Expression levels of UBE2 family members in patients with ovariancancer
We selected 12 gene members of the UBE2 family based on their possible association with ovarian cancer, and used the Oncomine database to analyze their mRNA expression levels in different cancer and normal tissue samples (Fig. 1a and Additional file 2). The database contained a total of 433, 467, 450, 299, 439, 452, 409, 444, 314, 392, 299, and 443 analyses for UBE2A, UBE2B, UBE2C, UBE2F, UBE2G, UBE2I, UBE2M, UBE2N, UBE2R2, UBE2S, UBE2T, and UBE2V2, respectively. Oncomine analysis revealed that the mRNA expression of UBE2C, UBE2M, UBE2N, UBE2S, UBE2T, and UBE2V2 was upregulated in patients with ovarian cancer. UBE2C mRNA levels were significantly higher in 8 ovarian cancer datasets: in Yoshihara’s dataset[23], UBE2C was significantly upregulated in 43 cases of ovarian serous adenocarcinoma compared with that in 10 normal tissue samples (P=5.73E-13, fold change=12.955); in Lu’s dataset[24], UBE2C was overexpressed in ovarian serous adenocarcinoma, endometrioid adenocarcinoma, and clear cell adenocarcinoma (P=1.39E-7, fold change=2.358; P=1.43E-4, fold change=2.298; P=0.001; fold change=2.156); in the TCGA, Adib’ s, and Welsh’ s dataset[25, 26], UBE2C was highly expressed in ovarian serous carcinoma compared with that in normal tissues. (P=2.24E-7, fold change=10.184; P=0.004, fold change=3.828; P=8.50E-8; fold change=4.020, respectively). Similar results were reported by Bonome et al. in ovarian carcinoma (P=2.50E-10, fold change=4.182)[27]. Moreover, in Bonome’s dataset[27], UBE2M (P=1.63E-7, fold change=2.479), UBE2N (P=6.86E-10, fold change=2.111), and UBE2S (P=5.22E-11, fold change=3.855) were upregulated in ovarian cancer. In the TCGA dataset, UBE2S (P=1.53E-5, fold change=3.267) and UBE2V2 (P=7.04E-8, fold change=2.031) were upregulated in ovarian serous cystadenocarcinoma. Significant UBE2T overexpression compared with that in normal tissues was found in Yoshihara’s dataset (ovarian serous adenocarcinoma; P=4.30E-8, fold change=8.877)[23], and in Lu’s dataset (ovarian serous adenocarcinoma and endometrioid adenocarcinoma; P=3.64E-8, fold change=2.751 and P=2.19E-6, fold change=2.814, respectively)[24] In addition, the mRNA levels of UBE2B, UBE2G, and UBE2I were reduced in patients with ovarian cancer. In Bonome’s dataset[27], UBE2B (P=2.98E-9, fold change=−2.405), UBE2G (P=6.86E-10, fold change=−2.111), and UBE2I (P=8.72E-8, fold change=−3.355) were downregulated in ovarian carcinoma compared with those in ovarian surface epithelium. A similar trend was found for UBE2I in Welsh’s[26] and TCGA datasets: the mRNA levels of UBE2I in ovarian serous surface papillary carcinoma (P=8.10E-5, fold change=−3.672) and ovarian serous cystadenocarcinoma (P=2.93E-7, fold change=−2.884) were significantly lower than those in normal tissues. Moreover, Oncomine analysis showed that the mRNA expression of UBE2A, UBE2F, and UBE2R2 was not significantly different between ovarian cancer and normal tissues.
Next, we used the GEPIA database to compare the mRNA levels of different UBE2s in ovarian cancer and normal ovarian tissues (Fig. 1b). The results indicated that the expression levels of UBE2C, UBE2F, UBE2N, UBE2S, and UBE2T were significantly higher in ovarian cancer tissues than in normal ovarian tissues, whereas those of UBE2A, UBE2B, UBE2G, UBE2I, UBE2M, UBE2R2, and UBE2V2 were not significantly different between the two groups. Combined analysis of the GEPIA and Oncomine databases revealed that the expression of UBE2C, UBE2N, UBE2S, and UBE2T was upregulated in ovarian cancer compared with that in normal tissues.
