Identification of DEGs

The GSE12470 dataset contained 43 OC tissue samples (8 early-stage tissues and 35 late-stage tissues) and 10 normal samples, while the GSE16709 dataset contained 17 OC tissue samples and 9 normal samples. The gene datasets were standardized and assessed for cross-comparability, suggesting that the selected samples were centered and the numerical distribution met the standard, indicating high microarray data quality and cross-comparability (Fig. 1A, B). Using GEO2R for analysis, we screened 2908 DEGs from the GSE12470 dataset and 3038 DEGs from the GSE16709 dataset, and clustering analysis was performed to obtain volcano plot of DEGs (Fig. 1C-D). The top 15 up-regulated and down-regulated genes were selected (Tables S2-3). The DEGs heatmaps revealed that the samples were clustered well with high confidence (Fig. 1E-F).

Fig. 1

A-B Cross-comparability evaluation of data set samples; C-D Volcano diagram: The x-coordinate is log2FoldChange and the y-coordinate is -log10 (p-value). Red dots represent up-regulated genes and blue dots represent down-regulated genes; E Heat map of GSE12470; F Heat map of GSE16709. Horizontal represents genes, and each column is a sample: red for high-expression genes, blue for low-expression genes

Venn diagrams were drawn to screen the common DEGs between the two datasets (Figure S1), demonstrating that there were 869 DEGs in common between the two datasets.

Function enrichment analysis of common DEGs

The results of the GO enrichment analysis of DEGs were divided into 3 parts: molecular function (MF), biological process (BP), and cellular component (CC). The most significantly enriched top 6 GO terms with the minimum p-values of each category were displayed in the histogram and the bubble diagram of GO enrichment analysis (Fig. 2A, Figure S2A). The DEGs were significantly enriched in “extracellular matrix organization”, "cell adhesion”, etc. of BP term; “extracellular matrix organization”, “cell adhesion”, etc. of CC term; and “extracellular matrix organization”, “cell adhesion”, etc. of MF term.

Fig. 2

Analysis of common differentially expressed genes (DEGs) A Gene Ontology (GO) enrichment analysis the histogram of DEGs, abscissa is GO term, and ordinate is -log10 (p-value) of enrichment in each term. B Bubble map for Kyoto Encyclopedia of Genes and Genomes enrichment analysis

According to the KEGG enrichment analysis results of DEGs, the top 10 pathways with the minimum p-value, which means the most significant enriched pathway, were selected for display. The results were shown in the histogram and the bubble diagram of KEGG enrichment analysis (Fig. 2B, Figure S2B). The results revealed that the DEGs were mainly enriched in “Pathways in cancer”, “Focal adhesion”, “ECM-receptor interaction”, “PI3K-Akt signaling pathway”, and “Human papillomavirus infection”, etc.

PPI network construction and hub gene identification

The STRING tool was employed to construct PPI networks based on the DEGs, which were visualized using Cytoscape software (Figure S3). Key modules were identified from the PPI network of DEGs, and Module 1 with the maximum score was selected (Figure S4). Then, genes in Module 1 were sorted according to the score, and the top 9 genes were selected as hub genes for analysis, including KIFC1, PCLAF, CDCA5, KNTC1, MCM3, OIP5, CENPM, KIF15, and ASF1B.

Analysis of hub genes

GO enrichment chord plot reflected the relationship between proteins and pathways, as well as the changes in the pathway functions (Fig. 3A). After processing with R language, we obtained two principal component variables, PC1 and PC2, which provided a total variance explanation rate of 90.6% and could distinguish between the control group samples and OC samples. The scatter plot suggested that the two groups were separated well, and further confirmed the effectiveness of PC1 and PC2 (Fig. 3B). The ridge line plot showed the expression of hub genes (Fig. 3C).

Fig. 3

A GO enrichment chord map, the data included 3 parts: genes; logFC is the multiple of gene change, used for sequencing and gene block color; The remaining columns are GO term, and different links of the gene indicate whether the gene is in this GO term; B Principal component analysis diagram of hub genes, in which axes PC1 and PC2 are the first and second principal components (i.e., explanation rate of difference by potential variables); Dots represent samples, and different colors represent different groups; C Hub gene ridge map, horizontal coordinate is gene expression level, mountain height represents sample abundance of corresponding expression level

The ROC curve of hub gene was drawn using the original sample data of GSE12470 and GSE16709 datasets (Figures S5, S6). The results demonstrated that the false positive rate of hub genes (KIFC1, PCLAF, CDCA5, KNTC1, MCM3, OIP5, CENPM, KIF15, and ASF1B) testing in GSE12470 was 0.5%, 0%, 0%, 0%, 0%, 0%, 4. 1%, 0%, and 0%, respectively, while the true positive rate was 99.5%, 100%, 100%, 100%, 100%, 100%, 95.9%, 100%, and 100%, respectively. The false positive rate of hub genes in GSE16709 dataset was 3.9%, 2.6%, 15%, 13.7%, 5.6%, 2.6%, 0%, 11.1%, and 13.1%, respectively, while the true positive rate was 96.1%, 97.4%, 85%, 86.3%, 95.4%, 97.4%, 100%, 88.9%, and 86.9%, respectively, indicating that hub genes were good indicators to distinguish OC from healthy controls. The differential expression of hub gene in tumor and non-tumor tissues analyzed by The Cancer Genome Atlas database revealed that the expression level of CENPM in tumor tissues was significantly higher than that in normal tissues (Figure S7).

