Background
Breast cancer is a heterogeneous disease that can be classified into several histological forms in current clinical practice. The molecular etiologies among different types of breast cancers are largely different, making the treatment of breast cancer difficult. Triple-negative breast cancer, which is characterized by the lack of expression of the estrogen receptor (ER), the progesterone receptor (PR), and the human epidermal growth factor receptor 2 (HER2), is a type of breast cancer with aggressive tumor behavior. Many targeted treatments, including endocrine therapies and HER2-targeted medicine, are not efficacious for triple-negative breast cancer [
1,
2]. Although previous studies have shown that the vast majority of triple-negative breast cancers display basal-like gene expression features [
3,
4] the molecular mechanisms driving tumor progression of triple-negative breast cancer still remain unknown.
MicroRNAs (miRNAs) are short (~21mer) non-coding RNA molecules that are important in gene expression regulation [
5,
6]. The primary role of miRNAs appears to be in the negative regulation of the expression of messenger RNA (mRNA) transcripts. The functional strand of a mature miRNA guides the RNA-induced silencing complex to bind a target mRNA in the 3′-untranslated region (3′-UTR), initiating translational repression, target mRNA cleavage, or mRNA deadenylation of the target gene. Emerging evidence has shown that aberrant miRNA expression plays a critical role in the tumorigenesis of many human cancers [
7-
9]. Some miRNAs are shown to possess oncogenic characteristics that promote malignancy of human cancers [
10], while some have tumor-suppressing abilities to reduce the production of oncogenic proteins [
11,
12].
The advent of deep sequencing technology allowed us to explore the largely unknown territory of the miRNA transcriptome in triple-negative breast cancer. Sequencing reads of miRNA expression data from 24 triple-negative breast cancers and 14 adjacent normal tissues were analyzed for the presence of deregulated miRNAs in this study. Differentially expressed miRNAs in triple-negative breast cancer were determined by statistical analyses of the sequencing data and were validated using the quantitative reverse transcription PCR (RT-PCR) method. We identified seven polycistronic miRNA clusters in the human genome harboring 29 deregulated miRNAs in triple-negative breast cancer. Furthermore, our work extends the potential target network of miRNAs by showing that the cyclin G2 gene (CCNG2) is a direct target of miR-130b-5p from the miR-301b-130b cluster. Forced expression of miR-130b-5p was found to significantly repress the endogenous expression levels of CCNG2 in triple-negative breast cancer cells. The findings described in this work may provide insights into the miRNA regulatory mechanisms underlying the tumorigenicity of triple-negative breast cancer.
Discussion
Sequencing data of miRNA expression reads from 24 triple-negative breast cancers and 14 adjacent normal tissues were analyzed in this study. Deregulated miRNAs were identified from the statistical analyses and a panel of the top 25 deregulated miRNAs was found to be an effective discriminator between triple-negative breast cancers and adjacent normal tissues. Deep sequencing technology allowed us to generate a comprehensive insight into the cellular transcriptome in triple-negative breast cancer that led to the identification of many more deregulated miRNAs not described in previous studies [
18,
19]. For example, aberrant expression from the miR-532-502 cluster in triple-negative breast cancer was first documented in this study. Interestingly, the expression level of the tumor suppressor gene
RUNX3 was found to be inversely correlated with that of miR-532-5p from the miR-532-502 cluster in primary melanomas [
20].
A number of the validated target genes of the deregulated miRNAs in our findings were shown to be involved in human cancer signaling cascades (Additional file
4). For example, we observed that both miR-143-5p and miR-145-5p in the miR-143-145 cluster were down-regulated in triple-negative breast cancer. Down-regulation of these two miRNAs was previously described in lung cancer, colon cancer, and bladder cancer [
21-
23]. The proto-oncogene c-Myc was a direct target for miR-145-5p, and introduction of miR-145-5p repressed c-Myc expression and tumor growth both
in vitro and
in vivo [
11]. The insulin receptor substrate-1 (
IRS-1), previously known as a major docking protein for both the type 1 insulin-like growth factor receptor and the insulin receptor in cancer cell growth and proliferation signaling, was also a validated target of miR-145-5p [
24]. In addition, Sachdeva
et al. reported that cell invasion ability was significantly inhibited by miR-145-5p, in part due to the silencing of the metastasis gene
MUC1 [
25].
