Identification of common 128 significant differentially expressed genes(DEGs)in PCOS and OC
Firstly, we found 1061 DEGs in the PCOS patients compared to normal women based on GSE34526 dataset of GEO database, and 2254 DEGs in the OC patient samples compared to normal ovary samples based on OC dataset of TCGA database (Figure1 A&B). Moreover, we found that there were 128 common DEGs in PCOS and OC progression (Figure1 C). We also constructed the protein–protein interactions (PPI) network to identify all of 120 genes in the dataset of PCOS and OC, which were visualized by GeneMANIA database (Figure1 D), which indicated that these genes have closely interactions. The PCA analysis found that these DEGs expression could well discriminates between ovarian cancer (blue) and normal (red) (Figure1 E). We extracted GO and KEGG pathway data for these genes based on DAVID database. In the term of GO enrichment, these genes were enriched in cell adhesion molecule binding, actin binding, cadherin binding, actin filament binding, cell-substrate junction, cell-substrate adherens junction, focal adhesion, collagen-containing extracellular matrix, antigen processing and presentation, and so on (Figure1 F). In the term of KEGG enrichment, these genes were enriched in cell adhesion molecules, staphylococcus aureus infection, hematopoietic cell lineage, viral myocarditis, and Asthma (Figure1 G). In summary, these results indicated that common DEGs highlighted the significant role of cell adhesion in the relationship between PCOS and OC.
Evaluation of clinical outcomes in OC based on the common 128 DEGs
The common 128 DEGs were used to analyse the prognosis in OC patients by the univariate Cox method. A number of twelve key genes were closely associated to the prognosis of OC patients, including RNF144B, LPAR3, CRISPLD2, JCHAIN, OR7E14P, IL27RA, PTPRD, STAT1, NR4A1, OGN, GALNT6 and CXCL11 (Figure 2A). Then, we used these expression profile to construct the prognosis model, and the risk score formula was as follow: Risk score = RNF144B*(-0.1441) + LPAR3*(-0.0187) + CRISPLD2*0.0701 + IL27RA*0.2226 + PTPRD*0.0055 + STAT1*(-0.0988) + NR4A1*0.0369 + OGN*0.0590 + GALNT6*(-0.0718) + CXCL11*(-0.0886). Next, we could divide these OC patients into high risk and low risk group with the median risk score based on the risk score formula (Figure 2B).
The survival score and status of the two groups in the training cohort based on TCGA database OC datasets were shown in Figure 3A&B. These twelve key genes expression profiles were shown by the heatmap (Figure 3C). Moreover, we used the GSE140082 dataset as a test cohort to validate the risk score formula, which survival score and status of high risk and low risk group were shown in Figure 3D&E. These key genes expression files in GSE140082 datasets was also visualization by the heatmap (Figure 3F).
In the training cohort, the survival time and rate were significantly decreased with the risk score increased (Figure 4A). The AUC at 1, 2, and 3 years under the ROC curve were 0.571, 0.607, and 0.554, respectively, indicating that a moderate incubation period could be utilized as a prognostic marker of twelve key genes expression profiles in survival monitoring (Figure 4B). However, t-SNE analysis showed that OC patients in different risk groups were not distributed in two group based on TCGA database, which suggested that the 12 signatures could not be an excellent subtype marker (Figure 4C). To validate the efficiency of the prognosis model constructed from the TCGA-OC cohort, we used the median value of training cohort to divide the OC patients into high risk and low risk group based on the GSE140082 cohort. Similar to the results of training cohort, OC patients with high risk had a poor prognosis compared to other OC patients in the low risk group (Figure 4D). The AUC value in 1, 2, and 3 years were 0.617, 0.682, and 0.651 in the test cohort (Figure 4E). The t-SNE analysis was also similar to the training cohort (Figure 4F).
The ectopic expression and prognosis significance of 12 signatures in OC patients
Next, we used the boxplot to visualize the mRNA level of 12 signatures in OC samples, which indicated that LPAR3, JCHAIN, IL27RA, GALNT6, CXCL11, RNF144B, STAT1, OR7E14P were significantly increased in OC patients, but CRISPLD2, PTPRD, OGN, NR4A1 were obviously decreased in patients with OC (Figure 5A). We also confirmed the overall survival rate of 12 signatures in OC patients based on TCGA database, suggesting that OGN was significantly and negatively correlated with OC patient’s prognosis, but JCHAIN, GALNT6, CXCL11, STAT1 were significantly and positively correlated with prognosis in OC patients (Figure 5B). These results suggested that JCHAIN, GALNT6, CXCL11, STAT1, and OGN might play a key role in the progression of OC patients.
