Identification of cuproptosis-related lncRNAs and differential analysis
A total of 532 EC samples and 35 normal samples from the TCGA database were finally enrolled. The detailed clinical characteristics of the 532 EC patients are summarized in Table 1. In total, 19 curoptosis-related genes were identified. The landscape and connections of the interactions among the 19 cuproptosis related genes and its related lncRNAs were illustrated in the network (Fig. 1A). The differentially expressed CRLs (DE-CRLs) in the patients with EC were visualized in a volcano map (Fig. 1B) and the heatmap (Fig. 1C). As the figure shown, there were 18 up-regulated and 55 down-regulated DE-CRLs.
Table 1
Characteristics of patients from TCGA dataset
Variables | | TCGA cohort |
Total number | | 532 |
Living status | Alive | 440 | 82.71 |
| Death | 92 | 17.29 |
Age | ༞=60 | 178 | 33.46 |
༜60 | 354 | 66.54 |
Tumor grade | G1 | 98 | 18.42 |
G2 | 119 | 22.37 |
G3 | 315 | 59.21 |
FIGO stage | I | 333 | 62.59 |
II | 52 | 9.77 |
III | 119 | 22.37 |
IV | 28 | 5.26 |
Histological type | EEA | 399 | 75.00 |
MIX | 22 | 4.14 |
SEA | 111 | 20.86 |
Menopausal status | post-menopause | 458 | 86.09 |
pre-menopause | 74 | 13.91 |
Recurrence | Yes | 106 | 19.92 |
No | 426 | 80.08 |
Peritoneal cytology | Negative | 458 | 86.09 |
Positive | 74 | 13.91 |
Cancer status | With tumer | 103 | 19.36 |
Tumor free | 429 | 80.64 |
LNM | Negative | 451 | 84.77 |
Positive | 81 | 15.23 |
FIGO, International Federation of Gynecology and Obstetrics; G, grade; EEA, endometrioid endometrial adenocarcinoma; LNM, lymph node metastasis. |
Table 2
Hub genes and correlated coeffificient value
Cuproptosis-related gene | Coeffificient |
AC008073.2 | 0.03360684 |
AC009229.1 | 0.11364294 |
AF131215.7 | 0.00094354 |
AL807761.3 | -0.0015656 |
CDKN2A.DT | 0.15255722 |
LINC01166 | 0.07124071 |
LINC01615 | -0.0138985 |
risk score | low: < 1.241 |
high: ≥ 1.241 |
Construction And Validation Of A Prognostic Model By The Crls
Using LASSO Cox regression analysis, we finally determined that 7 prognostic CRLs (AL807761.3, AF131215.7, AC008073.2, AC009229.1, CDKN2A.DT, LINC01615, LINC01166) were significantly associated with prognosis in EC (Fig. 2A-B). The signature, which consisted of the 7 key prognostic CRLs, was constructed using the respective regression coefficients. The risk scores of EC patients were calculated according to the following formula: Risk score = (AC008073.2×0.033606845)+ (AC009229.1×0.113642941) + (AF131215.7×0.000943541) + (AL807761.3×-0.001565595) + (CDKN2A.DT×0.152557217)+ (LINC01166×0.071240712)+ (LIN01615×-0.013898506). Then a prognostic model was built according to the coefficient for each which was obtained through the penalizing process. The relationships among the nine genes are shown in Fig. 2C. AF131215.7, AC008073.2, AC009229.1, CDKN2A.DT, LINC01615, and LINC01166 were positively correlated with each other, but AL807761.3 was negatively correlated with the other six genes. The expression of the 7 CRLs was compared between normal and tumor tissues (Fig. 2D and Figure S1A-B). The results showed that AC008073.2, AC009229.1, AF131215.7, AL807761.3, and LINC01166 were significantly up-regulated in normal groups, while CDKN2A.DT and LINC01615 were down-regulated in the normal groups. Survival curves revealed that patients in high expression of AC008073.2, AC009229.1, AF131215.7, and CDKN2A.DT have a better OS compared with low expression group. On the other hand, the survival of high expression of LINC01615 was poor. The survival curve had no significant difference in AL807761.3 and LINC01166 groups (Fig. 2E and Figure S1C-D).
Construct the prognostic characteristics of cuproptosis in patients with
Then we calculated the risk score of each patient, and the patients were stratified into high-risk group or low-risk group according to the median cut-off value. The expression levels of the 7 CRLs and the distribution of clinicopathological features in high- and low-risk groups were presented in the heatmap (Fig. 3A). The results revealed that the clinical features, including cancer status, peritoneal cytology, recurrence, tumor grade, menopausal status, histological type, stage, age, and living status are significant differently distributed in two risk groups. The EC patients were then divided into low- and high-risk groups based on the median risk score. Compared with patients in the low-risk group, those in the high-risk group indicated poorer survival probability based on our risk model (Fig. 3B). The K-M curve indicated patients in the low-risk group had a significantly better overall survival (OS) compared to those in the high-risk group (Fig. 3C). As shown in the time-dependent ROC analysis, the area under the curve (AUC) of the risk model reached 0.781 (Fig. 3D). Univariate and multivariate Cox regression analyses were carried out among the available variables to determine whether the predictive ability of the 7-CRL signature in predicting OS was independent of other traditional clinicopathological properties. It proved that the risk score of the prognostic signature (HR = 1.908, 95% CI = 1.179 − 3.089, P < 0.009) was significantly associated with OS in multivariate Cox regression analysis (Fig. 3E).