Genetic alterations of UBE2 family members in patients with ovarian cancer
Genetic alterations of UBE2 members were analyzed by using the cBioPortal database. A total of 1680 cases from three datasets (489 cases from TCGA, Nature 2011; 585 cases from TCGA, PanCancer Atlas; and 606 cases from TCGA, Provisional) of ovarian serous cystadenocarcinoma were analyzed (Fig. 2a,b). The alteration rates were 28.47%, 19.86%, and 13.5%, in the 3 ovarian cancer datasets, respectively. Twelve UBE2 genes showed various levels of genetic alteration. In all three datasets, only gene amplification was observed for UBE2C and UBE2V2; UBE2C exhibited the highest rate of amplification (6.35% in TCGA), while the rate of UBE2V2 amplification was 4.46% in TCGA. Most genetic variations in UBE2 family genes were amplifications and deep deletions, whereas mutations were found in UBE2A (0.17% in TCGA), UBE2M (0.17% in TCGA and TCGA pub), and UBE2T (0.17% in TCGA PanCan), and multiple alterations were found in UBE2A and UBE2I (both 0.17% in TCGA PanCan). In addition to gene amplification and deep deletions, UBE2G variations included gene fusions (1.37% in TCGA). The Kaplan-Meier curve indicated that there was no significant difference in OS and disease-free survival (DFS) regardless of the presence of alterations in one of the query genes (P values, 0.804 and 0.393, respectively, Fig. 2c,d)
Prognostic value of UBE2 genes in patients with ovarian cancer
Kaplan-Meier plotter analysis was applied to assess the relationship between the mRNA expression of individual UBE2 members and progression-free survival (PFS) in 1436 clinical ovarian cancer patients (Fig. 3). We found that the mRNA levels of UBE2A, UBE2B, UBE2C, UBE2G, UBE2N, UBE2R2, and UBE2T were associated with ovarian cancer prognosis, while those of the remaining members were not. Increased mRNA levels of UBE2A, UBE2B, UBE2C, UBE2N, UBE2R2, and UBE2T, and decreased mRNA levels of UBE2G were significantly associated with poor prognosis.
Ovarian serous tumors represents the most common histological subtype among ovarian tumors. The prognostic value of UBE2 genes with expression levels significantly correlated with ovarian cancer prognosis was further investigated in ovarian serous tumors. Increased mRNA levels of UBE2R2 and UBE2T were significantly correlated with poor OS and PFS in patients with ovarian serous tumors. Interestingly, increased expression of UBE2A and UBE2G was associated with shorter OS but longer PFS (Fig. 4).
By integrating the results of Oncomine, GEPIA, and KM plotter, we observed that UBE2T was significantly upregulated in ovarian cancer compared to normal tissues, and that high UBE2T mRNA expression was significantly correlated with shorter OS and PFS in patients with ovarian serous tumors. Moreover, we assessed the prognostic value of UBE2T in relation to the pathological grade, Federation of Gynecology and Obstetrics (FIGO) stage, and TP53 status of ovarian serous tumors. Increased expression of UBE2T was correlated with poor OS in patients with tumors of all FIGO stages (P<0.05, Fig. 5c,d). Moreover, UBE2T upregulation predicted poor OS in patients with both mutated and wild-type TP53 (P<0.05, Fig. 5e,f). Although among patients with well/moderate and poor differentiation, those with high UBE2T expression tended to have shorter OS compared to patients with low UBE2T expression, the difference was not statistically significant (P>0.05, Fig. 5a,b).
Analysis of the interactions between UBE2 family members
Spearman’s correlation coefficients were calculated by using the GEPIA platform to investigate the relationships between the expression levels of different UBE2 members in ovarian cancer (Fig. 5g). The coefficients ranged from 0.013 (UBE2B vs. UBE2M) to 0.67 (UBE2C vs. UBE2S). The results indicated a moderate positive correlation between the expression levels of UBE2A, UBE2N, UBE2S, UBE2T, and UBE2V2 (0.3<r≤0.7), and a low positive correlation between the expression levels of UBE2C, UBE2F, UBE2M, and UBE2R2 (0<r≤0.3). However, UBE2B showed a weakly positive correlation with other UBE2 genes, in addition to a weakly negative correlation with UBE2C (r=-0.017).
A protein-protein interaction (PPI) network consisting of 12 UBE2 family members and 20 UBE2-interacting proteins was visualized by using Cytoscape (PPI enrichment P value <1.0E−16). Among 32 nodes, the most significant 10 proteins were UBA2, UBC, NEDD8, RAD18, SUMO1/2, RANBP2, RABGAP1, ANAPC11, and CDC20. Biological process analysis showed that UBE2 proteins were mainly involved in anaphase-promoting complex-dependent catabolic processes, protein neddylation, protein sumoylation, postreplication repair, and positive regulation of ubiquitin protein ligase activity (Fig. 5h). Pathway analysis showed that the UBE2 enzymes were mainly implicated in ubiquitin-mediated proteolysis, cell cycle, oocyte meiosis, and progesterone-mediated oocyte maturation (Fig. 5i, Additional file 3).
