Integrating necroptosis and immune landscapes: a multi-omics-derived NecropImmScore stratifies prognosis and therapy in ovarian cancer

Abstract

Background

Ovarian cancer (OC) remains the deadliest gynecologic malignancy, largely due to its immunosuppressive tumor microenvironment (TME) and resistance to therapy. Necroptosis, a regulated lytic cell death pathway mediated by the RIPK1-RIPK3-MLKL axis, can trigger immunogenic cell death, but its specific role in shaping the OC immune landscape and its clinical translation potential are posorly understood.

Methods

We employed multi-omics analysis (transcriptomics, genomics, clinical data) from TCGA-OV (n = 380), ICGC OV-AU, and IMvigor210 cohorts, combined with rigorous in vitro functional validation using OC cell lines (SKOV3, HEY), macrophages (THP-1 derived), and T cells (Jurkat). Computational immunology approaches (ESTIMATE, CIBERSORT, ssGSEA) quantified immune infiltration. We identified MLKL-associated immune genes, performed survival analysis (Kaplan–Meier, Cox regression), and constructed a necroptosis-immune signature (NecropImmScore) using consensus clustering and PCA of 102 prognostic genes. Drug sensitivity was predicted via pRRophetic and CellMiner.

Results

MLKL emerged as a protective prognostic biomarker (p = 0.018), significantly correlated with enhanced immune infiltration (ImmuneScore, StromalScore, ESTIMATEScore; p < 2.22e-16), M1 macrophage polarization (p = 0.006), activated CD4 + T cells (p = 0.003), and elevated immune checkpoint expression (PD-L1, CTLA4, LAG3, TIGIT). In vitro, MLKL overexpression in OC cells promoted M1 polarization (p < 0.05), activated Jurkat T cells (upregulated CCR4/5/7/9, CD69, CD3D/E, GZMB; p < 0.05), and induced key chemokines (CXCL9/10/11/13) critical for immune cell recruitment. Integration of MLKL-related and immune-related DEGs (n = 632) revealed enrichment in T-cell activation, chemokine signaling, and antigen presentation pathways (FDR < 0.05). Consensus clustering based on 102 survival-associated genes defined three molecular subtypes (Clusters A–C) with divergent survival (p = 0.019), necroptosis activity, and immune infiltration (Cluster C: best prognosis, highest MLKL/ImmuneScore). The derived NecropImmScore robustly stratified patients: high-score correlated with superior overall survival (TCGA: p < 0.001; ICGC: p = 0.014), inflamed TME phenotype, elevated checkpoint expression, and improved response to anti-PD-L1 in IMvigor210. Critically, high NecropImmScore predicted higher BRCA1 mutation frequency (AUC = 0.802), synergy with BRCA1 status for prognosis, higher homologous recombination deficiency (HRD) score, sensitivity to cisplatin (p = 0.014), paclitaxel (p = 0.016), gemcitabine (p = 0.017), and provided superior prognostic stratification when combined with TMB and HRD score (p < 0.001).

Conclusion

This study establishes MLKL as a master regulator of anti-tumor immunity in OC, driving chemokine-mediated immune cell recruitment and TME reprogramming. The novel NecropImmScore is a multifaceted biomarker that effectively predicts prognosis, immunotherapy response, BRCA1 deficiency, and chemosensitivity, offering significant potential for guiding precision therapeutic strategies in OC.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-025-04192-z.

Introduction

Ovarian cancer (OC) persists as the most lethal gynecologic malignancy, characterized by insidious onset, frequent diagnosis at advanced stages (FIGO III/IV), and pervasive therapeutic resistance [1, 2]. Despite aggressive cytoreductive surgery and platinum-taxane chemotherapy, the 5-year survival rate for advanced disease remains dismal (20–30%), largely attributable to peritoneal dissemination, acquired chemoresistance, and the absence of effective early detection strategies [3]. A cornerstone of this clinical intransigence is the profoundly immunosuppressive tumor microenvironment (TME), which orchestrates systemic immune evasion through multifaceted mechanisms [4]. These include defective antigen presentation by dendritic cells, dominance of regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), physical exclusion of cytotoxic CD8 + T lymphocytes, and functional exhaustion of effector immune populations via chronic exposure to inhibitory signals [5]. This immunologically “cold” phenotype critically undermines the efficacy of emerging immunotherapeutic strategies, including immune checkpoint blockade (ICB) [6].

Within the landscape of tumor–immune interactions, regulated cell death (RCD) pathways, extending beyond canonical apoptosis, have emerged as pivotal modulators of immunogenicity [7]. Necroptosis, a lytic, caspase-independent form of programmed necrosis governed by the RIPK1-RIPK3-MLKL signaling axis, holds particular significance [8]. Upon activation, the terminal effector mixed lineage kinase domain-like pseudokinase (MLKL) undergoes phosphorylation, oligomerization, and membrane translocation, executing plasma membrane permeabilization [9]. This lytic event triggers the release of damage-associated molecular patterns (DAMPs), such as HMGB1, ATP, and mitochondrial DNA, which act as potent endogenous adjuvants [10]. DAMP recognition by pattern recognition receptors (PRRs) on antigen-presenting cells (APCs) can theoretically initiate immunogenic cell death (ICD), bridging innate sensing to adaptive T-cell priming and fostering anti-tumor immunity [11, 12]. However, the precise mechanistic integration of necroptotic signaling within the complex adaptive immune circuitry of OC, and its net impact on disease progression versus immune control, remains enigmatic. Furthermore, evidence suggests MLKL possesses non-canonical functions beyond membrane disruption, potentially modulating inflammasome activation, cytokine production (e.g., type I IFNs), and direct T-cell stimulation, adding layers of complexity to its immunomodulatory role [13].

While pan-cancer analyses hint at correlations between MLKL expression and immune cell infiltration [14], a systems-level understanding of how MLKL dynamically sculpts the OC-specific TME—influencing immune cell composition, functional states, and spatial organization—is critically lacking. Moreover, a significant translational gap exists: No validated biomarkers currently capture the intersection of functional necroptotic signaling and productive anti-tumor immune activation. This deficiency is acutely felt in OC, where molecular subtypes (e.g., immunoreactive, mesenchymal) exhibit divergent baseline immunogenicity yet lack predictive signatures to guide the application of ICB or other immunomodulatory therapies.

This study comprehensively investigates MLKL as a central orchestrator of the OC immune landscape and leverages this understanding to develop a novel prognostic and therapeutic biomarker. Utilizing multi-omics analyses across the TCGA-OV, ICGC OV-AU, and IMvigor210 cohorts, complemented by rigorous in vitro functional validation, we demonstrate that MLKL functions as a protective prognostic biomarker associated with enhanced immune infiltration and adaptive anti-tumor immunity. We further elucidate that MLKL reprograms the TME by promoting macrophage M1 polarization, augmenting T-cell activation, and driving chemokine-mediated recruitment of immune effector cells. Building upon these findings, we construct a necroptosis-immune signature (NecropImmScore) derived from MLKL-associated genes and immune-related genes, which effectively stratifies OC patients into molecular subtypes exhibiting divergent survival outcomes, distinct TME phenotypes, and differential responses to immunotherapy.

