E2F1-regulated long non-coding RNA RAD51-AS1 promotes cell cycle progression, inhibits apoptosis and predicts poor prognosis in epithelial ovarian cancer

Long non-coding RNA RAD51 antisense RNA 1 (RAD51-AS1, also known as TODRA) has been shown to be down-regulated by E2F1, a key cell cycle and apoptosis regulator, in breast cancer. Little is known regarding the role of RAD51-AS1 in disease. Here, we investigate the role of RAD51-AS1 in epithelial ovarian cancer (EOC). Using luciferase reporter and chromatin immunoprecipitation experiments, we verified RAD51-AS1 as a target of E2F1 under negative regulation in EOC. We then examined RAD51-AS1 expression in EOC samples using in situ hybridization (ISH). RAD51-AS1 was localized to the nucleus and found to be a critical marker for clinical features that significantly correlated with poor survival in EOC patients. RAD51-AS1 was also an independent prognostic factor for EOC. Overexpression of RAD51-AS1 promoted EOC cell proliferation, while silencing of RAD51-AS1 inhibited EOC cell proliferation, delayed cell cycle progression and promoted apoptosis in vitro and in vivo. RAD51-AS1 may participate in carcinogenesis via regulation of p53 and p53-related genes. Our study highlights the role of RAD51-AS1 as a prognostic marker of EOC. Based on its regulation of the tumor suppressor p53, RAD51-AS1-based therapy may represent a viable therapeutic option for EOC in the near future.


RAD51-AS1 is expressed in the nucleus and correlates with malignant clinicopathological features in EOC.
-AS1 expression in tissue microarrays of 163 patients (129 cases of EOC, 22 cases of borderline ovarian tumors and 12 cases of benign ovarian tumors) was semi-quantitatively examined by ISH ( Fig. 1A and Supplementary Fig. S2A). RAD51-AS1 was robustly expressed in the nucleus, with only very few punctate dots (one or fewer in a 40× magnification field) found outside the nucleus. Both EOC and borderline ovarian tumor tissues showed a significant increase in RAD51-AS1 expression compared with that observed in benign ovarian tissues (Fig. 1B). There were significant differences in RAD51-AS1 expression between each two grades (grade I, II, III and IV) of EOC, except between grade II and III (Fig. 1C).
Furthermore, the median value of RAD51-AS1 expression was used as a cut-off to divide the 129 EOC patients into a high-RAD51-AS1 expression group (n = 64) and a low-RAD51-AS1 expression group (n = 65). Clinicopathological features were compared between the two groups. High RAD51-AS1 expression correlated with advanced FIGO stage, histological subtype, high histological grade, number of tumors and distant metastasis ( Table 1). These findings suggest that RAD51-AS1 may function in the nucleus and has a strong relationship with a more malignant ovarian cancer phenotype.
In a previous study, there was opposite regulation of RAD51-AS1 and RAD51 in breast cancer 11 , so we detected the expression of RAD51 by IHC on the tissue microarrays ( Fig. 1E and Supplementary Fig. S2B). However, the result demonstrated that RAD51-AS1 expression had no correlation with RAD51 expression (P = 0.755, r = −0.026, Spearman correlation). It was found that borderline ovarian tumor had a higher expression of RAD51 than EOC, and both of them showed a significant increase in RAD51 expression compared with that observed in benign ovarian tissues (Fig. 1F). There was no significant difference in RAD51 expression among different FIGO stages of EOC ( Supplementary Fig. S2C). High level of RAD51 correlated with smaller tumor size in EOC (P = 0.023). However, no statistical significance was found between RAD51 expression and other clinicopathological features of EOC patients (Supplementary Table S1). Excluding 23 EOC patients who were lost to follow-up, complete outcome data were obtained in 106 EOC patients (follow-up rate, 82.2%). Survival analysis showed that EOC patients with high RAD51-AS1 expression levels had a significantly decreased overall survival (OS) than those with low RAD51-AS1 expression levels (P < 0.001, Fig. 1D). In addition, multivariate Cox regression analysis revealed that FIGO stage and RAD51-AS1 expression were independent predictors of OS in EOC (P < 0.05, Table 2).

