Disulfidptosis signature predicts immune microenvironment and prognosis of gastric cancer

Abstract

Background

Disulfidptosis is a newly identified mechanism of cell death triggered by disulfide stress. Thus, gaining a comprehensive understanding of the disulfidptosis signature present in gastric cancer (GC) could greatly enhance the development of personalized treatment strategies for this disease.

Methods

We employed consensus clustering to identify various subtypes of disulfidptosis and examined the distinct tumor microenvironment (TME) associated with each subtype. The Disulfidptosis (Dis) score was used to quantify the subtype of disulfidptosis in each patient. Subsequently, we assessed the predictive value of Dis score in terms of GC prognosis and immune efficacy. Finally, we conducted in vitro experiments to explore the impact of Collagen X (COL10A1) on the progression of GC.

Results

Two disulfidptosis-associated molecular subtypes (Discluster A and B) were identified, each with distinct prognosis, tumor microenvironment (TME), immune cell infiltration, and biological pathways. Discluster A, characterized by high expression of disulfidptosis genes, exhibited a high immune score but poor prognosis. Furthermore, the Dis score proved useful in predicting the prognosis and immune response in GC patients. Those in the low Dis score group showed better prognosis and increased sensitivity to immunotherapy. Finally, our experimental findings validated that downregulation of COL10A1 expression attenuates the proliferation and migration capabilities of GC cells while promoting apoptosis.

Conclusions

This study demonstrates that the disulfidptosis signature can assist in risk stratification and personalized treatment for patients with GC. The results offer valuable theoretical support for anti-tumor strategies.

Introduction

Gastric cancer (GC), a malignant tumor of the digestive tract, is associated with high morbidity and mortality rates [1]. In recent years, the treatment options for GC have been continuously improved, and the number of deaths has been significantly reduced. However, the long-term survival rate of GC patients is still limited [2, 3]. At present, more and more clinical evidences confirm the effectiveness of immunotherapy against malignant tumors [4, 5]. Despite this, only a small percentage of GC patients show sensitivity to immunotherapy [6]. Drug resistance caused by the high heterogeneity of GC limits the efficacy of immunotherapy [7]. Therefore, it is crucial to identify patients who are sensitive to immunotherapy.

Disulfidptosis is a novel form of cell death characterized by rapid cell demise triggered by disulfide stress resulting from an accumulation of excess cystine within cells [8]. Solute carrier family 7 member 11 (SLC7A11) functions as a transporter facilitating cystine uptake and plays a crucial role in combating oxidative stress and preventing ferroptosis. Recent research indicates that in conditions of glucose deprivation, cells with elevated SLC7A11 levels experience NADPH depletion and disulfide stress, leading to the binding of intracellular cysteine sulfhydryl groups with actin and cytoskeletal proteins to form disulfide sulfur bonds. This process causes cytoskeletal contraction and triggers disulfidptosis. Neither ferroptosis nor apoptosis inhibitors can inhibit this cell death, but disulfide reducing agents can effectively inhibit cell death [8, 9]. Proteomics research indicates that the accumulation of disulfide bonds leading to cell death primarily affects actin cytoskeletal proteins, specifically MYH9/10 and actin [8]. Furthermore, studies have demonstrated that GLUT1 inhibitors can successfully trigger disulfidptosis. The identification of the disulfidptosis mechanism provides important insights into cancer therapy. Therefore, we identified two molecular subtypes based on the disulfidptosis signature of GC. We analyzed the molecular characteristics, extent of immune infiltration, and prognostic significance of disulfidptosis subtypes. Overall, this study offers valuable insights for tailoring personalized treatment strategies for tumors.

Materials and methods

Data acquisition and processing

Transcriptome data and somatic mutation information of GC patients were downloaded from TCGA-STAD(N = 317) and GEO databases (GSE84437) (N = 431) (Table 1). The “SVA” package was utilized to correct for batch effects in the TCGA-STAD and GSE84437 datasets.

Table 1 Clinical information of patients with STAD in this study

Identification of disulfidptosis subtypes and their correlation with the TME

23 disulfidptosis-related genes were obtained from Liu et al [8]. The ConsensuClusterPlus package was utilized to categorize patients into distinct disulfidptosis subtypes according to the expression levels of disulfidptosis genes. Variations in biological functions among these subtypes were examined using GSVA and ssGSEA algorithms.