Significantly, the NecropImmScore demonstrates predictive power beyond the immune context. It correlates with genomic instability features, identifies BRCA1 deficiency, forecasts sensitivity to conventional chemotherapeutics, and synergizes with tumor mutational burden (TMB) and homologous recombination deficiency (HRD) score to provide superior prognostic stratification. Collectively, this work establishes MLKL as a master regulator of anti-tumor immunity in OC and introduces the NecropImmScore as a multifaceted clinical tool with significant potential for enhancing risk stratification and guiding precision therapeutic strategies.

Materials and methods

Data collection

Comprehensive molecular profiles—including transcriptomes, somatic mutations, and clinical annotations—from 380 OC patients were retrieved from The Cancer Genome Atlas (TCGA). For independent survival validation, gene expression data with matched clinical records of the ICGC OV-AU cohort were acquired. Comprehensive molecular profiles from various cancer types were gathered from the GDC Data Portal. Prognostic stratification divided OC cases equally into expression-based subgroups for overall survival (OS) analysis. The IMvigor210 cohort (anti-PD-L1-treated urothelial carcinoma) was used for immunotherapy response analysis [15].

Identification of differentially expressed genes (DEGs)

OC samples (n = 380) were divided into two groups based on median ImmuneScore and MLKL expression. DEGs were identified using the limma package with false discovery rate (FDR)-adjusted p < 0.05 and |log2FC|> 1 [16]. Further analysis focused on DEGs between MLKL high/low groups to pinpoint genes directly influenced by MLKL, which may shape the immune response. These DEGs were then analyzed for functional enrichment in immune processes.

Immunity analysis

Utilizing the ESTIMATE algorithm, we quantified TME constituents in OC patients via ImmuneScore, StromalScore, and ESTIMATEScore, with higher values indicating greater abundance of immune/stromal components [17]. The distribution of tumor-infiltrating immune cells (TICs) was analyzed using two computational approaches: CIBERSORT deconvolution and single-sample gene set enrichment analysis (ssGSEA) [18–20]. Further analysis with the GSVA package, leveraging MSigDB gene sets, characterized TIC dynamics across expression subgroups and evaluated immune-related functional pathways. ssGSEA scores were normalized by Z-scaling to ensure comparability across samples and groups [21]. Immune checkpoint (ICP) profiles from published literature were correlated with gene expression (p < 0.05) [22, 23], and immunophenoscores (IPS) from The Cancer Immunome Atlas (TCIA) were similarly assessed for expression correlations at the same significance level.

Functional enrichment analysis and gene set enrichment analysis (GSEA)

DEGs underwent functional annotation via integrated R packages (clusterProfiler, enrichplot, ggplot2), identifying significantly enriched Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) terms under dual-threshold criteria (p < 0.05 & q < 0.05). Parallel pathway interrogation employed GSEA v4.0.3 to analyze HALLMARK collections (C2. CP. KEGG.v7.2. symbols) across expression cohorts, with significance defined by 5% nominal p-value and FDR thresholds.

Survival analysis, univariate cox (uniCox), and multivariate cox regression analysis

Survival analyses were performed using Kaplan–Meier methodology with log-rank testing (p < 0.05) to compare high and low expression groups [24]. To identify genes associated with overall survival, univariate Cox regression was applied to protein-coding genes from the TCGA-OV cohort, excluding lowly expressed genes (< 0.1 in ≥ 10% of samples). Hazard ratios (HR) with 95% confidence intervals (CI) were calculated, and p-values were adjusted using the Benjamini–Hochberg method (FDR < 0.05). The independent prognostic value of the NecropImmScore was further evaluated by multivariate Cox regression, adjusting for age (≥ 60 vs. < 60 years), grade (G1-2 vs. G3-4), and stage (I–II vs. III–IV).

Construction of NecropImmScore

The NecropImmScore was constructed by first identifying 102 prognosis-associated differentially expressed genes through univariate Cox regression analysis. Unsupervised consensus clustering of these genes was performed using the ConsensusClusterPlus package (1000 iterations) to define molecular subtypes with distinct survival outcomes and immune infiltration characteristics. Principal component analysis was then applied to the expression matrix of the 102 genes, and the first two principal components (PC1 and PC2) were selected to capture major expression variations. The NecropImmScore was calculated for each sample as the sum of the contributions of each gene to PC1 and PC2 using the formula:

The NecropImmScore was constructed by first identifying 102 prognosis-associated differentially expressed genes through univariate Cox regression analysis. Unsupervised consensus clustering of these genes was performed using the ConsensusClusterPlus package (1000 iterations) to define molecular subtypes with distinct survival outcomes and immune infiltration characteristics. Principal component analysis was then applied to the expression matrix of the 102 genes, and the first two principal components (PC1 and PC2) were selected to capture major expression variations [25]. The NecropImmScore was calculated for each sample as the sum of the contributions of each gene to PC1 and PC2 using formula [26]:

NecropImmScore = ∑ (PC1_i + PC2_i)

where i represents expression of NecropImmune-related genes. This composite score integrates the expression patterns of the selected genes into a single metric that reflects the overall necroptosis–immune status of the tumor.

Mutation analysis for different NecropImmScore groups

Somatic mutation profiles of OC samples were analyzed using MAF-formatted data processed with the MAFTOOLS package [27]. This toolkit enabled simultaneous identification of differentially mutated genes across expression cohorts. TMB was quantified per case according to established methodology: (total mutations/covered bases) × 10⁶ [28]. DNA damage response (DDR) regulators were curated through systematic literature review, with subsequent correlation analysis performed against gene expression patterns [29].

Receiver operating characteristic (ROC) curve analysis

The predictive power of the NecropImmScore for BRCA1 mutation status was evaluated using ROC curve analysis. The area under the curve (AUC) was calculated to quantify the overall accuracy, with statistical significance assessed by DeLong’s test using the R package pROC [30].

HRD score analysis

To assess the HRD status in OC patients, we utilized a comprehensive approach to calculate the HRD score based on genomic scar signatures. These signatures include Loss of Heterozygosity (LOH), Telomeric Allelic Imbalance (TAI) [31], and Large-Scale State Transitions (LST), which are indicative of HRD. We obtained raw data from the Affymetrix Genome-Wide Human SNP Array 6.0 and the HT Human Genome U133 Array (Affymetrix) for OC from TCGA dataset. Additionally, processed data from the Infinium HumanMethylation27 BeadChip (Illumina) were also retrieved. For each tumor sample, matched normal tissue data were used for comparison.

For cases with matched tumor and normal samples, the log R ratio and B-allele frequency of each probe were calculated using Affymetrix Power Tools and PennCNV. Segmented copy number data were generated using the allele-specific copy number analysis of tumors (ASCAT) algorithm, which adjusts for biases in copy number data [32]. The HRD score was computed as the sum of the TAI, LST, and LOH scores [33]. Specifically, LOH was defined as the number of chromosomal regions longer than 15 Mb where LOH occurs [34]. TAI was measured as the number of regions of allelic imbalance that extend to the telomeres [31]. LST was calculated as the number of breakpoints between regions longer than 10 Mb after filtering out regions shorter than 3 Mb [35].

Assessment of the sensitivity to chemotherapy drugs for different NecropImmScore groups

Drug sensitivity correlations with NecropImmScore were evaluated through integrated computational approaches. IC50 predictions using the “pRRophetic” R package quantified chemotherapeutic agent responses [36]. Intergroup IC50 comparisons employed Wilcoxon signed-rank tests (significance threshold: p < 0.05). Clinically relevant drug target screening was subsequently performed via the CellMiner database (https://discover.nci.nih.gov/cellminer), focusing on FDA-approved compounds and investigational agents in active clinical trials [37].

