Comprehensive analysis of the value of angiogenesis and stemness‐related genes in the prognosis and immunotherapy of ovarian cancer

Abstract

Tumor angiogenesis and the presence of cancer stem cells (CSCs) are critical characteristics of tumors. Previous research has demonstrated that cancer stem cells promote tumor angiogenesis, while increased vascularity, in turn, fosters the growth of cancer stem cells. This creates a detrimental cycle that contributes to tumor progression. However, studies investigating the angiogenesis and stemness characteristics in ovarian cancer (OV) are limited. In this study, we employed cluster analysis and LASSO methods to assess the significance of angiogenesis‐ and stemness‐related genes in the efficacy of OV immunotherapy. Through multivariate Cox regression analysis and Friends analysis, we identified TNFSF11 as the most significant prognostic gene associated with angiogenesis and stemness. Additionally, molecular docking results confirmed that TNFSF11 exhibits a high affinity for sorafenib and sunitinib. In summary, for the first time, we conducted a comprehensive analysis of the roles of angiogenesis and stemness‐related genes in the prognosis and immunotherapy of OV patients, revealing TNFSF11 as a novel therapeutic target.

Tumor angiogenesis and the presence of cancer stem cells (CSCs) are critical characteristics of tumors. In this study, we utilized cluster analysis and LASSO methods to assess the significance of angiogenesis‐ and stemness‐related genes in the efficacy of oncolytic virus (OV) immunotherapy. Through multivariable Cox regression analysis and Friends analysis, we identified TNFSF11 as the most significant prognostic gene associated with angiogenesis and stemness. Furthermore, molecular docking results confirmed that TNFSF11 demonstrates a high affinity for sorafenib and sunitinib. In summary, we conducted the first comprehensive analysis of the role of angiogenesis and stemness‐related genes in the prognosis and immunotherapy of OV patients, highlighting TNFSF11 as a novel therapeutic target.

Abbreviations

CA125

carbohydrate antigen 125

CHRDL1

chordin‐like 1

CNVs

copy number variations

CSCs

cancer stem cells

GSCA

gene set cancer analysis

GSEA

gene set enrichment analysis

HE4

human epididymis protein 4

ICB

immune checkpoint blockade

KEGG

Kyoto Encyclopedia of Genes and Genomes

LASSO

least absolute shrinkage and selection operator

OV

oncolytic virus

qRT‐PCR

quantitative real‐time polymerase chain reaction

SNVs

single nucleotide variations

TCGA

The Cancer Genome Atlas

TGF‐β

transforming growth factor‐beta

VECs

vascular endothelial cells

1. INTRODUCTION

Ovarian cancer (OV) is a highly aggressive gynecological malignancy, and early diagnosis and treatment are often inadequate. ¹ While the overall five‐year relative survival rate is approximately 49%, this rate drops to only 29% for patients with advanced OV. ² Despite the clinical successes of progressive treatment approaches, mortality rates remain high. ³ The classic OV biomarkers, carbohydrate antigen 125 (CA125) and human epididymis protein 4 (HE4), can facilitate preliminary diagnosis; however, it is essential to acknowledge issues such as false positives and their limited prognostic predictive capability. ⁴ In clinical practice, treatment methods for OV are becoming increasingly diverse, particularly with the emergence of immunotherapy, which has provided new hope for patient survival. Nevertheless, accurately assessing patient prognosis, predicting treatment response, and selecting appropriate treatment options continue to present significant challenges for the medical community.

Initially, Professor Folkman proposed that tumor growth and metastasis depend on angiogenesis, positing that inhibiting this process could serve as an effective strategy for tumor treatment. ⁵ In recent years, targeting angiogenic factors has emerged as a significant approach for both tumor treatment and prevention. Vascular endothelial growth factor (VEGF) is a critical molecule in tumor angiogenesis, and inhibiting VEGF has demonstrated promising therapeutic effects across various cancers. ⁶ For instance, parecoxib has been shown to inhibit hepatocellular carcinoma tumorigenesis and angiogenesis via the ERK‐VEGF/MMPs signaling pathway. ⁷ Additionally, chordin‐like 1 (CHRDL1) influences the angiogenesis and metastasis of colorectal cancer cells through the transforming growth factor‐beta (TGF‐β)/VEGF pathway. ⁸ Cancer stem cells (CSCs) are a subset of cancer cells that possess the ability to self‐renew and differentiate into various cell types that make up the tumor. They are believed to be responsible for tumor initiation, maintenance, metastasis, and recurrence after treatment. CSCs are often characterized by their resistance to conventional therapies, which typically target rapidly dividing cells, leaving CSCs intact and capable of regenerating the tumor. ⁹ Although CSCs constitute only a small fraction of the overall tumor, they play a crucial role in sustaining tumor progression, malignant proliferation, invasion, drug resistance, metastasis, and recurrence. ¹⁰ For example, MUC1‐EGFR modulates the stemness of lung adenocarcinoma and impacts paclitaxel resistance by activating the NF‐κB and MAPK pathways. ¹¹ Furthermore, inhibiting HMGCR reduces the stemness and metastasis of hepatocellular carcinoma through Hedgehog signaling. ¹² Increasing evidence suggests that the interplay between tumor angiogenesis and CSCs has a synergistic effect on tumor growth. On one hand, CSCs directly promote the formation of tumor blood vessels by secreting pro‐angiogenic factors and differentiating into vascular endothelial cells (VECs). ¹³ Additionally, the tumor vasculature supplies oxygen, nutrients, and factors that support CSCs, thereby creating an ecological niche for them. ¹⁴ Consequently, a mutually reinforcing vicious cycle emerges between CSCs and tumor vasculature, which is critical for tumor growth.