TME scores and tumor purity of Discluster subtypes were calculated using the ‘ESTIMATE’ package. The ssGSEA algorithm was employed to assess immune cell infiltration in patients with GC.

Constructing the risk score related to disulfidptosis

The ‘limma’ package was utilized to identify gene expression differences among different Disclusters. Genes with expression differences having an absolute Log2 fold change greater than 1 and a corrected P value less than 0.05 were considered differentially expressed genes (DEGs). Univariate Cox regression analysis was conducted to pinpoint DEGs associated with prognosis. Subsequently, Lasso Cox regression was employed to construct a risk score related to disulfidptosis. This risk score for each sample was calculated as the expression level of each characteristic gene multiplied by its corresponding coefficient. Finally, KM survival analysis was used to assess differences in OS.

To further assess the independent prognostic significance of the risk model, we included clinical variables like age, gender, and tumor stage. Subsequently, we constructed a nomogram to predict OS in GC using the ‘rms’ package. The accuracy of the predictive model was evaluated using calibration curves, ROC curves, and DCA curves.

Assessment of immunological characteristics

To assess immune infiltration in GC patients, we employed the CIBERSORT and ESTIMATE algorithms for analysis. The somatic mutation profiles of high and low risk groups were analyzed using the ‘maftools’ package. Additionally, we investigated the correlation between Dis score, TMB, and MSI.

Immunological efficacy and drug sensitivity

Sensitivity to immunotherapy was evaluated using the TIDE score. The Cancer ImmunoAtlas database (TCIA) was utilized to assess the sensitivity of various subgroups to immune checkpoint inhibitors (ICIs). The IC50 value of chemotherapy drugs was calculated using the ‘PROPHIC’ package.

Cell culture and transfection

GC cell lines (AGS, HGC27, and MKN45) and human gastric mucosal epithelial cells (GSE1) were procured from the Cell Bank of the Chinese Academy of Sciences. BGC823 and SGC7901 were sourced from Jiangxi Provincial Key Laboratory. These cell lines were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, under conditions of 5% CO2 and 37 °C. Following the manufacturer’s guidelines, si-Col10A1 was transfected into AGS cells using Lipofectamine 3000 (Invitrogen, United States). Lipofectamine 3000 and si-Col10A1 (1:1) should be added to the serum-free medium separately. The mixture should then be incubated at room temperature for 15 min before being added to the cells for transfection. The siRNA fragments were obtained from Shanghai Hanheng Biotechnology Company. The siRNA sequences were shown below: siCOL10A1, 5′-CCAAGACACAGUUCUUCAUTT-3′ sense and 5′-AUGAAGAACUGUGUCUUGGTT-3′ antisense; negative control (NC), 5′-GUGGAUAUUGUUGCCAUCATT-3′sense and 5′-UGAUGGCAACAAUAUCCACTT-3′antisense.

Real-time qPCR and Western blot

Total RNA was isolated from cell lines following the protocol provided by the TRIzol kit (Invitrogen, United States). The extracted RNA was then reverse transcribed into cDNA using the PrimeScript RT kit (Takara, Japan). Subsequently, cDNA and gene primers were prepared in a 10 µl system according to the TAKARA kit instructions, followed by a Real-time qPCR reaction. The primer sequences for the genes can be found in Table 2.

Table 2 Primer sequences used for RT-qPCR

Clinical information and biological specimens were collected from patients who underwent radical gastric cancer surgery at the Second Affiliated Hospital of Nanchang University in December 2023 (supplementary material). RIPA lysis buffer was utilized to extract total protein from GC cells and clinical samples. In accordance with the protein concentration determination kit instructions (Beyotime, Shanghai), protein standards with varying concentrations should be prepared. An appropriate volume of BCA working solution should be prepared based on the number of samples, and added to both the protein sample and standards. Following a 2-hour incubation at room temperature, the absorbance at a wavelength of 562 nm should be measured using a microplate reader (Varioskan LUX, United States). The protein concentration of the sample was determined using the BCA method, followed by the addition of an appropriate amount of Loading buffer. Subsequently, electrophoresis was performed on a 10% SDS-PAGE gel, and the proteins were transferred to a PVDF membrane. The PVDF membrane was then sealed with a 5% BSA blocking solution for 1 h at room temperature. Next, the membrane was incubated overnight on a 4 °C shaker with Col10A1 antibody (1:500, Collagen X Antibody, abmart, TD13214) and GAPDH antibody (1:8000, GAPDH Monoclonal antibody, Proteintech, 60004-1-Ig).

