Abstract The mystery about the mechanistic basis of disulfidptosis has recently been unraveled and shows promise as an effective treatment modality for triggering cancer cell death. However, the limited understanding of the role of disulfidptosis in tumor progression and drug sensitivity has hindered the development of disulfidptosis-targeted therapy and combinations with other therapeutic strategies. Here, we established a disulfidptosis signature model to estimate tumor disulfidptosis status in approximately 10,000 tumor samples across 33 cancer types and revealed its prognostic value. Then, we characterized disulfidptosis-associated molecular features and identified various types of molecular alterations that correlate with both drug-resistant and drug-sensitive responses to anti-tumor drugs. We further showed the vast heterogeneity in disulfidptosis status among 760 cancer cell lines across 25 cancer types. We experimentally validated that disulfidptosis score-high cell lines are more susceptible to glucose starvation-induced disulfidptosis compared to their counterparts with low scores. Finally, we investigated the impact of disulfidptosis status on drug response and revealed that disulfidptosis induction may enhance sensitivity to anti-cancer drugs, but in some cases, it could also lead to drug resistance in cultured cells. Overall, our multi-omics analysis firstly elucidates a comprehensive profile of disulfidptosis-related molecular alterations, prognosis, and potential therapeutic therapies at a pan-cancer level. These findings may uncover opportunities to utilize multiple drug sensitivities induced by disulfidptosis, thereby offering practical implications for clinical cancer therapy. Keywords: Regulated cell death, Disulfidptosis, Pan-cancer, Drug sensitivity, Multi-omics, Targeted therapy Graphical abstract Disulfidptosis was a recently discovered regulated cell death, characterized by NADPH depletion, aberrant disulfide bonding among actin cytoskeleton proteins, and actin network collapse. In this study, we innovatively constructed a disulfidptosis signature to estimate the disulfidptosis status and observed that patients with active disulfidptosis status had significantly better survival. Our findings provide biological insight into the consideration of cancer combination therapeutic strategies based on disulfidptosis. Image 1 [41]Open in a new tab Highlights * • Disulfidptosis is another new breakthrough in the field of regulated cell death. * • The robustness of disulfidptosis signature was verified in 6 independent glucose-starved datasets. * • Prognostic significance and molecular alterations were characterized at multidimensional levels across 26 cancer types. * • Disulfidptosis score-high cell lines are more susceptible to glucose starvation-induced disulfidptosis. 1. Introduction Maintenance of the oxidation-reduction (redox) balance is vital for cell survival [[42]1]. Cancer cells often experience elevated levels of oxidative stress compared to non-tumor cells, primarily due to genetic mutations and metabolic reprogramming [[43]2]. To ensure cellular life and proliferation, cancer cells must maintain sufficient glutathione (GSH) levels to counteract excessive intracellular reactive oxygen species (ROS) [[44]1,[45]3,[46]4]. During the GSH synthesis process, cysteine is the key substrate with low content and needs to be supplied through multiple pathways. Most tumor cells mainly rely on solute carrier family 7 member 11 (SLC7A11) to transport extracellular cystine into the cytoplasm and reduce it to cysteine using nicotinamide adenine dinucleotide phosphate (NADPH). Therefore, SLC7A11 is considered a critical oncogene in maintaining cell survival and antioxidant defense of cancer cells [[47]5]. However, recent studies have proposed an unexpected role of SLC7A11 in inducing cell death under glucose deprivation conditions, termed “disulfidptosis” [[48]6,[49]7]. Research on disulfidptosis began in 2017, when Gan and his team found that SLC7A11 overexpression increased glucose dependence in cancer cells and triggered cell death upon glucose starvation [[50]8]. In 2020, they further indicated that protein regulator of cytokinesis 1 (PRC1) inhibition coupled with activating transcription factor 4 (ATF4) induction promotes cell death under glucose starvation conditions [[51]9]. Meanwhile, they demonstrated that key enzymes, including phosphogluconate dehydrogenase (PGD), glucose-6-phosphate