MicroRNA-9 Inhibition of Cell Proliferation and Identification of Novel miR-9 Targets by Transcriptome Profiling in Breast Cancer Cells*

Background: Dysregulation of miRNAs is associated with breast cancer. Results: MiR-9 overexpression and transcriptome analysis reveals novel miR-9 targets, including MTHFD2, which can recapitulate anti-proliferative effects of miR-9 overexpression. Conclusion: MiR-9 displays tumor suppressor-like activity in breast cancer cells; MTHFD2 contributes to this activity. Significance: Understanding miR-9-directed regulation of the breast cancer transcriptome is important for diagnosis and therapeutics. Although underexpression of miR-9 in cancer cells is reported in many cancer types, it is currently difficult to classify miR-9 as a tumor suppressor or an oncomir. We demonstrate that miR-9 expression is down-regulated in MCF-7 and MDA-MB-231 breast cancer cells compared with MCF-10-2A normal breast cell line. Increasing miR-9 expression levels in breast cancer cells induced anti-proliferative, anti-invasive, and pro-apoptotic activity. In addition, microarray profiling of the transcriptome of MCF-7 cells overexpressing miR-9 identified six novel direct miR-9 targets (AP3B1, CCNG1, LARP1, MTHFD1L, MTHFD2, and SRPK1). Among these, MTHFD2 was identified as a miR-9 target gene that affects cell proliferation. Knockdown of MTHFD2 mimicked the effect observed when miR-9 was overexpressed by decreasing cell viability and increasing apoptotic activity. Despite variable effects on different cell lines, proliferative and anti-apoptotic activity of MTHFD2 was demonstrated whereby it could escape from miR-9-directed suppression (by overexpression of MTHFD2 with mutated miR-9 binding sites). Furthermore, endogenous expression levels of miR-9 and MTHFD2 displayed inverse expression profiles in primary breast tumor samples compared with normal breast samples; miR-9 was down-regulated, and MTHFD2 was up-regulated. These results indicate anti-proliferative and pro-apoptotic activity of miR-9 and that direct targeting of MTHFD2 can contribute to tumor suppressor-like activity of miR-9 in breast cancer cells.

regulate gene expression post-transcriptionally. They mediate fundamental cellular processes such as differentiation, proliferation, and apoptosis and are actively involved in carcinogenesis (1,2). Dysregulation of miRNAs can contribute to tumorigenesis (3), and miRNA expression patterns can classify tumor subtypes (4). Several miRNAs have been experimentally validated to function as oncomirs (5)(6)(7) or tumor suppressor miRNAs (8,9). MiR-9, initially shown to function in neurogenesis (10), is recently implicated in cancer. Although most evidence indicates tumor suppressor activity for miR-9 in cancer cells (11), conflicting findings exist (12). Currently it is difficult to classify miR-9 as a tumor suppressor miRNA or an oncomir. For example, miR-9 is overexpressed in brain tumors (13) and induced by MYC in breast cancer cells enhancing metastasis (12,14), and knockdown of miR-9 decreased invasion of hepatocellular carcinoma cells (15), all of which is indicative of an oncogenic potential. Conversely, in gastric cancer tissues miR-9 was shown to be down-regulated (16). Decreased expression of miR-9 has also been reported in gastric adenocarcinoma (17), ovarian cancer (11,18), and in malignant mesothelioma (19). Moreover, epigenetic inactivation of the miR-9 gene has been shown in breast cancer (20,21), colorectal cancer (22), renal cell carcinoma (23), and several other cancer cell lines and tumor tissues (24), all of which is indicative of a tumor suppressor potential. To date, few miR-9 targets of relevance to cancer have been experimentally validated (25,26): CBX7 in human glioma (27), CDX2 in gastric cancer (28), NFKB1 in ovarian cancer (18) and in gastric adenocarcinoma (17), and E-cadherin when induced by MYC (29). It remains unclear how miR-9 contributes to cancer cell growth or how cancer transcriptomes respond to loss/gain of miR-9 expression. This study investi-gates the effects of miR-9 overexpression on cancer cell growth and apoptosis and further identifies global changes and novel direct targets in the transcriptome of the MCF-7 breast cancer cell line using a microarray profiling approach. We demonstrate that several cancer-associated genes are responsive to miR-9 overexpression and that miR-9 directly targets AP3B1, CCNG1, LARP1, MTHFD1L, MTHFD2, and SRPK1 post-transcriptionally. We also demonstrate that targeting MTHFD2 can be partially responsible for the anti-proliferative activity of miR-9 in breast cancer cells.

