RNAseq (lncRNA, mRNA and miRNA) and relevant clinical data for kidney renal clear cell carcinoma (KIRC) were obtained from the TCGA database (https://portal.gdc.cancer.gov/).

Based on the median expression of BAP1, patients were divided into the BAP1^high and the BAP1^low groups. For differential expression analysis, the cutoff value of DElncRNA was set at |log₂ FC| >0.5 and P. adj <0.05; DEmiRNA with cutoff value of |log₂ FC| >0.7 and P. adj <0.05; DEmRNA with cutoff value of |log₂ FC| >0.5 and P. adj <0.05.

Highly conserved microRNA family files were downloaded from the miRcode database (http://mircode.org/). Subsequently, the obtained DElncRNAs were used to find potential miRNAs targeting these DElncRNAs. Then, the selected miRNAs were inputted into the miRBD database (http://mirdb.org/) (31) and the Targetscan database (http://www.targetscan.org/vert_72/) (32) to explore target mRNAs. Finally, the screened DElncRNAs, DEmiRNAs, and DEmRNAs were put into Cytoscape (version 3.6.1) for ceRNA network construction, using the plug-in ‘cytoHubba’ for hub genes network construction (33, 34).

To investigate the functional annotation implicated with DEmRNAs, the Gene Ontology annotation and the Kyoto Encyclopedia of Genes and Genomes pathway analysis were performed using the Metascape website (https://metascape.org/) (35).

The expression of the hub genes in tumor and normal tissues based on the ccRCC dataset were compared using the Wilcoxon rank-sum test (P <0.05). Overall survival (OS) analysis for the expression of the hub genes between the high- and low-expression groups was performed, with P <0.05 indicating statistical significance.

To explore the clinical relevance of the ceRNA axis, the expression levels of NEAT1, miR-10a-5p, and SERPINE1 with different clinical characteristics were evaluated, and the Bonferroni method was used to correct the results of multiple hypothesis testing (Dunn’s test, P. adj <0.05). Moreover, univariate and multivariate Cox regression analyses were conducted to investigate the prognostic significance of clinical features.

The ACHN and 786-O cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA). Both cell lines were cultured in Dulbecco’s modified Eagle’s medium (Gibco) containing 10% fetal bovine serum (Gibco) and 1% penicillin/streptomycin and incubated at 37°C in a humidified 5% CO₂ atmosphere.

Cells were seeded into six-well plates and cultured until the cell density reached approximately 60%. The cells were then transfected with small-interfering RNA (siRNA-NEAT1#1, #2) or miR-10a-5p mimic/inhibitor, using jetPRIME^® transfection reagent (NY, USA) following the protocol of the manufacturer. Additionally, a relevant negative control (NC) was used. These siRNAs, miR-10a-5p mimic/inhibitor and relevant NC were designed by GenePharma (Shanghai, China).

Total RNA was extracted using RNAeasy™ (Beyotime; Shanghai, China), following the instructions of the manufacturer. HiScript^® II reverse transcriptase (Vazyme; Nanjing, China) was used to convert RNA to cDNA. RT-qPCR was performed in a LineGene 9600 Plus system (Bioer Technology, Hangzhou, China), using 2 × SYBR Green qPCR Master Mix (High ROX; Servicebio; Wuhan, China), following the protocol of the manufacturer. The expression of miR-10a-5p was normalized to control U6, while that of the others were normalized to GAPDH, and relative expression was calculated using the 2^−ΔΔCt method. Western blotting was performed as previously described (36), and anti-SERPINE1 (Rabbit Polyclonal Antibody, AF7965, Beyotime), anti-GAPDH (Mouse Monoclonal Antibody, AF5009, Beyotime) antibodies were used for further experimentation.

After 48 h of transfection, 100 µl cell suspension containing 1 × 10³ cells were seeded into 96-well plates and cultured. After 24, 48, 72, and 96 h, 10 µl of CCK-8 (Beyotime; Shanghai, China) was added to each well and incubated at 37°C for another 2 h. The absorbance was detected using a microplate reader at 450 nm.

After transfection for 48 h, the cell suspension was added to six-well plates with approximately 1 × 10³ cells per well and incubated at 37°C in a humidified 5% CO₂ atmosphere for 2 weeks. After culturing, the cells were washed twice with PBS, fixed with 4% paraformaldehyde for 15 min, stained with 0.1% crystal violet for 15 min, washed thrice with PBS again. The number of cell colonies with a diameter of >0.1 mm was further observed under the microscope.

The treated cells were seeded into six-well plates. When the cell density was close to 90%, a linear scratch was made using a 200 µl plastic pipette tip. The wound-healing time in different treatment groups were observed under a microscope and photographed at 0 and 12 h. The wound closure rate was measured thrice and averaged.