dehydrogenase (G6PD), transaldolase 1 (TALDO1), and transketolase (TKT) involved in the glucose-pentose phosphate pathway (PPP), could rescue glucose deprivation-induced cell death in SLC7A11^high cancer cells [[52]10]. In 2023, Gan and his team introduced the term disulfidptosis to define this unknown cell death [[53]6], and found glycogen synthase 1 (GYS1), 3-oxoacyl-ACP synthase, mitochondrial (OXSM), nicotinamide adenine dinucleotide hydrogen (NADH): ubiquinone oxidoreductase core subunit S1 (NDUFS1), NADH: ubiquinone oxidoreductase subunit A11 (NDUFA11), nucleotide binding protein-like (NUBPL), leucine rich pentatricopeptide repeat containing (LRPPRC), SLC7A11, solute carrier family 3 member 2 (SLC3A2), ribophorin I (RPN1), and NCK associated protein 1 (NCKAP1) as important participant hits in disulfidptosis through genome-wide CRISPR/Cas9 lose-of-function screening analysis [[54]6]. Notably, glucose starvation is a prerequisite for disulfidptosis in SLC7A11^high cancer cells, and as expected, inhibition of glucose uptake by glucose transporter 1/3 (GLUT1/3) inhibitors leads to glucose starvation and disulfidptosis in SLC7A11^high cancer cells both in vitro and in vivo [[55]6,[56]10]. Biologically, SLC7A11^high cancer cells enhance antioxidant defense capabilities by increasing cystine import and GSH synthesis. This beneficial effect consumes large amounts of NADPH, which is provided by glucose through the PPP [[57]1]. Therefore, when glucose supply is insufficient, tumor cells will experience NADPH depletion and abnormal accumulation of intracellular cystine and other disulfide molecules such as γ-glutamyl-cystine and glutathionyl-cysteine, which results in disulfide stress and aberrant disulfide bonding among actin cytoskeleton proteins, leading to actin network collapse and cell death [[58]6]. This process was verified through direct observation of disulfidptosis-related F-actin contraction and detachment from the plasma membrane by co-staining and unbiased bio-orthogonal chemical proteomic analyses [[59]6]. Cystine is the common amino acid with extremely low solubility, and its large accumulation can lead to the formation of highly toxic crystals in the intracellular lysosomes. However, no clear cystine crystal was observed in cells under disulfide stress through transmission electron microscopy [[60]10]. Additionally, elevated ATP levels, and inability of ROS scavengers to rescue the cell death ruled out the possibility of cell death associated with ATP depletion and ROS accumulation. Thus, this novel form of regulated cell death (RCD) is distinct from other existing forms of cell death, and shed light on new frontiers in RCD and reveal novel mechanisms by which organisms counteract malignant progression of tumors. RCD plays fundamental roles in cancer therapeutics, including apoptosis, necroptosis, autophagy-dependent cell death, pyroptosis, ferroptosis, and cuproptosis [[61][11], [62][12], [63][13]]. Like other RCDs, the elucidation of disulfidptosis will provide a critical framework for understanding and targeting this unique cell death in cancer therapy [[64]7]. However, current research on disulfidptosis is still limited to the molecular level, and the disulfidptosis status in large populations remains unclear. The hallmark of disulfidptosis status is F-actin contraction and detachment from the plasma membrane, which is difficult to detect in human physiological and pathological conditions. Therefore, it is urgent to develop a robust predictive signature to assess tumor disulfidptosis status in large-scale cancer patients and explore the impact of disulfidptosis on molecular alternations across multiple dimensions. In this study, we innovatively constructed a disulfidptosis signature to estimate the disulfidptosis status and observed that patients with active disulfidptosis status had significantly better survival. We next depicted the molecular characteristics of disulfidptosis across different cancer types from a multi-omics perspective. Moreover, we analyzed the correlation between disulfidptosis status and sensitivities to anti-cancer drugs, and experimentally validated the sensitization to several drugs by disulfidptosis induction in cancer cells. Our findings highlight the role of disulfidptosis in patient prognosis and provide biological insights into the consideration of cancer combination therapeutic strategies based on disulfidptosis. 