Gastric cancer (GC) represents the fifth most prevalent malignancy globally, with its mortality rate ranking fourth among all malignant tumors worldwide (1). In recent years, the ongoing advancements in tumor immunotherapy, epitomized by inhibitors targeting programmed death receptor 1 (PD-1) and its ligand PD-L1, have demonstrated remarkable efficacy across various solid tumors, including gastric cancer (2, 3). However, the overall effective rate of PD-1/PD-L1 inhibitors in the unselected population of solid tumors is less than 20% (2, 3). Consequently, an imperative need arises to devise precise and robust models to identify gastric cancer patients who are susceptible to tumor immunotherapy, thereby enabling individualized clinical interventions.

Programmed cell death (PCD), also known as regulated cell death, refers to the self-regulating process in which cells die under the control of specific genes and the precise coordination of various mechanisms. The main goal of PCD is to maintain the stability of the internal cellular environment. According to triggering stress, morphological characteristics, regulatory signaling pathways and effector molecules, PCD can be divided into apoptosis, necroptosis, pyroptosis, ferroptosis, entotic cell death (entosis), netotic cell death (NETosis), parthanatos, lysosome-dependent cell death (LDCD), autophagy-dependent cell death (ADCD), alkaliptosis, oxeiptosis, cuproptosis and paraptosis, immunogenic cell death (ICD) (4, 5). The activation of apoptosis mainly includes the extrinsic and the intrinsic pathway (5). Extrinsic apoptosis is mediated by the activation of plasma membrane-localized death receptors (such as TNFR1, Fas) by their cognate ligands (such as TNF, FasL) (6). Intrinsic apoptosis can be activated by BCL-2 family proteins induced mitochondrial outer membrane permeabilization by releasing cytochrome c and mediated by caspase-3/7/9 (6). Necroptosis is a form of death triggered by extracellular stimuli activating death receptors, which causes phosphorylation of receptor-interacting protein kinase (RIPK), leading to the recruitment of mixed lineage kinase domain-like (MLKL), necroptosis mainly depends on the activation of RIPK1 and RIPK3 (7). Pyroptosis is triggered by caspase-1-driven cleavage of the pore-forming protein gasdermin D (GSDMD), then the N-terminal fragment of GSDME will lead to the formation of cell membrane pores and thereby induce pyroptosis (8). Ferroptosis is an iron-dependent form of cell death characterized by the accumulation of lipid peroxides (9). The process includes iron accumulation, reactive oxygen species (ROS) activation, reduced cysteine uptake, depletion of glutathione (GSH), and activation of the mitogen-activated protein kinase (MAPK) system (10). Cuproptosis is a newly discovered type. It is related to the imbalance of intracellular copper metabolism. Excessive copper directly binds to lipoylated proteins in the tricarboxylic acid (TCA) cycle of mitochondria, leading to the abnormal aggregation of lipoylated proteins and the loss of iron-sulfur cluster proteins in respiratory chain complexes, causing a protein toxic stress response and ultimately leading to cell death (11). Entosis is triggered by autophagosomes formed by the cell membrane engulfing its cytoplasmic proteins or organelles. Its regulation mainly depends on the mTOR pathway, including signal pathways such as PI3K-AKT-mTOR and AMPK-TSC1/2-mTOR (12). LDCD is mainly achieved through changes in lysosomal membrane permeability (LMP). When LMP increases, the release of cathepsin B and cathepsin D in lysosomes will trigger lysosome-dependent cell death (6). NETosis is a form of regulated cell death (RCD) driven by neutrophil extracellular trap (NET), which is regulated by NADPH oxidase-mediated ROS production and histone citrullination (13). Parthanatos is mediated by poly (ADP-ribose) polymerase 1 (PARP1), which is caused by DNA damage. PARP1 hyper-activation stimulates apoptosis-inducing factor (AIF) nucleus translocation, and accelerates nicotinamide adenine dinucleotide (NAD+) and adenosine triphosphate (ATP) depletion, leading to DNA fragmentation (14). Alkaliptosis is a pH-dependent cell death process triggered by the small molecular compound JTC801 (15). ADCD, a phagocytic biological process, can disintegrate damaging proteins or organelles through lysosomal fusion (16). Oxeiptosis is activated in response to oxidative stress induced by ROS or ROS-generating agents and characterized by the activation of the KEAP1/PGAM5/AIFM1 signaling pathway (4). ICD is triggered by the release of damage-associated molecular patterns (DAMPs) from dying cells, which can trigger an adaptive immune response, release antigens, reverse the tumor immunosuppressive microenvironment, and improve the sensitivity of immunotherapy (17). Paraptosis is characterized by the swelling and vacuolization of the endoplasmic reticulum (ER) and mitochondria, resulting in the formation of large cytoplasmic vacuoles (4).

