Endometrial cancer (EC) was the sixth most common cancer diagnosed in women in 2020, with a total of 417,000 new cases documented worldwide. The median age of diagnosis of EC patients is 61 years, and the lifetime risk of EC is around 3% (1). EC is associated with a high mortality rate, and over 76,000 women die annually as a result of EC (2). The mortality rate due to EC further increases with advanced tumor stage, invasive histology, and metastasis (3). There is currently a paucity of biomarkers or models that can effectively predict the prognosis and survival of EC patients (4). Therefore, it is crucial to elucidate the molecular mechanisms underlying EC occurrence and progression in order to identify novel prognostic markers and therapeutic targets.

Cell death plays a crucial role in several biological processes, and can be classified as programmed cell death (PCD) and accidental cell death (ACD) (5). PCD is the culmination of ordered, gene-controlled pathways following the spontaneous loss of cellular function, whereas ACD is an uncontrolled process that is triggered in response to certain harmful stimuli. PCD is primarily responsible for maintaining intracellular homeostasis (6). Various types of PCD have been documented so far, including autophagy-dependent cell death, necroptosis, apoptosis, ferroptosis, pyroptosis, entosis, parthanatos, NETosis, alkaliptosis, lysosome-dependent cell death (LCD), and oxeiptosis (7).

Lysosomal degradation during autophagy-mediated cell death facilitates metabolic adaptation and nutrient recycling (8). Necrosis has long been considered an involuntary form of cell death, although recent evidence indicates that necrosis can be initiated and sustained, resulting in necroptosis that centers around the formation of necrosomes (9, 10). Apoptosis is an intrinsic mechanism for eliminating damaged cells and involves a series of events including condensation, nucleolysis, and nuclear fragmentation, which culminate in the engulfment of apoptotic vesicles by macrophages (11). Ferroptosis is characterized by the accumulation of lipid hydroperoxides in an iron-dependent manner that ultimately reaches a lethal threshold (12, 13). Cuproptosis is a recently identified mode of PCD that is triggered by copper imbalance and is closely associated with disease progression (14). Pyroptosis is an inflammatory form of PCD that is typified by the creation of membrane pores that compromise cellular integrity, eventually leading to cell rupture (15). Entosis is a form of cell “cannibalism” wherein one live cell is engulfed and lysed by another cell without the activation of the apoptotic pathway (16). Parthanatos is induced by excessive activation of the nuclease PARP-1 (17), an RNA polymerase (RP) that interacts with and activates DNA or RNA, leading to replication or repair of damaged DNA and proteins. NETosis is initiated by the release of neutrophil extracellular traps (NETs), i.e., interconnected structures that cells release in response to infection or injury (18). Alkaliptosis, a newly recognized type of PCD, heavily relies on the intracellular alkalization process (19). LCD is dependent on hydrolase, which facilitates lysosomal transport to the cytoplasm by means of membrane penetration. This process is regulated by intracellular signaling systems and membrane proteins (20). KEAP1, a detector of reactive oxygen species (ROS), has recently been shown to be involved in a unique type of cell death known as oxygen apoptosis (21).

Mutations in the PCD pathways have been detected in the early stages of cancer, which endow the tumor cells with resistance to anti-cancer treatments (7). Therefore, targeted interventions that activate PCD pathways using single or combined therapies are an effective anti-cancer strategy. For instance, the FDA-approved BCL-2 inhibitor triggers apoptosis in lymphoma cells (22). In addition, activation of GSDME-mediated pyroptosis has proved to be highly effective against many cancers (23). Cancer cells can resist PD-1/PD-L1 checkpoint inhibition by blocking ferroptosis through the regulation of ferritin and other proteins (24), indicating that activation of the ferroptosis pathway in these cells can sensitize them to PD-1/PD-L1 blockers.

In this study, we used bioinformatic approaches to establish a PCD-related prognostic gene signature for EC and found that higher risk scores were associated with a worse prognosis. Our findings provide new insights into the molecular basis of EC progression. Furthermore, the genes associated with PCD have significant potential as prognostic biomarkers and therapeutic targets for EC.

Since irregular vaginal bleeding is an early symptom of EC, most patients are diagnosed at an early stage. Although early diagnosis confers some survival advantage, the annual mortality rate due to EC is still high (4, 26), which warrants the identification of novel prognostic biomarkers. The aim of this study was to identify the prognostic PCD-associated genes in EC in order to construct a predictive model to guide clinical decision-making. A total of 43 PCD-related genes were associated with the prognosis of EC patients, and two distinct clusters were identified.

