Dual Strands of Pre-miR-149 Inhibit Cancer Cell Migration and Invasion through Targeting FOXM1 in Renal Cell Carcinoma

Our recent studies revealed that dual strands of certain pre-microRNAs, e.g., pre-miR-144, pre-miR-145, and pre-miR-150, act as antitumor microRNAs (miRNAs) in several cancers. The involvement of passenger strands of miRNAs in cancer pathogenesis is a novel concept in miRNA research. The analysis of a miRNA expression signature in clear cell renal cell carcinoma (ccRCC) has revealed that the guide strand of pre-miR-149 is significantly downregulated in cancer tissues. The aims of this study were to investigate the functional significance of miR-149’s guide strand (miR-149-5p) and passenger strand (miR-149-3p), and to identify the oncogenic genes regulated by these miRNAs in ccRCC cells. The ectopic expression of these miRNAs significantly inhibited cancer cell migration and invasion in ccRCC cells. Forkhead box protein M1 (FOXM1) was directly regulated by miR-149-5p and miR-149-3p in ccRCC cells. Knockdown studies using si-FOXM1 showed that the expression of FOXM1 enhanced RCC cell aggressiveness. Interestingly, the analysis of a large number of patients in the The Cancer Genome Atlas (TCGA) database (n = 260) demonstrated that patients with high FOXM1 expression had significantly shorter survival than did those with low FOXM1 expression (p = 1.5 × 10−6). Taken together, dual strands of pre-miR-149 (miR-149-5p and miR-149-3p) acted as antitumor miRNAs through the targeting of FOXM1 in ccRCC cells.


Introduction
Renal cell carcinoma (RCC) is the most common neoplasm of the kidney, and approximately 70-80% of RCCs are classified as clear cell-type (ccRCC) [1]. More than 270,000 patients are diagnosed with RCC every year, which is estimated to constitute approximately 3% of adult malignancies. The incidence and mortality rates of RCC are increasing at a rate of 2-3% per decade [2]. Although surgical resection remains the only known curative treatment for localized RCC, a significant percentage of these patients develop relapses or metastatic disease, which are associated with a poor prognosis [3]. Recently developed targeted molecular therapies and immunomodulatory agents are currently being widely used for patients with metastatic or recurrent RCC [4]. However, those curative approaches are restricted to RCC patients in advanced stages of disease and the 5-year survival rate of the patients is only 5-10% [5]. Therefore, to improve outcomes in patients with RCC, it is necessary to fully elucidate the molecular mechanisms of metastatic RCC based on new genomic approaches.
MicroRNAs (miRNAs) belong to a class of noncoding RNA molecules that fine-tune the expression of protein coding/noncoding RNAs by repressing translation or cleaving RNA transcripts in a sequence-dependent manner [6]. Overexpression of oncogenic miRNAs and dysfunction of antitumor miRNAs are associated with human cancer pathogenesis [7]. A single miRNA species can regulate the expression of hundreds or thousands of different mRNAs, and an individual species of mRNA can be regulated by multiple different miRNAs in normal cells [8]. Therefore, aberrantly expressed miRNAs can disrupt regulatory miRNA-mRNA networks in cancer cells.
The construction of expression signatures of miRNAs in human cancers is an effective strategy to identify aberrantly expressed miRNAs in cancer cells. We have constructed miRNA signatures in several types of human cancers, including RCC [9][10][11][12]. From analyses of the signatures, we have found that some passenger strands of miRNAs were significantly reduced in cancer tissues, suggesting that passenger strands of miRNAs have an antitumor function in cancer cells, as do guide strands of miRNAs. In miRNA biogenesis, one strand of mature miRNA (the guide strand) is loaded into the miRNA-induced silencing complex (RISC) that targets mRNA degradation and translational repression in processing bodies [13]. In contrast, the passenger strand of miRNA was previously thought to be degraded and to have no function [14][15][16]. Interestingly, our recent studies demonstrated that the dual strands (guide strand and passenger strand) of certain pre-miRNAs, such as pre-miR-144, pre-miR-145, pre-miR-139, and pre-miR-150, acted as antitumor miRNAs through the targeting of several oncogenic genes in several cancers [17][18][19]. Novel approaches to the analysis of dual strand miRNA-regulated RNA networks in cancer cells may provide new insights into the pathogenic development of human cancers.
Our original miRNA signature of RCC led us to focus on the dual strands of pre-miR-149, i.e., miR-149-5p (the guide strand) and miR-149-3p (the passenger strand) [9]. The aims of this study were to investigate the functional significance of these miRNAs and to identify coordinately regulated oncogenic genes in RCC cells. The identification of the function of passenger strands of miRNAs and novel mechanisms of miRNA-mediated gene regulation enhance our understanding of the molecular pathways underlying RCC initiation, development, and metastasis.