Cell culture, RNA extraction, and real-time qPCR

The human OC cell lines SKOV3 and HEY (both from ATCC, Manassas, VA, USA), the T-cell line Jurkat Clone E6-1 (Procell #CL-0129), and the monocytic cell line THP-1 (ATCC) were cultured in RPMI 1640 (SKOV3, Jurkat, THP-1; Procell) or DMEM (HEY; Procell). All media were supplemented with 10% fetal bovine serum (Biological Industries) and 1% penicillin/streptomycin (New Cell & Molecular Biotech), and cells were maintained at 37 °C under 5% CO₂. SKOV3 and HEY cells were transfected for 48 h with MLKL overexpression plasmid (MLKL-pcDNA3.1; Public Protein/Plasmid Library) or empty vector control (pcDNA3.1) using Lipofectamine 2000 (Invitrogen), according to the manufacturer’s protocol. Total RNA was extracted using TRIzol reagent (Invitrogen), reverse-transcribed with 5X All-In-One RT Master Mix (Applied Biological Materials Inc.), and analyzed by qPCR using TB Green Premix Ex Taq (Takara) with GAPDH as the endogenous control. Primer sequences are provided in Supplementary Table 1.

Co-culture of OC cell lines with THP-1 cells and human jurkat clone E6-1 T cells

THP-1 monocytes (5 × 10⁴ cells/well) were plated in 6-well culture dishes (Corning, USA) and polarized into M0 macrophages via 48 h exposure to 100 ng/ml phorbol-12-myristate-13-acetate (PMA; MedChemExpress, China). Subsequently, M1 macrophage differentiation was induced by stimulating M0 cells with 100 ng/ml lipopolysaccharide (LPS; PeproTech, USA) for 48 h, while M2 macrophages were generated by exposing M0 macrophages to 20 ng/ml interleukin-4 (IL-4, MedChemExpress, China) and 20 ng/ml interleukin-13 (IL-13, MedChemExpress, China) for 48 h.

For co-culture experiments, a Transwell® system (Corning, USA) featuring 0.4 μm microporous membranes was employed. OC cells (SKOV3/HEY) were introduced into the upper chambers, while effector cells—including M0 macrophages, M1 macrophages, and Jurkat Clone E6-1 T cells—were seeded in the lower compartments at 2 × 10⁵ cells/well. Following 24 h coculture, effector cells were harvested for downstream analyses.

Western blot (WB) assay

Cell lysates were prepared using radioimmunoprecipitation assay (RIPA) buffer containing protease inhibitor cocktail (PMSF) and then quantified using a BCA assay (Beyotime, China). Samples (30 µg per lane) were separated by 10% SDS-PAGE and transferred to PVDF membranes (Millipore, USA). After blocking with 5% non-fat milk in TBST for 90 min at room temperature, membranes were incubated overnight at 4 °C with the following primary antibodies diluted in 5% BSA-TBST: CD3D (Abclonal, A9770; 1:1000), CD3E (Abclonal, A19017; 1:1000), CD69 (Proteintech, 10,803–1-AP; 1:1000), CD8 (Proteintech, 66,868–1-Ig; 1:1000), GZMB (Abclonal, A2557; 1:1000), and β-actin (Abclonal, AC026; 1:10,000). Following TBST washes, membranes were incubated with HRP-conjugated goat anti-rabbit secondary antibody (1:10,000) for 60 min at room temperature. Signals were detected using ECL reagent (Thermo Fisher Scientific, USA) and imaged on a Bio-Rad ChemiDoc XRS + system.

Statistical analysis

All analyses were conducted using R (v4.0.5), GraphPad Prism 8, and SPSS 26.0. Intergroup comparisons employed unpaired t-tests (parametric) or Wilcoxon rank-sum tests (nonparametric), with p-values < 0.05 defining statistical significance.

Results

MLKL emerged as a central prognostic biomarker in OC via integrated necroptosis pathway analysis

In this study, a comprehensive Protein–Protein Interaction (PPI) network was generated based on the STRING database, incorporating only nodes with an interactive relationship confidence exceeding 0.95. This stringent criterion ensured the robustness and reliability of the network, allowing for an accurate ranking of gene importance within the necroptosis pathway (Fig. 1A). The necroptosis-related genes included in this study were ZBP1, TRAF5, NLRP3, FASLG, FAS, TRAF2, TNFSF10, TLR4, TLR3, TICAM1, MLKL, TRADD, TNF, FADD, RIPK3, RIPK1, and CASP8.

Fig. 1.

Identification of MLKL as a core prognostic biomarker in ovarian cancer through integrated necroptosis pathway analysis. A Protein–Protein Interaction (PPI) network of necroptosis-related genes (confidence score > 0.95) from the STRING database. Node size reflected the number of interacting partners. B Univariate Cox regression (uniCox) analysis of necroptosis genes (ZBP1, TRAF5, NLRP3, FASLG, FAS, TRAF2, TNFSF10, TLR4, TLR3, TICAM1, MLKL, TRADD, TNF, FADD, RIPK3, RIPK1, and CASP8) in the TCGA-OV cohort, highlighting MLKL and ZBP1 as significant prognostic factors (hazard ratios (HR) = 0.812 (95% CI: 0.668–0.986, p = 0.036); HR = 0.863 (95% CI: 0.768–0.971, p = 0.014), respectively, for MLKL and ZBP1). C Integration of PPI centrality and uniCox results identified MLKL as the pivotal node linking necroptosis signaling to survival outcomes. D Kaplan–Meier analysis confirmed high-MLKL expression predicts superior overall survival in OC patients (log-rank p = 0.018). Numbers at risk are shown below the plot. E Pan-cancer analysis of MLKL expression across 33 malignancies in TCGA. Tumor vs. normal tissue comparison revealed MLKL as a protective factor in most cancers, including urothelial cancer and ovarian cancer (red rectangle). (***p < 0.001; **p < 0.01; ns, not significant; two-tailed t-test). Abbreviations: HR, hazard ratio

Utilizing univariate Cox regression analysis on the above 17 necroptosis-related genes, MLKL and ZBP1 were identified as significant prognostic factors in the TCGA-OV cohort. The HR and their 95% CI were calculated for each gene. MLKL exhibited a hazard ratio of 0.812 (95% CI: 0.668–0.986, p = 0.036), indicating a protective effect on overall survival. Similarly, ZBP1 showed a hazard ratio of 0.863 (95% CI: 0.768–0.971, p = 0.014), also suggesting a protective effect (Fig. 1B). The integration of the PPI network and uniCox analysis revealed a critical intersection at MLKL (Fig. 1C). This gene emerged as a pivotal node in the necroptosis pathway, with its high expression patterns closely linked to improved survival outcomes in OC patients (p = 0.018, Fig. 1D).

To elucidate the broader role of MLKL across various cancer types, we conducted a pan-cancer analysis using data from TCGA. This analysis included 33 different malignancies, with tumor vs. normal tissue comparisons revealing MLKL as a protective factor in most cancers (Fig. 1E). The results showed that MLKL expression was significantly lower in tumor tissues compared to normal tissues in many cancer types, including urothelial cancer and ovarian cancer. This suggested that MLKL’s influence extends beyond OC, potentially offering a broader therapeutic target for cancer treatment strategies.