The patients in Cluster B generally had higher tumor purity, higher T stage, and worse prognosis, while Cluster A was associated with higher ESTIMATE scores, ImmuneScore, and StromalScore. The ImmuneScore is a measure of the percentage of cytotoxic and memory T cells at the tumor core and tumor margin (27), and the StromalScore is indicative of the ratio of stromal cells in the tumor microenvironment (TME) (28). A higher ImmuneScore or StromalScore suggests the presence of more immune substances or matrix components in the TME. Thus, Cluster B was associated with a general decrease in stromal and immune cells and an overall increase in tumor purity compared to Cluster A, indicating that immune infiltration and tumor purity are determinants of EC development and that enhancing the degree of immune infiltration may slow tumor growth. Furthermore, patients in Cluster A generally expressed genes characteristic of immune cell subsets, while patients in Cluster B showed high expression levels of STAT1 and interferon genes, which suggests an immune active state corresponding to favorable prognosis in Cluster A. Consistent with this, Cluster A showed a significant increase in the infiltration of T cells (such as CD8⁺ T cells, cytolytic activity, inflammation-promoting, and T cell co-stimulation) compared to Cluster B. On the other hand, Cluster B showed a significant increase in aDCs, para-inflammation, and type I IFN response compared to Cluster A. Furthermore, WGCNA indicated that the brown module has the strongest correlation with the clusters, and the genes in the brown module are upregulated in Cluster B. The WGCNA algorithm is a powerful tool for identifying co-expressed gene modules and their relationships with phenotypic traits. One of the main advantages of WGCNA is that it can handle large-scale gene expression data sets and identify biologically meaningful gene modules that are associated with specific traits or conditions. Additionally, WGCNA can be used to identify key hub genes that play important roles in regulating gene expression networks. Overall, WGCNA is a valuable tool for understanding the complex relationships between genes and phenotypes. Thus, the brown module was thought to be reliable. In addition, GO analysis showed that the 43 prognostic PCD-related genes were negatively regulated in the pathway of “lipids and atherosclerosis”. Based on these findings, we hypothesize overexpression of PCD-related genes may allow cancer cells to proliferate rapidly by accelerating lipid metabolism.

A LASSO-based PCD-related gene signature comprising 11 genes, including GZMA, ASNS, GLS, PRKAA2, VLDLR, PRDX6, PSAT1, CDKN2A, SIRT3, TNFRSF1A, and LRPPRC, was developed to predict the OS of EC patients. The LASSO algorithm is a type of linear regression that is used for feature selection and regularization. One of the main advantages of the LASSO algorithm is that it can help prevent overfitting by shrinking the coefficients of less important variables to zero. This can lead to a more parsimonious model that is easier to interpret and less prone to errors. Additionally, the LASSO algorithm can handle high-dimensional data sets with many variables, which can be useful in fields like genetics and finance. The PCD-related gene signature showed good performance in the TCGA-UCEC cohort and two external validation cohorts, TCGA-CESC and TCGA-BRCA. This finding strongly supported the accuracy, sensitivity, and specificity of the PCD-related gene signature in predicting prognosis. GZMA is predominantly expressed in the cytosolic granules of NK cells and cytotoxic T-cells. It cleaves gasdermin-B (GSDMB), which releases the pore-forming moiety of GSDMB and triggers pyroptosis (29–31). Asparagine synthetase (ASNS) catalyzes the de novo synthesis of asparagine by transferring amino groups from glutamine to aspartic acid. Inhibiting ASNS expression in cancer cells impairs nutrient uptake and promotes apoptosis (32). GLS hydrolyzes glutamine to produce glutamate (33), and its inhibition can induce apoptosis in tumor cells (34). PRKAA2 encodes the catalytic subunit of AMPK, a key enzyme that senses cellular energy status (35). VLDLR is a cell surface receptor with multiple functions, such as binding to very low-density lipoprotein and facilitating its endocytosis, which contributes to energy metabolism (36). PRDX6 is a mercaptan-specific peroxidase that reduces hydrogen peroxide and organic hydroperoxide to water and alcohol respectively (37, 38), and can attenuate apoptosis induced by oxygen-glucose deprivation/reoxygenation (39). PSAT1 is a member of the V-class pyridoxal phosphate ester-dependent transaminase family that fuels tumor cells by generating serine. Inhibition of PSAT1 expression can suppress serine synthesis in tumors, thereby inhibiting their growth (40). CDKN2A prevents MDM2-induced degradation of p53, and promotes p53-dependent apoptosis (41). SIRT3 is an exclusive mitochondrial member of the Sirtuin family of class III histone deacetylases, similar to the yeast Sir2 protein. It can eliminate reactive oxygen species, prevent malignant transformation, and inhibit apoptosis (42). TNFRSF1A belongs to the TNF receptor superfamily of proteins that plays a role in TNFα-mediated cell apoptosis and necrosis (7). LRPPRC is a mitochondrial protein that regulates RNA metabolism and transcription. Loss of LRPPRC affects the electron transport chain in the mitochondria, which increases mitochondrial permeability and generation of reactive oxygen species (43). LRPPRC, CDKN2A, PSAT1, PRDX6, PRKAA2, GLS, and ASNS were highly expressed in the high-risk group, while TNFRSF1A, SIRT3, VLDLR, and GZMA showed high expression in the low-risk group. Consistent with our findings, Tian et al. reported high expression of LRPPRC in EC tissues (44). Upregulation of CDKN2A in the extracellular matrix can promote EC progression by releasing cytokines and proteases in the TME (45, 46). Roh et al. demonstrated that silencing EZH2 in EC cells inhibited PRDX6, leading to the activation of the exogenous homocysteine pathway and eventually cell death (47). Zhou et al. found that estrogen activates Grn metabolism in estrogen-sensitive EC, depending on the up-regulation of GLS (48). However, the functions of LRPPRC, PSAT1, PRKAA2, ASNS, TNFRSF1A, SIRT3, VLDLR, and GZMA in EC progression, and the underlying mechanisms, remain to be elucidated.