Deregulation of the miR-497-195 cluster has been previously addressed in breast cancer. Li
et al. described that miR-497-5p and miR-195-5p in this cluster were both down-regulated in human breast cancer tissues and cell lines [
26]. However, the study did not specifically determine whether the deregulation of these two miRNAs observed in breast cancer cell lines was also observed in triple-negative breast cancer tissues. Our results provide evidence showing
in vivo that both miR-497-5p and miR-195-5p are down-regulated in triple-negative breast cancer tissues. Furthermore, the methylation state of CpG islands in the promoter region upstream of the miR497-195 cluster was responsible for the down-regulation of those two miRNAs in breast cancer, and direct targets of miR-195-5p included
CCND1 and
RAF1 [
26]. Moreover, introduction of miR-195-5p was shown to inhibit cancer cell colony formation
in vitro [
26] and tumor development in nude mice [
27], suggesting that ectopic expression of miR-195-5p may be involved in the tumorigenesis of breast cancer.
We were able to identify
CCNG2 as a direct target of miR-130b-5p in triple-negative breast cancer. Luciferase reporter assays revealed that miR-130b-5p-mediated repression of
CCNG2 is dependent on the sequence of the 3′-UTR.
CCNG2 is known as a negative regulator of cell cycle progression. Previous microarray analyses have shown that elevated CCNG2 expression induces cell cycle arrest during responses to various types of growth inhibitory effects, such as hypoxia, oxidative stress, and heat shock [
17,
28,
29]. The CCNG2 protein directly interacts with the catalytic subunit of protein phosphatase 2A to form active complexes that inhibit cell cycle progression [
15]. Increasingly, the evidence suggests that CCNG2 is crucially involved in human cancer signaling pathways [
30-
32]. For example,
CCNG2 was a primary target gene of estrogen-occupied estrogen receptor and that its expression was rapidly down-regulated by estrogens in MCF-7 breast cancer cells [
31]. Moreover,
CCNG2 promoter activity was found to be regulated by Nodal signaling in ovarian cancer cells and silencing of CCNG2 expression significantly increased cell proliferation [
32].
Gene expression levels of
CCNG2 between triple-negative breast cancers and normal breast tissues were further investigated using a published microarray dataset [GEO:GSE53752]. The gene expression levels of
CCNG2 in triple-negative breast cancers (n = 51) were significantly lower (
p <0.001; fold change 1.9) than those in normal breast tissues (n = 25) in the microarray data (Additional file
5). Of added interest, recently
CCNG2 was found to be an important prognostic factor for triple-negative breast cancer patients. In an analysis of a cohort of 250 primary triple-negative breast cancer samples from eight clinically annotated gene expression datasets, triple-negative breast cancer patients with low expression levels of
SHARP1 and
CCNG2 had a significantly higher probability of developing metastases and of reduced survival [
33]. It is thus important that we identified
CCNG2 as a direct target of miR-130b-5p. The ability of miR-130b-5p to repress CCNG2 expression may enhance malignancy by accelerating cell cycle transition in triple-negative tumor cells.
Methods
Breast cancer and normal tissue samples
Twenty-four triple-negative breast cancer samples and 14 adjacent normal tissue samples were collected from breast cancer patients during surgeries at National Taiwan University Hospital (NTUH, Taipei, Taiwan). All triple-negative breast cancer samples were invasive ductal carcinomas and were negative in immunohistochemical analyses of ER, PR, and HER2. AJCC/UICC TNM staging system was used for tumor classification. Treatment of each patient followed the National Comprehensive Cancer Network (
http://www.nccn.org/) guidelines. All samples were neoadjuvant-free and were collected before systemic chemotherapy treatments. Written informed consent was obtained from each patient who participated in this study. All human tissues used in this study were approved by the institutional review board at NTUH.
