Long non‐coding RNA EPB41L4A‐AS2 suppresses progression of ovarian cancer by sequestering microRNA‐103a to upregulate transcription factor RUNX1T1

What is the central question of this study? What is the specific mechanism by which EPB41L4A‐AS2 exerts a regulatory role in ovarian cancer? What is the main finding and its importance? Overexpressed EPB41L4A‐AS2 inhibited cell proliferation, migration and invasion in ovarian cancer by upregulating RUNX1T1 through downregulation of miR‐103a. This study provides new insight into the role of EPB41L4A‐AS2 in ovarian cancer.

co-functioning with protein molecules, DNA, RNA and/or their combinations (Yang, Lu, & Yuan, 2014a,b). It was reported that high expression of EPB41L4A-AS2, a novel lncRNA, contributes to favourable disease outcomes in breast cancers with a suppressive role in the proliferation of tumour cells (Xu et al., 2016). Using online analysis software in the present study, binding sites were found between the EPB41L4A-AS2 gene sequence and the sequence of microRNA-103a (miR-103a). Defined as a type of small non-coding RNA molecule, miRNAs can inhibit the expression of their target genes in a sequence-dependent manner, thereby functioning in cellular processes of cancers (Wozniak, Sztiller-Sikorska, & Czyz, 2015;Zhou & Rigoutsos, 2014). One study found that when miR-103a-3p or miR-107 was sequestered, their 13 target genes were upregulated, thereby changing the proliferation, migration and invasion of bladder cancer cells (Zhong, Lv, & Chen, 2016). Also, miR-103a-3p was demonstrated to be involved in the proliferation, migration and invasion process of OC cells (Bignotti et al., 2016). The gene Runt-related transcription factor 1 (RUNX1T1) is a putative target of miR-103a in the present study as demonstrated by the RNA22 website. Low RUNX1T1 expression was observed in primary pancreatic endocrine neoplasms (Nasir et al., 2011). Moreover, it was reported in a previous study that RUNX1T1 played a tumour suppressor role during OC (Yeh et al., 2011). Based on the aforementioned information, we hypothesized that there is an interaction between EPB41L4A-AS2, miR-103a and RUNX1T1 and that this has a function in the occurrence and development of OC. Therefore, our purpose in the present study was to investigate how EPB41L4A-AS2, miR-103a and RUNX1T1 interact with each other, and the role their interaction plays in the progression of OC.

Ethics approval
The study protocol was approved by the Ethics Committee and Experimental Animal Ethics of Jining No. 1 People's Hospital (approval reference number: 201201005 (human) and 201806004 (animal)).
Informed written consent was obtained from each patient prior to the study. All experiments involving human specimens in the present study were conducted in strict accordance with the Declaration of Helsinki  Grundy (2015). Every effort was made to minimize the pain, suffering and discomfort of the experimental animals.

Microarray-based gene expression profiling
With OC as a key word, the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) was used to search the OC-related miRNA microarray data GSE83693 (Nam et al., 2016) New Findings

• What is the central question of this study?
What is the specific mechanism by which EPB41L4A-AS2 exerts a regulatory role in ovarian cancer?
• What is the main finding and its importance?
Overexpressed EPB41L4A-AS2 inhibited cell proliferation, migration and invasion in ovarian cancer by upregulating RUNX1T1 through downregulation of miR-103a. This study provides new insight into the role of EPB41L4A-AS2 in ovarian cancer. and gene microarray data GSE18520 (Mok et al., 2009), both of which consisted of normal samples and OC samples. A differential analysis was performed using the R language. Standardization and difference analysis of microarray data were conducted using the limma package (http://master.bioconductor.org/packages/release/bioc/html/ limma.html; Ritchie et al., 2015) with the conditions of an absolute fold change (FC) value of greater than 1.5 and adjusted P-value < 0.05 (the P-value was corrected by the false discovery rate  -AS2  GTCGCAGTTAGGGGAGACAC  TGGCTACCCAGCTAACAAGC-3   RUNX1T1  GAAAGCCCACGACATGATCAC  CAGCCACTGCAGGTTTCACTC   miR-103a  AGCAGCATTGTACAGGGCTATGA  GCAGGGTCCGAGGTATTC   U6  TGCTCGCTTCGGCAGC  AAAAATATGGAACGC-TTCACG   -Actin  GTGGATCAGCAAGCAGGAGT  ATCCTGAGTCAAGCGCCAAA miR-103a, microRNA-103a; RT-qPCR, reverse transcription quantitative polymerase chain reaction; RUNX1T1, runt-related transcription factor 1.
collected fresh tissues were temporarily stored in at −80 • C in an ultra-low-temperature freezer, and long-term preserved in liquid nitrogen.

