Serum samples of ovarian cancer patients and patients with uterine fibroids in our hospital were collected, exosomes were extracted, and whole transcriptome sequencing was performed to detect differences in expression. The experimental group was ovarian cancer patients (the samples were preoperative serum samples of ovarian cancer patients), the postoperative histopathological results were malignant, and the control group was patients with uterine fibroids (the samples were preoperative serum samples) who had no previous ovarian disease. The inclusion criteria were as follows: 1) those who met the diagnostic criteria of ovarian cancer (Lokshin, 2012) and 2) those with a telephone follow-up. The exclusion criteria were 1) those with severe cardiac, hepatic, and renal dysfunction, 2) those with other malignant tumors or systemic infectious diseases, 3) those with other gynecological diseases, 4) those who withdrew from the study before completion, and 5) those with incomplete clinical data.

The serum was thawed in medium at 37°C for 30 min and centrifuged at 2000 × g and 4°C, and the supernatant was transferred to a new centrifuge tube and centrifuged again at 10,000 × g and 4°C for 45 min to remove larger vesicles. The supernatant was extracted and filtered through a 0.45 μM filter membrane, and the filter was collected. The filter was transferred to a new centrifuge tube, and an ultrarotor was selected and used at 4°C and 100,000 × g for 70 min. For removal of the supernatant, the supernatant was resuspended in 10 ml of prechilled 1 × PBS, and an ultrarotor was selected for centrifugation at 4°C and 100,000 × g and ultracentrifugation for 70 min. The supernatant was removed and resuspended in 100 μl of prechilled 1 × PBS. A total of 20 μl was collected for identification under electron microscopy, and the remaining exosomes were stored at −80°C.

Ten microliters of the exosomes was removed, 10 μl of the sample was pipetted dropwise onto a copper grid to precipitate for 1 min, and filter paper was pipetted off the floating liquid. Uranium acetate (10 μl) was added dropwise onto the copper grid to precipitate the sample for 1 min, and filter paper was pipetted off the floating liquid. The samples were dried at room temperature for several minutes at 100 kV for electron microscopy imaging (Rikkert et al., 2019). TEM imaging results were obtained.

The sequencing data were background corrected and normalized, and expression values were calculated using the Bioconductor R package in R. The DESeq2 package in R was used to calculate the differentially expressed lncRNAs and mRNAs between the two groups, with p < 0.05 and the magnitude of the expression change ≥ twofold [|log2 fold change (FC)| ≥ 1.0]. We screened the differentially expressed genes (DEGs), log2 FC ≥ 1.0 represents lncRNAs and mRNAs with upregulated expression, and log2FC ≤ −1.0 represents lncRNAs and mRNAs with downregulated expression. Finally, the differentially expressed lncRNAs (DElncRNAs) and mRNAs, that is, DElncRNAs as well as DEGs in the ovarian cancer and uterine leiomyoma control groups, were obtained. Heatmap plotting and cluster analysis of the screened DElncRNAs and DEGs were performed using the Heatmap package, and the p values in the data were −log10 transformed and −log10 (p values) grouped according to log2 FC (the upregulated lncRNA group, downregulated lncRNA group, and lncRNA group without statistical significance as well as the upregulated DEG group, downregulated DEG group, and DEG group without statistical significance). The post-treatment data were imported into GraphPad Prism 8 to generate volcano plots.

The sequencing datasets were differentially analyzed using the miRcode database, and the resulting DElncRNAs were predicted to be bound by miRNAs to further explore the underlying pathogenesis of the disease. The processed data were compared with the miRcode database for the DElncRNAs with upregulated and downregulated expression according to the set filtering criteria (http://www.mircode.org). Among the “Highly conserved microRNA families” dataset, miRNAs with potential binding to DElncRNAs were comparatively analyzed.

With the miRNAs derived from the previous step that were potentially bound by DElncRNAs, miRNA target gene prediction was performed. The miRNAs that potentially bound to the DElncRNAs were separately input into the online miRNA target gene prediction websites miRDB, miRTarBase, and TargetScan for target gene prediction, and the resulting predicted target genes were aligned and mapped with the DEGs obtained from sequencing. The differentially expressed lncRNAs—miRNA—mRNAs were further obtained by Cytoscape 3.7.2 with the competing endogenous RNA (ceRNA) interaction network (Shannon et al., 2003).

The common potential target genes predicted from the previous step were entered into the DAVID database with species set as human (Homo sapiens) for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling pathway analyses (Huang et al., 2009a; Huang et al., 2009b). The GO analysis included the cellular component (CC), molecular function (MF) and biological process (BP) categories of the DEGs. Target genes were screened at p < 0.05 to identify the biological processes of potential target genes and major signaling pathways. Pathway diagrams were established using Bioconductor-Pathview in R software (version R x64 3.5.1).

