2.1 Clinical samples

Sixty-four tumor tissues were obtained from OvCa patients after surgical resection. Twenty-two normal fallopian tubal and ovarian tissues were collected from patients who received a total hysterectomy and bilateral salpingo-oophorectomy with benign gynecological diseases for use as controls in this study. All clinical samples were prepared into paraffin-embedded sections. Written consent was obtained before surgery. The study protocol was approved by the Ethics Committee. Table 1 summarizes the clinical information of the patients.

2.2 Immunohistochemistry (IHC) staining

After dewaxing and rehydrating paraffin sections, antigen retrieval was carried out in sodium-citrate buffer at 95°C. The endogenous peroxidase was then inactivated with 3% H₂O₂, and nonspecific binding sites were blocked with 10% goat serum. The sections were incubated with primary antibodies against m⁶A (1:200, Abcam, ab151230) at 4°C overnight and cultured with the biotin-conjugated IgG the next day. The sections were then stained with hematoxylin and 3,3'diaminobenzidine (DAB). By multiplying the intensity score by the percentage staining area, the total score was calculated (total from 0 to 9).

2.3 Source of datasets

The Cancer Genome Atlas (TCGA; http://tcga-data.nci.nih.gov/tcga) database was used to identify m⁶A-related miRNAs in OvCa. The expression datasets of the normal tissues were obtained from the Genotype-Tissue Expression (GTEx; https://gtexportal.org/home/) database. Other OvCa datasets, including GSE101976, GSE119055, GSE83693, GSE27651, GSE66957, and GSE27651, were downloaded from the Gene Expression Omnibus (GEO; www.ncbi.nlm.nih.gov/geo/) database.

2.4 Cell culture and transfection

Human OvCa cell lines (SKOV3, ES2, A2780, CAOV3, OVCAR3, and OVCAR4) were obtained from the China Center for Type Culture Collection (Wuhan University). OVCAR3 cells were cultivated in RPMI-1640 medium containing 20% fetal bovine serum (FBS, Gibco), whereas SKOV3, ES2, A2780, CAOV3, and OVCAR4 cells were cultured in DMEM/F12 medium supplemented with 10% FBS in a humidified incubator (37°C, 5% CO₂). The HNRNPA2B1 small interfering RNAs (siRNAs) and miR-30c-5p agomirs/antagomirs were obtained from RiboBio. Lipofectamine 3000 (Invitrogen) was used for transfection assays, which were performed according to the manufacturer's instructions. The corresponding sequences are listed in Table S1.

2.5 Quantitative real-time PCR (qRT-PCR)

Total RNA from OvCa cells was isolated using TRIzol reagent (Takara). The HiScript III qRT SuperMix Kit (Vazyme) and Prime-Script RT Master Mix Kit (Takara) were used to reverse transcription, and the relative level of each RNA was determined using SYBR Green (Vazyme). Table S1 listed the corresponding primer sequences. Each experiment has three replicates.

2.6 RNA m⁶A quantitative assay

The EpiQuik m⁶A RNA Methylation Quantification Kit (Epigentek) was used to measure the m⁶A level of total RNAs. The m⁶A level was quantified by reading the absorbance at 450 nm using a Spectramax plate reader (SpectraMax i3). Each experiment has three replicates.

2.7 Transwell assay

Forty-eight hours after transfection, 4 × 10⁴ cells in serum-free medium were seeded into the upper chamber without (Transwell migration assay) or with (Transwell invasion assay) Matrigel (BD Biosciences). After 24 h of incubation, non-migrated or invaded cells were scraped off using a cotton swab, and cells on the bottom of the chamber were fixed with methanol for 10 min and stained using 0.1% crystal violet. Then, five fields were selected and photographed randomly. Each experiment was repeated three times.

2.8 Wound healing assay

OvCa cells formed a confluent monolayer in six-well plates, wounded with a 200-μl pipette tip. After replacing the culture media with serum-free medium, the wound was closed after 24 h. Each experiment has three replicates. The wound healing areas were observed by and measured by ImageJ software (version 1.51).

2.9 5-Ethynyl-2′-Deoxyuridine (EdU) cell proliferation assay

According to the manual (RiboBio) to perform the assays, all images were taken with an Olympus fluorescence microscope of five random fields. All experiments have three replicates at least.

2.10 Colony formation assay

Five hundred cells per well were plated in 6-well plates and cultured until colonies were visible. After being fixed with 4% formaldehyde, the colonies were treated with 0.1% crystal violet. All experiments have three replicates at least.