Effects of Ectopic Expression of miR-149-5p and miR-149-3p on Cell Proliferation, Migration, and Invasion Assays in RCC Cell Lines
To examine the functional roles of miR-149-5p and miR-149-3p, we performed gain-of-function studies by using A498 and 786-O cells transfected with mature miRNAs.

Effects of Ectopic Expression of miR-149-5p and miR-149-3p on Cell Proliferation, Migration, and Invasion Assays in RCC Cell Lines
To examine the functional roles of miR-149-5p and miR-149-3p, we performed gain-of-function studies by using A498 and 786-O cells transfected with mature miRNAs.

Both miR-149-5p and miR-149-3p Bind to Ago2
We hypothesized that both miR-149-5p and miR-149-3p may be incorporated into and function as part of the RISC. To test this hypothesis, we performed immunoprecipitation with antibodies targeting Ago2, which plays a central role in the RISC (Figure 2A). After transfection with miR-149- and miR-149-3p in RCC clinical specimens were determined by qRT-PCR. Data were normalized to RNU48 expression; (C) Correlation between the relative expression levels of miR-149-5p and miR-149-3p; (D) Cell proliferation was determined by XTT assays. * p < 0.0001; (E) Cell migration activity was determined by wound-healing assays. * p < 0.0001; (F) Cell invasion activity was determined using Matrigel invasion assays. * p < 0.0001.

Both miR-149-5p and miR-149-3p Bind to Ago2
We hypothesized that both miR-149-5p and miR-149-3p may be incorporated into and function as part of the RISC. To test this hypothesis, we performed immunoprecipitation with antibodies targeting Ago2, which plays a central role in the RISC (Figure 2A). After transfection with miR-149-5p or miR-149-3p, Ago2-bound miRNAs were isolated, and qRT-PCR was carried out to determine whether miR-149-5p and miR-149-3p were bound to Ago2 ( Figure 2B).

Screening of Target Genes Regulated by miR-149-5p and miR-149-3p in RCC Cells
Next, we sought to obtain further insights into the molecular mechanisms regulated by antitumor miR-149-5p and miR-149-3p in RCC cells. Thus, we screened the genes regulated by those miRNAs using in silico and genome-wide gene expression analyses. First, we performed in silico analyses. An analysis of the TargetScan database showed that 4750 genes and 7832 genes had putative target sites for miR-149-5p and miR-149-3p, respectively, in their 3′-UTRs. Next, we merged the data for the gene expression analysis data in miR-149-5p and miR-149-3p transfectants (GSE100746). Finally, we found 14 genes that were upregulated (fold-change log2 > 1.0) in cancer tissues by Gene Expression Omnibus (GEO) database analyses (GEO accession number: GSE22541 and GSE36895). We examined the Kaplan-Meier plot for these genes in RCC using the OncoLnc and cBioPortal database [20][21][22]. Our strategy for analysis is shown in Figure 3. The putative target genes regulated by dual strands of miR-149, i.e., miR-149-5p and miR-149-3p, are summarized in Table 2. Among them, we focused on FOXM1 because it showed the most significant difference in log rank tests using the OncoLnc database.