Immunohistochemistry (IHC)

The specific steps of immunohistochemistry refer to the previous operation [10]. GC tissues were paraffin-embedded, followed by dewaxing and hydration. Antigen retrieval was performed, and the tissues were then blocked. Subsequently, the tissues were incubated overnight with Col10A1 antibody (1:50, abmart), labeled with secondary antibodies, stained and photographed. Score according to the degree of cell staining: negative (0 points), weakly positive (1 point), positive (2 points) and strong positive (3 points). Scoring is based on the percentage of positive cells: ≤25% (1 point), 26-50% (2 points), 51%~75% (3 points) and > 75% (4 points). The IHC score was calculated based on the two scores.

CCK-8

Transfected AGS cells were initially seeded in a 96-well plate with a density of 3,000 cells per well. Following the cells’ recovery, 10 µl of CCK8 working solution (Uelandy, Suzhou) was introduced into each well at 0 h, 24 h, 48 h, and 72 h. Subsequently, after a 2-hour incubation period, utilize a microplate reader (Varioskan LUX, United States) to measure the absorbance at a wavelength of 450 nm for the purpose of detecting cell proliferation.

Wound healing experiments

Transfected cells were seeded at an appropriate density in a 6-well plate. Once the cells reached confluence, scratches were made using a 20 µl pipette tip. The progress of wound healing was then assessed every 24 h, and ImageJ was used to calculate the area of ​​the wound. The wound healing rate was calculated using the formula: (original wound area - unhealed wound area) / original wound area * 100%.

Transwell experiment

A suitable number of cells were seeded in the upper chamber with free medium. The lower chamber was filled with DEME medium containing 20% fetal calf serum. After 24 h of incubation, the cells at the bottom of the chamber were fixed using 4% paraformaldehyde, followed by staining with crystal violet.

EDU detection

Transfected cells were seeded in 96-well plates at a density of 2 × 10^4 cells per well. After incubating for 24 h, the EDU solution was diluted to a concentration of 50 µM in Edu medium. Following this, 100 µL of the diluted Edu medium was added to each well and incubated for 2 h. The cells were then fixed, and 100 µL of Click-iT working solution (Uelandy, Suzhou) was added for staining. Finally, DNA counterstaining was performed, and pictures were taken using a fluorescence microscope.

Clone formation experiment

A total of 800 transfected cells were evenly distributed in a 6-well plate. Following a 14-day incubation period, the cells were washed twice with PBS, fixed with 4% paraformaldehyde, and then stained with crystal violet. The number of cell clones was then determined.

Apoptosis analysis

Cells were digested using EDTA-free trypsin, followed by transfer to a flow tube. After washing twice with pre-cooled PBS, as per the FITC-Annexin V/PI cell apoptosis kit instructions (Uelandy, Suzhou), resuspend 1 × 10^5 cells, then add 5 µL of FITC-Annexin V and PI staining solution to each tube. Incubate the tubes for 15 min at room temperature in the dark. Cell apoptosis was subsequently assessed using Beckman flow cytometry (CytoFLEX, United States).

Transfected AGS cells were seeded into 96-well plates and allowed to recover their shape. Subsequently, the cells were washed with PBS and fixed with 4% paraformaldehyde. Following this, permeabilization was achieved by adding an appropriate amount of 0.2% Triton X-100 at room temperature. 50 µL of TUNEL reaction solution (Uelandy, Suzhou) was added to each well and the cells were incubated for 1 h at room temperature in the dark. Finally, the cells were counterstained with DAPI staining solution and observed for cell apoptosis using immunofluorescence microscopy.

