The recognized functioning principles of the circRNA–miRNA–mRNA axis guided the selection. Before selecting circRNAs, miRNAs, and mRNAs from the relevant Gene Expression Omnibus (GEO) datasets, Log2 Fold Change (log2FC) and p-values were considered. The overexpressed circRNAs were the intended focus of this research. Downregulated miRNAs were considered for sponge miRNAs. The selection of mRNAs was also limited to elevated ones.

GEO is a publicly available functional genomic data resource. Microarrays or sequence-based studies are accepted. Users can use the supplied tools to query and download experiments and determine gene expression profiles. Identifying molecules (circRNA, miRNA, mRNA, etc.) whose expressions alter in common among various datasets may be useful to obtain more accurate information about the association between these molecules and diseases before in vitro or in vivo studies.

Gene Expression Omnibus 2R (GEO2R) is an interactive web application that enables users to compare two or more sets of samples from a GEO series to find differentially expressed genes, miRNAs, circRNAs, etc. GSE102686 dataset (including 5 CC patient’s tissue samples and 5 adjacent normal tissue samples) and GSE169057 (including 3 CC patient tissue samples and 3 adjacent normal tissue samples) datasets were downloaded from the National Center for Biotechnology Information (NCBI) Geodatabase and the differential expression of circRNAs between tumor and normal samples was analyzed via GEO2R. When selecting upregulated circRNAs from the datasets, first we paid attention to the log2FC and p-value of circRNAs (log2FC > 1, p < 0.001). circRNAs that met these criteria and overlapped in both datasets were selected.

The Cancer Genome Atlas (TCGA) (https://cancergenome.nih.gov/) is a fundamental cancer genomics program, that molecularly analyzed approximately 20,000 primary cancers and matched normal samples from 33 different types of cancer. In the TCGA database, miRNAs with log2FC > 2 and p < 0.05 were identified in 307 CESC tissue samples and matched 3 normal tissue samples. UALCAN [15] and CancerMIRNome [16] are user-friendly platforms for investigating miRNA dysregulation in various cancer samples from the TCGA. For miRNA expression analysis these web tools were used.

The Gene Expression Profiling Interactive Analysis (GEPIA2) [17] is beneficial and well accepted online tool for gene expression, survival analysis, etc., using tumor and normal specimens from the TCGA and Genotype-Tissue Expression (GTEx) databases. Overexpressed genes with log2FC > 2 and p < 0.01 criterion were chosen in TCGA constructed from 307 CESC patients and 3 adjacent normal tissue samples using GEPIA2 tool (ANOVA method was selected using GEPIA2 tool).

circRNA interactome (CircInteractome) identifies miRNAs that might possibly target the circRNAs using the Targetscan prediction tool [18]. The CircInteractome web tool was used to identify potential sponge miRNAs of selected circRNAs from the GSE102686 and GSE169057 datasets. miRNet 2.0 is a user-friendly, web-based tool that integrates user data with existing information using network-based visual analytics to assist in unraveling miRNA functions [19]. The putative target genes of the selected miRNAs were defined via miRNet 2.0. Then, those matched genes with the log2FC > 2, p < 0.01 criterion in the TCGA database were chosen.

The biological processes, cellular components, and molecular activities in which the selected genes were involved were determined by creating the EnrichR tool (https://maayanlab.cloud/Enrichr/, pathways with p < 0.05 were considered).

Selected genes’ correlation analysis was constructed via Correlation AnalyzeR [20] and GEPIA2 (http://gepia2.cancer-pku.cn/) tools.

Kaplan-Meier plotter (KM plotter) is an online platform for determining genes/miRNAs’ expression and survival rates in many cancer types. It also enables publicly accessible transcriptome data, such as TCGA. The prognostic values of chosen miRNAs in CESC were determined utilizing the KM plotter database. In brief, the miRNAs were entered into the database. Then KM survival plots were constructed, with the hazard ratio (HR), 95% confidence intervals (CI), and log-rank p-value provided on the website. p < 0.05 were considered statistically significant.