The DNA alteration and immune infiltration of 5 key genes in OC progression
We found the 5 genes were genetically altered, such as missense mutation, amplification and deep deletion (Figure 6A). The CNV of JCHAIN was significantly correlated with CD8+ T cell, Neutrophil, and Dendritic cell. CXCL11 CNV was closely associated with CD8+ T cell, CD4+ T cell, Neutrophil, and Dendritic cell. The CNV level of OGN was markedly related to Macrophage. STAT1 CNV level had a closely relationship with CD8+ T cell and Dendritic cell. The CNV of GALNT6 was significantly associated with B cell, CD8+ T cell and CD4+ T cell (Figure 6B). Furthermore, we found the mRNA expression of GALNT6 was not obviously correlated with immune infiltration in any immune cell types. JCHAIN level was closely associated with purity, CD8+ T cell, CD4+ T cell, Neutrophil, and Dendritic cell. CXCL11 expression was correlated with the infiltration of purity, B cell, CD8+ T cell, CD4+ T cell, Neutrophil, and Dendritic cell. OGN level was significantly correlated with purity. STAT1 mRNA level had a closely relationship with purity, CD8+ T cell, Neutrophil, and Dendritic cell (Figure 6C). Taken together, the expression and alteration of these 5 key genes was involved in the immune infiltration progression of OC.
The drug sensitivity of hub gene
We further used the drug sensitivity analysis to confirm these 5 key genes. The result showed that OGN was closely correlated with chemotherapy resistance based on GSCALite database (Supplementary Figures S1). Therefore, targeting OGN could be a potential target in the treatment of patients with OC or PCOS.
The characteristics of OGN in OC and PCOS
For elucidate the expression, function and structure of OGN, we used the PDB databased to confirm the OGN structure, as shown in Figure 7A. The OGN has an LRR_8 domain and multiple phosphorylation, acetylation and N-linked glycosylation site. The protein expression was significantly decreased in OC tissues samples compare to normal ovary samples (Figure 7B&C). We further utilized GSVA and GSEA analysis to predict the potential function of OGN, as shown in Figure 7D&E. OGN might be involved in the progression of steroid hormone biosynthesis and steroid hormone response. Furthermore, we found the level of OGN was significantly and positively correlated with the level of FSHR in OC (Figure 7F). We overexpressed OGN in KGN and SKOV3 cell lines (Figure 7G), and confirmed the effect of OGN on FSHR expression by IF (Figure 7H). The results indicated that OGN played a key role in PCOS and OC progression by upregulating FSHR level.
OGN level is correlated with regulators of ferroptosis and m6A methylation in OC
The ferroptosis and m6A methylation was involved in the development and progression of OC. We firstly used the TCGA database to analyze whether OGN level is correlated with ferroptosis. We make a correlation analysis between the OGN expression and 25 ferroptosis genes in OC and ovary tissue samples based on TCGA and GTEx database (Figure 8A). The results showed that the expression of 25 ferroptosis genes were significantly between OC and normal ovary. Furthermore, we also confirmed the correlation of ferroptosis genes with OGN in OC samples. We found that the level of OGN was positively correlated with MT1G, HSPB1, GPX4, FDFT1 and ATP5MC3, but negatively correlated with CDKN1A, HSPA5, SLC1A5, NCOA4, LPCAT3, DPP4, ALOX15, ACSL4, and ATL1 (Figure 8B). Then, we extracted the expression profiles for the 20 m6A methylation genes between OC and normal ovary samples based on TCGA and GTEx database (Figure 8C), which indicated that these m6A methylation regulators were played a key role in OC progression. We further made a correlation analysis between these m6A regulators and OGN expression. The result showed that METTL14, WTAP, VIRMA, RBM15, RBM15B, ZC3H13, YTHDC1, YTHDC2, YTHDF3, YTHDF1, IGF2BP2, HNRNPA2B1, FTO, and ALKBH5 were positively and significantly associated with OGN expression in OC patients (Figure 8D). Taken together, OGN might be had another important function on OC ferroptosis and m6A methylation modifications.