In this study, we conducted a screening of angiogenic and stemness genes that exhibited differential expression in OV and normal tissues, and that were associated with prognosis. We analyzed the differential expression, prognostic value, and relevance to immunotherapy of angiogenesis‐ and stemness‐related genes in OV using subcomponent typing and LASSO methods. Our results indicate that TNFSF11 is the most significant prognostic gene among the angiogenesis‐ and stemness‐related genes, highlighting its correlation with OV immunotherapy. Furthermore, we confirmed the critical role of TNFSF11 and angiogenesis‐targeted drugs through molecular docking studies. Overall, our findings suggest that TNFSF11 is pivotal in the progression of OV, positioning it as a promising therapeutic target and a reliable prognostic indicator.

2. MATERIALS AND METHODS

2.1. Data acquisition

We identified a total of 1524 angiogenesis‐related genes and 10,809 stemness‐related genes (score >1) from the GeneCards website. Transcriptional data and clinical information for 376 OV cases and 180 normal tissues were obtained from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/). Additionally, data from 57 primary OV samples and 12 normal ovarian tissue samples from GSE69957 were utilized as the validation cohort. In addition, clinical prognostic information from 80 samples in the GSE14764 dataset was used as a validation of the prognostic model.

2.2. Mutation frequency, somatic copy number, and pathway activity analyses

Data from the Gene Set Cancer Analysis (GSCA, http://bioinfo.life.hust.edu.cn/GSCA/#/) were used to evaluate the occurrence of gene mutations linked to angiogenesis and stemness. This analysis included single nucleotide variations (SNVs), amplifications, and both homozygous and heterozygous deletion data. ¹⁵ Additionally, GSCA performed correlation analyses on well‐known cancer‐related pathways, determining the strength of these correlations via Spearman correlation analysis. The CancerSEA database was also utilized to explore the roles of genes connected to angiogenesis and stemness.

2.3. Consistency cluster analysis

An analysis of consistency was performed utilizing the ConsensusClusterPlus package in R (version 1.54.0). A sample size of 80% was selected 100 times, producing a maximum of six clusters. The hierarchical clustering approach was implemented using clusterAlg="hc" with innerLinkage set to ‘ward.D2’.

2.4. Construction of prognostic models

A total of twelve prognostic genes associated with angiogenesis and stemness were selected, and a prognostic model was developed utilizing the TCGA‐OV dataset through the least absolute shrinkage and selection operator (LASSO) method. Validation of the model was performed with the GSE14764 dataset. To filter the features, the LASSO regression technique was applied, and the analysis was executed employing 10‐fold cross‐validation. ¹⁶ The glmnet package in R was used to conduct the analysis.

2.5. Functional enrichment analysis

To enhance the verification of the possible roles of genes related to angiogenesis and stemness, functional enrichment analysis was performed on the data. The Kyoto Encyclopedia of Genes and Genomes (KEGG) serves as a useful tool for examining gene functions and provides high‐level information related to the genome. ¹⁷ , ¹⁸ Furthermore, Gene Set Enrichment Analysis (GSEA) was conducted using the clusterProfiler package.

2.6. Immune infiltration analysis

To assess the immune scores associated with angiogenesis and stemness‐related genes in OV, we employed immunedeconv, an R package that incorporates six sophisticated algorithms: TIMER, xCell, MCP‐counter, CIBERSORT, EPIC, and quanTIseq. For the purpose of our analysis, we focused on implementing the xCELL algorithm. The xCELL algorithm is designed to quantify immune cell types in the tumor microenvironment using gene expression data. It offers high specificity and sensitivity for various immune cell types, including T cells, B cells, macrophages, dendritic cells, and more. This makes it particularly suitable for studies aiming to dissect the immune landscape of tumors. This particular algorithm was selected due to its capability to evaluate a diverse array of immune cell types, making it highly suitable for our investigation. Additionally, we utilized the TIDE algorithm to predict the possible response to immune checkpoint blockade (ICB).

2.7. Cell culture and transient transfection

Cell lines IOSE80, A2780, HEY, SK‐OV‐3, and CoC1 were purchased from Beijing Bena Biotechnology Co. (Beijing, China). Cells were cultured in DEME F‐12 medium with 10% FBS (Gibco). The negative control (NC) and siRNA for TNFSF11 (Sagon, China) were transfected into the cells utilizing Lipofectamine 2000 (Invitrogen, Thermo Fisher, USA).