MLKL reprogrammed the tumor–immune microenvironment and predicted immunotherapy response of OC patients

In order to explore the role of MLKL in the OC TME, a detailed correlation analysis of MLKL expression with ImmuneScore, StromalScore, and ESTIMATEScore was then conducted. These scores are indicative of the immune and stromal components within the TME. The significant correlation observed between MLKL expression and these scores suggests that MLKL may play a crucial role in modulating the immune and stromal components of the OC microenvironment (Fig. 2A–C, p < 2.22e-16). Pan-cancer analysis further supported MLKL as a key modulator of the immune TME, showing strong covariation with ImmuneScore across multiple cancer types, particularly in immunogenic tumors such as cutaneous melanoma (SKCM, R = 0.67, p < 2.2e-16), sarcoma (SARC, R = 0.8, p < 2.2e-16), and ovarian cancer (OV, R = 0.72, p < 2.2e-16) (Supplementary Fig. 1A).

Fig. 2.

MLKL reprogrammed the tumor–immune microenvironment and predicted immunotherapy response in ovarian cancer. A–C MLKL expression positively correlated with ImmuneScore A, StromalScore B, and ESTIMATEScore C in the TCGA-OV cohort (Wilcoxon test; p < 2.22e-16). These scores, calculated using the ESTIMATE algorithm, reflect the abundance of immune and stromal components within the tumor microenvironment. Dots represented individual samples. D CIBERSORT analysis revealed significant associations between high-MLKL expression and increased M1 macrophages (p = 0.006) and activated CD4 + T cells (p = 0.003). Violin plots showed immune cell fractions. E–F ssGSEA demonstrated enrichment of immune-related pathways in high-MLKL tumors. Heatmap displayed normalized enrichment scores (NES) for 29 immune signatures. G-J In vitro co-culture assays: MLKL overexpression in SKOV3 G and HEY H OC cells induced M1 polarization markers (CD80/CD86/IL12B/NOS2) in M0/M1 macrophages. qPCR data; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; t-test. K-N MLKL-overexpressing OC cells activated Jurkat T cells: upregulation of chemokine receptors (CCR4/5/7/9, IL2R) by qPCR K-L and T-cell activation markers (CD3D/E, CD8, CD69, GZMB) by Western blot (M–N). O Elevated MLKL expression correlated with upregulated immune checkpoint molecules (PD-1, PD-L1, CTLA4, LAG3, TIGIT; ***p < 0.001, Wilcoxon test). P High-MLKL expression associated with “inflamed” immune phenotype in IMvigor210 cohort. Q-T Correlation with immunophenoscore (IPS) predicted enhanced response to anti-PD-1/CTLA4 therapy across combinatorial ICB scenarios

CIBERSORT analysis revealed that MLKL expression was significantly associated with M1 macrophages (p = 0.006) and activated memory CD4 + T cells (p = 0.003) (Fig. 2D). The ssGSEA results further support MLKL’s role in immune-related processes. The enrichment of gene sets related to immune pathways and functions indicates that MLKL expression was associated with the activation of specific immune pathways within individual OC samples (Fig. 2E–F). Pan-cancer immune infiltration analysis confirmed broad associations between MLKL and M1 macrophages in 21 cancer types, including breast invasive carcinoma (BRCA) and ovarian cancer (OV) (Supplementary Fig. 1B), as well as enhanced CD8 + T-cell infiltration in 12 malignancies such as uterine corpus endometrial carcinoma (UCEC, p < 0.001) and thyroid carcinoma (THCA, p < 0.001) (Supplementary Fig. 1C).

Co-culture experiments demonstrated that MLKL-overexpressing OC cells promoted polarization of macrophages toward M0 and M1 phenotypes (p < 0.05), indicating that MLKL could directly influence macrophage behavior (Fig. 2G–J). To further investigate the impact of MLKL on M2 macrophage polarization, we performed additional analyses and experiments. However, no significant correlation was observed with M2 macrophages (Supplementary Fig. 2A–C). Similarly, MLKL overexpression in OC cells enhanced T-cell activation, as evidenced by upregulation of activation markers (C–C Motif Chemokine Receptor 4 (CCR4), CCR5, CCR7, CCR9, IL2R; Fig. 2K–L) and elevated protein levels of GZMB, CD69, CD8, CD3D, and CD3E (Fig. 2M–N).

To evaluate the predictive utility of MLKL in immunotherapy outcomes, we next examined its correlation with ICPs. Our analysis revealed that MLKL expression was positively associated with key inhibitory receptors, including PD-L1, CTLA-4, LAG-3, and TIGIT, etc. (Fig. 2O). Immune subtype analysis in the IMvigor210 cohort revealed distinct clusters based on MLKL expression (Fig. 2P). Furthermore, integrated IPS analysis revealed that elevated MLKL expression correlated with significantly higher IPS values for both anti-PD-1 and anti-CTLA-4 therapies (Fig. 2Q–T), indicating enhanced immunotherapy responsiveness. Collectively, these findings established MLKL as a promising biomarker for predicting immunotherapy efficacy in OC patients. Moreover, pan-cancer analysis revealed that MLKL expression covaried with TMB in many kinds of cancers (e.g., colorectal adenocarcinoma (COAD), stomach adenocarcinoma (STAD); Supplementary Fig. 1D) and microsatellite instability (MSI) in BRCA (p < 0.001), colon adenocarcinoma (COAD, p < 0.001), OV (p < 0.001), and lung adenocarcinoma (LUAD, p < 0.01) (Supplementary Fig. 1E) This genomic context underpinned MLKL’s immunotherapy predictive value. These findings indicated that MLKL expression could reflect the immune landscape of TME, potentially offering predictive value for immunotherapy outcomes in OC patients.

MLKL drove chemokine-mediated immune cell recruitment to shape an anti-tumor microenvironment in OC

To deepen our understanding of MLKL’s contribution to the immune microenvironment, a series of analysis was performed. The GSEA analysis (Fig. 3A) highlighted the enrichment of gene sets related to MLKL expression, indicating its significant influence on chemokine signaling pathways. The PPI network analysis and reconstruction by cytoscape software (Fig. 3B) identified a cluster of chemokines, including C-X-C motif chemokine ligand 9 (CXCL9), CXCL10, CXCL11, and CXCL13, that were differentially expressed in MLKL high and low expression groups, suggesting their central role in MLKL-mediated biological processes. Next, uniCox analysis (Fig. 3C, Supplementary Table 2) demonstrated that higher expression levels of these chemokines were associated with improved survival outcomes in OC patients. This underscored their potential as prognostic markers. The correlation analysis (Fig. 3D) revealed a positive correlation between these chemokines and M1 macrophages, indicating their role in macrophage recruitment and activation, which was crucial for an anti-tumor–immune response. Similarly, the chemokines also showed positive correlation with CD8 + T cells (Fig. 3E).

Fig. 3.

