Patient characteristics
As shown in Table 1, 376 cases of primary OSC with clinical and gene expression data were downloaded from TCGA data. Of these 178 patients (47.3%) were over 60 years old. FIGO stage I disease was found in 1 patient (0.3%), stage II in 22 (5.9%), stage III in 293 (78.6%), and stage IV in 57 (15.3%). The histologic grades, G1, G2, G3 and G4 were 0.3%, 11.5%, 88.0%, and 0.3%, respectively. The cancer status included 71 without (21.3%) and 262 with (78.7%) tumors. Patients were divided in to four groups according to primary therapy outcome: progressive disease (PD), 27(8.9%); partial response (PR), 43 (14.1%); stable disease (SD), 22 (7.2%); and complete response (CR), 213 (69.8%). Residual tumors were found in 267 of 376 total cases (23.5%), lymphatic invasion in 100 of 148 (67.6%), and venous invasion in 63 of 103(61.2%). Lesions occurred in bilateral ovaries in 253 cases. Mutation of TP53 and BRCA genes were found in 248 and 25 cases, respectively. Of the patients in this cohort 61.2% finally succumbed to OSC (Table 1).
We compared hazard ratios (HR) with 95% confidence interval (CI) ranges among patient information parameters. Univariate analysis identified primary therapy outcome (p < 0.001), age (p = 0.032), residual tumor (p < 0.001), ATP1A3 (p < 0.001) and ATP1A4 (p = 0.03) as prognostic factors for OS (Table 2). Multivariate analysis showed that primary therapy outcome (p < 0.001), residual tumor (p = 0.004), ATP1A2 (p = 0.006), ATP1A3 (p < 0.001) and ATP1A4 (p < 0.001) were independent risk factors for OS (Table 2).
Univariate analysis identified primary therapy outcome (p < 0.001), residual tumor (p < 0.001), ATP1A3 (p = 0.002) and ATP1A4 (p = 0.023) as prognostic factors for DSS (Table 3). Multivariate analysis showed that primary therapy outcome (p < 0.001), residual tumor (p = 0.001), ATP1A3 (p = 0.006) and ATP1A4 (p = 0.017) were independent risk factors for DSS (Table 3).
3.2 Correlation between expression of ATP1A genes and clinical characteristics
Expression of ATP1A gene family members in OSC tissues and normal tissues contained were significant differences. ATP1A1 and ATP1A3 were highly expressed in tumor tissues, while ATP1A2 and ATP1A4 were highly expressed in normal tissues (all p < 0.001) (Figure 1A-D). To evaluate the expression of ATP1A gene members in human pan-cancers, the RNA-seq data from TCGA and GTEx databases were further analyzed. Differential expression between the tumor and adjacent normal tissues for ATP1A1-4 across all TCGA tumors are shown in Supplementary Figure 1. ATP1A1 expression was significantly higher in Cholangiocarcinoma (CHOL), skin cutaneous melanoma (SKCM) and thymoma (THYM) (all p-value < 0.001) compared with adjacent normal tissues. ATP1A2 expression was significantly lower in breast invasive carcinoma (BRCA), bladder urothelial carcinoma (BLCA) and colon adenocarcinoma (COAD) (all p < 0.001) compared with adjacent normal tissues. ATP1A3 expression was significantly higher in adrenocortical carcinoma (ACC) (p < 0.001), thymoma (THYM) (p < 0.001) and pheochromocytoma and paraganglioma (PCPG) (p < 0.01). ATP1A4 expression was significantly higher in breast invasive carcinoma (BRCA), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC) and lung squamous cell carcinoma (LUSC) (all p < 0.001) (Sup. Fig. 1). In addition, the mRNA expression of ATP1As gene members in pan-cancer cell lines were also analyzed based on the CCLE database (Sup. Fig. 2). As is observed, ATP1A1 was highly expressed in Ewing’s-sarcoma, melanoma, and colorectal carcinoma, while ATP1A3 was highly expressed in neuroblastoma, Burkitt lymphoma, and T-lymphocytic leukemia. By contrast, ATP1A2 and ATP1A4 were expressed in lower amounts in most tumor cell lines (Sup. Fig.2).
Next, we analyzed the sensitivity and specificity of ATP1As gene family members in predicting the diagnosis of OSC by ROC curve. The area under the curve (AUC) of ATP1A1, ATP1A2, ATP1A3, and ATP1A4 were 0.829, 0.963, 0.892 and 0.778, respectively (Figure 1E-H). These results suggested that ATP1A gene family may be potential diagnostic biomarkers for OSC patients.
Moreover, we analyzed the correlation between the expression of ATP1A family genes and clinical features by Kruskal-Wallis test and Wilcoxon signed-rank test. Significant correlation arose between ATP1A3 (p = 0.019) and FIGO stage, but no significant correlations were found between ATP1A1 (p = 0.395), ATP1A2 (p = 0.492), ATP1A4 (p = 0.07), and FIGO stage (Figure 2A-D). In addition, ATP1A3 (p < 0.001) was significantly associated with histological grade, while ATP1A1 (p = 0.913), ATP1A2 (p = 0.716), and ATP1A4 (p = 0.727) were not (Figures 2E-H). Although no correlation was found between ATP1As and lymphatic invasion, the p-value of ATP1A3 was close to 0.05, suggesting that the lack of significance may be due to a limitation of the number of patients (Figures 2I-L).
