The human OC cell lines A2780, CAOV3, and SKOV3 were purchased from the National Infrastructure of Cell Line Resources in China (Beijing, China). Cells were cultured in DMEM or RPMI-1640 medium with 10% FBS (Thermo Fisher Scientific, Inc.) and incubated at 37°C with 5% CO₂. To block Notch signaling in OC cells, the γ-secretase inhibitor N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT, #HY-13027; MedChem Express) dissolved in DMSO was added in the medium.

Thirty-five epithelial serous OC and 16 pathologically confirmed normal ovary tissues were collected from patients (medium age 54 years; range 30–72 years and medium age 51 years; range 35–64 years, respectively) undergoing surgery at the Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (Beijing, China) from July 2018 to June 2019. All OC cases were diagnosed by pathological evaluation and had complete clinical information. All tissue samples were immediately frozen in liquid nitrogen after resection from patients and stored at −80°C for RNA extraction. This study was conducted according to the ethical guidelines of the 1975 Declaration of Helsinki and approved by the Ethics Committee of Peking Union Medical College Cancer Hospital (Beijing, China) (grant no: NCC2017G-115). Written informed consent was obtained from all of the subjects.

The OC samples were divided into high and low expression of miR-1299 and NOTCH3 groups by the cut-off value, which was defined as the cohort median.

miR-1299 mimics, inhibitors, and scramble miRNA controls were purchased from GenePharma (Shanghai, China). The lentivirus-containing short hairpin RNA (shRNA) targeting TUG1 or NOTCH3, and the plasmid expressing the active intracellular domain of NOTCH3 (NICD3) were purchased from GeneChem (Shanghai, China). The wild-type and mutated 3′UTR sequence of NOTCH3 was cloned into the pmirGLO vector (Promega) to construct recombinant plasmids named pmirGLO-NOTCH3-WT and pmirGLO-NOTCH3-MUT. Sequences of lncRNA TUG1 and XIST containing the predictive binding sites of miR-1299 and containing point mutations at the site region were cloned into the pmirGLO vector and were named pmirGLO-TUG1-WT, pmirGLO-XIST-WT, and pmirGLO-TUG1-MUT. OC cells were transfected with oligonucleotides using Lipofectamine^® RNAiMAX reagent (Thermo Fisher Scientific, Inc.) and plasmids using TurboFect™ transfection reagent (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions.

Total RNA was extracted from OC tissues and cultured cell lines using TRIzol reagent (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. mRNA and lncRNA levels were quantified with RT-qPCR using the SYBR Premix Ex Taq reverse transcription PCR kit (Takara, China), and β-actin was used as an internal control. miRNA was quantified with RT-qPCR using the hairpin-it™ microRNA and U6 snRNA normalization RT-PCR quantitation kit (GenePharma). Each assay was carried out in triplicates in a Light Cycler 480 Instrument (Roche), and the relative expression of mRNA and miRNA was calculated using the 2^−ΔΔCq method (33). The primers for RT-qPCR are shown in Table SI.

For cell cycle analysis, cells were collected and adjusted to 1×10⁶/ml 48 h after transfection. Ethanol (70%) was used to fix the cells, and RNase A (100 µg/ml) was used to remove RNAs. Finally, cells were stained with propidium iodide (PI) (50 µg/ml) at room temperature for 30 min and analyzed using LSRII flow cytometry (BD Biosciences). Cell viability was detected using the Cell Counting Kit-8 (CCK-8) (Dojindo, Japan) according to the manufacturer's instructions. Twenty-four hours after transfection, the cells were seeded in 96-well plates at 3,000 cells/well. The proliferative ability of cells was determined at 0, 24, 48, 72 and 96 h by measuring the absorbance values at a wavelength of 450 nm. For colony formation assay, OC cells (800 cells per well) were seeded into 6-well plates 24 h after transfection. After 10 days of incubation, the cells were fixed in methanol and stained with 0.1% crystal violet. The colonies were counted using a GBOX F3 gel documentation system (Syngene, Cambridge, UK).

A total of 10,000 cells were seeded on glass coverslips in 24-well plates and grown overnight. For the EdU incorporation assay, 10 µM EdU was used to treat the cells, and an EdU cell proliferation kit (Sangon) was used for measurement. Cellular nuclei were stained using 5 µg/ml Hoechst 33342 at 26°C for 30 min. Images were observed under a DMI 4000 confocal laser scanning microscope at ×200 magnification (Leica, Frankfurt, Germany).

Cells were seeded in 96-well plates at 1.5×10⁴ cells/well. When the cells reached 60% confluence, vectors containing the wild-type or mutant 3′UTR of NOTCH3 and wild-type or mutant binding site sequence of TUG1 were co-transfected with miR-1299 mimics and scramble miRNA using Turbofect (Thermo Fisher Scientific, Inc.). Forty-eight hours after transfection, luciferase activity was measured using the Dual-Luciferase Reporter assay system (Promega) and expressed as the ratio of firefly and Renilla luciferase activities.