Increasing evidence shows that PCD plays a critical role in cancer initiation and progression (18–23). Cancer cell death has been confirmed as fundamental in the remodeling of the tumor immune microenvironment (TIME) (24). For instance, tumor cell fragments serve as antigens, which are captured, processed, and presented by conventional dendritic cells (cDCs) (25). Certain types of cell death, such as necroptosis and pyroptosis, release DAMPs and inflammatory cytokines due to cell membrane rupture (26). Conversely, some studies suggested that cell death can also directly or indirectly cause immunosuppression by recruiting myeloid cells (such as immunosuppressive macrophage subsets) (27). Considering the inherent connection between the TIME and the efficacy of immunotherapy, the recognition and induction of PCD forms to potentiate the immune response against cancer, particularly in the context of immune checkpoint inhibitors (ICIs), is of paramount importance (26). Furthermore, mounting evidence indicates that cancer patients with varying prognoses often exhibit distinct differences in TIME and their response to ICIs (28, 29).

In the past few years, many scholars have developed prediction models with characteristic genes of a single form of PCD, and have achieved moderate prediction accuracy in predicting cancer prognosis and drug resistance (30). However, interactions have been identified within signaling pathways that regulate different forms of cell death (31). While each form of cell death has its mechanism, they are not independent individuals and still have connections with each other (31). For instance, reactive oxygen species (ROS) is an indispensable component in the process of ferroptosis and can also participate in apoptosis, autophagy, necroptosis, and pyroptosis (32, 33). Lysosomal membrane permeabilization (LMP) not only participates in lysosome-dependent cell death but also amplifies cell death signals and increases the complexity of cell death under the induction of autophagy, necroptosis, and ferroptosis (16).

Therefore, a prediction model incorporating multiple forms of PCD may provide a more comprehensive representation of tumor characteristics compared to a model focusing on a single type. In this study, we collected 14 PCD pattern-related genes to identify biomarkers and establish a PCD score (PCDS) signature, aiming to predict the TIME, prognosis, and responsiveness to immunotherapy in GC. In the future, this may assist doctors in making individualized clinical treatment.

Firstly, characteristic gene sets of 14 forms of PCD including apoptosis, necroptosis, pyroptosis, ferroptosis, cuproptosis, entosis, NETosis, parthanatos, LDCD, ADCD, alkaliptosis, oxeiptosis, paraptosis, and ICD were collected. The TCGA-STAD database was selected as the training cohort, while the GSE84437 and GSE84433 databases served as the validation cohorts for the prognostic prediction model.

Integration of the results generated from these three methods identified 239 DEGs (FDR q value < 0.05, |log2FC| > 1), comprising 97 upregulated and 142 downregulated genes between tumor and normal samples ( Figure 1A , Supplementary Table 3 ). The heatmap of DEGs was shown in Figure 1B . Besides, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment analysis revealed that these PCD-related DEGs were involved in multiple biological pathways such as regulation of IL-17, JAK-STAT and p53 signaling pathways, etc ( Figures 1C, D ). Additionally, we assessed the mutations for PCD-related genes, revealing that approximately 93.17% (409/439) of GC patients exhibited mutations, predominantly missense mutations ( Figure 1E ). Notably, among the top 20 mutated PCD-related genes, TNN and TP53 exhibited mutation frequencies exceeding 30%. Analysis of CNV status indicated that PCD-related genes frequently underwent alterations. It was noted that the CNV deletion of SLC25A4 was the most extensive, while the copy number amplification of GSDMC was the most significant ( Figure 1F ).