The dual-luciferase reporter gene plasmids containing wild-type and mutant sequences were synthesized using Promega (Madison, WI, USA). The wild-type or mutant (NEAT1, SERPINE1) reporter plasmid and miR-10a-5p mimic or mimic-NC were co-transfected into 293T cells. Luciferase activity was measured 48 h after transfection.

FISH analysis on the ACHN cells was performed to determine the subcellular localization of NEAT1 and miR-10a-5p. The RNA probes were designed and synthesized via Servicebio (Wuhan, China). Briefly, the cells were fixed with 4% paraformaldehyde for 20 min and washed with PBS. Then, proteinase K (20 ug/ml) was digested for 8 min, followed by PBS washes. Subsequently, a pre-hybridization solution was added for 1 h at 37°C and then discarded. RNA probes containing a hybridization solution was added to the cells and hybridized overnight at 37°C. Next, DAPI staining solution was added after the washes and incubated for 8 min, followed by rinsing and the addition of an anti-fluorescence quenching blocker dropwise to seal the slice. Finally, the slices were visualized under a Nikon fluorescence microscope.

To investigate the DNA methylation level of SERPINE1, co-expression analysis of SEPRINE1 was conducted using three DNA methyltransferases (DNMT1, DNMT3A, DNMT3B). Following this, methylation analysis of SERPINE1 was performed between tumor and normal tissues using the online database DiseaseMeth version 2.0 (http://bio-bigdata.hrbmu.edu.cn/diseasemeth/). MEXPRESS (https://mexpress.be/) was used to further determine the relationship between SERPINE1 expression and DNA methylation status (37).

TIMER (https://cistrome.shinyapps.io/timer/) was used for the comprehensive analysis of the relationship between SERPINE1 expression and tumor-infiltrating immune cell levels, namely, neutrophils, macrophages, dendritic cells, B cells, CD4⁺ T cells, and CD8⁺ T cells (38).

R software (x64, version 4.0.3) was used for the statistical analyses (https://www.r-project.org/).

Figure 1 presents a flow diagram of ceRNA construction and analysis. A pan-cancer analysis was performed to evaluate the RNA and protein expression level of BAP1 using data from the UCSC XENA (https://xenabrowser.net/datapages/) and the Human Protein Atlas (HPA) databases (https://www.proteinatlas.org/), respectively. BAP1 RNA expression was found to be significantly downregulated in ccRCC ( Figure 2A ), while BAP1 protein level was the lowest in renal cancer ( Figure S1 ). Furthermore, immunohistochemistry staining data obtained from the HPA validated BAP1 downregulation in tumor tissue ( Figure 2B and Table S1 ). Additionally, Kaplan–Meier survival curves indicated that the low expression of BAP1 was associated with poor OS in ccRCC ( Figure 2C ).

Moreover, the cBioPortal (http://www.cbioportal.org/) was used to explore the potential mechanisms underlying the abnormally low expression of BAP1 in ccRCC (39). Figure 2D shows that the genetic alteration rate of BAP1 was found to be 13%, with gene deletions (deep deletion and shallow deletion) accounting for more than half of the copy number alterations in ccRCC samples ( Figure 2E ). Additionally, the mRNA expression level of BAP1 was found to be positively correlated with the copy number value ( Figure 2F ).

DERNAs were screened according to the cut-off value ( Figure 3 ) and intersected with the predicted RNAs. The shortlisted genes were inputted into Cytoscape for hub genes network construction ( Figure 4A ). Finally, lncRNAs (NEAT1, HELLPAR, PURPL), miRNAs (miR-10a-5p, miR-508-3p, miR-135a-5p) and mRNAs (IRS1, SERPINE1, KCAN1, TRIM2, RORB, SIX4) were identified ( Figure 4B ). A functional enrichment analysis of DEmRNAs demonstrated their involvement in mesenchyme development, transmembrane receptor protein tyrosine kinase signaling pathway and cell adhesion regulation ( Figure 4C ).

Expression and survival analyses of hub genes are shown in Figures 5 , 6 . In total, two DElncRNAs (NEAT1, HELLPAR), one DEmiRNA (miR-10a-5p) and four DEmRNAs (SERPINE1, TRIM2, RORB, SIX4) were found to be prognostic-related genes. Moreover, the lncLocator (www.csbio.sjtu.edu.cn/bioinf/lncLocator/) showed that NEAT1 was mainly distributed in the cytoplasm ( Figure S2 ), indicating its role as a ceRNA in enhancing SERPINE1 expression by sponging miR-10a-5p. Finally, a prognostic-related NEAT1/miR-10a-5p/SERPINE1 ceRNA network was constructed ( Figure 7A ).