Follow-up study
All patients with OC were followed up by means of telephone or subsequent visit before February 2018. The overall survival (OS) was recorded, which was defined as starting from randomization until the patients died for any reason. Up to the deadline, the follow-up period was 36 months, and 14 out of 126 patients were lost in the follow-up with a follow-up rate of 88.89%.

Reverse transcription quantitative polymerase chain reaction (RT-qPCR)
The total RNA from OC tissues was extracted by Trizol (cat. no. 16096020, Thermo Fisher Scientific Inc., Waltham, MA, USA), and 5 µg RNA was reverse transcribed into cDNA according to the instructions of the cDNA kit (K1622; Fermentas Inc., Burlington, Ontario, Canada).
With cDNA as a template, RT-qPCR was conducted according to the instructions provided by TaqMan Gene Expression Assays protocol (Applied Biosystems/Thermo Fisher Scientific). U6 served as an internal control for miR-103a, and -actin for other genes. Three duplicate wells were set for each RT-qPCR. Primer sequences are shown in Table 1. Quantification of relative gene expression was calculated by the 2 −ΔΔC T method (Livak & Schmittgen, 2001).

Western blot analysis
The total protein of tissues or cells was lysed with lysis buffer containing phenylmethanesulfonyl fluoride on ice for 30 min and centrifuged at 11,180 g at 4 • C for 15 min. The supernatant was transferred into a new Ependorf tube, and the protein concentration was estimated using a bicinchoninic acid kit (Thermo Fisher Scientific).

Cell treatment
Finally, the original transfection solution was renewed with normal cell culture medium, and the cells were further cultured for subsequent experiments.

Cell counting kit-8 (CCK-8) assay
The transfected cells were detached and resuspended with the concentration adjusted to 1 × 10 5 cells ml −1 . Subsequently, the cells were seeded into a 96-well plate at a density of 100 µl per well and cultured overnight. Then, the cell viability was determined in accordance with the instructions of a CCK-8 kit (Beyotime Biotechnology Co., Ltd, Shanghai, China) at the 24th, 48th, 72nd, and 96th hour post-seeding, during which 10 µl CCK-8 solution was added for a 4 h incubation. After that, the optical density (OD) value at 450 nm was measured with a microplate reader, and the growth curve was plotted.

Colony formation assay
The cells in the logarithmic growth phase were dissociated to create a single cell suspension and counted. A total of 1000 cells were seeded into 60 mm culture dishes and then cultured with 5% CO 2 , during which the culture medium was renewed every 3 d. After 14 d, the medium was discarded, and the cells were washed 3 times with PBS, fixed with methanol for 15 min, and stained with crystal violet for 15 min. Finally, the cell colonies formed containing over 50 cells was counted under a microscope (Wang et al., 2019a).