RAD51-AS1 promotes EOC cell proliferation, regulates cell cycle progression and cell apoptosis.
We evaluated RAD51-AS1 expression in EOC cell lines ( Fig. 2A), and the results indicated general expression in these cells. The highest expression levels were detected in SKOV3.ip cells and the lowest in Hey cells. SKOV3, SKOV3.ip and HO8910 cells were then chosen for knockdown experiments, while Hey and OVCAR3 were for overexpression experiments. We designed three ASOs (ASO1, ASO2 and ASO3) targeting RAD51-AS1. In this study, we used an ASO-mix, a 'cocktail' combination of ASOs with equimolar ASO1, ASO2 and ASO3, to reduce the off-target effects of ASOs while silencing the expression of RAD51-AS1. Their silencing efficiencies were quantified by qRT-PCR analysis ( Fig. 2B-a) and proliferation assay ( Fig. 2B-b). As shown in Fig. 2B-c, the ASO mix exerted a satisfactory silencing effect on RAD51-AS1 expression in SKOV3, SKOV3.ip and HO8910 cells. Additionally, the efficiency of RAD51-AS1 overexpression was confirmed by qRT-PCR (Fig. 2C).
Thereafter, we conducted a series of functional experiments on cells in which RAD51-AS1 expression had been knocked down or overexpressed. Cell proliferation assays showed that RAD51-AS1 silencing significantly inhibited the proliferation of SKOV3, SKOV3.ip and HO8910 cells, and RAD51-AS1 overexpression increased the proliferation of Hey and OVCAR3 cells (Fig. 2D). Cell cycle assays showed that RAD51-AS1 knockdown increased the cells in G1/G0 phase in all 3 cell lines (Fig. 2E). Changes in S-phase were cell line-specific, with SKOV3 showing an increase, SKOV3.ip showing no change, and H08910 showing a decrease. Decrease change in G2/M was only shown in the SKOV3 cells, and not the others. In addition, knockdown of RAD51-AS1 remarkably increased the percentage of apoptotic cells ( Fig. 2F and Supplementary Fig. S3D). These results indicate that RAD51-AS1 promotes proliferation in vitro.
group were greatly lower than in the control group (Fig. 3A). The Matrigel invasion assay indicated that the suppression of RAD51-AS1 expression significantly inhibited the invasive ability of SKOV3, SKOV3.ip and HO8910 cells (Fig. 3B). What's more, overexpression of RAD51-AS1 increased cell invasion in Hey and OVCAR3 cells (Fig. 3C).
In order to normalize the potential effect of cell proliferation, Mitomycin C was added to the culture media to make the cells uniformly arrested. The results showed that all migration cell indexes reduced after adding Mitomycin C than those observed without Mitomycin C. Knockdown of RAD51-AS1 still inhibited cell migration ( Supplementary Fig. S3A) and invasion of SKOV3, SKOV3.ip and HO8910 cells ( Supplementary Fig. S3B), while overexpression of RAD51-AS1 only increased cell invasion capacity of the OVCAR3 cells, not in the Hey cells ( Supplementary Fig. S3C). Overall, these findings illustrated that RAD51-AS1 promotes EOC cell migration and invasion in vitro. p53 and p53-related genes are key downstream mediators of RAD51-AS1 regulation in EOC. To further identify the mechanisms and downstream regulators of RAD51-AS1, we performed a microarray analysis. Three SKOV3.ip RAD51-AS1-knockdown samples and three control samples were used. After silencing RAD51-AS1 expression, we identified 962 differentially expressed coding genes (Q value < 0.05, Fold change ≥1.5, 401 up-regulated and 561 down-regulated), which are represented in the heat map (Fig. 4A) and volcano plot (Fig. 4B).