MLKL drove chemokine-mediated immune recruitment to shape an anti-tumor microenvironment in ovarian cancer. A GSEA enrichment plot revealed significant upregulation of chemokine signaling pathways in high-MLKL tumors (FDR < 0.05). B Protein–Protein Interaction (PPI) network identified CXCL9/10/11/13 as core chemokine hubs linked to MLKL signaling. Node size reflected interaction degree. C Univariate Cox regression confirmed chemokines as protective prognostic biomarkers in OC (HR < 1; p < 0.05). D-E Positive correlations between key chemokines and M1 macrophage infiltration D/CD8 + T-cell abundance E (***p < 0.001, Pearson). F-I In vitro validation: Conditioned medium (CM) from MLKL-overexpressing SKOV3 F, G and HEY H, I cells significantly upregulated CXCL9/10/11/13 in M0/M1 macrophages (qPCR; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; t-test; ns, not significant). J-K MLKL-induced CM enhanced chemokine expression in Jurkat T cells (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Data represented mean ± SD of triplicates

In vitro functional validation using OC cell lines (SKOV3, HEY) confirmed the modulatory role of MLKL on chemokine production. MLKL overexpression in SKOV3 and HEY cells triggered significant upregulation of CXCL9, CXCL10, CXCL11, and CXCL13 in co-cultured macrophages. Specifically, conditioned medium (CM) from MLKL-overexpressing SKOV3 cells induced significant increases in CXCL10, CXCL11, and CXCL13 in M0 macrophages (Fig. 3F), and CXCL9 and CXCL10 in M1 macrophages (Fig. 3G). Similarly, CM from MLKL-overexpressing HEY cells significantly upregulated all four chemokines (CXCL9, CXCL10, CXCL11, CXCL13) in M0 macrophages and M1 macrophages (Fig. 3H–I). Furthermore, the impact of MLKL extended to T cells. Specifically, CM derived from MLKL-overexpressing SKOV3 and HEY cells markedly elevated CXCL9, CXCL10, CXCL11, and CXCL13 levels in co-cultured Jurkat T cells (Fig. 3J–K). However, the results demonstrated no significant differences in M2 macrophages, suggesting that MLKL does not promote chemokine secretion in M2 macrophages co-cultured with ovarian cancer cells (Supplementary Fig. 2D–F).

Identification of MLKL-associated immune DEGs underpinning anti-tumor immunity in OC patients

Differential gene expression analysis comparing tumors with high versus low MLKL expression identified 738 significantly DEGs (|log2FC|> 1, FDR < 0.05) (Fig. 4A). Similarly, analysis comparing tumors stratified by high versus low ImmuneScore yielded 1135 significantly DEGs (|log2FC|> 1, FDR < 0.05) (Fig. 4B). Additionally, we focused on identifying genes that are specifically associated with MLKL expression and its impact on the immune microenvironment. Intersection analysis revealed a core set of 632 genes common to both DEG lists, establishing them as MLKL-associated immune-related genes (Fig. 4C).

Fig. 4.

Identification and functional annotation of MLKL-associated immune genes in ovarian cancer. A Heatmap of differentially expressed genes (DEGs) between high and low MLKL groups. B Heatmap of DEGs between high and low ImmuneScore expression groups. C Venn diagram identified 632 overlapping genes from immune-related (n = 1135) and MLKL-related DEGs (n = 738). D Gene Ontology (GO) enrichment of intersecting genes revealed significant enrichment in immune processes: Biological Processes (BP; e.g., T-cell activation, leukocyte migration), Cellular Components (CC; e.g., MHC complex, immunological synapse), and Molecular Functions (MF; e.g., chemokine binding, cytokine activity). Bar length indicated gene ratio; color intensity reflects -log₁₀ (q-value). E KEGG pathway analysis confirmed enrichment in anti-tumor immunity pathways including T-cell receptor signaling, chemokine signaling, antigen presentation, and cytokine-cytokine receptor interaction

Functional enrichment analysis of these 632 genes revealed significant involvement in immune-related processes. Gene Ontology (GO) terms included regulation of leukocyte apoptosis, T-cell activation, lymphocyte differentiation, and leukocyte migration (Biological Process); inflammasome complex, MHC class II complex, and immunological synapse (Cellular Component); and chemokine binding, cytokine receptor activity, and immunoglobulin binding (Molecular Function) (Fig. 4D). KEGG pathway analysis further indicated enrichment in Primary immunodeficiency, T-cell receptor signaling, Th1/Th2/Th17 cell differentiation, antigen processing and presentation, NK cell-mediated cytotoxicity, chemokine signaling, and cytokine–cytokine receptor interaction (Fig. 4E).

Collectively, these findings delineate a specific set of 632 genes whose expression is linked to both MLKL levels and the overall immune microenvironment (ImmuneScore) in OC. Their robust enrichment in fundamental immunological processes and pathways strongly implicates MLKL as a key modulator of the tumor–immune landscape through the regulation of these critical immune-related genes.

Molecular stratification based on necroptosis-immune signatures revealed distinct prognostic and microenvironmental phenotypes in OC

To further elucidate the clinical significance of the 632 intersecting genes identified above, uniCox regression analysis was conducted, yielding 102 genes associated with OC survival (Supplementary Table 3). Subsequently, consensus clustering analysis (k = 3) performed on 102 survival-associated genes derived from uniCox regression identified three robust molecular subtypes designated Cluster A, B, and C (Fig. 5A). Kaplan–Meier survival analysis demonstrated significant prognostic stratification among these clusters (Log-rank p = 0.019). Cluster C exhibited the most favorable OS, followed by Cluster B, with Cluster A displaying the poorest prognosis (Fig. 5B).

Fig. 5.

Molecular stratification of ovarian cancer based on necroptosis-immune signatures revealed distinct prognostic and microenvironmental phenotypes. A Consensus clustering matrix (k = 3) derived from 102 survival-associated genes (number of cluster A = 118, cluster B = 182, cluster C = 80). Increasing color intensity (white to blue) indicated higher cluster stability. B Kaplan–Meier analysis confirmed Cluster C (red) had superior overall survival vs. Cluster B (green) and A (blue) (log-rank p = 0.019). Numbers at risk were shown. C MLKL expression gradient across clusters. D Coordinated upregulation of 17 necroptosis pathway genes (FAS, RIPK1/3, MLKL, etc.) in Cluster C (**p < 0.01, ***p < 0.001). E ImmuneScore stratification: Cluster C > B > A (p < 0.001). F Differential immune infiltration across 23 immune cell subsets. Cluster C showed elevated activated CD4 + /CD8 + T cells, dendritic cells, and macrophages (***p < 0.001). G Clinicopathological associations: Cluster C correlated with low-grade tumors, early-stage disease (FIGO II), alive status, and high MLKL/ImmuneScore

Expression analysis revealed a strong association between cluster assignment and key necroptosis regulators. Cluster C demonstrated significantly elevated expression of the core necroptosis executor MLKL compared to Clusters B and A (Fig. 5C). Furthermore, a comprehensive assessment of 17 established necroptosis-related genes (including FAS, FASLG, FADD, RIPK1, RIPK3, TLR3, TNF, TLR4, TICAM1, ZBP1, TRADD, TRAF2, TRAF5, CASP8, TNFSF10, NLRP3) showed a consistent, cluster-dependent expression gradient. Cluster C exhibited the highest overall expression of these genes, followed by Cluster B, with Cluster A showing the lowest expression levels (Fig. 5D), indicating a profound link between cluster identity and the necroptosis pathway.