2.11 miRNA target prediction and protein–protein interaction (PPI) network

Online prediction tools included miRWalk databases (http://mirwalk.umm.uni-heidelberg.de/), miRDB databases (http://mirdb.org/), miRTarBase databases (http://mirtarbase.mbc.nctu.edu.tw/index.html), and TargetScan databases (http://www.targetscan.org/vert_80/). The online websites Kaplan–Meier plotter (KM plotter) and UALCAN were used to determine the prognostic significance and the expression of genes.²⁴ In addition, the interaction among the genes was analyzed by the STRING database (https://string-db.org/). The Hub genes were analyzed by the Molecular Complex Detection (MCODE) plugin in Cytoscape (version 3.8.2). (degree cut-off = 2, node score cut-off = 0.2, k-core = 2, and max. Depth = 100).

2.12 Functional enrichment analysis

The clusterProfiler package was used to reveal the functions and pathways of miR-30c-5p target genes using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment. In addition, miRNACancerMAP was utilized to investigate the miR-30c-5p cancer pathways.²⁵

2.13 HNRNPA2B1 immunostaining

For immunocytofluorescence (ICF) assays, the cells were fixed with 4% formaldehyde. After incubating with 0.1% Triton X-100 and 3% bovine serum albumin (BSA), the cells were incubated with anti-HNRNPA2B1 antibody (1:400, Proteintech, 14,813-1-AP) overnight at 4°C. Then, Cy3–conjugated IgG and FITC phalloidin (Yeasen) were used to stain the cells, followed by DAPI staining. For immunocytochemistry (ICC) assays, the cells were treated with biotin-conjugated IgG at 37°C and stained with DAB, hematoxylin, and eosin. A fluorescent microscope was used to image the cell morphology (Olympus). All experiments have three replicates at least.

2.14 Western blotting

RIPA buffer was used to extract total cellular protein. Protein extracts were separated in 10% SDS-PAGE and transferred to PVDF membranes. Membranes were blocked with 5% defatted milk. After incubation with the primary antibodies at 4°C overnight, the membrane was treated with HRP-conjugated secondary antibody (1:5000, CST). Protein bands were detected using an advanced chemiluminescence kit (Pierce, ThermoScientific) in Molecular Imager ChemiDoc XRS+ and Image Lab software. The primary antibodies used were anti-HNRNPA2B1 (1:8000, Proteintech, 14,813-1-AP) and anti-β-actin (1:5000, Proteintech, 66,009-1-Ig). All experiments were repeated thrice at least.

2.15 Luciferase reporter gene assay

The wild-type or mutated sequence of the HNRNPA2B1 mRNA 3'UTR was cloned into pmirGLO vector. Cells were co-transfected with the specific luciferase reporter plasmids. The assays were detected by the Dual-Luciferase Reporter Assay (E1910, Promega). Each experiment was repeated at least three times.

2.16 Prediction of RNA-protein interaction and m⁶A site

The potential interaction of HNRNPA2B1 and miR-30c-5p was assessed by use of the RNA–protein interaction prediction tool (RPISeq), which is based on random forest (RF) or support vector machine (SVM). Predictions with probabilities >0.5 were considered positive. The online URL: http://pridb.gdcb.iastate.edu/RPISeq/references.php. The potential m⁶A sites were predicted using an online tool, SRAMP (http://www.cuilab.cn/sramp/).

2.17 RNA stability analysis

The cells were treated with mRNA transcription inhibitor actinomycin D (5 μg/ml) (MCE, HY-17559) for 0, 2, 4, and 6 h. Cells were collected, and the total RNA was extracted by TRIzol Reagent (Invitrogen) and analyzed by qPCR. Each experiment was repeated at least three times.

2.18 Statistical analysis

Statistical analyses were carried out by using R Statistical Software (version 3.6.3) and GraphPad Prism (version 8.0.1). All data are expressed as mean ± standard deviation (mean ± SD). Pearson's correlation analysis was used to measure correlation. Student's t test or one-way ANOVA was used to evaluate the differences between two or multiple groups, respectively. The Mann–Whitney test and Kruskal-Wallis test were used for nonnormally distributed data. Statistical significance was set at p < 0.05.

3.1 Dysregulation of m⁶A level and m⁶A regulators in OvCa

Analysis of the IHC revealed that the m⁶A level was markedly upregulated in OvCa tissues than in normal controls (Figure 1A,B and Figure S1). Moreover, the overall survival (OS) of patients with high m⁶A level exhibited was shorter than of those in the low group (Figure 1C). These results suggest that m⁶A modification involves in the progression of OvCa.

Considering that m⁶A modification is mediated by m⁶A regulators, we identified the levels of the 22 m⁶A regulators between OvCa tissues and normal controls using datasets from the TCGA and GTEx databases. The findings revealed that the 22 m⁶A regulators were expressed variably, indicating that the m⁶A alternation might participate in OvCa tumorigenesis and development (Figure 1D and Figure S2A).