Screening of Target Genes Regulated by miR-149-5p and miR-149-3p in RCC Cells
Next, we sought to obtain further insights into the molecular mechanisms regulated by antitumor miR-149-5p and miR-149-3p in RCC cells. Thus, we screened the genes regulated by those miRNAs using in silico and genome-wide gene expression analyses. First, we performed in silico analyses. An analysis of the TargetScan database showed that 4750 genes and 7832 genes had putative target sites for miR-149-5p and miR-149-3p, respectively, in their 3 -UTRs. Next, we merged the data for the gene expression analysis data in miR-149-5p and miR-149-3p transfectants (GSE100746). Finally, we found 14 genes that were upregulated (fold-change log2 > 1.0) in cancer tissues by Gene Expression Omnibus (GEO) database analyses (GEO accession number: GSE22541 and GSE36895). We examined the Kaplan-Meier plot for these genes in RCC using the OncoLnc and cBioPortal database [20][21][22]. Our strategy for analysis is shown in Figure 3. The putative target genes regulated by dual strands of miR-149, i.e., miR-149-5p and miR-149-3p, are summarized in Table 2. Among them, we focused on FOXM1 because it showed the most significant difference in log rank tests using the OncoLnc database.

Analysis of FOXM1 Expression in ccRCC Clinical Specimens by qRT-PCR and Immunohistochemistry
We used qRT-PCR and immunohistochemical staining to examine the expression levels of FOXM1 in ccRCC specimens. The expression of FOXM1 was significantly upregulated in cancer tissues compared with normal tissues (p < 0.0001; Figure 4A). Spearman's rank tests showed a negative correlation between the expression levels of FOXM1 and miR-149-5p (p = 0.0347 r = −0.373; Figure 4B), but there was no correlation between FOXM1 and miR-149-3p ( Figure 4C). FOXM1 protein expression was strongly expressed in several cancer tissues (patient No. 9,19,20), whereas low expression was observed in normal tissues using a tissue microarray ( Figure 4D).

TCGA Database Analysis of Dual Strands of Pre-miR-149 and FOXM1
We analysed Kaplan-Meier overall survival (OS) curves according to the expression levels of dual strands of pre-miR-149 and FOXM1 and the relationships between FOXM1 expression and tumor stage and metastasis in ccRCC using the TCGA-KIRC database. The TCGA dataset showed that high expression of miR-149-5p was associated with a poor prognosis of the patients with RCC. The miR-149-3p data was not available in this database (data not shown). The Kaplan-Meier curves for OS rates showed that the group with high expression of FOXM1 had a significantly shorter survival than the low expression group in ccRCC (p = 1.5 × 10 −6 , Figure 5A). The expression levels of FOXM1 were significantly increased in advanced T stage cases and metastatic cases ( Figure 5B,C).

Analysis of FOXM1 Expression in ccRCC Clinical Specimens by qRT-PCR and Immunohistochemistry
We used qRT-PCR and immunohistochemical staining to examine the expression levels of FOXM1 in ccRCC specimens. The expression of FOXM1 was significantly upregulated in cancer tissues compared with normal tissues (p < 0.0001; Figure 4A). Spearman's rank tests showed a negative correlation between the expression levels of FOXM1 and miR-149-5p (p = 0.0347 r = −0.373; Figure 4B), but there was no correlation between FOXM1 and miR-149-3p ( Figure 4C). FOXM1 protein expression was strongly expressed in several cancer tissues (patient No. 9,19,20), whereas low expression was observed in normal tissues using a tissue microarray ( Figure 4D).

TCGA Database Analysis of Dual Strands of Pre-miR-149 and FOXM1
We analysed Kaplan-Meier overall survival (OS) curves according to the expression levels of dual strands of pre-miR-149 and FOXM1 and the relationships between FOXM1 expression and tumor stage and metastasis in ccRCC using the TCGA-KIRC database. The TCGA dataset showed that high expression of miR-149-5p was associated with a poor prognosis of the patients with RCC. The miR-149-3p data was not available in this database (data not shown). The Kaplan-Meier curves for OS rates showed that the group with high expression of FOXM1 had a significantly shorter survival than the low expression group in ccRCC (p = 1.5 × 10 −6 , Figure 5A). The expression levels of FOXM1 were significantly increased in advanced T stage cases and metastatic cases ( Figure 5B,C).

Effects of Silencing FOXM1 on Cell Proliferation, Migration, and Invasion in RCC Cells
We evaluated the knockdown efficiency of si-FOXM1 transfection in RCC cells. Our present data showed that si-FOXM1 transfection effectively downregulated FOXM1 expression in A498 and 786-O cells (Figure 7A,B). Functional assays demonstrated that cell proliferation, migration, and invasion were all inhibited in si-FOXM1 transfectants compared with mock-or miR control-transfected cells (p < 0.0001, Figure 7C-E).