2. Materials and methods 2.1. Estimating disulfidptosis status across cancer samples by a gene signature A total of 23 disulfidptosis-related genes were collected from published articles ([65]Table S1) [[66]6,[67]9,[68]10]. Then these genes were divided into two categories according to their regulatory direction, including promoting disulfidptosis genes and inhibiting disulfidptosis genes. The disulfidptosis score model to represent the disulfidptosis status was establish based on the expression data for genes of core promoting disulfidptosis components (pro) including SLC7A11, SLC3A2, RPN1, NCKAP1, cytoplasmic FMR1 interacting protein 1 (CYFIP1), WASP family member 2 (WAVE2), abl interactor 2 (ABI2), haematopoietic stem/progenitor cell protein 300 (HSPC300), Rac family small GTPase 1 (RAC1), ATF4; and negative core components (anti) of NUBPL, NDUFA11, LRPPRC, OXSM, NDUFS1, GYS1, G6PD, PGD, TALDO1, TKT, PRC1, GLUT1, GLUT3. We calculated enrichment score (ES) of pro-disulfidptosis genes and anti-disulfidptosis genes using single sample gene set enrichment analysis (ssGSEA) in the R package ‘GSVA’ [[69]14], the disulfidptosis score to computationally dissect the disulfidptosis status of the tissue samples, and cancer cell line was defined by the differences of ssGSEA score between the ES of pro-disulfidptosis genes minus anti-disulfidptosis genes. The disulfidptosis score model was validated in six independent datasets with known glucose starved status from the Gene Expression Omnibus (GEO): [70]GSE209636, [71]GSE184452, [72]GSE121378, [73]GSE62663, [74]GSE16157, and [75]GSE95097. Wilcoxon rank sum test was used to assess the statistical difference between glucose starved/GLUT1 inhibitor and normal conditions in different cancer cell lines. 2.2. Integration of multi-omics data and clinical data for TCGA samples Multi-omics data including mRNA expression, miRNA expression, protein expression, somatic mutations, somatic copy number alteration (SCNA) and clinical data across 33 cancer types were downloaded from The Cancer Genome Atlas (TCGA) data portal ([76]https://portal.gdc.cancer.gov/) [[77]15]. The tumor purity of TCGA-tumor samples was obtained from TIMER: Tumor Immune Estimation Resource ([78]http://cistrome.org/TIMER/download.html) and [79]https://doi.org/10.5281/zenodo.253193 [[80]16,[81]17]. 2.3. Stratification and multi-omics analysis of tumor samples from TCGA pan cancer cohort TCGA pan cancer samples were divided into three parts based on the disulfidptosis score distribution of tertiles, defining the top and bottom samples as disulfidptosis-score high and disulfidptosis-score low samples, respectively. We retained a total of 26 cancer types with≥30 samples in both disulfidptosis score-high and disulfidptosis score-low groups for further analysis. METAbolic Flux balance analysis (METAFlux) was used to calculate the glucose metabolic fluxes to investigate the association between the glucose uptake level and disulfidptosis status across cancer types [[82]18]. We further used the matching weights (MW) method of propensity score matching (PSM) algorithm to balance the effects of potential confounders [[83]19,[84]20], including age, gender, tumor purity, race, tumor stage, and examined the balance by comparing standardized difference before and after PSM (standardized difference <0.05). Subsequently, we compared the molecular difference of multi-omics between high disulfidptosis score and low disulfidptosis score in TCGA cohorts. In order to decrease random noise in feature identification, permutation test was repeated 100 times via randomly selecting the high disulfidptosis score or low disulfidptosis score samples. Significant features for four molecular types were identified by the criterion: mRNA expression |fold change| > 2, FDR <0.05; miRNA expression |fold change| > 1.5, FDR < 0.05; somatic mutation and SCNA: FDR <0.05; total protein and DNA methylation: |difference| > 0.1, FDR < 0.05. 2.4. Differential abundance (DA) score The DA score evaluates the differential regulation of a metabolic pathway between groups with high and low disulfidptosis scores. First, by performing Pearson correlation analysis, we identified 167 metabolites that were significantly positively correlated (pos_cor) with disulfidptosis score (Rs > 0.18, FDR <0.05) and 135 metabolites that were significantly negatively correlated (neg_cor) with disulfidptosis score (Rs < −0.18, FDR <0.05). The DA score for each pathway is calculated as: [MATH: DAscor< mi>e=No.ofmeta< mi>bolites(pos_cor)−No.ofmeta< mi>bolites(neg_cor)No.ofmeas< mi>uredmeta< mi>bolitesinpath< mi>way :MATH] Thus, the DA score varies from −1 to 1. A score of −1 indicates that all metabolites in a pathway are negatively correlated with disulfidptosis, while a score of 1 indicates that all metabolites are positively correlated with disulfidptosis. Only pathways with 3 or more significantly altered metabolites were scored. 