This investigation presents an extensive preliminary analysis of 14 different PCD patterns in GC. The study involved the construction of a PCDS signature within the TCGA-STAD cohort, subsequently validated by the GSE84437 and GSE84433 cohorts, affirming its robust efficacy. A nomogram model incorporating clinical characteristics and PCDS was developed, yielding promising outcomes for predicting OS for GC patients. Furthermore, the study examined potential associations between PCDS and TIME via various methodologies, suggesting implications for immunotherapy strategies in GC. Additionally, the correlation between PCDS and drug responsiveness was assessed, revealing that patients with a high PCDS may exhibit resistance to immunotherapy and standard adjuvant chemotherapy regimens; however, they may benefit from drugs such as NU7441, AZD8055, Dasatinib, JAK-, BMS-754807, and JQ1, with “Dasatinib” being an FDA-approved medication. The development of this comprehensive PCDS signature enhances the understanding of the intricate biological processes underlying GC. It provides a valuable tool for evaluating the prognosis of GC patients and guiding treatment decisions. Integrating PCDS into prognostic models offers promising prospects for personalized medicine, enabling the development of tailored treatment strategies for individual patients.

According to reports, SERPINE promotes malignant progression and correlates with poor prognosis in GC (48). Teng et al. found that the NKX2-1-AS1/miR-145-5p axis induces SERPINE1 translation, thus activating the VEGFR-2 signaling pathway to promote tumor progression and angiogenesis in GC (49). PLPPR4, also named LPPR4, Zhang et al. found that LPPR4 could promote the migration, invasion and adhesion of GC cells to facilitate peritoneal metastasis through the Sp1/integrin α/FAK pathway (50). CDO1 possesses functionally oncogenic aspects through modification of mitochondrial membrane potential (51). Mouse experiments have revealed that inhibiting CDO1 production facilitates ferroptosis by increasing oxidative stress and inhibiting GPX4 production (52). Harada et al.’s study suggested that abnormal CDO1 expression in GC may indicate distant metastatic ability (53). Current investigations into the mechanistic role of MID2 in GC remain limited. Nonetheless, MID2 inhibition could largely abrogate MORC4-induced drug resistance to adriamycin, 5-fluorouracil, and cisplatin in breast cancer (54). Abnormal NOX4 expression results in the production of ROS, contributing to various oncogenic processes (55). Tang et al. found that NOX4 promotes GC cell growth and apoptosis through the generation of ROS and subsequent activation of GLI1 signaling (56). DYNC1I1, as an important binding subunit of cytoplasmic dynein, primarily participates in cell cycle regulation (57). Gong et al. revealed that DYNC1I1 could upregulate IL-6 expression by increasing NF-κB nuclear translocation, and then trigger the DYNC1I1-driven IL-6/STAT pathway to promote GC proliferation and migration (47). High PDK4 expression is closely related to poor prognosis, and might participate in the proliferation, migration and invasion of GC cells by modulating the glycolysis level in GC cells (58). Miao et al. discovered that miR-5683 represses GC glycolysis and progression through targeting PDK4 (59). Furthermore, high MYB expression is positively associated with activated CD4+ T cell infiltration and poor prognosis in GC (60). Yan et al. demonstrated that SNHG3 binds and sequesters miR-139-5p, which can indirectly promote the upregulation of the miR-139-5p target gene MYB and drive the proliferation, migration, and invasion in GC (61). Additionally, high TUBB2A expression is linked to reduced immune cell infiltration and poor prognosis in triple-negative breast cancer (62).

