Integrative bioinformatics and experimental analysis revealed down-regulated CDC42EP3 as a novel prognostic target for ovarian cancer and its roles in immune infiltration

Cell culture

The human ovarian epithelial cell IOSE80 and ovarian cancer cells A2780 and TOV112D were obtained from Center for Molecular Medicine, Xiangya Hospital, Central South University. All cell lines were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, USA) with 10% fetal bovine serum (FBS, Gibco, USA) and 1% penicillin/streptomycin (Gibco, USA). The cultures were placed in a sterile incubator maintained at 37 °C with 5% CO2.

Western blot

To begin this procedure, cell lysates were processed with RIPA buffer supplemented with protease inhibitor. The total protein concentration was detected by BCA protein assay kit. Equivalent amounts of protein were loaded in each lane of SDS-polyacrylamide gel electrophoresis. Besides, the bands were transferred to PVDF membranes. Next, the membranes were blocked for 1 h at room temperature with TBST supplemented with 5% non-fat dry milk and then incubated at 4 °C overnight with primary antibodies. The primary antibodies displayed as follows: CDC42EP1 (1:500; Abcam, USA) and β-actin (1:5000; Santa Cruz, USA). Finally, protein bands were detected through the Immobilon Western Chemiluminescent HRP reagents (WBKLS0500; Millipore, USA).

Quantitative PCR (qPCR)

Total RNA was extracted from disparate cell lines using TRIzol reagent (Invitrogen, USA) following the manufacturer’s protocol and then converted to cDNA using a PrimeScriptTM RT reagent kit (6210; Takara, China). The qPCR assay was performed with iTaqTM Universal SYBR green Supermix (1725121, Bio-Rad, USA) to determine the relative RNA levels. β-actin was used as an internal control for quantification of each gene. The sequences of gene primers displayed as follows: CDC42EP3 forward: 5′-AGCAGTCTGTTGGAGAATGGG-3′and reverse: 5′-AGGAGGGAACCTGTAAGGTCAG-3′; β-actin forward: 5′-CATGTACGTTGCTATCCAGGC-3′and reverse, 5′-CTCCTTAATGTC ACGCACGAT-3′. The relative expression levels of RNAs were determined using the 2-ΔΔCT method (Buford et al., 2020; Kiss et al., 2019; Pandelides et al., 2020).

Data acquisition and processing

The integrative bioinformatics analysis of differential expressed genes was obtained from several bioinformatics databases (Table 1).

The gene expression profiles from two public datasets regarding ovarian cancer were obtained from the Oncomine datasets (Rhodes et al., 2007), Yoshihara Ovarian and Bonome Ovarian. The GEO2R, an R-based web application (Barrett et al., 2013), was applied to assess the co-differential expressed genes in these two datasets. The screening criteria to identify the differential expression genes were as follows: p-value < 0.01 and — log₂Fold Change —≥ 1.5. In addition, the N⁶-methyladenosine (m⁶A) target genes of ovarian cancer were acquired from the RNA modification-associated variants database (RMVar) (Luo et al., 2021).

UALCAN (Chandrashekar et al., 2017) integrates the RNA-seq data and clinical information of pan-cancer types from The Cancer Genome Atlas (TCGA). This database could enable researchers to explore the expression level of candidate genes between primary tumor and normal tissue samples.

PrognoScan (Mizuno et al., 2009) is utilized to analyze the relation between gene expression and clinical prognosis of patients, including overall survival (OS). In this research, PrognoScan was applied for the survival analysis of differentially expressed genes (DEGs) in patients with ovarian cancer. In addition, the Kaplan Meier (K-M) plotter (Nagy, Munkacsy & Gyorffy, 2021) is capable to assess the effect of candidate genes on the prognosis in pan-cancer. The K-M survival curves were applied to exhibit the prognostic effects of CDC42EP3 in ovarian cancer patients.