Publicly available platforms were used to conduct functional enrichment assessments. When choosing upregulated circRNAs from the datasets, we evaluated the log2FC and p-value (for GSE102686 analysis: log2FC > 1, p < 0.001; for GSE169057 dataset: log2FC > 1; p < 0.001). For DEmiRs log2FC > 2 and p < 0.05; overexpressed genes with log2FC > 2 and p < 0.001 were used. The KM method was utilized for determining overall survival (OS), and the log-rank test was used to determine differences (via KM plotter tool). The statistical cut-off for OS analysis and enrichment analysis in tools was set at a p < 0.05. GraphPad Prism 10 was employed to perform statistical calculations and visualize figures.

In GSE102686, 137 upregulated circRNAs met the log2FC > 1, p < 0.001 criteria among the circRNAs whose expression changed in cancer tissue samples compared to normal tissue samples; in GSE169057, 145 upregulated circRNAs met the log2FC > 1, p < 0.001 condition. In both datasets, three overlapping circRNAs (hsa_circ_0000711, hsa_circ_0000515, and hsa_circ_0001855) have been chosen by the required parameters (Table 1).

Following study on the circinteractome tool that suggested potential targets of the three selected circRNAs, 26 miRNAs were identified as common targets as seen in Figure 1. Three hundred fifty-nine upregulated miRNAs met the log2FC > 2 and p < 0.05 requirements in the TCGA database in CESC tissue samples compared to normal tissues. The 26 miRNAs were screened among these 359 miRNAs to detect overlapping miRNAs. Six possibly sponge miRNAs were overlapped. Among these 6 miRNAs, 3 miRNAs (hsa-miR-296-5p, hsa-miR-338-3p, and hsa-miR-361-3p) with decreased expression were selected (because selected circRNAs were upregulated) (Table 2).

Among the overexpressed genes in CESC tumors compared to normal tissues in the TCGA database, it was defined that 778 genes met the conditions of log2FC > 2 and p < 0.05.

Using miRNet 2.0 and miRTarBase v8, the target genes of the 3 selected miRNAs were examined and 355 genes matching these miRNAs were identified (Figure 2). Seven hundred seventy-eight genes which were identified from the TCGA database were compared to 355 genes shown as probable targets of 3 chosen miRNAs, in the end, 10 overlapping genes were selected (Table 3). The heatmap of these 10 genes that are upregulated in CESC is shown in Figure 3. In summary, three circRNAs were chosen because they had the potential to regulate the expression of ten genes via three miRNAs.

Pathway analysis showed that selected potential 10 target genes have been associated with many critical molecular cellular processes like cell cycle, and DNA replication Figure 4A. Enrichment analysis results showed that these genes could have crucial roles in the cancer process as seen in Figure 4B. They are also connected with many essential hub proteins, including the MYC oncogene and the TP53 tumor suppressor gene, as demonstrated in Figure 4C. A correlation study of selected genes revealed that four of them are connected with CDK1 and CDK2 genes, which are known key cancer-related genes in CESC, as can be seen in Figure 5.

The all 3 selected miRNAs were found to have reduced expression in CESC samples compared to the control group (Figure 6). OS analysis in TCGA for CESC patients was performed for 3 selected miRNAs and 10 genes. It was found that the miR-361-3p and miR-296-5p miRNAs have prognostic importance for the OS rate of CESC patients (Figure 7). However, any one of the genes did not have a statistically significant effect on the OS of the CESC patients.

According to the findings, this reserch suggest that hsa_circ_0000711 may be involved in the CESC process hsa-miR-338-3p/HMGA1, HIST1H2BJ, ZWINT, MMP9, KIFC1 axis and also via hsa-miR-361-3p/KIFC1, UBALD2, CDC6 axis. Moreover, the results suggest that hsa_circ_0000515 may affect CESC processes by changing the expressions of S100A2, PLK1, SLC7A5, and HMGA1 genes through hsa-miR-296-5p. These circRNAs–miRNAs–mRNAs axes, which are considered to be associated with CESC, should be confirmed in vitro and in vivo with new investigations.

Although this study was conducted using TCGA data containing a large number of patient samples in mRNA and miRNA selection, GSE102686 and GSE169057 GEO datasets containing a small number of patient samples were used in determining circRNAs. The small number of patient samples can be considered as a limitation of the study. Despite this, circRNA is a new topic, we considered that our study can be considered a clue for in vitro and in vivo studies.