Immunofluorescence (IF) staining

Cells were evenly seeded in a 12-well plate and fixed with 4% paraformaldehyde, followed by permeabilization with 0.1% Triton X-100. Subsequent to blocking with a 5% BSA solution, the cells were incubated overnight at 4 °C with Col10A1 antibody (1:50, abmart). Fluorescent secondary antibodies were then utilized for staining at room temperature. Finally, the cell nuclei were counterstained with DAPI, and the expression of Col10A1 was visualized using confocal fluorescence microscopy.

Cytoskeletal staining

Transfected cells were seeded evenly in a 12-well plate at an appropriate density, then fixed with 4% paraformaldehyde and permeabilized using a 0.2% Triton X-100 solution at room temperature. Subsequently, the cells were incubated with 200 µL of phalloidin working solution for 20 min and counterstained with DAPI. Finally, immunofluorescence microscopy was employed to observe the distribution of F-actin.

Statistical analysis

All bioinformatics analyses were performed using R version 4.1.3. The normality of continuous variables was tested using the single-sample KS test. If the data followed a normal distribution, data analysis was conducted using the t-test or one-way ANOVA analysis of variance. The experiments were repeated three times. A significance level of P < 0.05 was considered statistically significant.

Results

Genetic changes in disulfidptosis-related genes in GC

Our research route is depicted in Figure S1. Twenty-three disulfidptosis genes (Supplementary Material) were obtained from Liu et al [11]. Most of these genes were upregulated in GC (Fig. 1A). Univariate COX regression analysis revealed that MYH10, DSTN, MYL6, TLN1, and FLNA were genes associated with a poor prognosis in GC (Fig. S2). Additionally, the gene mutation frequency of MYH10 was found to be the highest (Fig. 1B). The situation of CNV amplification and deletion is shown in Fig. 1 (Fig. 1C, D).

Fig. 1

Genetic mutational landscape of disulfidptosis genes in GC. (A) Expression distributions of DEGs between GC and normal tissues. (B) Genetic alterations in the disulfidptosis gene. (C) Alteration frequency of CNV in disulfidptosis gene. (D) Location of CNV alterations in disulfidptosis genes on chromosomes

Identification of disulfidptosis subtypes

A consensus clustering algorithm was utilized to categorize GC patients into two disulfidptosis subtypes, namely Discluster A and B (Fig. 2A). Disulfidptosis genes were significantly different between the two subtypes (Fig. 2B, C). Notably, these genes showed upregulation in Discluster A, despite patients in this group exhibiting a poor prognosis (Fig. 2D).

Fig. 2

2 Discluster identified by consensus clustering. (A) A consensus matrix heatmap defining three clusters (k = 2) and their associated regions. (B) Differences in clinical features and disulfidptosis gene expression levels among different Discluster. (C) PCA analysis showed significant transcriptome differences between the two subgroups. (C) KM curve analysis of OS differences among different Discluster

Furthermore, we analyzed the differences between the two subtypes of TME and biological pathways. The GSVA results revealed a strong association between Discluster A and pathways related to tumor proliferation and metastasis, such as gap junctions, regulation of the actin cytoskeleton, and focal adhesion (Fig. 3A). There was a notable disparity in the level of enrichment between the two subtypes within the hallmark set (Fig. 3B). ssGSEA analysis revealed a significantly difference in the abundance of immune infiltrates between the two subtypes (Fig. 3C). Discluster A exhibited higher TME score, lower tumor purity, and increased infiltration levels of immunosuppression-related lymphocyte subpopulations (Fig. 3D). In summary, we can define Discluster A as an immune exclusion phenotype.

Fig. 3

Biological pathways and TME characteristics associated with Discluster. (A) GSVA analysis between subtypes. (B) Enrichment scores of 2 Discluster in the Hallmark pathway. (C) Infiltration levels of immune cells in 2 Discluster. (D) Correlation of Discluster with TME scores

Discluster-related DEGs and their biological functions

Utilizing the ‘Limma’ package, 365 disulfidptosis-related DEGs were identified. GO and KEGG analysis revealed that these DEGs were primarily associated with the extracellular matrix and cell migration in tumors (FigureS3A ,B). Subsequent univariate COX regression analysis pinpointed 332 prognostic-related DEGs. By categorizing patients into 2 genetic subtypes based on the expression of these DEGs using consensus cluster analysis (FigS4A, B). It was observed that these DEGs were highly expressed in genotype subtype B, leading to a poor prognosis (FigS4C).