Molecular mechanisms related to UBE2T expression in ovarian cancer
A total of 536 genes were obtained from the cBioPortal database, based on Spearman’s correlation coefficients higher than 0.35 between their expression and that of UBE2T. Information about the co-expressed genes is shown in Additional file 4. The Metascape portal was used, and P<0.01 was set as the threshold value to screen the top GO annotations and KEGG pathway results regarding UBE2T and its interactors. The top 20 GO enrichment items were classified into three functional groups: biological process (BP, 13 items), molecular function (MF, 1 items), and cellular component (CC, 6 items) (Fig. 6a). Regarding the BP, the genes co-expressed with UBE2T were mainly involved in cell division, cell cycle phase transition, DNA replication, DNA repair, and DNA conformation changes. Based on the MF, the identified genes were mainly associated with catalytic activity, acting on DNA. In terms of CC, the genes were enriched in the following categories: chromosomal region, nuclear chromosome, spindle, microtubule organizing center, and replication fork. The top 14 KEGG pathways for genes that were co-expressed with UBE2T are shown in Fig. 6b. Among these, cell cycle signaling, DNA replication, spliceosome, Fanconi anemia, and p53 signaling pathways were found to be involved in ovarian cancer tumorigenesis and progression. To further determine the relationship between the enriched terms, a similarity network was constructed, where terms with a similarity >0.3 were connected by edges. The network was visualized by using Cytoscape. Each node represented an enriched term and was colored according to its cluster (Fig. 6c), and P-value (Fig. 6d). In patients with high UBE2T expression, significantly upregulated gene sets with nominal P<0.05 and FDR<0.25 included “HALLMARK_E2F_TARGETS” and “HALLMARK_G2M_ CHECKPOINT”. The enrichment plots are shown in Fig. 6e.
The presence of somatic mutations was investigated in cases with high and low UBE2T expression. TP53 and TTN were the top two mutated genes in both groups, and a high frequency of mutations in DST, CSMD1, MUC17, and NEB genes was specifically found in patients with high UBE2T expression. Moreover, mutations in TOP2A, VPS13B, NF1, AHNAK, and FLG2 were significantly enriched in patients with low UBE2T expression (Fig. 6f,g). KEGG pathway analysis further demonstrated that in patients with high UBE2T expression the mutations mainly affected focal adhesion, calcium signaling, and ECM-receptor interactions (Fig. 6h). To explore the cause of UBE2T upregulation in ovarian cancer, we analyzed its correlation with gene copy number and methylation level. We found that UBE2T expression increased with the copy number. Therefore, the high level of UBE2T could be partially attributed to gene amplification (Fig. 6i). The expression of UBE2T was negatively correlated with the methylation level, as hypomethylation was associated with high UBE2T expression (Pearson’ s (372)=-0.125, P=0.0162, Fig. 6j).
Correlation between immune factors and UBE2T expression in ovariancancer
Recently, immunotherapy has received increasing attention, becoming a new treatment strategy for ovarian cancer. We further explored the correlation between immune factors, including tumor-infiltrating lymphocytes (TILs) and immune modulators, and UBE2T expression in ovarian cancer by using the TISIDB database. Immune modulators can be further classified into immune inhibitors, immune stimulators, and MHC molecules. TILs with the highest correlation with UBE2T expression included act_CD4 cells (Spearman: rho=0.331, P=3.57e-09), eosinophils (Spearman: rho=-0.424, P<2.2E-16), neutrophils (Spearman: rho=-0.274, P=1.23E-06), and memory B cells (Spearman: rho=-0.271, P=1.61E-06). Immune inhibitors with the highest correlation with UBE2T expression included CSF1R (Spearman: rho=-0.345, P=6.53E-10), TGFB1 (Spearman: rho=-0.257, P=5.53E-06), KDR (Spearman: rho=-0.228, P=5.95E-05), and CD160 (Spearman: rho=-0.155, P=0.0064). Immune stimulators with the strongest correlation with UBE2T expression included C10orf54 (Spearman: rho=-0.336, P=1.87E-09), NTSE (Spearman: rho=-0.308, P=4.13E-08), TNFSF14 (Spearman: rho=-0.298, P=1.23E-07), and TNFSF15 (Spearman: rho=-0.257, P=5.37E-06). MHC molecules most strongly correlated with UBE2T expression included HLA-E (Spearman: rho=-0.173, P=0.00245), HLA-DOA (Spearman: rho=-0.171, P=0.00265), HLA-DQB1 (Spearman: rho=-0.129, P=0.0236), and HLA-DQA1 (Spearman: rho=-0.121, P=0.0341, Additional file 5). Therefore, UBE2T may affect the immune activity in the ovarian cancer microenvironment by regulating the above immune factors.