Effects of Silencing FOXM1 on Cell Proliferation, Migration, and Invasion in RCC Cells
We evaluated the knockdown efficiency of si-FOXM1 transfection in RCC cells. Our present data showed that si-FOXM1 transfection effectively downregulated FOXM1 expression in A498 and 786-O cells ( Figure 7A,B). Functional assays demonstrated that cell proliferation, migration, and invasion were all inhibited in si-FOXM1 transfectants compared with mock-or miR control-transfected cells (p < 0.0001, Figure 7C-E). GAPDH was used as an internal control; (B) FOXM1 protein expression in RCC cell lines was evaluated by Western blot analysis. GAPDH was used as a loading control; (C) Cell proliferation was determined using XTT assays, * p < 0.0001; (D) Cell migration activity was determined by woundhealing assays, * p < 0.0001; (E) Cell invasion activity was determined using Matrigel invasion assays. * p < 0.0001.

Kaplan-Meier Survival Curves and Genes Affected by miR-149-5p and miR-149-3p
To investigate the contribution of miR-149-5p and miR-149-3p to the coordinate regulation of genes (Table 2) in ccRCC pathogenesis, we searched the TCGA database. Among 14 genes, the expression of 6 genes (FOXM1, SIPA1, CXCL5, LMNB2, GBP2, and APOBEC3C) was significantly associated with a poor prognosis of patients with ccRCC ( Figure 8). GAPDH was used as an internal control; (B) FOXM1 protein expression in RCC cell lines was evaluated by Western blot analysis. GAPDH was used as a loading control; (C) Cell proliferation was determined using XTT assays, * p < 0.0001; (D) Cell migration activity was determined by wound-healing assays, * p < 0.0001; (E) Cell invasion activity was determined using Matrigel invasion assays. * p < 0.0001.

Discussion
Currently approved targeted therapies have improved the management of patients with metastatic RCC. However, the use of these targeted therapies is restricted to advanced stages of RCC [23]. The elucidation of molecular mechanisms of relapse, metastasis and drug resistance is indispensable to improve the prognosis of RCC patients. To identify novel pathologic molecular mechanisms in RCC, our research group has identified RCC oncogenic pathways based on antitumor miRNAs that were identified by examining RCC miRNA signatures [9]. For example, lysyl oxidase homolog 2 (LOXL2) was overexpressed in RCC clinical specimens. Importantly, the antitumor miR-29 family (miR-29a/b/c) targeted LOXL2 and inhibited cancer cell migration and invasion [24]. More recently, it was found that miR-101 was significantly reduced in sunitinib-treated RCC tissues and the ectopic expression of miR-101 inhibited RCC cell aggressiveness through its targeting of ubiquitin-like with PHD and ring finger domains 1 (UHRF1) [9]. Our antitumor miRNAs-based strategy is an attractive approach for the identification of novel regulatory networks in cancer cells.
Our present data demonstrated that dual strands of pre-miR-149, miR-149-5p and miR-149-3p, acted as antitumor miRNAs in RCC cells. The antitumor function of miR-149-5p was reported in several other types of cancer [25,26]. In breast cancer, miR-149-5p was downregulated by the hypermethylation of its promoter region and was found to be involved in Adriamycin-resistant breast cancer cells targeting GlcNAc N-deacetylase/N-sulfotransferase-1 (NDST1) [27]. In hepatocellular carcinoma, miR-149-5p was significantly downregulated and its expression was correlated with distant metastases and TNM classification [26]. An overexpression of miR-149-5p suppressed the migration and invasion activities of hepatocellular carcinoma cells [26]. In colorectal carcinoma, the expression of miR-149-5p regulated cell growth, migration, and invasion through its targeting of the