2.5. Analysis of clinically actionable genes and drug response associated with disulfidptosis status The area under the dose-response curve (AUC) data and gene expression matrix for cancer cell lines were downloaded from the Genomics of Drug Sensitivity in Cancer (GDSC) ([85]http://www.cancerrxgene.org/downloads) and Cancer Cell Line Encyclopedia (CCLE). Imputed drug response of 138 antitumor drugs in TCGA cancer patients were download from a previous study [[86]21]. The information of clinically actionable genes targeted by Food and Drug Administration (FDA)-approved drugs was downloaded from a previous study [[87]22]. The drug repurposing information with drug-target was downloaded from The Drug Repurposing Hub ([88]https://clue.io/repurposing-app). To assess drug response in cancer cell lines, we calculated the Spearman correlation between the AUC and gene expression of cancer cell lines from GDSC for drug responsiveness (|Rs| > 0.3; FDR <0.05). To assess the effect of disulfidptosis status on drug sensitivity in TCGA tumor samples, Spearman correlation between imputed drug response and disulfidptosis score we calculated (|Rs| > 0.2; FDR <0.05). Among them, drug resistance showed a positive Spearman correlation, while drug sensitivity showed a negative Spearman correlation. 2.6. Pathways enrichment analysis and miRNA-target regulatory networks In order to study the differences of different disulfidptosis status in cancer hallmarks pathway and Gene Ontology Biological Processes (GOBP). We firstly downloaded the gene set of GOBP and cancer hallmark gene sets from the MSigDB database ([89]https://www.gsea-msigdb.org/gsea/msigdb/). Then, we used the GSVA method to evaluate these gene sets activity (enrichment scores) of GOBP and cancer hallmark pathways from in each cancer sample. Finally, we calculated the differences of each pathway between the disulfidptosis-high and low groups and screened out the pathways with consistent and significant changes in various cancers. And miRNA-target pairs were downloaded from the miRTarBase [[90]23]. The significantly altered miRNAs and mRNAs (as targets) were used to identify the miRNA-target relationships. Pathway Enrichment analysis of miRNA-target genes was performed by using the clusterProfiler Package [[91]24]. Based on the miRNA-target pairs, we constructed miRNA-target regulatory networks, where the nodes are the miRNAs, or target genes, and the edges are the regulatory pairs. 2.7. Bulk RNA-seq deconvolution analysis Cell abundance was measured using CIBERSORT with the LM22 matrix from Newman et al. [[92]25] ([93]https://cibersort.stanford.edu/) to quantify the relative abundance of 22 types of immune cells in TCGA pan-cancer. 100 times for permutation test. Expression matrix normalized by FPKM as the input data. 2.8. Cell culture The human malignant melanoma cell lines A375 and SK-MEL-28, human cervical carcinoma cell line Hela, and human hepatocellular carcinoma cell line Huh-7 were cultured in Dulbecco's modified Eagle medium (DMEM; Thermo Fisher Scientific) containing 10% fetal bovine serum (FBS; Biological Industries) and 1% penicillin-streptomycin solution at 37 °C in an incubator with humid air of 5% CO[2]. All cell lines were obtained from American Type Culture Collection and free of Mycoplasma contamination (tested by the vendor). None of the cell lines have been found in the International Cell Line Authentication Committee database of commonly misidentified cell lines, based on short tandem repeat profiling performed by the vendor. For the glucose deprivation experiments, cells were incubated with glucose-free DMEM with 10% FBS. The glucose-free DMEM (11966025) was purchased from Thermo Fisher Scientific. The cystine-free and glucose/cystine-double-deprived DMEM were customized from Procell Life Science&Technology. 2.9. Chemicals Necrostatin-1s (HY-14622A), chloroquine (HY-17589A), Z-VAD-FMK (HY–16658B), liproxstatin-1 (HY-12726), dithiothreitol (HY-15917), etoposide (HY-13629), methotrexate (MTX, HY-14519), and CMK (HY-52101), GW-441756 (HY-18314), ABT-263 (HY-10087), NSC-87877 (HY-18756), staurosporine (HY-15141), and 2-Deoxy-d-glucose (2DG, HY-13966) were purchased from MedChemExpress. 