EXPERIMENTAL PROCEDURES
Patient Samples and Cell Lines-Primary tumor (n ϭ 16) and normal breast (n ϭ 14) samples were obtained from patients at National University of Ireland Galway University hospital (supplemental Table 3). Human breast adenocarcinoma cell lines MCF-7 and MDA-MB-231 were cultured as previously described (30), and normal breast epithelial cell line MCF-10-2A was cultured based on ATCC recommendations.
Quantitative RT-PCR-Total RNA and small RNAs were isolated using Nucleospin miRNA kit (Macherey-Nagel, Germany) and reverse-transcribed using a RevertAid H-Minus First Strand cDNA Synthesis kit (Fermentas). Quantitative RT-PCR was performed using SYBR Green assays (ABI), and the reaction was run in a CFX96 system (Bio-Rad). ␤-Actin and GAPDH were used as housekeeping controls. qRT-PCR analysis of miR-9 and RNU6B (housekeeping control) were conducted using Taqman miRNA assays (Applied Biosystems). Relative quantification of expression was calculated using the 2 Ϫ⌬⌬Ct method in CFX96 system software. A t test was used to determine statistical significance.
Microarray Profiling and Data Analysis-RNA integrity number (RIN) was measured using Bioanalyzer (Agilent), and good quality samples (RIN Ͼ 8) were processed for hybridization into whole human genome Illumina bead arrays (Illumina) in biological triplicates (Wellcome Trust Sanger Institute microarray core facility). All data were variance-stabilizingtransformed and robust spline-normalized using the lumi package (Bioconductor). Differentially expressed genes were identi-fied using LIMMA (p Ͻ 0.05, signal log ratio Ͼ1.2). Putative microRNA targets were identified using MAMI microRNA target meta-predictor and TargetScan. MiR-9 seed region enrichment was identified using Sylamer. Enrichment analysis was performed using conditional hypergeometric tests.
Cell Viability and Caspase-3/7 Assays-For viability, cytotoxicity, and caspase-3/7 analyses, cells were grown in 96-well plates and 48 h post-transfection of overexpression constructs or oligonucleotides, ApoTox-Glo Triplex Assay was performed according to the manufacturer's protocol (Promega). This assay detects fluorogenic, cell-permanent peptide substrate (GF-AFC) for live-cell protease activity and cell-impermanent, fluorogenic peptide substrate (bis-AAF-R110) for dead-cell protease activity, resulting in ratiometric, inversely correlated measures of cell viability, and cytotoxicity. Each well is normalized as live/dead cells, making results comparable (well-to-well, plate-to-plate, and day-to-day) as the ratio of viable cells to dead cells is independent of cell number. This followed the addition of luminogenic DEVD-peptide substrate for caspase-3/7 and Ultra-Glo™ Recombinant Thermostable Luciferase onto cells. The luminescence output was measured with a luminometer. One M staurosporine treatment for 3 h was used as a positive control for apoptosis induction (33) to confirm the accuracy of the assay.
Wound Healing and Invasion Assays-Cells were grown to full confluence, transfected with pre-9 or pre-ctr, and wounded by a sterile pipette tip. Images of the cells were taken under an inverted microscope at 0, 24, and 48 h time points after wounding. For invasion, after transfections, 2.5 ϫ 10 4 MCF-7 or MDA-MB-231 cells (in serum starving medium; 0.1% FBS DMEM) were transferred into invasion chambers (BD Biosciences), fixed, and stained according to manufacturer's protocol. The images of invaded and stained cells were taken from an inverted microscope and counted.