The PCD risk score was also calculated across 32 types of cancer and showed marked organ specificity. For instance, uterine carcinosarcoma, brain lower-grade glioma, and ovarian serous cystadenocarcinoma had higher PCD risk scores, while uveal melanoma, mesothelioma, and skin cutaneous melanoma usually had lower PCD risk scores. The PCD risk scores of other cancers were similarly distributed. The PCD-related genes are likely overexpressed in the cancers with higher risk scores, and relatively lowly expressed in cancers with lower scores. We found that cancers with higher mutation frequencies in PCD-related genes tended to have lower PCD risk scores. The mutations may affect the normal expression of PCD-related genes, thereby affecting the level of the risk score. Furthermore, mutations in PCD-related genes may reduce their expression in some cancers, leading to dysregulation of cell death pathways and malignant development. CDKN2A was the most frequently mutated PCD gene across all cancer types, and exhibited the most diverse mutation profile, indicating that SNVs in CDKN2A are ubiquitous in multiple cancers.

The EC patients in the high-risk group were older and have more advanced-stage tumors, which corresponded to a worse prognosis. The PCD gene signature exhibited good predictive performance for 1-, 3- and 5-year survival. Furthermore, the prognosis of high-risk patients worsened with age and tumor grade. GSEA results showed that signaling pathways related to DNA replication were abundant in the high-risk group, while the low-risk group was enriched in pathways related to chemotaxis. This indicated that the tumor cells in the high-risk population proliferate actively, while an active immune response characterizes the tumors in the low-risk population. There were remarkable discrepancies in the distribution of specimens between high-risk and low-risk groups for the four different types of immunization. The patients in the high-risk group were sensitive to AT3148, elephantirin, I-BET-762, and Niraparib, suggesting their potential clinical applicability in EC patients. We also established a nomogram consisting of age, stage, grade, and risk score, which predicted the 1-, 3- and 5-year survival rates with high accuracy.

The PCD genes also showed a significant correlation with several proliferation-related genes. WNT5A is known to regulate the proliferation, invasion, and metastasis of tumor cells. In a previous study, we found that low expression of WNT5A in gastric cancer tissues was significantly associated with the invasion and metastasis of tumor cells, and poor prognosis (49, 50). MKI67 is a typical marker of cell proliferation that remains on the single mitotic chromosome after the breakdown of the nuclear membrane. It is expressed at low levels in EC tissues (51). Mutations in the CTNNB1 gene promote the development of esophageal cancer by upregulating the Wnt/beta-catenin pathway and the downstream target genes (52). CDH1 encodes E-cadherin, an epithelial marker that regulates cell adhesion, migration, and proliferation. It is downregulated during epithelial-mesenchymal transformation (EMT), which is the driver of tumor cell metastasis (53). Low expression of E-cadherin is linked to worse prognosis and survival (54). No study so far has reported any interaction between LRPPRC and these proliferation-related genes in EC. Based on the expression patterns and clinical association of LRPPRC, this gene is likely a risk factor in EC patients and therefore a potential therapeutic target. LRPPRC showed a strong positive correlation with GLS, PRKAA2, and PSAT1. There is a possibility of a synergistic interaction between these genes in EC development, which warrants further research. Wang et al. have identified biomarkers in different tumors by combining computational biology methods such as WGCNA, opening up more possibilities for researching tumorigenesis mechanisms (55, 56). LRPPRC expression showed a positive correlation with Th2 cells and a negative correlation with NK CD56^bright cells, which may accelerate tumor progression. High expression of LRPPRC may inhibit NK cell activity, thereby suppressing the immune response and promoting cancer progression. Further studies are needed to assess the role of LRPPRC in tumorigenesis and development, and the underlying molecular mechanisms involved in programmed cell death. Single-cell sequencing (scRNA-seq) has been widely used to explore the mechanisms and biomarkers of gynecological tumors (57, 58). In our subsequent study, we will focus on the mechanism of LRPPRC in EC development through the scRNA-seq approach.

PCD-related genes are involved in the development of EC and can predict patient prognosis. We developed a PCD-related cluster system for discriminating EC patients with different prognoses. We also constructed an 11-gene PCD-related signature with high predictive performance in prognosis, mutation, and drug response. LRPPRC, an adverse prognostic gene in EC and a member of the model genes, could predict the clinical status and immune infiltration level of EC patients. Our findings provide new insights into the mechanisms underlying EC development and highlight potential therapeutic targets.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by Department of Obstetrics and Gynaecology, Guangzhou Women and Children’s Medical Center, Guangzhou Medical University. The patients/participants provided their written informed consent to participate in this study.

KG contribute to the method and the results, JX wrote the manuscript. Other authors confirmed their contribution to this research. All authors contributed to the article and approved the submitted version.