LinkedOmics (Vasaikar et al., 2018) provides a unique platform to analyze multi-omics data across pan-cancer types. The DEGs associated with CDC42EP3 were screened from the TCGA-OV cohort through LinkFinder module in this database. The statistical method used in correlation analysis was Pearson correlation and the results of correlation were presented in volcano plot and heat maps, respectively. The LinkInterpreter module was performed to analyze the Gene Ontology molecular function (GO_MF) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway.

The TISIDB (Ru et al., 2019) is a web portal for tumor and immune system interplay, which contains multiple heterogeneous data. Here, TISIDB was employed to investigate the roles of CDC42EP3 in several immune-associated signaling pathways, such as tumor-infiltrating immune cells, immunomodulators, chemokines and chemokine receptors.

Statistical analyses

All experimental findings were shown as mean ± standard deviation (SD). Student’s t-test was used to explore the difference between two groups. P < 0.05 was considered as statistically significant difference.

Identification of differentially expressed genes

To screen for DEGs in ovarian cancer, we compared normal ovary tissues and tumor samples in two datasets from Oncomine, Yoshihara Ovarian and Bonome Ovarian. According to the filtering conditions of p-value < 0.01 and —log₂Fold Change—≥1.5, we identified 225 up-regulated genes in Bonome Ovarian and 1085 up-regulated genes in Yoshihara Ovarian, respectively. Meanwhile, 536 genes in Bonome Ovarian and 623 genes in Yoshihara Ovarian have been identified to be significantly down-regulated in ovarian cancer tissues (Table S1), and 757 m⁶A target genes of ovarian cancer acquired from the RMVar database were shown in Table S2.

Next, the co-DEGs between Yoshihara Ovarian and Yoshihara Ovarian were analyzed by Venn diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/). Studies have demonstrated that the m⁶A RNA modification, a reversible and dynamic epigenetic modification, affects the RNA metabolism and homeostasis, and plays critical functions in multiple of important physiological and pathological bioprocesses, including tumorigenesis and progression (Xu et al., 2020a). In order to investigate the possible roles of m⁶A target genes in ovarian cancer, Venn diagram was used to display the overlapping genes between co-DEGs and m⁶A target genes. The results showed that one m⁶A target gene, regulatory factor X subunit B (RFXANK), was highly-expressed, and two m⁶A target genes, CDC42EP3 and BMP and activin membrane bound inhibitor (BAMBI), were lowly-expressed (Figs. 1A–1B). These three genes were presumed to have a significance impact on the development of ovarian cancer.

CDC42EP3 shows the promising prognostic value in ovarian cancer

The correlations between the expression levels of RFXANK, CDC42EP3 and BAMBI and prognosis in ovarian cancer patients were analyzed by the PrognoScan database. The results revealed that patients with high level of RFXANK and CDC42EP3 displayed favorable OS (Figs. 2A–2B). However, patients with high level of BAMBI displayed poor OS (Fig. 2C). Given the up-regulated RFXANK and down-regulated CDC42EP3 and BAMBI in two independent ovarian cancer datasets from Oncomine (Figs. 1A–1B), we expected CDC42EP3 as a potential prognostic target that needs further investigation.

Associations between CDC42EP3 expression and clinicopathological parameters in ovarian cancer patients

The expression profiles of CDC42EP3 were further confirmed using several independent bioinformatics databases, such as TNM plot (Bartha & Gyorffy, 2021) and UALCAN. Firstly, two above-mentioned datasets, Yoshihara Ovarian and Bonome Ovarian, indicated that CDC42EP3 transcriptional level in ovarian cancer samples was significantly lower than adjacent noncancerous tissues (Figs. 3A–3B). Next, TNM plot revealed that CDC42EP3 mRNA expression levels were significantly down-regulated in ovarian cancer samples (Fig. 3C). In addition, qPCR and western blot were used to confirm the down-regulation of CDC42EP3 expression in ovarian cancer cells A2780 and TOV112D compared with the normal ovarian epithelial cell IOSE80 (Figs. 3D–3F). All these data revealed that the expression of CDC42EP3 was significantly downregulated in ovarian cancer tissues and cell lines.

