Deregulated microRNAs in triple-negative breast cancer revealed by deep sequencing

MicroRNAs (miRNAs) are short, non-coding RNA molecules that play critical roles in human malignancy. However, the regulatory characteristics of miRNAs in triple-negative breast cancer, a phenotype of breast cancer that does not express the genes for estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2, are still poorly understood. In this study, miRNA expression profiles of 24 triple-negative breast cancers and 14 adjacent normal tissues were analyzed using deep sequencing technology. Expression levels of miRNA reads were normalized with the quantile-quantile scaling method. Deregulated miRNAs in triple-negative breast cancer were identified from the sequencing data using the Student’s t-test. Quantitative reverse transcription PCR validations were carried out to examine miRNA expression levels. Potential target candidates of a miRNA were predicted using published target prediction algorithms. Luciferase reporter assay experiments were performed to verify a putative miRNA-target relationship. Validated molecular targets of the deregulated miRNAs were retrieved from curated databases and their associations with cancer progression were discussed. A novel 25-miRNA expression signature was found to effectively distinguish triple-negative breast cancers from surrounding normal tissues in a hierarchical clustering analysis. We documented the evidence of seven polycistronic miRNA clusters preferentially harboring deregulated miRNAs in triple-negative breast cancer. Two of these miRNA clusters (miR-143-145 at 5q32 and miR-497-195 at 17p13.1) were markedly down-regulated in triple-negative breast cancer, while the other five miRNA clusters (miR-17-92 at 13q31.3, miR-183-182 at 7q32.2, miR-200-429 at 1p36.33, miR-301b-130b at 22q11.21, and miR-532-502 at Xp11.23) were up-regulated in triple-negative breast cancer. Moreover, miR-130b-5p from the miR-301b-130b cluster was shown to directly repress the cyclin G2 (CCNG2) gene, a crucial cell cycle regulator, in triple-negative breast cancer cells. Luciferase reporter assays showed that miR-130b-5p-mediated repression of CCNG2 was dependent on the sequence of the 3′-untranslated region. The findings described in this study implicate a miR-130b-5p-CCNG2 axis that may be involved in the malignant progression of triple-negative breast cancer. Our work delivers a clear picture of the global miRNA regulatory characteristics in triple-negative breast cancer and extends the current knowledge of microRNA regulatory network.


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][8][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 triplenegative breast cancers and 14 adjacent normal tissues were analyzed for the presence of deregulated miRNAs in this study. Differentially expressed miRNAs in triplenegative 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 miR-NAs 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.

Results
A 25-miRNA signature discriminating triple-negative breast cancer from adjacent normal tissue Clinical information on the patients was recorded as shown in Table 1. A total of 113,412,568 miRNA reads were identified from 24 triple-negative breast cancers and 14 adjacent normal tissues after sequence alignment with the human miRNA reference, yielding a median of 2,670,242 miRNA reads per sample from our data (range: 393,305-7,906,634; Additional file 1). Setting a threshold to filter out the miRNAs with extremely low reads (mean expression <5 reads across 38 samples), we identified 707 mature miRNAs. The reads from those 707 miRNAs constituted 99.97% of all miRNA reads in our data for further analysis. The miRNA reads in each sample were normalized using the quantile-quantile scaling method on a log 10 -scale. Normalized expression data from each of the 707 miRNAs was used as input to generate a PCA plot to visually assess the intrinsic variation in the global miRNA profiles among the samples. The PCA plot showed that the global miRNA portrait was able to roughly separate the majority of triple-negative breast cancers from the adjacent normal tissues ( Figure 1A).
Differentially expressed miRNAs between the triplenegative breast cancer group (N = 24) and the adjacent normal tissue group (N = 14) were identified using Student's t-test and the Holm step down procedure for p-value adjustment. A list of the top 25 significantly differentially expressed miRNAs (adjusted p <0.05) was determined as shown in Table 2. Average-linkage hierarchical clustering analysis was performed on the samples using the expression data from the 25 miRNAs ( Figure 1B). All triple-negative breast cancers and adjacent normal tissues were distinctively divided into two clusters without any misclassification. To examine whether the 25-miRNA signature contained relevant information in classification of triple-negative breast cancers and normal breast tissues in PCA, each sample was projected onto the principal components using the expression data from the 25 miRNAs ( Figure 1C). The PCA plot revealed that the triple-negative breast cancers and the adjacent normal tissues were clearly separated into two groups using the 25-miRNA signature.