3.2 Identification of m⁶A-related miRNAs in OvCa

To identify m⁶A-related miRNAs in OvCa, the TCGA database was used to identify the matrix expression of 22 m⁶A regulators and 2155 miRNAs. Then, we defined m⁶A-related miRNAs as miRNAs that had a strong relationship with one or more of the 22 m⁶A regulators. Pearson's analysis was performed to determine the correlation (|Pearson R| > 0.3 and p < 0.05), and the m⁶A regulators-miRNAs co-expression network was visualized by a Sankey diagram (Figure 2A). A total of 126 miRNAs were identified as m⁶A-related miRNAs, and univariate Cox regression analysis was implemented to distinguish prognostic-related miRNAs. As shown in Figure 2B, miR-30c-5p, miR-1248, miR-199a-5p, and miR-6504-5p were significantly related to the OS of OvCa patients. The correlation between 22 key m⁶A regulators and four OS-associated miRNAs in TCGA is shown in Figure 2C.

To single out the key m⁶A-related miRNAs in OvCa, survival analyses from the TCGA database were performed on the four miRNAs selected above. miR-30c-5p (p < 0.001) and miR-1248 (p = 0.004) were associated with OS, but miR-199a-5p (p = 0.169) and miR-6504-5p (p = 0.088) were not (Figure 2D and Figure S2B). The GSE101976 dataset showed that the OvCa patients with high level of miR-30c-5p had longer OS and progression-free survival (PFS), which is consistent with the TCGA database. Considering that the International Federation of Gynecology and Obstetrics (FIGO) staging of OvCa is closely related to prognosis, the level of miRNAs in different stages was analyzed, revealing that miR-30c-5p expression was decreased in patients with advanced-stage disease (FIGO stages III–IV) compared with early-stage disease (FIGO stages I–II); however, this was not the case for miR-1248 (Figure 2E and Figure S2C). The GSE119055 and GSE83693 datasets validated our results, revealing that the expression of miR-30c-5p is decreased in OvCa tissues compared with normal controls, especially in recurrent OvCa lesions (Figure 2H,I). These findings suggest that miR-30c-5p, an m⁶A-related miRNA, inhibits the progression of OvCa.

3.3 miR-30c-5p decreases m⁶A level and inhibits OvCa cells proliferation, migration and invasion in vitro

To investigate how miR-30c-5p regulates the biological function in vitro, we transfected miR-30c-5p agomir (an agonist) and miR-30c-5p antagomir (an inhibitor) into OvCa cells and performed the qRT-PCR assays to validate the transfection efficiency (Figure 3A). The RNA m⁶A quantitative experiment showed that the m⁶A level in the agomiR-30c-5p group was lower than that of the negative control (NC) group and vice versa (Figure 3B). In SKOV3 and ES2 cells, overexpression of miR-30c-5p reduced motility and invasion (Figure 5C-F). Meanwhile, the miR-30c-5p agomir markedly reduced the number of clones and EdU-positive proliferating cells (Figure 5G-I). Accordingly, the migrative, invasive, and proliferative abilities of OvCa cells increased after downregulation of miR-30c-5p. These results suggested that miR-30c-5p could decrease m⁶A level and inhibit the progression of OvCa in vitro.

3.4 miR-30c-5p targets m⁶A reader HNRNPA2B1

Four online miRNA target analysis websites (TargetScan, miTarBase, miRDB, and miWALK) were used to predict the potential target genes, and the 22 overlapping genes were recognized (Figure 4A). PPI networks of the 22 overlapping genes were then visualized in CytoScape, and the top three hub genes (XPO1, AGO1, and HNRNPA2B1) were clustered using MCODE (Figure 4B,C). As presented in Figure 4D,E, GO and KEGG enrichment analyses of the 22 overlapping genes were mainly enriched in the transcription and posttranscriptional modification and the JAK–STAT/HIF-1 signaling pathway. Moreover, the miRNACancerMAP online website showed that miR-30c-5p is closely related to the tumor-associated signaling pathways, including the AKT pathway and ErbB pathway (Figure S3A). These target genes appeared to play important roles in OvCa carcinogenesis by regulating cancer cell proliferation and motility. Hub genes are highly connected genes in gene expression networks and are inclined to play important roles in biological mechanisms. Further studies were performed to investigate the prognostic significance of the three hub genes. Kaplan–Meier survival curves and log-rank test analyses indicated that patients with high HNRNPA2B1 protein expression were related to decreased PFS and OS (Figure 4F). Meanwhile, HNRNPA2B1 expressed more highly in OvCa lesions than in normal tissues (Figure 4G), which was consistent with the results from GSE23554, GSE27651, and GSE66957 (Figure 4H,I), while XPO1 and AGO1 had the opposite expression and prognosis pattern (Figure S3B-E). In general, m⁶A reader HNRNPA2B1 plays oncogenic roles in OvCa.