RESULTS
Increasing miR-9 Levels Leads to a Decrease in Cell Viability and an Increase in Apoptosis in MCF-7 Cells-Quantitative analysis (Taqman qRT-PCR) of miR-9 expression levels in breast cancer cell lines MCF-7 and MDA-MB-231 and normal epithelial breast cell line MCF-10-2A indicated lower levels of miR-9 in cancer cell lines compared with MCF-10-2A (Fig. 1A). The effect of miR-9 overexpression on tumor cell growth was investigated. MiR-9 expression levels could be increased up to 250-fold in the pre-9 synthetic oligonucleotide-transfected MCF-7 cells compared with pre-ctr (scrambled control)-transfected cells (Fig. 1B). MiR-9-overexpressing cells led to decreased viability and increased caspase-7 activity as an indicator of apoptosis (Fig. 1C). Although staurosporine treatment did not cause a significant change in the viability of the MCF-7 cells, it was sufficient to increase caspase-7 activity as a positive control of apoptosis (Fig. 1C). MCF-7 cells lack caspase-3 (34), and previous reports indicate that activation of caspase-7 occurs independent of caspase-3 in MCF-7 cells (35,36). Hence, similar to staurosporine treatment effects observed, in the miR-9 overexpression (pre-9) caspase-7 activity was increased compared with control (pre-ctr), indicating caspase-3 independent caspase-7 activation in MCF-7 cells (Fig. 1C). In addition to anti-proliferative and pro-apoptotic effects, miR-9-overexpressing MCF-7 cells (pre-9) had slower motility (wound healing) compared with control (pre-ctr) (Fig.  1D). To further examine the phenotypic effects of miR-9 overexpression on breast cancer cells, the invasion of MCF-7 (noninvasive, ERϩ) and MDA-MB-231 (highly invasive, ERϪ) cells was tested after miR-9 overexpression. No cell invasion was observed in MCF-7 cells. However, miR-9 overexpression (pre-9) significantly decreased the invasiveness of the highly invasive cell line MDA-MB-231 compared with control (prectr) treatments (Fig. 1E).
Genome-wide Effects of miR-9 Overexpression on MCF-7 Transcriptome-To determine the global effects of miR-9 overexpression on the transcriptome of MCF-7 breast cancer cells, microarray profiling was performed. MiR-9 was delivered at two dosages, designated 1ϫ and 2ϫ. Quantitative analysis of miR-9 before the microarray analysis showed that the overexpression of miR-9 by providing a 2ϫ dosage of the microRNA (pre-9) did not increase intracellular levels by 2-fold. However, there was a significant increase in miR-9 levels up to 1.5-fold compared with the 1ϫ microRNA dosage (data not shown). Differentially expressed genes were identified for the 1ϫ and 2ϫ treatments separately (compared with their corresponding controls). A total of 516 differentially expressed genes were identified in miR-9 overexpressing MCF-7 cells compared with controls (p Ͻ 0.05, signal log ratio Ͼlog 2 1.2). Among these, 145 differentially expressed genes were specific to 1ϫ dosage, and 121 genes were specific to 2ϫ dosage treatments. 250 genes were commonly differentially expressed in 1ϫ and 2ϫ (124 up-regulated and 126 down-regulated). There was no relationship observed between -fold changes and dosages in the common differentially expressed genes, with both dosages displaying approximately the same -fold change differences (supplemental Fig. S1), possibly due to saturation of miR-9 overexpression levels. Furthermore, a direct comparison of 1ϫ (pre-9) versus 2ϫ (pre-9) was performed that identified no differential expression, indicating no observable dosage effect.
