The human breast cancer cell line MDA-MB-231 was purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA). The cells were cultured in RPMI-1640 medium (HyClone, Hudson, NH, USA) supplemented with 10% fetal bovine serum (FBS) at 37°C and 5% CO₂.

The knockdown of RAGE in MDA-MB-231 cells was performed using siRNA. The sequence targeting RAGE was 5′-CACTGCAGTCGGAGCTAATGG-3′. A scrambled oligo siRNA (Santa Cruz Biotechnology, Inc., Dallas, TX, USA) was used as a control at the same concentration. The overexpression of RAGE was accomplished using a pcDNA3.1 vector. The overexpression of miR-328-5p in MDA-MB-231 cells was performed using miR-328-5p mimics, and scramble mimics were used as the control at the same concentration. MDA-MB-231 cells were initially plated in media containing 10% FBS. After 24 h, the cells were washed once with phosphate-buffered saline (PBS) and transfected with vector/mimics/siRNA or scrambled siRNAs by Nucleofector (Lonza Group, Ltd., Basel, Switzerland) according to the manufacturer's instructions.

MDA-MB-231 cells transfected with siRAGE and scrambled siRNAs were subjected to Agilent Human miRNA microarray (8×60 K) analysis at the Oebitech Co. (Shanghai, China). Total RNA was quantified using NanoDrop ND-2000 (Thermo Fisher Scientific, Inc., Waltham, MA, USA), and the RNA integrity was assessed using Agilent Bioanalyzer 2100 (Agilent Technologies, Inc., Santa Clara, CA, USA). The sample labelling, microarray hybridisation and washing were performed according to the manufacturer's standard protocols. Briefly, total RNA was dephosphorylated, denatured and then labelled with cyanine-3-CTP. After purification, the labelled RNAs were hybridised onto the microarray. After washing, the arrays were scanned with the Agilent Scanner G2505C (Agilent Technologies Inc.). Feature Extraction software (version 10.7.1.1; Agilent Technologies, Inc.) was then used to analyse array images to obtain the raw data. GeneSpring software (version 13.1; Agilent Technologies) was employed to complete the basic analysis of the raw data. The threshold set for upregulated and downregulated genes was a fold change of ≥2.0 and a P-value ≤0.05. The target genes of differentially expressed miRNAs were predicted with three databases (TargetScan, microRNA.org and PITA). Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) analyses were performed to determine the roles of these target genes. Hierarchical clustering was performed to reveal the distinguishable miRNA expression patterns among samples.

Total RNA was extracted from cell lines and tissues using TRIzol (Dongshen Biotech, Co., Ltd. Guangzhou, China). RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Inc.) was used for reverse transcription. qPCR was performed using an ABI Prism 7300 Sequence Detection System (Applied Biosystems; Thermo Fisher Scientific, Inc.) with SYBR-Green PCR mixture (Dongshen Biotech, Co., Ltd.). The miRNA and U6 primers were purchased from GeneCopoeia (Guangzhou, China). The RAGE primers were as follows: forward, GGTGCCTAATGAGAAGGGAGTA and reverse, GAAGCTACAGGAGAAGGTGGG. The β-actin primers were as follows: forward, AGGGGCCGGACTCGTCATACT and reverse, GGCGGCACCACCATGTACCCT. The relative expression level was determined using the 2^−ΔΔCt analysis method, in which β-actin and U6 were used as internal standards. All reactions were conducted in triplicate, and all experiments were performed in three independent replicates.

The 13 paired patient tissues (breast cancer and adjacent tissues) were collected during surgical resection at the Third Xiangya Hospital (from May 2015 to September 2017). All patient tissues were confirmed by histopathological evaluation. The study was approved by the Ethical Committee of the Third Xiangya Hospital of Central South University. No patients received preoperative radiotherapy and/or chemotherapy. The collected tissues were immediately stored at −80°C until use.

The cells were collected, and the concentration was adjusted to 5×10⁴ cells/well before culture. The cells were cultured in 35-mm dishes at 5% CO₂ and 37°C for 24, 48 and 72 h. For chemosensitivity detection, the cells were treated with cisplatin (DDP; 0–200 µM) for 48 h. Then, 10 µl of MTT was added to the cells and maintained at 5% CO₂ and 37°C for another 4 h before MTT was removed. The optical density value was measured at 570 nm while the cells were suspended in 150 µl of dimethyl sulfoxide (DMSO).