Then, the significant GO analysis was performed (P = 0.05, FDR = 0.05), and the top ten GO terms are listed. In biological process analysis, mitotic cell cycle changed most significantly, followed by apoptotic process and transcription (with 46, 61 and 107 differentially expressed gene counts in GO, respectively) (Fig. 4D) Table 1. Association of RAD51-AS1 expression with clinicopathological features of EOC patients. NOTE: The median value of RAD51-AS1 expression was used as a cut-off for the low/high RAD51-AS1 expression groups. Data were analyzed using the χ 2 test. All tests were two-sided, and P < 0.05 was considered significant.
interesting to note that these findings are consistent with our original hypothesis. In the molecular function analysis, protein binding presented 351 differentially expressed gene counts in GO, far more than ATP binding and DNA binding (Fig. 4E). In the cellular component analysis, the nucleus was the most represented component with an apparent advantage over others (with 294 differentially expressed gene counts in GO), consistent with our ISH results on RAD51-AS1 expression in the nucleus (Fig. 4F). These results suggest that RAD51-AS1 mainly regulates cell cycle and apoptosis progression in EOC. Expressed in the nucleus, RAD51-AS1 may function through protein binding to exert downstream effects. Next, a pathway relation network was developed based on the KEGG database to help define the synergistic effects of the most important pathways (Fig. 4C). The result highlighted that the p53 pathway was closely involved in those changes. Accordingly, we focused on genes in the p53 pathway, particularly p53-related genes associated with cell cycle and apoptosis, in the downstream regulation of RAD51-AS1. We used qRT-PCR to confirm the interference efficiency of RAD51-AS1 in the microarray (Fig. 5A) and detected the transcription levels of p53 and p53-related genes (Fig. 5B). We determined that the mRNA of p53 increased 4.56-fold after silencing RAD51-AS1 (1.56-fold change in the microarray). The expression of p53-related genes involved in cell cycle or apoptosis, such  Table 2. Univariate and multivariate analysis of overall survival in 106 EOC patients. All tests were two-sided, and P < 0.05 was considered significant.
as CDKN1A, TNFRSF10B, FADD, CASP3, CASP8 and CASP9 was higher, while CCNE2 levels were lower in the RAD51-AS1-knockdown group compared with the control group. Their expression confirmed our microarray results, thus ensuring the reliability of the HTA2.0 array. In addition, Western blot analysis was used to examine the protein levels of these genes. Changes of P53, CCNE2, CDKN1A, CASP8 and CASP9 were found statistically significant (Fig. 5C). Finally, we evaluated p53 expression in patient samples (Fig. 5D). The correlation analysis demonstrated that p53 expression negatively correlated with RAD51-AS1 expression (P < 0.001, r = −0.269, Fig. 5E). These results reinforce the possibility that p53 and p53-related genes are key downstream mediators of RAD51-AS1.

RAD51-AS1 knockdown impairs tumor growth in vivo.
To evaluate the effect of RAD51-AS1 in vivo, xenograft tumors were established in nude mice. As shown in Fig. 6A, tumor growth in the RAD51-AS1 knockdown group was substantially slower than that in the control group, and the gap between the two groups increased with time. After the mice were sacrificed, we performed qRT-PCR to confirm the silencing of RAD51-AS1 in vivo (Fig. 6C). By visual comparison, it was obvious that tumors in the RAD51-AS1-knockdown group were substantially smaller than those in the control group (Fig. 6B). In addition, the tumor weight in the RAD51-AS1-knockdown group decreased by almost 50% compared with the control group (Fig. 6D). These results indicate that silencing of RAD51-AS1 impairs tumor growth in vivo.