Proteins were extracted from cells or tissues using RIPA lysis buffer (Solarbio) with protease inhibitors and phosphatase inhibitors. A total of 30 µg proteins were loaded per lane, separated on 10% SDS-PAGE gels, and blotted on polyvinylidene difluoride membrane. After being blocked with 5% skim milk for 2 h at room temperature, the membranes were incubated with primary antibodies (dilution 1:1,000) overnight at 4°C. The primary antibodies used in this study were NOTCH3 antibody (product code ab23426; Abcam) and GAPDH (cat. no. 2118; Cell Signaling Technology, Inc.). Anti-mouse IgG, peroxidase-linked antibody was used as secondary antibody (dilution 1:2,500; cat. no. 5174; Cell Signaling Technology, Inc.) and was incubated at room temperature for 1 h. The blots were detected using a chemiluminescence kit (cat. no. 34577; Thermo Fisher Scientific, Inc.) and imaged using MiniChemi 610 system (Sage Creation Science, Co., Ltd.).

Eight four-week-old female BALB/c nude mice (body weight range, 17.1–18.2 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. and randomly divided into two groups. The mice were housed with filtered air, 12 h light/dark cycle, constant temperature (25°C) and had free access to sterilized food and water. After 24-h transfection, 2×10⁶ cells containing miR-1299 or NC agomir were infected into the right armpit of the mice. miR-1299 agomir or NC agomir (RiboBio) was directly injected into the implanted tumor at the dose of 2 nmol/30 µl phosphate-buffered saline (PBS) per mouse every 6 days for 6 times. Tumor growth was monitored by measuring the tumor volume (V) every 6 days with a Vernier caliper and calculated as: V=length × width²/2. All the mice were anesthetized with sodium pentobarbital (500 mg/kg, intraperitoneally) and sacrificed by cervical vertebra dislocation on day 42 or when the tumor volume reached the threshold of 1,500 mm³, and tumors were weighed and snap-frozen for protein and RNA extraction.

Animal experiments were conducted with the approval of the Animal Ethics Committee of Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing (ACC2019A060) and in accordance with the Guide for the Care and Use of Laboratory Animals by the US National Institutes of Health.

SPSS software version 25.0 (IBM Corp.) and GraphPad Prism 8 (GraphPad Software, Inc.) were used for statistical analysis. Each experiment was performed at least in triplicate, and numerical data are expressed as means ± SD. The difference in clinicopathological features between two groups was determined by independent-sample Student's t-test or Mann-Whitney U test for continuous variables and the Chi-square test or Fisher exact test for categorical variables. One-way ANOVA analysis with Tukey post hoc test was used for comparisons among multiple groups. Pearson's correlation analysis was used to examine the relationship between two gene expression levels. A value of P<0.05 was indicative of statistical significance.

To investigate the potential miRNAs that regulate NOTCH3, we firstly performed a bioinformatics analysis using the miRWalk database (34), which integrated predicted gene-miRNA target information from 13 databases and compared the results with miRNAs reported to be significantly downregulated in previously published miRNA profiles (NCBI/GEO/GSE47841) (35) (Table SII). The two screening methods overlapped on only 11 miRNAs among which miR-1299 had the highest score in miRWalk. Therefore, we focused our subsequent analysis on miR-1299 as a putative regulator of NOTCH3 in OC. We measured the level of mature miR-1299 and NOTCH3 in 35 fresh OC tissues and 16 normal ovarian tissues, as well as in four OC cell lines. miR-1299 expression was significantly downregulated in the OC tissues compared with that in the normal tissue (P<0.05, Fig. 1A). There was an inverse correlation between the level of miR-1299 and NOTCH3 both in OC tissues (Fig. 1B) and in OC cell lines (Fig. 1C).

To confirm the regulation of miR-1299 on NOTCH3 expression, we first transfected miR-1299 mimics and NC scramble into miR-1299 low-expressing cell A2780 and CAOV3. RT-qPCR results showed that NOTCH3 and its pathway genes were largely reduced in the A2780 (Fig. 1D) and CAOV3 cells (Fig. S1) when transfected with miR-1299 mimics, and western blotting results confirmed the decreased protein level of NOTCH3 (Fig. 1E). In turn, when we knocked down NOTCH3 or overexpressed NICD3 in OC cells, we observed a subsequent increase or decrease in the miR-1299 level (Fig. S2). Furthermore, we cloned wild-type 3′UTR sequence of NOTCH3 and 3′UTR containing point mutations in the putative binding sites into a luciferase reporter plasmid (pmirGLO-NOTCH3-WT and pmirGLO-NOTCH3-MUT) (Fig. 1F). Both plasmids were co-transfected with miR-1299 mimics or NC mimics separately in A2780 cells. The results showed that overexpression of miR-1299 suppressed luciferase activity significantly in cells transfected with pmirGLO-NOTCH3-WT plasmid but not in those transfected with pmirGLO-NOTCH3-mutant vectors (Fig. 1G). These observations suggested that miR-1299 bound directly to the predicted binding sites in the NOTCH3 3′UTR region and negatively regulated NOTCH3 expression.