The clonogenicity of single cells was detected by colony assay. Cells were collected by adding 0.25% trypsin and were adjusted to a concentration of 400 cells/petri dish, which was loaded with 2 ml of preheated culture media before culturing with 5% CO₂ at 37°C for 2–3 weeks. Colony formation was terminated when the colonies were visible to the naked eye. Subsequently, the cells were washed twice with PBS, and then 4% paraformaldehyde was added for 15 min to fix the cells before staining with Giemsa for 10 min. Then the number of colonies were counted, and the colony formation rate was calculated as follows: Colony formation rate = (number of colonies/inoculation cell number) × 100%.

The invasion ability was determined using a Transwell assay. MDA-MB-321 cells (5×10⁴) were plated in Transwell chambers. The chambers were placed in a 24-well plate and incubated for 24 h at 37°C. The cells that invaded the membrane were stained using a crystal violet staining solution kit (Beyotime Institute of Biotechnology, Shanghai, China). The stained cells were photographed using an inverted microscope (Motic Inc., Ltd., Causeway Bay, Hong Kong). After capturing the images, the stain from the membrane was eluted in 500 µl of DMSO. The eluted staining agent was then shifted into a 96-well plate and the absorbance was assessed using a microplate reader (Thermo Fisher Scientific, Inc.) at 570 nm.

The cell cycle was assessed via flow cytometry. Cells were trypsinised (Auragene Bioscience Co., Changsha, China) and washed twice with PBS before fixation in 500 µl 75% precooled ethanol at 4°C overnight. The cells were then washed twice with PBS and incubated in a water bath at 37°C for 30 min with 100 µl RNase A (Auragene Bioscience Co.). The cell suspensions were incubated for 30 min in the dark using propidium iodide (PI) (Beijing Solarbio Science & Technology Co., Ltd., Beijing, China) or for 15 min using an Annexin V/PI detection kit (Molecular Probes, Inc.; Thermo Fisher Scientific, Inc., Eugene, OR, USA), according to the manufacturer's instructions. Finally, the cell cycle and apoptosis were analysed by FACScanto™ (BD Biosciences, Franklin Lakes, NJ, USA).

Cells were lysed in cold RIPA buffer, and protein was separated with 10–12% SDS-PAGE, which was then transferred to NC membranes (Thermo Fisher Scientific, Inc.). The membranes were blocked with 5% non-fat dried milk (Beijing Yili Fine Chemical Co., Ltd., Beijing, China) in PBS Tween-20 solution (0.1% v/v) for 3 h at 4°C. Then, the membranes were incubated with primary antibodies rabbit polyclonal rage (1:1,000; cat. no. ab37647), rabbit polyclonal anti-NF-κB p65 (1:50,000; cat. no. ab32536), rabbit monoclonal anti-cyclin D1 (1:10,000; cat. no. ab134175) and mouse monoclonal anti-β-actin (1:1,000; cat. no. ab8226; all were from Abcam, Cambridge, MA, USA) overnight at 4°C, and then with the appropriate secondary antibodies goat anti-rabbit (1:5,000; cat. no. ab6721) and goat anti-mouse (1:10,000; cat. no. ab6789; both were from Abcam) for 1 h at room temperature. The immune complexes were detected using ECL Western Blotting kit (Thermo Fisher Scientific, Inc.). The relative protein expression was analysed using Image-Pro Plus software 6.0, using β-actin as the internal reference.

The possibility of miR-328-5p and RAGE target binding was predicted using the online software TargetScan (http://www.targetscan.org/vert_71/). The RAGE mRNA predicted binding site region and mutated site region were cloned into the psi-CHECK2 vector. MDA-MB-231 cells were co-transfected with psi-CHECK2 vectors and miR-122 mimics or scrambled controls. After 48 h, the supernatants were collected and the luciferase activities were assessed using a Dual-Luciferase Reporter Assay system (Promega, Fitchburg, WI, USA).

Statistical analysis was performed using GraphPad Prism 6 software (GraphPad Software, Inc., La Jolla, CA, USA) or SPSS 18.0 software (SPSS, Inc., Chicago, IL, USA). All experiments were repeated at least thrice, and the data are shown as the mean ± standard deviation. An unpaired two-tailed Student's t-test and one-way analysis of variance with Bonferroni post hoc test were used to analyse the data, depending on conditions and group number. The relationship between RAGE and miR-328-5p expression was determined by Pearson correlation analysis. P<0.05 was considered to indicate a statistically significant difference.