Quantitative RT-PCR validations of the deregulated miRNAs
To confirm the results obtained from the sequencing data, quantitative RT-PCR validations were carried out to examine the expression levels of the miRNAs residing in the miR-532-502 cluster, the miR-301b-130b cluster, and the miR-497-195 cluster in 19 triple-negative breast cancers and 4 adjacent normal tissues using the comparative C T method [13]. A conceptual diagram that depicts the specific 5p/3p miRNA forms located at each locus in the miR-532-502 cluster is shown in Figure 2A. miR-532-5p ( Figure 2B; mean fold change 2.4), miR-188-3p ( Figure 2C; mean fold change 2.5), miR-362-5p ( Figure 2D; mean fold change 4.0), miR-501-3p ( Figure 2E; mean fold change 5.3), miR-660-3p ( Figure 2F; mean fold change 2.2), and miR-502-5p ( Figure 2G; mean fold change 3.0) were all markedly up-regulated in the triplenegative breast cancers compared to the normal breast tissue controls.  Figure 3A shows the specific miRNA forms at each locus in the miR-301b-130b cluster. Quantitative RT-PCR data confirmed that miR-301b-5p, miR-130b-5p, and miR-130b-3p were all up-regulated in 19 triplenegative breast cancer samples ( Figure 3B-D, respectively). Two 5p miRNA forms, miR497-5p and miR-195-5p, located at two different loci in the miR-497-195 cluster are shown in Figure 3E. Compared with the normal breast tissue controls, the expression levels of miR497-5p were markedly down-regulated ( Figure 3F; mean fold change 0.3) in triple-negative breast cancers. Furthermore, miR-195-5p in the same cluster was down-regulated To understand how these deregulated miRNAs might function in triple-negative breast cancer, we performed target prediction of the miRNAs as described in the Methods section. Among those putative target candidates, five potential tumor suppressors, including CCNG2, FOXP1, NRG1, LRIG1, and TNFSF10, were shown to have a direct binding site with miR-130b-5p. We were particularly interested in CCNG2, because it was a crucial regulatory gene in cell cycle control [14,15]. Previous studies have functionally linked the decrease or complete loss of CCNG2 in human cancer to uncontrolled cell proliferation [16,17]. We thus further characterized the regulatory mechanism of miR-130b-5p in relation to CCNG2 in cells.
A schematic diagram of the putative miR-130b-5p binding site in the 3′-UTR of CCNG2 is shown in Figure 4A. Two luciferase reporter constructs, Luc-CCNG2-UTR and Luc-CCNG2-UTR-mt3, were made to test whether miR-130b-5p might directly target CCNG2 in the putative binding site. The possible binding sequences in the 3′-UTR of CCNG2 were carried by the CCNG2 gene reporter, Luc-CCNG2-UTR. The other luciferase reporter construct, Luc-CCNG2-UTR-mt3, was made to carry three mutated nucleotides at the putative binding site to disrupt binding of miR-130b-5p to CCNG2. The luciferase reporter assays revealed that miR-130b-5p significantly reduced the luciferase activity of the wild-type CCNG2 gene reporter, but not that of the site-directed mutant ( Figure 4B), suggesting that miR-130b-5p had a suppressive effect at the predicted binding site in the 3′-UTR of CCNG2.
Next, we examined whether miR-130b-5p might repress endogenous CCNG2 expression in triple-negative breast Normalized miRNA reads from each sample were obtained after quantile-quantile scaling. The raw p-value of each miRNA was calculated using the log 10 -scaled miRNA expression data after quantile-quantile scaling and was adjusted using the Holm step down procedure. Relative expression levels are shown. cancer cells. Overexpression of miR-130b-5p was carried out in MDA-MB-231 cells using plasmid DNA transfection of a lentiviral vector containing the precursor sequence of miR-130b-5p. Quantitative RT-PCR data confirmed that the endogenous expression levels of CCNG2 were significantly repressed in MDA-MB-231 cells after overexpression of miR-130b-5p ( Figure 4C). Compared with negative controls, a marked reduction (>50%) in CCNG2 expression level was observed in the miR-130b-5p-overexpressed MDA-MB-231cells. The functional relevance of miR-130b-5p in cell cycle regulation with triple-negative breast cancer cells was analyzed in this study. After overexpression of miR-130b-5p, MDA-MB-231 cells showed significant alterations (p <0.05) in cell cycle profile (Additional file 2). A significant decrease in G1 phase and increases in S and G2/M phases were observed with the miR-130b-5p-overexpressed MDA-MB-231cells in the cell cycle analysis. These results strongly support that miR-130b-5p plays a role in cell cycle progression through suppression of CCNG2 expression in triple-negative breast cancer.
We then investigated an association between miR-130b-5p expression and malignancy of triple-negative breast cancer from the sequencing data in our cohort. Sequencing data of miR-130b-5p expression were shown to be significantly associated with tumor progression in both early-stage triple-negative breast cancers (p = 0.002; n = 7) and advanced stage triple-negative breast cancers (p <0.001; n = 17) in Figure 4D. The expression levels of CCNG2 in our cohort of 38 samples were also measured Figure 2 RT-PCR validations of six differentially expressed miRNAs in the miR-532-502 cluster in triple-negative breast cancer. (A) A conceptual diagram illustrates the genomic loci harboring the 5p miRNA forms (brown) and 3p forms (orange) in the miR-532-502 cluster. The names of specific miRNA forms are provided under each locus. Relative expression levels of (B) miR-532-5p, (C) miR-188-3p, (D) miR-362-5p, (E) miR-501-3p, (F) miR-660-3p, and (G) miR-502-5p in triple-negative breast cancer tissues (n = 19) and adjacent normal tissues (n = 4) are shown. In each panel, paired tissue samples from the same patient are indicated with the same digits in the sample ID. The miRNA expression level in 557-N served as the calibrator control. All six miRNAs examined in the miR-532-502 cluster were markedly up-regulated in triple-negative breast cancer. Each bar represents the mean of triplicate measurements ± the SD.
using Agilent microarray (Human 1A v2). A direct correlation between high miR-130b-5p expression (p <0.001; fold change 2.8) and low CCNG2 expression (p <0.001; fold change 0.5) was observed in our triple-negative breast cancer samples as compared with normal breast tissue controls (Additional file 3A and B).