Small RNA library preparation
Total RNA was extracted from each sample for the preparation of a small RNA library. The small RNA library was constructed from total RNA using the SOLiD Total RNA-Seq Kit (Applied Biosystems, Foster City, CA, USA). Integrity of each small RNA library was examined using an RNA 6000 Nano Chip (Agilent, Santa Clara, CA, USA), a Small RNA Chip (Agilent), and the Bioanalyzer (Agilent) according to the manufacturer’s instructions.
Deep sequencing experiments
Upon completion of PCR amplification, the small RNA libraries were purified using the SOLiD Library Micro Column Purification Kit (Applied Biosystems) and hybridized to the template beads using the SOLiD EZ bead system (Applied Biosystems). The template beads were amplified and deposited onto a tray for small RNA ligation sequencing by the SOLiD 4 System (Applied Biosystems). The sequencing data were uploaded to the Gene Expression Omnibus (GEO) with an accession number of GSE40049.
Sequence alignment of miRNA reads
All reads obtained from ligation sequencing were first screened to filter out the reads containing ribosomal RNA, transfer RNA, and adaptor sequences. The remaining reads were then aligned to the human miRNA reference (miRBase v17.0) and the human genome reference (RefSeq Hg19) using the Small RNA Analysis Tool (Applied Biosystems). In the sequence alignment, only one mismatch was allowed for the first 16 bases of a miRNA read. The maximum number of permitted mismatches of a miRNA read was set at 4. Those reads that were not uniquely mapped to the miRBase reference were disregarded to eliminate ambiguous alignments.
Statistical analyses
The quantile-quantile scaling method [
34] was performed for the normalization of miRNA expression reads in log
10-scale. All miRNA expression reads in each dataset were linearly scaled to fit into a miRNA expression reference made of the mean expression value of each miRNA from 38 samples. Principal component analysis (PCA) was performed to analyze the miRNA expression profiles between triple-negative breast cancers and adjacent normal tissues with the Partek Genomics Suite (Partek Incorporated, St. Louis, MO, USA). Significant differences in expression of miRNAs from the triple-negative breast cancers and adjacent normal tissues were identified using two-tailed Student’s t-tests. The Holm step down procedure was used to counteract multiple comparisons. A
p-value of <0.05 was considered significant. The miRNAs with a fold change >2 and mean expression difference >100 reads between the two groups were investigated in the hierarchical clustering analysis using Genesis software (version 1.7.5).
Quantitative RT-PCR validation of miRNA expression
Differentially expressed miRNAs identified from our sequencing data were validated using quantitative RT-PCR in 19 triple-negative breast cancer samples and 4 adjacent normal tissue samples. Total RNA was extracted from each sample and then reverse transcribed into miRNA-specific cDNA following the standard protocol of the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems). Relative quantification of miRNA expression in each sample was obtained using the comparative threshold (C
T) method [
13]. Expression of U6 small nuclear RNA was used as the endogenous control.
Putative target prediction
Potential mRNA target genes of a miRNA were searched using the miRanda [
35] and Diana [
36] target prediction algorithms. Putative target candidates having complementary base-pairing matches in the 3′-UTR for the indicated miRNA seed region were obtained. Biological functions associated with the target genes were investigated using Ingenuity Pathway Analysis software (Ingenuity Systems, Redwood City, CA, USA).
Experimentally validated miRNA target genes
Experimentally validated miRNA target genes were retrieved from TarBase [
37] and miRecords [
38] using the miRSystem search engine [
39]. The molecular pathways encompassing the validated target genes were investigated using Ingenuity Pathway Analysis software.