Considering that the protein function is determined by its location, ICF and ICC assays were performed and revealed the nuclear location of HNRNPA2B1 in different OvCa cell lines (Figure 5A and Figure S4A). And, qRT-PCR assays and Western blotting assays were performed and revealed that miR-30c-5p agomir reduced HNRNPA2B1 expression and vice versa (Figure 5B,C). After validating transfection efficiency (Figure S4B), similar results were validated in other OvCa cells, including A2780, CAOV3, OVCAR3, and OVCAR4 cells (Figure S4C). Then, dual-luciferase reporter assays showed that miR-30c-5p significantly reduced the luciferase activity of the wild-type group of HNRNPA2B1, but no significant reduction was observed in the HNRNPA2B1 mutation group (Figure 5D,E). In conclusion, miR-30c-5p targets m⁶A reader HNRNPA2B1.

3.5 miR-30c-5p inhibits m⁶A level and OvCa cells proliferation, migration and invasion by targeting HNRNPA2B1 in vitro

To further study the biological function of HNRNPA2B1 in OvCa in vitro, small interfering RNAs (siRNAs) were used to decrease m⁶A reader HNRNPA2B1 expression. The silencing efficiency was validated by qRT-PCR and Western blotting assays (Figure S4D,E), and si#2 and si#3 were chosen to conduct the following experiments because of their relatively high silencing efficiency. Transwell migration assays and wound-healing assays showed that the number of migrated ES2 and SKOV3 cells significantly decreased after HNRNPA2B1. Moreover, Transwell invasion assays demonstrated that HNRNPA2B1 knockdown significantly reduced the invasion abilities (Figure 6A-D). Thus, HNRNPA2B1 knockdown inhibited the migration and invasion capacity of OvCa cells. EdU assays and cloning formation assays showed that the proliferation ability of OvCa cells, including ES2 and SKOV3 cells, decreased after knocking down HNRNPA2B1 (Figure 6E-G). The RNA m⁶A quantitative experiments were performed and showed that downregulation of HNRNPA2B1 reduced the m⁶A level (Figure 6H).

To further confirm that the function of miR-30c-5p in OvCa progression is dependent on HNRNPA2B1, EdU assays were performed and demonstrated that the miR-30c-5p antagomir increased the proliferation of OvCa cells (Figure 7A,C), which was attenuated by the downregulation of HNRNPA2B1. Transwell migration and invasion assays and m⁶A quantitative assay showed that HNRNPA2B1 downexpression rescued the cell mobility (Figure 7B,D) and the m⁶A level (Figure 7E) of miR-30c-5p knockdown cells. The miR-30c-5p antagomir increased the proliferation, migration, and invasion of OvCa cells (Figure 7A-D) and the m⁶A level (Figure 7E), which was attenuated by the downregulation of HNRNPA2B1. Collectively, we demonstrate that miR-30c-5p restrains OvCa progression by inhibiting the oncogenic activity of HNRNPA2B1 and the m⁶A level, suggesting that miR-30c-5p acts as the upstream regulator for the m⁶A modification system by targeting m⁶A reader HNRNPA2B1.

3.6 HNRNPA2B1 regulates the stability of CDK19 mRNA

Considering that HNRNPA2B1 has an important effect on the stabilization of mRNA, Pearson's correlation analysis was used to determine the correlation between the HNRNPA2B1 and mRNA in the TCGA database (Pearson R > 0.3 and p < 0.05). Then, univariate Cox regression analysis was performed and showed that eight genes had negative effects on survival among them (Figure 8A). Co-expression analysis of the eight genes in OvCa patients in the TCGA dataset was carried out using Pearson's correlation coefficients (Figure 8B). As shown in Figure 8C,D, qRT-PCR results revealed that HNRNPA2B1 knockdown could significantly downregulate CDK19 expression in ES2 and SKOV3 cells, indicating that CDK19 might act as a downstream target of HNRNPA2B1. RNA stability assays further identified that the half-life of CDK19 mRNA was reduced in ES2 and SKOV3 cells after knockdown of HNRNPA2B1 (Figure 8E,F). As shown in Figure 8G, we demonstrated that the possible interaction between HNRNPA2B1 and CDK19 mRNA was very high, which was assessed by RPISeq websites (RF classifier = 0.8, SVM classifier = 0.98). Meanwhile, several potential m⁶A sites were found on CDK19 mRNA using the online tool SRAMP (Figure 8H). These data suggest that HNRNPA2B1 might bind to the m⁶A site of CDK19 mRNA to promote the stability of CDK19 mRNA, which alters the m⁶A level of total RNA.