Construction and validation of Dis score

Four genes were selected through screening with LASSO and multivariate COX regression techniques, resulting in the development of a prognostic model associated with disulfidptosis (Fig S1 A, B). Dis score = PPP1R14A×0.105 + HEYL×0.199 + COL10A1 × 0.098 + TFF2 × 0.087. The KM survival analysis was performed on both the training set and the validation set, which showed that the low-risk group had a more favorable prognosis (Fig. 4A, B). The ROC curve of the training set shows that the AUCs of Dis score predicting 1-year, 3-year and 5-year OS were 0.606, 0.657 and 0.694, respectively. The AUC of the validation set also confirmed the good performance of Dis score in predicting the prognosis of GC (Fig. 4C, D). The risk curve demonstrates that patients with a high Dis score experience shorter OS and higher mortality rates (Fig. 4E, F).

Fig. 4

Constructing Dis score. (A, B) KM curve analysis of differences in OS between two subgroups. (C, D) ROC curves of Dis score predicting 1-year, 3-year and 5-year OS. (E, F) Risk curve of Dis score

In vitro experimental verification of COL10A1 function

Initially, we analyzed the expression of core genes in GC. HETL and COL10A1 showed higher expression levels in GC tissues compared to normal tissues, while PPP1R14A and TFF2 exhibited lower expression in tumors (Fig S5D). qPCR analysis revealed significant differences in mRNA expression of these 4 characteristic genes among GC cell lines (Fig. 5A). Among these genes, COL10A1 showed the most significant difference in expression between normal tissues and GC tissues, and thus was chosen as a representative gene.

Fig. 5

Expression level of COL10A1. (A) Expression of four characteristic genes in gastric cancer cell lines (GSE 1, AGS, HGC27, BGC823 and MKN45). The experiment was repeated three times and analyzed by one-way ANOVA, P < 0.05. (B) Immunohistochemical results of 4 pairs of gastric cancer tissues. (C) Expression of COL10A1 protein in 8 pairs of gastric cancer. Using t-test analysis, P < 0.05. (D) Expression level of COL10A1 protein in gastric cancer cell line. The experiment was repeated three times and analyzed by one-way ANOVA, P < 0.05. (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001)

Immunohistochemistry and Western blot experiments were used to analyze the expression of COL10A1 protein, both showing abnormally high expression in GC tissues (Fig. 5B, C). The mRNA and protein expression levels both indicated significantly abnormal expression of COL10A1 in AGS cells (Fig. 5A, D). Therefore, AGS cells were selected as the experimental cell line. IF staining revealed that COL10A1 was primarily expressed in the cytoplasm (Fig. 6A).

Fig. 6

Effect of COL10A1 on the progression of gastric cancer. (A) The expression of COL10A1 was analyzed by immunofluorescence. (B) The expression of COL10A1 in AGS was knocked down by siRNA, and the efficiency was verified by WB. (C, D, E) EDU, colony formation and CCK8 assay showed that the proliferative ability of AGS cells decreased after knockout of COL10A1. The results of CCK8 are shown as a line chart, and the other experimental results are shown as a bar chart. The above experiments were repeated three times, and the expression of COL10A1 was analyzed by one-way ANOVA. Other experiments were analyzed by t test. (*P < 0.05, **P < 0.01, ***P < 0.001)

We effectively knocked down COL10A1 using siRNA (Fig. 6B). Our experiments on cell proliferation using CCK8, EDU, and colony formation assays showed a significant decrease in the growth rate of AGS cells after downregulating COL10A1 expression (Fig. 6C, D, E). Furthermore, Transwell and wound healing assays demonstrated a notable inhibition in the migration ability of AGS cells after COL10A1 downregulation (Fig. 7A, B). Additionally, flow cytometry and TUNEL experiments were conducted to analyze the impact of COL10A1 on cell apoptosis, revealing that reducing COL10A1 expression promoted cell apoptosis (Fig. 7C, D). Interestingly, a decrease in F-actin expression was observed in AGS cells after knocking down COL10A1 (Fig. 7E). These findings provide further evidence of the close association between COL10A1 and the regulation of the cytoskeleton in the development of GC. Previous research has demonstrated that disulfidptosis is a form of cell death triggered by disulfide stress-induced cytoskeletal collapse. Our study provided indirect evidence supporting the involvement of COL10A1 in the disulfidptosis mechanism in GC.

