Ovarian cancer is the second most common cause of gynecologic cancer death in women around the world but accounts for the highest mortality rate among these cancers . According to the Global Cancer Observatory (GCO, https://gco.iarc.fr/), there are a total of 313,959 patients with ovarian cancer patients and 207,252 cases died from it.
In recent years, the potential of snoRNA as biomarkers has been graduated recognized, for example, SNORD89 was identified as a prognostic biomarker and prospective therapeutic in ovarian cancer patients and breast cancer patients [17, 18]. However, there is a lack of systematic and comprehensive research on snoRNA in ovarian cancer. In our study, we comprehensively analyzed the snoRNA in patients with ovarian cancer, and screened out 14 prognostic snoRNAs by Univariate Cox survival analysis (Table 4). Then, prognostic model was constructed by Multivariate Cox survival analysis, and 9 snoRNAs were included in the prognostic model (Table 5). Each patients with ovarian cancer has a unique RiskScore, and patients with high RiskScore had higher deaths and lower overall survival time (Figure 1A, 1B and 1E).
A good prognostic marker is often associated with multiple clinicopathological parameters. Hence, we analyzed the correlation of the RiskScore derived from the prognostic model with age, tumor size, lymphatic invasion, stage and tumor status of ovarian cancer patients. Results showed that RiskScore was significantly increased in patients with high-age group, large tumor, high-grade and with tumor status (Figure 2A, 2B, 2D-2E). These results suggested that the RiskScore was higher in the group of ovarian cancer patients with high risk factors for the prognosis.
In our research, we analyzed the correlation of snoRNAs in the prognostic model with their hostgenes. Among of them, 7 of 9 snoRNAs had positive correlation with their hostgenes in ovarian cancer tissues (Figure 3). The expression abundance of SNORA70J is very low, alike its host gene SNORA70J. CNV has been reported occurred in various cancers, and some snoRNAs were associated with their CNVs . In our research, 5 of 9 snoRNAs in our prognostic model had correlation with their CNVs (Figure 4A).
And, the specificity and sensitivity of the RiskScore were verified by ROC curve, and we found that the area of 7 years achieved 0.785. These results showed that the model has the best effect in predicting the prognosis of 7 years in ovarian cancer patients. Univariate and multivariate Cox survival analysis showed the RiskScore was an independent prognostic factor in ovarian cancer patients (Figure 4).
Stratified analysis of survival according to different clinical parameters were conducted. We found that RiskScore predict prognosis well in diverse ages and tumor size. However, RiskScore, in tumor free and no lymphatic invasion patients, could not predict patients’ prognosis well (Figure 5). We speculated that these results may be caused by the small number of experimental cases.
Moreover, patients with ovarian cancer were randomly divided into two groups, and validate the RiskScore in each subgroup. All of the results showed that patients with high RiskScore had poorer prognosis versus to patients with low RiskScore (Figure 6). Further, we detected the expression of snoRNAs in 7 paired tissues, all of them, except SNORD3C and SNORD89, down regulated in ovarian cancer tissues compared to ovarian normal tissues (Figure 6I). And, this result in accord with the previous research . The RiskScore of sample 1 and sample 2 are 46.47 and 2.469, and this result indicate sample 1 had poorer prognosis versus sample 2. Moreover, the results of H&E staining and immunohistochemistry of Ki67, P53 and P16 confirmed that patients with high RiskScore are more malignant. The positive rate of Ki67 in sample 1 was 64.9%, and higher than that 20.5% in sample 2 (Figure 6K). And, P16 block expression was found in sample 1 in contrast to mottled expression in sample2 (Figure 6J).