Transwell assay
Cell migration experiments were firstly conducted. In brief, the OC cells in logarithmic growth phase were starved for 24 h and then subjected to detachment, centrifugation and resuspension at a final concentration of 2 × 10 5 cells ml −1 . Then, 0.2 ml suspension was added to the apical chamber, while 700 µl pre-cooled DMEM containing 10% FBS was added to the basolateral chamber. Subsequently, the cells were cultured at 37 • C with 5% CO 2 . After 24 h, the cells on the apical chamber and the basement membrane were wiped off using a wet cotton swab, fixed for 30 min using methanol, and stained for 20 min with 0.1% crystal violet. Finally, the cells were observed and photographed under an inverted microscope with five fields of view (×200) randomly selected to count the number of transmembrane cells.
Cell invasion experiments were then conducted. In brief, extracellular matrix (ECM) Matrigel was placed at 4 • C overnight at 4 • C, and the next day, the Matrigel was diluted at a ratio of 1: 9 using serum-free medium to a final concentration of 1 mg ml −1 (all the pipettes and chambers were pre-cooled on ice for 30 min in advance). A total of 40 µl ECM Matrigel was added to the polycarbonate membrane in the apical chamber of a 24-well Transwell chamber, incubated for 5 h in an incubator at 37 • C with 5% CO 2 to polymerize the ECM Matrigel. Then, 70 µl pure DMEM was added to each chamber for incubation at 37 • C for 30 min to rehydrate the Matrigel. After starvation for 24 h, the cells were subjected to detachment, centrifugation and resuspension in FBS-free DMEM at a final concentration of 2.5 × 10 5 cells ml −1 . In the next step, 0.2 ml suspension was added to the apical chamber in which the basement membrane had been hydrated, and to the basolateral chamber, 700 µl pre-cooled DMEM with 10% FBS was added. Thereafter, the cells were cultured at 37 • C with saturated humidity and 5% CO 2 for 24 h.
Subsequently, wet cotton swabs were used to wipe off the cells on the chamber and basement membrane. The chamber was fixed with methanol for 30 min, stained for 20 min using 0.1% crystal violet and dried in an inverted position. Finally, the invasive cells were counted in five randomly selected visual fields with ×200 magnification under an inverted microscope.

Fluorescence in situ hybridization
The subcellular localization of EPB41L4A-AS2 in OC lines was identified by fluorescence in situ hybridization (FISH). According to the instructions of Ribo TM lncRNA FISH probe Mix (Red) (Guangzhou RiboBio Co., Ltd, Guangzhou, China), OC cells were seeded onto the coverslips, which were placed into a six-well culture plate in advance, followed by 1 day of culture to ensure 80% confluence. Then, the slides were fixed at room temperature using 1 ml 4% paraformaldehyde and treated with 2 µg ml −1 proteinase K, glycine and acetamidine reagent. Subsequently, the cells were incubated with 250 µl prehybridization solution for 1 h at 42 • C and hybridized with 250 µl hybridization solution containing probes (300 ng ml −1 ) overnight at 42 • C. Subsequently, 4 ′ -6-diamidino-2-phenylindole (DAPI; 1: 800) diluted by PBST was added to stain the nucleus for 5 min. Then the cells were sealed with anti-fluorescent quencher after being washed 3 times with PBST (3 min each time). Finally, five different visual fields were selected to photograph cells under a ×400 fluorescence microscope (Olympus Optical Co., Ltd, Tokyo, Japan).

RNA immunoprecipitation assay
The binding of miR-103a, EPB41L4A-AS2 and RUNX1T1 was detected by an RNA immunoprecipitation (RIP) kit (Millipore, Billerica, MA, USA). The OC cells were washed with pre-cooled PBS, centrifuged and the supernatant discarded. Next, an equal volume of lysis buffer (P0013B, Beyotime Biotechnology Co.) was used to lyse the cells in an ice bath for 5 min, and the supernatant was obtained after centrifugation at 22,000 g for 10 min at 4 • C. Part of the cell extracts was collected as input and the rest was incubated with the antibody for co-precipitation. In specific terms, 50 µl magnetic beads in each coprecipitation reaction system were resuspended in 100 µl RIP wash buffer. Then 5 µg antibody was added for the incubation depending on the grouping to form a magnetic bead-antibody complex. Subsequently, the magnetic bead-antibody complex was washed and then resuspended in 900 µl RIP wash buffer, followed by overnight incubation with 100 µl cell extract at 4 • C. Next, the magnetic beadprotein complex was collected from the sample on a magnetic stand.
The antibodies used in the experiment included anti-Argonaute2 (Ago2; ab32381, 1: 50, Abcam) for 30 min at room temperature, and IgG (1: 100, ab109489, Abcam), which was taken as a negative control (NC). Finally, RNA was extracted from the sample and input by proteinase K for subsequent RT-qPCR detection.