UBE2T expression in different ovarian tissues
The IHC staining of UBE2T in tissues from ovarian epithelial malignant tumor, ovarian epithelial borderline tumor, ovarian epithelial benign tumor, and normal ovary is shown in Fig. 7a–d, demonstrating that UBE2T was mainly localized in the cytoplasm and the plasma membrane. The rates of positive and highly positive expression of UBE2T in ovarian epithelial malignant tumors (89.53% and 72.09%, respectively) were significantly higher than in ovarian epithelial borderline tumors (55.00% and 25.00%, respectively), ovarian epithelial benign tumors (26.37% and 13.34%, respectively), and normal ovarian tissues (20.00% and 10.00%, respectively, Fig. 7e,f). Consistently, GSE51088 analysis results showed a significant upregulation of UBE2T in malignant tumors compared to borderline tumors, benign tumors, and normal ovary (Fig. 7g).
Relationship between UBE2T expression and clinicopathological features of epithelial ovarian cancer
According to the median expression value of UBE2T, ovarian cancer patients in GSE datasets were divided into a high-expression and a low-expression group. We found a significantly higher UBE2T high expression rate of UBE2T in FIGO stage III–IV patients than in FIGO stage I–II patients in both clinical samples (85.11% vs. 56.41 %, P=0.003) and GSE73614 (61.02% vs. 37.50 %, P=0.016). However, although UBE2T expression did not significantly correlate with age, stage of differentiation, presence of lymph node metastasis or pathologic subtype in our clinical samples, it was strongly associated with the pathological subtype of ovarian cancer in GSE73614 (P=0.001, Table 1, 2). UBE2T displayed the highest high positive rate in serous carcinoma (100%) and the lowest high positive rate in clear cell carcinoma (27.03%).
Table 1 Correlation between UBE2T expression and clinicopathological features in ovarian cancer clinical samples
Characteristics
|
Cases
|
Low
|
|
High
|
|
High positive rate
|
Chi square Value
|
|
(n=86)
|
−
|
+
|
++
|
+++
|
(%)
|
(P value)
|
Age (years)
|
|
|
|
|
|
|
0.031 (0.861)
|
<60
|
67
|
9
|
10
|
26
|
22
|
71.64
|
|
≥60
|
19
|
0
|
5
|
7
|
7
|
73.68
|
|
FIGO Stage
|
|
|
|
|
|
|
8.724 (0.003*)
|
I-II
|
39
|
7
|
10
|
11
|
11
|
56.41
|
|
III-IV
|
47
|
2
|
5
|
22
|
18
|
85.11
|
|
Differentiation
|
|
|
|
|
|
|
1.939 (0.164)
|
Well-moderately
|
47
|
7
|
9
|
21
|
10
|
65.96
|
|
Poorly-undifferentiated
|
39
|
2
|
6
|
12
|
19
|
79.48
|
|
Lymph node metastasis
|
|
|
|
|
|
|
1.418 (0.492)
|
No
|
57
|
6
|
12
|
19
|
20
|
68.42
|
|
Yes
|
22
|
2
|
2
|
10
|
8
|
81.82
|
|
No lymphadenectomy
|
7
|
1
|
1
|
4
|
1
|
71.43
|
|
Pathological subtype
|
|
|
|
|
|
|
7.963 (0.093)
|
Mucinous
|
6
|
2
|
2
|
2
|
0
|
33.33
|
|
Endometrioid
|
7
|
1
|
2
|
3
|
1
|
57.14
|
|
Serous
|
57
|
5
|
10
|
23
|
19
|
73.68
|
|
Poorly differentiated adenocarcinoma
|
10
|
1
|
1
|
3
|
5
|
80.00
|
|
Clear cell carcinoma
|
6
|
0
|
0
|
2
|
4
|
100.00
|
|
Table 2 Correlation between UBE2T expression and clinicopathological features in ovarian cancer patients of the GSE73614 dataset
Characteristics
|
Cases
(n=107)
|
Low High
|
High positive rate
(%)
|
Chi square Value
(P value)
|
Age (years)
|
|
|
|
|
0.231 (0.631)
|
<60
|
52
|
27
|
25
|
48.08
|