Cell lines
MDA-MB-231 and HEK-293 T were obtained from Bioresource Collection and Research Center (Taiwan). The cell lines were tested and authenticated by Genelabs Life Science (Taiwan) using STR-PCR profiling.
Vectors
miR-130b-5p was cloned into a lentiviral vector PreMiR-130b (System Biosciences, Mountain View, CA, USA) that was used to overexpress the miRNA in MDA-MB-231 cells. Expression of miR-130b-5p was verified and quantified using KAPA PROBE Fast qPCR Master Mix (Kapa Biosystems, Boston, MA, USA), and the LightCycler 480 System (Roche, Basel, Switzerland).
A
CCNG2 luciferase reporter construct was made by introducing the
CCNG2 3′-UTR carrying a predicted miR-130b-5p binding site (5′-CCTTGGAGATACTGAAAGAGA-3′) into the pmirGLO control vector (Promega, Madison, WI, USA). Site-directed mutagenesis of the putative miR-130b-5p binding site was made using a facile PCR procedure [
40]. All PCR products were verified by DNA sequencing before use.
Luciferase assay
Luciferase assays were performed with HEK-293 T cells using the Dual-Glo® Luciferase Assay System (Promega). Cells were transfected with PreMiR-130b lentiviral vectors using TransIT®-2020 transfection reagent (Mirus Bio, Madison, WI, USA). Forty-eight hours after transfection, the cells were then harvested and lysed for the luciferase assay. Renilla luciferase signals were used for normalization according to the manufacturer’s protocol.
Cell cycle analysis
3×104 cells of MDA-MB-231 cells were seeded in a 24-well plate. miR-130b-5p or empty vector (control) were overexpressed in the MDA-MB-231 cells. Cell synchronization was performed using double thymidine block. 2 mM of thymidine was added into cells and cells were incubated at 37°C for 16 hours. To remove thymidine, cells were washed with PBS and incubated with fresh media at 37°C for 8 hours. 2 mM of thymidine was added into cells and cells were incubated at 37°C for 16 hours. To release the cells from thymidine block, cells were washed with PBS and incubated with fresh media and collected after 14 hours. The cells were harvested by trypsinization and washed twice with cold PBS. The cells were fixed by 0.5 mL of cold 95% ethanol and were kept at -20°C overnight. The ethanol was removed and cells were washed twice with PBS. The cells were resuspended in 500 μL of PI solution (10 μg/mL of propidium iodide, 0.2 mg/mL of RNaseA and 0.1% Triton X-100). Cell cycle profile was analyzed by fluorescence-activated cell sorter (FACS) analysis.
Gene expression microarray experiments
Gene expression data of CCNG2 in our cohort of 38 samples (14 normal breast tissues and 24 triple-negative breast cancer samples) were analyzed using Agilent Human 1A (version 2) microarray platform. Microarray experiments were performed following the manufacturer’s instructions. Microarray data were normalized using Quantile normalization before statistical analyses.
The sequencing data were uploaded to the Gene Expression Omnibus (GEO) with an accession number of GSE40049.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
YYC and WHK carried out the miRNA transcriptome study, performed bioinformatics analyses, and drafted the manuscript. JHH and CYL participated in the miRNA sequencing alignment. YHL, WCL, and CYS performed RT-PCR validation and luciferase assay experiments. CSH, FJH, LCL, and MHT participated in the design of the study and high-throughput sequencing data acquisition. KJC and EYC provided intellectual criticisms and gave final approval to the manuscript to be published. All authors read and approved the final manuscript.
WHK and CSH are breast cancer surgeons at National Taiwan University Hospital. LCL is a professor of Physiology at National Taiwan University. MHT is a professor of Biotechnology at National Taiwan University. KJC is a professor of Breast Surgery at National Taiwan University. EYC is a professor of Biomedical Electronics and Bioinformatics at National Taiwan University.