The association of miR-1299 level and clinicopathological features was also analyzed in OC patients (Table I). Low expression of miR-1299 was significantly correlated with low tumor differentiation. Decreased miR-1299 level was associated with younger age, advanced tumor stage, and necessity to receive neoadjuvant chemotherapy before surgery; however, the association was not statistically significant because of the limited number of cases. Thus, we considered miR-1299 as a regulator of NOTCH3 and hypothesized that decreased expression of miR-1299 promoted OC progression and development.

Subsequently, we investigated the role of miR-1299 in OC by transfecting miR-1299 mimics in OC cell lines A2780 and CAOV3 (Fig. 2A), both of which had an endogenous low miR-1299 expression. Overexpression of miR-1299 in OC cells significantly reduced cell proliferation, as observed with a CCK-8 assay compared to cells transfected with scrambled miRNA (NC mimic) (Fig. 2B). Colony formation assay showed that treatment with miR-1299 mimics decreased the number of colonies formed by OC cell lines (Fig. 2C). EdU incorporation assay revealed that miR-1299 inhibited DNA synthesis in cell proliferation (Fig. 2D). Additionally, cell cycle analysis showed that transfection of miR-1299 mimics significantly blocked the cells in the G0/G1 phase, accompanied by a smaller population in S and G2 phases (Fig. 2E).

We further examined the impact of miR-1299 on OC in a xenograft nude mouse model. A2780 cells transfected with miR-1299 or NC mimics were implanted subcutaneously in BALB/c nude mice. Starting on day 8 post-implantation, miR-1299 or NC mimics were injected intratumorally every 6 days for 6 times. In vivo tumor growth was evaluated by measuring tumor volume and final weight. We observed that treatment with miR-1299 mimics significantly inhibited tumor growth in vivo. The tumor volume and weight were significantly lower in the miR-1299 mimic group than in the NC group (Fig. 3A-C), substantiating the tumor suppressor function of miR-1299 in OC tumorigenesis. Moreover, decreased NOTCH3 protein level in the xenograft tissues with overexpression of miR-1299 was further confirmed (Fig. 3D).

To investigate the potential regulators of NOTCH3 by ceRNA mechanism, we first performed a comprehensive bioinformatics analysis using predicted miR-1299-targeted lncRNAs from DIANA databases (36), and OC-relevant lncRNAs from Starbase database (37) and lncRNA disease database (38) (Fig. 4A). The three-screening data overlapped on only two lncRNAs, TUG1 and XIST (X-inactive specific transcript). Therefore, we constructed luciferase reporter plasmids containing the two highest scoring putative miR-1299 binding sites in TUG1 (pmirGLO-TUG1-WT1 and pmirGLO-TUG1-WT2) and XIST (pmirGLO-XIST-WT1 and pmirGLO-XIST-WT2). After co-transfecting the plasmids with miR-1299 mimics or NC mimics in A2780, we found that overexpression of miR-1299 suppressed luciferase activity significantly in two wild-type plasmids of TUG1 (Fig. 4B and C) but not in those of XIST (Fig. S3). We also constructed plasmids carrying point mutations in either binding site of TUG1 (pmirGLO-TUG1-MUT1 and pmirGLO-TUG1-MUT2), and no significant changes in luciferase activity was observed in cells transfected with the two mutant plasmids (Fig. 4B and C). RT-qPCR further revealed that miR-1299 could be upregulated in OC cells after knockdown of TUG1 by lentiviral shRNA particles (Fig. 4D and E). Taken together, these results indicated that the lncRNA TUG1 was a direct sponge of miR-1299 in OC.

We next evaluated the potential tumorigenicity of TUG1 in OC. As expected, downregulation of TUG1 by shRNA resulted in the inhibition of OC cells in terms of cell proliferation, colony formation, EdU incorporation, and cell cycle, which could be partially rescued by inhibition of miR-1299 (Fig. 4F-I). Moreover, knockdown of TUG1 reduced NOTCH3 expression in OC cells, and miR-1299 inhibitors partially rescued the NOTCH3 level (Fig. 4J). This suggested that TUG1, acting as a ceRNA, affects NOTCH3 expression and promotes cell proliferation by sponging miR-1299.

In the NOTCH signaling pathway, after the cleavage of γ-secretase, the intracellular domain of NOTCH3 protein (NICD3) is translocated to the nucleus and binds the transcription factor complex RBP Jκ (39). When we analyzed the transcriptional start site (TSS, −2 kb to +1 kb) of TUG1 in the JASPAR CORE database (40), we found 5 RBP Jκ binding motifs around its TSS (Fig. 5A), indicating that TUG1 may be directly regulated by NOTCH3. Additionally, there was a significant positive correlation between the expression of TUG1 and NOTCH3 both in our tissues and in the TCGA database (41) (Fig. 5B and C). When we treated A2780 cells with DAPT, an inhibitor of γ-secretase, that can block the NOTCH signaling pathway without interference in NOTCH3 mRNA activity, we found a subsequent decrease in TUG1 expression in a concentration gradient manner (Fig. 5D). Therefore, we considered TUG1 as a potential downstream target of NOTCH3, and that there may exist a miR-1299/NOTCH3/TUG1 feedback loop in the development of OC (Fig. 5E).