Immune characterization and functional analysis in different risk signatures
To verify the feasibility of the grouping strategy, we conducted an ESTIMATE analysis to profile the immune characteristics of EC based on the expression of immune cell types. Stromal, immune, and ESTIMATE scores were calculated and then we conducted a correlation analysis between these scores and risk scores. The results showed that the risk score decreased with the increase of stromal score (Fig. 4A), immune score (Fig. 4B), and ESTIMATE score (Fig. 4C). The violin plot also further verified that there the high risk signature was significantly positively correlated with the ESTIMATE and immune and stromal scores (Fig. 4D). To further explore the biological behaviors among the two immune subtypes, we performed GSVA enrichment analysis. As shown in Fig. 4E, low risk group was related to starch and sucrose metabolism, folate biosynthesis, fructose and mannose metabolism, amino sugar and nucleotide sugar metabolism, and butanoate metabolism. High risk group was enriched in glycosphingolipid biosynthesis ganglio series and autoimmune thyroid disease. GSEA was then performed to test whether the genes of patients in different risk were enriched in previously defined biological pathways. We found that the genes of patients in low risk was enriched in cell cycle, durg metabolism cytochrome 450, mismatch repair, regulation of the cytoskeleton, and ubiquitin mediated proteolysis. Patients in high risk group were enriched in endometrial cancer, fatty acid metabolism, glycerolipid metabolism, type I diabetes, tyrosine metabolism (Fig. 4F). Next, to investigate the correlation between TIICs and risk signature in EC, we used CIBERSORT to calculate infiltration of 22 immune cells in the EC cases. Then, we compared the infiltration of 22 immune cells in distinct signature. The differentially analytical results showed that CD8 T cells, T cells CD4 memory activated, Tregs, NK cells resting, NK cells activated, and Mast cells resting were significantly different in two groups (Fig. 4G).
Distribution of risk score and correlation with mRNSsi and TMB
We next analyzed the distribution of risk score in different clinicopathological features, named cancer status, grade, lymph node metastasis, recurrence, peritoneal cytology, and tumor stage (Fig. 5A-F). The results revealed that invasive and poor clinical features had higher risk scores, such as patients with tumor, tumor grade 3, positive LNM, recurrence, positive peritoneal cytology, and tumor stage IV, which is significant. Correlative analysis indicated that mRNAsi is positive associated with risk score (Fig. 5G), while tumor mutation burden (TMB) is negative related with risk score (Fig. 5H). We then combined the risk signature with mRNAsi or TMB score, and divided the cohort into for subgroups according to its corresponding features. The results showed that patients in low risk group and high mRNAsi subgroup had the best prognosis, and patients from high risk group with high mRNAsi subgroup tended to survive shorter than the patients from other subgroups (Fig. 5I). On the other analysis, patients with low risk score and high TMB had a longer survival time, while high risk score with low TMB patients had a worse prognosis (Fig. 5J).
Construction And Evaluation Of The 7-crl Related Nomogram
Based on the multivariate Cox regression analyses, age, tumor grade, stage, and 7-CRL signature were proved to be independent predictors for OS. Then a nomogram that integrated the risk score model and these clinicopathological characteristics was constructed to quantify the predictive results of OS probability of the EC patients (Fig. 6A). The C-index for the nomogram was 0.760, and the calibration curves of the nomogram showed great consistency between the predicted OS rates and actual survival situation at 1, 3, and 5 years (Fig. 6B). We calculated the total score of each patient according to the nomogram (Table 3) and then categorized the patients into three subgroups evenly as low-, moderate-, and high score subgroups in the basis of total score. Additionally, ROC curve analysis showed that the nomogram provided a more accurate prediction for OS at 1-year survival (AUC = 0.729), 3-year survival(AUC = 0.792) and 5-year survival (AUC = 0.822, Fig. 6C). Kaplan-Meier analysis indicated that the survival status of the patients in the low score group was significantly better than that in the patient of medium-risk group and high-risk group (p < 0.001, Fig. 6D). To further evaluate and test the nomogram, we conducted the survival analysis in different clinical subgroups, including tumor grade (G1-2, G3) and FIGO stage (Stage I and Stage II-IV). The test of survival model showed the similar results, and these patients presented the same predictive tendencies, which suggested that this nomogram could accurately differentiate patients in the whole groups and in different clinicopathological subgroups (Fig. 6E-H). These results indicated that our nomogram not only have high accuracy in predicting OS for EC patients, but also can find specific patients in different clinicopathological characteristics.
Table 3
Corresponding risk score for each variable and total score.
Variables | Score |
age | < 60 | 0 |
≥ 60 | 30 |
tumor grade | G1 | 0 |
G2 | 35 |
G3 | 100 |
FIGO stage | I | 0 |
II | 20 |
III | 50 |
IV | 75 |
risk score | low | 0 |
high | 70 |
total score | low | 0-100 |
moderate | 110–160 |
high | ≥ 170 |
FIGO, International Federation of Gynecology and Obstetrics. |
In Vitro Experimental Validation For Linc01615
We first performed the western blot (WB) validation in different cell lines following the steps described above. We chose LINC01615 as our target because this molecule is the only one that have obvious promotion function for cancer progression. We then verified the expression of LINC01615 in five cell lines including HEC-1A, HEC-1B, AN3CA, ishikawa, and RL-952, and find that the expression of LINC01615 is higher in HEC-1B, AN3CA, and ishikawa (Fig. 7A). We then knockdown LINC01615 with siRNA in AN3CA (Fig. 7B), and conducted transwell experiments to find whether LINC01615 had an influence on the invasion of EC cell lines. The result showed that knockdown of LINC01615 significantly inhibited the invasive ability of EC cell line AN3CA (Fig. 7C).