2.10. Cell viability assay To measure cell viability, 6000 cells per well were seeded in 96-well plates and allowed to adhere. For cell death inhibitor rescue experiments, cells were cultured with glucose-free medium with different cell death inhibitors for about 16 h. For drug sensitivity experiments, cells were cultured with glucose-free or glucose-containing DMEM in indicated concentrations of drugs for about 10–14 h. After treatment, the culture medium in each well of the plate was replaced with 100 μl fresh medium containing 10 μl Cell Counting Kit-8 (CCK-8) (Bimake, [94]B34302). And the culture was returned to the incubator for 2–3 h at 37 °C. Measure the absorbance at 450 nm using a microplate reader and calculate the cell viability according to the manufacturer's instructions. 2.11. Cell proliferation assay Cell proliferation assay was performed using BeyoClick™ EdU Cell Proliferation Kit with Alexa Fluor 488 (Beyotime Biotechnology, C0071L) according to the manufacturer's instructions. Briefly, cells (4 × 10^4 cells/well) were seeded in 24-well plates and cultured with indicated treatments for about 12 h. Subsequently, cells were incubated with EdU for 2 h, fixed with 4% paraformaldehyde for 15 min, and then permeated with 0.3% Triton X-100 for 15 min. The cells were incubated with the Click Reaction Mixture in the dark for 30 min at room temperature, and then incubated with Hoechst (Beyotime Biotechnology, 33342) for 10 min. Images was detected and captured by Fluorescent Microscopy (Nikon, ECLIPSE Ts2R), and quantification was done using Image J. The results are shown as the ratio of the EdU-positive cells to Hoechst-positive cells. 2.12. Cell cycle Cell cycle was implemented by cell cycle kit (Beyotime Biotechnology, C1052). After receiving the indicated treatments, cells were collected and fixed overnight in cold 70% ethanol. Then, the cells were stained with propidium iodide and measured by flow cytometry according to the manufacturer's protocols. Cell cycle distribution was assessed by Flow Jo software (version 10.4). 2.13. Caspase-3 activity assay The activity of caspase-3 was determined using the caspase-3 activity kit (Beyotime Biotechnology, C1116). According to the manufacturer's protocol, cell lysates of SK-MEL-28 cells after indicated treatments were centrifuged at 12, 000×g at 4 °C for 15 min, and protein concentrations were determined by Bradford protein assay. Cellular extracts were incubated in a 96-well plate with Ac-DEVD-pNA (2 mM) for 6 h at 37 °C. Then caspase-3 activity was quantified in the samples with a microplate reader by the absorbance at a wavelength of 405 nm. 2.14. Evaluation of apoptosis by flow cytometry The apoptosis ratio was analyzed using the Annexin V-FITC Apoptosis Detection Kit (Beyotime Biotechnology, C1062). After indicated treatments, the cells were washed with PBS, digested by trypsin (EDTA depleted), and collected by centrifugation. After being washed with PBS, the cells were resuspended and stained by the binding buffer containing PI and Annexin-V-FITC for 15 min according to the manufacturer's instructions, measured by flow cytometry and finally analyzed by Flow Jo software (version 10.4). 2.15. LDH release assay Cellular cytotoxicity was monitored by LDH release assay using the LDH Cytotoxicity Assay Kit (Beyotime Biotechnology, C0016) according to the manufacturer's instructions. Briefly, cells were seeded in 96-well plates and then cultured with indicated treatments. Subsequently, the supernatants were collected and transferred to a new plate and incubated with Reaction Mixture in the dark for 30 min at room temperature. Absorbance at 490 nm was measured using a microplate reader to determine the amount of LDH released from cells. 