Fig. 7

Effect of COL10A1 on the progression of gastric cancer. (A, B) Transwell and wound healing experiments showed that the migration ability of AGS cells decreased after knocking down COL10A1. (C, D) After knocking down COL10A1, the apoptosis rate of AGS cells increased. (E) After knocking down COL10A1, the expression of F-actin in AGS cells decreased. The above experiments were analyzed by t test. (*P < 0.05, **P < 0.01, ***P < 0.001)

Development of a nomogram to predict prognosis

We analyzed the association of Dis score with clinical variables. Dis score is significantly related to T and N stages of GC (Fig. 8A). The study integrated Dis score with clinical variables to assess the prognosis of patients with GC. The findings indicated that age, N stage, and risk score were significant factors associated with a poor prognosis (Fig. 8B). Furthermore, a prediction model was developed to enhance the accuracy of survival evaluations for GC patients (Fig. 8C). The calibration curve demonstrated that the nomogram had a high level of accuracy in predicting OS (Fig. 8D). The AUC and the clinical decision curve (DCA) indicated that the nomogram exhibited excellent predictive capability and clinical utility (Fig. 8E, F).

Fig. 8

Construction and validation of a nomogram. (A) Dis score was positively correlated with T and N stages. The later the T and N stages, the higher the DIS scores. (B) Results of univariate and multivariate COX regression. (C) A nomogram predicting 1-, 3-, and 5-year OS in STAD patients in the entire cohort. (D) Calibration curve of the nomogram. (E, F) ROC curve and DCA curve of nomogram predicting 3-year OS of patients

Immune landscape between different subgroups

The distribution of immune cell subpopulations between the two subgroups was found to be significantly different using the CIBERSORT algorithm (Fig. 9A). The low-risk group showed enrichment of anti-tumor lymphocyte subpopulations, including plasma cells, CD8 T cells, activated CD4 memory T cells, and M1 macrophages. However, the low-risk group exhibited lower TME scores compared to the high-risk group (Fig. 9B).

Fig. 9

Levels of immune cell infiltration and TME characteristics between the two ICDRS groups. (A) Correlation between Dis score and immune cell types. In addition, correlation of signature genes and immune cell infiltration in the Dis score model. (B) Correlation between Dis score and TME score. (C, D) Correlation between Dis score and TMB. (E) KM curve analysis of OS between high and low TMB groups. The prognosis of patients with high TMB and low risk group is obviously better than other groups. (F) Differences in the distribution of Dis score in TNN wild and mutant types. (G) Correlation of Dis score with MSI. (H) The relationship between Dis score and tumor stemness index

TMB combined with MSI can be used to identify patients who are sensitive to tumor immunotherapy. Patients with high TMB and MSI may have a higher objective response rate after receiving immunotherapy. The study results indicate a significant increase in mutation rate in the low-risk group (Fig. 9C, D). Additionally, we also observed a significant survival advantage in low-risk group patients who also had high TMB (Fig. 9E). Interestingly, we observed a higher occurrence of TNN gene mutations in low-risk patients (Fig. 9F). Additionally, our study revealed that patients with lower Dis scores were more likely to have MSI-H status (Fig. 9G). Furthermore, a clear relationship was observed between Dis score and RNAss (Fig. 9H).

Immunotherapy efficacy and drug sensitivity analysis

Chemotherapy combined with immunotherapy has been shown to effectively improve the prognosis of patients with advanced GC. Analysis of patient immune therapy responses using the TIDE database revealed that the risk score in the responsive group was significantly lower than that in the non-responsive group (Fig. 10A). Additionally, the TIDE score in the high-risk group showed a marked increase (Fig. 10B), suggesting that patients in the high-risk group were more likely to exhibit resistance to immune therapy. Interestingly, we observed a significant increase in the expression of PDL1 and CTLA4 in the low-risk group, while other ICP genes were highly expressed in the high-risk group (Fig. 10C). Results from IPS score analysis also indicated that patients in the low-risk group showed higher sensitivity to PD1 and CTLA4 therapy (Fig. 10D).