Down-regulated Genes Responsive to miR-9 Overexpression Are Enriched with miR-9 Seed Region Binding Sequences-To identify enrichment of miR-9 seed region, binding sites in the 3Ј-UTRs of differentially expressed genes Sylamer (SylArray) (37,38) analysis was performed. Briefly, all genes on the microarray were first ranked based by t-statistic. Sylamer detected significantly over-and under-represented words in the 3Ј-UTRs of the ranked list of genes. Results showed significant enrichment (p value Ͻ 1 ϫ 10 Ϫ15 ) for canonical miR-9 seeds (6-, 7-, and 8-mer) in down-regulated genes in both dosages; 7-mer enrichment is shown ( Fig. 2A). Analysis of all the differentially expressed genes (union of 1ϫ and 2ϫ; 516 genes) in the MAMI microRNA meta-predictor identified predicted miR-9 targets (41 down-regulated and 6 up-regulated) (Fig. 2B). The specific effects of direct miR-9 targeting are highlighted by the enrichment of miR-9 seed region binding sites in the 3Ј-UTRs of genes observed to be down-regulated, indicating that the post-transcriptional regulatory role of miR-9 likely involves the direct targeting of multiple genes. 25 differentially expressed genes (19 down-regulated and 6 up-regulated genes) were selected for further analysis based on genes classified for enriched cancer-related GO terms (e.g. cell proliferation, apoptosis, cell death) (supplemental Table 1) or those that are predicted targets and displayed the greatest extreme in differential expression (independent of the GO term, based on lowest p values) including 13 predicted targets of miR-9 (12 down-regulated, 1 up-regulated) (supplemental Table 2). The differential regulation of 25 genes (19 down-regulated and 6 up-regulated) including NFKB1 as a known miR-9 target (18) in response to miR-9 overexpression was confirmed by SYBR Green qRT-PCR for pre-9-treated samples compared with pre-ctr (t test; p Ͻ 0.05) (Fig. 2C).
Identification of Six Novel Targets of miR-9-In most instances microRNAs negatively regulate gene expression (i.e. reduce levels of mRNAs/proteins) (39,40). From the validated (down-regulated) predicted targets, AP3B1, CCNG1, LARP1, MTHFD1L, MTHFD2, and SRPK1 were tested for direct miR-9 targeting using a 3Ј-UTR luciferase-based reporter system (where one luciferase construct would contain a wild type 3Ј-UTR while the other construct contains a disruption of the seed region of the miR-9 binding site) (Fig. 3A). CCNG1, MTHFD2, LARP1, and SRPK1 were selected on the basis of their previous implications in breast cancer (41)(42)(43)(44), whereas AP3B1 and MTHFD1L were selected as potential biomarkers. All of these genes were predicted to have one miR-9 binding site, except MTHFD2, which has two predicted miR-9 binding sites (Fig. 3A). All the predicted sites are annotated as con-served except for the miR-9 binding site in SRPK1, which is poorly conserved (TargetScan Release 5.2 June 2011) (Fig. 3A). For the luciferase reporter analysis, in addition to construction of 3Ј-UTR constructs where wild type miR-9 binding occurs, site-directed mutations in the miR-9 seed region in full-length 3Ј-UTRs were introduced (MUT UTR) to perturb miR-9 bind- , and in MCF-7 cells transfected with pre-ctr (control precursor) or pre-9 (miR-9 precursor) synthetic oligonucleotides (B) is shown. All data were analyzed by relative quantification method (2 Ϫ⌬⌬Ct ) using RNU6B small RNA as the housekeeping control. Statistical significance was calculated using Student's t test (*, p Ͻ 0.05; **, p Ͻ 0.001). C, viability and caspase-7 activity in MCF-7 cells 48 h post-transfection with pre-9 or pre-ctr (analyzed by ApoTox-Glo Triplex Assay). Cytotoxicity (dead cell) fluorescence measurements at wavelengths 485 EXT and 520 EM were used for normalizations. Viability (live cell) fluorescence measurements at wavelengths 400 EXT and 505 EM were represented in the graph as normalized percentage (%) change (live/dead cell ratio). Caspase-7 activity luminescence measurements were represented in the graph as normalized percentage (%) change using viability (live cell) measurements in well-to-well normalization (apoptotic/live cell ratio). Three hours of staurosporine (STS) treatment (1 M) was used as a control for the assay sensitivity to detect caspase-3/7 activation as an indication of apoptosis.  AUGUST 24, 2012 • VOLUME 287 • NUMBER 35

JOURNAL OF BIOLOGICAL CHEMISTRY 29519
