The RNA-sequencing (RNA-seq) results and corresponding clinical data from 611 tissues and 530 cases of ccRCC samples were obtained from TCGA database (https://portal.gdc.cancer.gov). These data were current as of May 7, 2019. RNA-seq results of 72 normal samples and 539 cancer samples were combined into a matrix file using a merge script in Perl language (http://www.perl.org/). Next, the Ensembl database (http://asia.ensembl.org/index.html) was used to convert gene names from the Ensembl IDs to a matrix of gene symbols. The data was normalized using the R package edgeR and genes with expression levels of greater than one were retained. In the GEO database, GSE36133 (containing the expression profiles of 20 RCC cell lines determined by array) was selected and downloaded.

Using TCGA data downloaded in the previous step, the information was grouped according to sample type, tumor stage, and pathological grade to compare the expression of RUNX2. In the comparison process, 3 samples without stage information and 6 samples without grade information were excluded. Four cases lacked relevant follow-up information and were excluded, and the remaining 526 cases were divided into high and low groups according to the RUNX2 expression level for survival analysis by the log-rank test.

Both tumor samples from TCGA and data from 20 RCC cell lines were divided into two groups according to the median expression of RUNX2. GSEA (http://software.broadinstitute.org/gsea/index.jsp) was used to identify the potential functions correlated with RUNX2 by assessing whether a series of a priori-defined biological processes was enriched in the gene rank derived from whole genes between the two groups. In the molecular signature database v6.2, hallmark gene sets of curated gene sets were used. Terms with |normalized enrichment score| >1, nominal P-value <0.05, and false discovery rate of q-value <0.25 were identified.

ccRCC and corresponding noncancerous tissues were obtained from patients at the Department of Urology at the First Hospital of China Medical University, between December 2016 and January 2018. There was a total of 40 patients, 28 males and 12 females, aged 41–78 years, with mRNA samples from 14 patients (10 males, 4 females, age range 45–74 years) and protein samples from 26 other patients (18 males, 6 females, age range 41–78 years). The protocols used in the study were approved by the Hospital's Protection of Human Subjects Committee and written informed consent was obtained from all patients. The collected tissue samples were stored at −80°C prior to use.

Cells from the human RCC cell lines 769-P, 786-O and OS-RC-2 were cultured in RPMI-1640 medium (HyClone; GE Healthcare Life Sciences), ACHN cells were cultured in minimal essential medium with Earle's balanced salts (HyClone; GE Healthcare Life Sciences), and Caki-1 cells were cultured in McCoy's 5A medium (Gibco; Thermo Fisher Scientific, Inc.). HK-2 cells (normal cortex/proximal tubule cells) were cultured in Dulbecco's modified Eagle's medium/F12 (HyClone; GE Healthcare Life Sciences). All culture media were supplemented with 10% fetal bovine serum. All cell lines were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China) and cultured under humidified air containing 5% CO₂ at 37°C. When the cells exceeded 80% confluence, they were washed with 1X phosphate-buffered saline (PBS) and trypsinized at 37°C for a specified amount of time for cell passage cultivation.

Total RNA was extracted using ice-cold TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.) following the manufacturer's protocol. The concentration of total RNA was measured using Thermo Fisher Scientific NanoDrop ND-100. Total RNA was subjected to reverse transcription using the SYBR PrimeScript RT-PCR kit (Perfect Real-Time) (Takara Bio, Inc.). Real-time PCR was carried out using a Thermal Cycler Dice™ Real-Time system TP800 (Takara Bio, Inc.). The primer sequences designed for RUNX2 and β-actin were as follows (5′-3′): RUNX2 forward, GCGCATTCCTCATCCCAGTA and reverse, GGCTCAGGTAGGAGGGGTAA; and β-actin forward, CATGTACGTTGCTATCCAGGC and reverse, CTCCTTAATGTCACGCACGAT. The mRNA expression of the target gene was analyzed using the 2^−ΔΔCq method (8).