Gene	Target sequences (5′–3′)
si‐TNFSF11#1	CTCCATGAAAATGCAGATTTTCA
si‐TNFSF11#2	TGCAAAAGGAATTACAACATATC

2.8. Quantitative real‐time polymerase chain reaction (qRT‐PCR)

TRIzol (Thermo Fisher, USA) reagent was used to extract total RNA. Using FastStart Universal SYBR Green Master, quantitative reverse transcription‐polymerase chain reaction (qRT‐PCR) was performed on the RNA extracted from each sample (2 μg) on a LightCycler 480 PCR System (Roche, USA). The cDNA was utilized as a template with a reaction volume of 20 μL (2 μL of cDNA template, 10 μL of PCR mixture, 0.5 μL of forward and reverse primers, and an appropriate water volume). The following procedures were utilized for the PCR reactions: Cycling conditions started with an initial DNA denaturation phase at 95°C for 30 s, followed by 45 cycles at 94°C for 15 s, 56°C for 30 s, and 72°C for 20 s. Three separate analyses were performed on each sample. Based on the 2‐ΔΔCT method, data from the threshold cycle (CT) were obtained and standardized to the levels of GAPDH in each sample. The expression levels of mRNA were compared to controls obtained from normal tissues. The following is a list of the sequenCSE of primer pairs for the genes that were being targeted:

Gene	Forward primer sequence (5′–3′)	Reverse primer sequence (5′–3′)
ANGPTL4	GATGGCTCAGTGGACTTCAACC	TGCTATGCACCTTCTCCAGACC
COL8A2	CGGGCGTCCTCTGGTGGCG	TCCATCGGCAGCAGCGGTAGA
CX3CR1	CACAAAGGAGCAGGCATGGAAG	CAGGTTCTCTGTAGACACAAGGC
ERBB2	GGAAGTACACGATGCGGAGACT	ACCTTCCTCAGCTCCGTCTCTT
JUP	ACCAGCATCCTGCACAACCTCT	GGTGATGGCATAGAACAGGACC
TNFSF11	GCCTTTCAAGGAGCTGTGCAAAA	GAGCAAAAGGCTGAGCTTCAAGC
GAPDH	GTCTCCTCTGACTTCAACAGCG	ACCACCCTGTTGCTGTAGCCAA

2.9. Determination of 5‐ethynyl‐2′‐deoxyuridine (EdU)

EdU assay was performed using BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 594 (Biotek, Shanghai, China). After rinsing with PBS, cells were incubated with EdU solution for 2 h and then stained with DAPI solution for nuclei. After washing, the samples were observed with an inverted microscope (Olympus).

2.10. Western blot

Total protein was extracted with RIPA buffer. After protein quantification, aliquots of protein samples were loaded and separated by 10% SDS‐PAGE and then transferred to a 0.45‐μm PVDF membrane. 5% skimmed milk was blocked for 2 h. The corresponding primary antibodies were incubated at 4°C overnight. Subsequently, membranes were incubated with secondary antibodies for 1 h at room temperature. Immunoblots were displayed with an ECL fluorescence detection kit (Beyotime, Shanghai) and visualized with a Tanon 4600 system (Tanon Science and Technology Co., Ltd.).

1	TNFSF11/RANKL Polyclonal antibody	23408‐1‐AP
2	AKT Polyclonal antibody	10176‐2‐AP
3	Phospho‐AKT (Ser473) Monoclonal antibody	66444‐1‐Ig
4	ERK1/2 Recombinant antibody	83533‐1‐RR
5	Phospho‐ERK1/2 (Thr202/Tyr204) Polyclonal antibody	28733‐1‐AP
6	CREB1 Polyclonal antibody	12208‐1‐AP
7	Phospho‐CREB1 (Ser133) Recombinant antibody	81871‐1‐RR
8	STAT6 Recombinant antibody	82630‐1‐RR
9	Phospho‐Stat6 (Tyr641) (D8S9Y) Rabbit mAb	56554
10	HRP‐conjugated beta actin monoclonal antibody	HRP‐66009

2.11. Cell viability

Cell viability was detected using the Cell Counting Kit‐8 assay (Beyotime, China), according to the manufacturer protocol. Cells from different treatments were cultured in 96‐well plates at a density of 1 × 103 cells per well. CCK‐8 solution was applied at the indicated time points. After incubation at 37°C for 2 h, the O.D 450 values of each well were detected using a microplate reader (BioTeK, USA).

2.12. Tube formation assay

Tube formation assay was performed using Matrigel Basement Membrane Matrix (Corning, 356234, USA). To investigate the effect of TNFSF11 on vessel formation. First, Matrigel was added to 96‐well plates using an ice‐cold pipette tip at approximately 40 μL per well and incubated at 37°C to solidify. Then, 1.2 × 104 HUVEC (si‐NC vs. si‐TNFSF11) were seeded into the solidified Matrigel. After 24 h of incubation, the cells were photographed using an inverted light microscope (Lecial, Germany).