Tumour xenograft in nude mice
A total of 20 female BALB/c nude mice (aged 3-4 weeks, weighing 14 ± 2 g) were purchased from Hui Ao Biotechnology Co., Ltd (Beijing, China). All mice were housed in an environment with constant temperature of (25-27 • C) and humidity (45%-50%) with free access to water and food. These mice were then respectively injected with OC cells expressing oe-NC and oe-EPB41L4A-AS2 (10 mice for each injection). Subsequently, stably transfected cell lines were constructed, and the cells were adjusted to a concentration of 1 × 10 7 cells ml −1 .
Then 20 µl cell suspension was subcutaneously inoculated into nude mice, after which the growth of tumours was observed and recorded every day after photographing. The volume of tumours was also recorded at an interval of 6 d, followed by plotting a growth curve based on a formula of (a × b 2 )/2 (where a represents the longest diameter of tumours and b represents the shortest diameter of tumours) (Zhu, Ma, & Zhang, 2017). After 30 days, the nude mice were killed by excessive inhalation of CO 2 , and the xenograft tumours were extracted and weighed for subsequent RT-qPCR and western blot analysis.

EPB41L4A-AS2 is poorly expressed in OC tissues and cells
Initially, the R language was used to perform differential analysis of OC microarray data. From microarray data GSE18520, 1260 DEGs were screened as they met the conditions of an absolute FC value of greater than 1.5 and adjusted P-value < 0.05. A heatmap for the top 25 DEGs was plotted (Figure 1a), in which we noted that EPB41L4A-AS2 was expressed at a low level in OC. Besides, EPB41L4A-AS2 was identified as suppressing the progression of breast cancer (Pang et al., 2019), hepatocellular carcinoma (Wang, Wang, Shi, & Zhai, 2019b) and non-small cell lung cancer (Shu, Li, Chen, Zhu, & Yu, 2018). RT-qPCR was conducted to detect the expression of EPB41L4A-AS2 in 126 OC tissues and paracancerous tissues, the results of which documented that EPB41L4A-AS2 was poorly expressed in OC tissues compared with in paracancerous tissues (P < 0.05; Figure 1b). These specimens were divided into the EPB41L4A-AS2-high group (63 cases) and the EPB41L4A-AS2-low group (63 cases) by taking the median expression of EPB41L4A-AS2 in 126 OC specimens of 2.169 as the cut-off point.
As shown in Table 2

Overexpressed EPB41L4A-AS2 inhibits the proliferation, migration and invasion of OC cells
The aforementioned results showed that EPB41L4A These results collectively documented that upregulated EPB41L4A-AS2 could inhibit the proliferation, migration and invasion of OC cells.