Whole-cell lysates were extracted in radioimmunoprecipitation assay buffer containing 1 mM phenylmethylsulfonyl fluoride (Beyotime Institute of Biotechnology), protease and phosphatase inhibitors. The concentration of protein samples was measured using a bicinchoninic acid assay (Beyotime Institute of Biotechnology). Equal amounts of protein (30 µg/lane) were separated by 10% SDS-PAGE (140 V) and the resolved proteins were transferred (350 mA) to polyvinylidene fluoride membranes (0.2 µm) using a Mini-Trans-Blot apparatus (Bio-Rad Laboratories, Inc.). The membranes were blocked with 5% non-fat milk at 37°C for 1 h, incubated with the primary antibody at 4°C overnight, and then incubated for 1 h with the appropriate secondary antibody. Proteins were finally detected with the EasySee^® Western Blot kit (Beijing Transgen Biotech Co., Ltd.). The immunoblots were quantified using ImageJ software (version 1.51; National Institutes of Health). The optical density values from the immunoblots for samples and cells were normalized to the density values acquired for β-actin and GAPDH, respectively. The monoclonal primary antibodies were: Anti-E-cadherin (dilution 1:5,000; product code ab40772), anti-N-cadherin (dilution 1:5,000; product code ab76011) and anti-vimentin (dilution 1:1,000; product code ab92547; all from Abcam), anti-p44/42MAPK (anti-ERK; dilution 1:1,000; product no. 4695), anti-phospho-p44/42MAPK (anti-p-ERK; dilution 1:2,000; product no. 4370), anti-RUNX2 (dilution 1:1,000; product no. 12556) and anti-GAPDH (dilution 1:1,000; D16H11; all from Cell Signaling Technology, Inc., and anti-β-actin (dilution 1:5,000; cat. no. 66009-1-Ig; ProteinTech Group, Inc.).

The 786-O and Caki-1 cells (cultured to 70% confluence) were seeded into 6-well plates at a density of 4×10⁵ cells/well and transfected with two effective RUNX2 siRNAs (JTS Scientific). siRNA transfection was performed using Lipofectamine 3000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the product's protocol. After 6 h, the medium was replaced, and the cells were cultured for another 24 or 48 h for further assays. The siRNA and negative control (NC) siRNA sequences used were as follows (5′-3′): siRNA-1 sense, CACGCUAUUAAAUCCAAAUTT and antisense, AUUUGGAUUUAAUAGCGUGTT; siRNA-2 sense, CAAGUCCUUUUAAUCCACATT and antisense, UGUGGAUUAAAAGGACUUGGT; and NC sense, UUCUCCGAACGUGUCACGUTT and antisense, ACGUGACACGUUCGGAGAATT.

Cell migration was measured using Transwell chambers with 8-µm pores in 24-well tissue culture plates (Corning Costar, Corning, Inc.). For cell invasion assays, Transwell chambers coated with Matrigel (BD Biosciences) were inserted into a 24-well plate. After transfection of siRNA for 24 h, the cells were re-suspended in medium without fetal bovine serum and 0.2 ml of cell suspension (2×10⁴ cells/well for cell migration; 4×10⁴ cells/well for cell invasion) was seeded into the top chamber, whereas the lower chamber of each well was filled with 0.6 ml of medium containing 10% fetal bovine serum to act as a chemoattractant. After 12–24 h of incubation at 37°C, the cells that remained on the upper side of the filter were removed using cotton swabs and those that had migrated to the lower side were fixed with 4% paraformaldehyde for 10 min, and stained with 1.0% crystal violet for 10 min at room temperature. Images were captured with an EVOS™ XL Core Imaging system (Invitrogen; Thermo Fisher Scientific, Inc.) and cells were counted using ImageJ software (version 1.51; National Institutes of Health).

786-O and Caki-1 cells were seeded into 24-well plates at a density of 1×10⁴ cells/well and cultured for 24 h. The solution of an EdU Kit (BeyoClick™ EDU Cell Proliferation Kit with Alexa Fluor 488; Beyotime Institute of Biotechnology) was diluted 1:1,000 in cell medium. The cells were incubated with the EdU solution for 2 h at 37°C. Following fixation in 4% paraformaldehyde for 15 min at room temperature and treatment with 0.3% Triton-X for 15 min at room temperature, the cells were incubated for 30 min with Click reaction cocktail in a dark room at room temperature. Prior to observation under a fluorescence microscope, nuclei were stained with Hoechst 33342.

STRING (https://www.string-db.org) version 11.0 was used with multiple protein modules for protein interaction network analysis.

The co-expression relationships between RUNX2 and related genes were determined based on gene expression levels to determine the strength of the relationships at the transcriptional level. The R corrplot package was used to calculate the Pearson correlation between genes.

All data were expressed as the mean ± standard deviation (SD) and represented as the average of at least three experiments with each experiment performed in triplicate. Statistical significance was determined using the Student's t-test (two-tailed) for 2 groups, one-way analysis of variance or (and) Tukey's test for more than 2 groups. Correlations between the expression levels of different genes were evaluated by Spearman correlation analysis. In the bar graphs *, **, ***, ****, and ns indicate P<0.05, P<0.01, P<0.001, P<0.0001, and not significant, respectively. A P-value <0.05 was considered to indicate a statistically significant difference.

Box-and-Whisker plots revealed RUNX2 expression levels (based on sample type) in normal and tumor samples (Fig. 1A). The expression was higher in all individual cancer stages than in adjacent normal tissues, and stage 4 exhibited the highest RUNX2 expression (Fig. 1B). Comparisons of the grades revealed that RUNX2 expression had a more evident increasing trend (Fig. 1C). Survival analysis revealed that the RUNX2 gene significantly affected prognosis (Fig. 1D). The aforementioned results indicated that RUNX2 is an important gene in ccRCC and its mRNA expression is upregulated during tumor development.