In total, 2,550 miRNAs were detected by Agilent Human miRNA microarray. In addition, 10 miRNAs were found to be differentially expressed after knockdown of RAGE, as shown in Fig. 1A. The expression changes and reported functions of the miRNAs are listed in Table I (9–18).

Target genes of differentially expressed miRNAs were predicted using three databases (TargetScan, microRNA.org and PITA), as shown in Fig. 1B. GO and KEGG analyses were used to determine the roles of these databases with crossed target genes. According to the results shown in Fig. 2, we found that the following targets were principally involved: i) biological processes (transcription from RNA polymerase II promoter, response to mechanical stimulus and regulation of insulin secretion); ii) cellular components (i.e. nucleoplasm, cytoplasm and nuclear chromatin); and iii) molecular functions (protein and chromatin binding, ionotropic glutamate receptor binding). These genes are mainly involved in the following pathways or conditions: Hippo signalling, alcoholism, Rap1 and various cancer proliferation-related pathways.

qPCR was used to assess the microarray results. The results of all 10 differentially expressed miRNAs were the same with the microarray results and had almost similar fold changes, but only slightly lower (Fig. 3).

Since miR-328-5p has been reported less in breast cancer, it was chosen for further studies. To investigate the role of miR-328-5p in breast cancer, we collected 13 samples of breast cancer tissue and compared these with healthy adjacent tissues. The qPCR results revealed that the expression of miR-328-5p was significantly decreased in breast cancer tissue compared with adjacent tissues (Fig. 4A). To assess the function of miR-328-5p in breast cancer, we used miR-328-5p mimics to construct a miR-328-5p overexpression cell model in MDA-MB-231 cells, with scrambled mimics as a negative control. The representative qPCR assay results, displayed in Fig. 4B, indicated that our miR-328-5p overexpression model was successfully constructed.

miR-328-5p mimics inhibit cell proliferation, drug resistance and the cell cycle. To elucidate the functions of miR-328-5p in breast cancer, MDA-MB-231 cells were transfected with miR-328-5p and scramble mimics for 24, 48 and 72 h. MTT and colony formation assays revealed that miR-328-5p mimics significantly suppressed cell viability (Fig. 4C and E). However, the Transwell assay revealed that miR-328-5p mimics did not affect cell invasiveness (data not shown). Moreover, chemotherapy resistance against DDP was suppressed in the miR-328-5p mimic group, as previously described by Pan et al (11) (Fig. 4D). To investigate the mechanisms of miR-328-5p inhibition on cellular proliferation and drug resistance, we analysed the effects of miR-328-5p on cell apoptosis and the cell cycle. As shown in Fig. 4F, we found that the cell cycle was significantly arrested in phase G1/S after overexpression of miR-328-5p. However, apoptosis was not influenced (data not shown).

The microarray was conducted after knockdown of RAGE; hence, we analysed the relationship between RAGE and miR-328-5p. We found that the relative expression level of RAGE was markedly decreased when miR-328-5p was overexpressed in MDA-MB-231 cells (Fig. 5A and B). Then we analysed the correlation between RAGE and miR-328-5p expression in 13 samples of breast cancer tissue. We found that there was a negative correlation between RAGE and miR-328-5p expression in breast cancer tissue (Fig. 5C). Online (http://www.targetscan.org/) bioinformatics predictions revealed that RAGE was a target of miR-328-5p. The Wt-RAGE-3′UTR or Mut-RAGE-3′UTR luciferase reporter vector was generated, respectively (Fig. 5D). We further performed a luciferase reporter assay to confirm their relationship in MDA-MB-231 cells. As shown in Fig. 5E, luciferase activity was significantly downregulated only in cells co-transfected with the miR-328-5p mimics and the Wt-RAGE-3′UTR vector, which indicated that miR-328-5p could directly bind to the 3′ untranslated region (3′UTR) of RAGE in breast cancer cells.

After confirming that RAGE was the target of miR-328-5p, we investigated whether the function of miR-328-5p in MDA-MB-231 cells was associated with RAGE by co-transfecting the overexpression vector of RAGE in cells overexpressing miR-328-5p to further observe cellular function changes. We found that RAGE overexpression could reverse the inhibition of cell growth, chemotherapy resistance, clone formation ability and cell cycle progression caused by the miR-328-5p mimics (Fig. 6A-D). We used western blotting to assess the protein expression level of RAGE, nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB) and cyclin D1. We found that RAGE enhanced the expression levels of these proteins (Fig. 6E).