EPB41L4A-AS2 upregulates RUNX1T1 by binding to miR-103a
In order to further explore the mechanism of EPB41L4A-AS2 in OC, the subcellular localization data of EPB41L4A-AS2 was obtained The cell experiments were repeated 3 times. Measurement data were expressed as mean ± standard deviation. A paired t test was used for comparison between cancer and paracancerous tissues, and ANOVA was used for comparisons among multiple groups, followed by Tukey's post hoc test with LncATLAS; the results of this showed that EPB41L4A-AS2 was mainly expressed in the cytoplasm (Figure 3a). In addition, the results obtained from FISH were consistent with those from LncATLAS ( Figure 3b). Subsequently, 1196 miRNAs were predicted to have binding sites with EPB41L4A-AS2 in RNA22, while 25 differentially expressed miRNAs with high expression in OC were screened through analysis of miRNA microarray data GSE83693. Then, differentially overexpressed miRNAs related to OC were compared with miRNAs that bound to EPB41L4A-AS2, and five intersected miRNAs were observed ( Figure 3c). As shown in Table 3, the P-value of hsa-miR-103a-3p was the lowest, and the expression of miR-103a-3p in OC tissues was significantly higher than that in paracancerous tissues. The effect of miR-103a on OC was studied further. The potential target genes of miR-103a were predicted in mirDIP, TargetScan, miRDB, miRWalk, RNA22 and DIANA. There were 60 predicted target genes of miR-103a according to the criteria of integrated score >0.9 in mirDIP, 208 genes based on Total context ++ score <−0.4 in TargetScan, 220 target genes based on target score >70 in miRDB, 7678 target genes based on energy <−22 in miRWalk, 296 target genes based on miTG score >0.9 in DIANA and 11,656 target genes of miR-103a predicted in RNA22. These target genes were subsequently compared, and a total of 12 genes (SNRK, ATP13A3, RNF38, RUNX1T1, ARMC1, CPEB3, AGO1, N4BP1, DICER1, NDEL1, ZHX1 and ANO3) were observed to be intersected ( Figure 3d). Subsequently, the 1260 DEGs in OC microarray data were compared with the predicted target genes of miR-103a, in which only the gene RUNX1T1 overlapped (Figure 3e). Moreover, RUNX1T1 was poorly expressed in OC through OC microarray data GSE18520 (Figure 3f). Collectively, this shows that RUNX1T1 may be differentially expressed in OC under the regulation of miR-103a.
The binding sites between EPB41L4A-AS2 and miR-103a were predicted by the RNA22 website (Figure 3g). The results of a dual luciferase reporter gene assay showed that miR-103a mimic could significantly decrease the luciferase activity of EPB41L4A-AS2-WT (P < 0.05), but had no effect on the luciferase activity  bind to miR-103a. Subsequently, mimic NC and miR-103a mimic were respectively co-transfected with RUNX1T1-WT and RUNX1T1-MUT, followed by the performance of a dual luciferase reporter gene assay to detect luciferase activity. The results documented that miR-103a mimic can significantly reduce the luciferase activity of RUNX1T1-WT (P < 0.05; Figure 3k), but had no effects on RUNX1T1-MUT (P > 0.05), which showed that miR-103a can target RUNX1T1 and downregulate its expression. Afterwards, we conducted RT-qPCR to detect the expression of miR-103a and RUNX1T1 after oe-EPB41L4A-AS2 transfection, which demonstrated that the expression of miR-103a was decreased while that of RUNX1T1 was increased upon oe-EPB41L4A-AS2 transfection (P < 0.05; Figure 3l). Besides, we further conducted RT-qPCR to determine the expression of miR-103a and RUNX1T1 in response to miR-103a mimic transfection. As depicted in Figure 3m, the expression of miR-103a was increased while that of RUNX1T1 was decreased upon miR-103a mimic transfection (P < 0.05). Overexpressed EPB41L4A-AS2 led to higher expression of RUNX1T1, but this promoting effect was reversed in response to both miR-103a mimic and overexpressed EPB41L4A-AS2 (P < 0.05; Figure 4a). Taken together, EPB41L4A-AS2 may upregulate RUNX1T1 expression by binding to miR-103a.

Overexpressed EPB41L4A-AS2 prevents the progression of OC by activating RUNX1T1 via miR-103a
The Expression of miR-103a and RUNX1T1 upon miR-103a mimic transfection measured by RT-qPCR. *P < 0.05. The cell experiments were repeated 3 times. Measurement data were expressed as means ± standard deviation. Unpaired t test was used for comparisons between two groups

Overexpressed EPB41L4A-AS2 inhibits tumour growth of OC cells in vivo
The tumour growth of nude mice was observed, with the volume and