2.13. Statistical analysis

All the analysis methods and R package were implemented by R version 4.0.3. A statistically significant difference is indicated by p < 0.05. *p < 0.05, **p < 0.01 and ***p < 0.001.

3. RESULT

3.1. Screening of angiogenesis and stemness‐related genes in OV

To investigate the significance of angiogenesis‐related and stemness‐related genes in OV, we collected relevant genes from Genecard website, applying an inclusion criterion of a score greater than 1. These genes were then integrated with differentially upregulated genes and prognostic risk genes identified in the TCGA‐OV dataset, resulting in a total of 14 differentially prognostic genes associated with stemness and angiogenesis. The prognostic value of these genes was illustrated through forest plots (Figure 1A,B). Additionally, box plots were utilized to compare the expression levels of these 14 genes in cancerous and non‐cancerous samples from the TCGA‐OV and GSE66957 datasets (Figure 1C,D). Furthermore, we conducted a correlation analysis of these 14 genes within the TCGA‐BLCA and GSE66957 datasets, revealing significant positive correlations among most of the genes (Figure 1E,F).

FIGURE 1.

Identification of angiogenesis and stemness‐related differential prognostic genes in OV. (A) Venn diagram identifying angiogenesis and stemness related differential genes. (B) Forest plot demonstrating prognostic differences for 14 genes. (C) Expression of angiogenesis‐ and stemness‐related differential prognostic genes in the TCGA‐OV dataset. (D) Expression of angiogenic and dryness‐related differential prognostic genes in the GSE66957 dataset. (E) Correlation of angiogenesis and stemness‐related differential prognostic genes in the TCGA‐OV dataset. (F) Correlation of angiogenesis and stemness‐related differential prognostic genes in the GSE66957 dataset.

3.2. Functional analysis and genetic alterations in angiogenesis and stemness‐related differential prognostic genes

To investigate the potential roles of angiogenesis and stemness‐related genes in OV, we utilized GSCALite to analyze the associated pathways, single nucleotide variations (SNVs), and copy number variations (CNVs) of these 14 genes in OV. Our functional analysis revealed that these genes were linked to several processes, including cell cycle inhibition, DNA damage response, epithelial‐mesenchymal transition (EMT) activation, androgen receptor inhibition, and RAS/MAPK pathway activation (Figure 2A). The SNV percentage heatmap indicated that ERBB2 emerged as the most frequently mutated gene with deleterious effects in OV (Figure 2B). Additionally, we assessed the distribution proportions of various CNV categories—heterozygous amplification, heterozygous deletion, homozygous amplification, and homozygous deletion—among angiogenesis‐ and stemness‐related genes in OV patients (Figure 2C). Notably, we found that the mRNA levels of these genes exhibited a positive correlation with heterozygous amplified CNVs, while showing a negative correlation with heterozygous deleted CNVs (Figure 2D). Lastly, we further explored the functions of these 14 genes from a single‐cell perspective using the CancerSEA website, revealing their association with critical cell behaviors such as metastasis, hypoxia, and angiogenesis (Figure 2E).

FIGURE 2.

Functional network and genetic alterations of angiogenesis and stemness‐related genes in OV. (A) Functions of angiogenesis and stemness‐related genes. (B) Mutation frequency analysis of angiogenesis and stemness related genes in OV. (C) The CNV information of angiogenesis and stemness‐related genes in OV. (D) The percentage of heterozygous CNV for angiogenesis and stemness‐related genes in OV. (E) CancerSEA database analyzes functions of key genes.

3.3. Cluster analysis of angiogenesis and stemness‐related genes in OV

To investigate the roles of angiogenesis‐ and stemness‐related genes in OV, we conducted subgroup typing of these key genes. CDF curve and cluster heatmap analyses indicated that optimal grouping was achieved by dividing the samples into two clusters (Figure 3A,B). Additionally, the PCA plot clearly illustrates the distribution of samples between these two clusters (Figure 3C). In terms of gene expression, all 14 angiogenesis and stemness‐related genes included in the study exhibited higher levels in cluster 2 compared to cluster 1 (Figure 3D,E). Furthermore, we assessed the survival outcomes between the two clusters and found that cluster 2 exhibited worse overall survival, as well as reduced Disease‐Free Survival and Disease‐Specific Survival, compared to cluster 1 (Figure 3F–H). To explore the underlying reasons for the differences in prognosis among OV patients in these two clusters, we performed KEGG enrichment analysis (Figure 3I,J). The ribosome, neutrophil extracellular trap formation, PPAR signaling pathway, and oxidative phosphorylation pathway were significantly enriched in cluster 1, while the PI3K−Akt signaling pathway, MAPK signaling pathway, TNF signaling pathway, and PD−L1 expression and PD−1 checkpoint pathway were significantly enriched in cluster 2.

FIGURE 3.

Subgroup typing of angiogenesis and stemness‐related genes. (A, B) Heatmaps and CDF curves for cluster analysis (C) Sample distribution between the two clusters. (D, E) Differences in expression of angiogenesis and stemness‐related genes between the two clusters. (F) KM curves of overall survival between the two clusters. (G) KM curves of disease‐free survival between the two clusters. (H) KM curves of disease specific survival between the two clusters. (I, J) Enrichment analysis of upregulated genes between the two clusters.