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). Relative expression levels of (B) miR-301b-5p, (C) miR-130b-5p, and (D) miR-130b-3p in triple-negative breast cancer tissues (n = 19) and adjacent normal tissues (n = 4) are shown. All three miRNAs in the miR-301b-130b cluster were markedly up-regulated in triple-negative breast cancer. (E) A conceptual diagram illustrates the genomic loci harboring the 5p miRNA forms (brown) in the miR-497-195 cluster. Relative expression levels of (F) miR-497-5p and (G) miR-195-5p in 23 samples are shown. Both miRNAs in the miR-497-195 cluster were down-regulated in triple-negative breast cancer. Each bar represents the mean of triplicate measurements ± the SD.
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][22][23]. The protooncogene 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 Luciferase assays of CCNG2 in miR-130b-5p-overexpressed HEK-293 T cells. Two luciferase reporter constructs, Luc-CCNG2-UTR and Luc-CCNG2-UTR-mt3, were made to test whether miR-130b-5p directly targets the putative binding site in the 3′-UTR of CCNG2. The mutated nucleotides are indicated in red. Compared with the negative control, overexpression of miR-130b-5p was shown to significantly inhibit luciferase reporter activity with Luc-CCNG2-UTR, but not with Luc-CCNG2-UTR-mt3. Each bar represents the mean of triplicate measurements ± the SD; *p <0.05. (C) Endogenous expression analysis of CCNG2 in MDA-MB-231 cells transfected with lentiviral vectors containing miR-130b-5p precursor sequences. The endogenous expression levels of CCNG2 were significantly repressed after the forced expression of miR-130b-5p in MDA-MB-231 cells. (D) Comparisons of miR-130b-5p expression in normal breast tissues (n = 14), early-stage triple-negative breast cancers (n = 7), and advanced-stage triple-negative breast cancers (n = 17). MiRNA expression reads were normalized and analyzed in log 10 scale. miR-130b-5p expression was significantly associated with both early-stage and advanced-stage triple-negative breast cancers.
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][31][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 triplenegative 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, triplenegative 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.

Conclusions
Our work clearly depicts the global miRNA regulatory characteristics in triple-negative breast cancer. The 25-miRNA signature determined in this study may be used as a functional tool to distinguish triple-negative breast cancer tissues from normal breast tissues. Seven polycistronic miRNA clusters preferentially harboring deregulated miRNAs were identified in triple-negative breast cancer. Moreover, we extended the current knowledge of microRNA regulatory network by showing that miR-130b-5p in the miR-301b-130b cluster directly silences CCNG2 in triple-negative breast cancer. It would be interesting to test whether this novel regulatory mechanism of miR-130b-5p and its CCNG2 target in triple-negative breast cancer may be involved in other malignancies as well. Our findings not only provide insight into the miRNA regulatory mechanisms in triple negative breast cancer, but also shed light on the identification of potential therapeutic targets for this disease.

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 triplenegative 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 10scale. 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 miRNAspecific 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.
A CCNG2 luciferase reporter construct was made by introducing the CCNG2 3′-UTR carrying a predicted miR-130b-5p binding site (5′-CCTTGGAGATACTGA AAGAGA-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.