DISCUSSION
In terms of morbidity and mortality, OC ranks in the top eight leading cancers around the world (Coburn et al., 2017). At present, several approaches to treat OC have had a significant favourable initial response, but still many OC patients with advanced disease will show recurrence within 18 months (Jayson, Kohn, Kitchener, & Ledermann, 2014;Oza et al., 2015), thus emphasizing the urgent need for treatments contributing to the improvement of clinical outcome.
In the present study, we performed different kinds of experiments to study the effects EPB41L4A-AS2 on OC and reveal the related mechanism. The results obtained from our experiments documented that EPB41L4A-AS2 expression was poor in OC tissues and cells, and EPB41L4A-AS2 overexpression impeded the proliferation, migration and invasion of OC cells by promoting the expression of RUNX1T1 via miR-103a.
Initial findings from our study revealed poor EPB41L4A-AS2 expression in OC tissues and cells. It has been demonstrated that overexpressed lncRNAs such as HAND2 antisense RNA 1 and nicotinamide nucleotide transhydrogenase-antisense RNA1 play a suppressive role in the proliferation and metastasis of OC cells (Huang et al., 2018a(Huang et al., , 2018b(Huang et al., , 2018cYang et al., 2018aYang et al., , 2018b. EPB41L4A-AS2 was proven to be poorly expressed in non-small cell lung cancer (NSCLC) tissues and cells, and the apoptosis of NSCLC cells was promoted by highly expressed EPB41L4A-AS2 (Shu et al., 2018). It has been demonstrated that tumours of a higher degree and more malignant characteristics possess lower expression of EPB41L4A-AS2, and overexpressed EPB41L4A-AS2 leads to better OS of several malignant tumours including breast cancer, renal cancer and lung cancer (Xu et al., 2016). To further verify our results, we performed in vitro experiments that demonstrated that the proliferation, invasion and migration of OC can be prevented when the expression of EPB41L4A-AS2 is upregulated. These data were supported by a prior study suggesting that ectopic expression of EPB41L4A-AS2 repressed cell proliferation in lung, breast and renal cancer (Xu et al., 2016), which indicated the suppressive effects of EPB41L4A-AS2 on tumour formation.
EPB41L4A-AS2 was also reported to be poorly expressed in head and neck squamous cell carcinoma (HNSCC), and played a suppressive role in the invasion and metastasis of HNSCC by suppressing the expression of transforming growth factor receptor 1 (Huang et al., 2018a(Huang et al., , 2018b(Huang et al., , 2018c. This was in accordance with the results from Huang et al. that demonstrated that EPB41L4A-AS2 was poorly expressed in resistant breast cancer cells and might be regarded as a possible biomarker for docetaxel sensitivity of breast cancer (Huang et al., 2018a(Huang et al., , 2018b(Huang et al., , 2018c. All of these studies have verified that EPB41L4A-AS2 acts as a suppressor in the progression of OC.
Another important finding was that overexpressed EPB41L4A-AS2 inhibited the development of OC through increasing RUNX1T1 by combining with miR-103a. In a previous study, RUNX1T1 protein   . Measurement data were expressed as means ± SD. Unpaired t-test was used for comparisons between two groups. Repeated measures one-way ANOVA was utilized for comparisons of data at different time points, followed by a Bonferroni's post hoc test was verified as a promising biomarker for the diagnosis of liver metastases since it was poorly expressed in cells of well-differentiated metastatic primary pancreatic endocrine tumours and associated with non-metastatic emerges and primaries (Nasir et al., 2011). Besides, it has been verified in a previous study that the expression of RUNX1T1 is downregulated in OC cell lines, and the restoration of RUNX1T1 could inhibit the growth of OC cells (Yeh et al., 2011). Multiple investigations have reported the suppressive function of the interaction between lncRNAs, miRNAs and mRNAs in human cancers. For example, overexpressed lncRNA LINC00312 can inhibit the proliferation, migration and invasion of thyroid cancer cells by inhibiting miR-197-3p expression (Liu, Huang, Yan, Luo, & Min, 2017). The migration of glioma cells can be suppressed after overexpressed ADAM metallopeptidase with thrombospondin type 1 motif, 9 antisense RNA 2 treatment, which can be reversed by silencing the expression of DNMT1 (Yao et al., 2014). Besides, it has been documented that by targeting RUNX2, upregulated miR-103a contributes to inhibition of bone formation (Zuo et al., 2015). Another study demonstrated that a serous epithelial OC cohort was correlated with high expression of miR-103 (Kan et al., 2012). Furthermore, microarray data showed that miR-103a was overexpressed, but RUNX1T1 was diminished in OC. The highly expressed miR-103 is also presented in OC cells and acts a promoter in OC cell metastasis by targeting Dicer1 (Yang et al., 2014a(Yang et al., , 2014b. Moreover, another study also identified that miR-103a-3p exhibited high expression in OC cells and mediated proliferation, migration and invasion of OC cells (Bignotti et al., 2016). Collectively, the interaction of EPB41L4A-AS2, miR-103a and RUNX1T1 functions in the occurrence and progression of OC.
Altogether, the key findings from the experiments of the present study suggest that EPB41L4A-AS2 can play a tumour inhibitory role in OC. EPB41L4A-AS2 was downregulated in OC tissues and cells, and highly expressed EPB41L4A-AS2 could impede the proliferation, colony formation, migration and invasion of OC cells by inducing RUNX1T1 expression by combining with miR-103a. Investigation of EPB41L4A-AS2 in OC cells and its function yields a better understanding of the in-depth mechanisms and may have potentially important therapeutic implications in the treatment of OC. In the future, further experiments will be warranted on how the expression of miR-103a and RUNX1T1 affects the development of OC in vivo.