Discussion
Currently approved targeted therapies have improved the management of patients with metastatic RCC. However, the use of these targeted therapies is restricted to advanced stages of RCC [23]. The elucidation of molecular mechanisms of relapse, metastasis and drug resistance is indispensable to improve the prognosis of RCC patients. To identify novel pathologic molecular mechanisms in RCC, our research group has identified RCC oncogenic pathways based on antitumor miRNAs that were identified by examining RCC miRNA signatures [9]. For example, lysyl oxidase homolog 2 (LOXL2) was overexpressed in RCC clinical specimens. Importantly, the antitumor miR-29 family (miR-29a/b/c) targeted LOXL2 and inhibited cancer cell migration and invasion [24]. More recently, it was found that miR-101 was significantly reduced in sunitinib-treated RCC tissues and the ectopic expression of miR-101 inhibited RCC cell aggressiveness through its targeting of ubiquitin-like with PHD and ring finger domains 1 (UHRF1) [9]. Our antitumor miRNAs-based strategy is an attractive approach for the identification of novel regulatory networks in cancer cells.
Our present data demonstrated that dual strands of pre-miR-149, miR-149-5p and miR-149-3p, acted as antitumor miRNAs in RCC cells. The antitumor function of miR-149-5p was reported in several other types of cancer [25,26]. In breast cancer, miR-149-5p was downregulated by the hyper-methylation of its promoter region and was found to be involved in Adriamycin-resistant breast cancer cells targeting GlcNAc N-deacetylase/N-sulfotransferase-1 (NDST1) [27]. In hepatocellular carcinoma, miR-149-5p was significantly downregulated and its expression was correlated with distant metastases and TNM classification [26]. An overexpression of miR-149-5p suppressed the migration and invasion activities of hepatocellular carcinoma cells [26]. In colorectal carcinoma, the expression of miR-149-5p regulated cell growth, migration, and invasion through its targeting of the EPH receptor B3 gene (EphB3), and the knockdown of EphB3 inhibited tumor growth by in vivo assays [28].
In contrast to miR-149-5p, few reports have described the function of miR-149-3p in cancer cells. Dioscin is a natural product that induces apoptosis and suppresses tumor growth in pancreatic cells [29]. The expression level of miR-149-3p was upregulated by dioscin, and miR-149-3p inhibited the Akt1 signaling pathway [29]. A major bioactive component in licorice root is 18β-glycyrrhetinic acid (GRA). It reportedly possesses antitumor effects in several cancers [30][31][32]. In gastric cancer tissues, miR-149-3p was downregulated, but its expression was upregulated by GRA treatment [33]. The overexpression of miR-149-3p inhibited gastric cancer cell proliferation and cell cycle progression [33]. In these analyses, we did not detect the downregulation of the passenger strand of miR-149-3p in RCC tissues. However, the ectopic expression of miR-149-3p inhibited cancer cell aggressiveness in RCC cells. This fact indicates that the passenger strand of miR-149-3p may possess cancer-suppressing functions in RCC cells. These findings strongly suggest that dual strands of pre-miR-149 act as antitumor miRNAs in human cancers.
In this study, we focused on FOXM1 as a target of the dual strands of pre-miR-149 and validated the direct binding of these miRNAs to the 3 -UTR using luciferase reporter assays. Past studies have shown that FOXM1 was regulated by miR-149-5p in colorectal cancer and non-small cell lung cancer [36,37]. This is the first study to report that dual strands of pre-miR-149 coordinately regulate FOXM1 in RCC cells. FOXM1 is a member of the forkhead transcription family of proteins, and it plays pivotal roles in cell cycle progression in normal cells [38]. An overexpression of FOXM1 has been detected in a broad range of human cancers, including RCC [39][40][41]. Our present data demonstrated that the knockdown of FOXM1 in tumor cells by RNAi reduced cell proliferation, migration, and invasion, suggesting that FOXM1 acts as an oncogene in RCC cells. Moreover, recent meta-analyses have indicated that high expression of FOXM1 was significantly associated with poor overall survival in most solid tumors [42,43]. Taken together, the expression of FOXM1 is a useful biomarker for the prognosis of human cancers and a potential therapeutic target in RCC cells. Authors should discuss the results and how they can be interpreted in perspective of previous studies and of the working hypotheses. The findings and their implications should be discussed in the broadest context possible. Future research directions may also be highlighted.

Clinical ccRCC Specimens
Clinical ccRCC specimens were obtained from patients who were admitted to Chiba University Hospital and had undergone radical nephrectomy from 2012 to 2015. A total of 16 pairs of ccRCC specimens and adjacent non-cancerous specimens were collected. The clinicopathological characteristics of the patients are summarized in Table 1. These samples were staged according to the UICC TNM classification [44]. All patients in this study provided written informed consent for tissue donation for research purposes. The protocol was approved by the Institutional Review Boards of Chiba University Hospital (identification code: No. 484, 30 August 2011).