Fig. 10

Relationship between Dis score and immunotherapy and chemotherapy drug sensitivity. (A) Differences in risk scores between immunotherapy responders and non-responders. (B) TIDE scores between different Dis score subgroups. (C) Expression levels of ICPs in high and low risk groups. (D) IPS scores between different ICDRS subgroups. (E) Correlation between Dis score and chemotherapeutic drug sensitivity

The IC50 values of chemotherapy drugs were calculated using the “PRROPHIC” software package. The findings indicated that the low-risk group exhibited greater sensitivity to dasatinib, lapatinib, and imatinib, whereas the high-risk group displayed higher sensitivity to chemotherapy drugs like paclitaxel, bosutinib, gefitinib, and sorafenib (Fig. 10E). In conclusion, the Dis score serves as a valuable reference for personalized tumor treatment.

Discussion

Regulated cell death (RCD) is a cell death mode in which cells are regulated by specific molecular pathways [12]. RCD dysregulation is closely related to many diseases and tumors [13, 14]. Resistance to cell death and unlimited proliferation are key characteristics of cancer. Various modes of cell death, including ferroptosis, apoptosis, pyroptosis, and cuproptosis, have been extensively researched in recent years [15,16,17]. Recently, Liu et al. discovered a new type of cell death - disulfidptosis [8]. This new RCD provides a new direction for cancer therapy.

This study proposes a stratification method for GC based on disulfidptosis characteristics, which divides GC into two subtypes. The two subtypes exhibited notable differences in prognosis and TME characteristics. Discluster A is characterized by high expression of disulfidptosis genes, poor prognosis, enrichment of lymphocyte subsets associated with immunosuppression, high immune score and matrix score. We define ClusterA as an immune-exclusive type, or “cold tumor”. However, Dsicluster B has the opposite characteristics. Therefore, this GC stratification is beneficial for individualized treatment of tumors.

To comprehensively analyze the disulfide death characteristics of GC patients, we developed a scoring system specifically focused on disulfidptosis. This Dis score is derived from four key characteristic genes: PPP1R14A, HEYL, COL10A1, and TFF2. The low-risk group was characterized by favorable prognosis, enriched subsets of anti-tumor lymphocytes, high TMB and high MSI. Therefore, we defined the low-risk group as immunoinflammatory, or “hot” tumors. The high-risk group showed the opposite characteristics. “Hot” tumors are generally more responsive to ICIs, while “cold” tumors, which lack infiltration of anti-tumor lymphocytes, exhibit low or no response to ICIs [18, 19]. This conclusion is further supported by the TIDE score and IPS score. Therefore, the risk score we developed can assist in identifying GC patients who could potentially benefit from immunotherapy. It can also aid in the selection of appropriate chemotherapy drugs and ultimately improve patient outcomes.

In our further investigation, we examined the impact of Dis score on the advancement of GC using in vitro experiments. Col10A1, a significant member of the collagen family, plays a crucial role in regulating the extracellular matrix [20]. Previous research has shown that Col10A1 facilitates the progression of different malignant tumors through various signaling pathways [11, 21]. Col10A1 promotes pancreatic cancer progression through MEK/ERK signaling pathway [22]. Col10A1 promotes lung adenocarcinoma progression by remodeling the extracellular matrix [23]. However, the research of Col10A1 in gastric cancer is mainly based on bioinformatics analysis. This study reveals that Col10A1 plays a significant role in the malignant progression of GC. Silencing Col10A1 has shown to effectively suppress the proliferation and migration of GC cells while enhancing cell apoptosis. Additionally, Col10A1 is closely associated with tumor cytoskeletal remodeling. These findings indicate that Col10A1 could serve as a promising target for the treatment of GC.

This study has several limitations. Firstly, our analysis only focused on the correlation between disulfidptosis characteristics and tumor immunity using bioinformatics methods. Further experiments are necessary to explore this relationship in the future. Secondly, we only investigated the role of Col10A1 in GC and did not delve into the relationship between Col10A1 and disulfidptosis.

Conclusions

Our study revealed that the disulfidptosis signature plays a crucial role in evaluating the prognosis and TME of GC. We investigated the impact of Col10A1 in enhancing the advancement of GC, thereby offering valuable theoretical backing for anti-tumor strategies.