3.4. Analysis of angiogenesis and stem cell‐related genes with immune infiltration in OV

xCell was utilized to evaluate multiple immune cell types in order to analyze the relationship between angiogenesis, stemness‐related genes, and immune infiltration in OV. Significant differences were observed in various immune cell types between the two clusters, including CD8+ naive T cells, common lymphoid progenitors, natural killer (NK) T cells, stroma scores, M1 macrophages, monocytes, microenvironment scores, activated myeloid dendritic cells, CD4+ memory T cells, gamma delta T cells, and CD4+ naive T cells (Figure 4A). The proportion of immune cell infiltration in TCGA‐OV samples was visually demonstrated (Figure 4B). Additionally, we analyzed the expression variation of immune checkpoint‐related genes between the two clusters and found significant differences in all immune checkpoint‐related genes except for IGSF8 (Figure 4C). These results indicate a significant association between immune infiltration in OV and angiogenesis‐ and stemness‐related genes. TIDE employs a gene expression marker set to assess two mechanisms of tumor immune evasion: cytotoxic T lymphocyte (CTL) dysfunction within the tumor infiltrate and CTL rejection driven by immunosuppressive factors. High TIDE scores correlate with poor efficacy of ICB and shorter survival following ICB treatment. We found that the TIDE score in cluster 2 was significantly higher than that in cluster 1, indicating that patients in cluster 2 experienced shorter survival after ICB treatment compared to those in cluster 1 (Figure 4D).

FIGURE 4.

Angiogenesis and stemness‐related genes are significantly associated with the immune microenvironment in OV. (A) Heatmap of immune cell scores. (B) Percentage abundance of tumor‐infiltrating immune cells per sample. (C) Heatmap of immune checkpoint‐related gene expression. (D) Upper panel: statistical table of immune response of samples in different groups in the predicted results; lower panel: distribution of immune response scores in different groups in the predicted results.

3.5. Prognostic modeling based on angiogenesis and stemness‐related genes

To identify the most relevant prognostic genes associated with angiogenesis and stemness in OV, we applied a narrow filter to a set of 12 selected genes using the LASSO algorithm. The resulting prognostic model included ANGPTL4, COL8A2, CX3CR1, ERBB2, JUP, PI3, and TNFSF11 (Figure 5A,B). We visualized the relationship between risk score, survival time, survival status, and the expression levels of ANGPTL4, COL8A2, CX3CR1, ERBB2, JUP, PI3, and TNFSF11 in the TCGA‐BLCA dataset using a heat map (Figure 5C). Compared to patients in the low‐risk group, those in the high‐risk group exhibited a significantly worse overall survival rate (Figure 5D). Additionally, we assessed the predictive capability of this model for 3‐year, 4‐year, and 5‐year survival outcomes of OV patients, yielding AUC values of 0.642, 0.652, and 0.659, respectively. Furthermore, we validated the model with the GSE14764 dataset, and the results were consistent with those obtained from the TCGA‐BLCA dataset (Figure 5F–H).

FIGURE 5.

Construction and validation of prognostic models for angiogenesis and stemness. (A, B) LASSO algorithm identified the most robust prognostic genes. (C–E) Construction of prognostic models related to angiogenesis and stemness in the TCGA‐OV dataset. (F–H) Construction of angiogenesis and dryness‐related prognostic models in the GSE14764 dataset.

3.6. Analysis of prognostic models correlating with immune infiltration in OV

We investigated the relationship between a prognostic model that combines angiogenesis and stemness, developed using the LASSO algorithm, and the immune microenvironment of OV. Our results revealed significant differences in the expression of immune‐related checkpoint genes between high‐risk and low‐risk groups. Specifically, there was no significant difference in the expression of ITPRIPL1 and IGSF8 between these groups, while the remaining immune‐related checkpoint genes exhibited higher expression levels in the high‐risk group (Figure 6A). Additionally, the TIDE score was significantly elevated in the high‐risk group compared to the low‐risk group, suggesting that patients in the high‐risk group experienced shorter survival after immune checkpoint blockade (ICB) treatment (Figure 6B). Furthermore, we analyzed the correlation between risk scores and immune infiltration‐related cells using the xCell method (Figure 6C).

FIGURE 6.

Prognostic models were significantly associated with the immune microenvironment of OV. (A) Differential expression of immune checkpoint‐related genes in high‐ and low‐risk groups. (B) TIDE‐based algorithm predicts responsiveness of high and low risk groups to predicted immune checkpoint inhibitors. (C) Analysis of risk score correlation with immune infiltration‐associated cells based on the xCELL method.