RT-qPCR results suggested that compared to the normal ovarian cells IOSE80, the mRNA expression levels of KIFC1, PCLAF, CDCA5, KNTC1, MCM3, OIP5, CENPM, KIF15, and ASF1B were significantly up-regulated in SKOV3 and A2780 (Fig. 4).

Fig. 4

Real-time quantitative PCR (RT-qRCR) detected the mRNA expression levels of KIFC1, PCLAF, CDCA5, KNTC1, MCM3, OIP5, CENPM, KIF15, and ASF1B in cells. *P < 0.05, **P < 0.01, ***P < 0.001 vs. IOSE80 cells

Knockdown of CENPM suppresses proliferation, migration, and invasion of OC cells

RT-qPCR results revealed that in SKOV3 and A2780 cells, the CENPM expression level in the si-NC group suggested no significant difference, compared with the control group, while the CENPM expression levels in the si-CENPM-1, si-CENPM-2, and si-CENPM-3 groups were significantly decreased (Fig. 5A). Among them, si-CENPM-1 had the highest transfection efficiency. Western blot was performed and showed the same results as RT-qPCR detection, further confirming the RT-qPCR results (Fig. 5B). Therefore, the subsequent experiments were conducted using si-CENPM-1.

Fig. 5

Effect of CENPM on ovarian cancer cell. A RT-qPCR was used to detect transfection efficiency; B Western blot was used to detect the protein expression level of CENPM; C Cell counting kit was used to detect viability of SKOV3; D Cell counting kit was used to detect viability of A2780. E Transwell detected cell migration; F Transwell assessed the cell invasion. *P < 0.05, **P < 0.01, ***P < 0.001 vs. Control

CCK-8 results suggested that in SKOV3 and A2780 cells, compared to the control group, there were no significant changes in si-NC group, while cell proliferation of SKOV3 and A2780 cells in the si-CENPM group was significantly suppressed (P < 0.001), indicating that knocking down CENPM significantly inhibited the proliferation ability of OC cells (Fig. 5C-D). Additionally, in SKOV3 and A2780 cells, the cell migration, and invasion abilities in the si-CENPM group were notably decreased compared with the control group (Fig. 5E-F).

Knockdown of CENPM promotes pyroptosis in OC cells

Compared to the control group, the concentrations of TNF-α, IL-1β, and IL-6 in the si-CENPM group were significantly increased (Fig. 6A-B). Western blot assay demonstrated that in SKOV3 and A2780 cells, compared with the control group, the expression levels of pyroptosis-related proteins in the si-CENPM group were significantly increased (Fig. 6C).

Fig. 6

Effect of CENPM knockdown on pyroptosis. A The level of inflammatory factors in SKOV3 cells; B The level of inflammatory factors in A2780 cells; C Western blot was used to detect the expression of pyroptosis-related proteins in SKOV3 and A2780 cells. *P < 0.05, **P < 0.01, ***P < 0.001 vs. Control

Knockdown of CENPM inhibits tumor growth, metastasis, and pyroptosis in mice

The mice in the LV-CENPM group presented significantly lower tumor volumes and weights than those in LV-NC group (Fig. 7A-C). Immunohistochemical staining showed that Ki67 expression level was decreased in LV-CENPM group (Fig. 7D). HE staining suggested that knockdown of CENPM inhibited malignancy of OC (Fig. 7E).

Fig. 7

Effect of CENPM knockdown on tumor. A Subcutaneous tumor in each group; B Growth curve of subcutaneous tumor in mice; C Weight of subcutaneous tumor in mice; D Immunohistochemical analysis detected Ki67 levels; E Hematoxylin & eosin staining of subcutaneous tumor (Amplification: 200 × , Scale: 100 μm); F Western blot detected epithelial-mesenchymal transition-related proteins; G Western blot to detect the expression of CENPM and pyroptosis-related proteins in tumor tissues. **P < 0.01 ***P < 0.001 vs. LV-NC group

Epithelial-mesenchymal transition-related proteins were detected to evaluate tumor metastasis by western blot assay (Fig. 7F). The results showed that knockdown of CENPM increased the protein expression level of E-cadherin, while decreasing the protein expression level of N-cadherin and vimentin. Western blot results demonstrated that the expression level of CENPM protein in tumor tissues of mice was significantly reduced after CENPM knockdown and the expression levels of pyroptosis-related proteins were significantly increased (Fig. 7G). These results further confirmed the inhibitory effect of CENPM knockdown on tumor progression and pyroptosis.

Knockdown of CENPM suppresses the development of OC via activating cGAS-STING pathway

Western blot results revealed that the expression levels of pathway proteins, cGAS, STING, and p-STING in cGAS-STING pathway were significantly higher in si-CENPM group than those in si-NC group (Fig. 8A). Moreover, the treatment of RU.52 reversed the inhibitory effect of CENPM knockdown on proliferation, migration, and invasion of OC cells (Fig. 8B-D).

Fig. 8

Effect of CENPM knockdown on cGAS-STING pathway. A Western blot was used to detect the protein expression levels of pathway proteins; B Cell counting kit was used to detect cell viability. C Transwell detected cell migration; D Transwell assessed the cell invasion. **P < 0.01, ***P < 0.001 vs. Control; ##P < 0.01, ###P < 0.001 vs. si-CENPM