Based on the findings in vitro, we then detected certain key regulatory factors by IHC to further clarify the function of RAD51-AS1 in vivo (Fig. 6E-F). As expected, p53 expression was increased following RAD51-AS1 silencing. In addition, the expression of the proliferation marker Ki67 was significantly lower in RAD51-AS1-knockdown tumors. Similarly, PH3 (an M-phase marker of the cell cycle) and CCNE2 (a core component of cell cycle machinery, particularly the G1/S phase transition) presented with lower expression levels, consistent with the cell cycle assay. In contrast, the expression levels of apoptotic factors (CASP3 and CASP9) were significantly higher in RAD51-AS1-knockdown tumors than in the control ones. Taken together, these results support our findings in vitro and suggest that RAD51-AS1 promotes cell proliferation and cell cycle progression and inhibits cell apoptosis in vivo.

Discussion
Numerous studies have revealed that lncRNAs have distinct biological functions and essential roles in tumorigenesis 22 . LncRNAs, such as HOTAIR, ANRIL and MALAT1, have been suggested as potential prognostic biomarkers and therapeutic targets in cancers [5][6][7][8] . Therefore, a better understanding of lncRNAs may help in the identification of new biomarkers and effective therapeutic strategies that can improve patient outcomes. RAD51-AS1 is a novel lncRNA that is transcribed 69 bp upstream of RAD51, in the opposite direction. Herein, we determined that the transcription factor E2F1, a key regulator of cell cycle and apoptosis [13][14][15][16][17][18][19]23 , negatively regulated RAD51-AS1 expression in EOC cell lines by binding its promoter. These findings are consistent with a previous study in breast cancer 11 .
In breast cancer, there was opposite regulation of RAD51-AS1 and RAD51 expression, and higher RAD51-AS1/lower RAD51 expression was associated with a less aggressive histological tumor phenotype 11 . However, in our study, the correlation analysis demonstrated that RAD51-AS1 expression has no correlation with RAD51 expression in EOC. Moreover, we found that RAD51-AS1 expression correlated with a more malignant phenotype and a poorer prognosis of EOC. RAD51-AS1 was an independent prognostic indicator for EOC patients. These clinical findings enhance the potential value of RAD51-AS1 as a prognostic marker for EOC.
In support of the hypothesis posed at the beginning of this study, in vitro and in vivo experiments provide evidence that RAD51-AS1 is involved in the regulation of cell cycle or apoptosis and plays a role in promoting cell proliferation in EOC. The microarray results reinforce our findings in cytobiology experiments. Additionally, RAD51-AS1 also regulates cell migration and invasion in EOC cells. We found there was enrichment for genes expressed in the nucleus through GO analysis. Using ISH, RAD51-AS1 was found to be strongly expressed in the nucleus, where it most likely functions through binding to proteins. Mechanistically, the most well-known lncRNAs regulate transcription through interactions with protein, DNA, or other cellular macromolecules 24 . In addition, recent studies have shown that lncRNAs expressed in the nucleus mostly regulate cell processes by facilitating the epigenetic repression of downstream genes 4,25 . For instance, ANRIL, HOTAIR, H19 and XIST all play a repressive function by coupling with histone modifying or chromatin remodeling protein complexes [26][27][28][29][30] . Thus, we speculated that RAD51-AS1 may function through binding to proteins, such as transcription factors, to achieve downstream effects; some key genes might be repressed by RAD51-AS1.