3.7. TNFSF11 as the best angiogenesis and stemness‐related prognostic gene in OV

Based on the previously mentioned subgroup typing and constructed prognostic model analysis, we conclude that genes associated with angiogenesis and stemness are significantly correlated with the prognosis and immune infiltration treatment of OV patients. Further screening of the six genes included in the prognostic model, alongside clinical factors, is planned. Initially, we ranked the similarity of the seven prognostic genes using Friend analysis, which revealed that TNFSF11 is the most critical gene among them (Figure 7A). Subsequently, we performed univariate and multivariate Cox regression analyses on the seven genes in the prognostic model, as well as on patient age, TNM stage, and grading factors using data from the TCGA‐OV database (Figure 7B,C). The results of the multivariate Cox regression identified CX3CR1, PI3, TNFSF11, and patient age as feasible prognostic biomarkers for OV (Figure 7D). Finally, we plotted a calibration curve based on the results of the multivariate Cox regression, as depicted in Figure 7E. This curve illustrates the relationship between the predicted risk and the observed risk of the outcome variable, serving as a tool to evaluate the calibration performance of the model. In conjunction with the results from the Friend analysis, we identified TNFSF11 as the most crucial gene for the prognosis of OV patients among those related to angiogenesis and stemness.

FIGURE 7.

Cox regression analysis of angiogenesis and dryness‐related prognostic genes. (A) Friend analyzes similarities in angiogenesis and stemness genes. (B, C) Cox regression analysis of the prognostic value of angiogenesis‐ and dryness‐related genes. (D) Column line plot based on multifactorial Cox results. (E) Calibration curves plotted based on multifactorial Cox results.

3.8. Functional analysis of TNFSF11 in OV

We first conducted a gene enrichment analysis of TNFSF11 within the TCGA‐OV dataset, revealing a significant association with the pathway of immunoregulatory interactions between lymphoid and non‐lymphoid cells, which includes antigen activation of the B cell receptor (BCR), leading to the generation of second messengers, angiogenesis, and DNA methylation (Figure 8A). Subsequently, we utilized the SMART database to investigate the role of TNFSF11 methylation in OV. This analysis encompassed the examination of the chromosomal distribution of methylation probes associated with TNFSF11, alongside a comprehensive exploration of genomic information related to TNFSF11 (Figure 8B,C). We also assessed the correlation of all methylation probes with TNFSF11, finding that cg18090487 exhibited a positive correlation with TNFSF11 (Figure 8D). Furthermore, we analyzed the relationship between the angiogenesis pathway and TNFSF11, confirming that TNFSF11 is positively correlated with the angiogenesis pathway, which underscores its significant regulatory role in angiogenesis (Figure 8E). Finally, we investigated the binding affinity between TNFSF11 and angiogenesis‐targeted drugs through molecular docking, revealing that both drugs included in the analysis displayed strong binding capabilities to TNFSF11 (Figure 8F–H).

FIGURE 8.

TNFSF11 plays an important role in OV. (A) Gene enrichment analysis of TNFSF11 in OV. (B) Methylation probes linked to TNFSF11 and their chromosomal distribution. (C) Detailed genomic information of TNFSF11. (D) Correlation of the expression of different methylated probes with TNFSF11 expression. (E) Analysis of the correlation between TNFSF11 and angiogenesis. (F–H) Molecular docking of TNFSF11 with angiogenesis‐related drugs.

3.9. TNFSF11 is strongly associated with immune infiltration in OV

The correlation between TNFSF11 and immune checkpoint‐related genes was meticulously investigated. Our results indicated significant differences in the expression levels of HAVCR2, PDCD1LG2, CTLA4, TIGIT, and ITPRIPL1 between high‐ and low‐TNFSF11 expression subclasses (Figure 9A). Notably, the low TNFSF11 expression group exhibited a significantly lower TIDE score compared to the high TNFSF11 expression group, suggesting that patients with high TNFSF11 expression experience shorter survival following immune checkpoint blockade (ICB) treatment (Figure 9B). Subsequently, we employed the xCELL method to analyze the relationship between TNFSF11 and immune infiltration‐related cells in OV (Figure 9C). Finally, we presented a heatmap illustrating the differences in TNFSF11 expression and immune cell infiltration (Figure 9D).

FIGURE 9.

TNFSF11 and OV immune microenvironment and chemotherapy sensitivity analysis. (A) Difference in expression of immune checkpoint‐related genes in high TNFSF11‐expressing and low TNFSF11‐expressing groups. (B) Prediction of responsiveness to predicted immune checkpoint inhibitors in high and low TNFSF11 expression groups based on the TIDE algorithm. (C, D) Analysis of TNFSF11 correlation with immune infiltration‐associated cells based on the xCELL method.

3.10. TNFSF11 is elevated in ovarian cancer cell lines and correlates with poor prognosis

Firstly, we detected the transcript levels of ANGPTL4, COL8A2, CX3CR1, ERBB2, JUP, and TNFSF11 in A2780 and HEY cell lines by PCR. TNFSF11 transcription was significantly upregulated in the A2780 and HEY cell lines compared to the ovarian normal epithelial cell line IOSE80 (Figure 10A,B).

FIGURE 10.

