Development of a senescence-related lncRNA signature in endometrial cancer based on multiple machine learning models

Abstract

Background: Senescence-related lncRNAs (srlncRNA) mediate carcinogenesis in various malignancies. However, its roles in endometrial cancer (EC) remain unknown. Our research aims to construct a predictive srlncRNA model with prognostic and therapeutic significance in EC.

Methods: We first downloaded the gene expression and medical information from the TCGA, as well as senescence-related lncRNAs (srlncRNAs) from the CellAge databases. Then, a co-expression network of cell senescence-related mRNA-lncRNA was explored with R. Subsequently, we performed Cox and Lasso regression and machine learning analysis to identify srlncRNAs related to the prognosis of EC and built a predictive model. Continually, we drew a nomogram to improve its ability to predict prognosis. Further, GSEA was used to explore potential mechanisms. Differences in TME, immune infiltrating cells, and checkpoints of the two risk groups were compared using GSEA and CIBERSORT. Finally, the drug sensitivity of patient-derived tumor organoids (PDOs) was investigated.

Results: We first built a prognostic model based on seven srlncRNAs (AL121906.2, AP002761.4, BX322234.1, LINC00662, LINC00908, VIM-AS1, and ZNF236-DT). The model, which was screened by machine learning, functioned well in three sets with good stability and accuracy. Furthermore, the nomogram based on age, grade, and risk scores could precisely predict the prognosis of EC patients. The AUC of risk scores was highest compared to other clinical parameters (AUC risk score = 0.769, AUC age = 0.615, and AUC grade = 0.681). This srlncRNAs were enriched in the cell cycle, certain malignant tumors, and cancer-associated regulatory pathways. Afterward, low-risk EC patients had more immune-infiltrating cells and may benefit from anti-PD-1 and anti-CTLA4 treatment. Paclitaxel, gemcitabine, and cisplatin (all p < 0.05) may be more useful in EC patients with high expression of targeted srlncRNAs in the GDSC database. The levels of targeted srlncRNAs and drug sensitivity varied significantly among different EC PDOs. The EC-18 PDO was more resistant to three drugs, which aligned with clinical observation.

Conclusion: The srlncRNA signature (AL121906.2, AP002761.4, BX322234.1, LINC00662, LINC00908, VIM-AS1, and ZNF236-DT) could guide prognosis prediction and treatment choices for EC patients.

Keywords: TCGA; cell senescence; endometrial cancer; immunotherapy; patient-derived tumor organoids; prognosis; signature.

FIGURE 1

Flow diagram of the study.

FIGURE 2

Identification of srlncRNAs with significant prognostic value in EC. (A) The forest showed the HR (95%CI) and p-value of selected lncRNAs by univariate Cox proportional-hazards analysis. (B,C) Lasso regression.

FIGURE 3

Survival curve of EC patients in different groups. (A) Comparison of survival rates between high and low VIM-ASI levels in EC patients using Kaplan-Meier analysis. (B) Comparison of survival rates between high and low LINC00908 levels in EC patients using Kaplan-Meier analysis. (C) Comparison of survival rates between high and low ZNF236-DT levels in EC patients using Kaplan-Meier analysis. (D) Comparison of survival rates between high and low AP002761.4 levels in EC patients using Kaplan-Meier analysis. (E) Comparison of survival rates between high and low BX322234.1 levels in EC patients using Kaplan-Meier analysis. (F) Comparison of survival rates between high and low LINC00662 levels in EC patients using Kaplan-Meier analysis. (G) Comparison of survival rates between high and low AL121906.2 levels in EC patients using Kaplan-Meier analysis. EC, endometrial cancer.

FIGURE 4

Screening of prognostic srlncRNA in EC. (A) A co-expression network of srlncRNAs and mRNAs. (B) The Sankey diagram of the relationship between lncRNA and mRNA.

FIGURE 5

The prognostic value of our predictive model in the entire, train, and test sets. (A) Exhibition of risk scores of the low and high-risk groups in three sets. (B) Survival time and survival status of EC patients in low and high-risk groups of three sets. (C) Kaplan-Meier survival curves for EC patients in low- and high-risk groups across three sets. (D) Time-dependent ROC curves between low and high-risk groups in three sets.

FIGURE 6

Assessment of the prognostic survival model based on seven srlncRNAs. (A,B) Univariate and Multivariate Cox regression analysis of risk score and clinical factors. (C) The nomogram of risk score and clinical factors. Clinical factors: age and grade. (D) The AUC for risk model score, age, and grade. Clinical factors: age and grade. (E) The 1-year OS calibration curve. (F) The 3-year OS calibration curve. (G) The 5-year OS calibration curve.

FIGURE 7

(A) Heat map of seven srlncRNAs’ expression. (B) Survival analysis of the high-risk and low-risk groups in patients under different subgroups (age ≤ 65, age >65, grades 1, and grades 2–3).

FIGURE 8

The results of functional analysis based on seven srlncRNAs model by GSEA. (A) GO enrichment analysis. (B) KEGG enrichment analysis.

FIGURE 9

The investigation of tumor immune factors and immunotherapy. (A–C) Estimate score, immune score, and stromal score in high-and low-risk groups. (D,E) The varied proportions of immune cells and immune functions in high-and low-risk groups by ssGSEA. (F) The 29 immune checkpoint inhibitor (ICI) levels in high-and low-risk groups.

FIGURE 10

(A) The organoid morphology of P0, P1, and P2. (B) Comparison of HE and IHC between organoids and EC tissues. HE, Hematoxylin-eosin staining. IHC, Immunohistochemistry.

FIGURE 11

(A) Relative mRNA levels of targeted genes from four organoids; (B) Drug sensitivity analysis, including paclitaxel, gemcitabine, and cisplatin, based on the GDSC database; (C) The inhibition ratio plot of organoids treated with various concentrations of paclitaxel, gemcitabine, and cisplatin. GDSC, Drug Sensitivity in Cancer.