Furthermore, we analyzed the correlation between CDC42EP3 expression levels and clinicopathological features of ovarian cancer patients. According to UCLCAN database, we discovered that the higher tumor grade, the lower expression levels of CDC42EP3 in ovarian cancer (Fig. 3G). However, the Xiantao tool (http://www.xiantao.love/products) was analyzed to reveal that CDC42EP3 expression has no correlation with the other clinical parameters, including tumor status, stage, age and invasion (Table 2). In addition, the K-M plotter uncovered that the expression of CDC42EP3 was significantly positively correlated with OS in high-grade tumor patients (p < 0.05) (Fig. 3H). However, the correlation between CDC42EP3 level and OS in low-grade patients was of no statistical significance (Fig. 3I). Collectively, these results indicate that CDC42EP3 could be a potential prognostic biomarker for ovarian cancer, especially like the high-grade patients.

CDC42EP3 co-expression network in ovarian cancer

In order to explore the biological functions of CDC42EP3 in ovarian cancer, we performed the co-expression pattern of CDC42EP3 screened from the TCGA-OV cohort by the LinkFinder module of LinkedOmics. As shown in Fig. 4A, 4227 genes (red dots) were positively correlated with CDC42EP3, and 3122 genes (green dots) were negatively correlated with CDC42EP3. Figures 4B–4C showed the heatmaps of top 50 genes positively and negatively associated with CDC42EP3, respectively (Tables S3–S4). 14/50 positively correlated genes had protective hazard ratio (HR), while 13/50 genes negatively correlated with CDC42EP3 had unfavorable HR (Fig. 4D).

To further elaborate the potential biological roles of CDC42EP3 in ovarian cancer development, we performed GO and KEGG functional enrichment analysis. GO enrichment analysis revealed that co-expression genes of CDC42EP3 mainly participated in extracellular matrix structural constituent, growth factor binding, collagen binding, extracellular matrix binding, Wnt-protein binding, etc (Fig. 4E). KEGG pathway analysis indicated these genes were significantly related to several cancer-associated signaling pathways, such as focal adhesion, extracellular matrix (ECM)-receptor interaction, Hippo signaling pathway, hedgehog signaling pathway, basal cell carcinoma, Wnt signaling pathway, etc (Fig. 4F). These findings suggested that important functions of CDC42EP3-associated network in ovarian cancer.

Relation between CDC42EP3 with immune molecules

We explored whether CDC42EP3 expression can affect the abundance of tumor-infiltrating lymphocytes (TILs) in ovarian cancer using Xiantao Tool. The significantly positive relations could be obtained between CDC42EP3 expression and several tumor-infiltrating immune cells, including natural killer cells (NK), T central memory cells (Tcm) and T gamma delta cells (Tgd) (Figs. 5A–5B). Similar results were obtained by analyzing the immune-related signatures of 28 TIL types through TISIDB algorithm (Figs. S1A–S1B).

Furthermore, we used TISIDB database to analyze the correlation between CDC42EP3 expression and other immune signatures, including immunostimulators, immunoinhibitors, chemokines and chemokine receptors. Figs. S2A–S2B exhibited the positively correlated immunostimulators with CDC42EP3, including 5′-nucleotidase ecto (NT5E), C-X-C motif chemokine ligand 12 (CXCL12), tumor necrosis factor superfamily member 4 (TNFSF4), and TNF receptor superfamily member 8 (TNFRSF8). Figures S2C–S2D uncovered the negatively correlated immunoinhibitors with CDC42EP3, including galectin 9 (LGALS9), indoleamine 2,3-dioxygenase 1 (IDO1), cluster of differentiation 274 (CD274) and V-set domain containing T cell activation inhibitor 1 (VTCN1). Figures S3A–S3B showed the significant associations between CDC42EP3 expression and chemokines, such as C-C motif chemokine ligand 14 (CCL14). Figures S3C–S3D revealed the associations between CDC42EP3 expression and chemokine receptors, including C-C motif chemokine receptor 1 (CXCR1). Generally, we concluded that CDC42EP3 might participated in modulating diverse of immunoregulatory molecules in ovarian cancer, which could affect the infiltration and functions of immune cells in TME.