Evaluation of the tumor–immune microenvironment using the ImmuneScore metric revealed a significant hierarchical correlation with the gene clusters. Cluster C was associated with the highest ImmuneScore, followed by Cluster B, while Cluster A displayed the lowest ImmuneScore (Fig. 5E). Subsequent immune infiltration analysis across 23 distinct immune cell subsets further corroborated this relationship. Significant differences in estimated immune cell abundance were observed among the clusters, with Cluster C generally demonstrating the highest levels of immune infiltration across multiple cell types, including activated B cells, activated CD4/CD8 T cells, activated dendritic cells, various natural killer cell subsets, and macrophages (Fig. 5F).

Finally, the clinical relevance of the clusters was substantiated by their significant associations with established clinicopathological parameters, including tumor grade, disease stage (FIGO stage II, III, IV), vital status (Alive/Dead), and MLKL expression levels (Fig. 5G). These findings collectively establish the identified gene expression clusters (A, B, C) as robust molecular subtypes with distinct prognostic implications, necroptosis pathway activity, immune microenvironment composition, and clinical characteristics in OC.

Construction of the NecropImmScore for prognostic, immune microenvironment, and immunotherapy response stratification of OC patients

Based on the molecular clusters defined by 102 intersected genes, we established a quantitative NecropImmScore to evaluate the necroptosis-immune signature in individual OC tumors. Patients with a high NecropImmScore showed significantly longer overall survival than those with a low score (log-rank test, p < 0.001; Fig. 6A), consistent with the favorable prognosis of cluster C (Fig. 6B).

Fig. 6.

NecropImmScore stratified ovarian cancer patients by prognosis, immune microenvironment, and immunotherapy response. A Kaplan–Meier analysis confirmed superior overall survival in high NecropImmScore patients (log-rank p < 0.001). Numbers at risk were shown. B NecropImmScore distribution across molecular subtypes (Clusters A–C). C ssGSEA demonstrated enhanced immune pathway activity in high-score tumors. Heatmap showed normalized enrichment scores (NES) for 29 immune signatures. D Differential immune cell infiltration: High-score tumors exhibited increased T cells CD8 (p < 0.001), activated CD4 + T cells (p < 0.001), M0 macrophages (p < 0.001), M1 macrophages (p < 0.001), etc. Boxplots showed cell fractions. E Coordinated upregulation of immune checkpoints (PD-1, PD-L1, CTLA4, LAG3, TIGIT; ***p < 0.001) in high-score group. F–I Immunophenoscore (IPS) analysis predicted enhanced response to combinatorial immune checkpoint blockade: CTLA4 + /PD-1 + F, CTLA4 + /PD-1- G, CTLA4-/PD-1 + H, and CTLA4-/PD-1- I

The NecropImmScore effectively stratified tumors based on distinct immune microenvironment features. Analysis of immune cell infiltration revealed significant differences between high and low NecropImmScore groups across multiple immune cell subsets, including activated CD4 memory T cells, activated NK cells, macrophage subtypes (M0, M1, M2), and dendritic cell states (resting and activated) (Fig. 6C–D). This underscored the score’s ability to delineate OC tumors with fundamentally divergent immune landscapes, consistent with the hierarchical ImmuneScore association seen across the original clusters (Cluster C > B > A).

Furthermore, evaluation of key ICPs showed significantly elevated expression levels in the high NecropImmScore group. This coordinated upregulation encompassed PD-1 (PDCD1), PD-L1 (CD274), PD-L2 (PDCD1LG2), CTLA4, CD86, LAG3, TIM3 (HAVCR2), TIGIT, ICOS, and CTSS (Fig. 6E). This pattern suggested an immune-active, potentially checkpoint-responsive TME associated with a high NecropImmScore.

Notably, the NecropImmScore strongly correlated with immunophenotype score (IPS), with high-score tumors exhibiting significantly higher IPS across all combinations of CTLA-4 and PD-1/PD-L1 status (all p-values highly significant, e.g., p < 2.22e-16, Fig. 6F–I), indicating enhanced potential response to immune checkpoint blockade.

Collectively, NecropImmScore was established as a robust clinical tool that quantifies the necroptosis–immune axis in OC. It validates its significant prognostic power, its ability to reflect distinct and biologically relevant tumor–immune microenvironments (including checkpoint expression), and its strong potential as a predictive biomarker for immunotherapy response.

The NecropImmScore as a multifaceted biomarker for prognosis, immunotherapy, and genomic instability in OC patients

To comprehensively evaluate the clinical utility of the NecropImmScore, we first confirmed its independent prognostic value. Multivariate Cox regression analysis, adjusting for age, grade, and stage, established the NecropImmScore as a significant independent prognostic factor for overall survival (HR = 0.9406, 95% CI [0.9147, 0.9673], p = 1.74e-05; Fig. 7A). This prognostic robustness was further validated in the independent ICGC OC cohort, where patients with high NecropImmScores exhibited significantly superior overall survival (Fig. 7B, Log-rank p = 0.037).

Fig. 7.

NecropImmScore predicts clinical outcome, reflects genomic instability, and synergizes with molecular biomarkers in ovarian cancer. A Multivariate cox regression analysis of NecropImmScore and clinical covariates in ovarian cancer. B ICGC validation cohort: High NecropImmScore predicted superior overall survival (log-rank p = 0.037). C–F Mutational landscape analysis of high versus low NecropImmScore groups. C–D Waterfall plot visualizing differential mutation profiles in high versus low NecropImmScore groups. E Forest plot displayed the significantly differentially mutated genes between high and low NecropImmScore cohorts, highlighting BRCA1 as the most differentially altered gene (*p < 0.05; **p < 0.01; ***p < 0.001). F Ovarian cancer patients with BRCA1 mutated-type status correlated with elevated NecropImmScore (p = 0.00023). G ROC analysis confirmed NecropImmScore accurately predicted BRCA1-mutated status (AUC = 0.802). H Kaplan–Meier analysis showed improved overall survival in BRCA1-mutated (mut) versus wild-type (WT) patients (log-rank p = 0.031). I Combined NecropImmScore and BRCA1 status stratified patients into four prognostic subgroups, with high-score/BRCA1-mut tumors exhibiting optimal survival (log-rank p < 0.001). J Correlation heatmap linking NecropImmScore to DNA damage response (DDR) pathway activity (e.g., homologous recombination, Fanconi anemia; *p < 0.05; **p < 0.01; ***p < 0.001). K Scatter plot demonstrating a positive correlation between HRD score and NecropImmScore (p = 0.00016). L Kaplan–Meier survival analysis showing improved overall survival in patients with high HRD scores (log-rank p < 0.001). M Kaplan–Meier survival analysis of combined HRD score and NecropImmScore groups, showing the best survival outcomes in patients with both high HRD and high NecropImmScores (log-rank p < 0.001). N Tumor Mutational Burden (TMB) alone predicted improved survival (log-rank p < 0.001). O Combined NecropImmScore/TMB stratification revealed synergistic prognostic power: High-score/high-TMB patients showed optimal survival versus other subgroups (log-rank p < 0.001)

We next characterized the genomic landscape associated with the NecropImmScore. Mutational analysis revealed that while the overall mutation rate was high in both groups, the high NecropImmScore group was characterized by a significantly higher frequency of BRCA1 mutations (Fig. 7C–F). The NecropImmScore itself demonstrated strong predictive power for BRCA1 deficiency, as evidenced by ROC curve analysis (AUC = 0.802; Fig. 7G). Survival analysis confirmed the favorable prognosis associated with BRCA1 mutations (Fig. 7H, p = 0.031), and, importantly, the integration of NecropImmScore with BRCA1 status provided superior prognostic stratification, with patients exhibiting both a high score and a BRCA1 mutation having the best outcomes (Fig. 7I, p < 0.001).