ing. For all the six genes tested, the wild type 3Ј-UTR (WT UTR) constructs displayed significant (p Ͻ 0.001) down-regulation in luciferase activity when co-transfected with pre-9 (normalized to pre-ctr) (Fig. 3B) in MCF-7 cells. In contrast, in the mutated 3Ј-UTR constructs (MUT UTR) luciferase activity was restored significantly, confirming the functionality of miR-9 binding sites in the 3Ј-UTR of these genes (Fig. 3B). In addition, the effect of miR-9 binding was also tested at the protein level for all six genes, which indicated miR-9 overexpression-induced reductions in protein levels for AP3B1, CCNG1, LARP1, MTHFD1L, MTHFD2, and SRPK1 (Fig. 3C). Because MTHFD2 3Ј-UTR contains two putative miR-9 binding sites and there is no detailed research reported on its activity in cancer cells, we focused on identification of the effects of MTHFD2 on three cell lines: MCF-7, MDA-MB-231 (breast cancer) , and MCF-10-2A (normal breast) (see supplemental Fig. S2 for all the MCF-10-2A results). In addition to MCF-7 cells (Fig. 3B), luciferase assay for MTHFD2 3Ј-UTR was also performed in MDA-MB-231 and MCF-10-2A cells. In MCF-10-2A cells, despite the decreased luciferase activity in wild type construct (WT), the mutated construct (MUT) displayed no significant restoration of luciferase activity (supplemental Fig. S2A). However, a slight FIGURE 2. Microarray profiling of miR-9 overexpression in MCF-7 cells and qRT-PCR validation of differentially expressed genes. A, shown is a landscape plot of Sylamer analysis for miR-9 seed (7-mer; 1A and 2) enrichment in down-regulated genes is shown. Colored lines represent miRNA seeds with the highest and lowest peaks in gene list ranked by t-statistic (x axis). B, shown is a heat map representation of differentially expressed genes in two dosages of miR-9 expression (1ϫ and 2ϫ pre-9) compared with controls (1ϫ and 2ϫ pre-ctr); each treatment was performed in triplicate, and results represent cutoffs p value 0.01 and -fold change Ͼ1.2. C, relative expression analysis (qRT-PCR) of 25 selected genes from microarray results, validating differential expression of 19 down-regulated and 6 up-regulated genes, shown as normalized relative expression ϮS.E. GAPDH and ␤-Actin were used as housekeeping controls in normalizations.
suppression at mRNA and protein levels of MTHFD2 was observed upon miR-9 overexpression in MCF-10-2A cells (supplemental Fig. S2B). In contrast, in MDA-MB-231 cells, the wild type construct (WT UTR) displayed lower luciferase activity, whereas mutated construct (MUT UTR) restored the luciferase activity significantly (Fig. 3D), suggesting that miR-9; MTHFD2 regulatory interaction occurs in two different breast cancer cell lines. The suppressive effect of miR-9 on MTHFD2 in MDA-MB-231 cells was also confirmed at mRNA and protein levels (Fig. 3E). Overall, these results indicate that AP3B1, CCNG1, LARP1, MTHFD1L, MTHFD2, and SRPK1 are direct targets of miR-9, negatively regulated by miR-9 at the mRNA and protein levels.
Overexpression of MTHFD2 with Mutated miR-9 Binding Sites Displays Proliferative Activity-To test for opposite effects to those observed in MTHFD2 knockdown and to identify the contribution of miR-9-directed regulation on the potential oncogenic activity of MTHFD2, overexpression constructs for MTHFD2 with wild type 3Ј-UTR (pCMV-MTHFD2-UTR) or 3Ј-UTR with mutated miR-9 binding sites (pCMV-MTHFD2-UTRmut) or empty vector control (pCMV-AC,EV) (Fig. 5A) were transfected in MCF-7, MDA-MB-231, and MCF-10-2A cells. The overexpression of MTHFD2 (via both wild type 3Ј-UTR and mutated 3Ј-UTR) was detected at mRNA and/or protein levels in all the cells (Fig. 5, A and B) (supplemental Fig. S2E for MCF-10-2A). The MCF-10-2A cells ectopically overexpressing miR-9 (pre-9) and MTHFD2 with mutated miR-9 binding sites (UTRmut) displayed higher MTHFD2 expression that may indicate the escape of MTHFD2 from miR-9-directed suppression due to the mutated binding sites. However, the viability and caspase-3/7 activity displayed no significant change upon MTHFD2 overexpression (supplemental Fig. S2F). In contrast, despite the similar ectopic expression profile of MTHFD2 in UTR-or UTRmut-transfected samples, both MCF-7 and MDA-MB-231 cells displayed significantly greater viability and reduced apoptosis when MTHFD2 was overexpressed with mutated miR-9 binding sites   AUGUST 24, 2012 • VOLUME 287 • NUMBER 35 (UTRmut) (Fig. 5C). These findings highlight the significant contribution of miR-9-directed regulation on reducing the proliferative and anti-apoptotic capacity of MTHFD2. Furthermore, compared with control (EV), the invasiveness of MDA-MB-231 cells was increased when MTHFD2 was overexpressed with wild type 3Ј-UTR (UTR) (Fig. 5C), and this effect was greater (i.e. more invasiveness) when MTHFD2 was overexpressed with ablated miR-9 binding sites (UTRmut) (Fig. 5C), indicating that miR-9-directed regulation can limit the proinvasive potential of MTHFD2.