TNFSF11 was upregulated in ovarian cancer cell lines. (A, B) PCR was performed to detect the transcript levels of ANGPTL4, COL8A2, CX3CR1, ERBB2, JUP, and TNFSF11 in A2780 and HEY cell lines. (C–F) Western‐blot detection of TNFSF11 protein concentration in A2780 and HEY cell lines. (G, H) Transcript levels of TNFSF11 in SK‐OV‐3 and CoC1 cell lines compared to IOSE80 cell line. (I) Ovarian cancer patients with high expression of TNFSF11 have a poor prognosis.

The expression levels of TNFSF11 protein in IOSE80, A2780 and HEY cell lines were detected by western‐blot (Figure 10C–F). TNFSF11 protein levels were significantly upregulated in ovarian cancer cell lines A2780 and HEY. TNFSF11 protein levels were significantly upregulated in ovarian cancer cell lines A2780 and HEY. We also validated TNFSF11 transcript levels in two other ovarian cancer cell lines, SK‐OV‐3 and CoC1, and TNFSF11 transcription was upregulated in SK‐OV‐3 and CoC1 cell lines (Figure 10G,H). Results from the GEPIA2 database suggest that ovarian cancer patients with high TNFSF11 expression are associated with a worse prognosis (Figure 10I).

After clarifying that TNFSF11 expression is upregulated in ovarian cancer cell lines, we examined the efficiency of small interfering RNAs to inhibit TNFSF11 in A2780 and HEY cell lines (si‐TNFSF11_1 and si‐TNFSF11_2). si‐TNFSF11_2 showed the highest inhibition efficiency in A2780 and HEY cell lines (Figure 11A,B,D,E). For this reason, we selected si‐TNFSF11_2 for subsequent studies. Inhibition of TNFSF11 in A2780 and HEY cell lines resulted in a significant decrease in cell viability (Figure 11C,F), a significant decrease in proliferation (Figure 11G,H), and a decrease in the ability of inducing endothelial cells’ tube formation in vitro (Figure 11I,J). In contrast, after overexpressing TNFSF11 in the human ovarian normal epithelial cell line IOSE80, the viability and proliferative capacity of the cell line were significantly upregulated (Figure 11K–N).

FIGURE 11.

TNFSF11 promotes a malignant phenotype in ovarian cancer cell lines. (A, B) PCR and western‐blot were used to detect the inhibition efficiency of small interfering RNA in A2780 cell line. (C) Alterations in cell viability following inhibition of TNFSF11 expression in the A2780 cell line. (D, E) PCR and western‐blot were used to detect the inhibition efficiency of small interfering RNA in HEY cell line. (F) Alterations in cell viability following inhibition of TNFSF11 expression in the HEY cell line. (G, H) Alterations in cell proliferative capacity after inhibition of TNFSF11 in A2780 and HEY cell lines. (I, J) Altered ability to promote angiogenesis after inhibition of TNFSF11 in A2780 and HEY cell lines. (K) Validation of the efficiency of TNFSF11 overexpression plasmid in IOSE80. (L–N) Alterations in cell viability and cell proliferation capacity after overexpression of TNFSF11 in IOSE80.

3.11. TNFSF11 activates tumor‐associated signaling pathways

We examined alterations in tumor‐associated signaling pathways following inhibition of TNFSF11 transcription in the A2780 and HEY cell lines. Upon inhibition of TNFSF11 transcription in ovarian cancer cell lines, phosphorylation levels of AKT, ERK1/2, CREB, and STAT6 were significantly suppressed, suggesting a promotional role of TNFSF11 for tumor‐associated signaling pathways (Figure 12).

FIGURE 12.

TNFSF11 activates tumor‐associated signaling pathways. (A–H) Alterations in phosphorylation levels of AKT, ERK1/2, CREB, and STAT6 following inhibition of TNFSF11 expression in the A2780 cell line. (I–P) Alterations in phosphorylation levels of AKT, ERK1/2, CREB, and STAT6 following inhibition of TNFSF11 expression in the HEY cell line.