Beyond BRCA1, the NecropImmScore demonstrated broader associations with DDR pathway activity. Analysis revealed significant correlations between the score and the expression levels of key genes involved in other critical DDR pathways (e.g., homologous recombination, Fanconi anemia pathway, mismatch repair), suggesting the score reflected a wider genomic instability profile influencing the TME of OC patients (Fig. 7J). Furthermore, the NecropImmScore showed a significant positive correlation with the HRD score, a comprehensive measure of genomic instability (Fig. 7K). Patients with high HRD scores had improved survival (Fig. 7L), and the combination of high HRD and high NecropImmScore identified a subgroup with the most favorable prognosis (Fig. 7M, p < 0.001), underscoring a synergistic relationship between genomic instability and the necroptosis–immune axis.

The predictive utility of the NecropImmScore was then assessed in the context of immunotherapy. In the IMvigor210 cohort of anti-PD-L1-treated patients, a high NecropImmScore was associated with significantly improved overall survival (Supplementary Fig. 3A, Log-rank p = 0.014) and was strongly correlated with objective clinical response, where patients achieving a Complete Response (CR) displayed the highest scores (Supplementary Fig. 3B) This enhanced benefit was mechanistically linked to a tumor microenvironment characterized by an “inflamed” immune phenotype (Supplementary Fig. 3C, p < 0.001). The distribution of immune phenotypes across gene clusters revealed that gene cluster C predominantly exhibited the “inflamed” phenotype, indicating an immunologically “hot” tumor environment, while gene cluster A was enriched for the “desert” phenotype, suggesting an immunologically “cold” environment, with gene cluster B showing a mixed phenotype (Supplementary Fig. 3D, p < 2.2e-16). These findings highlight the role of NecropImmScore in identifying tumors with distinct immune microenvironments. Additionally, a higher tumor neoantigen burden (TNB) was observed in tumors with a high NecropImmScore, further supporting the association with an immunologically active tumor environment (Supplementary Fig. 3E, p < 0.001). The integrative prognostic power of the NecropImmScore was further highlighted when combined with TMB. While high TMB alone predicted improved survival (Fig. 7N, p < 0.001), patients with both high TMB and a high NecropImmScore constituted a subgroup with the most favorable outcomes, significantly outperforming all other combinations (Fig. 7O, p < 0.001).

Finally, we explored the score’s ability to predict response to conventional chemotherapy. In silico drug sensitivity analysis indicated that patients with high NecropImmScores exhibited significantly increased sensitivity to standard chemotherapeutic agents, including cisplatin, paclitaxel, and gemcitabine (Supplementary Fig. 3F–H).

Collectively, these analyses position the NecropImmScore as a multifaceted biomarker that independently predicts prognosis, identifies responders to immunotherapy, reflects underlying genomic instability (including BRCA1 deficiency and HRD), and indicates susceptibility to conventional chemotherapy.

Discussion

This study provides compelling evidence that necroptosis, specifically orchestrated by its terminal executor MLKL, functioned as a pivotal regulator of anti-tumor immunity in OC. We demonstrated that MLKL expression was a protective prognostic biomarker associated with enhanced immune infiltration, M1 macrophage polarization, activated CD4 + T cells, and an elevated expression of critical ICP molecules. Crucially, we established that MLKL reprogrammed the TME through the induction of key chemokines (CXCL9/10/11/13), facilitating the recruitment and activation of immune effector cells. Building upon this mechanistic foundation, we developed and validated the NecropImmScore, a novel integrated biomarker derived from MLKL-associated immune genes, which effectively stratified OC patients into distinct molecular subtypes with divergent clinical outcomes, TME phenotypes, therapeutic vulnerabilities, and immunotherapy responses.

Our findings position MLKL as a central orchestrator of the OC immune landscape, extending beyond its canonical role as an executor of lytic cell death. This dual functionality—mediating ICD and actively reshaping the TME—aligns with emerging evidence across malignancies. Recent studies in pancreatic ductal adenocarcinoma (PDAC), bladder urothelial carcinoma (BLCA), and lung cancer highlight MLKL-driven necroptosis as a potent inducer of T-cell infiltration and a predictor of immunotherapy response [38]. The robust positive correlation between MLKL expression and ImmuneScore across multiple cohorts, including pan-cancer analyses, underscored its fundamental role in modulating TME composition. The functional validation, demonstrating MLKL overexpression in OC cells directly promotes M1 macrophage polarization and activates Jurkat T cells (evidenced by upregulated CCR4/5/7/9, CD69, CD3D/E, GZMB), provides direct experimental support for its immunomodulatory capacity. This aligned with the concept of ICD triggered by necroptosis, where the release of DAMPs like HMGB1 and ATP can activate APCs [39]. However, our data extend beyond passive DAMP release. We identify a novel chemokine-centric mechanism: MLKL induces robust expression of CXCL9, CXCL10, CXCL11, and CXCL13 in co-cultured macrophages and T cells. These chemokines are critical ligands for CXCR3 and CXCR5 receptors, known to recruit cytotoxic T lymphocytes (CTLs) and T helper 1 (Th1) cells into tumors [40]. This chemokine-driven recruitment represents a proactive strategy by which MLKL transforms the OC TME from an immunologically “cold” state to an “inflamed” phenotype, directly countering the immunosuppressive mechanisms (e.g., Treg/MDSC dominance, T-cell exclusion) that characterize advanced OC [41].

The integration of MLKL-related DEGs and immune-related DEGs revealed a core set of 632 genes significantly enriched in pathways essential for anti-tumor immunity: T-cell activation, chemokine signaling, antigen presentation, and cytokine–cytokine receptor interaction. This genetic signature underpins the observed immune phenotypes and provided the foundation for defining molecular subtypes via consensus clustering of 102 survival-associated genes. The identification of three distinct clusters (A–C) with progressively improving prognosis (Cluster C > B > A), paralleled by increasing MLKL expression, necroptosis pathway activity, and immune infiltration (ImmuneScore), solidified the interconnection between functional necroptosis, a robust immune response, and favorable clinical outcomes. Cluster C, characterized by the highest NecropImmScore, represents an immunologically “hot” phenotype primed for immune recognition and attack.