The resulting common potential target genes from 1.2.5 were subjected to survival analysis using human clinical and expression data from TCGA (https://portal.gdc.cancer.gov/). R software was utilized to collate summary clinical data and further derive its expression matrix. Finally, the survival package was utilized to perform survival analysis on the common potential target genes obtained in the previous step.

To determine whether the particles isolated from serum were exosomes, we characterized the vesicles by TEM, size assessment, and analysis of protein quality. TEM images showed that the serum exosomes were morphologically intact, spherical, and uniform in size, with diameters ranging from 30 to 200 nm, which corresponds to the conventional size range of exosomes (Figure 1). As expected, the results of protein quantity assessment showed that the commonly used exosomal markers, such as CD9, CD63, CD81, and TSG101, were abundantly expressed in the isolated pellets, and all the above results confirmed the successful isolation of exosomes from serum samples.

After establishment of screening conditions, we obtained three preoperative serum samples from ovarian cancer patients (aged 39.0 ± 6.0 years) and three serum samples from patients with uterine fibroids (aged 58.0 ± 13.0 years) collected in our hospital, and the baseline characteristics of the samples are shown in Table 1. The sequencing data were background corrected, normalized, and normalized using the R package, and the principal component analysis (PCA) of the corrected data distribution is shown in Figure 2. We used p < 0.05 and a ≥ twofold change in expression (|log₂ FC| ≥ 1.0) as the criteria for the differentially expressed lncRNAs and coding RNAs, and 117 DElncRNAs—36 lncRNAs with upregulated expression and 81 lncRNAs with downregulated expression—and 513 differentially expressed mRNAs—231 mRNAs with upregulated expression and 282 mRNAs with downregulated expression—were selected in the sequencing data according to the p value. The top 50 significantly DElncRNAs and coding RNAs were screened and plotted as a heatmap (see Figure 3). Red represents upregulation of gene expression, and green represents downregulation of gene expression. The p value in the sequencing data after differential analysis was −log10 transformed, and −log10 (p value) was grouped according to log₂ FC (the upregulated lncRNA group, downregulated lncRNA group, and lncRNA group without significant difference as well as the upregulated DEG group, downregulated DEG group, and DEG group without statistical significance). The post-treatment data were imported into GraphPad Prism 8 to generate a volcano plot (see Figure 4).

Analysis of sequencing data using the miRcode database was used to obtain DElncRNAs and the predicted binding miRNAs to further explore the underlying pathogenesis of the disease. The processed data were compared with the miRcode database according to the set filtering criteria for DElncRNAs. The DElncRNAs with up- and downregulated expression were (http://www.mircode.org/) in the “highly conserved microRNA families” dataset and were comparatively analyzed to obtain the miRNAs potentially bound to the DElncRNAs. The results are shown in Table 2.

Using the miRNAs potentially bound to the DElncRNAs from the previous step, we performed miRNA target gene prediction and compared the data with the DEGs derived from the previous differential expression analysis (see Figure 5) to further explore the underlying pathogenesis of the disease. The miRNAs potentially bound to the DElncRNAs were input into the online miRNA target gene prediction websites miRDB, miRTarBase, and TargetScan for target prediction. The results are shown in Table 3, and Cytoscape 3.7.2 software was used to obtain the ceRNA interaction network of differentially expressed lncRNAs—miRNA—mRNA (see Figure 5B).

Utilizing TCGA (https://portal.gdc.cancer.gov/), we downloaded clinical and transcriptomic expression data related to ovarian cancer, 379 related datasets were generated according to the screening conditions set in the previous period, the pooled summary clinical data were collated using R software, and the expression matrix was further derived. Finally, the common potential target genes obtained in the previous period were subjected to survival analysis using the survival package, which revealed that the survival curve of the HOXA10 gene had statistical significance (p < 0.05), but none of the other indexes showed statistical significance (see Figure 6).

GO and KEGG pathway enrichment analyses of potential target genes were performed using the DAVID database, and the GO functions of the core genes were mainly involved in transmembrane receptor protein kinase activity, structural molecule activity with elasticity, transforming growth factor-β activated receptor activity, and GABA receptor binding, as shown in Figure 7. The results of KEGG pathway enrichment analysis showed that the KEGG pathways were mainly involved in signaling pathways regulating stem cell pluripotency, bladder cancer, glycerolipid metabolism, central carbon metabolism in cancer, and resistance to EGFR tyrosine kinase inhibitors (see Figure 7B). A schematic diagram of the signaling pathways related to the HOXA10 gene is shown in Figure 7C.