Tissue Collection and Cell Culture
Clinical specimens were immersed in RNAlater (Thermo Fisher Scientific, Waltham, MA, USA) and stored at 4 • C until RNA was extracted. Human ccRCC cells (A498 and 786-O cells) were obtained from the American Type Culture Collection (Manassas, VA, USA).

Quantitative Real-Time Reverse Transcription Polymerase Chain Reaction (qRT-PCR)
Stem-loop RT-PCR (TaqMan MicroRNA Assays; product ID: 002255 for miR-149-5p and 002164 for miR-149-3p; Applied Biosystems, Foster City, CA, USA) was used for these assays. TaqMan probes and primers for FOXM1 (product ID: Hs01073586_m1; Applied Biosystems) were assay-on-demand gene expression products. We used GUSB (product ID: Hs00939627_m1; Applied Biosystems), GAPDH (product ID: Hs02758991_g1; Applied Biosystems), and RNU48 (product ID: 001006; Applied Biosystems) as internal controls. We first verified the transfection efficiency of miRNA in cell lines based on downregulation of TWF1 (PTK9) mRNA following transfection with miR-1 (This method was recommended by the manufacturer).

Cell Proliferation, Migration, and Invasion Assays
A498 and 786-O cells were transfected with 10 nM miRNAs or siRNAs by reverse transfection. Cell proliferation was determined by XTT assay using a Cell Proliferation Kit II (Sigma-Aldrich, St. Louis, MO, USA). Cell migration was evaluated with wound healing assays. Cell invasion was analyzed using modified Boyden chambers containing Transwell-precoated Matrigel membrane filter inserts. These assays were performed as described previously [45][46][47].

miRNA Incorporated into RISC by Ago2 Immunoprecipitation
To confirm that exogenous miR-149-5p or miR-149-3p were incorporated into RISC, we performed immunoprecipitation assays using a microRNA Isolation Kit, Human Ago2 (Wako, Osaka, Japan) as described previously [19]. The expression levels of miRNAs bound to Ago2 were measured by TaqMan RT-qPCR. The miRNA expression data were normalized to the expression of miR-26a (product ID: 000404; Applied Biosystems), which was not affected by miR-149-5p or miR-149-3p expression.

TCGA Database Analysis of ccRCC
To investigate the clinical significance of target genes, we used the TCGA database from OncoLnc and cBioPortal (http://www.cbioportal.org/) [20][21][22]. We analyzed Kaplan-Meier survival curves and log-rank tests for target genes, and evaluated the lower expression group and the higher expression group of FOXM1 by T stage and metastatic cases.

Immunohistochemistry Using Tissue Microarrays
We used a tissue microarray of renal cell carcinoma samples obtained from US Biomax (Derwood, MD, USA; cat no. KD806), which contained a total of 80 renal tissues (clear cell carcinoma, n = 68; carcinoma sarcomatodes, n = 2; normal renal samples, n = 10). Detailed information on these samples is available at http://www.biomax.us/tissue-arrays/Kidney/KD806.

Statistical Analysis
The relationships between the two groups and expression values obtained by RT-PCR were analyzed using Mann-Whitney u-tests. The correlations between miR-149-5p and miR-149-3p expression were evaluated using Spearman's rank test. The relationships among more than three variables and numerical values were analyzed using Bonferroni-adjusted Mann-Whitney u-tests. Survival analysis was carried out using the Kaplan-Meier method and log-rank tests with JMP software (version 12, SAS Institute Inc., Cary, NC, USA). Other analyses were performed using Expert StatView software (version 5.0, SAS Institute Inc., Cary, NC, USA) for these analyses.

Conclusions
Both strands of pre-miR-149, miR-149-5p (guide strand) and miR-149-3p (passenger strand), acted as antitumor miRNAs in RCC cells. FOXM1 was regulated by these miRNAs, and an overexpression of FOXM1 was observed in RCC clinical specimens. The aberrant expression of FOXM1 enhanced cancer cell aggressiveness, and high expression of FOXM1 was significantly associated with a poor prognosis of this disease. The identification of antitumor-mediated oncogenic networks may lead to a better understanding of RCC pathogenesis.