MiR-9 Inhibits Proliferation of Breast Cancer Cells
MiR-9 and MTHFD2 Show Inverse Expression Levels in Primary Breast Cancer Samples-To test whether miR-9 and MTHFD2 expression levels are inversely correlated in clinical cancer samples, expression levels of miR-9 and MTHFD2 in 30 samples (16 primary breast cancer samples (8 non-metastatic tumor and 8 metastatic tumor) and 14 normal breast samples) were analyzed. The relative -fold changes of miR-9 and MTHFD2 in tumor samples were normalized against the mean expression of normal breast tissue samples, which were set to one and are presented as relative-fold changes (log 2 Ϯ S.E.) (Fig. 6A). In six of eight non-metastatic samples, miR-9 and MTHFD2 showed an inverse expression profile, whereas in metastatic samples the trend was less obvious; three of eight samples showed an inverse expression profile. Furthermore, two publically available datasets, Richardson_breast_2 (n ϭ 47) (45) and Zhao_breast (n ϭ 64) (46), available through the Oncomine TM data base (v4.4 Compendia Bioscience, Ann Arbor, MI) (47), confirmed higher expression levels of MTHFD2 in breast tumor samples compared with normal breast tissue (Fig. 6B). This is consistent with increased MTHFD2 expression levels in breast cancer that was recently reported in a proteomics study with 21T breast cancer cell lines (derived from a single patient with metastatic breast cancer) (42).

DISCUSSION
The role of miR-9 in cancer biology is not well understood. This study demonstrates that endogenous miR-9 levels are lower in breast cancer cells compared with normal cells. It also provides evidence of direct miR-9 targets previously implicated in cancer. Although miR-9 overexpression has been associated with colorectal cancer metastasis (48), its significant down-regulation in gastric and clear cell renal carcinomas (16,23) has also been reported. These conflicting findings on miR-9 expression (down-regulation or overexpression in cancer cells) may be due to differential mature miR-9 processing from three dif- ferent primary precursors where down-regulation of miR-9 due to aberrant hypermethylation of any of the miR-9-1, miR-9-2, and miR-9-3 precursor regions have been reported in independent studies in different cancer types (23,24,49). Our findings that endogenous miR-9 levels are lower in breast cancer cells compared with normal cells (for both cell lines and primary tumor samples) is similar to previous reports of lower levels of miR-9 associated with ovarian cancer, gastric carcinoma, and clear cell renal carcinomas (16,23).
Changes in cell proliferation and apoptosis are key phenotypes observed in malignant transformation (50). Our results demonstrate that miR-9 overexpression decreases cell viability and increases apoptosis in MCF-7 breast cancer cells. As a caspase-3 null cell line (51), previous studies have shown that DEVDase activity can be increased in MCF-7 cells independent of caspase-3 whereby caspase-7 cleaves DEVD substituting caspase-3 (35,(52)(53)(54). In this study caspase-7 activity was increased, indicating pro-apoptotic activity of miR-9 in MCF-7 cells. The anti-proliferative and pro-apoptotic effects of miR-9 are supportive of a tumor suppressor-like role of miR-9 in the context of breast cancer cell proliferation, migration, and apoptosis.
