Involved microRNAs in alternative polyadenylation intervene in breast cancer via regulation of cleavage factor “CFIm25”

Cleavage factor “CFIm25”, as a key repressor at proximal poly (A) site, negatively correlates to cell proliferation and tumorigenicity in various cancers. Hence, understanding CFIm25 mechanism of action in breast cancer would be a great benefit. To this aim four steps were designed. First, potential miRNAs that target 3′-UTR of CFIm25 mRNA, retrieved from Targetscan web server. Second, screened miRNAs were profiled in 100 breast cancer and 100 normal adjacent samples. Third, miRNAs that their expression was inversely correlated to the CFIm25, overexpressed in MDA-MB-231 cell line, and their effect on proliferation and migration monitored via MTT and wound healing assays, respectively. Fourth, interaction of miRNAs of interest with 3′-UTR of CFIm25 confirmed via luciferase assay and western blot. Our results indicate that CFIm25 considerably down-regulates in human breast cancer tissue. qRT-PCR assay, luciferase test, and western blotting confirm that CFIm25 itself could be directly regulated by oncomiRs such as miR-23, -24, -27, -135, -182 and -374. Besides, according to MTT and wound healing assays of cell lines, CFIm25 knockdown intensifies cell growth, proliferation and migration. Our results also confirm indirect impact of CFIm25 on regulation of mRNA’s 3′–UTR length, which then control corresponding miRNAs’ action. miRNAs directly control CFIm25 expression level, which then tunes expression of the oncogenes and tumor proliferation. Therefore, regulation of CFIm25 expression level via miRNAs is expected to improve treatment responses in breast cancer.


Results
Correlation of CFIm25 expression level with oncogenic miRNAs in breast cancer clinical samples and cell lines. According to bioinformatics analysis, miR-23, miR-24, miR-27, miR-96, miR-135, miR-182, miR-221, miR-222 and miR-374 found to target CFIm25 beside their oncogenic potential. In order to determine whether CFIm25 expression is controlled by miRNAs of interest in breast cancer, we measured the expression levels of potentially involved miRNAs in breast cancer patients (n = 100) and normal adjacent tissues (n = 100) using miR-quantitative qRT-PCR analysis. Results showed considerable elevated level of miRNAs in samples acquired from patients (at least in 6 out of 9 monitored miRNAs), while CFIm25 expression level was diminished significantly (Fig. 1a, b). In the next step, patients categorized as low or high CFIm25-expressing tumors, using the median expression for CFIm25 in the normal sample as a cut-off.
Since CFIm25 level differs in samples acquired from patients, their correspondingly correlated miRNAs might be variable too. By this regard, distribution of the median expression of nine miRNAs of interest evaluated respect to both low and high levels of CFIm25 expression (Fig. 1c).

Some of the potentially involved miRNAs in breast cancer, down-regulate
In the next step, we investigated whether 3′-UTR of CFIm25 has functional targets for predicted miRNAs. To this aim, full-length sequence of CFIm25 3′-UTR cloned to psiCHECK-2 vector, which is placed at downstream of Renilla luciferase reporter gene (CFIm25-psiCHECK2). In a parallel experiment, as negative controls, the seed sequence of miR- 23 Table 1.. Proto-oncogenes that CFIm25 and CPSF1 target their 3′UTR mRNA. The cases that are highlighted in dark grey represent proto-oncogenes with positive interaction probability. The RF (Random Forest) classifier is a probability index, which falls between 0 (the lowest probability) to 1 (the highest probability).  Results of migration assay showed that number of migrated cells in the miR-Off-23/miR-Off-374 transduced group, was less than that in the shCFIm25 group. This last finding confirmed that the migration ability of MDA-MB-231 cells was inhibited by miR-Off-23/miR-Off-374, which had been found previously to be under influence of CFIm25 expression level (Fig. 5).