In this study, NEAT1 was correlated with gender (P <0.05, Figure S3A ). The lower expression level of miR-10a-5p was associated with higher T stage, M stage, pTNM stage, tumor grade and gender (P <0.05, Figure S3B ). Moreover, SERPINE1 was strongly correlated with T stage, N stage, pTNM stage, tumor grade and gender (P <0.05, Figure S3C ). The multivariate Cox regression analysis showed that NEAT1 (hazard ratio (HR) = 1.488, P = 0.011), SERPINE1 (HR = 1.456, P = 0.015) and miR-10a-5p (HR = 0.681, P = 0.014) were independent prognostic factors in ccRCC ( Tables S2–4 ). The AUC (area under the curve) of the receiver operating characteristics (ROC) analysis ( Figure S4 ) indicated a good prognostic performance of SERPINE1 (AUC = 0.789) and miR-10a-5p (AUC = 0.892). Additionally, the pan-cancer analysis showed that SERPINE1 mRNA was highly expressed in kidney cancer ( Figure S5A ). Furthermore, immunohistochemical analysis revealed that SERPINE1 was located in the cytoplasmic/membranous area ( Figure S5B ).

RT-qPCR analysis showed that NEAT1 was highly expressed in kidney cancer cell lines compared to that in HK-2, with more significant differences in 786-O and ACHN cell lines ( Figure 7B ). The knockdown efficiency of the two siRNAs was verified using RT-qPCR ( Figure 7C ). CCK-8 and colony formation assays showed that 786-O and ACHN cell proliferation was significantly suppressed after NEAT1 knockdown ( Figures 7D, E ). Wound healing assays showed that the migration ability of the cells was inhibited after transfection with siNEAT1 ( Figure 7F ).

The target site of NEAT1 to miR-10a-5p was predicted using the starBase (https://starbase.sysu.edu.cn/), and the wild-type and mutant sequences are shown in Figure 8A . A dual fluorescein reporter gene plasmid (NEAT1-WT/NEAT1-MUT) was constructed and co-transfected into 293T cells with miR-10a-5p mimic and miR-NC. Overexpression of miR-10a-5p significantly reduced luciferase activity in the NEAT1-WT group but not in the NEAT1-MUT group ( Figure 8B ), confirming that miR-10a-5p binds directly to NEAT1. Moreover, FISH assays revealed that miR-10a-5p co-localized with NEAT1 in the cytoplasm of ACHN cells ( Figure 8C ). Therefore, NEAT1 serves as a sponge for miR-10a-5p and inhibit its function in kidney cancer cells.

Next, the efficiency of the miR-10a-5p mimic and inhibitor on miR-10a-p expression was verified using RT-qPCR ( Figure 8D ). On transfection with siNEAT1 in ACHN and 786-O cells, miR-10a-5p expression was significantly enhanced ( Figure 8E ). Therefore, co-transfection siRNA and miR-inhibitor (siNC + inhibitor NC; siNC + inhibitor; siNEAT1 + inhibitor NC; siNEAT1 + inhibitor) in ACHN and 786-O cells, CCK-8 and colony formation assays suggested that the knockdown of NEAT1 on the suppression of cell proliferation and colony formation in ACHN and 786-O cells could be reversed by a miR-10a-5p inhibitor ( Figures 8F, G ).

To confirm that miR-10a-5p regulates the expression of SERPINE1 by binding directly to the target sequence, a luciferase reporter gene plasmid was constructed for SERPINE1-WT/MUT ( Figure 9A ). The results showed that miR-10a-5p mimic significantly reduced the luciferase activity of SERPINE1-WT; however, no significant changes were observed for SERPINE1-MUT ( Figure 9B ). siNEAT1 was co-transfected with miR-10a-5p inhibitor into ACHN and 786-O cells and evaluated using RT-qPCR and western blot to verify the regulation of SERPINE1 expression. The results demonstrated that the knockdown of NEAT1 significantly downregulated the expression of SERPINE1; however, this effect could be reversed by a miR-10a-5p inhibitor ( Figures 9C, D ).

To elucidate the mechanism of the abnormally high expression of SERPINE1 in ccRCC, a series of methylation analyses of SERPINE1 was performed. Co-expression analysis suggested that SERPINE1 expression was positively correlated with DNMT1, DNMT3A, and DNMT3B expression levels (P <0.05, Figure 10A ). Additionally, the methylation level of SERPINE1 in normal tissues was much higher than ccRCC tissue samples (P <0.001, Figure 10B ). Moreover, 12 DNA methylation sites that were negatively correlated with SERPINE1 expression were identified ( Figure 10C ).

To further investigate the relationship between SERPINE1 expression and the immune microenvironment in ccRCC, an immune infiltration analysis was performed using TIMER. ‘SCNA’ module analysis indicated that the immune infiltration level of CD 4⁺ T cell was associated with the altered copy numbers of SERPINE1 ( Figure 11A ). Moreover, ‘Gene’ module analysis showed that the expression of SERPINE1 was positively related to the immune infiltration level of CD4⁺ T cell, CD8⁺ T cell, macrophages, dendritic cells and neutrophils ( Figure 11B ).