|
≥60
|
55
|
26
|
29
|
52.73
|
|
FIGO Stage
|
|
|
|
|
5.856 (0.016*)
|
I-II
|
48
|
30
|
18
|
37.50
|
|
III-IV
|
59
|
23
|
36
|
61.02
|
|
Differentiation
|
|
|
|
|
3.645 (0.162)
|
Moderately
|
29
|
17
|
12
|
41.37
|
|
Poorly
|
69
|
34
|
35
|
50.72
|
|
Undifferentiated
|
9
|
2
|
7
|
77.78
|
|
Pathological subtype
|
|
|
|
|
14.772 (0.001*)
|
Clear cell carcinoma
|
37
|
27
|
10
|
27.03
|
|
Endometrioid
|
66
|
26
|
40
|
60.61
|
|
Serous
|
4
|
0
|
4
|
100
|
|
Relationship between UBE2T expression and prognosis in patients with epithelial ovarian cancer
During the follow-up period, of 86 patients with epithelial ovarian cancer, 32 died (37.20%) and 12 were lost to follow-up (13.95%). The 5-year OS rate was 69.7% and the mean survival time 89.7 months (95% confidence interval (CI): 80.6~98.8 months). Kaplan-Meier survival analysis revealed that in patients with high UBE2T expression the mean OS was significantly lower than that in those with low UBE2T expression (P=0.00093, Fig. 7h). Univariate Cox regression analysis indicated that the OS of ovarian cancer patients was markedly correlated with UBE2T expression, FIGO stage, and differentiation status (HR:6.462, 4.029, and 0.372, respectively, all P<0.05). Multivariate analysis showed that high UBE2T expression, advanced stage, and poorly differentiated or undifferentiated status were independent risk factors in patients with ovarian cancer (HR: 4.095, 3.210, and 0.422, respectively, all P<0.05, Table 3). Furthermore, high UBE2T expression was confirmed to be a predictor of poor OS in the GSE63885 (P=0.03, Fig. 7i) datasets and GSE73614 (P=0.035, Fig. 7j). Moreover, application of the Cox regression model to the GSE63885 dataset confirmed high UBE2T expression as an independent risk factor for OS in ovarian cancer patients (HR=1.717, P=0.031, Table 4).
Table 3 Univariate and multivariate analyses of overall survival in ovarian cancer clinical samples
Variable
|
Univariate analysis
|
|
Multivariate analysis
|
|
|
HR (95% CI)
|
P value
|
HR (95% CI)
|
P value
|
Age
|
|
|
|
|
<60 vs. ≥60
|
1.345 (0.604-2.995)
|
0.469
|
|
|
UBE2T expression
|
|
|
|
|
High vs. low
|
6.462 (1.875-22.266)
|
0.003
|
4.095 (1.154-14.533)
|
0.029
|
FIGO Stage
|
|
|
|
|
I-II vs. III-IV
|
4.029 (1.733-9.367)
|
0.001
|
3.210 (1.348-7.640)
|
0.008
|
Differentiation
|
|
|
|
|
Poorly-undifferentiated vs. well-moderately
|
0.372 (0.179-0.772)
|
0.008
|
0.422 (0.201-0.884)
|
0.022
|
Pathological subtype
|
|
|
|
|
Serous vs. nonserous
|
1.177 (0.575-2.409)
|
0.656
|
|
|
Lymph node metastasis
|
|
|
|
|
Yes vs. no
|
1.964 (0.959-4.022)
|
0.065
|
|
|
|
|
|
|
|
|
|
Table 4 Univariate and multivariate analyses of overall survival in ovarian cancer patients of the GSE63885 dataset
Variable
|
Univariate analysis
|
|
Multivariate analysis
|
|
|
HR (95% CI)
|
P value
|
HR (95% CI)
|
P value
|
UBE2T expression
|
|
|
|
|
High vs. low
|
1.711 (1.048-2.793)
|
0.032
|
1.717 (1.052-2.803)
|
0.031
|
FIGO Stage
|
|
|
|
|
II-III vs. IV
|
2.214 (1.093-4.485)
|
0.027
|
2.226 (1.102-4.495)
|
0.026
|
Differentiation
|
|
|
|
|
Poorly-undifferentiated vs. moderately
|
2.321 (1.040-5.178)
|
0.04
|
|
|
Pathological subtype
|
1.293 (0.517-3.233)
|
0.582
|
|
|
Serous vs. nonserous
|
|
|
|
|