Discussion
It is conformed that most of the human genes have multiple polyadenylation signals, which dynamically regulate their variations in 3′-UTR length 26 . The CFIm25, as a newly discovered repressor of proximal poly (A) site usage, has a significant role in UTR-APA site regulation. The CFIm25, as a subunit of cleavage factor complex, recruits CPSF6 and/or CPSF7 along with CPSF5, and activates 3′-mRNA cleavage site and polyadenylation processing machinery 3,11 . Down-regulation of CFIm25, significantly diminishes "distal poly-A signal" length, and enhances tumorigenic properties and size of the tumor cells. However, CFIm25 overexpression reduces aforementioned properties and inhibits tumor growth. An explanation for this is that by choosing a proximal PAS, the 3′-UTR length of the mRNAs shortens, which then eliminates miRNAs binding sites and revokes regulation through miRNAs 11,[27][28][29] . Here, we approved CFIm25 is indirectly involved in regulation of many genes through changing www.nature.com/scientificreports/ 3′-UTR length, which then eliminates binding sites of their corresponding miRNAs. Nonetheless, more studies are still needed to clarify comprehensive regulatory mechanism of CFIm25 gene by miRNAs.
Although breast cancer is the most common cancer diagnosed in women, survival rates of breast cancer have improved recently, which is mostly due to factors such as a new personalized approach to cancer treatment and a better understanding of the molecular mechanism of the disease 30 . In this study, we used human breast cancer specimens to analyze their CFIm25 expression profile. We confirmed that CFIm25 is significantly down-regulated in breast cancer samples, compared to the controls. This phenomenon is consistent with the previous study 15 that introduced CFIm25 as APA regulator that was also down-regulated in glioblastoma and came up with cancer proliferation and tumorigenicity 16 . Besides, it was reported that expression level of CFIm25 in hepatocellular carcinoma (HCC) was negatively correlated to the metastatic potential of HCC cell line through increasing E-cadherin level 31 , while glutaminase mRNA isoforms, which contain distinct 3′-UTR (KGA and GAC), showed a complex interplay between RNA processing and microRNA repression in controlling glutamine metabolism in cancer cells 14 .
In the next step of the current study, we selected nine oncomiRs using bioinformatics analysis. They all confirmed to have recognition sites in the 3′-UTR of the CFIm25 (miR-23: 2 sites, miR-24: 2 sites, miR-27: 3 sites, miR-96: 1 site, miR-135: 2 sites, miR-182: 1 site, miR-221: 2 sites, miR-222: 2 sites and miR-374: 4 sites). So far, several studies investigated role of oncomiRs in different cancers including breast cancer 20,24 . However, no experimental work has performed to clarify effects of these miRNAs on CFIm25 expression level and, as a result, in regulation of UTR-APA site. When clinical samples divided into two groups, whose CFIm25 expression was more or lower than the median expression of control samples, a negative correlation between level of miR-23, miR-24, miR-27, miR-135, miR-182 and miR-374 and CFIm25's expression was found. This negative correlation in turn suggests that CFIm25 is being regulated by aforementioned miRNAs. Besides, in vitro data from transduced MDA-MB-231 cells confirmed that high levels of miR-23, miR-24, miR-27, miR-135, miR-182 and miR-374 was negatively correlated with either CFIm25 mRNA or its protein content.