Genome browser UCSC hg19 (http://genome.ucsc.edu/) was used to get DNA sequences. Target genes under cis-regulatory control were defined as genes whose transcription was regulated by lncRNAs in nearby genomic locations (≤10 kbp upstream or downstream) 31 . Based on this, we identified two predicted target genes of RAD51-AS1: Tyro3 and IVD. However, neither the mRNA nor protein levels of these molecules changed after silencing RAD51-AS1 expression. Then, the p53 pathway highlighted by KEGG pathway analysis stimulated our interest. p53 activation can cause cell cycle arrest and apoptosis 32,33 , which is the exact phenotype observed upon RAD51-AS1 silencing. We found that both mRNA and protein levels of p53 were elevated by RAD51-AS1 silencing. Furthermore, the expression of RAD51-AS1 and p53 showed reverse correlation in patient tissues, raising the possibility that p53 is a key downstream gene repressed by RAD51-AS1. In general, elevated levels of p53 protein will in turn induce CDKN1A transcription and lead to cell cycle arrest at the G1 phase 34,35 . As expected, we detected elevated CDKN1A after silencing RAD51-AS1. In addition, RAD51-AS1 silencing activates apoptotic regulators associated with p53 up-regulation, which may explain the pro-apoptotic effect of RAD51-AS1 silencing. These findings not only further illustrate that RAD51-AS1 regulates cell cycle and apoptosis but also support that p53 and p53-related genes are key downstream mediators of RAD51-AS1. Their dysreg8ulation may partially explain the involvement of RAD51-AS1 in EOC development.
As the number of well-described lncRNAs grows and along with the development of RNAi-based therapeutics 36 , the value of lncRNAs in cancer therapy has been attracting increasing attention 9, 10, 37 . Based on the reverse regulation of the tumor suppressor p53, one might anticipate that the inhibition of RAD51-AS1 might have a therapeutic effect by restoring the expression of p53 and p53-related genes. However, it is beyond the scope of this study to examine the direct target genes of RAD51-AS1. Therefore, future studies of the detailed mechanisms of RAD51-AS1 are needed. In addition, studies with larger samples are required to enhance the feasibility and reliability of RAD51-AS1 as a prognostic biomarker for EOC.

Materials and Methods
Cell lines and cell culture. The human ovarian cancer cell lines (SKOV3, SKOV3.ip, HO8910 and HO8910-PM) were provided by the University of Texas MD Anderson Cancer Center (Houston, TX, USA) and were authenticated by Short Tandem Repeat (STR) profiling. The human ovarian cancer cell lines Hey and OVCAR3 were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). All cells were cultured in RPMI 1640 (Gibco BRL, Gaithersburg, MD, USA) containing 10% fetal bovine serum (FBS; Gibco) with 100 units/ml penicillin and 100 mg/ml streptomycin in a 5% CO 2 humidified incubator at 37 °C. Mitomycin C was purchased from MedChem Express (China, HY-13316). In the migration assay and transwell invasion assay, the cells were mitotically inactivated by treating with mitomycin C (10 µg/mL, 2 h).
Tissue microarray and patient data. This study was approved by the ethics committee of the Obstetrics and Gynecology Hospital of Fudan University (No. [2015]33) and complies with the REMARK guidelines for biomarker studies 38 . All paraffin-embedded tissue samples were obtained from the Tissue Bank of the Obstetrics and Gynecology Hospital of Fudan University. This study included 163 patients, who underwent surgery in the Obstetrics and Gynecology Hospital of Fudan University between January 2009 and December 2011. Patients with two or more different malignancies and those who had received chemotherapy, preoperative radiotherapy, or hormonal therapy were excluded from the study. Tissue microarrays were manufactured. For each tissue sample, we collected tissues from two different areas to reduce the influence of tumor heterogeneity and to be more representative of the tumor. Clinical and histopathologic features were retrieved by reviewing medical charts and pathology reports. The outcomes of each patient were retrieved from the Follow-up Database in our hospital. Informed consent was obtained from all of the patients, according to the principles of the Declaration of Helsinki.