To identify the potential function of RUNX2, GSEA was conducted to search for pathways enriched in the higher-RUNX2-expressed samples. The functions and pathways potentially enriched by RUNX2 were separately analyzed in the previously downloaded TCGA cancer samples and GEO cell line data. The purity of the cancer samples affected our analysis results, however the cell lines were not influenced by purity, but had a small total sample size. Therefore, the screening conditions for the cell lines were relaxed and the results obtained to determine the functions and pathways potentially affected by RUNX2 were combined (Fig. 1E). The IL6/JAK/STAT3 signaling axis in cancer drives the proliferation, survival, invasiveness, and metastasis of tumor cells while strongly suppressing the antitumor immune response (9). In the IL2/STAT5 signaling pathway, STAT5 was revealed to play a critical role in the function and development of regulatory T cells (Tregs), and consistently activated STAT5 was associated with the suppression of antitumor immunity and increased proliferation, invasion, and survival of tumor cells (10). Epithelial-mesenchymal transition (EMT) is associated with tumor stemness, metastasis, and therapy resistance (11). The three pathways and mechanisms aforementioned were correlated with RUNX2 and promoted tumor progression. Overexpression of RUNX2 was related to the inflammatory response and allograft rejection (Fig. 1F).

The clinical tissue specimens collected from our hospital were analyzed to verify the results of bioinformatics analyses and it was confirmed that RUNX2 mRNA expression was significantly upregulated in tumor tissues compared to normal kidney tissues in 14 paired samples (Fig. 2A). To examine RUNX2 protein expression in clinical ccRCC specimens, another 26 pairs of cancer tissues and normal tissues were assessed. Western blot results indicated a significant increase of RUNX2 protein expression in tumor tissues compared to adjacent normal tissues (Fig. 2B and C). These data indicated that RUNX2 was highly expressed in kidney tumors and may be a reliable target for further investigation. Next, RUNX2 mRNA and protein levels were assessed in RCC cell lines and the cell line with the highest RUNX2 expression (780-O) and that with the lowest expression (Caki-1) were selected for subsequent experiments (Fig. 2D and E). Notably, the EMT pathway-associated proteins in 786-O were significantly different from those in Caki-1, consistent with the high activation of the RUNX2-mediated EMT mechanism predicted by GSEA (Fig. 2F).

Transwell assays were performed to explore the effect of RUNX2 on the migration and invasion of RCC cells. Decreasing RUNX2 expression with siRNA in 786-O and Caki-1 cells inhibited cell migration and invasion (Fig. 3A and B). Combined with upregulation of E-cadherin and downregulation of N-cadherin, vimentin and phosphorylated extracellular signal-regulated kinase (p-ERK) proteins after transfection, RUNX2 was revealed to be associated with the EMT pathway in renal clear cell carcinoma, which is consistent with the previous results of GSEA. The gray value of the western blot of RUNX2 in Caki-1 (Fig. 3A) was significantly higher than that of RUNX2 (Fig. 2E) in the previous cell line comparison. This was because the sample load was increased by 3-fold and the exposure time was extended to 6 min (exposure time in Fig. 3E is 2 min).

EdU assays revealed that cell proliferation was decreased when RUNX2 was silenced, indicating that RUNX2 enhanced tumor growth in vitro (Fig. 3C).

Overexpression of RUNX2 is related to ccRCC progression. Tumor progression is a complex biological process involving multiple cellular pathways and mechanisms. RUNX2 has been revealed to be involved in the EMT pathway, the IL6/JAK/STAT3 signaling pathway, and the IL2/STAT5 signaling pathway (screened by GSEA), wherein the relationship with the core enrichment genes may be direct or indirect. IL6/JAK/STAT signaling has been revealed to promote EMT (12). These enriched pathways and mechanisms have similar biological functions and thus the genes shared among them may be critical. The core enrichment genes of the EMT, IL6/JAK/STAT3 signaling, and IL2/STAT5 signaling pathways were analyzed to identify the genes revealing significant relationships with RUNX2, and then the genes were combined from TCGA and cell lines for subsequent analysis (Fig. 4A).

In STRING, the proteins from the 19 core genes mostly revealed protein interactions (Fig. 4B). Among them, IL6 exhibited the most extensive effect, whereas RUNX2 exhibited protein interactions with 5 other proteins. According to TCGA ccRCC tumor data, RUNX2 was also positively correlated with most core-related gene transcriptional levels normalized by log2 (Fig. 4C). CSF2 was ruled out because its expression was too low.