According to previous reports, low level of CFIm25 leads to cancer proliferation and migration 14,28,31 , which is in agreement with our MTT and wound healing assay results. Importantly, functional regulatory role of miR-23 and miR-374 on CFIm25 mRNA, approved via MTT assay. This assay confirmed increased growth of MDA-MB-231 cells, which were co-transduced with miR-Off-23/miR-Off-374 and CFIm25 knock-down vectors. CFIm25 is known to be crucial in controlling invasion and metastasis of HCC 32 . It is also confirmed that www.nature.com/scientificreports/ activation of APA sites can affect the migratory capacity of cancer cells 33 as well as fibroblasts 34 . According to the results of wound healing assay, regulation of CFIm25 expression by miR-23 and 374 directly controls breast cancer cells' migration and invasion. Since invasion and metastasis are the underlying causes of poor long-term survival of breast cancer patients 35 , regulation of CFIm25 via miRs could be considered as a promising strategy for inhibition of metastasis in future clinical studies. Parental genes of mRNAs with shorter 3′-UTRs, are more prone to up-regulation during tumorigenesis due to increased escaping from miRNA repression. It is determined that almost 67% genes with shorter 3′-UTRs in tumors, have lost at least one predicted miRNA-binding site 36 . In some cases, overexpression of oncogenic proteins comes up without regulatory intervenience of the proto-oncogenes. One example of this phenomenon is cancer cell lines (such as MDA-MB-231) that express substantial amounts of mRNA isoforms with shortened 3′-UTRs 30,31 . CFIm25 is a regulatory protein that controls key factors such as EGFR and MYC that are involved in NF-κB and MAPK/ERK pathways 31,37,38 . Our finding suggests that regulation of proto-oncogenes by CFIm25, could be mediated via APA and shortening of 3′UTR of their corresponding mRNAs, which eventually result in progression of breast cancer. The MYC, PTEN and AKT2 proto-oncogenes are more responsive to CFIm25 than EGFR and AKT3, since they have several potential interaction sites either at proximal or distal of their mRNA's 3′-UTRs (Table 1). Tuning impact of CFIm25 on MYC, PTEN and AKT2 will then control downstream phenomena such as glutamine metabolism in cancer cells 14 , which could be applied as an strategy for targeted therapy 39 . Besides, 3′-UTR of EGFR's mRNA contain multiple microRNA target sites, which are associated with cell cycle arrest and cell death. As long as CFIm25 interacts with distal region of EGFR's 3′-UTR mRNA (Table 1), Bioinformatic analysis. The miRNAs that were susceptible to target CFIm25 mRNA 3′-UTR acquired using prediction web servers such as TargetScan, miRWalk (version 3.0) and DianaTool. The standard cutoff for screening predicted miRNAs was 8 mer, at the most, while the percentage of context ++ score (CS) did not exceed 95%. Predicted miRNAs with the high score were investigated in the miRCancer database. Finally, miRNAs with the oncogenic roles were selected for the rest of the research, including miR-23, miR-24, miR-27, miR-96, miR-135, miR-182, miR-221, miR-222 and miR-374.
In order to find out if CFIm25 efficiently targets 3′-UTR of proto-oncogenes' mRNAs, three steps were followed. First, 3′-UTR mRNA of twenty proto-oncogenes that known to be involved in breast cancer 42 , retrieved from UTR database 43   . The PAS that is closest to the 5′ end of the 3′-UTR is considered as "proximal" and the other PAS(s) in the downstream of 3′-UTR considered as distal 44 . Third, the retrieved 3′-UTRs beside CFIm25 (O43809) and CPSF1 (Q10570) submitted to the RNA-Protein Interaction Prediction (RPISeq) server to estimate their interaction probabilities 45 . Interaction probabilities ranged from zero to one. Optimal RNA/protein interactions were the ones with interaction probabilities of more than 0.5, which represented accuracy of the Random Forest (RF) classifiers in cross-validation evaluation experiments on benchmark datasets.

RNA extraction and qRT-PCR.
Total RNA extraction of cell line, human breast cancer tissues and normal specimens was performed using Trizol (Invitrogen) in accordance with manufacturer's protocol and previous study 22 . Then, 500 ng of total RNA was reverse transcribed to synthesize cDNA with random hexamer primers and Fermentas kit (for CFIm25 gene) or BON RT adaptor primer and BONmiR kit (for miRNA genes). A total of 1 μL of cDNA was amplified using SYBR Green Real-time PCR Master Mix (Takara) and performed on ABI 157 PRISM 7,500 real-time PCR System (Applied Biosystems). The relative expression levels of the CFIm25 and miRNA genes calculated using the 2 −∆∆Ct method, with normalization against β-actin and SNORD47 internal controls, respectively. Three biological replicates were performed for each set of experiments. The qRT-PCR primer sequences are listed in supplementary Table 1 and 2.
In clinical samples, the mean of CFIm25 and β-actin Cts between all of the normal samples was determined and compared with the obtained Cts of each patient. This procedure adopted from our previous studies with minor modifications 20,46 .
Based on the results, tumors were classified as low and high/equal CFIm25-expressing tumors using the median of the normal samples as a cut-off. Next, the relative gene expression of predicted miRNAs in two groups of patients was calculated by comparing obtained Cts with the mean Cts of normal samples.

