in tumour tissue compared with in the adjacent tissue of 45 patients (Fig. 1A). Subsequently, miR-574-5p expression levels in TNBC (n=29) and non-TNBC (n=16) were compared. As presented in Fig. 1B, miR‑574‑5p expression level in TNBC was significantly lower compared with that in non-TNBC. Furthermore, the expression of miR-574-5p was lower in both TNBC and non-TNBC cell lines (MDA-MB-436, MDA-MB-231, MDA-MB-468) compared with that in the normal breast epithelial cell line MCF10A (Fig. 1C). In summary, the expression of miR-574-5p was shown to be lower in both BC tissues and cells, particularly in TNBC tissue.

To explore the effects of miR-574-5p on TNBC progression, miR-574-5p was overexpressed in TNBC cells via transfection with miR-574-5p mimics or NC mimic. Compared with the NC mimic group, miR-574-5p expression was significantly increased in the miR‑575‑5p mimic‑transfected group (Fig. 2A). The MTT assay revealed that cell viability was significantly suppressed in cells that overexpressed miR-574-5p (Fig. 2B and C). A colony formation assay was performed to detect the effect of miR-574-5p overexpression on colony formation. The results demonstrated that the colony formation ability of miR-574-5p-overexpressing cells was significantly inhibited compared with that of cells transfected with NC mimic (Fig. 2D). In addition, Transwell assays showed that miR‑574‑5p overexpression significantly inhibited cell migration and invasion (Fig. 2E and F). The effects of miR-574-5p on the expression of EMT-related proteins, E-cadherin, vimentin, Slug and N-cadherin, were analysed by western blotting. Overexpression of miR‑574‑5p significantly inhibited the expression of Slug, N-cadherin and vimentin, and significantly increased the expression of E‑cadherin (Fig. 2G). Taken together, these findings suggested that miR-574-5p inhibits proliferation, migration and EMT in TNBC cells.

To explore the mechanism by which miR-574 regulates the malignant phenotype, the present study examined the proteins involved in the SKIL pathway. The results of the western blot analysis revealed that overexpression of miR‑574‑5p significantly decreased the expression of SKIL, TAZ and CTGF (Fig. 3A). Next, how miR-574 regulates SKIL was examined. The effect of miR-574-5p overexpression on BCL11A and SOX2 expression levels was investigated, which have been reported to regulate SKIL transcription (22). The data demonstrated that the expression levels of BCL11A and SOX2 in cells transfected with miR-574-5p mimics were significantly lower compared that in control cells (Fig. 3B). Additionally, the relative mRNA expression levels of BCL11A and SOX2 in cells transfected with miR-574-5p mimics were significantly lower compared with in the NC mimic group (Fig. 3C and D). Bioinformatics prediction revealed that miR-574-5p targets the 3′-untranslated regions (3′-UTRs) of BCL11A and SOX2 (Fig. 3E). Subsequently, a luciferase assay was used to verify that miR-574-5p binds directly to the 3′-UTRs of BCL11A and SOX2. In the WT group, miR-574-5p overexpression significantly inhibited the luciferase activity of the reporter gene containing the WT BCL11A and SOX2 3′-UTR fragments, whereas for the Mut group, there was no significantly difference in luciferase activity between the NC mimic and miR-574-5p mimic groups, demonstrating that miR-574-5p binds directly to the predicted sequences of BCL11A and SOX2 (Fig. 3F and G). These combined data suggested that miR-574-5p targets BCL11A and SOX2, which may lead to miR-574-5p-mediated inhibition of the SKIL/TAZ/CTGF axis.

As presented in Fig. 4A-C, miR-574-5p overexpres-sion significantly inhibited tumour size and weight in vivo. Additionally, miR-574-5p overexpression inhibited pulmonary migration and significantly reduced the number of metastatic nodules (Fig. 4D-F). Hence, it can been suggested that miR-574-5p may inhibit tumorigenesis and lung metastasis in vivo.

Co-overexpression of miR-574-5p and SOX2 or BCL11A was performed by co-transfection with miR-574-5p mimics and pcDNA3.1-SOX2 or pcDNA3.1-BCL11A in MDA-MB-231 and MDA-MB-468 cells. Transfection efficiency is presented in Fig. S1A and B. Western blot analysis revealed that co-overexpression of miR-574-5p and SOX2 or BCL11A significantly reversed the inhibition of SKIL by miR-574-5p (Fig. 5A and B). Overexpression of SOX2 restored both BCL11A and SOX2 expression (Fig. 5A and B). However, BCL11A overexpression only restored its own expression and had no effect on the expression of SOX2 (Fig. 5A and B). Co-overexpression of miR-574-5p and SOX2 or BCL11A restored the inhibition of cell viability by miR-574-5p (Fig. 5C and D). miR-574-5p overexpression inhibited cell migration ability, while overexpression of miR-574-5p and SOX2 or BCL11A restored migration (Fig. 5E). Furthermore, western blot analysis demonstrated that co-overexpression of miR-574-5p and SOX2 or BCL11A reversed the inhibition of vimentin, Slug, and N-cadherin expression and the promotion of E-cadherin expression by miR-574-5p (Fig. 5F and G). In summary, the inhibition of proliferation, migration and EMT by miR-574-5p in TNBC cells was at least partly dependent on targeting SOX2 and BCL11A.

As presented in Figs. 6A and S1C and D, the knockdown efficiency of BCL11A shRNAs was examined and shRNA#3 was selected for subsequent experiments as it demonstrated the best knockdown efficiency in MDA‑MB‑231 cells. To further investigate the underlying interactive regulation of SOX2 and BCL11A, pcDNA3.1-SOX2 and shBCL11A were co-transfected into MDA-MB-231 and MDA-MB-468 cells. Western blot assays indicated that transfection with pcDNA3.1‑SOX2 significantly upregulated the expression of SOX2, SKIL and BCL11A. BCL11A knockdown reversed this effect but had no effect on SOX2 expression (Fig. 6B and C). SOX2 overexpression significantly promoted cell viability and migration, while silencing BCL11A reversed this effect (Fig. 6D‑F). Moreover, overexpression of SOX2 significantly increased the expression of Slug, vimentin and N-cadherin, and decreased the expression of E-cadherin, while BCL11A knockdown reversed this effect (Fig. 6G and H). Collectively, these results demonstrated that the oncogenic effect of SOX2 is partly dependent on BCL11A.