Invasiveness is one of the hallmarks of cancer (55). Overexpression of some miRNAs (e.g. miR-21) has been shown to increase the invasiveness of normal and cancer cells (56,57). In some instances non-invasive cells can be transformed into invasive cells, whereas in other cases the invasiveness of a cell can be increased due to changes in the expression levels of specific miRNAs (or genes) (58,59). Although our study demonstrates no effect of miR-9 overexpression on the invasiveness of the non-invasive MCF-7 cell line, the overexpression of miR-9 elicited a significant reduction in the invasiveness of the highly invasive MDA-MB-231 cell line. The lack of invasion observed in MCF-7 cells could be due to the poorly invasive nature of MCF-7 cells (60). The reduction of invasion observed in the MDA-MB-231 cell line could be due to miR-9-directed downregulation or inactivation of pro-invasive genes and indicates that miR-9 levels can affect invasiveness. Although miR-9 has been implicated in cancer biology in several studies, no genome-wide transcriptome analysis of miR-9 expression effects in cancer cells has been reported to date. This study demonstrates that miR-9 overexpression in breast cancer cells in two dosages (1ϫ and 2ϫ) identifies 250 genes whose expression levels are responsive to miR-9, including 47 predicted miR-9 targets. The cancer relevance of miR-9-induced effects on the MCF-7 transcriptome is highlighted by cancer-related GO terms (e.g. cell proliferation, apoptosis) being enriched in the genes differentially regulated.
The identification of predicted miR-9 target genes displaying reduced expression levels in MCF-7 cells when miR-9 is overexpressed suggests such genes are potential direct targets of miR-9. This study identifies and validates six genes as novel direct miR-9 targets, including five targets with previously identified roles in cancer (AP3B1, CCNG1, LARP1, MTHFD2, and SRPK1) and one target that was not previously implicated in cancer (MTHFD1L) in MCF-7 cells. MTHFD2 was tested also in MDA-MB-231 and MCF-10-2A cells and confirmed as a direct target of miR-9 in MDA-MB-231 cells in addition to MCF-7. Interestingly, in MCF-10-2A cells luciferase activity remained suppressed in the construct with mutated miR-9 binding sites, indicating no significant miR-9 contribution on MTHFD2 suppression in these cells, which may be a result of different cellular context in combination with other regulatory relationships.
For CCNG1, LARP1, and SRPK1, a functional role in cancer has been previously reported by knockdown experiments. CCNG1 (cyclin-G1), a p53 target, has been implicated in cancer (41,(61)(62)(63)(64), and silencing of CCNG1 displayed decreased cell growth rate and invasion in hepatocarcinoma cells (61). LARP1 (La ribonucleoprotein domain family, member 1), functions in vesicle transport. Silencing of LARP1 increased apoptosis in HeLa cells (43). SRPK1 (SRSF protein kinase 1) functions in alternative splicing and has been implicated in many cancer types (44,(65)(66)(67)(68). Silencing of SRPK1 increased apoptotic potential in breast and colonic tumor cell lines (44) and decreased proliferative capacity in leukemic cells (69). The observed reduction in mRNA and protein levels highlights that due to direct targeting by miR-9, and the reduced activity of these oncogenic targets can potentially lead to suppression of tumorigenesis.
In addition, our study also provides the first knockdown experiments for three additional direct targets of miR-9, namely AP3B1, MTHFD1L, and MTHFD2. We considered that the lack of experimental evidence of loss-of-function effects of these genes on cancer cells made them good candidates to investigate their possible involvement in breast carcinogenesis using a siRNA approach. AP3B1 (AP-3 complex subunit beta-1), involved in organelle biogenesis, has been investigated in relation to cancer in only a few studies (70,71). MTHFD1L has been previously implicated in neural tube defects (72). MTHFD2, a mitochondrial bifunctional enzyme, is known to involve in folate metabolism (73), and loss-or gain-of function studies have not been described yet. A possible link between MTHFD2 and cancer, such as a SNP in MTHFD2 associated with increased risk of bladder cancer (74), and a correlation of MTHFD2 expression with artesunate resistance in tumor cell lines (75) have been reported.