plasmids, viral vectors construction and luciferase assay. For the construction of miRNA-express-
ing vectors, genomic fragment consists of the stem-loop structure of each miRNA and theirs flanking genomic sequences, cloned into the mammalian expression vector pLEX.jred and pCDH.TurboGFP at downstream of the cytomegalovirus (CMV) promoter. The PCR primers used for miRNAs cloning were listed in supplementary Table 2. Empty vector without any cloned sequence (pLEX-Ctrl/ pCDH-Ctrl) was used as control in accordance with Furukawa et al. method 47 .
HEK293T cells were transiently co-transfected by miRNA-expressing lentivectors (or empty backbones), pPAX2 plasmid (packaging plasmid), and pMDG plasmid (containing vsv-G) using calcium phosphate. Lentivirus supernatants were harvested for two or three times, every 12 h, concentrated by ultracentrifuge at 47,000×g for 2 h at 4 °C. Lentivirus titer examined by flow cytometry analysis of reporter positive 293 T cells (Attune Acoustic Focusing Flow Cytometer, ABI, FlowJo software) 48 .
To down-regulate miR-23 and miR-374 expression, pLenti-III-miR-Off constructs containing GFP marker purchased from ABM Company and packaged into lentiviruses particles. For luciferase assay, full length of CFIm25 3′-UTR amplified by PCR using the forward 5′-GCA AGC TTA GCT GTT CTT CTGCC-3′ and reverse 5′-GCA CTA GTG TCA TCA AAA TAT TTA TTAA-3′, respectively. The CFIm25 3′-UTR was then cloned to downstream of the luciferase gene in the pSICHECK2 vector (Promega). Mutant stem-loop forms of each miRNA, harboring two mutations in seed sequence, cloned in the lentiviral vector and named Mut-miR, which was considered as negative control. The HEK293 cells were transiently transfected with wild type or mutant miRNA-expressing vector and CFIm25-pSICHECK2 vector using lipofectamine 2000 (Invitrogen) 49 .
Luciferase assay performed after 48 h of post-transfection, using the dual-luciferase reporter assay system (Promega). Renilla luciferase signal normalized against firefly luciferase activity to monitor transfection efficiency. Data are the means of experiments performed in triplicate.
Western blot. The MDA-MB-231 cells were lysed using cell lysis buffer (50 Mm Tris, pH = 8.0, 150 mM NaCl,1% NP-40, 0.5% sodium deoxycholat, 1 mM sodium fluoride, 1 mM sodium orthovanadate, 1 mM EDTA). The polypeptide component were resolved by electrophoresis at 200 V using 12% SDS-polyacrylamide gel electrophoresis (PAGE) transferred into poly vinyl den fluoride membrane and immersed in 5% non-fat milk powder over one hour at room temperature. Upon completion of the transfer, membranes were incubated with CFIm25 and β-actin primary antibodies (Abcam). After washing, membranes probed with horseradish peroxidase-conjugated secondary antibodies (1:1,000, Abcam) and developed for detection by chemiluminescence on Kodak X-film. In order to quantify bands in western blot gel, acquired images of CFIm25 and beta actin bands with quality of 75 out of 100 compared as a matter of expression level, using Image J software. Higher number of averaged white pixels considered as higher amount of expression. www.nature.com/scientificreports/ group, MDA-MB-231 cells that were co-transduced with miR-Off-23/miR-Off-374 expressing vectors were transfected by pCDH-shCFIm25. In the third group, MDA-MB-231 cells transfected merely by pCDH-shCFIm25. At 48, 72 and 96 h after transduction/transfection, 10 μL of MTT reagent added into wells and incubated for next three hours. Next, 100 μL DMSO was added and the resulting absorbance measured at 550 nm using a multi-well spectrophotometer (Bio-Tek).
Wound healing assay. Wound healing assay performed to evaluate MDA-MB-231 cell migration. The MDA-MB-231 cells seeded in 24-well tissue culture plate until their growth rate reached a confluence of ~ 80%. The monolayer scratched gently with a 20-μL pipette tip across the center of the well. After scratching, wells washed twice with medium to remove the detached cells. Cell migration to the scratch area monitored for additional 48 h and photographed every 12 h. Number of cells moved from the border of scratch toward the empty area considered as the "migration index". Related calculations performed by ImageJ software.