The NecropImmScore emerged as a powerful and multifaceted clinical tool. Its robust prognostic value was validated across independent cohorts (TCGA-OV, ICGC OV-AU), with high-score patients exhibiting significantly superior OS. Notably, the continuous NecropImmScore demonstrated a highly significant prognostic value (p < 0.001), which is substantially stronger than that of the discrete clusters (p = 0.019). This highlights the robustness and precision of the continuous score in capturing the necroptosis-immune landscape and its impact on patient outcomes. The continuous score allows for a more nuanced and comprehensive representation of the necroptosis-immune signature, enabling finer stratification of patient prognosis by accounting for the full spectrum of variability in the underlying gene expression patterns. Critically, in the IMvigor210 immunotherapy cohort, high NecropImmScore predicted improved survival outcomes following anti-PD-L1 therapy and was strongly associated with objective clinical responses (CR/PR vs. SD/PD). This predictive power stems from its ability to identify tumors with an “inflamed” TME phenotype, characterized by abundant immune cell infiltration, elevated ICPs expression (PD-L1, CTLA4, LAG3, TIGIT), and higher TNB. These features are hallmarks of ICB-responsive tumors [42]. It is important to note that while the IMvigor210 cohort is primarily focused on urothelial carcinoma, the immune phenotypes (desert, excluded, inflamed) described in this cohort have been widely recognized across various cancer types, including OC. These phenotypes describe the general state of immune cell infiltration and immune activity within the TME, which can be similarly assessed in OC. The synergy between high NecropImmScore and high TMB provided superior prognostic stratification, highlighting its complementary value to established genomic biomarkers. This positions the NecropImmScore as a potentially transformative tool for selecting OC patients most likely to benefit from immunotherapy, addressing a critical unmet need in this malignancy.

Beyond immunotherapy prediction, the NecropImmScore demonstrated significant associations with genomic instability and therapeutic vulnerabilities. BRCA1 loss induces genomic instability and cytosolic DNA accumulation, activating the cGAS-STING pathway and downstream type I interferon responses, which can foster an immunogenic TME [43], yet its integration with immune biomarkers remains underexplored. Our data revealed the strong predictive capacity of NecropImmScore for BRCA1 deficiency (AUC = 0.802) and synergistic prognostic value when combined with BRCA1 mutation status reveals a deep connection to DDR pathways. This link was further supported by correlations with expression levels of other key DDR genes.

Furthermore, our analysis established a significant positive correlation between the NecropImmScore and HRD score, a comprehensive genomic scar signature reflecting a state of profound genomic instability. This finding mechanistically links the necroptosis–immune axis to the core biological feature of a subset of OCs. A high HRD score, indicative of defective DNA double-strand break repair, was itself a strong prognosticator for superior survival, consistent with the notion that genomic instability fosters the generation of neoantigens and promotes an immunogenic TME [44]. The synergy observed between high HRD scores and high NecropImmScores for patient prognostication is particularly compelling. It suggests that tumors with concurrent defects in DNA repair (creating immunogenic potential) and an active necroptosis-driven immune recruitment mechanism (enabling immune effector cell infiltration and activation) represent a distinct biological entity with the most favorable outcomes. This aligns with the paradigm that DDR deficiency, particularly in BRCA1/2 genes, can activate the cGAS-STING pathway, leading to type I interferon production and subsequent T-cell priming [45]. Our results posit that the NecropImmScore effectively captures the immune-reactive consequences of this genomic instability, identifying those HRD-high tumors that have successfully overcome local immunosuppression to mount an effective anti-tumor response. This synergistic stratification not only refines risk assessment but also holds significant implications for therapy selection, as patients with high HRD and high NecropImmScore are likely the optimal candidates for combining DNA-damaging agents (like platinum chemotherapy or PARP inhibitors) with immunotherapy.

Importantly, high NecropImmScore predicted increased sensitivity to standard-of-care chemotherapeutics (cisplatin, paclitaxel, gemcitabine). This aligns with the concept that DDR-deficient tumors are more vulnerable to DNA-damaging agents, but crucially, our score integrates the immune context of this vulnerability, potentially reflecting enhanced immunogenic cell death induced by chemotherapy in inflamed TMEs [23]. This differential sensitivity profile suggests the NecropImmScore could guide personalized treatment selection, not only for immunotherapy but also for conventional chemotherapy, based on the underlying molecular and immunological landscape of the tumor.

While this study provides significant insights, certain limitations warrant consideration. The reliance on retrospective cohorts necessitates validation in prospective clinical trials. The in vitro co-culture systems, though informative, cannot fully recapitulate the complexity of the human TME in vivo. Future investigations should employ in vivo models (e.g., immunocompetent OC mouse models) to further elucidate the causal mechanisms by which MLKL modulates anti-tumor immunity and validate the therapeutic implications of the NecropImmScore. Tumor heterogeneity poses significant challenges for the reproducibility of molecular subtyping signatures. The coexistence of multiple signals within the same tumor tissue and organ-specific environmental differences between primary and metastatic sites can affect the consistency of the NecropImmScore. Future studies should include pathological validation through tumor macro- or microdissection to assess the spatial heterogeneity of the NecropImmScore within tumors, including immune-active areas, immune-sparse “cold” regions, tumor-free stroma, and primary versus metastatic sites. Besides, exploring the non-lytic functions of MLKL in immune modulation and investigating potential crosstalk between necroptosis and other forms of RCD (e.g., pyroptosis, ferroptosis) within the OC TME represent promising future directions. Additionally, while the in silico drug sensitivity predictions using pRRophetic provide valuable insights into the potential therapeutic vulnerabilities associated with the NecropImmScore, it is important to note that these predictions require experimental validation. Future studies are warranted to confirm these findings through in vitro and in vivo experiments, which will further elucidate the clinical utility of the NecropImmScore in guiding treatment decisions for OC patients. Furthermore, translating the NecropImmScore into a clinically feasible assay (e.g., targeted RNA-seq panel) is essential for its practical application in routine diagnostics and therapeutic decision-making.

In conclusion, this work established MLKL as a master regulator of anti-tumor immunity in OC, driving chemokine-mediated immune cell recruitment and TME reprogramming toward an immunogenic state. The novel NecropImmScore, derived from the integration of necroptosis and immune signatures, transcended a simple prognostic indicator. It served as a multifaceted biomarker capable of predicting OS, stratifying molecular subtypes, identifying inflamed TMEs responsive to immunotherapy, forecasting BRCA1 deficiency, and indicating sensitivity to conventional chemotherapy. By capturing the critical intersection of necroptotic signaling and productive anti-tumor immunity, the NecropImmScore offers a powerful framework for advancing precision oncology in OC, ultimately aiming to improve patient stratification and therapeutic outcomes.

Conclusion

Based on the integrative multi-omics and functional validation in this study, MLKL-driven necroptosis orchestrates anti-tumor immunity in OC by reprogramming the immunosuppressive microenvironment through chemokine induction (CXCL9/10/11/13), enhancing immune infiltration, M1 macrophage polarization, and T-cell activation. Leveraging this mechanism, the novel NecropImmScore biomarker robustly predicted superior patient prognosis (validated in TCGA/ICGC cohorts), identified immunotherapy responders (correlating with inflamed TME, elevated checkpoint expression, and improved survival/response in IMvigor210), forecasted BRCA1 deficiency (AUC = 0.802) while synergizing with BRCA1/TMB/HRD score for enhanced risk stratification, and indicated chemosensitivity to cisplatin, paclitaxel, and gemcitabine. This signature transcends conventional biomarkers, offering a precision tool to optimize therapeutic strategies for OC.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Shaohua Xu conceived and designed the manuscript, Fangfang Xu analysed and interpretated the data. Fangfang Xu and Shufeng Kang performed the basic experiments. Fangfang Xu and Shufeng Kang wrote the manuscript. Shaohua Xu helped with manuscript and data review. All authors read and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 81772762), Clinical science, the Natural Science Foundation of Shanghai (Grant No. 21ZR1450900), and the Shanghai Science and Technology Planning Project (21Y11907000).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Conflict of interest

The authors declare no conflict of interests.