Based on these mRNA expression patterns of ATP1A family genes in OSC, we next explored the protein expression patterns of ATP1As in OSC by the Human Protein Atlas. As shown in Figure 3, ATP1A1 is highly expressed in OSC and low in normal ovarian tissues (Figure 3A). In addition, ATP1A2 and ATP1A3 protein expression was low in both normal ovarian tissues and OSC tissues (Figure 3B, C). A lack of protein expression patterns of ATP1A4 in the database at present precluded our analysis of this protein.
Prognostic value of mRNA expression of ATP1As gene family in OSC patients
We used a Kaplan-Meier plotter (http://kmplot. com/analysis/) to analyze the prognostic values of the mRNA expression of ATP1A genes in OSC patients. First, we analyzed the relationship between mRNA expressions of distinct ATP1A family members and prognoses of OSC patients. As shown in Figure 4, for OS and DSS, higher mRNA expression of ATP1A3 (HR = 1.60, CI: 1.23-2.08, p < 0.001; and HR = 1.57, CI: 1.19-2.08, p = 0.002, respectively) (Figure 4C, D) and ATP1A4 (HR = 0.75, CI: 0.57-0.97, p =0.030; and HR = 0.72, CI: 0.54-0.96, p = 0.023, respectively) (Figure 4G, H) were significantly associated with shorter OS and DSS of OSC patients. However, neither ATP1A1 (HR = 1.15, CI: 0.88-1.49, p =0.298; and HR = 1.13, CI: 0.86-1.50, p =0.377, respectively) (Figure 4A, B) nor ATP1A2 (HR = 1.26, CI: 0.97-1.63, p =0.086; and HR = 1.30, CI: 0.98-1.73, p =0.064, respectively) (Figure 4E, F) mRNA expression showed any correlation with OS or DSS of OSC patients. These results suggested that mRNA expressions of ATP1A3 and ATP1A4 were closely related to the prognosis of OSC and may be used as potential biomarkers to predict the survival of patients.
Secondly, we analyzed the effects of distinct ATP1A family member expression on patients’ prognosis in the subgroups of FIGO stage III, histological grade-G3, TP53 mutation, and age ≥ 60 years. The results showed that in patients with FIGO stage III, no significant correlation was found in ATP1A1 (HR = 1.16, CI: 0.87-1.56, p = 0.315), ATP1A2 (HR = 1.24, CI: 0.92-1.66, p = 0.157) and ATP1A4 (HR = 0.77, CI: 0.58-1.04, p = 0.085), while the high expression of ATP1A3 (HR = 1.70, CI: 1.27-2.28, p < 0.001) was significantly associated with shorter OS (Figure 5A-D). Higher mRNA expression of ATP1A2 (HR = 1.34, CI: 1.01-1.78, p = 0.042) and ATP1A3 (HR = 0.75, CI: 1.30-2.30, p < 0.001) were significantly associated with poor OS in patients with Histological grade-G3, while ATP1A1 (HR = 1.20, CI: 0.91-1.59, p = 0.202) and ATP1A4 (HR = 0.79, CI: 0.60-1.06, p = 0.113) were not (Figure 5E-H). In addition, the high expression of ATP1A3 (HR = 1.62, CI: 1.17-2.26, p = 0.004) was significantly associated with decreased OS in patients with TP53 mutation, but there was no correlation with ATP1A1 (HR = 1.33, CI: 0.95-1.84, p = 0.094), ATP1A2 (HR = 0.98, CI: 0.70-1.36, p = 0.881) and ATP1A4 (HR = 0.68, CI: 0.49-0.95, p = 0.023) (Figure 5I-L). Furthermore, we found that the high expression of ATP1A2 (HR = 1.48, CI: 1.02-2.15, p = 0.039) and ATP1A3 (HR = 2.23, CI: 1.54-3.24, p < 0.001) were significantly associated with poor OS in patients of ≥ 60 years old, while there was no correlation with ATP1A1 (HR = 1.15, CI: 0.80-1.65, p = 0.460). By contrast, we found that the high expression of ATP1A4 (HR = 0.66, CI: 0.46-0.95, p = 0.027) was correlated with better OS (Figure 5M-P). Finally, the effects of ATP1As members on DSS in patients with subgroups are shown in Sup. Fig. 3.
Based on the results of the multivariate analysis, a genomic-clinical nomogram including the primary tumor therapy outcome, residual tumor, ATP1A2, ATP1A3, and ATP1A4, was established to predict the 1-, 3-, and 5-year OS and DSS for OC patients (Figure 6A, C). The calibration curves of the nomogram for predicting these survival times indicated that it performed well (Figure 6B, D).
Mutation frequency of ATP1As gene family
Based on the cBioPortal database, we explored the mutation frequency of the ATP1As gene family. The mutation rates of ATP1A1, ATP1A2, ATP1A3 and ATP1A4 were 4%, 5%, 3% and 4%, respectively (Figure 7).