According to prior research, the tumor microenvironment plays crucial roles in tumor initiation, progression, metastasis, and response to therapies. Moreover, tumor cells can survive because the tumor microenvironment allows them to evade immune surveillance and drug interference (63, 64). In this study, PCDS was correlated with the infiltration abundance of stromal and immune cells, notably M2 macrophages, Tregs, CAFs, and endothelial cells. Among these, M2 macrophages, also referred to as tumor-associated macrophages, facilitate tumor progression by fostering cancer invasion and metastasis, promoting neovascularization, and contributing to the development of an immunosuppressive TME (65). Endothelial cells mainly provide nutrition for tumor development. By responding to angiogenic factors such as VEGF, they promote new blood vessel formation and provide oxygen and nutrition for tumors, and playing a key role in the angiogenesis of gastric cancer (66). CAFs induce hypoxia in the tumor microenvironment, leading to ECM hardening and degradation, which in turn affects tumor cell proliferation, migration and invasion as well as angiogenesis (67). Tregs can be divided into two types: natural regulatory T cells (nTregs) and induced regulatory T cells (iTregs) (68). nTregs originate from the thymus and play a role in mediating immune tolerance through transcription factors such as nuclear factor κB (NF-κB), while iTregs develop in the peripheral environment and are stimulated by inhibitory cytokines IL-2 and TGF-β in the tumor microenvironment, which in turn helps GC cells evade immune surveillance (69). Tregs regulate immune cell activity in the tumor microenvironment, suppressing cytotoxic T and natural killer cells, reducing immune responses, enabling tumor cells to evade immune surveillance, and fostering a tumor growth-permissive environment (70). Increasing evidence indicates that high TMB can send signals to activate immune responses, thereby making tumors more sensitive to immunotherapy (71). GC patients with high TMB exhibit superior OS compared to those with low TMB (72). These findings align with our results. Our results showed that TMB was significantly negatively correlated with PCDS. In terms of molecular pathways, compared with the low PCDS group, cancer-related pathways such as Wnt/beta-catenin and cell adhesion molecules were overactivated in the high PCDS group. Therefore, immunotherapy may be an effective treatment approach for patients with low PCDS, while those with high PCDS may not benefit as much.

Drug sensitivity analysis indicates that GC patients with high PCDS may exhibit resistance to immunotherapy and standard adjuvant chemotherapy regimen. Notably, PCDS showed a significant negative correlation with the IC50 values of NU7441, AZD8055, dasatinib, JAK-8517, BMS-754807, and JQ1, implying these drugs may have potential benefits for GC patients with high PCDS. BMS-754807, a selective IGF-1R inhibitor, exhibits potent inhibitory effects on GC cells (73). Another study also demonstrated the activation of the IGF1/IGF1R pathway in mesenchymal gastric tumors, which showed sensitivity to another selective IGF-1R inhibitor, Linsitinib (OSI-906) (74). Likewise, mTOR inhibitors such as 2,6-DMBQ (AZD8055) have also been previously reported for their inhibitory efficacy in GC (75). AZD8055 inhibits the phosphorylation of mTORC1 substrates p70S6K and 4E-BP1 as well as phosphorylation of the mTORC2 substrate AKT and downstream proteins, thereby leading to tumor growth inhibition (76). Dasatinib plays a synergistic role with oxaliplatin in inhibiting gastric cancer cell growth both in vitro and in vivo, via inhibiting Src activity stimulated by oxaliplatin (77). Wang Shi et al. found dasatinib also showed potential in sensitizing cancer cells to cisplatin, and the PI3K/AKT pathway was involved in the anti-cancer effect of dasatinib or combined with cisplatin (78). In addition, dasatinib is FDA-approved drug. BET protein inhibitor JQ1 downregulates chromatin accessibility and suppresses metastasis of gastric cancer via inactivating RUNX2/NID1 signaling (79). Furthermore, JQ1 augments the antitumor efficacy of abemaciclib (ABE) in preclinical models of gastric carcinoma. Mechanistically, the combination of ABE and JQ1 enhances the cell cycle arrest of GC cells and induces unique characteristics of cellular senescence through the induction of DNA damage (80). NU7441, a DNA−PKcs inhibitor, increases the sensitivity of GC cells to radiotherapy (81). This inhibitor increases the sensitivity of radioresistant BGC823 and MGC803 cells to radiotherapy through the cleaved−caspase3/γH2AX signaling pathway, thus presenting a potential treatment method for GC (81). JAK_8517 is a small molecule inhibitor that targets Janus kinase (JAK), which is involved in cell signaling. Thus, GC patients with high PCDS patients may benefit from the above six candidate drugs, especially the FDA-approved drug dasatinib.

Although our model has demonstrated excellent performance in both the training and validation cohorts, it is important to acknowledge certain limitations. First, retrospective recruitment of patients may introduce some inherent biases to a certain extent. Second, more experiments are needed. Therefore, additional validation through high-quality, multicenter randomized controlled trials with sufficient sample sizes and adequate follow-up is necessary.

In conclusion, through a comprehensive analysis of 14 PCD pattern-related genes, a new PCDS signature has been established. This innovative signature accurately predicts the prognosis and drug sensitivity of GC. The findings indicated that PCDS can serve as a valuable tool for evaluating the prognosis and guiding immunotherapy treatment decisions for GC patients.