In situ hybridization (ISH). Paraffin sections of the tissue microarrays were analyzed according to a protocol developed by Advanced Cell Diagnostics 39 . ISH was performed using the RNAscope ® 2.0 Assay (ACD, Inc. Catalog No. 320497) and HybEZ ™ Hybridization System (ACD, Inc. 110 VAC, Cat. no. 310010). RAD51-AS1 target probes and positive and negative control probes were designed and purchased from ACD. Technical and slide quality control was certified using a positive control probe targeting the common housekeeping gene PPIB and a negative control probe targeting the bacterial gene DapB (PPIB > 2 and DapB < 1). Finally, slides were imaged using an Olympus Dual-CCD microscope digital camera, and semi-quantitative scores were obtained by estimating the number of punctate dots. See Supplementary Materials and Methods for details. RNA interference and RAF51-AS1 overexpression. We used 6 siRNAs targeting RAD51-AS1, but none knocked down RAD51-AS1 expression by more than 50%. Therefore, three antisense oligonucleotides (ASO) targeting RAD51-AS1 and a negative control (NC) were designed and synthesized by Ribobio (Guangzhou, China). Cells were transfected with 50 nM RAD51-AS1-targeting ASO or 50 nM NC using Lipofectamine 3000 (Invitrogen). Silencing efficiency was evaluated by qRT-PCR. Each ASO had a satisfactory interference effect on RAD51-AS1 expression. In this study, we used an ASO-mix (a 'cocktail' combination of ASOs with equimolar concentrations of ASO1, ASO2 and ASO3) to reduce off-target effects. The target sequences of RAD51-AS1-targeting ASOs are represented in the following sequences: ASO1-TCCGCGAGTTCTCACCATCG, ASO2-TCCTAGCTACTCGAAAGGCT, and ASO3-GCATGGAAACGAACTACATG. Additionally, the lentiviruses expressing the RAD51-AS1 sequence (RAD51-AS1-overexpression, OE) and the negative control lentivirus (NC) were constructed by Hongli Co. (Shanghai, China).
Proliferation and migration assays. Twenty-four hours after transfection, cells were seeded at 5 × 10 3 cells/well in cell culture E-Plate 16 (ACEA Biosciences Inc., USA) for proliferation assays and at 2 × 10 4 cells/well in cell culture CIM-Plate 16 (ACEA Biosciences Inc., USA) for migration assays according to the manufacturer's instructions. The plates were incubated at 37 °C with 5% CO 2 . The cell growth and migration curves were automatically recorded on the xCELLigence RTCA System (Roche, USA) in real-time. The cell index of the proliferation and migration assays was followed for 4-5 days.
Flow cytometric analysis of apoptosis and cell cycle. Cells were harvested 48 hours after transfection.
For apoptosis analysis, the transfected cells were stained using the Dead Cell Apoptosis Kit with Annexin V Alexa Fluor 488 and propidium iodide (PI) (Invitrogen, catalog number V13241) according to the manufacturer's recommendations. For cell cycle analysis, after fixing and permeabilizing the transfected cells, 0.5 ml of PI/RNase Staining Buffer (BD Biosciences, USA, catalog number 550825) was used per sample (1 × 10 6 cells) and incubated for 30 minutes at room temperature. Thereafter, the cell apoptosis ratio and the percentage of cells in each phase of cell cycle were measured using a FACStation (FV500, Beckman Coulter, Brea, CA, USA). The results were analyzed using FlowJo 7.6.2 software.
Transwell invasion assay. Cells were harvested 24 hours after transfection. Cell invasion was evaluated using the Transwell invasion assay with inserts of 8-µm pore size (Coring Costar), as previously described 7 .
Microarray assay. Twenty-four hours after transfection of 50 nM RAD51-AS1-ASOs or NC, total RNA from SKOV3.ip cells was extracted using TRIzol reagent (Life Technologies). Then, the Affymetrix Human Transcriptome Array 2.0 (HTA 2.0) was used to compare the transcriptome profiling of RAD51-AS1 knockdown (KD) and control samples. Differentially expressed genes, Gene Ontology (GO) and KEGG pathway analyses were used to investigate the functional significance of the molecular changes. The Gene Expression Omnibus accession numbers are (GSE89374).

Western blot analysis.
Western blot analysis was performed as previously described 7 . The antibodies used are described in the Supplementary Materials and Methods section. The quantitative protein analysis was performed using ImageJ software.