2.16. Statistical analysis Wilcoxon rank sum test was used to compare the differences. Assessment of statistical significance (p-value) for the association between disulfidptosis-biased mRNA expression and methylation patterns was calculated by Fisher's exact test. Univariate Cox regression model was used to calculate the hazard ratio (HR) of disulfidptosis score. Survminer package was used to determine the cutoff point of survival information for each dataset based on the association between disulfidptosis score and patient overall survival (OS) time, progression free survival (PFS) time, disease free survival (DSS) time and the log-rank test was used to determine the significance of the differences. To find the maximum rank statistic and reduce the calculated batch effect, the “surv-cutpoint” function was used to dichotomy disulfidptosis score and all potential cutting points were repeatedly tested, then the patient samples were divided into the high-disulfidptosis score group and the low-disulfidptosis score group according to the maximum selected log-rank statistics. Kaplan-Meier comparative survival analyses for prognostic analysis were generated, and the log-rank test was used to determine the significance of the differences. Univariate Cox regression model was used to calculate the hazard ratio (HR) of the disulfidptosis status (the high-disulfidptosis score group and the low-disulfidptosis score group). All statistical analysis was two-side. The experimental data was analyzed by GraphPad Prism software and presented as mean ± SD. P values of <0.05 were considered statistically significant and represented as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. All experimental data were obtained from at least three independent biological replicates. 3. Results 3.1. Identification of a gene signature to estimate disulfidptosis status across cancer samples To estimate the disulfidptosis status in tumor patients, we firstly collected 23 genes coding disulfidptosis regulators and markers from current relevant research [[95]6,[96]9,[97]10], including 10 positive regulators and 13 negative regulators of disulfidptosis ([98]Fig. 1A, [99]Table S1). Next, we collected 6 independent gene expression datasets of cancer cell lines of multiple cancer types under glucose starved/GLUT1 inhibitor-treated and control conditions. To dissect the complexity and heterogeneity of disulfidptosis levels, we then used the ssGSEA algorithm to calculate the disulfidptosis scores based on the enrichment score (ES) of pro-disulfidptosis genes calculated by ssGSEA minus that of the anti-disulfidptosis genes (see Methods) [[100]26]. Notably, the disulfidptosis scores can accurately distinguish the disulfidptosis status in tumor cell lines, characterized by higher disulfidptosis scores in cells under GLUT1 inhibition or glucose starvation, compared with control cells in all 6 independent datasets ([101]Fig. 1B). Consistently, GLUT1 inhibitor BAY-876 or glucose starvation has been reported to induce intracellular disulfidptosis in SLC7A11^high cells [[102]6]. To further simulate this state in cancer patient samples, we divided approximately 10,000 tumor samples into two groups: “high-SLC7A11/low-GLUT1” and “low-SLC7A11/high-GLUT1” according to the expression level of SLC7A11 and GLUT1 across 33 cancers from TCGA. We observed that patients in high-SLC7A11/low-GLUT1 group have a significantly higher disulfidptosis scores than those in low-SLC7A11/high-GLUT1 group ([103]Fig. 1C). Furthermore, we calculated glucose metabolic fluxes using METAFlux, a framework for inferring metabolic fluxes from bulk or single-cell transcriptome data [[104]18], to investigate the association between the glucose uptake level and disulfidptosis status across cancer types. As expected, we observed the glucose uptake level was significantly lower in patients with high disulfidptosis score across tumor lineages, comparing differences in glucose uptake levels between the disulfidptosis score-high and disulfidptosis score-low groups ([105]Fig. S1). These results demonstrate the robustness of the 23-gene signature to define disulfidptosis status across different cancer types. Fig. 1. [106]Fig. 1 [107]Open in a new tab Validation of a gene signature for estimating disulfidptosis status among cancer samples. (A) Cell model diagram of positive and negative disulfidptosis regulators. (B) Disulfidptosis scores of cancer cell lines under glucose deprivation/fasting-mimicking/CLUT1 inhibitor-treated conditions (navy blue) and control diets conditions (magenta) in six datasets. Wilcox test was used to assess the difference. (C) Difference of disulfidptosis score between “high-SLC7A11/low-GLUT1” and “low-SLC7A11/high-GLUT1” groups among 33 cancer types. Pie charts show the percentage of cancer types with significant (magenta) and non-significant (grey) alteration. Wilcox test was used to assess the difference. (For interpretation of the references to colour in this figure legend, the reader is referred to