The individual effects of knockdown of MTHFD1L and AP3B1 showed only a small 15% decrease in viability in MCF-7 cells. In contrast, the knockdown of MTHFD2 significantly decreased cell viability and increased apoptosis in both MCF-7 and MDA-MB-231 cells, mimicking the phenotypic effect of miR-9 overexpression. Apart from the significant increase in caspase-3/7 activation in MCF-10-2A cells, no significant effect was observed in viability of MCF-10-2A cells or viability and caspase-3/7 activity of MDA-MB-231 cells after MTHFD2 knockdown. Similar findings indicating that overexpression or knockdown of a gene may not have an impact on cell proliferation but can affect cancer cell invasion have been described (76). Although the extent of the effect of MTHFD2 knockdown may not completely account for the overall extent of the phenotypic effects of miR-9, it is clear that knockdown of MTHFD2 can contribute to the phenotypic effects as seen for miR-9 overexpression in MCF-7 cells.
The overexpression of MTHFD2 with the wild type (UTR) or mutated (UTRmut) miR-9 binding sites displayed similar expression levels. MTHFD2 overexpression with mutated miR-9 binding sites (UTRmut) could be expected to display higher MTHFD2 levels due to "escaping" from miR-9-directed suppression. In the UTR-transfected cells ectopically expressed miR-9 can target both endogenous and ectopically introduced MTHFD2, whereas in UTRmut-transfected cells miR-9 can only target endogenous MTHFD2. We consider that due to high endogenous levels of MTHFD2, ectopically expressed miR-9 can still be captured by the endogenous MTHFD2, resulting in overall similar MTHFD2 expression compared with UTRmut. Potentially, much greater miR-9 ectopic expression in UTR-transfected cells should be sufficient for suppression of both endogenous and ectopically expressed MTHFD2, resulting in less MTHFD2 expression compared with UTRmuttransfected cells. MTHFD2 overexpression with wild type 3Ј-UTR (UTR) had no effect on viability or apoptosis compared with EV control. In contrast, MTHFD2 lacking miR-9 binding sites (UTRmut) displayed increased viability and decreased apoptosis. This effect may indicate that the escape of MTHFD2 from miR-9-directed suppression potentiates its anti-proliferative activity. Indeed, the increased invasiveness observed for MDA-MB-231 cells during wild type MTHFD2 3Ј-UTR (UTR) overexpression and the highest invasiveness observed with MTHFD2-mutated 3Ј-UTR (UTRmut) highlight a contribution of miR-9 in balancing MTHFD2 levels (via suppression) to reduce invasiveness. Overall, the results suggest that, despite the variable effects of MTHFD2 on cell viability, apoptosis, and invasion, the decreased MTHFD2 levels may be partially responsible for the effects observed during miR-9 overexpression in MCF-7 cells. In addition, MTHFD2 escape of the miR-9-directed suppression (3Ј-UTR with no miR-9 binding sites) leads to a trend toward proliferative, anti-apoptotic, and proinvasive activity, indicating that miR-9-directed MTHFD2 regulation is a contributor to anti-proliferative activity of miR-9.
The results in our study suggest a tumor suppressor-like activity of miR-9 when ectopically expressed in breast cancer cells. Moreover, transcriptome profiling of miR-9 overexpression identifies six novel direct targets of miR-9, including MTHFD2 (which when subject to knockdown recapitulated the anti-proliferative and pro-apoptotic activity of miR-9 overexpression in MCF-7 breast cancer cells). Furthermore, overexpression of MTHFD2-escaping miR-9 regulation suggests that miR-9 contributes to suppression of the potential oncogenic activity of MTHFD2. Taken together, our study highlights underexpression of miR-9 and overexpression of MTHFD2 in breast tumor tissues as potential expression level biomarkers. The expression levels of MTHFD2 can be controlled (downregulated) via increasing miR-9 expression in cancer cells, thereby reducing cell viability and enhancing apoptosis. In addition to the individual and negatively correlated expression profiles of miR-9 and MTHFD2, the interaction between miR-9 and MTHFD2 is significant in breast cancer cells and next will be subjected to further screening in population-size sample groups and possibly in combination with other miRNAs, which display promise as targets for breast cancer diagnosis and therapeutics.