4. DISCUSSION

OV is the second most common malignancy in women's health and the leading cause of gynecological cancer deaths. ¹⁹ , ²⁰ Uncontrolled tumor growth necessitates a continuous supply of oxygen and nutrients. ²¹ Tumor development is heavily reliant on the vasculature, which serves as a crucial transport pathway for essential nutrients. However, tumor blood vessels differ structurally and functionally from normal tissue blood vessels, exhibiting characteristics such as pericyte relaxation, distorted distribution, and excessive expansion. These abnormal blood vessels can create a hypoxic microenvironment within the tumor, impeding drug penetration. ²² In this context, anti‐angiogenic therapies designed to ‘starve’ tumor cells have been developed and implemented in the treatment of various tumors. ²³ , ²⁴ CSCs possess the ability to self‐renew within the tumor microenvironment. Through self‐renewal, CSCs facilitate tumor growth by generating new CSCs. Moreover, differentiation leads to the production of non‐CSC tumor cells, contributing to tumor heterogeneity and the stratified organization of cells within the tumor. ²⁵ The relationship between CSCs and angiogenesis is complex and bidirectional. Recent studies have shown that CSCs can influence angiogenesis. For instance, CSCs have been found to secrete angiogenic factors that promote the formation of new blood vessels, thereby enhancing their own survival and proliferation within the tumor microenvironment. ²⁶ This interaction is particularly evident in the context of hypoxia, a common feature of solid tumors, which can induce the expression of hypoxia‐inducible factors (HIFs) that promote both angiogenesis and the maintenance of stemness in CSCs. Moreover, the presence of CSCs can alter the characteristics of the tumor vasculature. For example, CSCs can differentiate into endothelial cells, contributing directly to the formation of new blood vessels, a process known as vasculogenic mimicry. ²⁷ This phenomenon complicates the traditional understanding of angiogenesis, as it suggests that tumor cells themselves can participate in creating their vascular supply, further supporting their growth and metastasis. On the other hand, the tumor microenvironment, shaped by angiogenesis, can also influence the behavior of CSCs. The newly formed blood vessels can create niches that support CSC maintenance and proliferation. For instance, the interaction between endothelial cells and CSCs can enhance the stem‐like properties of these cells, promoting their survival and resistance to therapies. ²⁸ In addition, the interplay between angiogenesis and CSCs is implicated in the mechanisms of drug resistance. Anti‐angiogenic therapies, which aim to inhibit blood vessel formation, have shown limited success in clinical settings, partly because they can inadvertently promote the survival of CSCs by depriving the tumor of oxygen and nutrients, leading to a more aggressive tumor phenotype. ²⁹ A comprehensive investigation of the interplay between OV angiogenesis and stemness characteristics is critical for improving patient outcomes in cancer therapy.

In our study, we initially analyzed the TCGA‐OV dataset and the GSE66957 dataset to identify angiogenesis‐ and stemness‐related genes with differential expression and prognostic significance. We utilized the GSCALite website as a tool to assess the importance of the genome across various tumors. ³⁰ Our functional analysis revealed that these genes were associated with several biological processes, including cell cycle inhibition, DNA damage response, epithelial‐mesenchymal transition (EMT) activation, androgen receptor inhibition, and RAS/MAPK pathway activation. We categorized the TCGA‐OV samples into two groups based on angiogenesis and stemness‐related gene expression, revealing distinct pathogenic mechanisms and clinical prognostic characteristics between the two groups. Compared to cluster 1, patients in cluster 2 exhibited a poorer prognosis regarding overall survival, disease‐specific survival, and disease‐free survival. The MAPK, PI3K‐Akt, TNF, and NF‐kappaB signaling pathways may contribute to the observed prognosis in the deteriorating cluster 2, as these pathways are well‐established regulators of tumor angiogenesis and stemness. ³¹ , ³² , ³³ Furthermore, we noted significant differences in immune infiltration and immunotherapy response between cluster 1 and cluster 2. The LASSO method is a widely utilized machine learning technique for constructing prognostic models, particularly in the context of the TCGA‐OV dataset. The efficacy of this model was validated using the GSE14764 dataset. Additionally, we examined the association between the OV prognostic model and immunotherapy. Through multivariate Cox regression analysis and Friends analysis, TNFSF11 emerged as the most promising prognostic gene. Ultimately, we established a significant correlation between TNFSF11 and both OV immunotherapy and angiogenesis‐targeted therapies. Nevertheless, our research has certain limitations. In future studies, we will further investigate the specific mechanism of TNFSF11 in ovarian cancer, including its role within the tumor microenvironment and its impact on the characteristics of cancer stem cells. Additionally, future research may focus on the clinical application of TNFSF11‐targeted therapy and assess its combined effects with existing treatments, such as chemotherapy and immunotherapy, to facilitate personalized treatment.

5. CONCLUSION

In this study, we emphasize the significance of angiogenesis‐ and stemness‐related genes in predicting OV prognosis. Additionally, we examined their roles in immune infiltration and immunotherapy. Among these genes, TNFSF11 demonstrated the most promising results as a prognostic marker. Notably, TNFSF11 exhibits a high affinity for angiogenesis‐targeted drugs, highlighting its potential as a novel targeted therapeutic option for OV.

FUNDING INFORMATION

This work was funded by General Project of Jiangsu Provincial Health Commission (H2023090) and General Project of Nantong Municipal Health Commission (MS2023081).

CONFLICT OF INTEREST STATEMENT

The authors declare that the research was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest.

Zhou L, Min Y, Cao Q, Tan X, Cui Y, Wang J. Comprehensive analysis of the value of angiogenesis and stemness‐related genes in the prognosis and immunotherapy of ovarian cancer. BioFactors. 2025;51(1):e2155. 10.1002/biof.2155

DATA AVAILABILITY STATEMENT

The datasets obtained from TCGA database (https://portal.gdc.cancer.gov/) and Genecards database (https://www.genecards.org), GEO database (https://www.ncbi.nlm.nih.gov/geo/), partial analysis by GSCALite website (http://bioinfo.life.hust.edu.cn/web/GSCALite/).